I

شــركـت رهــروان طــب ))با مسئوليت محدود((ثمين

Medical Journals

Print & CD-Rom

Full Text Archives of 250 Medical

Journals of CD-Rom

Printable , Color

For More Info. Please Contact Us :

Tel & Fax : 021- 6429069 - 6432484 – 6947341

Second Floor , NO.135,Keshavarz Blvd./

Tehran 14187 , Iran – P.O.Box : 14185 - 157

AJRCCM -- Table of Contents (April 15 2004, 169 [8])

Contents: Volume 169, Issue 8; April 15, 2004

EDITORIALS OCCASIONAL ESSAYS A. Hypersensitivity, Pneumonitis and Nicotine B. Community-Acquired Pneumonia C. Asthma, Pranlukast, and Bone Marrow D. Infantile Airway Function and Asthma E. Cystic Fibrosis and Airway Function F. Cardiovascular Effects of Particulate Matter G. Maximal Pulmonary Function: Determinants H. Sleep, Breathing, Blood Pressure, Children I. Pressure-Volume Curve and Lung Injury NHLBI WORKSHOP CORRESPONDENCE

EDITORIALS:

Martin J. Tobin ATS Centenary: Four-Century Prologue to a Century of Progress Am. J. Respir. Crit. Care Med. 2004; 169: 891-893 [Full Text]

Lisa A. Maier Is Smoking Beneficial for Granulomatous Lung Diseases? Am. J. Respir. Crit. Care Med. 2004; 169: 893-895 [Full Text]

Scott F. Dowell Surviving Pneumonia—Just a Short-Term Lease on Life? Am. J. Respir. Crit. Care Med. 2004; 169: 895-896 [Full Text]

OCCASIONAL ESSAYS:

John B. West A Century of Pulmonary Gas Exchange Am. J. Respir. Crit. Care Med. 2004; 169: 897-902 [Full Text]

AJRCCM -- Table of Contents (April 15 2004, 169 [8])

ARTICLES:

A. Hypersensitivity, Pneumonitis and Nicotine:

Marie-Renée Blanchet, Evelyne Israël-Assayag, and Yvon Cormier Inhibitory Effect of Nicotine on Experimental Hypersensitivity Pneumonitis In Vivo and In Vitro Am. J. Respir. Crit. Care Med. 2004; 169: 903-909. First published online December 30 2003 as doi:10.1164/rccm.200210-1154OC [Abstract] [Full Text]

B. Community-Acquired Pneumonia:

Grant W. Waterer, Lori A. Kessler, and Richard G. Wunderink Medium-Term Survival after Hospitalization with Community-Acquired Pneumonia Am. J. Respir. Crit. Care Med. 2004; 169: 910-914. First published online December 23 2003 as doi:10.1164/rccm.200310-1448OC [Abstract] [Full Text]

C. Asthma, Pranlukast, and Bone Marrow:

Krishnan Parameswaran, Richard Watson, Gail M. Gauvreau, Roma Sehmi, and Paul M. O'Byrne The Effect of Pranlukast on Allergen-induced Bone Marrow Eosinophilopoiesis in Subjects with Asthma Am. J. Respir. Crit. Care Med. 2004; 169: 915-920. First published online January 23 2004 as doi:10.1164/rccm.200312-1645OC [Abstract] [Full Text] [Online Supplement]

D. Infantile Airway Function and Asthma:

Stephen W. Turner, Lyle J. Palmer, Peter J. Rye, Neil A. Gibson, Parveenjeet K. Judge, Moreen Cox, Sally Young, Jack Goldblatt, Louis I. Landau, and Peter N. Le Souëf

The Relationship between Infant Airway Function, Childhood Airway Responsiveness, and Asthma Am. J. Respir. Crit. Care Med. 2004; 169: 921-927. First published online February 5 2004 as doi:10.1164/rccm.200307-891OC [Abstract] [Full Text] [Online Supplement]

E. Cystic Fibrosis and Airway Function:

AJRCCM -- Table of Contents (April 15 2004, 169 [8])

Sarath C. Ranganathan, Janet Stocks, Carol Dezateux, Andrew Bush, Angie Wade, Siobhán Carr, Rosemary Castle, Robert Dinwiddie, Ah-Fong Hoo, Sooky Lum, John Price, John Stroobant, and Colin Wallis

The Evolution of Airway Function in Early Childhood Following Clinical Diagnosis of Cystic Fibrosis Am. J. Respir. Crit. Care Med. 2004; 169: 928-933. First published online January 30 2004 as doi:10.1164/rccm.200309-1344OC [Abstract] [Full Text]

F. Cardiovascular Effects of Particulate Matter:

Michael Riediker, Wayne E. Cascio, Thomas R. Griggs, Margaret C. Herbst, Philip A. Bromberg, Lucas Neas, Ronald W. Williams, and Robert B. Devlin

Particulate Matter Exposure in Cars Is Associated with Cardiovascular Effects in Healthy Young Men Am. J. Respir. Crit. Care Med. 2004; 169: 934-940. First published online February 12 2004 as doi:10.1164/rccm.200310-1463OC [Abstract] [Full Text] [Online Supplement]

G. Maximal Pulmonary Function: Determinants:

Xiaobin Wang, Tjeert T. Mensinga, Jan P. Schouten, Bert Rijcken, and Scott T. Weiss Determinants of Maximally Attained Level of Pulmonary Function Am. J. Respir. Crit. Care Med. 2004; 169: 941-949 [Abstract] [Full Text]

H. Sleep, Breathing, Blood Pressure, Children:

Raouf S. Amin, John L. Carroll, Jenny L. Jeffries, Charles Grone, Judy A. Bean, Barbara Chini, Ronald Bokulic, and Stephen R. Daniels

Twenty-four–hour Ambulatory Blood Pressure in Children with Sleep-disordered Breathing Am. J. Respir. Crit. Care Med. 2004; 169: 950-956. First published online February 5 2004 as doi:10.1164/rccm.200309-1305OC [Abstract] [Full Text]

I. Pressure-Volume Curve and Lung Injury:

John M. Downie, Arthur J. Nam, and Brett A. Simon Pressure–Volume Curve Does Not Predict Steady-State Lung Volume in Canine Lavage Lung Injury Am. J. Respir. Crit. Care Med. 2004; 169: 957-962. First published online February 5 2004 as

AJRCCM -- Table of Contents (April 15 2004, 169 [8])

doi:10.1164/rccm.200305-614OC [Abstract] [Full Text] [Online Supplement]

NHLBI WORKSHOP:

Scott T. Weiss and Stephanie Shore Obesity and Asthma: Directions for Research Am. J. Respir. Crit. Care Med. 2004; 169: 963-968. First published online January 23 2004 as doi:10.1164/rccm.200303-403WS [Full Text]

CORRESPONDENCE:

Shekhar Venkataraman, Adrienne Randolph, James Hanson, Peter Forbes, Ira Cheifetz, Rainer Gedeit, and Peter Luckett

All Roses Are Flowers, But Not All Flowers Are Roses Am. J. Respir. Crit. Care Med. 2004; 169: 969 [Full Text]

Douglas B. Trout, Elena H. Page, Diane R. Gold, Paul C. Stark, and Harriet A. Burge Fungal Exposure and Lower Respiratory Illness in Children Am. J. Respir. Crit. Care Med. 2004; 169: 969-971 [Full Text]

Daniel R. Goldstein, Bethany M. Tesar, Scott M. Palmer, Lauranell H. Burch, and David A. Schwartz Toll-like Receptors and Allograft Rejection Am. J. Respir. Crit. Care Med. 2004; 169: 971-972 [Full Text]

John S. Torday and Virender K. Rehan Forces in Emphysema: Newtonian v Quantum Mechanics Am. J. Respir. Crit. Care Med. 2004; 169: 972 [Full Text]

Editorials

ATS CentenaryFour-Century Prologue to a Century of Progress

The National Association for the Study and Prevention of Tuber-culosis, forerunner of the American Lung Association, wasfounded in 1904, with Edward Trudeau (1848–1915) as president,and William Osler vice-president (1, 2). During the first twomeetings, some sanatorium physicians decided to found a smallassociation as an appendage to the parent organization. OnDecember 1, 1905, 17 (of 49 invited) sanatorium physicians metat the Natural History Museum of New York, and the AmericanSanatorium Association was formed (1). In 1939, membershipwas broadened to include physicians with an interest in tubercu-losis who did not work in sanatoria, and the name was changedto American Trudeau Society (1). The name was later changedto American Thoracic Society in 1960. To celebrate our Society’sfounding, the Journal is publishing a series of articles coveringthe history of pulmonary and critical care medicine over the lastcentury.

The formation of a society of medical specialists at the turnof the twentieth century arose from three historical currents: theheritage of scientific societies that had been in existence forfour centuries; the maturation of medicine as a profession; andjustification for specialization. I will not cover the history of pulmo-nary and critical care medicine over the last century—that is thesubject of the upcoming articles. Instead, this editorial serves asprologue, tracing three overlapping, yet unique, historical currentsthat led to the emergence of phthisiology as a subspecialty.

The Scientific Revolution was the time of greatest change inthe last millennium, reducing the Renaissance and Reformationto the rank of mere internal displacements within medieval Eu-rope (3). Though indebted to the tradition of Greco-Islamicthought, scholars recognized the incompleteness of inheritedknowledge. Shaking off the shackles of scholasticism, a vanguardof thinkers made a fresh start at scholarship: henceforth, newobservations were to be criticized, validated, and replicated be-fore being incorporated into the body of knowledge (3, 4). Thefirst known scientific society was the Accademia dei Lincei(Academy of the Lynxes), founded in Rome by Duke FedericoCesi in 1601 (3, 5). Galileo (1564–1642), the originator of theempirical method, joined as a corresponding member in 1609and the Accademia published two of his works (5). When itspatron died in 1630, the Accademia ceased. In Florence, twoMedici brothers, Prince Leopoldo and Grand Duke FerdinandII, founded the Accademia del Cimento (Academy of Experi-mentation) in 1657 (5). When Leopoldo received a cardinal’s hatin 1667, the Accademia ended. The closing of these Accademiastogether with the obscurantist reaction to new thinking at thetrial of Galileo caused the Scientific Revolution to peter out inItaly (3, 5).

On the death of Galileo in 1642, the Scientific Revolutionmoved to a more hospitable climate in England, at that time inrevolt against a dictatorial king. English nonconformists gave asharper meaning to irreverence for inherited authority whenthey cut off the head of Charles I in 1649 (6). Enthusiasts forscience had started weekly meetings in 1645. The “invisible col-

Am J Respir Crit Care Med Vol 169. pp 891–896, 2004Internet address: www.atsjournals.org

lege” was the name Robert Boyle (1627–1691) gave to theseinformal gatherings (7). In 1660, the monarchy was restoredand later that year the Royal Society of London for ImprovingNatural Knowledge was founded. Charles II granted a royalcharter in 1662, but provided no money; he assigned the societyclaims on Irish lands, but these proved impossible to realize (5).Membership was open to anyone, not because of egalitarianprinciples but through financial necessity (8). Early membersincluded Robert Boyle and Robert Hooke, Thomas Willis andEdmund Halley, and Isaac Newton (9). Without a governmentgrant, the society had constant financial difficulties. They thoughtof raising money through teaching but Boyle blocked the idea,seeing it as a distraction from experimentation (5). Instead, theydecided to publish periodic reports of work done under theirauspices or by other scientists. And so, the first scientific journal,the Philosophical Transactions, was published in 1665 (10).

In Paris, Louis XIV (1638–1715) founded the Academie Roy-ale des Sciences in 1666, holding the inaugural meeting in hisprivate library (8). Unlike the penurious English king, Louisprovided substantial financial support for full-time investigation.After the death in 1683 of Colbert, Louis’ controller-general offinances, the Academie declined (5). The ending of religioustoleration, with the revocation of the Edict of Nantes in 1685,accelerated France’s decline (6).

The new approach to scholarship, involving telescopes andmicroscopes, was expensive. As such, there were cogent reasonsto form unions (3). Describing the ideal research institute inthe New Atlantis (1626), Francis Bacon—acknowledged as theinspiration for the new movement—stressed that progress couldbest be achieved through cooperation (communism he called it)(8, 11). And after 1660 unions of researchers dominated science.That the discoveries of men such as Newton and Boyle, andHuygens and Leibniz were closely connected with scientific soci-eties, contributed hugely to the societies’ immediate success (5).In contrast to art—the other field of creativity—science is ofvalue only when used by other scientists. Claude Bernard madethe distinction pithily: “Art is I; science is we.” In 1600, scienceis in the Middle Ages; by 1660, with the inception of the scientificsocieties, it is in modern times (5). The advances to learning inthe seventeenth century outshine those of any other (3, 5).

By the time of Newton’s death in 1727, the Royal Society haddegenerated into a group dominated by intellectual snobbery (6).Fellows nursed their awards, hobnobbed with amateur scientistsamong the aristocracy, and were more concerned with statusthan with making original discoveries (6). The most famousexample of conservative caution was the rejection of EdwardJenner’s experiments on vaccination by the Philosophical Trans-actions in 1796 (12). As such, the inventions of the IndustrialRevolution during the eighteenth century were made outsidethe scientific societies and universities (6, 10).

Medical knowledge, the second of our three historical cur-rents, was slow to build on the discovery of a Newtonian universeinvolving clockwork motion produced by discoverable laws (3,13). In the early eighteenth century, the Dutch city of Leidenwas preeminent, thanks to Hermann Boerhaave (1668–1738),who promoted clinical teaching on the wards of the local charityhospital (13). Boerhaave’s trainees brought Leiden methods toEdinburgh, which dominated British medicine for decades (13).

892 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

And Edinburgh served as the model for the founding of medicalschools in Britain’s North American colonies in the latter eigh-teenth century (13, 14). In 1761, Giovanni Battista Morgagni ofPadua, waiting till his seventy-ninth year to publish his majorwork (15), showed how symptoms could be traced back to diseasedorgans, opening the way for logical diagnosis. Throughout theeighteenth century, medicine was still operating more like atrade than the lofty profession to which it aspired (13).

If blood from the head of Charles I spurred on English sci-ence, the decapitation of Louis XVI did the same for Frenchmedicine. The French Revolution moved hospitals out of ecclesi-astical control and into the hands of the nation (13). Salariedphysicians turned the large public hospitals of Paris into institu-tions for medical research and teaching (16). Homage to yester-day’s authorities went the way of the ancien regime, and hands-onexperience replaced book learning. Symptoms were seen as su-perficial, and diseases as conditions with laws of their own tobe disclosed through autopsies. Xavier Bichat (1771–1802) per-formed 600 autopsies and built a new system of tissue pathology(though never obtaining a major hospital post) (13). The shiftfrom symptoms to pathological lesions emphasized the ontologi-cal model of disease: diseases as discrete entities, real things.At a minor and peripheral institution, Hopital Necker, ReneLaennec (1781–1826) invented the stethoscope in 1816 (15),which remained medicine’s chief diagnostic tool until the discov-ery of X-rays in 1895. Vast numbers of students from Europeand North America (up to 5,000 at a time) were flocking to earlynineteenth-century Paris, with Laennec alone teaching some 300foreign pupils (13, 14).

Around the mid-nineteenth century, preeminence in medicinecrossed to the German states where it stayed till dissipated bytwentieth-century war (13, 15). At last, science moved into theuniversities. The first laboratory for teaching science was estab-lished at the University of Giessen by Justus von Liebig in 1824(11). Germans established the first research organizations that wereboth large-scale and self-consciously professional (17). JohannesMuller (1801–1858) is regarded as the greatest physiologist—lessfor his discoveries than for developing a systematic approach toexperimental physiology. Muller’s trainees included Schwann,Traube, Henle (through him, Koch and Waldeyer), Ludwig(through him, Welch), Helmholtz (through him, Hertz, Planck,and Michelson), and Virchow (through him, Ehrlich and Pavlov)(11, 18). Rudolf Virchow (1821–1902) gave pride of place to thecell, as Morgagni had done for the organ and Bichat the tissue(15). Several factors contributed to the success of German sci-ence (14). The universities were engines of inquiry, whereasEnglish and French universities served more as finishing schoolsfor gentlemen (13). Shunning the dilettante attitude of the latter,German faculty recognized that the war on ignorance would bewon by painstaking, if pedestrian, investigations, thoroughness,and attention to detail (17). Their approach calls to mind thewords of Carlyle that genius “means transcendent capacity fortaking trouble.” For the first time, universities promoted facultyon the basis of original research. Academic freedom, originatingout of the failed Revolution of 1848, became paramount: profes-sors pursued research without constraint (19). Germany servedas the model for reform of American universities, still intellectualbackwaters in the late 1800s (16). About half of future leadersof American medicine graduating around this time studied in aGerman-speaking country (20). And Germany was the first tointroduce socialized medicine when Chancellor Bismarck em-barked on a national system of compulsory sickness insurancein 1883 (13).

By the late nineteenth century few therapies had been intro-duced: opium (around 3,000 BC), cinchona bark (1677), digitalis(1785), morphine (1805), quinine (1820), salicin (1826), ether(1842), chloroform (1847), and cocaine (1859) (21). New diagnos-

tic techniques, however, were strengthening the authority ofmedicine and altering the patient–doctor relationship. Early inthe nineteenth century, diagnosis depended on a patient’s recita-tion of symptoms (15). By mid-century, the stethoscope hadbeen joined by the ophthalmoscope and laryngoscope, the ther-mometer and spirometer (16). The discovery of X-rays by Wil-helm Konrad Roentgen in 1895 revolutionized the perceptionof physicians by patients. Whereas an ophthalmoscope could beused by only one physician at a time, and thus at risk of subjectivedistortion, X-rays could be viewed by several physicians simulta-neously, bolstering claims to objective judgment (16). By themid-1800s, successful peasants and petit bourgeois of Europewere routinely visiting a doctor, whereas they would rarely havedone so in 1700 (13).

Medicine today is organized through structures that allowphysicians considerable autonomy under state protection, whileclaiming to protect the public from charlatans and substandardcare (13). The rise of any profession depends ultimately on itsauthority, that is, its claim to the possession of specialized andvalidated knowledge, technical skills, and rules of behavior thatcompel trust (16). To acquire authority, a profession needs toachieve consensus among members (as to goals) and be seen aslegitimate by the state (16). All this crystallized only in the latenineteenth century.

In ancien regime France, licensure was in the hands of univer-sity faculties. The Jacobins judged such practices as efforts toreduce competition and raise rewards, and abolished them (13).In early nineteenth-century America, attempts to introduce li-censure were also viewed as efforts to gain a selfish monopolyand failed (14, 16). In 1858, the London Parliament created asingle register for all approved practitioners and a council toserve as an ethico-legal watchdog (13, 14). The United StatesSupreme Court finally ruled in favor of licensing in 1888 (16).A key precondition for the acquisition of medical authority wasthe construction of large hospitals. Once dreaded as cesspoolsof infection, hospitals began to be seen as temples of healingand citadels of science, affording them a new moral identity.America had only 178 hospitals in 1872, increasing to 4,000 in1910 and over 6,000 by 1920 (16).

Adam Smith’s views on the division of labor, in his Wealthof Nations, also applied to medicine. As the medical marketgrew, so did opportunity and incentive to specialize—the last ofour three historical currents. The satisfaction of mastering asingle field was attractive (22). Pediatrics, neurology, and derma-tology emerged as separate specialties by the late 1880s (23).Just as the craft guilds of the 1500s struggled to secure economicadvantage, so physicians of the early 1900s battled to defineboundaries of authority against emerging ancillary occupations(16). American nurses were strongly established as anesthetistsby the 1920s, and non-physicians were sometimes in charge ofradiology units (16, 24). Only in the late 1930s did physiciansgain control of these two departments. It is no accident that thefirst certification examinations were developed by two groupsthat felt the keenest competition from non-physicians: ophthal-mologists (in 1917) and otolaryngologists (in 1924) (16, 23, 24).The goals of specialty boards were to provide hallmarks of qualityin fields that were prime targets for non-physicians (such as optom-etrists) and general practitioners, and to establish monopoliesbased on specialist techniques (23). The American Board of Inter-nal Medicine was established in 1936, partly in response to thechallenge of separate boards in tuberculosis and gastroenterology,already in embryonic form. Subspecialty certification in tubercu-losis was introduced in 1941 (23).

Tuberculosis has ravaged mankind back to Neolithic times.It is estimated to have caused a seventh to a quarter of deaths ofworking people in the nineteenth century. Victims have includedFrederic Chopin and Niccolo Paganini, Friedrich Schiller and

Editorials 893

Franz Kafka, Emily, Anne and Charlotte Bronte, John Keats,Laurence Sterne, Anton Chekhov, D.H. Lawrence, George Or-well, and Simonetta Vespucci (the pale model for Botticelli’sVenus) (25). To make sense of such tragic loss of talent, wastingof the flesh was thought to inspire creativity of the soul—so-called spes phthisica. Pianists, poets, and playwrights notwith-standing, tuberculosis mainly affected the urban poor (13).

The great triumph was Koch’s discovery of the tubercle bacil-lus in 1882 (26). The tuberculin test (introduced in 1890 andrefined in 1907) revealed that latent infection was widespread.Satisfactory drug therapy would not arrive until the 1940s (andnone of the agents was to emerge from a great center of medicalresearch [25]). In 1854, Hermann Brehmer opened a sanatoriumin Silesia, and Edward Trudeau opened his cottage in the Adiron-dacks thirty years later (25). The first pneumothorax operationwas attempted by Carlo Forlanini of Pavia in 1888 (13). Theoperation was taken up by Ferdinand Sauerbruch (1875–1951),who tried solving the problem of lung collapse on opening thethorax through the use of a negative pressure chamber. (Sauer-bruch later served on the Committee of the Reich ResearchCouncil that approved grants for the Mengele twin experimentsin Auschwitz [27]–clear evidence that history, even medical his-tory, does not necessarily represent a progressively upward evo-lution.) By 1904, negative pressure chambers to contain a patientand full operating team were in use on both sides of the Atlantic.The three historical currents of this prologue—the heritage ofscientific societies, the maturation of medicine, and the justifica-tion for specialization—now converged in the battle against tu-berculosis. Local and national associations sprang up in Europeand America, and one of these was the National Associationfor the Study and Prevention of Tuberculosis (with the AmericanSanatorium Association as an appendage) (1, 2). In foundingthis association, Trudeau, Osler, and assembled phthisiologistswere linking each of us, as members of today’s ATS, with Galileo,Boyle, Newton, and the first unions of scientists that battledignorance for the betterment of mankind. The upcoming seriestakes off from this point and covers the history of pulmonaryand critical care medicine in the twentieth century.

History provides us with a scale, and with a time-constantfor extrapolating into the future. From our perch in the present,we look back and see how the complexities of today evolvedfrom relatively simple roots. A panoramic perspective of thepast enables us to understand change and its significance. Thoughhistory is about the past, it is our key to true understanding ofthe present—which is simply the edge of the past.

Conflict of Interest Statement : M.J.T. is editor of AJRCCM. He receives a fixedstipend from the American Thoracic Society. He does not receive financial supportfor research from pharmaceutical, biotechnology, or medical device companies.He does not serve as a consultant to or on the advisory board of any company.

Martin J. TobinEditor

References

1. Wilson JL. History of the American Thoracic Society. Part 1: The Ameri-can Sanatorium Association. Am Rev Respir Dis 1979;119:177–184.

Is Smoking Beneficial for GranulomatousLung Diseases?

In 2000 an estimated 4.8 million deaths worldwide were attribut-able to smoking (1). The major health hazards of tobacco useare well known and include cardiovascular disease, chronic ob-structive pulmonary disease, lung cancer, and increased risk of

2. Bliss M. William Osler: A life in medicine. New York: Oxford UniversityPress; 1999. p. 281.

3. Butterfield H. The origins of modern science 1300–1800. London: G Belland Sons; 1949. p. vii, 75, 94–95, 118, 175, 180, 187.

4. Hall AR. Scientific revolution. In: Bynum WF, Browne EJ, Porter R,editors. Dictionary of the history of science. London: The MacmillanPress, Ltd; 1981. p. 378–380.

5. Ornstein M. The role of the scientific societies in the seventeenth century.Chicago: University of Chicago Press; 1928. p. 3, 50, 68, 74–75, 77–78,90, 106, 122, 156.

6. Bronowski J, Mazlish B. The Western intellectual tradition. New York:Dorset Press; 1960. p. 171, 184, 308, 323.

7. Boorstin DJ. The discoverers. New York: Random House; 1983. p. 386–394.

8. Siegelman SS. The genesis of modern science: contributions of scientificsocieties and scientific journals. Radiology 1998;208:9–16.

9. Jardine L. Ingenious pursuits: building the scientific revolution. NewYork: Doubleday; 1999.

10. Kronick DA. A history of scientific and technical periodicals: the originsand development of the scientific and technological press 1665–1790.New York: The Scarecrow Press, Inc; 1962. p. 57, 74–75, 193.

11. Ziman J. The force of knowledge: the scientific dimension of society.Cambridge: Cambridge University Press; 1976. p. 48, 131.

12. Friedman M, Friedland GW. Medicine’s 10 greatest discoveries. NewHaven: Yale University Press; 1998. p. 57–60, 81.

13. Porter R. The greatest benefit to mankind: a medical history of humanity.New York: W.W. Norton & Co; 1998. p. 246–248, 286, 287, 289, 290–291, 304, 306, 307, 314, 320, 322, 354, 355, 402, 611, 669.

14. Bonner TM. Becoming a physician: medical education in Britain, France,Germany, and the United States, 1750–1945. Baltimore: Johns Hop-kins University Press; 2000. p. 43, 106–109, 193, 196, 232–235.

15. Nuland SB. Doctors: the biography of medicine. New York: Alfred AKnopf Inc.; 1988. p. 145–170, 214, 219–228, 321–324.

16. Starr P. The social transformation of American medicine. New York:Basic Books; 1982. p. 15, 30, 72, 79–81, 106, 112, 137, 169, 223, 356.

17. Ziman J. Public knowledge: an essay concerning the social dimension ofscience. Cambridge: Cambridge University Press; 1968. p. 83, 85–86.

18. Millar D, Millar I, Millar J, Millar M. The Cambridge dictionary ofscientists. Cambridge: Cambridge University Press; 1996. p. 237–238.

19. Ziman J. An introduction to science studies: the philosophical and socialaspects of science and technology. Cambridge: Cambridge UniversityPress; 1984. p. 124–125.

20. Bonner TN. American doctors and German universities: a chapter ininternational intellectual relations 1870–1914. Lincoln: University ofNebraska Press;1987. p. 69–106.

21. Weatherall M. Drug treatment and the rise of pharmacology. In: PorterR, editor. The Cambridge illustrated history of medicine. Cambridge:Cambridge University Press; 1996. p. 246–277.

22. Ludmerer KM. Time to heal: American medical education from the turnof the century to the era of managed care. New York: Oxford Univer-sity Press; 1999. p. 188.

23. Stevens R. Issues for American internal medicine through the last cen-tury. Ann Intern Med 1986;105:592–602.

24. Stevens R. American medicine and the public interest, updated ed. Berke-ley: University of California Press; 1998. p. 113–114, 225–230.

25. Ryan F. The forgotten plague: how the battle against tuberculosis waswon—and lost. Boston: Black Bay Books. 1992. p. 24, 26, 31.

26. Daniel TM. Pioneers of medicine and their impact on tuberculosis. Roch-ester, NY: University of Rochester Press. 2000. p. 63–97.

27. Pross C. Nazi doctors, German medicine, and historical truth. In: AnnasGJ, Grodin MA, editors. The Nazi doctors and the Nuremberg Code.New York: Oxford University Press; 1992. p. 32–52.

DOI: 10.1164/rccm.2402027

infections (2). Immunosuppressive properties of tobacco smokeand effects on innate and adaptive immunity may contribute tothe development of some of these diseases (3). Recent studies,however, also suggest that the same immunosuppressive proper-

894 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

ties may have beneficial effects on some inflammatory diseases,such as hypersensitivity pneumonitis, as noted by Blanchet andcolleagues (pp. 903–909) in this issue of the Journal (3, 4).

Since the 1970s, numerous epidemiologic studies have shownthat individuals with hypersensitivity pneumonitis or sarcoidosisare less likely to smoke compared with control subjects or thepopulation at large (5–8). Other studies indicate that smokershave a lower serum antibody response to hypersensitivity pneu-monitis-inducing agents (9). Collectively, these findings suggestthat smoking may be protective or inhibit the development ofgranulomatous inflammation; this observation has not been wellacknowledged by the medical community, probably for a numberof reasons. First, the epidemiologic studies are associative anddo not indicate a definite causal link between smoking and de-creased risk of disease. Thus, decreased smoking could be anepiphenomenon or even a result of the development of respira-tory symptoms or disease. Second, these studies do not providea potential mechanism(s) by which cigarette smoke may inhibitthe development of disease. Finally, clinicians who frequentlyconfront the devastating and common clinical sequelae of to-bacco abuse may be reluctant to acknowledge that tobacco, withall of its negative effects, could reduce the risk of granulomatouslung diseases. How would we counsel our patients with thesediseases or with exposures that might increase the risk of disease?

The study by Blanchet and coworkers provides additionalsupport and biological plausibility that smoking reduces the de-velopment of granulomatous lung disease (4). Using the antigenSaccharapolysora rectivirgula to induce granulomatous inflam-mation in mouse and cell line models of disease, the investigatorsdemonstrate that simultaneous nicotine exposure reduces thebronchoalveolar lavage cellular response, including the total la-vage white blood cell and lymphocyte cell count, the extent oflung inflammation on biopsy, and interferon-� but not interleu-kin-10 mRNA expression in the lung. Using a macrophage cellline treated concomitantly with lipopolysaccharide or Sacchara-polysora rectivirgula and nicotine, decreased tumor necrosisfactor protein, and tumor necrosis factor, interleukin-10, andinterferon-� mRNA were observed.

Expanding on the epidemiologic studies, this study begins toevaluate possible mechanisms by which nicotine and, potentially,cigarette smoke may inhibit development of granulomatous lungdisease. This study and others suggest that smoke and its compo-nents affect cellular constituents in bronchoalveolar fluid, witha relative increase in macrophages and a decrease in lymphocytesand dendritic cells (3, 4, 10, 11). Relevant cellular functions arelikely impaired, as macrophages demonstrate decreased antigenphagocytosis and clearance and T cells display altered antigen-mediated signaling and proliferation, indicative of impaired cell-mediated immunity (3, 10, 12, 13). In mouse and cell line modelsfor hypersensitivity pneumonitis in this study, nicotine alteredthe pattern of cytokine production (4). This change in the Th1/Th2 cytokine balance is thought to be crucial to the developmentof granulomatous lung disease and reflects another potentialcritical immunomodulatory effect of nicotine (3, 12, 14, 15).

There are a number of questions raised by the study ofBlanchet and coworkers (4). What are the specific mechanismsvital to tobacco’s immunosuppressive effects and its ability toinhibit hypersensitivity pneumonitis and sarcoidosis? Are thereother nonimmunogenic effects of tobacco or nicotine, whichreduce disease risk or development? Is nicotine the only agentin tobacco smoke that is able to inhibit the development ofhypersensitivity pneumonitis? There are over 4,500 componentsof tobacco smoke. Acrolein, hydroquinone, catechol, and benzo-[a]pyrene have demonstrated immunosuppressive effects in celland animal models (3, 13, 15). Can other components of cigarettesmoke alone or in combination completely abolish these granulo-

matous diseases or, as is noted in mice in this study and previoushuman studies, are these effects partial? In humans, might nico-tine or cigarette smoke incompletely suppress the inflammatoryresponse but make it difficult to obtain a definitive diagnosis?What happens to those who develop disease despite smoking?Epidemiologic studies suggest that smokers with hypersensitivitypneumonitis or sarcoidosis may have a more insidious and pro-gressive form of the disease (6, 16). Long-term effects of nicotineor smoke on the mice in this model have not been reported todate.

Finally, the question we are pondering but are hesitant toask: should we recommend smoking or nicotine as therapy forgranulomatous lung diseases? Although Blanchet and coworkerssuggest that we may want to consider nicotine as treatment forhypersensitivity pneumonitis, the studies to date do not supportthis approach (4). Primary prevention of disease with cigarettesmoke or nicotine is not likely to be beneficial as there aresignificant risks associated with this therapy to treat a diseasethat develops in a fraction of those exposed. Moreover, both inthis model and in human disease, nicotine or cigarette smokedoes not completely abolish the granulomatous immune re-sponse (4–7, 11). Most importantly, we do not know whethernicotine or cigarette smoke is able to modulate granulomatousinflammation once started. The studies from humans with sar-coidosis and hypersensitivity pneumonitis would suggest other-wise: although cigarette smoke may be able to inhibit diseasedevelopment, it may have deleterious effects on disease severityor prognosis (6, 16). By understanding the precise immunomodu-lating effects of cigarette smoke and/or its components, we maybe able to use this information to halt or reverse granulomatousinflammation once started with analogs of nicotine or othercomponents of cigarette smoke. Hopefully these issues will beexplored further in this or other relevant animal models andhuman studies to help us better understand how to inhibit granu-lomatous lung diseases. Until we gain more information, wecannot be ambiguous about our recommendations for smokingcessation to our patients.

Conflict of Interest Statement : L.A.M. does not have a financial relationship witha commercial entity that has an interest in the subject of this manuscript.

Lisa A. Maier, M.D., M.S.P.H.National Jewish Medical and Research CenterDenver, Colorado

References

1. Ezzati M, Lopez AD. Estimates of global mortality attributable to smok-ing in 2000. Lancet 2003;362:847–852.

2. Centers for Disease Control and Prevention. Cigarette smoking–attributable morbidity—United States, 2000. MMWR 2003;52:842–844.

3. Sopori M. Effects of cigarette smoke on the immune system. Nat RevImmunol 2002;2:372–377.

4. Blanchet M, Israel-Assayag E, Cormier Y. Inhibitory effect of nicotineon experimental hypersensitivity pneumonitis in vivo and in vitro. AmJ Respir Crit Care Med 2004;169:903–909.

5. Warren C. Extrinsic allergic alveolitis: A disease commoner in nonsmok-ers. Thorax 1977;32:567–573.

6. Ohtsuka Y, Munakata M, Tanimura K, Ukita H, Kusaka H, Masaki Y,Doi I, Ohe M, Amishima M, Homma Y, et al. Smoking promotesinsidious and chronic farmer’s lung disease, and deteriorates the clini-cal outcome. Intern Med 1995;34:966–971.

7. Douglas JG, Middleton WG, Gaddie J, Petrie GR, Choo-Kang YF, Pres-cott RJ, Crompton GK. Sarcoidosis: a disorder commoner in non-smokers? Thorax 1986;41:787–791.

8. Newman L, Bresnitz EA, Rose C, Rossman M, Terrin ML, Barnard J,and the ACCESS Group. Environmental and occupational factorsassociated with sarcoidosis risk: A case control etiologic study of sar-coidosis. Am J Respir Crit Care Med 2001;163(5):A959.

9. Baldwin CI, Todd A, Bourke S, Allen A, Calvert JE. Pigeon fanciers’

Editorials 895

lung: effects of smoking on serum and salivary antibody responses topigeon antigens. Clin Exp Immunol 1998;113:166–172.

10. Robbins CS, Dawe DE, Goncharova SI, Pouladi MA, Drannik AG,Swirski FK, Cox G, Stampfli MR. Cigarette smoke decreases pulmo-nary dendritic cells and impacts antiviral immune responsiveness. AmJ Respir Cell Mol Biol 2004;30:202–211.

11. Valeyre D, Soler P, Clerici C, Pre J, Battesti J-P, Georges R, Hance AJ.Smoking and pulmonary sarcoidosis: effect of cigarette smoking onprevalence, clinical manifestations, alveolitis, and evolution of thedisease. Thorax 1988;43:516–524.

12. Nouri-Shirazi M, Guinet E. Evidence for the immunosuppressive role ofnicotine on human dendritic cell functions. Immunology 2003;109:365–373.

13. McCue JM, Link KL, Eaton SS, Freed BM. Exposure to cigarette tar

Surviving Pneumonia—Just a Short-Term Lease on Life?

Considering the importance of pneumonia, it is remarkable howlittle is known about what happens to patients after they recover.Pneumonia is the leading infectious killer worldwide. It was theimmediate cause of an estimated 3.9 million deaths in 2002; thenumber of deaths occurring among patients who recover froman initial episode of pneumonia is not routinely measured (1).With more than a million hospitalizations in the United Stateseach year (2), caring for a relative after an episode of pneumoniais an experience many adults can anticipate during their lifetimes.In this issue of the Journal (pp. 910–914), Waterer and coworkers(3) provide a useful contribution to a growing body of evidenceindicating that patients who survive hospitalization for pneumo-nia can expect a mortality rate that is modestly to substantiallyincreased over the subsequent one to four years. Physicians andfamilies now have the means to reduce this delayed mortality,and the current study should help prompt them to do so moreeffectively.

Other investigators have reported substantially increasedmortality after hospitalization for pneumonia (4–8). One of thestrengths of the current study is the thorough follow-up. Byenrolling patients prospectively and by using social security num-bers to rigorously review death records, contact all treating physi-cians, and trace postal contacts, Waterer and coworkers (3) wereable to ascertain the survival status of a remarkable 97% ofpatients at an average of 3 years after discharge. Compared witha matched cohort of the US population, the observed mortalityof 34% was considerably elevated relative to the expected 7%mortality in this time period. When the subset of about half ofsubjects with no comorbidities was compared with the matchedU.S. cohort, mortality among the pneumonia patients was onlymodestly elevated (see their Table 3: 11% versus 5%, p � 0.03).Other investigators studying larger cohorts have found that mor-tality is significantly increased among pneumonia survivors, evenamong those with no preexisting comorbidities (7, 9).

Minor limitations of the study by Waterer and coworkers, inaddition to the relatively small number of patients, should bekept in mind before translating their results into recommenda-tions. Immune-compromised patients and those with a recenthospitalization were excluded from the analysis—groups thatmight comprise a substantial proportion of patients with pneu-monia in some settings. Despite rigorous tracing of almost allpatients, the longest follow-up period was 4 years, leaving thelonger-term prognosis unstudied. Still, the average time to deathamong those who died in this cohort was 435 days, and otherinvestigators have recorded even longer delays (5, 8). It appearsthat an increased risk of death may persist for several years afteran episode of pneumonia. To patients and their families, this

inhibits ribonucleotide reductase and blocks lymphocyte proliferation.J Immunol 2000;165:6771–6775.

14. Matsunaga K, Klein TW, Friedman H, Yamamoto Y. Involvement ofnicotinic acetylcholine receptors in suppression of antimicrobial activ-ity and cytokine responses of alveolar macrophages to Legionellapneumophila infection by nicotine. J Immunol 2001;167:6518–6524.

15. Ouyang Y, Virasch N, Hao P, Aubrey MT, Mukerjee N, Bierer BE,Freed BM. Suppression of human IL-1beta, IL-2, IFN-gamma, andTNF-alpha production by cigarette smoke extracts. J Allergy ClinImmunol 2000;106:280–287.

16. Strom KE, Eklund AG. Smoking does not prevent the onset of respira-tory failure in sarcoidosis. Sarcoidosis 1993;10:26–28.

DOI: 10.1164/rccm.2402023

argues for continued vigilance, and attention to preventive mea-sures with sustained benefits.

Specific features of the pneumonia episode may alert clini-cians to focus particular attention on the longer-term prognosisof certain patients. In addition to patients with classic comorbiddiseases, such as cardiovascular and cerebrovascular disease,Waterer and coworkers identified patients presenting with al-tered mental status or anemia as having independently increasedmedium-term mortality, even in the absence of recognized causesfor these findings. Other investigators also have identified thesehigh-risk features (4), and clinicians would do well to scrutinizesuch patients for underlying disease, counsel their families aboutthe more guarded prognosis, and seek preventive measures thathave stood the test of time.

It is ironic that William Osler died several years after an episodeof pneumonia, and that simple preventive measures available to-day might have prolonged his life. A father of modern medicine,Osler was also a lifelong student of pneumonia, which he famouslylabeled “the old man’s friend,” and “the captain of the men ofdeath” (10). At the time of his first bout with pneumonia he wasphysically rigorous and his writings and intellect placed him atthe pinnacle of his field. The pneumococcus had been described,but there was no vaccine or specific treatment, and it was stillbelieved that influenza was caused by a small bacterium (Loeffler’sbacillus—now known as Haemophilus influenzae). After his recov-ery, Osler continued his lifelong smoking habit and he experienceda series of respiratory illnesses over the subsequent years, eventu-ally contracting an influenza-like illness in October of 1919, andthen succumbing to an apparent bacterial superinfection. He was70 years of age, and despite his own contributions to the diagnosisand treatment of pneumonia, died without benefit of a properchest radiograph, antimicrobial therapy, or surgical intervention.Thankfully, we can do better for patients today.

The prognosis for modern day patients who recover from afirst bout of pneumonia may be substantially improved by offeringan effective smoking cessation program, influenza vaccine, andpneumococcal vaccine. The most recent guidelines for the man-agement of pneumonia from the U.S. and Canadian infectiousdisease and thoracic societies already endorse such recommenda-tions, but the implementation remains poor (11, 12). Only 55%of adults 65 years or older reported receiving the pneumococcalvaccine in 2002, and the proportion was even lower among nursing-home residents (38%) and 18- to 64-year-old subjects with highrisk conditions (17%) (13, 14). The Advisory Committee on Im-munization Practices does not currently identify patients dis-charged with pneumonia as a separate high-risk category, but theaccumulated data argue that this should be reconsidered. Influenzavaccination is clearly effective at reducing pneumonia, hospitaliza-

896 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

tion, and death among the elderly in general (15), and specificallyreduces the risk of readmission and death among those with aprevious hospitalization for pneumonia (16). Standing orders arean effective means of improving vaccination rates, and are ideallysuited to protecting patients being discharged after an episode ofpneumonia (13).

The new data from Waterer and coworkers should be usedby physicians to alert patients and their families that recoveryfrom pneumonia may provide only a short-term lease on a futurehealthy life. This sobering message may be accompanied byrecommendations on proven approaches to moderate the risk.All physicians caring for pneumonia patients should review theirinstitution’s standing orders policies, and work to include influ-enza vaccination, pneumococcal vaccination, and a smoking ces-sation program in standing discharge orders for patients whohave recovered from pneumonia.

Conflict of Interest Statement : S.F.D. has no declared conflict of interest.

Scott F. Dowell, M.D., M.P.H.Thai Ministry of Public Health and

U.S. Centers for Disease Controland Prevention

Nonthaburi, Thailand

References

1. WHO. The World Health Report. Geneva: World Health Organization;2002.

2. Peters K, Kochanek D, Murphy S. Deaths: final data for 1996. Natl VitalStat Rep 1998;47:1–100.

3. Waterer GW, Kessler LA, Wunderink RG. Medium-term survival afterhospitalization with community-acquired pneumonia. Am J Respir CritCare Med 2004;169:910–914.

4. Brancati FL, Chow JW, Wagener MM, Vacarello SJ, Yu VL. Is pneumo-nia really the old man’s friend? Two-year prognosis after community-acquired pneumonia. Lancet 1993;342:30–33.

5. Koivula I, Sten M, Makela PH. Prognosis after community-acquiredpneumonia in the elderly: a population-based 12-year follow-up study.Arch Intern Med 1999;159:1550–1555.

6. Hedlund JU, Ortqvist AB, Kalin ME, Granath F. Factors of importancefor the long term prognosis after hospital treated pneumonia. Thorax1993;48:785–789.

7. Kaplan V, Clermont G, Griffin MF, Kasal J, Watson RS, Linde-ZwirbleWT, Angus DC. Pneumonia: still the old man’s friend? Arch InternMed 2003;163:317–323.

8. Mortensen EM, Kapoor WN, Chang CC, Fine MJ. Assessment of mortalityafter long-term follow-up of patients with community-acquired pneumo-nia. Clin Infect Dis 2003;37:1617–1624.

9. Carriere KC, Jin Y, Marrie TJ, Predy G, Johnson DH. Outcomes andcosts among seniors requiring hospitalization for community-acquiredpneumonia in alberta. J Am Geriatr Soc 2004;52:31–38.

10. Bliss M. William Osler: a life in medicine. Toronto: University of TorontoPress; 1999.

11. Mandell L, Marrie T, Grossman R, Chow A, Hyland R. Canadian Guide-lines for the Initial Management of Community-Acquired Pneumonia:an evidence-based update by the Canadian Infectious Diseases Societyand the Canadian Thoracic Society. Clin Infect Dis 2000;31:383–421.

12. Mandell LA, Bartlett JG, Dowell SF, File TM Jr, Musher DM, WhitneyC. Update of practice guidelines for the management of community-acquired pneumonia in immunocompetent adults. Clin Infect Dis 2003;37:1405–1433.

13. CDC. Facilitating influenza and pneumococcal vaccination throughstanding orders programs. MMWR Morb Mortal Wkly Rep 2003;52:68–69.

14. Whitney CG, Schaffner W, Butler JC. Rethinking recommendations foruse of pneumococcal vaccines in adults. Clin Infect Dis 2001;33:662–675.

15. Gross PA, Hermogenes AW, Sacks HS, Lau J, Levandowski RA. Theefficacy of influenza vaccine in elderly persons. A meta-analysis andreview of the literature. Ann Intern Med 1995;123:518–527.

16. Herzog NS, Bratzler DW, Houck PM, Jiang H, Nsa W, Shook C, Weingar-ten SR. Effects of previous influenza vaccination on subsequent read-mission and mortality in elderly patients hospitalized with pneumonia.Am J Med 2003;115:454–461.

DOI: 10.1164/rccm.2402025

Occasional Essay

A Century of Pulmonary Gas ExchangeJohn B. West

Department of Medicine, University of California San Diego, La Jolla, California

The period from 1905 to 2004 was a particularly important era inthe history of pulmonary gas exchange. The dramatic increasein knowledge from rudimentary ideas at the beginning of the 20thcentury to today’s sophisticated analyses might be comparedwith the progress from the Wright flyer at Kitty Hawk in Decem-ber 1903 to modern supersonic aircraft. However, I cannot resistthe temptation to peek back one page to the beginning of the19th century because in some ways that sets the stage for 1905.

THE SETTING

Antoine Laurent Lavoisier (1743–1794) had the distinction ofidentifying the three respiratory gases in 1777: “Eminently respi-rable air [he later called it oxygine] that enters the lung, leavesit in the form of chalky aeriform acids [CO2] . . . in almost equalvolume. . . . Respiration acts only on the portion of pure air thatis eminently respirable . . . the excess, that is the mephitic portion[nitrogen], is a purely passive medium. . . . The respirable portionof air has the property to combine with blood and its combinationresults in its red color” (1). Subsequently, Lavoisier emphasizedthe similarity between respiration and combustion stating that“respiration is nothing but a slow combustion of carbon andhydrogen, similar in all respects to that of a lamp or a lightedcandle, and from this point of view, animals which breathe areequally combustible substances burning and consuming them-selves.” Incidentally, the English novelist Charles Dickens wasarguably influenced by this when he had one of his charactersin Bleak House die by spontaneous combustion (2).

However, Lavoisier with his colleague Laplace made onemajor error when they stated that the combustion (oxidation)took place in the lung itself. In fact, it is remarkable that theidentification of where the actual energy metabolism occurredproved to be extremely elusive, being a central problem in physi-ology for much of the 19th century. It was only in the 1870s thatEduard Pfluger (1829–1910) and his coworkers showed conclu-sively that metabolism takes place in peripheral tissues and thatthe blood simply transports the respiratory gases. By 1905, thegas exchange function of the lung was firmly established, andthe carriage of oxygen and carbon dioxide by the blood hadbeen largely worked out, although for example, the effects ofpH and temperature on the oxygen dissociation curve were notelucidated until the 2nd decade (3, 4), and it was not until as

(Received in original form December 30, 2003; accepted in final form January 26, 2004)

Supported by National Institutes of Health grant RO1 HL 60698.

Correspondence and requests for reprints should be addressed to John B. West,M.D., Ph.D., UCSD Department of Medicine 0623A, 9500 Gilman Drive, La Jolla,CA 92093–0623. E-mail: jwest@ucsd.edu

Am J Respir Crit Care Med Vol 169. pp 897–902, 2004DOI: 10.1164/rccm.200312-1781OEInternet address: www.atsjournals.org

late as 1967 that the role of 2–3 diphosphoglycerate on theoxygen affinity of hemoglobin was appreciated (5, 6).

OXYGEN SECRETION

One of the most colorful controversies in the first decade of the20th century concerned how oxygen moved across the pulmo-nary capillary wall into the blood. Was this by passive diffusionor active secretion? Christian Bohr (1855–1911) (Figure 1) was amajor proponent of the secretion ability of the lung, and in 1909,he referred to this as the lung’s “specific function” (7), althoughoddly enough in the same article he developed the mathematicalbasis for diffusion of oxygen across the pulmonary capillary, nowknown as the “Bohr integration.” Incidentally, Bohr has the un-usual distinction of having his name attached to three differentareas of pulmonary gas exchange: the Bohr integration, the Bohrdead space, and the Bohr effect (reduction of the oxygen affinity ofhemoglobin caused by an increase in Pco2). J.S. Haldane (1860–1936) visited Bohr’s laboratory and became one of the championsof oxygen secretion stating for example, “In the animals investi-gated the normal oxygen tension in the arterial blood is alwayshigher than the alveolar air, and in some animals higher thanthe inspired air. The absorption of oxygen by the lungs thuscannot be explained by diffusion alone” (8).

Haldane subsequently led the influential Anglo-AmericanExpedition to Pikes Peak, Colorado, in 1911 and believed thathe obtained further evidence for oxygen secretion (9). Eventoday it is not clear precisely where the measurements wereerroneous. In fact, Haldane continued to believe in oxygen secre-tion until his death in 1936 in spite of mounting evidence againstthe theory, and in the second edition of his book Respiration,he devoted an entire chapter to the subject (10). Haldane pointedout that secretion of several substances against concentrationgradients (i.e., by active transport) occurs in many glands andthat in the swim bladder of fishes the Po2 is often much higherthan in the surrounding water. Because the swim bladder, similarto the lung, is an outgrowth of the gut, he reasoned that oxygensecretion could be expected.

August Krogh (1874–1949) (Figure 2) was one of the mostarticulate opponents of the secretion theory, and there is a mem-orable passage at the beginning of one of his articles that waspublished a year before Bohr died (11). Krogh developed anaccurate tonometer in which a small air bubble was equilibratedwith flowing blood and showed that the arterial Po2 was alwayslower than the alveolar value in a variety of animal experiments.Because Bohr had been one of the most ardent supporters of thesecretion theory and Krogh was his student, Krogh’s introductionrequired all of the tact that he could muster: “I shall be obligedin the following pages to combat the views of my teacher Prof.Bohr on certain essential points and also to criticize a few ofhis experimental results. I wish here not only to acknowledgethe debt of gratitude which I, personally, owe to him, but alsoto emphasize the fact . . . that the real progress, made duringthe last twenty years in the knowledge of the processes in the

898 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

Figure 1. Christian Harald Lauritz Peter Emil Bohr (1855–1911). Thiseminent Danish physiologist is remembered in the Bohr dead space,Bohr effect, and Bohr integration. He strongly believed in oxygen secre-tion by the lung. Courtesy of the Medical History Museum, Universityof Copenhagen, Denmark.

lungs, is mainly due to his labours” (11). Subsequently, JosephBarcroft (1872–1947) led an expedition to Cerro de Pasco, Peruin 1921–1922 and showed that the arterial Po2 was always lessthan the alveolar value in humans (12). He also made the impor-tant observation that the arterial oxygen saturation fell duringexercise at high altitude and argued that this could be explainedby the failure of equilibration of Po2 between alveolar gas andpulmonary capillary blood. This was one of the first direct dem-onstrations of diffusion limitation in normal lungs at high alti-tude, a finding that has been confirmed many times since.

DIFFUSING CAPACITY OF THE LUNG

One of the reasons for the theory of oxygen secretion was thatit seemed impossible to explain oxygen transfer by the lung athigh altitude. For example, in 1909, which was the same year whenBohr’s classic article, cited previously here, appeared, the Duke ofthe Abruzzi reached the extraordinary altitude of 7,500 m in theKarakorum mountains without supplementary oxygen. This as-tonished alpinists and physiologists alike. Only a few years be-fore, an experienced mountaineer had reported that 21,500 ft(6,500 m) was “near the limit at which man ceases to be capableof the slightest further exertion” (13). Haldane and his colleaguescalculated the alveolar Po2 of the Duke to be only 30 mm Hg,and they concluded that adequate oxygenation of the bloodwould be impossible based on passive diffusion, and therefore,oxygen secretion must have occurred (9).

Figure 2. August Krogh (1874–1949) and his wife Marie (1874–1943).August made important contributions to several areas of pulmonarygas exchange and argued strongly against oxygen secretion. Marie Kroghdeveloped the measurement of diffusing capacity of the lung usingcarbon monoxide. Courtesy of Dr. Bodil Schmidt-Nielsen.

However, August Krogh’s wife, Marie (Figure 2), developeda method for measuring the diffusing capacity of the lung usinginhaled carbon monoxide. Her technique was essentially thesame as the single-breath method employed in pulmonary func-tion laboratories today. Her measurements showed that Haldaneand his colleagues had markedly underestimated the diffusingcapacity of the lung, and therefore, it was not necessary to resortto the oxygen diffusion hypothesis. Marie Krogh’s method ofmeasuring the diffusing the diffusing capacity did not becomepracticable for clinical work because of the difficulty of measur-ing the carbon monoxide concentration. However, with the intro-duction of the infrared CO meter, which was developed duringWorld War II, there was renewed interest in the diffusing capac-ity, and Bates developed a steady-state technique (14), whilethe original Krogh single-breath method was slightly modifiedand popularized (15). Subsequently, Roughton and Forster (16)showed that it was possible to separate the two components de-termining uptake of oxygen across the blood–gas barrier; thatcaused by the diffusive properties of the alveolar membrane onthe one hand and that due to the finite rate of uptake of oxygenby the hemoglobin in the red blood cell. The measurement ofthe diffusing capacity of the lung using carbon monoxide becamean important test in the pulmonary function laboratory and isstill extensively used today. A colorful application was the dem-onstration of an increased diffusing capacity in orbiting astro-nauts resulting from the movement of blood from dependent re-gions of the body into the lung as a result of the weightlessness (17).

VENTILATION/PERFUSION INEQUALITY

We can now turn from oxygen secretion and diffusion to anothermechanism that is fundamental to understanding pulmonary gasexchange, that is, ventilation–perfusion inequality. One of thefirst realizations that the gas exchange that takes place in anylung unit is determined by the ratio of ventilation to blood flowwas that by Krogh and Lindhard (18) when they wrote this: “Ifthe different lobes of the lungs are not equally dilated duringinspiration the air in them must obtain a different compositionand this must be true both with regard to O2 and CO2 during

Occasional Essay 899

normal breathing and with regard to other gases during specialmixing respirations.” They then added in a prescient footnotethis: “Unless, indeed, the circulation through each lobe shouldbe in proportional to its ventilation.” A little later Haldane (19)stated that mismatching of ventilation and blood flow was apotential cause of hypoxemia, but unfortunately, he concludedthat it would not cause carbon dioxide retention, a serious mis-conception that still surfaces from time to time.

However, the key advances in understanding ventilation–perfusion inequality came out of World War II, and the circum-stances are fascinating. One principal group led by Wallace Fenn(1893–1971) was in the Department of Physiology at the Univer-sity of Rochester, New York. They included Fenn himself whowas working on muscle contraction and potassium movementacross cell membranes, Hermann Rahn who was developing abioassay method in frogs for pituitary hormones, and ArthurOtis who was studying the activation of the enzyme tyrosinasein grasshopper eggs (Figure 3). This seems like an unlikely groupto revolutionize pulmonary gas exchange. Because of the de-mands of war, however, Fenn was asked by the U.S. Air Forceto investigate the physiologic effects of pressure breathing at highaltitude in the hope of improving the performance of airmen. Theresult was that although none of the three physiologists weretrained in human physiology and, it is alleged, were vague aboutthe definition of residual volume, in a few years, they had madefundamental discoveries in both pulmonary gas exchange andmechanics.

The other source was perhaps equally surprising. At the U.S.Naval School of Aviation Medicine in Pensacola, Joseph Lilien-thal was investigating possible carbon monoxide poisoning inpilots as an explanation for the high frequency of fatal accidentsduring training. To measure CO levels in blood, he was using themicrosyringe analyzer developed by Scholander and Roughton

Figure 3. From right to left: Wallace Fenn (1893–1971), Hermann Rahn(1912–1990), and Arthur Otis (1913–). Not only did they make keydiscoveries in pulmonary gas exchange, especially ventilation–perfusionrelationships, but they also laid the foundation of modern respiratorymechanics. Reprinted by permission from Otis AB, Rahn H. Developmentof concepts in Rochester, New York in the 1940s. In: West JB, editor.Pulmonary gas exchange. Vol. 1. Ventilation, blood flow and diffusion.New York: Academic Press; 1980.

(20) whereupon Richard Riley (1911–2001) (Figure 4), who wasworking across the hall, saw the syringes and wondered whetherit might be possible to determine the Po2 and Pco2 of arterialblood by equilibrating a small bubble of gas with the blood. Hewas successful in this (21) and thus developed his interest in themechanisms of hypoxemia in lung disease and the role of ventila-tion–perfusion inequality. There is a nice anecdote here thatresulted from Riley developing pulmonary tuberculosis in 1948.Fortunately, he was allowed to rest at home but of course hada good deal of time on his hands. His story is that he spent muchof his time exploring the intricacies of ventilation–perfusion rela-tionships, manipulating his four-quadrant diagram and occasion-ally contacting Rahn by mail. Riley remarked that “never wasforced confinement given more profitable psychotherapy” (22).An interesting aside is that although many of us always regardRiley as a giant in the area of pulmonary gas exchange, he alwaysclaimed that his major research contribution was on the airbornetransmission of pulmonary tuberculosis (23).

It is remarkable that the two groups worked essentially inde-pendently, although they communicated from time to time. Be-cause the Rochester group (who later continued the work in Buf-falo) started with the problem of pressure breathing, they tendedto emphasize the gas side of the blood–gas barrier and, for exam-ple, developed the enormously powerful oxygen–carbon dioxidediagram (24). This was initially used to analyze the effects ofhigh altitude, hyperventilation, and oxygen breathing on alveolar

Figure 4. Richard Riley (1911–2001). His contributions to pulmonarygas exchange included the bubble technique for measuring arterial Po2

and Pco2, the concept of ideal alveolar Po2, and the three-compartmentmodel. He also did important work on the airborne transmission ofpulmonary tuberculosis.

900 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

gas composition, although the group certainly recognized that itcould also be used to predict changes in the arterial blood gases.In contrast, Riley who came from a clinical training in AndreCournand’s laboratory at Bellevue Hospital in New York Citywas particularly interested in the causes of hypoxemia in lungdisease, and he and his colleagues tended to concentrate on theblood side of the barrier. They developed the elegant four-quadrant diagram based on the oxygen and carbon dioxide disso-ciation curves and went on to derive the three-compartmentmodel of pulmonary gas exchange where one compartment is“ideal” in the sense that gas exchange is optimal, a second com-partment has unperfused alveoli, and the third has unventilatedalveoli. This model was the gold standard for assessing ventila-tion–perfusion inequality in patients with lung disease until theintroduction of the multiple inert gas elimination technique,which allowed distributions of ventilation–perfusion ratios to bedescribed (25).

One of the reasons that diagrams such as the O2–CO2 diagramof Rahn and Fenn (24) and the four-quadrant diagram of Rileyand Cournand (26) were so valuable is that the oxygen andcarbon dioxide dissociation curves are not only nonlinear butalso interdependent. The result is that the equations relating Po2

and Pco2 to the ventilation–perfusion ratio cannot be solvedalgebraically, and therefore, graphical methods had to be em-ployed. An important breakthrough occurred in the mid-1960swhen first Kelman (27) and then Olszowka and Farhi (28) intro-duced digital computer procedures for describing the oxygenand carbon dioxide dissociation curves. Perhaps I can be alloweda little personal recollection here. When I first saw Kelman’ssubroutines demonstrated at a meeting of the Physiological Soci-ety in the United Kingdom in the early 1960s, I realized that thismeant that the tedious graphic methods of analyzing ventilation–perfusion relationships would soon be supplanted by computeranalyses. However, when I subsequently developed a computermodel of the lung containing ventilation–perfusion inequality dur-ing a sabbatical year at the NASA Ames Research Center (29),we had a visit from a Presidential Science Advisory Committee,including Wallace Fenn himself, and I proudly showed him acomputer solution for a ventilation–perfusion ratio equation.Much to my chagrin, he seemed to be completely unimpressed,but these computer techniques eventually led to the developmentof the multiple inert gas elimination technique (25). Here, amajor advance was the realization that the normal respiratorygases, oxygen and carbon dioxide, have a very limited ability togive information about patterns of ventilation–perfusion in-equality, and much more information became available whenthe gas exchange of a series of inert gases having a range ofsolubilities was exploited. However, the multiple inert gas-elimi-nation technique only became feasible because of the mathe-matic expertise of Peter Wagner and his colleague John Evans(30) but has now become the gold standard for assessing ventila-tion–perfusion inequality.

ARTERIAL BLOOD GASES

The multiple inert gas elimination technique for determiningdistributions of ventilation–perfusion ratios primarily remains aresearch tool because of its complexity. The main armament inthe trenches for assessing abnormal pulmonary gas exchangeremains the measurement of the arterial blood gases with blood–gas electrodes, and their development constitutes one of the mostimportant advances of the last 100 years. In fact, when I tell med-ical students that when I was a young resident we could not

measure arterial Po2, Pco2, or pH, they stare at me in disbelief.Those were the days when considerable importance was givento the detection of cyanosis as an index of impaired gas exchange.Who knows how many patients with chronic obstructive pulmo-nary disease died when they were given oxygen to relieve theircyanosis and they developed lethal carbon dioxide retention.

One of the first measurements of the amounts of oxygen andcarbon dioxide in blood was by Gustav Magnus (1802–1870),who used his new mercury pump to expose the blood to a partialvacuum (31). His demonstration that arterial blood had moreoxygen and less carbon dioxide than venous blood helped todispel the erroneous notion that the metabolic combustion (oxi-dation) took place in the lungs. Improvements in gasometricanalysis of blood were made by Carl Ludwig and Eduard Pfluger,but the methods were not used in clinical medicine. A majoradvance was the development of arterial puncture because be-fore this it was not practicable to obtain arterial blood. The firstarterial punctures in humans were made by Hurter (32), and hewas able to show that the arterial oxygen saturation in fournormal subjects was between 93% and 100%. However, thesignificance of his work was overlooked until Stadie introducedthe technique at the Rockefeller Institute where he investigatedthe relationship between arterial oxygen saturation and cyanosisin patients with pneumonia (33). These early measurements ofarterial oxygen saturation were made by vacuum extraction ofthe blood gases and their subsequent manometric or volumetricanalysis, as described by Van Slyke and O’Neill (34). Referencehas already been made to the Riley bubble technique for measur-ing both arterial Po2 and Pco2, and this gave a great stimulus tothe understanding of pulmonary gas exchange. However, thetechnique was technically demanding, and many potential “bub-blers” were forced to make the pilgrimage to Johns Hopkins tolearn the intricacies at firsthand.

The measurement of arterial Po2 was revolutionized by theintroduction of the polarographic oxygen electrode. The firstof these was the “dropping mercury cathode,” which workedbecause the continuously renewed surface of the drops avoidedinactivation of the electrode by proteins in the blood. A droppingmercury electrode was successfully used by Berggren (35) tomeasure alveolar–arterial Po2 differences in a series of normalsubjects. However, the device was difficult to use and neverbecame a clinical instrument. The breakthrough was the intro-duction of the platinum electrode by Leland Clark and col-leagues (36). In the original version, a small platinum electrodewas covered with cellophane and immersed in a sample of blood.However, these early devices suffered from errors caused byoxygen depletion near the electrode unless the blood was rapidlystirred. This problem was later avoided by using a very smallelectrode tip.

Shortly after the development of a clinically useful oxygenelectrode, Severinghaus and Bradley (37) described an electrodefor measuring Pco2 in blood. The principle was that carbondioxide diffused from the blood through a Teflon membraneinto a small volume of electrolyte solution in which the pH wasmeasured with a glass electrode. Both the Po2 and Pco2 electrodeswere incorporated into a common thermostat. In fact, Stow andcoworkers (38) had previously demonstrated the possibility ofmeasuring the Pco2 of blood by wrapping a thin rubber mem-brane over a film of distilled water around a glass pH electrode,but this suffered from severe instability. This problem was over-come by Severinghaus and Bradley when they added bicarbonateto the electrolyte.

The clinical measurement of blood pH predated both theoxygen and carbon dioxide electrodes. An important stimulushere was the polio epidemics of 1950–1953, which resulted inpatients with bulbar polio being treated by mechanical ventilation

Occasional Essay 901

in “respirators,” as they were then known. These were heroicdays before sufficient ventilators were available, and teams ofmedical students and others were employed to compress rubberbags manually during successive shifts over the 24 hours. It wasanyone’s guess what the arterial Pco2 was under these conditions.Poul Astrup in Copenhagen first computed the Pco2 from theHenderson–Hasselbalch equation by combining the plasma pHand CO2 concentration, the latter being determined in a VanSlyke apparatus (34). Actually, the first blood pH electrode wentback to the 1920s and 30s, but it was too awkward to be usedclinically. Astrup subsequently developed a simpler method formeasuring Pco2 by using the linear relationship between bloodpH and log Pco2, this being generated by equilibrating a plasmasample with gases of differing CO2 concentrations. As a result,the Pco2 could be read from the measured pH. This work ledto the introduction of the so-called standard bicarbonate, thatis, the bicarbonate concentration at a normal Pco2 (originallyobtained by having the technician exhale over the sample!) andlater the base excess (39).

OXIMETRY

Modern blood–gas electrodes play a critical role in a modernclinical respiratory laboratory, particularly in the intensive careunit. However, the relatively invasive arterial puncture is nowsupplemented by measurements of arterial oxygen saturation byoximetry. The principle of this goes back to Hoppe-Seyler, whocrystallized and named hemoglobin and showed that oxygenchanged its color (40). Various devices to measure the color ofthe blood were developed over the years, particularly by Millikanin the United States and Kramer in Germany during World WarII. Millikan (41) described an ear oximeter apparently largelybased on German research, and this was subsequently improvedby Earl Wood who added a pressure device to squeeze bloodout of the ear and thus obtain a zero setting. The arterial oxygensaturation was measured using a string galvanometer, and on apersonal note, I made a large series of measurements with thisvery fragile device in the Silver Hut expedition in the Himalayasat an altitude of 5,800 m in 1960–1961 (42).

The ear oximeter was cumbersome, and a big advance wasmade by Takuo Aoyagi in 1972 when he developed the pulseoximeter, which could be used on the finger (43). This deviceis now used very extensively in the clinical environment withenormous value.

CONCLUSIONS

Gas exchange is the primary function of the lung, and manypatients with pulmonary disease have impaired gas exchangethat can progress to respiratory failure and death. It is thereforevery satisfying to be able to review the enormous advances thathave taken place in the last 100 years. Nevertheless, it is worthpointing out that important areas of ignorance remain. For exam-ple, a critically ill patient in the intensive care setting often hasgrossly disturbed pulmonary gas exchange, but many aspects ofthis are poorly understood. Just to take one example, alteringthe levels of positive end-expiratory pressure and/or the inspiredoxygen concentration will often improve the arterial Po2, butmany of the changes within the lung after these interventions areobscure. Furthermore, with the modern emphasis on molecularbiology, much of the attention that was devoted to pulmonarygas exchange has moved away, and an entire generation of youngpulmonary physicians are sadly ignorant of this area. Presum-ably, the pendulum will swing back in due course, and moreinterest will be directed at this critically important and fascinatingarea of pulmonary medicine.

Conflict of Interest Statement : J.B.W. has no declared conflict of interest.

References

1. Lavoisier AL. Experiences sur la respiration des animaux, et sur leschangements qui arrivent a l’air en passant par leur poumon; 1777.[Reprinted in Lavoisier AL. Oeuvres de Lavoisier. Paris: ImprimerieImperial; 1862–1893.]

2. West JB. Spontaneous combustion, Dickens, Lewes, and Lavoisier. NewsPhysiol Sci 1994;9:276–278.

3. Barcroft J. The respiratory function of the blood. Cambridge: UniversityPress; 1914.

4. Christiansen J, Douglas CG, Haldane JS. The absorption and dissocia-tion of carbon dioxide by human blood. J Physiol (London) 1914;48:244–271.

5. Chanutin A, Curnish RR. Effect of organic and inorganic phosphataseson the oxygen equilibrium of human erythrocytes. Arch Biochem Bio-phys 1967;121:96–102.

6. Benesch R, Benesch RE. The effect of organic phosphates from thehuman erythrocyte on the allosteric properties of hemoglobin. Bio-chem Biophys Res Commun 1967;26:162–167.

7. Bohr C. Uber die spezifische Tatigkeit der Lungen bei der respira-torischen Gasaufnahme und ihr verhalten zu der durch die Alveo-larwand stattfindenden Gasdiffusion. Skand Arch Physiol 1909;22:221–280. [English translation in: West JB, editor. Translations in respiratoryphysiology. Stroudsburg, PA: Dowden, Hutchinson & Ross; 1975,pp. 691–735.]

8. Haldane JS, Lorrain Smith J. The absorption of oxygen by the lungs.J Physiol (London) 1897;22:231–258.

9. Douglas CG, Haldane JS, Henderson Y, Schneider EC. Physiologicalobservations made on Pike’s Peak, Colorado, with special referenceto adaptation to low barometric pressures. Philos Trans R Soc LondB Biol Sci 1913;203:185–381.

10. Haldane JS, Priestley JG. Respiration, 2nd ed. London, New York: Ox-ford University Press (Clarendon); 1935.

11. Krogh A. On the oxygen-metabolism of the blood. Skand Arch Physiol1910;23:192–199.

12. Barcroft J, Binger CA, Bock AV, Doggart JH, Forbes HS, Harrop G,Meakins JC, Redfield AC. Observations upon the effect of high alti-tude on the physiological processes of the human body, carried outin the Peruvian Andes, chiefly at Cerro de Pasco. Philos Trans R SocLond B Biol Sci 1923;211:351–480.

13. Hinchliff TW. Over the sea and far away. London: Longmans Green;1876.

14. Bates DV, Boucot NG, Dormer AE. Pulmonary diffusing capacity innormal subjects. J Physiol (London) 1955;129:237–252.

15. Ogilvie CM, Forster RE, Blakemore WS, Morton JW. A standardizedbreath holding technique for the clinical measurement of the diffusingcapacity of the lung for carbon monoxide. J Clin Invest 1957;36:1–17.

16. Roughton FJW, Forster RE. Relative importance of diffusion and chemi-cal reaction rates in determining the rate of exchange of gases inthe human lung, with special reference to true diffusing capacity ofpulmonary membrane and volume of blood in the lung capillaries.J Appl Physiol 1957;11:290–302.

17. Prisk GK, Guy HJ, Elliott AR, Deutschman RA III, West JB. Pulmonarydiffusing capacity, capillary blood volume and cardiac output duringsustained microgravity. J Appl Physiol 1993;75:15–26.

18. Krogh A, Lindhard J. The volume of the dead space in breathing and themixing of gases in the lungs of man. J Physiol (London) 1917;51:59–90.

19. Haldane JS. Respiration. New Haven: Yale University Press; 1922.20. Scholander PF, Roughton FJW. Micro gasometric estimation of the blood

gases: II: carbon monoxide. J Biol Chem 1943;148:551–563.21. Riley RL, Proemmel DD, Franke RE. A direct method for determination

of oxygen and carbon dioxide tensions in blood. J Biol Chem 1945;161:621–633.

22. Riley RL. Development of the three-compartment model for dealingwith uneven distribution. In: West JB, editor. Pulmonary gas exchange:ventilation, blood flow, and diffusion, Vol. 1. New York: AcademicPress; 1980. p. 67–85.

23. Riley RL, Mills CC, Nyka W, Weinstock N, Storey PB, Sultan LU, RileyMC, Wells WF. Aerial dissemination of pulmonary tuberculosis: a two-year study of contagion in a tuberculosis ward. 1959. Am J Epidemiol1995;142:3–14.

24. Rahn H, Fenn WO. A graphical analysis of the respiratory gas exchange.Washington, DC: American Physiological Society; 1955.

25. Wagner PD, Saltzman HA, West JB. Measurement of continuous distri-

902 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

butions of ventilation-perfusion ratios: theory. J Appl Physiol 1974;36:533–537.

26. Riley RL, Cournand A. “Ideal” alveolar air and the analysis of ventilation-perfusion relationships in the lungs. J Appl Physiol 1949;1:825–847.

27. Kelman GR. Digital computer subroutine for the conversion of oxygentension into saturation. J Appl Physiol 1966;21:1375–1376.

28. Olszowka AJ, Farhi LE. A system of digital computer subroutines forblood gas calculations. Respir Physiol 1968;4:270–280.

29. West JB. Ventilation-perfusion inequality and overall gas exchange incomputer models of the lung. Respir Physiol 1969;7:88–110.

30. Evans JW, Wagner PD. Limits on VA/Q distribution from analysis ofexperimental gas elimination. J Appl Physiol 1977;42:889–898.

31. Magnus HG. Ueber die im blute enthaltenen gase, sauerstoffe, stickstoff,und kohlensaure. Ann Phys Chem (Leipzig) 1837;12:583–606.

32. Hurter. Untersuchungen am arteriellen menschlichen Blute. Dtsch ArchKlin Med 1912;108:1–34.

33. Stadie WC. The oxygen of the arterial and venous blood in pneumoniaand its relationship to cyanosis. J Exp Med 1919;30:215–240.

34. Van Slyke DD, O’Neill JM. The determination of blood gases in bloodand other solutions by vacuum extraction and manometric measure-ment. J Biol Chem 1924;61:523–573.

35. Berggren S. The oxygen deficit of arterial blood caused by non-ventilatingparts of the lung. Acta Physiol Scand 1942;4(Suppl 11).

36. Clark LC, Wolf R, Granger D, Taylor Z. Continuous recording of bloodoxygen tension by polarography. J Appl Physiol 1953;6:189–193.

37. Severinghaus JW, Bradley AF. Electrodes of blood Po2 and Pco2 determi-nation. J Appl Physiol 1958;13:515–520.

38. Stow R, Baer RF, Randall B. Rapid measurement of the tension of car-bon dioxide in blood. Arch Phys Med Rehabil 1957;38:646–650.

39. Siggaard-Andersen O, Engel K, Jørgensen K, Astrup P. A micro methodfor determination of pH, carbon dioxide tension, base excess andstandard bicarbonate in capillary blood. Scand J Clin Lab Invest 1960;12:172–176.

40. Hoppe-Seyler F. Uber das verhalten des blutfarbstoffes im spektrum dessonnenlichtes. Virchows Arch Path Anat Physiol 1862;23:446.

41. Millikan GA. The oximeter: an instrument for measuring continuouslyoxygen saturation of arterial blood in man. Rev Sci Instrum 1942;13:434–444.

42. West JB, Lahiri S, Gill MB, Milledge JS, Pugh LG, Ward MP. Arterialoxygen saturation during exercise at high altitude. J Appl Physiol1962;17:617–621.

43. Aoyagi T. Pulse oximetry: its invention, theory, and future. J Anesth 2003;17:259–266.

Inhibitory Effect of Nicotine on ExperimentalHypersensitivity Pneumonitis In Vivo and In VitroMarie-Renee Blanchet, Evelyne Israel-Assayag, and Yvon Cormier

Unite de Recherche, Centre de Pneumologie, Institut Universitaire de Cardiologie et de Pneumologie de l’Universite Laval, Hopital Laval,Quebec City, Quebec, Canada

The incidence of hypersensitivity pneumonitis (HP) is lower in smok-ers than in nonsmokers. Because nicotine is immunosuppressive,we hypothesized that it could have a protective effect on HP induc-tion in vivo. HP was induced in mice that were treated with nicotineeither intraperitoneally (IP) (0.5 to 2.0 mg/kg/day) or intranasally(IN) (0.025 to 2.0 mg/kg/day). Both IP- and IN-treated animals hadfewer bronchoalveolar lavage total cells and lymphocytes and adecreased lung tissue inflammation. IFN-� but not interleukin-10mRNA expression was reduced in lung tissue of 2.0-mg/kg IN-treated animals. To test the effect of nicotine on alveolar macro-phages, AMJ2-C11 cells were treated with nicotine and stimulatedwith lipopolysaccharide or Saccharopolyspora rectivirgula, a caus-ative agent of HP. Nicotine reduced tumor necrosis factor releaseand tumor necrosis factor, interleukin-10, and IFN-� mRNA expres-sion after stimulation and decreased CD80 expression by 55% inlipopolysaccharide-stimulated cells and by 41% in S. rectivirgula–stimulated cells. We conclude that nicotine could be, at least inpart, responsible for the protection observed in smokers againstHP. The inhibitory effect of nicotine on alveolar macrophages couldbe one of the mechanisms involved.

Keywords: alveolar macrophages; nicotine; hypersensitivity pneumoni-tis; cytokines

Nicotine, a tertiary cholinomimetic alkaloid, is a major compo-nent of cigarette smoke. This molecule is highly lipophilic andthus has the capacity to cross the blood/brain barrier and causeaddiction (1). This nicotinic agonist has some immunomodula-tory and antifibrotic effects. In fact, it inhibits lymphocyte prolif-eration (2), interleukin (IL)-1�, tumor necrosis factor (TNF),IL-6 and IL-12 production by macrophages (3, 4), the expressionof costimulatory molecules such as CD28 and CTLA-4 on Tcells (5), and fibroblast proliferation (6). Nicotine is effective inthe treatment of ulcerative colitis, an intestinal inflammatorydisease (7), and has been shown to have beneficial effects in atype 1 diabetes mouse model (8).

Interestingly, certain inflammatory diseases, such as sarcoido-sis and ulcerative colitis are less frequent in smokers than innonsmokers, and cigarette smoking protects against radiationpneumonitis (9, 10). When exposed to an environment that cancause hypersensitivity pneumonitis (HP), smokers have lowerlevels of specific antibodies to the causative antigen (11). Onthe other hand, when HP does occur in smokers, it promotesan insidious and more chronic form of the disease and worsensthe clinical outcome (12).

HP is a pulmonary inflammatory disease characterized by theaccumulation and proliferation of lymphocytes in the lung (13).This disease is caused by an immune reaction to inhaled antigens

(Received in original form October 8, 2002; accepted in final form December 30, 2003)

Correspondence and requests for reprints should be addressed to Yvon Cormier,M.D., Hopital Laval, 2725 Chemin Ste-Foy, Quebec City, PQ, G1V 4G5 Canada.E-mail: yvon.cormier@med.ulaval.ca

Am J Respir Crit Care Med Vol 169. pp 903–909, 2004Originally Published in Press as DOI: 10.1164/rccm.200210-1154OC on December 30, 2003Internet address: www.atsjournals.org

such as Saccharopolyspora rectivirgula, which is responsible forthe induction of farmer’s lung, a form of HP (14). The costimula-tion between T lymphocytes (CD28 and CTLA-4 molecules)and alveolar macrophages (AMs) (B7 molecules or CD80-86)plays a major role in the development of this disease: blockade ofthis pathway inhibits the inflammatory response to S. rectivirgulaantigen in mice (15). Cytokines such as TNF, IFN-�, and IL-10are involved in HP. Increased TNF serum bioactivity was re-ported in farmer’s lung patients (16). Animal models have alsoshown that IL-10 modulates inflammation and that IFN-� isnecessary for development of the disease (17, 18).

Generally, smokers have lung inflammation characterized byan increased number of AMs (up to 99% of total cells), asdemonstrated by bronchoalveolar lavage (BAL) (19). The aimof this study was to verify whether nicotine could, at least in part,be responsible for the protection against the development of HPobserved in smokers and to look at possible mechanisms of action.For this, we used a previously described mouse model of HPthat has been used extensively in our laboratory (15) and treatedmice with nicotine, either intraperitoneally (IP) or intranasally(IN).

In addition, an AM cell line stimulated with S. rectivirgulaantigen or Escherichia Coli lipopolysaccharide (LPS), a potentactivator of AM (3), was used to test the effect of nicotine onTNF, IFN-�, and IL-10 mRNA expression and B7 costimulatorymolecule expression in vitro. Some of the results in these studieshave been previously published in the form of an abstract (20).

METHODS

Induction of HP and BAL

C57Bl/6 female mice (Charles River, St-Constant, PQ, Canada) re-ceived 50 �l of IN S. rectivirgula antigen 3 consecutive days a week for3 weeks. Mice were simultaneously treated with either 100-�l IP or50-�l IN nicotine tartrate salt (Sigma, St. Louis, MO) daily or twice aday (Table 1). Four days after the last S. rectivirgula instillation, micewere sacrificed by overexposure to isoflurane and tracheotomized, anda BAL was performed using three aliquots of 1 ml phosphate-bufferedsaline. Total cells were counted. Cytospin preparations were stainedwith Hemacolor Stain Set (EM Diagnostic Systems, Middletown, VA),and differential counts were obtained.

Histopathologic Studies

A section of the left lung from the highest nicotine dose (2.0 mg/kg)IN-treated mice from a separate group that did not undergo BAL wascollected on the day of the sacrifice and stained with hematoxylin/eosin.Inflammatory parameters of the lung tissue were evaluated blindly bya pathologist. The histology score was graded from 0 to 4.

Semiquantitative Cytokine mRNA Expression

The reverse transcription-polymerase chain (RT-PCR) reaction experi-ments were done on lung sections (from the 2.0-mg/kg IN-treated group)lysed in TRIZOL reagent (GIBCO BRL, Grand Island, NY). TotalRNA was extracted and quantified with Ribo-Green reagent (MolecularProbes, Eugene, OR). One microgram of RNA was reverse transcribedwith MMLV-Reverse Transcriptase (GIBCO BRL), and a PCR wasperformed with a Peltier Thermal Cycler 200 for detection of IFN-�

904 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

TABLE 1. DESCRIPTION OF THE ANTIGEN AND NICOTINE ADMINISTRATION FOR EACH GROUPUSED FOR THE IN VIVO STUDIES

Frequency ofNicotine Treatment TreatmentAdministration Group Name Instillation (mg/kg) (times per day)

IP Sal Sal Sal 2Nic Sal Nic 20 S. rectivirgula Sal 2

0.5 (1�) S. rectivirgula Nic, 0.5 10.5 (2�) S. rectivirgula Nic, 0.5 21.0 (2�) S. rectivirgula Nic, 1.0 2

IN Sal Sal Saline 1Nic Sal Nic 10 S. rectivirgula Sal 1

0.025 S. rectivirgula Nic, 0.025 10.25 S. rectivirgula Nic, 0.25 10.5 S. rectivirgula Nic, 0.5 11.0 S. rectivirgula Nic, 1.0 12.0 S. rectivirgula Nic, 2.0 1

Definition of abbreviations: IN � intranasally; IP � intraperitoneally; Nic � nicotine; Sal � saline; S. rectivirgula � Saccharopolysporarectivirgula.

and IL-10 mRNA expression using Taq DNA polymerase (Promega,Madison, WI). Primers used for IFN-� and IL-10 were from BiosourceInternational (PQ, Canada).

The PCR reaction was run on a 1.5% agarose gel, stained withethidium bromide, and exposed to ultraviolet light. Densitometric anal-yses of the bands were performed (ScionImage; Scion Corporation,Frederick, MD).

In Vitro Studies

A mouse AM cell line, AMJ2-C11 (ATCC, Manassas, VA), was used.Cells (150,000) were plated in a 24-well plate, stimulated with S. rectivir-gula (50 �g/ml) or LPS (0.1 �g/ml) for 24 hours, and treated with 0,40, 80, or 160 �M of nicotine. Supernatant TNF levels were measuredby ELISA (RD Diagnostics, Minneapolis, MN), and cells were collectedfor analysis of the B-7 (CD80/86) costimulatory molecule expressionor lysed using TRIZOL reagent (GIBCO BRL) for RNA extraction.

Semiquantitative RT-PCR for In Vitro Studies

Cell lysis, RNA extraction, quantification, and RT-PCR were per-formed as for the in vivo studies. Primers used for TNF were 3�-CCTGGC TAG TGG GGC TTC AAG TCA TCT GTC TT-5� and 5�-GTATGA GAT AGC AAA TCG GCT GAC GGT GTG GG-3�. Primersused for IFN-� and IL-10 were the same as for the in vivo studies.

Flow Cytometry Analysis for CD80/86 Expression

Cells were stimulated with LPS or S. rectivirgula, treated with 40 �Mof nicotine for 24 hours, and incubated for 45 minutes with a mouseanti-CD80 antibody coupled to fluorescein isothiocyanate fluorochromeor a mouse anti-CD86 coupled to phycoerythrin fluorochrome or theirisotype control. The percentage of cells positive for CD80/86 was ana-lyzed in an Epics ELITE flow cytometer.

Statistical Analysis

Statistical analysis was made using an analysis of variance table followedby a Fisher’s post hoc test.

RESULTS

Total Cell and Differential Counts in BAL

Results of BAL total cell counts and differential counts arepresented in Figures 1–3. A significant reduction of total cellcounts was observed in mice treated with 0.5 mg/kg of nicotinegiven IP daily and twice a day and 1.0 mg/kg given twice a day(p � 0.02, p � 0.02, and p � 0.03; n � 8 mice per group), whereasthe number of lymphocytes decreased in all the IP-treated groups

(p � 0.003, p � 0.007, and p � 0.04) (Figure 1). IN administration(Figure 2) first resulted in a significant decrease in total cells ata very low dose of nicotine (0.025 mg/kg, p � 0.04, n � 16) andat the 0.25-, 0.5-, and 2.0-mg/kg dose (p � 0.01 for all groups, n �16 mice per group). The lymphocyte population was significantlydecreased in all IN-treated groups except for the 1.0-mg/kg group(p � 0.0001 for all groups). Because the 1.0-mg/kg IN-treatedgroup did not first respond to the treatment, a separate experi-ment was performed with this particular dose to verify whetherthis result was due to experimental variations. The results (Fig-ure 3) show that this dose did indeed reduce both total cells(p � 0.01) and lymphocytes (p � 0.002) in the BAL (n � 8 miceper group).

Histopathologic Studies

Results of the histopathologic studies are presented in Figure 4and show a marked peribronchial, perivascular, and parenchy-mal infiltration of inflammatory cells in the S. rectivirgula group(Figure 4B) compared with control mice (Figure 4A). Mice treatedwith 2.0 mg/kg nicotine (Figure 4C) showed a decreased tissueinfiltration of mononuclear cells in lung tissue compared with

Figure 1. Bronchoalveolar lavage (BAL) total cell and differential countsof intraperitoneally (IP) nicotine-treated mice. Nicotine (nic) administra-tion at doses of 0.5 mg/kg treated once or twice daily and 1.0 mg/kgtreated twice daily decreased the total cell accumulation compared withthe Saccharopolyspora rectivirgula (SR)-alone group. All nicotine-treatedgroups had significantly fewer BAL lymphocytes than the nontreatedmice. *p � 0.05 compared with the SR-alone group.

Blanchet, Israel-Assayag, and Cormier: Nicotine and Lung Inflammation 905

Figure 2. BAL total cell and differential counts of intranasally (IN) nico-tine-treated mice. All nicotine-treated groups had fewer total BAL cellsand BAL lymphocytes, except for the 1.0-mg/kg group. *p � 0.05compared with the SR-alone group.

nontreated mice. This result was confirmed by a reduction of totalhistologic score (p � 0.01, n � 9 mice per group) (Figure 4D).

RT-PCR for In Vivo Studies

The results for RT-PCR for detection of IFN-� and IL-10 mRNAexpression in mice lung sections from the highest dose–treatedmice (2.0 mg/kg) are presented in Figure 5. The ratio betweenthe intensity of the �-actin band and that of the cytokine bandwas calculated. Results are expressed as a percentage of expres-sion, with 100% being attributed to the S. rectivirgula alone ratio.Each band represents a different animal. The 2.0-mg/kg treat-ment significantly reduced IFN-� mRNA expression (p � 0.01,n � 4 mice per group). The expression of IL-10 was not affectedby the nicotine treatment at this dose.

TNF Concentration in AMJ2-C11 Cell Line Supernatants(In Vitro Studies)

Results on the effect of nicotine on TNF levels in LPS- andS. rectivirgula–stimulated AMJ2-C11 cells supernatants are pre-sented in Figure 6. Results are expressed as a percentage ofrelease, with 100% being attributed to LPS- or S. rectivirgula–

Figure 3. BAL total cell and differential counts of 1.0-mg/kg IN nicotine-treated mice. Nicotine administration significantly reduced both totalcells and lymphocytes in the BAL. *p � 0.05 compared with the SR-alone group.

Figure 4. Histopathology results (�10) of control (SAL; A ), nontreated(SR; B ), and 2.0-mg/kg treated (SR � nicotine; C ) mice. SR treatmentinduced a marked peribronchial, perivascular, and parenchymal infiltra-tion of inflammatory cells compared with saline-treated mice, whereasnicotine treatment reduced this infiltration. Results are expressed in Das the total histologic score (graded on an arbitrary scale, from 0 to 4:0, no inflammatory cells; 1, less than 10%; 2, 10–25%; 3, 25–50%;and 4, more than 50%), which was significantly reduced in nicotine-treated mice (total score of 1.6 0.6) compared with nontreated mice(total score of 2.46 0.68) (*p � 0.01; n � 9).

906 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

Figure 5. IFN-� and interleukin (IL)-10 mRNA expression in lung sec-tions of untreated (SR) and 2.0-mg/kg IN nicotine-treated (NIC) mice(n � 4). (A and C ) Lung mRNA expression was determined by reversetranscription-polymerase chain reaction (RT-PCR). �-Actin primers wereused as a positive control. Each lane represents a different animal. (B andD ) Semiquantification of mRNA expression was done by densitometry.Nicotine treatment had no effect on IL-10 mRNA expression but signifi-cantly reduced IFN-� mRNA expression in the 2.0-mg/kg group (SR �

NIC). *p � 0.05 compared with SR-alone group.

stimulated and untreated cells. The TNF release was significantlyreduced to 84% for 160-�M nicotine-treated and LPS-stimulatedcells (Figure 6A) (n � 4, p � 0.004). Similarly, TNF release wassignificantly reduced to 76% for 160-�M nicotine-treated and

Figure 6. Percentage of tumor necrosis factor (TNF) release by lipopoly-saccharide (LPS) (A )- and SR (B )-stimulated cells; 100% is attributedto the LPS- or SR-stimulated and untreated cells. Nicotine treatmentsignificantly reduced TNF release by 160-�M treated cells for both typesof stimulations (84% in LPS-stimulated and nicotine-treated cells and76% in SR-stimulated and nicotine-treated cells). *p � 0.05; n � 4.

S. rectivirgula–stimulated cells (Figure 6B) (n � 4, p � 0.02).AMJ2-C11 cells failed to release IL-10 and IFN-� (data notshown).

Flow Cytometry Analysis

In the AMJ2-C11 cell line, CD86 was expressed in 100% of thecells; no stimulation of expression was therefore achievable byLPS or S. rectivirgula. Nicotine had no effect on CD86 expres-sion. CD80 was expressed in 5% of unstimulated cells. Nicotinealone increased its expression to 20% of the cells, whereas LPSstimulated the expression to 38.1% and S. rectivirgula to 34.6%of cells. Despite the fact that nicotine alone stimulated CD80expression, nicotine treatment on LPS- or S. rectivirgula–stimulated cells lowered CD80 expression to 19.8% (p � 0.0015)and 26.1% (p � 0.01), respectively (n � 4) (Figure 7).

Figure 7. Expression of CD80 costimulatory molecule in AMJ2-C11 LPS-or SR-stimulated cells. Nicotine treatment (40 �M) reduced CD80expression by 51% in LPS-stimulated cells (LPS � Nic) and by 25% inSR-stimulated cells (SR � Nic). *p � 0.01, n � 4.

Blanchet, Israel-Assayag, and Cormier: Nicotine and Lung Inflammation 907

RT-PCR for In Vitro Studies (AMJ2-C11 Cell Line)

Results are presented in Figure 8 (TNF), Figure 9 (IL-10), andFigure 10 (IFN-�) and are expressed as a percentage of expres-sion, with 100% being attributed to the LPS- or S. rectivirgula–alone stimulated cells. The results are representative of fourdifferent experiments. Nicotine treatment had an inhibitory effecton TNF mRNA expression that was reduced by 98% (p � 0.0001),with the 40-�M dose in LPS-treated cells and 34% (p �0.03) with the 160-�M dose in the S. rectivirgula–stimulated cells.A similar effect was observed with the IL-10 mRNA expression.In fact, nicotine treatment reduced IL-10 mRNA expression by88% (p � 0.0001) in LPS-stimulated cells (40-�M nicotine) and62% (p � 0.01) in S. rectivirgula–stimulated cells (160 �M).AMJ2-C11 cells failed to express IFN-� mRNA with LPS stimu-lation, but it was detected with the S. rectivirgula antigen stimula-tion. Nicotine treatment reduced IFN-� mRNA expression by80% (p � 0.0001) in 40-�M treated cells.

Figure 8. TNF mRNA expression in AMJ2-C11 cells. (A ) Expression wasdetermined by RT-PCR after LPS or SR stimulation, and �-actin primerswere used as positive control subjects. The first band represents un-treated cells, and the second, third, and fourth bands represent, respec-tively, 40-, 80-, and 160-�M nicotine treatments. This is a representativeresult of four different experiments. (B and C ) Quantification of thebands was achieved by densitometry. Nicotine treatment reduced TNFmRNA expression by 98% in LPS-stimulated and 40-�M–treated cellsand by 34% in SR-stimulated and 160-�M–treated cells. (*p � 0.05compared with the nontreated group, n � 4).

DISCUSSION

This study was performed to verify the effect of nicotine on lunginflammation in an in vivo mouse model of HP and in vitro usinga mouse AM cell line. Nicotine treatment, either IP or IN, hada significant antiinflammatory effect in the mouse model. Totallung cells as well as tissue inflammation were significantly de-creased in nicotine-treated animals. The BAL cell populationthat was the most affected by the nicotine treatment was thelymphocyte population. These cells were significantly decreasedin BAL, either in IP- or IN-treated animals. Nicotine treatmentalso had an inhibitory effect on IFN-� mRNA expression buthad no effect on IL-10 mRNA expression by lung tissue.

The effect of nicotine in the IP-treated mouse model did notfollow a positive dose response pattern. For the IP administra-tion, the most effective dose on total cell accumulation in BALwas 0.5 mg/kg administered twice a day. This is a relatively highdose, corresponding to that of people smoking an average of

Figure 9. IL-10 mRNA expression in AMJ2-C11 cells. (A ) mRNA expres-sion was determined by RT-PCR, and �-actin primers were used as apositive control. The first band represents untreated cells, and the sec-ond, third, and fourth bands represent, respectively, 40-, 80-, and160-�M nicotine treatments. This is a representative result of four differ-ent experiments. (B and C ) Nicotine treatment reduced IL-10 mRNAexpression by 88% in LPS-stimulated and 40-�M–treated cells and by62% in SR-stimulated and 160-�M–treated cells (*p � 0.05 comparedwith the nontreated group, n � 4).

908 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

Figure 10. IFN-� mRNA expression in AMJ2-C11 SR-stimulated cells.(A ) Expression was determined by RT-PCR, and �-actin primers wereused as a positive control. The first band represents untreated cells, andthe second, third and fourth bands represent, respectively, 40-, 80-, and160-�M nicotine treatments. This is a representative result of four differentexperiments. (B) Densitometric quantification of the bands showed a 80%reduction of expression in 40-�M SR-stimulated and nicotine-treatedcells. *p � 0.05 compared with the nontreated group, n � 4).

one pack a day (21). The decrease of the antiinflammatory effectseen with increasing nicotine doses could be attributed to nico-tinic acetylcholine receptor desensitization at high doses or totoxic effects of nicotine at these very high doses. The IN adminis-tration first dose response also showed an unusual pattern. The0.025-, 0.25-, 0.5-, and 2.0-mg/kg doses had significant antiin-flammatory effects on total cells, whereas the 1.0-mg/kg dosedid not. Of interest is the significant effect of the lowest doseof 0.025 mg/kg. This dose corresponds to a single administrationof NT nasal spray in humans (22).

The results obtained with the 1.0-mg/kg group were surpris-ing. Such results could be explained by a double dose responsecurve. However, because in this group half of the animals seemedto have a response and half did not (data not shown), we ques-tioned whether this was due to a technical error. To verify this,an additional group of mice was treated at this IN dose. Thissecond trial showed that a 1.0-mg/kg IN administration signifi-cantly reduces both total cells and lymphocytes in the BAL; wecan conclude that the previous result for that dose was due toexperimental variability and that nicotine did not have a doubledose response curve, as it was suggested by the first results.

Nicotine had a striking inhibitory effect on lymphocyte accu-mulation in the lung, as demonstrated by BAL. This effect is inagreement with other studies showing an inhibitory effect ofnicotine on lymphocyte proliferation (23). The CD4/CD8 sub-type of lymphocytes found in the BAL was not affected by thenicotine treatment (result not shown). Because significant effectswere obtained at low doses in both IP and IN protocols, furtherstudies are needed to identify the optimal antiinflammatory dose.Similar results with low doses of nicotine being more effectiveto reduce inflammation than high concentrations have also beenreported in treatment of ulcerative colitis (24).

The effect of nicotine on tissue infiltration of inflammatorycells further supports the antiinflammatory effects of nicotine.

One of the reasons that we used IN administration was to

avoid the liver first pass effect and to achieve an antiinflammatoryeffect at lower doses than with IP administration. When adminis-tered IN, nicotine is rapidly absorbed by the vessels surroundingthe nasal sinuses. These vessels drain into the superior vena cava,and blood passes into pulmonary circulation before entering theperipheral circulation (22). Our research supports the bettereffect of direct administration into the lungs compared withsystemic administration; the 0.025-mg/kg dose administered IN hada significant antiinflammatory effect, whereas the 0.25-mg/kgdose administered IP did not.

The downregulating effects of nicotine on IFN-� mRNA ex-pression by lung tissue from the highest dose of nicotine-treatedmice could be one of the intracellular mechanisms involved inthe protection effect of cigarette smoking on the developmentof HP. In fact, IFN-� expression was reported to be essential inthe development of this disease (18). The lack of effect of nico-tine on IL-10 mRNA expression in lung tissue is not surprising.Nicotine-treated mice had lung tissue inflammation, and IL-10is an antiinflammatory cytokine induced by proinflammatorycytokines. The preservation of high levels of IL-10 mRNA couldalso be a positive effect added to other nicotine effects in control-ling the inflammation response to the S. rectivirgula antigen.

The inhibitory effect of nicotine on CD80 expression on themouse AM cell line suggests that AMs could be at least partiallyresponsible for the observed lymphocyte suppression. Becauseblockade of the CD80/86-CTLA-4/CD28 pathway inhibits lym-phocyte response to S. rectivirgula antigen (15), the decrease inexpression of CD80 that we observed could be another mecha-nism by which cigarette smoking could decrease the risk for HP.

TNF is a cytokine that is involved in the pathology of HP(16). The fact that TNF production by the AMJ2-C11 cell lineas well as TNF mRNA expression was reduced by the nicotinetreatment could explain in part the immunosuppressive effectof nicotine in vivo. Nicotine also downregulated IL-10 and IFN-�mRNA expression by AMJ2-C11 cells. These results are similarto those previously published on nicotine inhibitory effect oncytokine production by peripheral blood monocytes (3, 25).AMJ2-C11 cells failed to release IL-10 and IFN in vitro. This isconsistent with the ATCC technical data on this cell line, whichstates that AMJ2-C11 cells preferably produce IL-6 and a smallamount of TNF on stimulation with LPS.

Once again, the dose response curve of nicotine showed thatthe lowest dose (40 �M) had the best inhibitory effect on LPS-stimulated cells and that the highest dose (160 �M) had the besteffect on S. rectivirgula–stimulated cells. This could be explainedby a difference in the level of activation of the cells when stimu-lated with 0.1-�g/ml LPS compared with 50 �g/ml for S. rectivir-gula antigen stimulation and by the different mechanisms ofaction of the two antigens. S. rectivirgula is phagocytosed byAMs, whereas LPS acts through its own receptor, CD14 (26).This hypothesis is further supported by the fact that all the LPS-treated cells showed a better response with the lowest dose ofnicotine compared with S. rectivirgula–stimulated cells. We alsobelieve that a shorter time of incubation could prevent receptordesensitization and have a better effect on cytokine mRNAinhibition. Finally, the fact that AM produced IFN-� mRNAexpression on S. rectivirgula stimulation but did not with LPSstimulation is understandable because IFN-� is a TH1-type cyto-kine and S. rectivirgula is known to induce HP, a TH1-polarizeddisease (27).

Because nicotine did reduce IL-10 mRNA production by AMin vitro, the expression found in lung tissue from the mouselungs could come from other inflammatory cells or from struc-tural cells.

The inhibitory effect of nicotine could not only explain thelower prevalence of HP in smokers but also the different outcome

Blanchet, Israel-Assayag, and Cormier: Nicotine and Lung Inflammation 909

of HP that does occur in some smokers (12). The downregulatingeffect could be sufficient to prevent acute HP. Subjects could thusdevelop a more insidious form that could progress to irreversiblelung damage before the diagnosis is made, explaining the pooreroutcome in these subjects.

Lymphocytes are highly activated and recruited by the cyto-kines released by AMs. The inhibitory effect of nicotine oncytokines mRNA production in AMs could, together with inhibi-tion of the costimulatory pathway, explain the decrease in thelymphocyte population in the BAL of nicotine-treated animals.We are aware that inhibition of mRNA expression does notalways transfer to a protein production inhibition, but however,there was concordance between TNF mRNA levels and proteinrelease in AMJ2-C11 cells. Because in our study the decreasein mRNA and protein expression corresponds to cellular find-ings, we believe that the overall inhibitory effect of nicotine couldexplain, at least to some extent, the protection that smokers haveagainst the development of this disease.

Conclusions

Results of this study show that nicotine reduces the alveolarinflammatory response to S. rectivirgula antigen and affects someAM (stimulated with LPS or S. rectivirgula) functions in vitro.This influence could be, at least in part, responsible for theprotection that smokers have against development of HP. Be-cause nicotine is effective in the treatment of ulcerative colitis,it could also be of interest in the treatment of HP and otherpulmonary inflammatory diseases.

Conflict of Interest Statement : M-R.B. has no conflict of interest and the work inthis paper is part of the patent number PCT/CA02/00412 [March 2002] and isthe property of Laval University; E.I-A. has no conflict of interest and the work inthis paper is part of the patent number PCT/CA02/00412 [March 2002] and isthe property of Laval University; Y.C. has no conflict of interest and the workin this paper is part of the patent number PCT/CA02/00412 [March 2002] andis the property of Laval University.

Acknowledgment : The authors thank Jocelyne Simard and Alina Milahia for theirhelp with the mRNA extraction and RT-PCRs and Dr. Marcien Fournier for thehistopathologic studies. They also thank the Canadian Institutes of Health Re-search, the J-D. Begin foundation, and Institut Robert-Sauve en Sante et Securiteau Travail for financial support.

References

1. Katzung BG. Basic and clinical pharmacology. New York: McGraw-Hill/Appleton & Lange; 1998.

2. Kalra R, Singh SP, Savage SM, Finch GL, Sopori ML. Effects of cigarettesmoke on immune response: chronic exposure to cigarette smokeimpairs antigen-mediated signaling in T cells and depletes IP3-sensitiveCa(2�) stores. J Pharmacol Exp Ther 2000;293:166–171.

3. Payne JB, Johnson GK, Reinhardt RA, Dyer JK, Maze CA, DunningDG. Nicotine effects on PGE2 and IL-1 beta release by LPS-treatedhuman monocytes. J Periodontal Res 1996;31:99–104.

4. Matsunaga K, Klein TW, Friedman H, Yamamoto Y. Involvement ofnicotinic acetylcholine receptors in suppression of antimicrobial activ-ity and cytokine responses of alveolar macrophages to Legionellapneumophila infection by nicotine. J Immunol 2001;167:6518–6524.

5. Zhang S, Petro TM. The effect of nicotine on murine CD4 T cell re-sponses. Int J Immunopharmacol 1996;18:467–478.

6. Lahmouzi J, Simain-Sato F, Defresne MP, De Pauw MC, Heinen E,

Grisar T, Legros JJ, Legrand R. Effect of nicotine on rat gingivalfibroblasts in vitro. Connect Tissue Res 2000;41:69–80.

7. Louvet B, Buisine MP, Desreumaux P, Tremaine WJ, Aubert JP, PorchetN, Capron M, Cortot A, Colombel JF, Sandborn WJ. Transdermalnicotine decreases mucosal IL-8 expression but has no effect on mu-cin gene expression in ulcerative colitis. Inflamm Bowel Dis 1999;5:174–181.

8. Mabley JG, Pacher P, Southan GJ, Salzman AL, Szabo C. Nicotine re-duces the incidence of type I diabetes in mice. J Pharmacol Exp Ther2002;300:876–881.

9. Warren CP. Extrinsic allergic alveolitis: a disease commoner in non-smokers. Thorax 1977;32:567–569.

10. Johansson S, Bjermer L, Franzen L, Henriksson R. Effects of ongoingsmoking on the development of radiation-induced pneumonitis inbreast cancer and oesophagus cancer patients. Radiother Oncol 1998;49:41–47.

11. Cormier Y, Israel-Assayag E, Bedard G, Duchaine C. Hypersensitivitypneumonitis in peat moss processing plant workers. Am J Respir CritCare Med 1998;158:412–417.

12. Ohtsuka Y, Munakata M, Tanimura K, Ukita H, Kusaka H, Masaki Y,Doi I, Ohe M, Amishima M, Homma Y. Smoking promotes insidiousand chronic farmer’s lung disease, and deteriorates the clinical out-come. Intern Med 1995;34:966–971.

13. Sharma OP, Fujimura N. Hypersensitivity pneumonitis: a noninfectiousgranulomatosis. Semin Respir Infect 1995;10:96–106.

14. Fink JN. Hypersensitivity pneumonitis. J Allergy Clin Immunol 1984;74:1–10.

15. Israel-Assayag E, Fournier M, Cormier Y. Blockade of T cell costimula-tion by CTLA4-Ig inhibits lung inflammation in murine hypersensitiv-ity pneumonitis. J Immunol 1999;163:6794–6799.

16. Schaaf BM, Seitzer U, Pravica V, Aries SP, Zabel P. Tumor necrosisfactor-alpha-308 promoter gene polymorphism and increased tumornecrosis factor serum bioactivity in farmer’s lung patients. Am J RespirCrit Care Med 2001;163:379–382.

17. Gudmundsson G, Bosch A, Davidson BL, Berg DJ, Hunninghake GW.Interleukin-10 modulates the severity of hypersensitivity pneumonitisin mice. Am J Respir Cell Mol Biol 1998;19:812–818.

18. Gudmundsson G, Hunninghake GW. Interferon-gamma is necessary forthe expression of hypersensitivity pneumonitis. J Clin Invest 1997;99:2386–2390.

19. Israel-Assayag E, Dakhama A, Lavigne S, Laviolette M, Cormier Y.Expression of costimulatory molecules on alveolar macrophages inhypersensitivity pneumonitis. Am J Respir Crit Care Med 1999;159:1830–1834.

20. Blanchet M-R, Israel-Assayag E, Cormier Y. Nicotine and the cellularimmune response in hypersensitivity pneumonitis (HP) (abstract). AmJ Respir Crit Care Med 2001;A748.

21. Benowitz NL, Jacob P III. Daily intake of nicotine during cigarettesmoking. Clin Pharmacol Ther 1984;35:499–504.

22. Guthrie SK, Zubieta JK, Ohl L, Ni L, Koeppe RA, Minoshima S, DominoEF. Arterial/venous plasma nicotine concentrations following nicotinenasal spray. Eur J Clin Pharmacol 1999;55:639–643.

23. Mellon RD, Bayer BM. The effects of morphine, nicotine and epibatidineon lymphocyte activity and hypothalamic-pituitary-adrenal axis re-sponses. J Pharmacol Exp Ther 1999;288:635–642.

24. Sykes AP, Brampton C, Klee S, Chander CL, Whelan C, Parsons ME.An investigation into the effect and mechanisms of action of nicotinein inflammatory bowel disease. Inflamm Res 49:311–319.

25. Madretsma S, Wolters LM, van Dijk JP, Tak CJ, Feyerabend C, WilsonJH, Zijlstra FJ. In-vivo effect of nicotine on cytokine production byhuman non-adherent mononuclear cells. Eur J Gastroenterol Hepatol1996;8:1017–1020.

26. Dobrovolskaia MA, Vogel SN. Toll receptors, CD14, and macrophageactivation and deactivation by LPS. Microbes Infect 2002;4:903–914.

27. Yamasaki H, Ando M, Brazer W, Center DM, Cruikshank WW. Polar-ized type 1 cytokine profile in bronchoalveolar lavage T cells of patientswith hypersensitivity pneumonitis. J Immunol 1999;163:3516–3523.

Medium-Term Survival after Hospitalization withCommunity-Acquired PneumoniaGrant W. Waterer, Lori A. Kessler, and Richard G. Wunderink

Department of Medicine, University of Western Australia, and Department of Respiratory Medicine, Royal Perth Hospital, Perth, WesternAustralia, Australia; and Physicians Research Network, Methodist Le Bonheur Healthcare, and Clinical Research, Methodist Healthcare,Memphis, Tennessee

An episode of community-acquired pneumonia (CAP) has been sug-gested to predict greater than expected mortality after dischargefrom hospital. We ascertained the survival status as of December2002 of a cohort of patients with CAP identified prospectively be-tween November 1998 and June 2001. Cox regression analysis wasused to examine the impact of demographic factors, comorbid ill-nesses, and CAP severity on subsequent mortality. Of 378 CAP survi-vors we ascertained the survival status of 366 (96.9%), 125 (34.1%)of whom had died. The mean length of follow-up was 1,058 days(range, 602–1,500 days). Independent predictors of mortality wereincreasing age (p � 0.001), comorbid cerebrovascular (p � 0.002)and cardiovascular (p � 0.023) disease, an altered mental state (p �

0.001), a hematocrit of less than 35% (p � 0.035), and increasingblood glucose level (p � 0.025). In 41- to 80-year-olds withoutsignificant comorbidities there was a trend to greater than expectedmortality. In conclusion, an episode of CAP in young adults withoutsignificant comorbid illnesses does not appear to be an adverseprognostic marker of medium-term survival. The trend to greaterthan expected mortality in patients over 40 years of age needsfurther study and physicians should be particularly alert for underly-ing life-limiting disease processes in patients presenting with acuteconfusion or a hematocrit of less than 35%.

Keywords: hospitalization; pneumonia, survival long-term

Community-acquired pneumonia (CAP) is both common and amajor cause of morbidity and mortality in most Western coun-tries, including the United States (1). Risk factors for acquiringCAP include increasing age and comorbid illnesses, such as cardiacfailure, diabetes, neoplasia, and chronic obstructive pulmonarydisease (2, 3). When a predisposing factor cannot be identified,potential explanations include a particularly virulent pathogenor an underlying genetic predisposition. Some patients have noidentifiable cause despite exhaustive investigation.

Several studies have shown that patients surviving an intensivecare unit admission have a subsequent mortality rate substantiallygreater than that of age-matched control subjects (4–6). The greatermortality rate is largely explained by the more common occurrenceof comorbid illnesses in intensive care unit subjects (4–6).

Brancati and coworkers followed a cohort of patients surviv-ing an episode of pneumonia requiring hospitalization (7). Theonly independent predictors of 2-year mortality were comorbidillnesses and a hematocrit of less than 35%. As age was not an

(Received in original form October 24, 2003; accepted in final form December 22, 2003)

Supported by the Methodist Le Bonheur Healthcare Foundation and the Mid-South Pulmonary and Critical Care Research Foundation.

Correspondence and requests for reprints should be addressed to Grant W.Waterer, M.B.B.S., F.R.A.C.P., School of Medicine and Pharmacology, Universityof Western Australia, and Department of Respiratory Medicine, Royal Perth Hospi-tal, G.P.O. Box X2213, Perth 6847, Western Australia, Australia. E-mail: waterer@cyllene.uwa.edu.au

Am J Respir Crit Care Med Vol 169. pp 910–914, 2004Originally Published in Press as DOI: 10.1164/rccm.200310-1448OC on December 23, 2003Internet address: www.atsjournals.org

independent predictor of subsequent mortality, this raises thepossibility that for some patients an episode of pneumonia couldbe a sentinel event for an underlying life-limiting disease. Evenif the findings of Brancati and colleagues cannot be confirmed,identification of predictors of subsequent mortality after dis-charge with CAP may have important implications for bothpatients and their treating physicians.

We therefore ascertained the survival status up to 4 yearsafter discharge from hospital of a prospectively collected cohortof patients with CAP to study the risk factors for subsequentmortality. Some of the results of this study were reported inabstract form at the American College of Chest Physicians An-nual Scientific Meeting in 2003 (8).

METHODS

Study Design

A prospective cohort of patients admitted to the Methodist HealthcareMemphis Hospitals (Memphis, Tennessee) with CAP was recruitedbetween November 1, 1998 and June 30, 2001. An attempt was madeto enroll all patients admitted with CAP during this period; however,subjects were enrolled only after written informed consent was ob-tained. The Methodist Healthcare Institutional Review Board approvedthe study. Survival was ascertained for subjects as of December 31,2002, using social security number-linked death records, review of hospi-tal and outpatient pharmacy records, contact with all known treatingphysicians, and postal contact at the last known home address.

Inclusion Criteria

CAP was defined as an acute illness (fewer than 14 days of symptoms),the presence of a new chest radiographic infiltrate as confirmed byeither a radiologist or a pulmonary/critical care physician, and clinicalfeatures suggestive of acute pneumonia. The clinical features requiredwere one of Group A (fever � 37.8�C, hypothermia � 36�C, cough,sputum production); or two of Group B (dyspnea, pleuritic pain, physi-cal findings of lung consolidation, and leukocyte count � 12 � 10/Lor � 4.5 � 10/L). These criteria are consistent with published guidelinesfor the diagnosis of CAP (9).

Exclusion Criteria

Exclusion criteria included (1) patients with severe immunodeficiencyas defined by the Centers for Disease Control Criteria for patientswith acquired immune deficiency syndrome (10); (2) patients receivingchemotherapy in the past 60 days; (3) patients receiving treatment withcorticosteroids equivalent to prednisolone at more than 20 mg/day formore than 14 days; (4) patients receiving immunosuppression afterorgan transplantation; (5) patients receiving cyclosporine, cyclophos-phamide, or azathioprine; (6) nonambulatory nursing home patients;and (7) patients hospitalized within the past 30 days.

Data Collection

All clinical and outcome data were assessed and collated by a pulmonaryphysician (R.G.W./G.W.W.). Results of microbiological and other labo-ratory tests as ordered by the treating physician were recorded. Pneumo-nia Severity Index (PSI) scores as described by Fine and coworkers (11)were calculated at the time of admission to hospital. Applied physiologyand chronic health evaluation (APACHE) II scores were also calcu-lated, using the worst values over the first 24-hour period (12).

Waterer, Kessler, and Wunderink: Medium-Term Survival after Hospitalization with CAP 911

Figure 1. Survival by age group.

Definitions

Septic shock was defined on the basis of Society of Critical Care Medi-cine/American College of Chest Physicians criteria (13), and had tooccur within 48 hours of presentation. Mechanical ventilation was de-fined as any period of invasive ventilation via an endotracheal or naso-tracheal tube. Noninvasive ventilation (e.g., by face mask) was notdefined as mechanical ventilation. To be considered valid, blood cul-tures had to be obtained before antibiotic administration. Bacteremiawas defined as at least one blood culture positive for a known CAPpathogen. Patients with blood cultures positive for coagulase-negativestaphylococci or other common skin contaminants were not classifiedas bacteremic. Comorbid diseases were classified as present or absent,using the Pneumonia Severity Index criteria (11).

Statistical Analysis

Statistical calculations, including multivariate analysis, were performedwith the statistical package SPSS 10.1.0 (SPSS, Chicago, IL). Cox regres-

Figure 2. Survival by age group and comor-bidity status. Age groups: 18–40 (plus signs),41–60 (filled squares), 61–80 (open circles),and � 81 (filled triangles) years.

TABLE 1. RISK FACTORS FOR MORTALITY BY AGE GROUP

Age Group (yr)

Factor 18–40 41–60 61–80 81�

Cardiovascular disease 2 (2.7) 17 (14.2) 35 (27.6) 19 (41.3)Cerebrovascular disease 0 (0) 8 (6.7) 16 (12.5) 5 (10.9)Renal disease 0 (0) 3 (2.5) 7 (5.5) 2 (4.3)Hepatic disease 0 (0) 2 (1.7) 2 (1.6) 0 (0)Neoplastic disease 2 (2.7) 11 (9.2) 17 (13.4) 8 (17.4)Altered mental state 1 (1.4) 11 (9.2) 17 (13.4) 11 (23.9)Septic shock 0 (0) 9 (7.5) 8 (6.3) 3 (6.5)Mechanical ventilation 4 (5.5) 15 (12.5) 20 (15.8) 6 (13.0)Bacteremia* 4 (7.5) 17 (18.7) 8 (9.0) 5 (14.3)Total 73 120 127 46

* The denominator for bacteremia is the number of subjects in that age groupwith valid blood cultures.

Data presented are n (% of total).

sion modeling was used for multivariate analysis with models includingall significant interactions. All reported p values are two-tailed, with avalue of less than 0.05 considered significant.

RESULTS

Of an initial cohort of 404 subjects with CAP, 378 (93.6%)survived to hospital discharge. We were able to ascertain thesurvival status of 366 (96.9%) of the 378 survivors. These 366subjects had a mean age of 58.1 years (range, 18–99 years): 54.9%were female, 62.0% were African American, and 38.0% werewhite. Death occurred in 125 patients (34.1%) after dischargefrom hospital. The mean length of follow-up was 901 days (range,5–1,500 days) for all patients and 1,058 days (range, 602–1,500days) for all survivors. For nonsurvivors the mean time to deathwas 435 days (range, 15–1,310 days).

A Kaplan–Meier plot of mortality by age groups is shown inFigure 1. The age groups were deliberately selected to be identi-cal in design to those of Brancati and coworkers (7). Figure 2shows a Kaplan–Meier plot for each age group stratified by thepresence or absence of comorbid diseases. Table 1 shows theprevalence of other potential risk factors for subsequent mortal-ity in each age group.

912 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

A Kaplan–Meier plot of mortality by PSI grade is shown inFigure 3. A Cox regression analysis utilizing all components ofthe PSI score was performed. A summary of the final factorsincluded in the Cox regression model is shown in Table 2.Treated as a continuous variable, hematocrit was not predictiveof mortality (p � 0.14). However, as a categorical variable usingthe same cutoff of 0.35 as Brancati and coworkers (7), a strongassociation between hematocrit and subsequent mortality, whichremained after adjusting for age, cerebrovascular and cardiovas-cular disease, and altered mental state, was found (Table 2).Serum glucose on admission was also independently predictiveof mortality after correcting for the above-cited factors, but nocritical cutoff value was identifiable. Modeling various cutoffsfor other physiological variables in the PSI failed to find anysignificant association.

The development of shock (p � 0.007) or the need for me-chanical ventilation (p � 0.009) during the episode of CAP waspredictive of higher mortality rates after discharge. However,once age and cardiovascular and cerebrovascular disease wereincluded with these organ failures in a Cox regression analysis,neither shock (p � 0.22) nor mechanical ventilation (p � 0.17)remained a significant factor.

To enable comparison with published data on medium-termmortality after intensive care unit admission (4–6), we analyzedsurvival by APACHE II score (Figure 4). Survival curves ineach APACHE II score class are similar to those previouslyreported (4–6). A Cox regression analysis found that the pointtotal for all three components of the APACHE II score (A,physiological; B, chronic organ failure; and C, age) (12) wereindependent predictors of subsequent mortality (p � 0.001 inall cases).

Table 3 shows a comparison of observed mortality rates forall age groups compared with the expected mortality calculatedfrom published U.S. age-, sex-, and race-matched mortality rates(14). As underlying comorbidities are likely to have a significantinfluence on mortality and we were interested in whether mortal-ity rates were higher than expected in subjects without underly-ing life-limiting diseases, we defined the subgroup of subjectswith no significant comorbidities (including all comorbiditiesdefined in the PSI plus chronic obstructive pulmonary diseaseand diabetes). An expected mortality over the period of follow-up was again calculated for each subject from published age-,sex-, and race-matched U.S. population data (Table 3) (14).

Figure 3. Survival by pneumonia severity index class.

TABLE 2. FACTORS PREDICTING MORTALITY IN THE FINALCOX REGRESSION MODEL

Factor p Value Odds Ratio (95% Confidence Intervals)

Age � 0.001 —Cerebrovascular disease 0.002 2.52 (1.42–4.46)Cardiovascular disease 0.023 1.72 (1.08–2.77)Altered mental state � 0.001 3.13 (1.93–5.10)Hematocrit less than 35% 0.035 1.61 (1.03–2.52)Increasing blood glucose 0.025 —

DISCUSSION

We have examined the factors predicting mortality in the 2to 4 years after discharge from hospital with CAP. The majorpredictors of subsequent mortality were increasing age, the pres-ence of cardiovascular or cerebrovascular disease, and the pres-ence of an altered mental state at the time of presentation tohospital with CAP. Although some correlation between the se-verity of CAP and subsequent survival was seen, age and comor-bidities largely accounted for this association.

The primary concern addressed by our study was the issueraised by Brancati and colleagues (7) that an episode of CAPis an adverse prognostic marker for medium- to long-term survival.In all age groups it was not surprising that mortality exceededthat expected from population data because of the high prevalenceof life-limiting comorbid diseases. Our primary concern waswhether an episode of CAP has adverse implications for longerterm survival in the absence of known life-limiting comorbidities.As best demonstrated in Figure 2, no apparent excess medium-term mortality occurred in the 18- to 40-year age group in theabsence of comorbid diseases. Although our study is not poweredto exclude anything but large mortality differences, this findingis certainly reassuring, particularly given the earlier alarmingfindings of Brancati and coworkers (7) in this age group. How-ever, there was a trend toward higher mortality in all older agegroups with no comorbid illnesses, especially the 41- to 60-yearage group. This trend is even more concerning when the factthat the expected mortality rate is based on population data thatincludes subjects with significant comorbidities is considered.Although the lack of statistical significance may be reassuring,our data do not exclude the possibility that an episode of CAPmay be a sentinel event for increased mortality in a subgroupof patients.

Figure 4. Survival by applied physiology and chronic health evaluation(APACHE) II score.

Waterer, Kessler, and Wunderink: Medium-Term Survival after Hospitalization with CAP 913

TABLE 3. OBSERVED AND EXPECTED* MORTALITY

All Subjects Subjects with No Comorbid Illnesses

Age Group (yr) Observed Expected p Value Observed Expected p Value

18–40 7 (9.6) 0.4 (0.6) 0.01 1 (1.7) 0.4 (0.6) 1.041–60 29 (24.2) 2.2 (1.8) � 0.001 7 (11.3) 1.1 (1.7) 0.0661–80 56 (44.1) 10.3 (8.1) � 0.001 7 (16.3) 3.6 (8.3) 0.5� 81 33 (71.7) 12.0 (26.1) � 0.001 5 (29.4) 4.3 (25.4) 1.0

* Expected mortality calculated from age-, sex-, and race-matched U.S. population data (13).Parenthetic data indicate percent of mortality.

Although our cohort was both larger and monitored for alonger period of time than that of Brancati and coworkers, weobserved a similar overall mortality rate (38 versus 35%) (7).As other studies of medium-term survival after critical illnesssupport our finding of increasing age as a significant predictorof postpneumonia mortality (4–6), we believe the failure of Bran-cati and coworkers to detect age-related medium-term mortalitydifferences may have been due to an excess of significant comor-bidities in younger subjects, relative to our cohort.

The finding that both cardiovascular and cerebrovascular dis-eases increase the subsequent rate of mortality is not surprising.Similarly, poorly controlled diabetes, reflected in the correlationbetween absolute blood glucose level and mortality, also hasstrong biological plausibility. What is less clear and requiresfurther comment are the two other significant independent pre-dictors of mortality: an altered mental state at presentation anda hematocrit of less than 35%.

Increasing age and underlying organic brain disease are twomajor risk factors for confusion in any patient hospitalized withsepsis (15). However, the multivariate analysis showed that analtered mental state was a strong predictor of subsequent mortal-ity even after correcting for age and the presence of cerebrovas-cular disease. Consistent with our findings, a study of elderlypatients also identified an altered mental state as a strong pre-dictor of 6-month mortality after discharge from hospital (16).Confusion also appears to be an independent risk factor fordeath during hospitalization for CAP (17). One explanation isthat altered mental state is indicative of some underlying neuro-degenerative process not recognized at the time of admission(i.e., Alzheimer’s disease or cerebrovascular disease). Confusionat admission may also reflect active and excessive alcohol inges-tion (18), with the associated long-term health consequences.Confusion is also more likely to occur in patients with moresevere organ disease, especially more severe cerebrovasculardisease, which may not be adequately controlled for in our analy-sis. In any respect, for patients with CAP who present withan altered mental state, physicians should be vigilant for theopportunity to intervene in disease processes that may have beenpreviously unrecognized or the severity underestimated.

The association between hematocrit and mortality is muchharder to explain. The absence of any correlation when hemato-crit is treated as a continuous variable could suggest the associa-tion is spurious. However, that the arbitrary cutoff of 35% waschosen due to the identical findings of Brancati and colleaguesadds support to the possibility that the association is real (7).Because anemia is a common complication of chronic illness, lowhematocrit, like confusion, may function as a marker of moresevere chronic illness. Although further study is clearly required,a hematocrit of less than 35% in a patient with CAP, similar tothe occurrence of mental confusion, should alert the physician tothe possibility of comorbid life-limiting disease processes.

The primary limitation of our study is that the actual causeof death in our subjects was not collated. Although potentially

available from death certificates, the deficiencies in this approachare well documented (19, 20). As our primary objective wasto determine whether an episode of CAP per se had adverseimplications for medium-term survival, establishing the exactcause of death was not critical. A further limitation is that thediagnosis of comorbid diseases was based on clinical history,examination, and routine investigations, such as creatinine forrenal failure and liver function tests for hepatic failure. Subclini-cal disease, such as cerebrovascular disease as already discussedabove, could clearly have been missed as cranial imaging studieswere not routinely performed, nor were echocardiograms orgated heart scans performed to exclude impaired left ventricularfunction. Although this partially limits the conclusions that canbe drawn from our analysis, all the factors incorporated into ourstatistical models are readily available to physicians treatingpatients with CAP. Therefore our findings remain highly applica-ble in the usual clinical context.

In summary, we have shown that increasing age, comorbiddiseases (especially cardiovascular and cerebrovascular disease),the presence of an altered mental state, a hematocrit of less than35%, and poor glycemic control are significant and independentpredictors of mortality in the subsequent 2 to 3 years after dis-charge for CAP. Physicians treating patients with CAP shouldbe aware of the predictors of increased mortality, as they mayindicate previously unknown or underestimated comorbid dis-eases. Although the natural course of some of these diseaseprocesses may not be alterable, earlier recognition maximizesthe potential for interventions to impact on subsequent morbid-ity and mortality.

Conflict of Interest Statement : G.W.W. has no declared conflict of interest; L.A.K.has no declared conflict of interest; R.G.W. has no declared conflict of interest.

References

1. Hoyert DL, Arias E, Smith BL, Murphy SL, Kochanek KD. Deaths:final data for 1999. Natl Vital Stat Rep 2001;49:1–114.

2. Almirall J, Bolibar I, Balanzo X, Gonzalez CA. Risk factors for commu-nity-acquired pneumonia in adults: a population-based case-controlstudy. Eur Respir J 1999;13:349–355.

3. Farr BM, Bartlett CL, Wadsworth J, Miller DL. Risk factors for community-acquired pneumonia diagnosed upon hospital admission. British Tho-racic Society Pneumonia Study Group. Respir Med 2000;94:954–963.

4. Wright JC, Plenderlieth L, Ridley SA. Long-term survival following in-tensive care: subgroup analysis and comparison with the general popu-lation. Anaesthesia 2003;58:637–642.

5. Ridley S, Jackson R, Findlay J, Wallace P. Long term survival afterintensive care. BMJ 1990;301:1127–1129.

6. Niskanen M, Kari A, Halonen P. Five-year survival after intensive care:comparison of 12,180 patients with the general population. FinnishICU Study Group. Crit Care Med 2003;24:1962–1967.

7. Brancati FL, Chow JW, Wagener MM, Vacarello SJ, Yu VL. Is pneumo-nia really the old man’s friend? Two-year prognosis after community-acquired pneumonia. Lancet 1993;342:30–33.

8. Wunderink RG, Kessler LA, Waterer GW. Long-term outcome in pa-tients with community acquired pneumonia [abstract]. Chest 2003;124:189S.

914 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

9. Chow AW, Hall CB, Klein JO, Kammer RB, Meyer RD, Remington JS.Evaluation of new anti-infective drugs for the treatment of respiratorytract infections. Clin Infect Dis 1992;15:S62–S88.

10. Centers for Disease Control and Prevention. 1993 revised CDC HIVclassification system and expanded surveillance definition for adoles-cents and adults. MMWR Recomm Rep 1992;41:1–19.

11. Fine MJ, Auble TE, Yealy DM, Hanusa BH, Weissfeld LA, Singer DE,Coley CM, Marrie TJ, Kapoor WN. A prediction rule to identify low-risk patients with community-acquired pneumonia. N Engl J Med 1997;336:243–250.

12. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a se-verity of disease classification system. Crit Care Med 1985;13:818–829.

13. Bone RC, Balk RA, Cerra FB, Dellinger RP, Fein AM, Knaus WA,Schein RM, Sibbald WJ. Definitions for sepsis and organ failure andguidelines for the use of innovative therapies in sepsis. ACCP/SCCMConsensus Conference Committee, American College of Chest Physi-cians/Society of Critical Care Medicine. Chest 1992;101:1644–1655.

14. Arias E. United States Life Tables, 2000. Natl Vital Stat Rep 2002;51:1–38.

15. Roche V. Southwestern Internal Medicine Conference: etiology and man-agement of delerium. Am J Med Sci 2003;325:20–30.

16. Kakuma R, du Fort GG, Arsenault L, Perrault A, Platt RW, MonetteJ, Moride Y, Wolfson C. Delerium in older emergency departmentpatients discharged home: effect on survival. J Am Geriatr Soc 2003;51:443–450.

17. Karalus NC, Cursons RT, Leng RA, Mahood CB, Rothwell RP, HancockB, Cepulis S, Wawatai M, Coleman L. Community acquired pneumo-nia: aetiology and prognostic index evaluation. Thorax 1991;46:413–418.

18. Zalacain R, Torres A, Celis R, Blanquer J, Aspa J, Esteban L, MenendezR, Blanquer R, Borderias L; Pneumonia in the Elderly WorkingGroup, Area de Tuberculosis e Infecciones Respiratorias. Community-acquired pneumonia in the elderly: Spanish multicentre study. EurRespir J 2003;21:294–302.

19. Sington JD, Cottrell BJ. Analysis of the sensitivity of death certificatesin 440 hospital deaths: a comparison with necropsy findings. J ClinPathol 2002;55:499–502.

20. Smith-Sehdev AE, Hutchins GM. Problems with proper completion andaccuracy of the cause-of-death statement. Arch Intern Med 2001;161:277–284.

The Effect of Pranlukast on Allergen-induced BoneMarrow Eosinophilopoiesis in Subjects with AsthmaKrishnan Parameswaran, Richard Watson, Gail M. Gauvreau, Roma Sehmi, and Paul M. O’Byrne

Firestone Institute for Respiratory Health, St. Joseph’s Healthcare; and Department of Medicine, McMaster University, Hamilton,Ontario, Canada

We investigated the mechanisms by which leukotriene receptorantagonists decrease airway eosinophil number. In a randomized,double-blind crossover study, we examined the effects of 2 weeksof treatment with pranlukast 300 mg twice a day or placebo onallergen-induced changes in airway eosinophil number and bonemarrow eosinophil progenitors in 15 subjects with mild asthma.Pranlukast treatment for 2 weeks decreased mean sputum eosino-phil count from 0.15 � 106/g (5.3% of cells) before treatment to0.02 � 106/g (0.7% of cells) after treatment (p � 0.05), whereasplacebo did not. Pranlukast also decreased the eosinophil count(5.6% at 7 hours and 7.5% at 24 hours) (p � 0.05) after allergeninhalation compared with placebo (13.8% at 7 hours and 15.3% at24 hours). There was a similar trend for sputum cells immunostainingfor EG2, eotaxin, interleukin-5, and regulated upon activation, normalT cell expressed and secreted. Pranlukast also significantly attenuatedthe allergen-induced increase in the number of bone marrow eosino-phil/basophil cfu (mean 0.3) at 24 hours compared with placebo(mean 6.2). The proportion of CD34� cells expressing the eotaxinreceptor CC chemokine receptor 3, 24 hours after allergen inhalation,was also reduced by pranlukast. We conclude that, the cysteinylleukotriene receptor antagonist, pranlukast, attenuates allergen-induced increase in airway eosinophils by decreasing bone marroweosinophilopoiesis and airway chemotactic and eosinophilopoieticcytokines.

Keywords: leukotriene receptor antagonist; allergen; sputum eosino-phil; eosinophil progenitor

Allergen inhalation by patients with asthma causes, within 7hours, an increase in the number of inflammatory cells in theairway, notably eosinophils and basophils (1). Eosinophilic in-flammation, which is a hallmark of allergic inflammation, is asso-ciated with the late asthma response and an increase in airwayhyperresponsiveness (2, 3). This cellular infiltration is facilitatedby a number of cytokines such as interleukin (IL)-5, chemokinessuch as eotaxin and regulated upon activation, normal T cellexpressed and secreted (RANTES), and mediators such as cys-teinyl leukotriene C4, D4, and E4. Whereas IL-5 promotes eosino-phil proliferation, differentiation, priming, and survival (4),eotaxin (5) and RANTES (6) induce chemotaxis of eosinophilsto the airway and their activation.

There is strong evidence to suggest that the bone marrowplays an important role in the allergen-induced eosinophilic air-

(Received in original form December 3, 2003; accepted in final form January 17, 2004)

Supported by a grant-in-aid from Ono Pharmaceutical Co., Ltd. K.P. is a postdoc-toral fellow of the Canadian Institutes of Health Research.

Correspondence and requests for reprints should be addressed to Paul M. O’Byrne,M.B., F.R.C.P.I., F.R.C.P.(C), Department of Medicine, McMaster University, HealthSciences Centre, Room 3W10, 1200 Main Street West, Hamilton, ON, L8N 3Z5Canada. E-mail: obyrnep@mcmaster.ca

This article has an online supplement, which is accessible from this issue’s tableof contents online at www.atsjournals.org

Am J Respir Crit Care Med Vol 169. pp 915–920, 2004Originally Published in Press as DOI: 10.1164/rccm.200312-1645OC on January 23, 2004Internet address: www.atsjournals.org

way inflammation (7). After an allergen inhalation, eosinophillineage–committed progenitor cells expressing the membrane-bound isoform of IL-5 receptor �-subunit (CD34�IL5-R��) andthe IL-5–responsive eosinophil/basophil cfu (Eo/B cfu) increasein the bone marrow (8–10) and in peripheral blood (11). Progeni-tor cells are also observed in the airway mucosa of subjects withasthma (12). It is likely that allergen inhalation increases thenumber of eosinophil progenitors in the bone marrow, whichmigrate to the airways either as mature eosinophils or as imma-ture cells and undergo local differentiation. An increase in theeotaxin receptor, CC chemokine receptor 3 (CCR3), on thebone marrow progenitor cells after an allergen inhalation mayfacilitate the progenitor cell mobilization from the bone marrowto the peripheral circulation (13).

Cysteinyl leukotrienes promote airway eosinophilic inflam-mation. Inhalation of LTD4 and LTE4 causes sputum and tissueeosinophilia (14), and cysteinyl leukotriene receptor antagonistscan decrease sputum eosinophilia (15). Cysteinyl leukotrienescan cause airway eosinophilia by a number of possible mecha-nisms (16). They can promote eosinophil chemotaxis into theairway (17), increase the surface expression of adhesion mole-cules on eosinophils and blood vessels facilitating their migration(18), prolong eosinophil survival (19), and upregulate gene ex-pression of various cytokines/chemokines such as IL-4, IL-5,IL-13, eotaxin, and granulocyte–macrophage colony-stimulatingfactors that can promote eosinophil production and survival (20).We have shown previously that bone marrow Eo/B cfu culturesgrown in the presence of IL-5 were significantly increased byLTD4 and this was inhibited by montelukast (21). We thereforehypothesized that one of the mechanisms by which cysteinylleukotrienes promote airway eosinophilia is by promoting eosin-ophilopoiesis in the bone marrow.

We investigated this by studying the effect of a cysteinylleukotriene-1 receptor antagonist, pranlukast, on allergen-inducedchanges in bone marrow–derived eosinophil lineage–committedCD34� cells and IL-5–responsive Eo/B cfu. In addition, we stud-ied the effect of treatment on CCR3 receptor expression onthe progenitor cells. We also studied the effect of pranlukasttreatment on the numbers of total and activated eosinophils insputum and the immunoreactivity in sputum cells for EG2�,IL-5, eotaxin, and RANTES.

Some of the results of these studies have been reported pre-viously in the form of an abstract (22).

METHODS

Subjects

The subjects were 16 nonsmokers with atopy and mild asthma (Table1) using short-acting �-agonists infrequently. All subjects gave writteninformed consent, and the Research Ethics Committee of HamiltonHealth Sciences Corporation approved the study.

Design

This was a randomized, double-blind, crossover, two-period study com-paring 2 weeks of treatment with pranlukast, 300 mg tablet twice a day,or matching placebo, with at least 2 weeks of washout period between

916 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

TABLE 1. BASELINE SUBJECT CHARACTERISTICS

Mean (min, max)

Age, yr 27 (19, 53)FEV1, L 3.3 (2.3, 4.6)FEV1% predicted 91 (71, 105)PC20 methacholine, mg/ml* 2.4 (0.1, 18.4)

Definition of abbreviation: PC20 methacholine � provocative concentration ofmethacholine producing a 20% drop in FEV1.

* Geometric mean.

the treatment periods. The subjects underwent a screening allergeninhalation to identify those with a dual asthma response (greater than15% fall in FEV1 within 120 minutes, followed by a similar drop between3 and 7 hours after allergen inhalation). After 2 to 4 weeks, they hadspirometry, sputum induction, and a methacholine inhalation test. Theywere then randomized to one of the two treatment arms using computer-derived codes that were maintained off-site by an independent third-party pharmacist. The medications were taken at 8 p.m. and 8 a.m. Thefirst dose was in the evening of Day 1, and the last dose was on themorning of Day 15. On Day 8 (� 2 days), they visited the laboratoryfor a physical examination to evaluate for adverse effects and to evaluatecompliance. On Day 13 (� 2 days), subjects attended the laboratoryfor spirometry, sputum induction, and an iliac bone marrow aspiration.The next day, an allergen inhalation test was performed. At 7 hours afterallergen inhalation, sputum was induced. The next day, approximately24 hours after allergen inhalation, subjects attended the laboratory forspirometry, sputum induction, and another bone marrow aspiration.Compliance was evaluated weekly by pill counting. Any adverse effectwas evaluated by self-reported symptoms, physical examination, bloodchemistry, and urine examination.

Allergen Inhalation

FEV1 was measured using a Collins water-sealed spirometer (WarrenE. Collins, Braintree, MA) and kymograph according to the AmericanThoracic Society recommendations (23). Allergen inhalation was per-formed as described previously (2, 24).

Sputum Induction and Processing

Sputum was induced with hypertonic saline, separated from saliva, andprocessed as described by Pizzichini and coworkers (25).

Sputum Cytochemistry

Sputum cytospins were prepared on Aptex-coated slides (Sigma Chemi-cal Co., Mississauga, ON, Canada), fixed for 10 minutes in periodate–lysine–paraformaldehyde, and stained as described by Gauvreau andcoworkers (1).

Bone Marrow Progenitor Culture

Five milliliters of bone marrow sample was aspirated into heparinized(1,000 U/ml) syringes from the iliac crests after freezing the skin andperiosteum with 2% lidocaine. Low-density mononuclear cells wereisolated by sedimentation on Percoll density gradients (specific gravity1.08) and cultured in the presence of IL-5 (1 ng/ml) as described pre-viously (8–10).

Immunofluorescence Staining

Nonadherent mononuclear cells were stained with saturating amountsof biotin-conjugated anti–IL-5R�, and anti-CCR3, or the isotype-controlantibody in 100 �l of ice-cold, fluorescence-activated, cell-sorter stainingbuffer for 30 minutes at 4C (8, 9, 12, 13).

Flow Cytometry

The stained nonadherent mononuclear cells were analyzed using aFACScan flowcytometer equipped with an argon laser (Becton Dickin-son Instrument Systems, Mississauga, ON, Canada) using the CELL-QUEST program. CD34� blast cells were identified as described pre-viously (8, 13).

Analysis

Demographic data were summarized using descriptive statistics. Sputumeosinophil counts, expressed in absolute terms and as a percentage ofthe total cell count, and the immunopositive cells were analyzed usingrepeated measures analysis of variance. Treatment (pranlukast or pla-cebo), time (baseline, preallergen, 7 hours postallergen, 24 hours post-allergen), and period (first or second treatment) were the within-subjectfactors. The differences between pranlukast and placebo on allergen-induced changes in bone marrow progenitor colony counts and receptorexpression were analyzed using paired t test. All analyses were per-formed using the Statistical Package for Social Sciences, version 10(SPSS Inc., Chicago, IL). p Values less than 0.05 were considered tobe statistically significant.

RESULTS

One subject developed a facial urticaria after one dose of thestudy medication and was withdrawn from the study. Fifteensubjects completed the study. One subject had transient, mild,and reversible elevation of liver enzymes that was not attributedto pranlukast treatment. The overall compliance with treatmentwas 90%.

Sputum Cell Counts

Two weeks of treatment with pranlukast decreased mean sputumeosinophil count from 0.15 � 106/g (� 0.24) before treatmentto 0.02 � 106/g (� 0.02) after treatment (p � 0.05) (Figure 1),whereas placebo did not (0.11 � 106/g [� 0.16] before treatmentand 0.10 � 106/g [� 0.15] after treatment).

Allergen inhalation increased sputum eosinophil numbers,and this effect was attenuated by pranlukast. After placebo treat-ment, sputum eosinophils increased to 0.72 � 106/g (� 1.04) at7 hours and 0.54 � 106/g (� 0.59) at 24 hours. After pranlukasttreatment, the increase in sputum eosinophil was significantlyless (0.21 � 106/g at 7 hours [� 0.38] and 0.33 � 106/g [� 0.38]at 24 hours) (p � 0.05) (Figure 1). There were no significantchanges in the total sputum cell counts or other cell countswith either allergen inhalation or pranlukast treatment (data notshown).

Sputum Cytochemistry

Similar to the effect on eosinophil counts, 2 weeks of treatmentwith pranlukast decreased EG2� cells and attenuated the aller-

Figure 1. Changes in sputum eosinophil count. Two weeks of placebotreatment did not have any effect on the eosinophil count (expressedas a percentage of total cell count), whereas treatment with pranlukastdecreased it significantly. The increase in eosinophil count at 7 and24 hours after allergen inhalation was also significantly decreased bypranlukast treatment but not by placebo (*p � 0.05).

Parameswaran, Watson, Gauvreau, et al.: Leukotriene Receptor Antagonist and Eosinophilopoiesis 917

Figure 2. Changes in eosinophil/basophil cfu (Eo/B cfu) in the bonemarrow. Allergen inhalation after 2 weeks of treatment with placebocaused a significant increase in the number of cfu (per 2.5 � 105 nonad-herent mononuclear cells in the bone marrow). This was completelyattenuated by 2 weeks of treatment with pranlukast (*p � 0.05). How-ever, pranlukast treatment did not have an effect on the baseline numberof cfu.

gen-induced increase in EG2� cells (Table 2). The effect at 7hours was statistically significant compared with the placebotreatment. A similar pattern was observed on sputum cellsstained for IL-5, RANTES, and eotaxin. Whereas placebo pre-treatment caused a 2.4-fold increase in IL-5–positive cells anda 1.7- and 2-fold increase in RANTES and eotaxin-positive cells,respectively, at 7 hours after allergen challenge, the correspond-ing numbers after pranlukast treatment were 1.5-, 1.0-, and 1.2-fold, respectively. However, this difference was not statisticallysignificant (Table 2).

Bone Marrow Eo/B cfu

After 2 weeks of treatment with pranlukast, the number (mean �SD) of IL-5–responsive Eo/B cfu in the bone marrow was 17.3 �6.3. This was not different from placebo treatment (17.3 � 6.1).Allergen inhalation increased the bone marrow Eo/B cfu duringplacebo treatment to 23.6 � 6.2. This effect was significantlydecreased during pranlukast treatment to 17.7 � 7.4 (p � 0.05)(Figure 2).

Bone Marrow Progenitor Cell Receptor Expression

We examined the numbers of bone marrow cells expressingCD34 and the numbers of CD34� cells expressing CCR3. Aller-gen inhalation preceded by 2 weeks of placebo treatment did

TABLE 2. CHANGES IN SPUTUM CELLS IMMUNOSTAINING FOR CYTOKINES

Placebo Pranlukast

(�106/g) Baseline Pretreatment 7 h 24 h Baseline Pretreatment 7 h 24 h

EG2� 0.04 (0.05) 0.08 (0.10) 0.51 (1.20) 0.24 (0.33) 0.11 (0.21) 0.02 (0.02) 0.10 (0.18) 0.16 (0.23)IL-5 2.16 (1.80) 1.77 (1.33) 4.16 (4.17) 2.87 (2.70) 2.49 (2.82) 1.14 (1.10) 1.71 (1.45) 2.13 (2.16)Eotaxin 2.02 (1.98) 1.65 (1.21) 3.38 (3.61) 1.89 (1.73) 3.48 (2.47) 1.24 (1.23) 1.51 (1.30) 2.00 (3.05)RANTES 0.64 (0.86) 0.50 (0.90) 0.85 (1.51) 0.68 (1.06) 0.65 (0.92) 0.89 (1.93) 0.92 (1.89) 0.18 (0.30)

Definition of abbreviations: IL � interleukin; RANTES � regulated upon activation, normal T cell expressed and secreted.Immunocytochemistry on sputum cells was performed for EG2, eotaxin, IL-5, and RANTES at baseline, after 2 weeks of treatment

before the allergen inhalation (pretreatment), and 7 and 24 hours after allergen inhalation. Pretreatment with pranlukast decreasedthe allergen-induced increase in positively staining cells compared with placebo. Statistical significance was observed only for EG2staining. The values are given as mean (SD).

not cause a significant increase in CD34� cells. However, pranlu-kast significantly reduced the allergen-induced increase inCD34�CCR3� cells compared with placebo (p � 0.05) (Table 3).

DISCUSSION

This study confirms previous findings that treatment with a leu-kotriene receptor antagonist, pranlukast, decreases the numberof total and activated eosinophils in the airway and attenuatestheir increase after an allergen inhalation. The study also showedfor the first time that this was associated with a decrease in thenumbers of IL-5–responsive eosinophil cfu and CD34�CCR3cells in the bone marrow. This suggests that one of the mechanismsby which leukotriene antagonists decrease allergen-induced air-way eosinophil number is by a direct effect on allergen-inducedeosinophilopoiesis in the bone marrow.

The ability of specific antagonists of the Cys-LT1 receptor todecrease airway eosinophil number is well recognized. Six weeksof treatment with montelukast decreased sputum eosinophil num-ber in patients with mildly uncontrolled asthma (14). Four tosix weeks of treatment with pranlukast (26), montelukast (27),and zafirlukast (28) decreased allergen-induced increase in air-way eosinophil number. We observed a similar effect on eosino-phil number and activation after 2 weeks of treatment with pran-lukast and after an allergen inhalation. We hypothesized, for anumber of reasons, that one of the mechanisms would be a directeffect on eosinophilopoeisis and egress of the progenitor cellsfrom the bone marrow. First, allergen inhalation results in in-creased production of cysteinyl leukotrienes in the airway (29).Second, CD34� granulocytic precursors express Cys-LT1 recep-tors on their surface (30). Third, we had demonstrated previ-ously, in vitro, an increase in IL-5–responsive Eo/B cfu when non-adherent mononuclear cells from bone marrow of subjects withatopy were treated with LTD4 (21). Finally, LTD4 evoked cal-cium fluxes and actin polymerization in CD34� cells derived frombone marrow of normal subjects and promoted chemotaxis to-ward it that was inhibited by a Cys-LT1 receptor antagonist (31).

Consistent with our hypothesis, we observed two novel effectsof the Cys-LT1 receptor antagonist, pranlukast, on eosinophilprogenitor cells. First, the increase in the number of IL-5–responsive cfu after an allergen inhalation was significantly de-creased by 2 weeks of treatment with pranlukast, which did notseem to have an effect on the baseline number compared withthe placebo treatment arm. The study did not investigate themechanism of this effect. The effect of a leukotriene receptorantagonist may be direct or indirect. Because cysteinyl leuko-trienes may be involved in mediating some of the biologicaleffects of IL-5 (32), pranlukast may interfere with the ability ofIL-5 to promote eosinophilopoiesis. This seems unlikely in thisstudy because the nonadherent mononuclear cells from the

918 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

TABLE 3. CHANGES IN BONE MARROW PROGENITOR CELLS

Placebo Pranlukast

Receptor (/2.5 � 105 NAMNC) Preallergen Postallergen Preallergen Postallergen

CD34 4,557 � 2,110 4,777 � 2,214 6,568 � 5,097 6,452 � 4,712CCR3 63 � 42 79 � 65 93 � 66 57 � 40*

Definition of abbreviations: CCR3 � CC chemokine receptor 3; NAMNC � nonadherent mononuclear cells.Cells expressing CD34 and CD34CCR3 on nonadherent mononuclear cells in the bone marrow were enumerated by flow

cytometry. There was no effect of allergen or pranlukast on the total number of CD34� cells. The allergen-induced increase inCD34�CCR3� cells observed after placebo treatment was attenuated by pranlukast treatment. The data are presented as mean �

SEM.* p � 0.05.

pranlukast-treated subjects were grown in vitro in the presenceof optimal concentration of IL-5, unless pranlukast is able tomodulate IL-5 signal transduction pathways and make the cellsunresponsive or less responsive to the effect of IL-5. There aretwo major signaling pathways of IL-5 in eosinophils. IL-5 activatesLyn, Syk, and JAK2 and propagates signals through the Ras–mitogen-activated protein kinase and Janus kinase-signal trans-ducer and activator transcription factor pathways (33). It is notknown whether cysteinyl leukotrienes are involved in either ofthese pathways. On the basis of recent evidence that they maybe upstream of STAT6 signaling in the IL-13 signaling mecha-nism (34), this is a likely possibility that needs further investiga-tion. The effect of pranlukast on eosinophilopoiesis may also beindirect. Because IL-5 can upregulate Cys-LT1 receptor expres-sion on HL-60 cells that differentiated into eosinophils (35), theymay also increase the expression of Cys-LT1 receptors on CD34�

cells. Pranlukast may directly prevent IL-5–responsive eosino-phil differentiation of the CD34� cells with increased Cys-LT1

receptor expression.The second novel observation in this study was that pranlu-

kast attenuated the allergen-increased increase in the number ofCD34� cells in the bone marrow expressing the eotaxin receptor,CCR3. We confirmed previous observation of CCR3 expressionon CD34� cells (13); however, we did not examine the localiza-tion of CCR3 to CD34� cells that also express IL-5R�. TheCD34� cells expressing CCR3 show increased chemotaxis to-ward eotaxin (13), and this is augmented in the presence of IL-5.It is likely, therefore, that the allergen-induced increase in thelevels of IL-5 and eotaxin in the airways of patients with dualasthma response causes the migration of pluripotent undifferen-tiated hemopoeitic cell from the bone marrow to the airway(36), where it can undergo local maturation into eosinophils.Our results suggest that cysteinyl leukotrienes are involved inthe expression of CCR3 receptors on the progenitor cells. Themechanism was not investigated in this study. Because IL-5 isknown to increase CCR3 expression on leukemic cell lines (37)and because treatment with leukotriene receptor antagonists candecrease airway IL-5 levels in murine models of asthma (38, 39),we postulate that the effect of pranlukast may be indirect bydecreasing airway or perhaps bone marrow IL-5 levels. Indeed,airway and bone marrow IL-5 levels increase after an allergeninhalation (40), and in the present study, pranlukast treatmentcaused nearly 50% attenuation in the number of sputum cellsstaining positive for IL-5, 7 hours after an allergen inhalation,compared with placebo. Two weeks of treatment with pranlukast(preallergen) also seemed to cause a trend toward increasingthe number of CD34� cells and CD34�CCR3� cells. Becausethe bone marrow contains a dynamic progenitor cell pool andbecause we did not perform bone marrow aspiration before thestart of treatment, it is difficult to interpret whether this increaserepresents an increase in bone marrow production or whether

it is a reflection of the ability of pranlukast to prevent themfrom exiting the bone marrow. The latter seems to be the likelypossibility.

The effect of pretreatment with a leukotriene antagonist onallergen-induced bone marrow responses seems to be differentfrom that of pretreatment with an inhaled corticosteroid. Sevendays of treatment with budesonide decreased IL-5–responsiveEo/B cfu in subjects with asthma; however, unlike in this study,it did not prevent the allergen-induced increase in the numbersof cfu (41). In other words, although an allergen inhalation wasable to overcome the inhibitory effect of inhaled corticosteroidsin the growth of IL-5–responsive bone marrow progenitor cells,a leukotriene antagonist seems to be able to prevent it. The mostlogical explanation is that cysteinyl leukotriene levels increasesignificantly after an allergen inhalation and may contribute tothe stimulation of the bone marrow. It is also possible thatallergen may increase the expression of Cys-LT1 receptor on theprogenitor cells making them more responsive to a leukotrieneantagonist. The lack of significant effect of 2 weeks of treatmenton preallergen Eo/B cfu was surprising considering that thesputum eosinophil counts were decreased with 2 weeks of treat-ment. It is possible that the role of cysteinyl leukotrienes ineosinophilopoiesis is modest in patients with mild, stable asthmaand that other mechanisms such as eosinophil chemoattractionand effect of cytokines are more pronounced. Although it isknown that leukotrienes are produced in the bone marrow (42), itis not known whether their levels are increased after an allergeninhalation. We did not measure cysteinyl leukotrienes in thebone marrow in this study, but we plan to do it in future studies.Similar to the previous study (41), we did not find the totalnumber of CD34� cells to change significantly with allergeninhalation or with treatment.

We also examined the effects of pranlukast treatment on threecytokines/chemokines that are relevant in causing airway eosinophilinfiltration. Compared with placebo, pranlukast treatment attenu-ated the increase in sputum cells staining for IL-5, RANTES, andeotaxin 7 hours after an allergen inhalation. The difference, how-ever, was not statistically significant. One of the reasons is that thiswas not a primary outcome measure and the study was not pow-ered to show this difference. Second, the variability in the totalcell counts in sputum was high, similar to previous reports (1),increasing the noise to signal ratio. However, our observationsare consistent with previous reports that have shown leukotrienereceptor antagonists to decrease airway IL-5 (38, 39, 43) andRANTES (44) levels in murine models of allergic sensitization.

In summary, the cysteinyl leukotriene receptor antagonist,pranlukast, decreases allergen-induced increase in airwayeosinophil by decreasing both eosinophil progenitor cells in thebone marrow and the levels of eosinophilopoietic and chemotac-tic cytokines in the airway. The reduction in the numbers ofIL-5–responsive eosinophil cfu and CD34� cells expressingCCR3 in the bone marrow suggests a role for cysteinyl leuko-trienes in IL-5 signal transduction pathway.

Parameswaran, Watson, Gauvreau, et al.: Leukotriene Receptor Antagonist and Eosinophilopoiesis 919

Conflict of Interest Statement : K.P. has received honoraria for lectures from Merck($5,000 CAD), Altana ($5,000 CAD), 3-M ($2,000 CAD), and GlaxoSmithKline($1,500 CAD), has provided consulting for Sepracor ($6,000 USD), and is coappli-cant on a grant from AstraZeneca ($100,000 CAD); R.W. has no declared conflictof interest; G.M.G. has no declared conflict of interest; R.S. has received researchgrants from GlaxoSmithKline and AstraZeneca; P.M.O. is a consultant and sits onadvisory boards for AstraZeneca, Altana, Bristol-Meyers Squibb, GlaxoSmithKline,Topigen, Roche, and Merck, has received lecture fees from these companies andholds sponsored grants from Altana, AstraZeneca, Dynavax, GlaxoSmithKline,Ono, and Merck.

Acknowledgment : The authors thank the subjects who participated in this study.The authors are grateful to A. Fanat, T. Rerecich, T. Strinich, S. Behmann, and J.Otis, McMaster University, for their excellent technical assistance; F. Denning,Cato Research Ltd., NC, and M. Katayama, Ono Pharma USA, Inc., NJ, for monitor-ing the study; and G. Sobhi, McMaster University, for assistance with dispensingthe study medications.

References

1. Gauvreau GM, Watson RM, O’Byrne PM. Kinetics of allergen-inducedairway eosinophilic cytokine production and airway inflammation. AmJ Respir Crit Care Med 1999;160:640–647.

2. O’Byrne PM, Dolovich J, Hargreave FE. Late asthmatic response. AmRev Respir Dis 1987;136:740–756.

3. O’Byrne PM, Inman MD, Parameswaran K. The trials and tribulationsof IL-5, eosinophils, and allergic asthma. J Allergy Clin Immunol 2001;108:503–508.

4. Clutterback EJ, Sanderson CJ. Human eosinophil hematopoiesis studiedin vitro by means of murine eosinophil differentiation factor (IL-5):production of functionally active eosinophils from normal human bonemarrow. Blood 1988;71:646–651.

5. Garcia-Zepeda EA, Rothenberg ME, Ownbey RT, Celestin J, Leder P,Luster AD. Human eotaxin is a specific chemoattractant for eosinophilcells and provides a new mechanism to explain tissue eosinophilia.Nat Med 1996;2:449–456.

6. Kameyoshi Y, Dorschner A, Mallet AI, Christophers E, Schroder JM.Cytokine RANTES released by thrombin-stimulated platelets is apotent attractant for human eosinophils. J Exp Med 1992;176:587–592.

7. Denburg JA, Sehmi R, Saito H, Pil-Seob J, Inman MD, O’Byrne PM.Systemic aspects of allergic disease: bone marrow responses. J AllergyClin Immunol 2000;106:S242–S246.

8. Sehmi R, Woods L, Watson RM, Foley R, Hamid Q, O’Byrne PM,Denburg JA. Allergen-induced increases in IL-5�-subunit expressionon bone marrow derived CD34� cells from asthmatic subjects: a novelmarker of progenitor cell commitment towards eosinophil differentia-tion. J Clin Invest 1997;100:2466–2475.

9. Sehmi R, Howie K, Sutherland DR, Shragge W, O’Byrne PM, DenburgJA. Increased levels of CD34 progenitor cells in atopic subjects. AmJ Respir Cell Mol Biol 1996;15:645–654.

10. Wood LJ, Inman MD, Watson RM, Denburg JA, Foley R, O’Byrne PM.Bone marrow inflammatory progenitor cells after allergen inhalationin asthmatic subjects. Am J Respir Crit Care Med 1998;157:99–105.

11. Gauvreau GM, Wood LJ, Sehmi R, Watson RM, Dorman SC, SchleimerRP, Denburg JA, O’Byrne PM. The effects of inhaled budesonide oncirculating eosinophil progenitors and their expression of cytokinesafter allergen challenge in subjects with atopic asthma. Am J Respir CritCare Med 2000;162:2139–2144.

12. Robinson DS, Damia R, Zeibecoglou K, Molet S, North J, Yamada T,Kay AB, Hamid Q. CD34�/interleukin-5R� messenger RNA� cellsin the bronchial mucosa in asthma: potential airway eosinophil progen-itors. Am J Respir Cell Mol Biol 1999;20:9–13.

13. Sehmi R, Dorman S, Baatjes A, Watson R, Foley R, Ying S, RobinsonDS, Kay AB, O’Byrne PM, Denburg JA. Allergen-induced fluctuationin CC chemokine receptor 3 expression on bone marrow CD34� cellsfrom asthmatic subjects: significance for mobilization of haemopoieticprogenitor cells in allergic inflammation. Immunology 2003;109:536–546.

14. Gauvreau GM, Parameswaran KN, Watson RM, O’Byrne PM. Inhaledleukotriene E4, but not leukotriene D4, increased airway inflammatorycells in subjects with atopic asthma. Am J Respir Crit Care Med 2001;164:1495–1500.

15. Pizzichini E, Leff JA, Reiss TF, Hendeles L, Boulet LP, Wei LX, Efthimi-adis AE, Zhang J, Hargreave FE. Montelukast reduces airway eosino-philic inflammation in asthma: a randomized, controlled trial. EurRespir J 1999;14:12–18.

16. Peters-Golden M. Do anti-leukotriene agents inhibit asthmatic inflam-mation? Clin Exp Allergy 2003;33:721–724.

17. Spada C, Nieves A, Krauss A, Woodward D. Comparison of leukotrieneB4 and D4 effects on human eosinophil and neutrophil motility in vitro.J Leukoc Biol 1994;55:183–191.

18. Fregonese L, Silvestri M, Sabatini F, Rossi G. Cysteinyl leukotrienesinduce human eosinophil locomotion and adhesion molecule expres-sion via a Cys-LT1 receptor-mediated mechanism. Clin Exp Allergy2002;32:745–750.

19. Lee E, Roberston T, Smith J, Kilfeather S. Leukotriene receptor antago-nists and synthesis inhibitors reverse survival in eosinophils of asthma-tic individuals. Am J Respir Crit Care Med 2000;161:1881–1886.

20. Peters-Golden M, Sampson A. Cysteinyl leukotriene interactions withother mediators and with glucocorticoids during airway inflammation.J Allergy Clin Immunol 2003;111:S37–S48.

21. Braccioni F, Dorman SC, O’Byrne PM, Inman MD, Denburg JA, Para-meswaran K, Baatjes AJ, Foley R, Gauvreau GM. The effect ofcysteinyl leukotrienes on the growth of eosinophil progenitors fromperipheral blood and bone marrow of atopic subjects. J Allergy ClinImmunol 2002;110:96–101.

22. Parameswaran K, Watson R, Rerecich T, Otis J, Strinich T, O’ByrnePM. The CysLT1 receptor antagonist, pranlukast, attenuates allergen-induced increase in airway eosinophils and bone marrow derived eosin-ophil/basophil progenitors in subjects with atopic asthma [abstract].Eur Respir J 2003;22:349s.

23. American Thoracic Society. Standardization of spirometry, 1994 update.Am J Respir Crit Care Med 1995;152:1107–1136.

24. Cockcroft DW, Murdock KY, Kirby J, Hargreave FE. Prediction of airwayresponsiveness to allergen from skin sensitivity to allergen and airwayresponsiveness to histamine. Am Rev Respir Dis 1987;135:264–267.

25. Pizzichini E, Pizzichini MMM, Efthimiadis A, Hargreave FE, DolovichJ. Measurement of inflammatory indices in induced sputum: effectsof selection of sputum to minimize salivary contamination. Eur RespirJ 1996;9:1174–1180.

26. Obase Y, Shimoda T, Tomari SY, Mitsuta K, Kawano T, Matsuse H,Kohno S. Effects of pranlukast on chemical mediators in inducedsputum on provocation tests in atopic and aspirin-intolerant asthmaticpatients. Chest 2002;121:143–150.

27. Leigh R, Vethanayagam D, Yoshida M, Watson RM, Rerecich T, InmanMD, O’Byrne PM. Effects of montelukast and budesonide on airwayresponses and airway inflammation in asthma. Am J Respir Crit CareMed 2002;166:1212–1217.

28. Calhoun WJ, Lavins BJ, Minkwitz MC, Evans R, Gleich GJ, Cohn J.Effect of zafirlukast (Accolate) on cellular mediators of inflammation:bronchoalveolar lavage fluid findings after segmental antigen chal-lenge. Am J Respir Crit Care Med 1998;157:1381–1389.

29. Macfarlane AJ, Dworski R, Sheller JR, Pavord ID, Kay AB, BarnesNC. Sputum cysteinyl leukotrienes increase 24 hours after allergen in-halation in atopic asthmatics. Am J Respir Crit Care Med 2000;161:1553–1558.

30. Figueroa DJ, Breyer RM, Defoe SK, Kargman S, Daugherty BL, Wald-burger K, Liu Q, Clements M, Zeng Z, O’Neill GP, et al. Expressionof the cysteinyl leukotriene 1 receptor in normal human lung andperipheral blood leukocytes. Am J Respir Crit Care Med 2001;163:226–233.

31. Bautz F, Denzlinger C, Kanz L, Mohle R. Chemotaxis and transendo-thelial migration of CD34� hematopoietic progenitor cells inducedby the inflammatory mediator leukotriene D4 are mediated by the7-transmembrane receptor CysLT1. Blood 2001;97:3433–3440.

32. Underwood DC, Osborn RR, Newsholme SJ, Torphy TJ, Hay DW.Persistent airway eosinophilia after leukotriene D4 administration inthe guinea pig: modulation by the LTD4 receptor antagonist, pranlu-kast, or an interleukin-5 monoclonal antibody. Am J Respir Crit CareMed 1996;154:850–857.

33. Adachi T, Alam R. The mechanism of IL-5 signal transduction. Am JPhysiol Cell Physiol 1998;44:C623–C633.

34. Elias JA, Lee CG, Zheng T, Shim Y, Zhu Z. Interleukin-13 and leuko-trienes: an intersection of the pathogenetic schema. Am J Respir CellMol Biol 2003;28:401–404.

35. Thivierge M, Doty M, Johnson J, Stankova J, Pleszczynski MR. IL-5 up-regulates cysteinyl leukotriene 1 receptor expression in HL-60 cellsdifferentiated into eosinophils. J Immunol 2000;165:5221–5226.

36. Palframan RT, Collins PD, Williams TJ, Rankin SM. Eotaxin induces arapid release of eosinophils and their progenitors from the bone mar-row. Blood 1998;91:2240–2248.

37. Tiffany HL, Alkhatib G, Combadiere C, Berger EA, Murphy PM. CCchemokine receptors 1 and 3 are differentially regulated by IL-5 duringmaturation of eosinophilic HL-60 cells. J Immunol 1998;160:1385–1392.

920 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

38. Wu AY, Chik SC, Chan AW, Li Z, Tsang KW, Li W. Anti-inflammatoryeffects of high-dose montelukast in an animal model of acute asthma.Clin Exp Allergy 2003;33:359–366.

39. Eum S-Y, Maghni K, Hamid Q, Campbell H, Eidelman DH, Martin JG.Involvement of the cysteinyl-leukotrienes in allergen-induced airwayeosinophilia and hyperresponsiveness in the mouse. Am J Respir CellMol Biol 2003;28:25–32.

40. Wood LJ, Sehmi R, Dorman S, Hamid Q, Tulic MK, Watson RM, FoleyR, Wasi P, Denburg JA, Gauvreau G, et al. Allergen-induced increasesin bone marrow T lymphocytes and interleukin-5 expression in subjectswith asthma. Am J Respir Crit Care Med 2002;166:883–889.

41. Wood LJ, Sehmi R, Gauvreau GM, Watson RM, Foley R, Denburg JA,O’Byrne PM. An inhaled corticosteroid, budesonide, reduces baseline

but not allergen-induced increases in bone marrow inflammatory cellprogenitors in asthmatic subjects. Am J Respir Crit Care Med 1999;159:1457–1463.

42. Lindgren JA, Stenke L, Mansour M, Edenius C, Lauren L, Nasman-Glaser B, Ericsson I, Reizenstein P. Formation and effects of leuko-trienes and lipoxins in human bone marrow. J Lipid Mediat 1993;6:313–320.

43. Hojo M, Suzuki M, Maghni K, Hamid Q, Powell WS, Martin JG. Roleof cysteinyl leukotrienes in CD4(�) T cell-driven late allergic airwayresponses. J Pharmacol Exp Ther 2000;293:410–416.

44. Kawano T, Matsuse H, Kondo Y, Machida I, Saeki S, Tomari S, MitsutaK, Obase Y, Fukushima C, Shimoda T, et al. Cysteinyl leukotrienesinduce nuclear factor kappa b activation and RANTES production ina murine model of asthma. J Allergy Clin Immunol 2003;112:369–374.

The Relationship between Infant Airway Function,Childhood Airway Responsiveness, and AsthmaStephen W. Turner, Lyle J. Palmer, Peter J. Rye, Neil A. Gibson, Parveenjeet K. Judge, Moreen Cox,Sally Young, Jack Goldblatt, Louis I. Landau, and Peter N. Le Souef

School of Paediatrics and Child Health, University of Western Australia, and Department of Respiratory Medicine, Princess Margaret Hospitalfor Children, Perth, Australia

The relationship between reduced pulmonary function in early lifeand persistent wheeze (PW) in school-aged children remains uncer-tain. In this study, VmaxFRC was assessed at 1 month of age, andthe presence of wheeze up to 11 years of age was prospectivelyidentified. At 11 years of age, airway responsiveness (AR) to inhaledhistamine and atopy were assessed. Recent wheeze at 11 years ofage was associated with a reduced mean z score for VmaxFRC at1 month of age (�0.41 [SD 0.91], n � 31) compared with no recentwheeze (0.04 [SD 1.00], n � 153, p � 0.03). Wheeze between 4and 6 years that persisted at 11 years (PW) was most prevalentamong those with reduced VmaxFRC at 1 month and atopy aged11 years (p � 0.002) or reduced VmaxFRC and increased AR aged11 years (p � 0.015). When all factors were considered, reducedVmaxFRC at 1 month (p � 0.03) and increased AR aged 11 years(p � 0.001) were independently associated with PW (n � 17) com-pared with other outcomes (n � 129). Reduced airway functionpresent in early infancy is associated with PW at 11 years of age,and this relationship is independent of the effect of increased ARand atopy in childhood.

Keywords: respiratory sounds; respiratory function tests; longitudinalstudy; infant

Recurrent childhood wheeze is common (1), begins in early life(2), and may then persist into later life (3). In some individuals,factors present in early life might be lifelong determinants ofrespiratory outcome. Factors associated with persistent child-hood wheeze include male sex and a history of maternal asthmaor smoking (4, 5). Atopy (4, 6) and increased airway respon-siveness (AR) in young children (7–9) have also been associatedwith persistent wheeze (PW) in later life. The mechanism forthe development of persistent childhood wheeze remains incom-pletely understood but appears to be complex, and in children,increased AR and atopy may be particularly important.

In addition to increased AR and atopy, abnormalities of pul-monary function are also associated with increased wheeze inchildren (4, 10), and these abnormalities persist into adulthood(11). What remains uncertain is whether abnormalities of pulmo-nary function precede the development of respiratory symptomsor, alternatively, are a consequence of the disease process re-sponsible for respiratory symptoms. Several studies have con-firmed that infants with reduced pulmonary function, as evi-

(Received in original form July 3, 2003; accepted in final form January 25, 2004)

Supported by National Health and Medical Research Council of Australia grantnumber 9938107 (S.W.T.).

Correspondence and requests for reprints should be addressed to Stephen W.Turner, M.D., School of Medicine, Department of Child Health, Aberdeen Chil-dren’s Hospital, Foresterhill, Aberdeen AB25 2ZG, Scotland. E-mail: s.w.turner@abdn.ac.uk

This article has an online supplement, which is accessible from this issue’s tableof contents online at www.atsjournals.org

Am J Respir Crit Care Med Vol 169. pp 921–927, 2004Originally Published in Press as DOI: 10.1164/rccm.200307-891OC on February 5, 2004Internet address: www.atsjournals.org

denced by reduced VmaxFRC, before the onset of respiratorysymptoms appear to be at increased risk for the developmentof bronchiolitis (12), pneumonia (13), and increased wheeze(4, 14–16). Two groups have followed individuals with reducedVmaxFRC in early infancy and have demonstrated persistingabnormalities of pulmonary function, as evidenced by reducedFEF25–75% at 11 years of age (12, 13, 16). These data suggest thatVmaxFRC, in addition to increased AR and atopy, may also bean important determinant of respiratory symptoms and pulmo-nary function in children.

The relationship between reduced VmaxFRC and persistentchildhood wheeze is unclear, and this is in part due to the techni-cal and practical difficulties in undertaking a study of this nature.One study has reported that there was no association betweenreduced VmaxFRC in infancy and PW at 6 years of age (4). Nostudy has reported outcomes at 11 years. Investigators in ourdepartment have recruited a birth cohort that underwent anassessment of pulmonary function at 1 month of age, before theonset of any respiratory symptoms. The 11-year follow-up ofstudy subjects is now complete. We hypothesized that reducedVmaxFRC soon after birth would be associated with PW at 11years of age, independent of atopy and increased AR in child-hood. Some of the results of this study have been previouslyreported as abstracts (17, 18).

METHODS

SubjectsThe cohort was enrolled before birth and was selected from a whitepopulation attending an antenatal clinic between June 1987 and Novem-ber 1990 as described previously (19). There was no selection for parentalasthma. Enrolled individuals who were subsequently born prematurelyor who developed respiratory symptoms in the first month of life wereexcluded from the study. The study was approved by the Medical EthicsCommittee of Princess Margaret Hospital for Children. Informed paren-tal consent was obtained for each assessment.

ProtocolAt enrollment, parents received instructions that would assist in de-tecting wheeze, and a parental history of smoking and/or physician-diagnosed asthma (PDA) was noted. Infant pulmonary function wasassessed at 1 month of age. A history of recent wheeze or PDA wasidentified from monthly questionnaires completed by parents in thefirst year and annually on the child’s second, third, fourth and fifthbirthdays. Aged 6 and 11 years, individuals underwent an assessmentthat included questionnaire, spirometry, AR to inhaled histamine, andskin prick testing. At 11 years of age, the presence of reported previouswheeze and PDA was verified using previous questionnaire data.

Definitions“Recent wheeze” included wheeze caused by all causes present in thepast year. “Parental asthma” indicated that at least one parent had ahistory of PDA at enrollment. “Atopy” was defined as at least onepositive skin prick test. Children were grouped according to the pres-ence or absence of wheeze as follows: NW for no wheeze reported atany age; W0–3 for wheeze before but not after the third birthday; W4–6for wheeze between ages 4 and 6 years but not after; W11 for wheezeat 11 years but not previously; and finally, PW for those who wheezedbetween 4 and 6 years and at 11 years of age.

922 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

Infant Pulmonary Function Measurement

The techniques used have been described (19). After sleep was inducedwith chloral hydrate, VmaxFRC was determined from the rapid thora-coabdominal compression technique during tidal breathing. In accor-dance with published guidelines (20), VmaxFRC was expressed whenappropriate as a z score, after adjustment for sex, age, length, weight,and maternal smoking during pregnancy (21). AR was determined fromthe response of VmaxFRC to doubling concentrations of nebulizedhistamine solutions (from 0.125 to 8 mg/ml). The airway challengeended if a 40% reduction in VmaxFRC was provoked or if the finalconcentration had been administered. AR was expressed as the concen-tration of histamine provoking at least a 40% reduction in VmaxFRC(PC40). As previously (7), those individuals in whom VmaxFRC did notfall by 40% after inhalation of the maximal concentration of histaminewere assigned the value PC40 � 16 mg/ml.

Childhood Pulmonary Function and AR

Childhood pulmonary function was measured with a portable spirome-ter (Pneumocheck Spirometer 6100; Welch-Allyn, Skaneateles Falls,NY) in accordance with published guidelines (22). Childhood pulmo-nary function was expressed as a z score after adjustment for height,sex, current AR, and current parental smoking status (see Table E1 onthe online supplement). The rapid technique was used to determinechildhood AR (23). Briefly, increasing doses of inhaled histamine wereadministered from a handheld dosimeter until either a 20% reductionin FEV1 occurred or the maximal cumulative dose had been administered(7.8 �mol). The response of pulmonary function to inhaled histaminewas expressed using one of two methods: first, the dose of histamine(�M) that provoked at least a 20% fall in FEV1 (PD20) (increased ARwas defined as PD20 of less than 7.8-�M histamine) (24), and second, thedose–response slope (DRS), which was calculated as follows:

�FEV1 prechallenge � FEV1 postchallengeFEV1 prechallenge

cumulative dose of histamine

�� 100

The DRS was adjusted for the influence of reduced FEF25–75%.

Skin Prick Tests

The skin prick test described by Pepys (25) was used to determinesensitivity to these allergens: cows milk, egg white, rye grass, mixedgrass (No. 7), Dermatophagoides farinae, Dermatophagoides pteronyssi-nus, cat dander, dog dander, Alternaria alternans, and Aspergillus fumi-gatus (Hollister-Stier, Elkhart, IN). The positive control was histaminesulfate (10 mg/ml), and the negative control was 0.9% saline. A positiveskin test was defined as a weal of at least 3 mm in any dimension.

Statistical Analysis

The distributions of VmaxFRC and PC40 at 1 month and dose responseslope at 6 and 11 years were skewed with long right-handed tails andwere log10 transformed before analysis (a constant of three was addedto DRS to allow values of zero or less to be included). Chi-square test,

TABLE 1. COMPARISON OF CHARACTERISTICS OF THE 156 CHILDREN WHOSE DETAILS AREPRESENTED IN THIS STUDY WITH THE ORIGINAL COHORT

Individuals in whomIndividuals Placed into Pulmonary Function Was

a Wheeze Category Assessed on Each Occasion Original Cohort

Male 54% (85/157) 52% (49/95) 56% (136/243)Mother smoked during pregnancy 17%* (25/157) 18% (17/95) 24% (57/242)Parental asthma† 30% (46/153) 29% (27/93) 31% (73/232)VmaxFRC z score at 1 mo of age (SD) �0.01 (1.02), n � 157 0.00 (1.06), n � 95 �0.04 (0.99), n � 243PC40 at 1 mo of age,‡ (95% CI) 1.01 (0.82, 1.36), n � 131 1.10 (0.83, 1.34), n � 77 1.04 (0.80, 1.14), n � 202

Definition of abbreviations: CI � confidence interval; PC40 � concentration of histamine provoking at least a 40% reduction inVmaxFRC.

* p � 0.05 compared with the original cohort.† Data missing for 11 fathers.‡ Expressed as geometric mean and 95% confidence intervals (1.96 � SEM).

Student’s t test (equivariance not assumed), Mann-Whitney U test,Kruskal-Wallis test, or analysis of variance (with Bonferroni correction)were used where appropriate to compare differences between groups.

Logistic regression models were created to study the relationshipbetween current and previous PDA at 11 years of age (outcome vari-ables) and PC40 at 1 month (explanatory variable) adjusting for sex andVmaxFRC (used in previous analyses) (7). Longitudinal associationsbetween measurements of VmaxFRC in infancy and FEF25–75% in chil-dren were studied among those individuals in whom data were completeby comparing mean z scores for all measurements of pulmonary function(i.e., three measurements for each individual) between the differentwheezing groups.

A Cox proportional hazards model was created to determine therelationship between VmaxFRC aged 1 month, atopy and DRS aged11 years (explanatory variables), and PW (outcome variable). In thismodel, individuals with PW were compared with all other individuals.The following confounding variables were also considered in this model:PC40 aged 1 month, sex, length aged 1 month, maternal or paternalsmoking during pregnancy, and parental asthma. Variables were re-moved in a backward stepwise manner assuming significance at the 5%level.

All reported p values were two sided. Analyses were performedusing a standard statistical software package (SPSS release 10.0.7; SPSS,Chicago, IL).

RESULTS

Subjects

At 1 month of age, 243 infants underwent an assessment ofpulmonary function; VmaxFRC was measured in all individualsand PC40 in 202 infants. Questionnaire data were available from112 study subjects aged 1 year, 169 aged 2 years, 113 aged 3years, 126 aged 4 years, and 106 aged 5 years. At 6 years of age,117 children were assessed, and at 11 years of age, 185 cohortmembers were assessed, including 111 children seen at 6 yearsof age (see Figures E1, E2, and E3 in the online supplementfor figures showing the numbers of individuals where details ofwheeze, pulmonary function and AR were available during theperiod of follow-up). Ten individuals were recruited and notassessed aged 1 month but did participate in later assessments.One hundred fifty-seven children could be placed into one ofthe following groups: NW, n � 67; W0–3, n � 28; W4–6, n �39; W11, n � 6; or PW, n � 17. Table 1 compares details of these157 individuals with the original cohort. Wheeze was reported onat least one occasion during the first 3 years in 25 of the 37 (68%)individuals with W4–6 and for 13 of 16 (81%) of individuals withPW where questionnaire data were available.

Prevalence of Asthma and Wheeze

Recent wheeze was reported in 37 (33%) individuals in the firstyear, 61 (36%) individuals aged 2 years, 40 (35%) individuals

Turner, Palmer, Rye, et al.: Infant Lung Function and Asthma 923

TABLE 2. DETAILS IN EARLY INFANCY FOR GROUPS DEFINED BY CHILDHOOD WHEEZE

NW W0–3 W4–6 W11 PW Trend Test(n � 67) (n � 28) (n � 39) (n � 6 ) (n � 17 ) (p )

Male 54% (36/67) 43% (12/28) 59% (23/39) 50% (3/6) 65% (11/178) NSMother smoked

during pregnancy 15% (10/67) 29% (8/28) 21% (8/39) 17% (1/6) 18% (2/17) NSParental asthma* 23% (15/66) 30% (8/27) 26% (10/38) 50% (3/6) 63% (10/16) 0.03VmaxFRC z score 0.08 (0.96), 0.36 (1.28), �0.21 (0.94), 0.22 (0.41), �0.59 (0.90),

at 1 mo (SD) n � 67 n � 28 n � 39 n � 6 n � 17 0.02PC40 1 mo† 1.05 (0.77, 1.45), 1.34 (0.82, 2.21), 0.81 (0.59, 1.10), 1.04 (0.17, 6.34), 0.73 (0.49, 1.09),

n � 59 n � 26 n � 30 n � 6 n � 11 NS

Definition of abbreviations: NW � no wheeze; W0–3 � wheeze before but not after the third birthday; W4–6 � wheeze between ages 4 and 6 years but not after;W11 � wheeze at 11 years but not previously; PW � those who wheezed between 4 and 6 years and at 11 years of age; PC40 � concentration of histamine provokingat least a 40% reduction in VmaxFRC.

* Data missing for 11 fathers.† Expressed as geometric mean and 95% confidence interval (1.96 � SEM).

aged 3 years, 34 (27%) individuals aged 4 years, 33 (31%) aged5 years, 26 (22%) aged 6 years, and 31 (17%) aged 11 years.The incidence of wheeze was inversely related to age (�2

5 �24.84, p � 0.001). At 11 years of age, 55 (28%) children had a

history of wheeze ever. A history of PDA was confirmed byquestionnaires in 55 (45%) children aged 6 years of which 28(24%) children reported current PDA; at 11 years of age, therespective figures for PDA ever and PDA currently were 73(38%) and 28 (15%).

PC40 Aged 1 Month and Respiratory Symptoms at 11 Years

Where PC40 at 1 month of age was known, a history of diagnosedasthma ever by 11 years of age (n � 59) was associated withreduced PC40 at 1 month of age compared with no history ofdiagnosed asthma (n � 106) (geometric means 0.72 [95% confi-dence interval [CI], 0.56, 0.94] vs. 1.09 [95% CI, 0.86, 1.38] p �0.04). There was no reduction in PC40 at 1 month and a historyof wheeze ever (geometric mean, 0.86 [95% CI, 0.67, 1.11]; n �97) compared with NW ever (geometric mean, 0.97 [95% CI,0.74, 1.29]; n � 97). There was no relationship between PC40 at1 month and current wheeze and current PDA at 11 years of age.

VmaxFRC at 1 Month and Respiratory Symptoms betweenAges 4 and 11 Years: Cross-sectional Analyses

Wheeze between 4 and 6 years of age was associated with areduced mean VmaxFRC z score (�0.31 [SD 0.96], n � 64) com-

Figure 1. Box and whisker plot showing median and quartilesvalues for z scores of VmaxFRC at 1 month in groups definedby wheeze at different ages. Data from four individuals withz scores of 2.5, 3.8, 3.8, and 4.0 are not included but wereincluded in the analysis. NW � no wheeze reported at anyage; W0–3 � wheeze before but not after the third birthday;W4–6 � wheeze between ages 4 and 6 years but not after;W11 � wheeze at 11 years but not previously; PW � thosewho wheezed between 4 and 6 years and at 11 years of age.

pared with NW during this period (0.16 [SD 1.06], n � 95, p �0.005). Recent wheeze at 11 years of age was associated with areduced mean VmaxFRC z score aged 1 month (�0.41 [SD 0.91],n � 31) when compared with no recent wheeze (0.04 [SD 1.06],n � 153, p � 0.03). There was no significant reduction in meanVmaxFRC z score aged 1 month for those individuals with cur-rent PDA when compared with other children at 6 years of age(�0.21 [SD 0.99], n � 28, vs. 0.02 [SD 1.03], n � 89) and 11years of age (�0.21 [SD 1.00], n � 28, vs. 0.00 [SD 1.03],n � 157).

Factors Associated with Different Wheezing Outcomes

The mean VmaxFRC z score aged 1 month for individuals inthe NW group was 0.08 (SD 0.96, n � 67), for those in the W0–3group was 0.36 (SD 1.28, n � 28), for those in the W4–6 groupwas �0.21 (SD 0.94, n � 39), for those in the W11 group was0.22 (SD 0.41, n � 6) and �0.59 (SD 0.90, n � 17) for those inthe PW group (analysis of variance, p � 0.02; Table 2 and Figure1). When both VmaxFRC and increased AR at 11 years wereconsidered, PW was more likely to be present for those individu-als in both the lowest terctile for VmaxFRC z score and withincreased AR, �2

8 for trend across groups with increased AR �19.0, p � 0.015 (Figure 2), for trend across groups without in-creased AR �2

8 � 12.2 (p � 0.1). PW was also most prevalentamong those individuals in the lowest tercile for VmaxFRC z

924 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

Figure 2. The relationship between terciles of VmaxFRC at1 month of age, increased airway responsiveness (AR) aged11 years, and respiratory symptoms in childhood. The groupcontaining individuals with z scores of VmaxFRC aged 1 monthin the lowest tercile and increased AR aged 11 years alsocontained the majority of individuals with persistent wheeze(black). �2

6 for trend across groups with increased AR � 19.0,p � 0.015.

score aged 1 month and atopy aged 11 years (Figure 3), �28 for

trend across groups with atopy � 25.0, p � 0.002, and �28 for

trend across groups without atopy � 3.6, p � 0.8.

Comparisons between Wheezing Groups at 6 and11 Years of Age

At 6 and 11 years of age, the prevalence of atopy, the doseresponse slope, and prevalence of PDA differed significantlybetween groups (Tables 3 and 4). There was also a nonsignificanttrend for reduced FEF25–75% at 6 and 11 years of age to be associ-ated with PW when compared with NW. At 11 years of age,mean z scores for FEV1 were higher for the W11 and W0–3groups compared with W4–6 and PW groups (analysis of vari-ance, p � 0.005).

Longitudinal Tracking of Pulmonary Function

Pulmonary function data were available at 1 month, 6 years, and11 years of age in 95 individuals. Table 1 compares the details

Figure 3. The relationship between terciles of VmaxFRC at1 month of age, atopy aged 11 years, and respiratory symp-toms in childhood. The group containing individuals with zscores of VmaxFRC aged 1 month in the lowest tercile andatopic aged 11 years also contained the majority of individualswith persistent wheeze (black). �2

6 for trend across groups withatopy � 25.0, p � 0.002.

of these individuals with the entire cohort. Figure 4 illustrateshow the mean z scores for all measurements of pulmonary func-tion (i.e., VmaxFRC at one month and FEF25–75 at ages 6 and11 years) within each wheezing group were consistently lowerduring the period of follow-up for the group with PW (�0.57,SD 0.91) compared with the NW (0.19, SD 0.88) and W0–3groups (0.08, SD 1.04) (analysis of variance, p � 0.001).

Cox Proportional Hazards Model

When all variables were considered, VmaxFRC aged 1 month(hazards ratio � 0.18; 95% CI, 0.00, 0.73; p � 0.03) and DRSaged 11 years (hazards ratio � 8.68; 95% CI, 3.27, 23.1; p �0.001) were independently associated with PW.

DISCUSSION

This study was designed to determine the relationship betweenlung function in early life and respiratory outcome in later child-hood, and the data suggest that airway function in early infancywas associated with persistent childhood wheeze. Cross-sectional

Turner, Palmer, Rye, et al.: Infant Lung Function and Asthma 925

TABLE 3. DETAILS AT 6 YEARS OF AGE IN CHILDREN GROUPS DEFINED BY CHILDHOOD WHEEZE

NW W0–3 W4–6 W11 PW Trend Test

Atopic 31% (13/39) 24% (4/20) 32% (8/25) 0% (0/5) 77% (9/12) p � 0.009Dose response slope* 2.8 (1.8, 4.3), n � 38 4.1 (2.0, 7.0), n � 18 3.3 (2.0, 5.0), n � 22 3.2 (0.2, 9.0), n � 4 18.9 (7.1, 44.3), n � 12 p � 0.001Mean FEF25–75 z score (SD) 0.28 (0.85), n � 38 0.03 (1.11), n � 18 �0.01 (0.84), n � 24 �0.41 (0.51), n � 4 �0.43 (0.80), n � 12 NSMean FEV1 z score (SD) 0.09 (0.79), n � 40 �0.14 (0.91), n � 19 �0.09 (0.73), n � 25 �0.32 (0.29), n � 4 �0.19 (0.72), n � 12 NSDiagnosed asthma 5% (2/44) 10% (2/21) 36% (10/28) 20% (1/5) 72% (11/15) p � 0.001

Definition of abbreviations: NW � no wheeze; W0–3 � wheeze before but not after the third birthday; W4–6 � wheeze between ages 4 and 6 years but not after;W11 � wheeze at 11 years but not previously; PW � those who wheezed between 4 and 6 years and at 11 years of age.

* Expressed as geometric mean � 95% confidence interval (1.96 � SEM).

analyses demonstrated a relationship between reduced VmaxFRCaged 1 month and wheeze between ages 4 to 6 and also at 11years of age. Longitudinal analysis revealed that reduced neonatallung function was associated with wheezing at age 4 to 6 yearsthat persisted to 11 years of age. In the final analysis, reducedVmaxFRC at 1 month of age was shown to be associated withPW, and this relationship was independent of atopy and in-creased AR in infancy and childhood and, additionally, factorsthat may influence VmaxFRC. Individuals with atopy or in-creased AR at 11 years of age who also had reduced VmaxFRCaged 1 month were most likely to have PW, but the influenceof increased AR subsumed that of atopy. Because the groupwith PW has the usual phenotype for childhood asthma, the datasuggested that for many children asthma is associated with bothreduced VmaxFRC at 1 month of age and increased AR at 11years of age (Figure 5).

Current understanding of the relationship between infantpulmonary function and childhood asthma has been mostly in-fluenced by a study from Tucson, which demonstrated an associa-tion between reduced VmaxFRC and transient and not PW (4).The findings of this study are in contrast to the previous studybecause we have found individuals with transient wheeze to havenormal pulmonary function at 1 month of age. The techniquesused to determine infant pulmonary function in the two studieswere very similar. The wheeze outcomes at 11 years of age can-not be compared between the two studies because the Tucsongroup has not published the relationship between VmaxFRC ininfancy and wheeze at 11 years of age for their cohort. In keep-ing with our findings, a study of young children with recurrentwheeze with a follow-up to 6 years of age has reported that in-dividuals with persisting wheeze had reduced VmaxFRC at 17months of age when compared with individuals with transientwheeze (8). Both our study and the Tucson study agree that re-duced VmaxFRC in infancy is associated with reduced FEF25–75%

at 11 years of age (12, 13, 16). We are not able to account forthe different outcomes that we have observed from our cohortcompared with the cohort in Tucson.

TABLE 4. DETAILS AT 11 YEARS OF AGE IN CHILDREN GROUPS DEFINED BY CHILDHOOD WHEEZE

NW W0–3 W4–6 W11 PW Trend test

Atopic 53% (34/64) 46% (13/28) 42% (15/36) 50% (3/6) 88% (15/17) p � 0.03Dose–response slope*

(95% CI) 1.6 (1.2, 2.1), n � 66 2.3 (1.4, 3.4), n � 29 1.7 (1.2, 2.4), n � 34 2.4 (1.4, 3.6), n � 6 8.2 (3.2, 17.4), n � 17 p � 0.001Mean FEF25–75 z score

(SD) 0.08 (0.74), n � 64 �0.28 (0.60), n � 28 0.08 (0.86), n � 37 0.06 (0.73), n � 6 �0.46 (1.01), n � 16 NSMean FEV1 z score

(SD) �0.03 (0.97), n � 65 0.41 (0.97), n � 28 �0.21 (0.96), n � 37 1.10 (0.88), n � 6 �0.28 (1.0), n � 17 0.005Diagnosed asthma 3% (2/67) 4% (1/28) 10% (4/39) 50% (3/6) 82%* (14/17) p � 0.001

Definition of abbreviations: CI � confidence interval; NW � no wheeze; W0–3 � wheeze before but not after the third birthday; W4–6 � wheeze between ages 4and 6 years but not after; W11 � wheeze at 11 years but not previously; PW � those who wheezed between 4 and 6 years and at 11 years of age.

* Expressed as geometric mean � 95% confidence interval (1.96 � SEM).

A relationship between reduced VmaxFRC in infants, in-creased AR in childhood, and PW might explain why earlywheeze is transient in some children but persistent in others.Reduced VmaxFRC in infancy in the absence of increased AR inlater childhood has, in our cohort, been associated with transientwheeze (12). This study reports that persistent childhood wheezewas, for the majority of cases, present for those individuals withboth reduced VmaxFRC at 1 month and greater levels of ARat 11 years of age. Our data are consistent with other studiesthat have reported associations between abnormalities of lungfunction and wheeze in early childhood (4, 14) and between in-creased AR and PW or asthma in later childhood (24). Childhoodasthma is commonly considered to be a complex, multifactorialcondition, and the increased likelihood of persistent symptoms forindividuals with both reduced VmaxFRC in infancy and increasedAR in childhood is plausible.

The findings of this study suggested that increased AR ininfancy and childhood is associated with different wheezing phe-notypes. Increased AR present at 1 month of age was associatedwith future asthma that often resolved, whereas increased ARpresent at 11 years of age was associated with persisting asthma.There was a trend for individuals in groups W4–6 and PW tohave increased AR at 1 month of age compared with othergroups, and there is a possibility that with larger numbers ofstudy subjects this trend may have become significant. In thiscohort, at 4 weeks of age, increased AR was not influenced byatopy (7), and therefore, increased AR present in infancy maybe a nonatopic mechanism for wheeze in younger children. At11 years of age, increased AR was associated with persistentrespiratory symptoms and atopy. Stein and colleagues (26) haveproposed that childhood wheeze could be considered as earlynonatopic wheeze and later atopic wheeze. Our data would sup-port this concept of wheeze and suggest that the presence ofincreased AR in either or both infancy or childhood may be animportant determinant of wheezing phenotype.

In a previous report, we have reported an association betweenreduced VmaxFRC at 1 month of age, as evidenced by flow

926 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

Figure 4. Chart demonstrating mean z scores for VmaxFRC at 1 monthand FEF25–75% at 6 and 11 years of age for groups determined by wheezeoutcome.

limitation during tidal expiration, transient wheeze, and in-creased AR in childhood (16). In this study, we report thatreduced VmaxFRC and increased AR in childhood were inde-pendently related to PW. These outcomes, taken from the samecohort, appear to be contradictory. However, at 11 years of age,the formerly flow-limited individuals were no more likely to beatopic than other cohort members (16), whereas in this study,those individuals with PW were mostly atopic. The data fromour study therefore suggest that persistent respiratory symptomsare not associated with increased childhood AR per se but in-creased childhood AR associated with atopy.

This study has confirmed previous observations that measure-ments of VmaxFRC in infancy correlate with measurements ofFEF25–75% in childhood (12, 13, 16). Maximal flow at functionalresidual capacity and FEF25–75% are flow-related measurementswith relatively large intrasubject variability and as such a strong

Figure 5. Schematic representation of the relationship between re-duced VmaxFRC in infancy and AR in childhood, which appears to beimportant to persistent childhood wheeze and asthma.

interrelationship might not be expected. The coefficient of varia-tion for VmaxFRC in infants may vary between 11% and 36%(27), although measurement of VmaxFRC becomes less variablein older infants (28), and the coefficient of variation for FEF25–75%

in 7 year olds is 15% (29). Children with wheeze have reducedFEF25–75% but not reduced FEV1 or FVC, suggesting that FEF25–75%

is a sensitive measurement of pulmonary dysfunction despiteincreased variability (10).

There are at least two separate mechanisms that may explainthe relationship between reduced VmaxFRC at 1 month andreduced FEF25–75% and increased wheeze throughout childhood.First, wheeze may be the result of narrow, small airways. Thishypothesis is supported by studies of infant pulmonary functionthat have reported reduced total respiratory conductance in chil-dren that subsequently developed wheeze (30, 31). Alternatively,altered airway compliance may result in increased wheezing;abnormal airway wall properties have been demonstrated ininfants with a history of wheeze (32). VmaxFRC does not distin-guish between reduced airway caliber and altered airway compli-ance; therefore, this study is not able to determine the specificunderlying abnormality of pathophysiology.

The main findings of this study are based on a proportion ofthe original cohort with a relatively lower level of in utero smokeexposure, and this loss to follow-up may have affected the out-comes because antenatal exposure to tobacco products has beenassociated with reduced VmaxFRC in our cohort (21) and an-other (33). Despite adjusting VmaxFRC for exposure to in uterosmoke exposure, this study may not be able to exclude definitivelya relationship between exposure to in utero tobacco productsand increased childhood wheeze. Two recent studies involvinglarge numbers of 6- and 11-year old Perth children (34, 35) havereported prevalences of wheeze very similar to that reportedin this study, and this suggests that the symptom frequency re-ported by cohort members was representative of the generalpopulation.

Compared with our findings, two other studies have reporteda larger proportion of individuals with wheeze in the first 3 yearsof life compared with the second 3 years (4, 5). One possibleexplanation for this apparent difference between our study andother is that wheeze in the first 3 years was underreported inour study because questionnaire data were not available for allstudy subjects. A second consequence of incomplete question-naire data is that we cannot exclude the possibility that wheezewas present in the first 3 years but not reported for some childrenwith W4–6 and PW.

In summary, this study demonstrated that reduced VmaxFRCat 1 month was associated with PW at 11 years of age. The datasuggested that the mechanism for PW in many children involvesboth an intrinsic abnormality in pulmonary function, as evi-denced by reduced VmaxFRC, determined at an early age andthe later onset of increased AR associated with atopy. Our data

TABLE 5. THE FINAL OUTPUT FROM A COX PROPORTIONALHAZARDS REGRESSION MODEL IN WHICH THE OUTCOMEVARIABLE WAS CODED PERSISTENT WHEEZE (� 1) ANDNO WHEEZE (� 0)

RR 95% CI for RR p Value

Dose–response slope aged 11 yr 8.68 3.27, 23.1 � 0.001VmaxFRC aged 1 mo 0.18 0.00, 0.73 0.03

Definition of abbreviations: CI � confidence interval; RR � relative risk.Predictive variables included log VmaxFRC and log PC40 aged 1 month, log

dose response slope aged 11 years, atopy aged 11 years, sex, maternal smokingduring pregnancy, age (weeks), length (cm), and weight (kg) at 1-monthassessment.

Turner, Palmer, Rye, et al.: Infant Lung Function and Asthma 927

may therefore help to explain why asthma does not develop inall atopic children and why early childhood wheeze does notpersist in some individuals. The incidence of childhood wheezeis increasing (36), and mechanisms responsible for reducedVmaxFRC in early life and increasing atopy in childhood requirefurther study.

Conflict of Interest Statement : S.W.T. has no declared conflict of interest; L.J.P.has no declared conflict of interest; P.J.R. has no declared conflict of interest;N.A.G. has no declared conflict of interest; P.K.J. has no declared conflict ofinterest; M.C. has no declared conflict of interest; S.Y. has no declared conflictof interest; J.G. has no declared conflict of interest; L.I.L. received airfares, accom-modation and honorarium ($1500) from GSK for speaking at approximately twoconferences/workshops each year; P.N.L. has no declared conflict of interest.

Acknowledgment : The authors acknowledge the contribution of many colleaguesover the last 15 years and are indebted to the families involved in the OsbornePark family asthma study.

References

1. Park ES, Golding J, Carswell F, Stewart-Brown S. Preschool wheezingand prognosis at 10. Arch Dis Child 1986;61:642–646.

2. Yunginger JW, Reed CE, O’Connell EJ, Melton LJ III, O’Fallon WM,Silverstein MD. A community-based study of the epidemiology of asthma:incidence rates, 1964–1983. Am Rev Respir Dis 1992;146:888–894.

3. Kelly WJ, Hudson I, Phelan PD, Pain MC, Olinsky A. Childhood asthmain adult life: a further study at 28 years of age. Br Med J (Clin Res Ed)1987;294:1059–1062.

4. Martinez FD, Wright AL, Taussig LM, Holberg CJ, Halonen M, MorganWJ. Asthma and wheezing in the first six years of life: the GroupHealth Medical Associates. N Engl J Med 1995;332:133–138.

5. Rusconi F, Galassi C, Corbo GM, Forastiere F, Biggeri A, Ciccone G,Renzoni E. Risk factors for early, persistent, and late-onset wheezingin young children: SIDRIA Collaborative Group. Am J Respir CritCare Med 1999;160:1617–1622.

6. Clough JB, Keeping KA, Edwards LC, Freeman WM, Warner JA, War-ner JO. Can we predict which wheezy infants will continue to wheeze?Am J Respir Crit Care Med 1999;160:1473–1480.

7. Palmer LJ, Rye PJ, Gibson NA, Burton PR, Landau LI, Le Souef PN.Airway responsiveness in early infancy predicts asthma, lung function,and respiratory symptoms by school age. Am J Respir Crit Care Med2001;163:37–42.

8. Delacourt C, Benoist MR, Waernessyckle S, Rufin P, Brouard JJ, deBlic J, Scheinmann P. Relationship between bronchial responsivenessand clinical evolution in infants who wheeze: a four-year prospectivestudy. Am J Respir Crit Care Med 2001;164:1382–1386.

9. Lombardi E, Morgan WJ, Wright AL, Stein RT, Holberg CJ, MartinezFD. Cold air challenge at age 6 and subsequent incidence of asthma: alongitudinal study. Am J Respir Crit Care Med 1997;156:1863–1869.

10. Gold DR, Wypij D, Wang X, Speizer FE, Pugh M, Ware JH, Ferris BGJr, Dockery DW. Gender- and race-specific effects of asthma andwheeze on level and growth of lung function in children in six UScities. Am J Respir Crit Care Med 1994;149:1198–1208.

11. Oswald H, Phelan PD, Lanigan A, Hibbert M, Carlin JB, Bowes G,Olinsky A. Childhood asthma and lung function in mid-adult life.Pediatr Pulmonol 1997;23:14–20.

12. Turner SW, Young S, Landau L, Le Souef PN. Reduced lung functionboth before bronchiolitis and at 11 years. Arch Dis Child 2002;87:417–420.

13. Castro-Rodriguez JA, Holberg CJ, Wright AL, Halonen M, Taussig LM,Morgan WJ, Martinez FD. Association of radiologically ascertainedpneumonia before age 3 yr with asthmalike symptoms and pulmonaryfunction during childhood: a prospective study. Am J Respir Crit CareMed 1999;159:1891–1897.

14. Clarke JR, Salmon B, Silverman M. Bronchial responsiveness in theneonatal period as a risk factor for wheezing in infancy. Am J RespirCrit Care Med 1995;151:1434–1440.

15. Murray CS, Pipis SD, McArdle EC, Lowe LA, Custovic A, Woodcock A,National Asthma Campaign-Manchester A and Allergy Study Group.Lung function at one month of age as a risk factor for infant respiratorysymptoms in a high risk population. Thorax 2002;57:388–392.

16. Turner SW, Palmer LJ, Rye PJ, Gibson NA, Judge P, Young S, LandauLI, Le Souef PN. Infants with flow-limitation at 4 weeks: outcome at6 and 11 years. Am J Respir Crit Care Med 2002;165:1294–1298.

17. Le Souef PN, Turner SW, Palmer LJ, Rye PJ, Gibson NA, Judge P,Young S, Landau LI. Pulmonary function at four weeks correlates withpulmonary function at 6 and 12 years [abstract]. Am J Respir Crit CareMed 2001;163:A541.

18. Turner SW, Palmer LJ, Rye PJ, Gibson NA, Judge V, Young S, LandauLI, Le Souef PN. Reduced lung function at one month is associatedwith asthma at eleven years [abstract]. Arch Dis Child 2002;86:A37.

19. Young S, Le Souef PN, Geelhoed GC, Stick SM, Turner KJ, Landau LI.The influence of a family history of asthma and parental smoking onairway responsiveness in early infancy. N Engl J Med 1991;324:1168–1173.

20. Sly PD, Tepper R, Henschen M, Gappa M, Stocks J. Tidal forced expira-tions: ERS/ATS Task Force on Standards for Infant Respiratory Func-tion Testing. Eur Respir J 2000;16:741–748.

21. Young S, Sherrill DL, Arnott J, Diepeveen D, Le Souef PN, Landau LI.Parental factors affecting respiratory function during the first year oflife. Pediatr Pulmonol 2000;29:331–340.

22. Standardisation of spirometry: 1987 update: Statement of the AmericanThoracic Society. Am Rev Respir Dis 1987;136:1285–1298.

23. Yan K, Salome C, Woolcock AJ. Rapid method for measurement ofbronchial responsiveness. Thorax 1983;38:760–765.

24. Salome CM, Peat JK, Britton WJ, Woolcock AJ. Bronchial hyperrespon-siveness in two populations of Australian schoolchildren: I: relationto respiratory symptoms and diagnosed asthma. Clin Allergy 1987;17:271–281.

25. Pepys J. Skin tests for immediate, type I, allergic reactions. Proc R SocMed 1972;65:271–272.

26. Stein RT, Holberg CJ, Morgan WJ, Wright AL, Lombardi E, Taussig L,Martinez FD. Peak flow variability, methacholine responsiveness andatopy as markers for detecting different wheezing phenotypes in child-hood. Thorax 1997;52:946–952.

27. Le Souef PN. Forced expiratory manouvres: infant respiratory functiontesting. New York: Wiley-Liss; 1996. p. 379–402.

28. Henschen M, Stocks J. Assessment of airway function using partial expir-atory flow-volume curves: how reliable are measurements of maximalexpiratory flow at FRC during early infancy? Am J Respir Crit CareMed 1999;159:480–486.

29. Strachan DP. Repeatability of ventilatory function measurements in apopulation survey of 7 year old children. Thorax 1989;44:474–479.

30. Dezateux C, Stocks J, Dundas I, Fletcher ME. Impaired airway functionand wheezing in infancy: the influence of maternal smoking and agenetic predisposition to asthma. Am J Respir Crit Care Med 1999;159:403–410.

31. Martinez FD, Morgan WJ, Wright AL, Holberg CJ, Taussig LM. Dimin-ished lung function as a predisposing factor for wheezing respiratoryillness in infants. N Engl J Med 1988;319:1112–1117.

32. Frey U, Makkonen K, Wellman T, Beardsmore C, Silverman M. Alter-ations in airway wall properties in infants with a history of wheezingdisorders. Am J Respir Crit Care Med 2000;161:1825–1829.

33. Tager IB, Ngo L, Hanrahan JP. Maternal smoking during pregnancy:effects on lung function during the first 18 months of life. Am J RespirCrit Care Med 1995;152:977–983.

34. The ISAAC Steering Committee. Worldwide variations in the prevalenceof asthma symptoms: the International Study of Asthma and Allergiesin Childhood (ISAAC). Eur Respir J 1998;12:315–335.

35. Oddy WH, Holt PG, Sly PD, Read AW, Landau LI, Stanley FJ, KendallGE, Burton PR. Association between breast feeding and asthma in 6year old children: findings of a prospective birth cohort study. Br MedJ (Clin Res Ed) 1999;319:815–819.

36. Kuehni CE, Davis A, Brooke AM, Silverman M. Are all wheezing disor-ders in very young (preschool) children increasing in prevalence?Lancet 2001;357:1821–1825.

The Evolution of Airway Function in Early ChildhoodFollowing Clinical Diagnosis of Cystic FibrosisSarath C. Ranganathan, Janet Stocks, Carol Dezateux, Andrew Bush, Angie Wade, Siobhan Carr,Rosemary Castle, Robert Dinwiddie, Ah-Fong Hoo, Sooky Lum, John Price, John Stroobant, Colin Wallis,and The London Collaborative Cystic Fibrosis Group

Portex Anaesthesia, Intensive Therapy and Respiratory Medicine Unit; Centre for Paediatric Epidemiology and Biostatistics, Institute of ChildHealth; Department of Paediatric Respiratory Medicine, Royal Brompton Hospital; Department of Child Health, Royal London Hospital;Department of Paediatric Respiratory Medicine, Great Ormond Street Hospital; Department of Child Health, King’s College Hospital;Department of Child Health, University Hospital Lewisham; and Neonatal Unit, Homerton University Hospital, London, United Kingdom

This study aimed to investigate the evolution of airway function ininfants newly diagnosed with cystic fibrosis (CF). FEV0.5 was mea-sured soon after diagnosis (median age of 28 weeks) and 6 monthslater in subjects with CF and on two occasions 6 months apart(median ages of 7.4 and 33.7 weeks) in healthy infants, using theraised-volume technique. Repeated measurements were successfulin 34 CF and 32 healthy subjects. After adjustment for age, length,sex, and exposure to maternal smoking, mean FEV0.5 was signifi-cantly lower in infants with CF both shortly after diagnosis and atthe second test, with no significant difference in rate of increasein FEV0.5 with growth between the two groups. When comparedwith published reference data, FEV0.5 was reduced by an averageof two z scores on both test occasions in those with CF, with 72%of individuals having an FEV0.5 of less than 1.64 z-scores (i.e., lessthan the fifth percentile) on one or both test occasions. On longitu-dinal analysis, subjects with CF experienced a mean (95% confi-dence interval) reduction in FEV0.5 of 20% (11, 28). Airway functionis diminished soon after diagnosis in infants with CF and does notcatch up during infancy and early childhood. These findings haveimportant implications for early interventions in CF.

Keywords: cystic fibrosis; respiratory function tests; infant; early inter-vention; forced expiration

Cystic fibrosis (CF) is the most common lethal inherited diseaseaffecting Northern European populations, with a birth preva-lence of approximately 1:2,500. It is a multisystem disorder, withrespiratory morbidity secondary to chronic inflammation andinfection being the leading cause of death. Recent survival datafrom both the United Kingdom and the United States indicateimproving survival in successive birth cohorts (1, 2), suggestingthat events in early life exert an important influence on outcome.There has therefore been increased interest in the early naturalhistory of CF lung disease and in interventions that might preventor delay the progression of pulmonary disease.

Recent studies have suggested that inflammation in the CFlung develops very early in life, even in asymptomatic infants(3–5). We have reported diminished airway function soon after

(Received in original form September 30, 2003; accepted in final form January 25, 2004)

Supported by the Cystic Fibrosis Trust UK, Portex Ltd., the Dunhill Medical Trust,and the Foundation for the Study of Sudden Infant Death (research at the Instituteof Child Health and Great Ormond Street Hospital for Children National HealthService (NHS) Trust benefits from research and development funding receivedfrom the NHS Executive).

Correspondence and requests for reprints should be addressed to Sarath Rangana-than, M.B.Ch.B., M.R.C.P., M.R.C.P.C.H., Ph.D., Portex Unit, 6th Floor, CardiacWing, Institute of Child Health, 30 Guilford Street, London WC1N 1EH, UK. E-mail:drsarath@clara.net

Am J Respir Crit Care Med Vol 169. pp 928–933, 2004Originally Published in Press as DOI: 10.1164/rccm.200309-1344OC on January 30, 2004Internet address: www.atsjournals.org

diagnosis in infants with CF, even in the absence of any priorclinically recognized lower respiratory illness (6). Similarly, ab-normalities of airway function were detected in 25% of thosewhose respiratory status was considered to be normal by a CFspecialist (7). However, the natural history of these early lungfunction abnormalities is unclear, as few longitudinal studiesexist (8). The aims of this study were to determine whetherinitial impairment in airway function noted in infants after aclinical diagnosis of CF persists despite treatment in specialistcenters and if so whether this is independent of somatic growthimpairment when compared both with repeated measurementsmade in a group of healthy infants without CF and with recentlypublished cross-sectional reference data (9).

METHODS

Forty-seven infants and young children (less than 24 months) newlydiagnosed with CF by sweat test and/or by positive genotype for CFmutations (10) were recruited between January 1999 and May 2001 fromfive specialist centers in London as previously reported (7). Newbornscreening was not available for infants who were managed in the collab-orating centers for the duration of this study. Full details of methodsof recruitment, eligibility criteria, and background characteristics ofthis cohort have been published previously (6, 7). Details of the modeof presentation before hospital admission or current or prior treatmentwith intravenous, inhaled, or oral antibiotics were obtained by inspec-tion of medical records and parental report. A cough swab was takenapproximately every 2 months and at each attendance for lung functiontests. All past microbiological assessments were reviewed at the timeof the test. Healthy infants were recruited as part of an ongoing epidemi-ologic study (11). Healthy infants in whom at least one prior lungfunction test had been performed were eligible for follow-up if lessthan 15 months of age. Parents of both CF and healthy subjects gaveinformed written consent. The study was approved by the North ThamesMulticentre Research Ethics Committee and the Local Research EthicsCommittees of the participating hospitals.

Measurement of Airway Function

Both CF and healthy children were tested when well and clinically freefrom upper respiratory tract infections for at least 3 weeks. Measure-ments were made as soon as possible after diagnosis and 6 months laterin those with CF and on two occasions, approximately 6 months apart,in healthy subjects.

All measurements were performed by a single specialist team ac-cording to a standardized protocol as described previously (7, 11, 12).On the day of testing, subjects were weighed, and crown–heel lengthwas measured. Weight and length percentiles were calculated fromstandard UK growth reference data (13). Exposure to maternal smokingprenatally and postnatally was assessed from maternal report and cur-rent smoking habits confirmed by maternal salivary cotinine (14). Allsubjects were studied in the supine position after sedation with an oraldose of 60–100 mg/kg of chloral hydrate or an equivalent dose oftriclofos sodium.

Full-forced expiratory maneuvers were performed using the raised-volume rapid thoracoabdominal compression technique (7, 11, 12, 15)

Ranganathan, Stocks, Dezateux, et al.: Lung Function in Infants with CF 929

from which parameters of airway function similar to those obtainedusing spirometry in older subjects were measured. During the raisedvolume technique, expiration was forced from an inflation pressure of3 kPa. Maneuvers were repeated until a minimum of two (usually three)acceptable and reproducible flow-volume curves (sum of FVC andFEV0.5 being within 10% of each other) were obtained. FVC, FEV0.5,and forced expiratory flow when 75% of FVC had been expired (FEF75)were reported from the “best” flow-volume curve (defined as the techni-cally acceptable maneuver with the highest sum of FVC and FEV0.5)(11, 12, 16).

Statistical Analysis

Although our group has reported FEV0.4 previously (12), currently, noreference data exist for this parameter. Therefore, we calculated FEV0.5

in this study to compare our results with published cross-sectionalreference data for the raised-volume technique using similar equipmentand technique (9). Thus, airway function for subjects with CF wascompared directly with these reference data, using prediction equationsincorporating coefficients for both age and length, in addition to therepeated measurements made in the healthy control subjects. Lungfunction parameters were reported in absolute terms, and z scores werecalculated for FEV0.5, FVC, and FEF75 from the reference data. Valuesfalling below 1.645 z scores (i.e., less than the 5th percentile) wereconsidered to be unusually low (17). Multiple linear regression wasused as previously described (6) to estimate the reduction in airwayfunction in infants with CF on each of the two test occasions afteraccounting for differences in length, age, weight, sex, and exposure tomaternal smoking. Multilevel modeling (MLWin version 1.10; Instituteof Education, London, UK) was used to compare longitudinal trendsin measurements after adjusting for the same factors. All parametersof airway function were log transformed before multilevel modeling.CF-length and CF-age interactions were assessed for evidence thatproportionate growth in airway function differed between CF andhealthy infants. Spearman correlation was used to assess whether thevalue of airway function expressed as a z score on the second testoccasion was related to that obtained at the first test. Paired t testswere used to assess changes in z scores between tests. Airway functionat follow-up was thus assessed according to initial ranking to determinewhether this position was maintained with growth (“tracking”).

RESULTS

Sixty-three subjects less than 24 months of age were diagnosedwith CF during the study. Sixteen infants were not recruited.The reasons for this included parental refusal (n � 6), ventilationfor respiratory failure (n � 2), adverse social circumstances (n �1), distance from laboratory too great (n � 1), associated Arnold-Chiari malformation (n � 1), and Pierre-Robin syndrome (n �1). Four infants were recruited but failed to attend for the initiallung function test on at least two occasions. There was no differ-ence in the background characteristics of those who were andwere not recruited (data not shown). Forty-seven infants withCF were therefore studied at a median (range) of 12 (1 to 37)weeks after diagnosis. Thirty-seven were restudied after a median

TABLE 1. BACKGROUND DETAILS OF STUDY INFANTS

Difference: CF � controlCystic Fibrosis Control Subjects (95% CI)

n 37 33Male, n (%) 14 (38) 16 (48) �11% (�36, 13)White n (%) 36 (97) 32 (97) 0.3% (�2, 1)Maternal smoking, n (%) 12 (32) 7 (21) 11% (�13, 37)Mean (SD) gestational age, wk 38.8 (2.2) 39.9 (1.2) �1.1 (�2.0, �0.3)*Mean (SD) birthweight, kg 3.03 (0.58) 3.37 (0.37) �0.34 (�0.57, �0.10)*Mean (SD) birthweight percentile 42.1 (34.2) 44.7 (23.7) �2.6 (�16.8, 11.6)

Definition of abbreviations: CF � cystic fibrosis; CI � confidence interval.* p � 0.01.

(interquartile range) interval of 29 (26–32) weeks. The primarymode of presentation for these infants was recurrent chest infec-tions (n � 18), failure to thrive (n � 5), meconium ileus (n �10), and meconium ileus with antenatal bowel pathology (n �4). Twenty two (60%) of the infants were homozygous for the�F508 mutation. Of the 10 infants who were not followed, 2were considered too old for sedation by the time the second testwas due (i.e., more than 30 months). Two had moved out of thearea, and the families of six infants (13%) either declined furtherassessments or failed to attend for follow-up. The backgrounddetails of the 37 infants who were followed are shown in Table 1.There were no important differences in the distributions of sex,genotype, maternal smoking, age, body weight or length, or lungfunction measurements on first test occasion in those who wereand were not successfully followed (data not shown).

Twenty five (68%) of the 37 infants with CF who were fol-lowed had been admitted to hospital for a respiratory illness ona median (range) of two (1–8), occasions, and 24 of them hadreceived between one to seven courses of intravenous antibiotics(median � 1) at some time before the second lung functionassessment. Thirteen received intravenous antibiotics before thefirst test only, eight only in the interval between the lung functiontests, and three both before the first test and between the tests.

Pseudomonas aeruginosa had been identified in eight (22%)infants by the time of the first test and in an additional eightinfants by the second test (i.e., total of 43%). Other organismsidentified by the second lung function assessment were Staphylo-coccus aureus (n � 5), methicillin-resistant S. aureus (2), Escheriacoli (7), klebsiella sp. (4), Enterobacter sp. (2), Streptococcuspneumoniae (2), and Aspergillus species (1). In nine (24%) in-fants, no organisms had been cultured by the time of the secondlung function test.

Thirty-three healthy control infants had lung function testsrepeated after a median (interquartile range) interval of 25weeks (20–42) weeks. Twenty eight of these subjects were in-cluded in the cohort of 138 infants whose airway function, mea-sured on a single occasion, we reported previously (7). Therewere no important differences in the distributions of sex, age,body weight or length, or initial lung function measurements inthose that were and were not asked to return for repeat testing(data not shown).

There was a parental report of wheezing associated with lowerrespiratory illness in two control infants before the second test,but neither required hospitalization. None had crackles orwheeze on the day of test. The background details of the healthyinfants are shown in Table 1, with anthropometric details at timeof tests summarized in Table 2. At the time of the second test,infants with CF were older and slightly longer but of similarweight to healthy control subjects. When corrected for age andsex, by expressing as z scores (13), it can be seen that infants

930 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

TABLE 2. MEASUREMENTS OF ANTHROPOMETRY AND AIRWAY FUNCTION ON EACH TEST OCCASION

Test 1 Test 2

CF Healthy Difference: CF � Healthy CF Healthy Difference: CF � HealthyMean (SD) Mean (SD) (95% CI) Mean (SD) Mean (SD) (95% CI)

Age, wk* 28.4 (16.6–43.0) 7.4 (5.7–8.9) 13.3, 25.6§ 59.0 (48.3–69.0) 33.7 (28.1–50.1) 13.1, 27.6§

Weight, kg* 6.66 (4.90–8.12) 4.82 (4.45–5.61) 0.67, 2.45§ 8.90 (8.11–10.7) 8.72 (8.18–9.13) �0.40, 1.05Weight z score§§ �1.78 (1.42) �0.11 (1.1) �2.2, �1.2§ �0.97 (1.4) 0.44 (1.0) �1.8, �1.0§

Length, cm* 66.3 (60.1–71.9) 57.5 (55.3–60.1) 4.8, 11.4§ 75.1 (72.5–80.1) 71.4 (70.0–76.1) 0.9, 4.9‡

Length z score§§ �0.73 (1.6) 0.37 (1.0) �1.7, �0.5§ �0.18 (1.5) 1.1 (1.0) �1.9, �0.7§

FVC, mL 215 (81) 162 (60) �22† (�44, �0.4)¶ 365 (93)†† 367 (84)†† �37* (�71, �1)¶

FEV0.5, mL 178 (65) 144 (51) �30‡ (�50, �9)¶ 267 (64)** 299 (56)†† �51§ (�77, �25)¶

FEF75, ml · s�1 217 (100) 210 (87) �53† (�97, �8)¶ 305 (126)†† 389 (109)‡‡ �89‡ (�147, �31)¶

FEV0.5/FVC% 83 (10) 91 (5) �4 (�8, 0)|| 74 (13) 83 (7) �8† (�13, �3)||

FVC z score¶¶ �1.9 (1.4) �0.33 (1.3) �1.57§ (�2.2, �0.9) �1.5 (1.3) �0.37 (1.3) �1.1§ (�1.8, �0.5)FEV0.5 z score¶¶ �2.1 (1.8) �0.18 (1.5) �1.6§ (�2.7, �1.1) �2.0 (1.6) �0.13 (1.3) �1.9§ (�2.5, �1.1)FEF75 z score¶¶ �1.0 (1.7) 0.04 (1.3) �1.42‡ (�1.7, �0.28) �1.0 (1.7) 0.23 (1.0) �1.2‡ (�1.9, �0.56)FEV0.5/FVC z score¶¶ �0.01 (1.2) 0.25 (0.53) �0.26 (�0.70, 0.20) �0.49 (1.8) 0.51 (0.74) �1.0‡ (�1.7, �0.31)

Definition of abbreviations: CF � cystic fibrosis; CI � confidence interval; FEF75 � forced expiratory flow when 75% of FVC had been expired.Data are expressed as mean (SD) except * median (interquartile range).† p � 0.05.‡ p � 0.01.§ p � 0.001.¶ Mean difference adjusted for length, sex and exposure to maternal smoking.|| Mean difference adjusted for age, sex, and exposure to maternal smoking.** n � 34.†† n � 32.‡‡ n � 31.§§ Freeman and colleagues (13).¶¶ Jones and colleagues (9).

with CF were significantly lighter and shorter than the healthycontrol subjects on both test occasions. All infants had been freeof symptoms of upper respiratory tract infection for at least 3weeks on the day of test. However, crackles or wheeze werestill identified on auscultation in eight (22%) infants with CFon at least one test occasion. Wheeze was present in five infantswith CF at the first test. Four subjects with CF had wheezeat the second test, all of whom had been wheezy on the firstoccasion.

Airway Function

Paired measurements of FEV0.5 were obtained in 32 of the 33healthy infants and young children and in 34 of 37 subjects withCF. FVC and FEF75 were not reported in two of the subjectswith CF in whom FEV0.5 was available because of early inspira-tion before residual volume had been attained.

The association of airway function with length according todisease status for FVC, FEV0.5, and FEF75 is shown in Figure 1.

FEV0.5 and FEF75 were significantly diminished in male sub-jects (by 14% and 28%, respectively), and there was a 26%reduction in FEF75 among infants and young children exposedto maternal smoking (data not shown). These parameters weretherefore included in the model when comparing CF with healthysubjects on each test occasion.

The adjusted mean (95% confidence interval) difference inairway function in infants and young children with CF assessedcross-sectionally on each test occasion after accounting for sig-nificant factors (length, sex, and exposure to maternal smoking)by using multiple linear regression is shown in Table 2. FVC,FEV0.5, and FEF75 were significantly diminished on both testoccasions in those with CF when adjusted for such factors.FEV0.5/FVC was strongly dependent on age rather than onlength. It was not significantly reduced at the first test in thosewith CF; however, a small but significant reduction of 8% wasidentified at the time of the second test.

Figure 2 shows the airway function of healthy children andthose with CF on both test occasions in relationship to the pub-lished normal data of Jones and colleagues (9). When expressedas z scores, all parameters of lung function, except FEV0.5/FVCat the first test, were significantly lower among those with CFthan both published normal values and those for the healthyinfants and young children in this study. Among infants withCF, 14 (41%) had an FEV0.5 that was below �1.64 z scores onboth test occasions (compared with only 6% of healthy subjects).Of 16 infants with CF in whom the z score for FEV0.5 was �1.96or less (i.e., less than the 2.5 percentile) on the first occasion,10 also had an abnormally low z score at the second test. Zscores were reduced similarly for FVC and FEF75 (Table 2).

Despite considerable within-subject variability in the rate ofincrease of airway function in relationship to somatic growth be-tween the tests, especially among those with CF (Figure 1), therewas demonstrable tracking between all parameters of airway func-tion when differences in body size and growth were accountedfor. Spearman correlations between airway function measured onthe first and second test occasion were 0.55, 0.66, and 0.80 forFVC, FEV0.5, and FEF75, respectively (p � 0.001 for all), indicatinga significant correlation of ranking of airway function betweenthe two test occasions. Mean z scores for FEV0.5 in healthy subjectswere similar to predicted values on both test occasions, with nosignificant change in these z scores between occasions (p � 0.83)(Table 2). These data suggest that the ranking of lung functionand an individual’s z score are maintained when repeat measure-ments of airway function using the raised-volume technique aremade during the first 2 years of life.

Mean z scores among those with CF also remained unchangedover the 6-month period, being �2.0 and �2.2 on the first andsecond occasions, respectively (95% confidence interval of dif-ference between tests, �0.8, 0.4; p � 0.50). There was no signifi-cant change within subjects in z score for FVC, FEF75, or FEV0.5/FVC between tests in either healthy subjects or those with CF.

Ranganathan, Stocks, Dezateux, et al.: Lung Function in Infants with CF 931

Figure 1. Association of (A ) FVC, (B ) FEV0.5, and (C ) forced expiratoryflow when 75% of FVC had been expired (FEF75) with length. The dashedand solid lines indicate airway function of healthy infants and infantswith cystic fibrosis (CF), respectively.

On multilevel (i.e., longitudinal) modeling, subjects with CFexperienced an average (95% confidence interval) reduction of21% (12, 29), 20% (11, 28), and 30% (16, 42) in FVC, FEV0.5, andFEF75, respectively, compared with healthy subjects. Althoughabsolute differences were thus greater for older and larger subjectswith CF (e.g., a mean adjusted reduction in FEV0.5 of �51 ml at thesecond test compared with �30 ml at the first), the proportionatedifference (20%) was similar at each test. There was no signifi-cant difference between groups with respect to the relative rateof increase in any parameter of airway function between tests(data not shown), with FVC, FEV0.5, and FEF75 increasing by amean (95% confidence interval) of 5.9% (5.2, 6.5), 5.8% (4.5,7.0), and 4.0% (3.0, 4.6), respectively, per centimeter of growthin length after adjustment for age, sex, and smoking in bothhealthy infants and those with CF.

The age at diagnosis, mode of presentation, isolation ofP. aeruginosa at any stage before either the first or second tests,cough on the day of test, history of hospitalization, and genotypedid not appear to influence airway function on either test occa-sion or the magnitude of change between tests in those with CF.However, the mean z score for FEV0.5 was lower for the fiveinfants who were wheezy at the first test (�4.6 vs. �1.8 in those

with and without wheeze respectively, p � 0.06) and for four ofthese five infants who were noted to be wheezy again at thesecond test (�4.2 vs. �1.7, p � 0.02). There was no differencein the change in FEV0.5 z score between tests in those subjectswith CF who did or did not have evidence of wheeze on the dayof test (p � 0.7).

DISCUSSION

We have reported previously that airway function, as measuredby the raised-volume technique, is diminished soon after clinicaldiagnosis in infants with CF (6, 7). In this follow-up study per-formed 6 months after clinical diagnosis and management inspecialist centers, this diminution in airway function persisted,reflecting the fact that airway function increased at a similar butnot greater rate in infants and young children with CF comparedwith healthy subjects; in other words, there was no “catch-up”growth in lung function. This was evident in comparison to boththe healthy control group and previously published referencedata. This finding is consistent with previous reports of reducedairway function in infants with CF early in the course of disease(6, 7). There was strong evidence of “tracking” of airway functionin both healthy subjects and those with CF; that is, those with

932 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

Figure 2. Association of FEV0.5 with length for healthy subjects (opensquare, occasion one; closed square, occasion two) and infants with CF(closed inverted triangles, occasion one; closed triangles, occasion two)in this study and that of the cross-sectional reference population ofJones and colleagues (9) (open circles).

lowest lung function initially tended to maintain this positionwith growth.

To our knowledge, this is the first study to use the raised-volume technique to perform repeated measurements of airwayfunction in healthy infants and young children and those withCF. Longitudinal studies of airway function that have been re-ported (18, 19) used techniques that are now considered lessdiscriminative than the raised-volume technique for identifyingdiminished airway function in infants with CF (7) and did notrecruit a healthy control group prospectively. Hence, it is some-what difficult to relate these results to those from previous stud-ies (8). Although comparison of our results with cross-sectionalpublished reference data (9) provided additional evidence thatin CF airway function failed to improve during infancy and earlychildhood, comparing measurements obtained in subjects withCF directly with those measured longitudinally in a healthy pop-ulation of infants using identical methods and equipment re-mains a more powerful approach.

The infants with CF recruited into this study were typical ofthose followed at the five collaborating centers. Twenty-twopercent had evidence of P. aeruginosa infection by the first testand 43% by the second test at median ages of 28 and 59 weeks,respectively. Sixty-five percent had received intravenous antibi-otics at some point before the second test, and evidence forsignificant protein–energy malnutrition was present at diagnosis.These factors suggest that our cohort had relatively severe dis-ease and may not be representative of those diagnosed by neona-tal screening or following clinical diagnosis with milder diseaseat other centers. Repeating such a study in a cohort after adiagnosis made by neonatal screening and before the onset ofclinical disease would improve our understanding of the earlypathophysiology of CF.

A number of aspects of study design merit discussion. Theinterval between tests was selected for practical reasons. Al-though a shorter interval might have permitted more than twomeasurements to be made before the age limit for infant lung

function tests was exceeded, this might also have precluded iden-tification of changes associated with growth and treatment be-tween tests. In contrast, a longer interval might have increasedthe likelihood of demonstrating change in airway function butwould have also limited the number of infants in whom testscould be repeated. Selection of suitable healthy control subjectsposed problems as matching for age or crown–heel length wouldhave resulted in the control group being longer or younger,respectively, as many infants with CF are growth retarded andhave reduced length for age. Although there appeared to havebeen some catch-up growth in those with CF, a similar trendwas noted among healthy control subjects, particularly with re-spect to length. This suggests that published growth charts maybe underestimating body size in healthy infants (and potentiallyunderestimating any growth retardation among those with dis-ease) and emphasizes the importance of a prospective controlgroup. Although the healthy subjects were younger on averagethan those with CF, their lengths, which accounted for the highestproportion of the variance in lung function between infants,were broadly similar. However, the differences in age rangebetween healthy infants and those with CF in this study couldpotentially confound interpretation of the results, particularlywith respect to rate of change. For example, developmentalchanges in lung maturation and growth might not be linearlyrelated to body size even if airway function is. For this reason,we also calculated z scores for the infants with CF by comparingmeasurements of airway function with published reference datafor healthy subjects whose ages spanned the range of our entirecohort. Our healthy infants had airway function similar to thatpredicted from the reference population, but z scores were sig-nificantly lower on each test occasion in those with CF (Figure 2)after length, age, and sex were accounted for when calculatingthe z scores. There was no significant change in z score betweentests for individuals with CF for any of the parameters of airwayfunction measured in this study. This pattern of reduction in zscores of our CF cohort on both test occasions in relationship tothe reference population provides further evidence that airwayfunction does not catch up during infancy and early childhood.

The primary cause of morbidity and mortality in patientswith CF is progressive obstructive lung disease associated withinfection and an intense neutrophilic inflammation. Inflamma-tion has been identified in infants as young as 4 weeks of ageusing bronchoalveolar lavage (3–5, 20, 21). Inflammation couldcause airway obstruction as a result of airway wall thickening,airway wall destruction leading to increased airway wall compli-ance, increased airway tone caused by increased smooth musclemass or caused by attenuation of the airway–parenchymal teth-erings, or obstruction secondary to intraluminal mucus. In ourcenters, bronchoalveolar lavage is rarely undertaken to identifypulmonary inflammation in asymptomatic infants with CF, andthus, we are unable to comment on the association betweeninflammation and diminished lung function in this cohort ofinfants. However, in a recent study, an inverse correlation wasdemonstrated between either infection or inflammation and spe-cific respiratory system compliance, and a positive correlationdemonstrated with hyperinflation in children with CF of meanage 25 months, suggesting that infection and inflammation im-pact on lung function early in CF (22). In our study, FEV0.5/FVCwas reduced at the second test in infants and young childrenwith CF compared with the healthy control subjects and couldbe considered as evidence that airway obstruction is present.However, interpretation of changes in FEV0.5/FVC during thefirst year of life is notoriously difficult because of the negative agedependency of this parameter, making it difficult to distinguisheffects of disease from those of growth and development (12).

The longer-term implications of early diminution in airway

Ranganathan, Stocks, Dezateux, et al.: Lung Function in Infants with CF 933

function in those with CF are unclear as longitudinal studies arelacking. Although there are few longitudinal studies of airwayfunction in healthy infants, those available suggest that infantswith diminished airway function shortly after birth are at anincreased risk of diminished lung function and subsequent respi-ratory morbidity in later childhood (23–26). Thus, early impair-ment of airway function may have long-term consequences in adisease where the majority of patients die because of pulmonaryinvolvement (27). The findings of our study demonstrate thatdespite early treatment at specialist centers, airway function doesnot appear to improve in a population of infants with CF, notdiagnosed by newborn screening, relative to healthy infants.Early detection of presymptomatic changes in lung function,together with the ability to assess response to treatment objec-tively, should strengthen our ability to evaluate the effectivenessof therapeutic interventions to minimize or prevent lung damagein infants with CF during a critical period of growth and develop-ment. This in turn could increase longevity and contribute to animproved quality of life for these children.

In conclusion, we have shown that airway function is dimin-ished soon after diagnosis in infants with CF and does not catchup during infancy and early childhood despite treatment in cen-ters specializing in the management of CF. We have been able toaddress important questions about the natural history of airwayfunction in infants with CF and have highlighted the value ofsuch measurements for future clinical research.

Conflict of Interest Statement : S.C.R. has no declared conflict of interest; J.S. hasno declared conflict of interest; C.D. has no declared conflict of interest; A.B. hasno declared conflict of interest; A.W. has no declared conflict of interest; S.C. has nodeclared conflict of interest; R.C. has no declared conflict of interest; R.D. has nodeclared conflict of interest; A-F.H. has no declared conflict of interest; S.L. has no de-clared conflict of interest; J.P. has no declared conflict of interest; J.S. has no declaredconflict of interest; C.W. has no declared conflict of interest.

Acknowledgment : The authors thank the families who participated in this studyand Dr. Colin Feyerabend for his analysis of the cotinine samples. The membersof the London Collaborative Cystic Fibrosis Group are as follows: Beryl Adler, IanBalfour Lynn, Andy Bush, Siobhan Carr, Rosie Castle, Kate Costeloe, Sarah Davies,Charlotte Daman-Willems, Jane Davies, Carol Dezateux, Robert Dinwiddie, JackieFrancis, Iris Goetz, Ah Fong Hoo, Jane Hawdon, Sooky Lum, Su Madge, JohnPrice, Sarath Ranganathan, Mark Rosenthal, Gary Ruiz, Janet Stocks, John Stroo-bant, Angie Wade, Colin Wallis, and Hilary Wyatt.

References

1. Lewis PA, Morison S, Dodge JA, Geddes D, Coles EC, Russell G, Lit-tlewood JM, Scott MT. Survival estimates for adults with cystic fibrosisborn in the UK between 1947 and 1967. Thorax 1999;54:420–422.

2. Kulich M, Rosenfeld M, Goss CH, Wilmott R. Improved survival amongyoung patients with cystic fibrosis. J Pediatr 2003;142:631–636.

3. Armstrong DS, Grimwood K, Carzino R, Carlin JB, Olinsky A, PhelanPD. Lower respiratory infection and inflammation in infants withnewly diagnosed cystic fibrosis. BMJ 1995;310:1571–1572.

4. Khan TZ, Wagener JS, Bost T, Martiniez J, Accurso FJ, Riches DWH.Early pulmonary inflammation in infants with cystic fibrosis. Am JRespir Crit Care Med 1995;151:1075–1082.

5. Balough K, McCubbin M, Weinberger M, Smits W, Ahrens R, Fick R.The relationship between infection and inflammation in the early stagesof lung disease from cystic fibrosis. Pediatr Pulmonol 1995;20:63–70.

6. Ranganathan S, Dezateux CA, Bush A, Carr SB, Castle R, Madge SL,Price JF, Stroobant J, Wade AM, Wallis CE, et al. Airway function ininfants newly diagnosed with cystic fibrosis. Lancet 2001;358:1964–1965.

7. Ranganathan SC, Bush A, Dezateux C, Carr SB, Hoo AF, Lum S, MadgeS, Price J, Stroobant J, Wade A, et al. Relative ability of full and partialforced expiratory maneuvers to identify diminished airway function ininfants with cystic fibrosis. Am J Respir Crit Care Med 2002;166:1350–1357.

8. Gappa M, Ranganathan SC, Stocks J. Lung function testing in infantswith cystic fibrosis: Lessons from the past and future directions. PediatrPulmonol 2001;32:228–245.

9. Jones M, Castile R, Davis S, Kisling J, Filbrun D, Flucke R, GoldsteinA, Emsley C, Ambrosius W, Tepper RS. Forced expiratory flows andvolumes in infants. Am J Respir Crit Care Med 2000;161:353–359.

10. Rosenstein BJ, Cutting GR. The diagnosis of cystic fibrosis: a consensusstatement: Cystic Fibrosis Foundation Consensus Panel. J Pediatr 1998;132:589–595.

11. Lum S, Hoo AF, Dezateux C, Goetz I, Wade A, DeRooy L, CosteloeK, Stocks J. The association between birthweight, sex, and airway func-tion in infants of nonsmoking mothers. Am J Respir Crit Care Med2001;164:2078–2084.

12. Ranganathan SC, Hoo AF, Lum SY, Goetz I, Castle RA, Stocks J.Exploring the relationship between forced maximal flow at functionalresidual capacity and parameters of forced expiration from raised lungvolume in healthy infants. Pediatr Pulmonol 2002;33:419–428.

13. Freeman JV, Cole TJ, Chinn S, Jones PRM, White EM, Preece MA.Cross sectional stature and weight reference curves for the UK, 1990.Arch Dis Child 1995;73:17–24.

14. Jarvis MJ, Tunstall-Pedoe H, Feyerabend C, Vesey C, Saloojee Y. Com-parison of tests used to distinguish smokers from nonsmokers. Am JPublic Health 1987;77:1435–1438.

15. Lum S, Hoo AF, Stocks J. Effect of airway inflation pressure on forcedexpiratory maneuvers from raised lung volume in infants. Pediatr Pul-monol 2002;33:130–134.

16. American Thoracic Society. Standardization of spirometry 1994 Update.Am J Respir Crit Care Med 1995;152:1107–1136.

17. Jones MH, Howard J, Davis S, Kisling J, Tepper RS. Sensitivity ofspirometric measurements to detect airway obstruction in infants. AmJ Respir Crit Care Med 2003;167:1283–1286.

18. Beardsmore CS, Bar-Yishay E, Maayan C, Yahav Y, Katznelson D,Godfrey S. Lung function in infants with cystic fibrosis. Thorax 1988;43:545–551.

19. Tepper RS, Montgomery GL, Ackerman V, Eigen H. Longitudinal evalu-ation of pulmonary function in infants and very young children withcystic fibrosis. Pediatr Pulmonol 1993;16:96–100.

20. Armstrong DS, Grimwood K, Carlin JB, Carzino R, Gutierrez JP, HullJ, Olinsky A, Phelan EM, Robertson CF, Phelan PD. Lower airwayinflammation in infants and young children with cystic fibrosis. Am JRespir Crit Care Med 1997;156:1197–1204.

21. Muhlebach MS, Stewart PW, Leigh MW, Noah TL. Quantitation ofinflammatory responses to bacteria in young cystic fibrosis and controlpatients. Am J Respir Crit Care Med 1999;160:186–191.

22. Dakin CJ, Numa AH, Wang H, Morton JR, Vertzyas CC, Henry RL.Inflammation, infection, and pulmonary function in infants and youngchildren with cystic fibrosis. Am J Respir Crit Care Med 2002;165:904–910.

23. Martinez FD, Wright AL, Taussig LM, Holberg CJ, Halonen M, MorganWJ, Asthma and wheezing in the first six years of life. N Engl J Med1995;332:133–138.

24. Young S, Arnott J, O’Keeffe PT, Le Souef PN, Landau LI. The associa-tion between early life lung function and wheezing during the first 2years of life. Eur Respir J 2000;15:151–157.

25. Turner SW, Palmer LJ, Rye PJ, Gibson NA, Judge PK, Young S, LandauLI, Le Souef PN. Infants with flow limitation at 4 weeks: outcome at6 and 11 years. Am J Respir Crit Care Med 2002;165:1294–1298.

26. Dezateux C, Stocks J, Wade AM, Dundas I, Fletcher ME. Airway func-tion at one year: association with premorbid airway function, wheezing,and maternal smoking. Thorax 2001;56:680–686.

27. FitzSimmons SC. The changing epidemiology of cystic fibrosis. J Pediatr1993;122:1–9.

Particulate Matter Exposure in Cars Is Associated withCardiovascular Effects in Healthy Young MenMichael Riediker, Wayne E. Cascio, Thomas R. Griggs, Margaret C. Herbst, Philip A. Bromberg, Lucas Neas,Ronald W. Williams, and Robert B. Devlin

Center for Environmental Medicine, Asthma and Lung Biology, School of Medicine; Division of Cardiology, School of Medicine,University of North Carolina at Chapel Hill; North Carolina State Highway Patrol, Raleigh; U.S. Environmental Protection Agency,Office of Research and Development, National Health and Environmental Effects Research Laboratories, Research Triangle Park;and U.S. Environmental Protection Agency, Office of Research and Development, National Exposure Research Laboratories,Research Triangle Park, North Carolina

Exposure to fine airborne particulate matter (PM2.5) is associatedwith cardiovascular events and mortality in older and cardiac pa-tients. Potential physiologic effects of in-vehicle, roadside, and am-bient PM2.5 were investigated in young, healthy, nonsmoking, maleNorth Carolina Highway Patrol troopers. Nine troopers (age 23 to30) were monitored on 4 successive days while working a 3 P.M.

to midnight shift. Each patrol car was equipped with air-qualitymonitors. Blood was drawn 14 hours after each shift, and ambula-tory monitors recorded the electrocardiogram throughout the shiftand until the next morning. Data were analyzed using mixed mod-els. In-vehicle PM2.5 (average of 24 �g/m3) was associated withdecreased lymphocytes (�11% per 10 �g/m3) and increased redblood cell indices (1% mean corpuscular volume), neutrophils (6%),C-reactive protein (32%), von Willebrand factor (12%), next-morn-ing heart beat cycle length (6%), next-morning heart rate variabilityparameters, and ectopic beats throughout the recording (20%).Controlling for potential confounders had little impact on the effectestimates. The associations of these health endpoints with ambientand roadside PM2.5 were smaller and less significant. The observa-tions in these healthy young men suggest that in-vehicle exposureto PM2.5 may cause pathophysiologic changes that involve inflam-mation, coagulation, and cardiac rhythm.

Keywords: ambulatory electrocardiography; complete blood cell count;vehicle emissions

During the past decade, many epidemiologic studies have reportedstatistically significant positive correlations between daily meanconcentrations of air pollution particulate matter (PM) with anaerodynamic diameter equal to or less than 2.5 �m (PM2.5) andincreased mortality and morbidity attributable to respiratory andcardiovascular causes, as reviewed in the U.S. EnvironmentalProtection Agency Air Quality Criteria for Particulate Matterdocument (1). Exposure to PM2.5 was positively associated withacute myocardial infarction (2) and with daily deaths for citiesthroughout the United States, with a linear dose–response rela-tionship extending to very low PM2.5 concentrations (3). Because

(Received in original form October 28, 2003; accepted in final form February 10, 2004)

Supported by Environmental Protection Agency cooperative agreements CR-824915 and CR-829522 to the University of North Carolina at Chapel Hill, con-tract 68-D-00-206 to ManTech Environment, and the Swiss National ScienceFoundation.

Correspondence and requests for reprints should be addressed to Michael Riediker,Institut de Sante au Travail (Institute of Occupational Health Sciences), Rue duBugnon 19, CH-1005 Lausanne, Switzerland. E-mail: michael.riediker@alumni.ethz.ch

This article has an online supplement, which is accessible from this issue’s tableof contents online at www.atsjournals.org.

Am J Respir Crit Care Med Vol 169. pp 934–940, 2004Originally Published in Press as DOI: 10.1164/rccm.200310-1463OC on February 12, 2004Internet address: www.atsjournals.org

of the large fraction of the population potentially exposed to PM,it has been estimated that 800,000 excess deaths worldwide eachyear may be attributable to PM (4). Older individuals with under-lying cardiac or pulmonary disease appear to be at greatest risk(5, 6). Inhalation of PM might result in cardiopulmonary events,which then rapidly trigger increased mortality or morbidity ina small fraction of the population. Two recent studies reportassociations between elevated PM levels and serious ventriculararrhythmias (7) as well as myocardial infarctions (8). However,the underlying pathophysiologic mechanisms that link PM andcardiopulmonary mortality are still poorly understood.

Heart rate variability (HRV) is influenced by autonomic con-trol mechanisms and helps to identify patients with an increasedrisk of cardiac mortality (9). Therefore, several panel studies ana-lyzed ambulatory electrocardiogram recordings for PM-associ-ated changes in HRV. In samples of older individuals living inBaltimore (10) and Boston (11) and in cardiac patients (12), in-creased PM2.5 concentrations were associated with small de-creases in HRV. These changes were mostly seen in subjects withpreexisting cardiovascular conditions.

PM might also induce lung inflammation, which could causecardiovascular stress. Low levels of lower respiratory tract in-flammation were observed in humans exposed to ambient PM(13). Serum levels of C-reactive protein were increased in rela-tion to ambient PM exposure (8), and acute-phase reactantsinvolved in thrombotic processes, such as fibrinogen, were in-creased in association with exposure to PM (14).

We investigated the potential health effects of in-vehicle PM2.5

exposure in young, healthy North Carolina State Highway Patroltroopers. Members of this police force spend the majority oftheir working time in or near their cars and often in heavy traffic.Previous epidemiologic studies had indicated that exposure toparticles of vehicular origin might be especially potent with re-gard to increased mortality (15, 16), as well as increased acutecardiovascular morbidity (17). Our hypothesis was that exposureof these healthy young male individuals to PM2.5 from trafficwould lead to detectable changes in the autonomic regulationof the heart, as measured by HRV, and that PM2.5 would induceinflammatory responses large enough to be observable in theperipheral blood. Detailed in-vehicle exposure data for PM2.5,its components, gaseous co-pollutants and several organic com-pounds were collected (18). This article describes the effects ofPM2.5. Potentially confounding co-pollutants and activity param-eters were included in the analysis. Some of the results of thisstudy have been previously reported in the form of abstracts(19, 20).

METHODS

Ten nonsmoking, male North Carolina State Highway Patrol troopersvolunteered to participate in this study in the autumn of 2001 whileworking a 3 p.m. to midnight shift (but one had to be excluded becauseof the high frequency of supraventricular ectopic beats and high serumcholesterol levels). The institutional review board of the University of

Riediker, Cascio, Griggs, et al.: Cardiovascular Effects of PM2.5 935

North Carolina School of Medicine approved the study. Written consentwas obtained from all troopers before their participation in the study.The troopers refrained from alcohol, caffeine, and any medication 24hours before the start and until the end of their participation.

Two troopers were monitored in parallel each week on 4 consecutiveworkdays from Monday to Thursday. Venous blood was drawn beforethe first shift and 10 to 14 hours after each shift. Blood samples wereprocessed immediately for storage and for transport to the laboratory.Parameters analyzed included standard serum chemistry, differentialblood cell counts, and inflammation and coagulation markers. Thetroopers wore ambulatory electrocardiogram monitors throughout theshift and until the next morning. HRV was assessed during controlledresting periods before the shift, after the shift (before going to bed),and the next morning (immediately after awakening). The subjects satin a quiet room for 20 minutes, and data from the final 10 minuteswere used for the HRV analysis. Time–domain parameters includedthe mean cycle length of R-R intervals for normal beats, the standarddeviation of all normal R-R intervals (SDNN) and the percentage ofsuccessive normal R-R interval differences greater than 50 ms (PNN50).The frequency spectrum was divided into low-frequency (LF) power(0.04 to 0.15 Hz) and high frequency (HF) power (0.15 to 0.40 Hz).

The patrol cars were powered by a gasoline engine and wereequipped with air conditioning units. They had a fresh air intake withoutfiltration. Each patrol car was equipped with portable air-quality moni-tors to measure PM2.5 and other air pollutants during the shift as describedearlier (18). PM2.5 was measured by two different methods: particlecollection on filters with gravimetric analysis (PM2.5Mass) and real-timeparticle mass estimation from light scattering (PM2.5Lightscatter, most sensi-tive to particles in the size range of 0.2 to 2.0 �m). These measurementsserved as an estimate for the troopers’ exposure during their work. Theroute of the car was monitored with a global-positioning system trackingdevice, and the troopers’ activities were obtained from official logs anda diary. Data were also collected simultaneously at a site reflectingurban ambient concentrations (PM2.5Ambient) and from multiple stationaryroadside locations near busy roads in Wake County (PM2.5Roadside). Forthese outdoor sites, only gravimetric PM2.5 data are available.

Statistics were calculated using S-Plus 2000 for Windows (MathsoftInc., Cambridge, MA). Mixed-effects regression models with restrictedmaximum likelihood estimation were used to investigate the effects ofPM2.5 on HRV and blood parameters (21). The models included arandom intercept for each subject and fixed effects for the exposurevariables. All parameters as well as the residuals were tested for timetrends across the week. The effects of potential confounders were inves-tigated by including them as fixed effects in the models.

Additional details on the methods used to calculate HRV, the pro-cessing of the blood samples, and the data treatment procedures areprovided in an online supplement.

RESULTS

Subjects

Data from a total of nine individual male nonsmoking trooperswere used for the analysis, each with 4 consecutive days (10troopers were monitored, but 1 had to be excluded because ofthe high frequency of supraventricular ectopic beats and highserum cholesterol levels). The age of the nine subjects (eightwhite and one black) ranged from 23 to 30 years (average 27.3years), the average weight from 74 to 102 kg (average 87 kg �

192 pounds), the height from 168 to 191 cm (average 179 cm �

70.5 inches), and the body mass index from 24 to 31 kg/m2

(average 27 kg/m2). They were all in excellent physical condition,and all exercised several times per week.

Exposure

Pollutant levels in the patrol cars of the analyzed troopers werehighly variable but were always well below occupational thresh-

TABLE 1. IN-VEHICLE POLLUTANT CONCENTRATIONSDURING THE ANALYZED SHIFTS: MEAN OF ALL SHIFTS,RANGE OF THE SHIFT AVERAGES, AND CORRELATIONSOF THE POLLUTANTS TO FINE PARTICULATE MATTER (PM2.5)

Correlation CoefficientsConcentrations (Spearman rho)

Mean Range PM2.5Lightscatter PM2.5Mass

PM2.5Lightscatter, �g/m3 24.1 4.5–54.4 1.00 0.71†

PM2.5Mass, �g/m3 23.0 7.1–38.7 0.71† 1.00PM2.5Ambient, �g/m3 32.3 9.9–68.9 0.67† 0.63†

PM2.5Roadside, �g/m3 32.1 8.9–62.2 0.65† 0.58†

Ozone, ppb 12.4 �4.7*–57.6 �0.21 0.23CO, ppm 2.6 0.9–5.9 0.53† 0.55†

NO2, ppb 35 1.6–213 �0.30 0.06Relative humidity, % 36 25–46 0.40† 0.33†

Temperature, �C 25.6 20.3–29.1 0.17 0.15

Definition of abbreviations: ppb � parts per billion; PM2.5 � particulate matterwith an aerodynamic diameter equal to or less than 2.5 �m.

* After blank correction.† Significant correlation with p � 0.05.

old limits values (Table 1). Individual volatile organic compoundlevels (hydrocarbons and aldehydes) inside the investigated carswere in the parts per billion range. In-vehicle PM2.5 was 24% lowerthan ambient and roadside concentrations, whereas in-vehicle CO,NO2, aldehydes, hydrocarbons, and some metals were elevated.On average, troopers spent 35% of their shift away from thecar, mostly inside buildings (office, jail, hospitals, or for dinner).None of the investigated exposure parameters showed a timetrend for increasing or decreasing concentrations throughout theweekdays. A detailed description of pollutants measured insidethe vehicles, at the roadside, and at the fixed ambient site ispublished elsewhere (18).

Changes in Cardiac Parameters

Most HRV parameters increased significantly from the start ofthe shift in the afternoon until midnight, and SDNN (in the timedomain) and total frequency power (in the frequency domain)were further increased in the morning after awakening (see Ta-ble E1 in the online supplement). None of the investigated car-diac parameters showed a time–trend throughout the weekdays.

The associations between the PM2.5 concentrations inside thecars (averages of the 9-hour shifts) and the HRV parameters(at midnight and in the morning after the shift) are shown inTable 2. PM2.5Lightscatter during the shift was significantly positivelyassociated with all time domain parameters (PNN50, SDNN,and mean cycle length) as well as with HF power and the powerratio LF/HF on the morning after the shift. The effect estimates(slopes) for PM2.5Mass were similar; however, these associationshad a larger uncertainty, and only the effect estimate for meancycle length reached significance (Figure E1 in the online supple-ment shows exposure–response plots for a selection of the sig-nificant associations of health parameters to the PM2.5 measures).

The number of ectopic beats throughout the work shift andduring the contiguous night was low. Most supraventricular ec-topic beats occurred during the sleep, whereas the ventricularectopic beats were most frequent in the late evening and aroundwake-up. Both ventricular and supraventricular ectopic beatswere strongly increased in association with PM2.5Lightscatter but notwith PM2.5Mass (Table 2).

936 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

TABLE 2. HEART RATE VARIABILITY AND ECTOPY PARAMETERS AND THEIR ASSOCIATION WITH THE IN-VEHICLE FINEPARTICULATE MATTER (PM2.5)

Effect per 10 �g/m3 PM2.5Lightscatter Effect per 10 �g/m3 PM2.5Mass

Percentage PercentageParameter Mean SD Slope SE Change p Value Slope SE Change p Value

After shift resting periodMCL, ms 950 130 17.1 13.6 1.8 0.220 �1.51 23.0 �0.2 0.948SDNN, ms 86 30 0.05 4.10 0.1 0.990 �5.10 6.82 �5.9 0.461PNN50, % 41 19 5.04 2.46 12.4 0.051 �0.89 4.28 �2.2 0.838LF 8.5 5.2 0.32 0.70 3.8 0.652 0.64 1.16 7.5 0.588HF 5.1 3.4 0.54 0.45 10.6 0.244 �0.03 0.77 �0.6 0.968Total power 48 78 �15.6 9.5 �32.7 0.113 �19.4 17.0 �40.6 0.262Ratio LF/HF 1.9 1.1 �0.17 0.15 �8.9 0.252 0.20 0.24 10.4 0.411

Next morning resting periodMCL, ms 1001 140 41.9 12.7 4.2 0.003 63.4 20.9 6.3 0.005SDNN, ms 129 42 15.2 5.02 11.7 0.006 10.3 9.45 7.9 0.288PNN50, % 43 20 5.28 1.87 12.2 0.009 5.87 3.22 13.6 0.080LF 12 6.8 �0.60 0.90 �5.1 0.510 �0.25 1.49 �2.1 0.871HF 5.7 3.4 0.84 0.34 14.8 0.019 0.57 0.59 10.0 0.342Total power 86 62 13.4 8.20 15.7 0.113 20.7 13.7 24.1 0.145Ratio LF/HF 2.8 1.8 �0.59 0.20 �21.5 0.007 �0.51 0.36 �18.6 0.169

Ectopic beatsVentricular, per hr 14.5 40.5 2.76 1.31 19.1 0.045 3.00 2.22 20.7 0.189Supraventricular, per hr 4.48 4.52 1.03 0.39 23.0 0.014 0.42 0.70 9.4 0.552

Definition of abbreviations: HF � high frequency; LF � low frequency; MCL � mean cycle length; PM � particulate matter; PNN50 � percentage of successive normalR-R interval differences greater than 50 ms; SDNN � SD of all normal R-R intervals.

Effect estimates are given in absolute values (slopes with SE) and as percent change relative to the mean. p Values � 0.05 are significant.

Changes in Vascular Components

Standard serum chemistries were within the normal range forall subjects studied. Only serum osmolality was slightly elevatedwith an average of 296 mOsm/kg (SD of 3.1 mOsm/kg). None of theinvestigated vascular components showed a time–trend through-out the weekdays. PM2.5 concentrations inside the cars were sig-nificantly associated with changes in parameters of the blood thatwas collected approximately 14 hours after the shift (Table 3). Incontrast with the cardiac measurements, vascular componentswere more often significantly associated with PM2.5Mass than withPM2.5Lightscatter.

Small but significant (p � 0.0002) changes were seen for redblood cell volume (MCV) in association with PM2.5Lightscatter. This

TABLE 3. BLOOD PARAMETERS (MEAN AND SD) AND THEIR ASSOCIATION WITH THE IN-VEHICLE FINE PARTICULATE MATTER(PM2.5) CONCENTRATIONS

Effect per 10 �g/m3 PM2.5Lightscatter Effect per 10 �g/m3 PM2.5Mass

Blood Parameter Mean SD Slope SE % Change p Value Slope SE % Change p Value

Uric acid, serum, mg/dl 6.2 1.1 0.07 0.12 1.2 0.555 0.38 0.19 6.1 0.055Hematocrit, % 45 3.0 0.70 0.27 1.5 0.014 0.29 0.48 0.6 0.554MCV, fL 88 2.9 0.85 0.19 1.0 0.0002 0.78 0.37 0.9 0.045MCH, pg 30 0.9 0.10 0.08 0.3 0.242 0.28 0.13 0.9 0.033MCHC, g/dl 34 0.7 �0.19 0.08 �0.6 0.032 �0.22 0.14 �0.7 0.129Neutrophils, % 58 8.5 1.85 1.00 3.2 0.077 3.56 1.61 6.2 0.036Lymphocytes, % 31 7.4 �1.71 0.87 �5.5 0.059 �3.29 1.39 �10.5 0.025Lymphocyte count, 103/�l 1.6 0.4 �0.05 0.04 �3.0 0.281 �0.13 0.07 �8.4 0.063C-reactive protein, mg/L 1.0 1.2 0.21 0.08 21.1 0.013 0.32 0.13 31.9 0.023Plasminogen, IU/ml 1.7 1.7 0.27 0.22 15.9 0.221 0.67 0.36 39.6 0.073von Willebrand factor, % 153 37 5.50 4.60 3.6 0.242 18.0 7.13 11.8 0.018

Definition of abbreviations: MCH � mean red blood cell volume; MCHC � mean red blood cell hemoglobin; MCV � mean red blood cell hemoglobin concentration;PM � particulate matter.

Only parameters with p � 0.1 are listed (p � 0.05 are significant). Blood was sampled 14 hours after the shift ended. Effect estimates are in absolute values (slopeswith SE) and as a percentage change relative to the mean.

was true for each individual trooper (shown in Figure E1D inthe online supplement) as well as for the group as a whole. MCVwas also significantly associated with PM2.5Mass. The potentialeffect of osmolality on MCV and MCHC was assessed by stand-ardizing them to the average osmolality (MCV multiplied bythe osmolality of the sample and divided by the average osmolal-ity; the inverse for MCHC). This standardization did not changethe effect estimates listed in Table 3 but did increase the statisti-cal significance.

Associations with Ambient and Roadside PM2.5 Concentrations

Figure 1 summarizes the associations between time–domainHRV parameters, ectopic beats, and selected vascular compo-

Riediker, Cascio, Griggs, et al.: Cardiovascular Effects of PM2.5 937

Figure 1. Selected heart rhythm (A ) and blood (B ) parameters. Comparison of fine particulate matter (PM2.5) effect estimates for the two in-vehiclemethods (PM2.5Lightscatter, closed circles; PM2.5Mass, open circles) and for the gravimetric data from the ambient site (open squares) and the roadsidelocations (open triangles). Lines indicate the 95% confidence intervals of the effect estimates. The effect estimates for MCV were multiplied by 10to better fit the scale.

nents; and PM measured in-vehicle, at the roadside, and fromfixed-site community monitors. The most consistent effect esti-mates across all PM2.5 measurements were seen with ventricularectopic beats. Effect estimates for roadside PM2.5 and severaleffect estimates for ambient PM2.5 tended to be smaller than thecomparable in-vehicle effect estimates. Roadside mass resultedfor most parameters in the smallest effect estimates. For eachparameter, in-vehicle PM2.5Mass had the widest confidence inter-vals compared with all other PM2.5 measures, whereas PM2.5Roadside

had wider confidence intervals than PM2.5Ambient.

Potential Confounders

The potentially confounding effects of CO, NO2, and relativehumidity were investigated by including them in bivariate modelswith PM2.5. These co-pollutants were to some extent correlatedto PM2.5 (Table 1), which makes them potential confounders ofPM2.5 effects. The procedure led to minimal changes in mosteffect estimates for the HRV parameters (see Figure E2 in theonline supplement). However, for ventricular ectopic beats, in-cluding NO2 and relative humidity broadened the 95% confi-dence intervals and made the effect estimate for PM2.5Lightscatter

nonsignificant. The same relationship was observed with CO andsupraventricular ectopic beats. In both cases, the effect estimates

(slopes) themselves were little changed. None of the effect esti-mates for NO2, CO, or relative humidity were significant in anyof these models.

Controlling the blood parameters for the same confoundersdid not reduce the effect estimates for the PM2.5 associations withMCV, von Willebrand factor, and C-reactive protein. However,for CO and leukocyte percentages (percentage of neutrophils andpercentage of lymphocytes), the bivariate models showed a cleardrop in the PM2.5 effect estimates (especially for PM2.5Lightscatter),and the 95% confidence intervals were broadened to includezero. None of the effect estimates for NO2, CO, or relative hu-midity were significant in any of these models.

The potential influence of stress was controlled by includingthe number of accidents during the shift, the number of citationsissued, “other” activities, and the total number of law-enforce-ment activities. Controlling the HRV parameters for these fac-tors had virtually no effect on either the effect estimates or theconfidence intervals. However, significant effect estimates forthe association between PNN50 and PM2.5Lightscatter resulted aftercontrolling for the number of accidents (slope of 5.68 � 2.56,p � 0.04), “other” activities (6.20 � 2.24, p � 0.01), and totalnumber of law enforcement activities (5.54 � 2.56, p � 0.04).

938 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

DISCUSSION

This study investigated the effects of relatively low levels ofPM2.5 exposure inside vehicles in a selected group of young,healthy, nonsmoking, fit, male highway patrol troopers. Thetroopers showed significant and strong increases of HRV, ectopicbeats, blood inflammatory and coagulation markers, and MCVin association with the in-vehicle exposure to PM2.5. These novelfindings could be helpful in better understanding the pathophysi-ologic processes subsequent to PM2.5 exposure that can lead insusceptible populations to increases in cardiovascular morbidityand mortality.

Normal values for all baseline blood work attest to the excel-lent general health of the troopers. The fitness of the troopersis evidenced by their average body mass index of 27 and their lowresting pulse (approximately 60 beats per minute). The slightlyelevated serum osmolality (average of 296 mOsm/kg) possiblyreflects slight dehydration secondary to inadequate fluid replace-ment during the work shift. Overall, these troopers represent aclinical subgroup that appears to be at very low risk for cardiacor other adverse health conditions.

PM2.5 was relatively low inside the cars compared with the24-hour National Ambient Air Quality Standard for PM2.5 of65 �g/m3 and mostly in the “green” and “yellow” range of theair quality index (i.e., below 40 �g/m3 PM2.5). PM2.5 was measuredgravimetrically and also estimated with a light-scattering device.The two methods resulted in almost identical averages and werehighly correlated (Table 1). However, the PM2.5 levels estimatedby light scattering had a larger variability, as reflected by theinterquartile range. This is consistent with earlier studies com-paring these two devices (22). Gravimetric particle measuresare strongly influenced by large particles, whereas the light-scattering device estimates are more influenced by particle num-bers. Inside the patrol cars, the two methods appear to reflectthese differences: PM2.5Mass represents exposure to some coarsemode material from suspended road dust as well as to ambientaccumulation mode particles, whereas PM2.5Lightscatter representsambient particles and fresh combustion particles, which occurin high numbers in the smaller accumulation size range.

Inflammatory markers in the peripheral blood collected 10to 14 hours after the work shift were associated with the PM2.5Mass

concentration in the cars. von Willebrand factor, a marker forendothelial activation and thrombosis (23), the inflammatorymarker C-reactive protein, and the percentage of neutrophilswere increased, whereas the percentage of lymphocytes was de-creased. These data suggest that PM exposure causes slight vas-cular inflammation. It is possible that this inflammation wasrestricted to the peripheral lung, where a large proportion offine particles is deposited (24, 25). Our findings suggesting PM2.5

exposure to cause inflammatory processes is consistent withother studies, which reported elevated C-reactive protein levels(8) and inflammatory cell numbers (13) as a consequence ofPM2.5 exposure.

The inflammatory blood markers for which we observedchanges have been previously linked to increased morbidity. Arecent prospective study showed increased von Willebrand factorto be associated with an increased risk of coronary heart diseaseand nonfatal myocardial infarction (26). Increased C-reactiveprotein levels were also associated to cardiovascular diseases (27)and coronary calcification (28). It was reported that C-reactiveprotein was an even better long-term predictor for cardiovascu-lar events than cholesterol (29), although the predictive valueof increased levels of C-reactive protein for these diseases is stilldisputed (30). The low concentration of C-reactive protein inthe serum of the troopers (average of approximately 1 mg/L)suggests a relatively low cardiovascular risk (29). Nevertheless,

the consequences of several years of daily exposure to in-vehiclepollutants for the development of arteriosclerosis, particularlyif these pollutants increase markers of systemic inflammation,remain unclear.

The various inflammatory markers differed in the strengthand significance of their association to particle concentrationsat the various locations and in the effects of potential confound-ers on the association. These differences could reflect several atleast partly independent inflammatory processes in response tothe particles or their components: The association between theinflammatory markers and PM2.5Mass was much tighter comparedwith PM2.5Lightscatter. Consequently, the larger accumulation sizeparticles might be more important for the inflammatory processobserved. There was no association of neutrophils, lymphocytes,and C-reactive protein to roadside and ambient particle concen-trations, and PM2.5Mass lost significance in the model with CO(whereas CO had no significant effects in this model). This canbe explained by two possibilities: (1) Other combustion products(and not solely particle mass) might be important for the leuko-cyte response as well, or (2) mostly traffic-related particles (withCO as an indicator for traffic) were causing the response. Futureanalyses of co-pollutants and particle composition might answerthe question, whether particles from traffic sources might havecaused this strong inflammatory response.

The increase of the red blood cell volume MCV associatedwith increasing PM2.5 could not be explained by decreased serumosmolality and therefore could be a consequence of particlecomponents or inflammatory mediators interfering with the vol-ume regulatory ion channels and/or pumps in the erythrocytes(31). These changes in MCV also raise a question about potentialeffects of PM2.5 exposure on ion channels and/or pumps in othercell types. An earlier study (32) found PM-associated changesin red blood cell count, hematocrit, and hemoglobin concentra-tion. No results were reported for MCV. They proposed periph-eral sequestration of red cells as an explanation. However, ourresults favor a mechanism that affects the volume of red bloodcells.

The troopers’ resting HRV after the shift (at midnight beforegoing to bed) was influenced very little by the PM2.5 exposureduring the shift. Instead, the HRV changes across the shiftseemed to reflect mostly a diurnal pattern and a normal physio-logical change from daytime to late evening. The changes arein a range similar to those observed for preshift and postshiftHRV of emergency physicians on a night shift schedule (33), insteel workers on evening shift (34), and between sleep and workin nurses (35). Only PNN50 showed some association withPM2.5Lightscatter (p � 0.051, significant after controlling for some ofthe activity parameters). This might indicate that some HRVchanges had already started at midnight, although the changeswere mostly too small to be detected within the noise from thedaily activities and the daytime variation.

The HRV parameters were highest in the morning on arising.This is in contrast to a study of HRV in nurses (35), whichreported that HRV at the time of awakening was equal to day-time levels. The mixed model analysis of the trooper’s datasuggests that this increase of HRV was mostly a consequenceof the preceding in-vehicle PM2.5 exposure and not simply adiurnal pattern. Furthermore, the increase in the HRV from theafternoon to the next morning was similar to that reported forsteel workers (34), who (like the troopers) have occupationalexposure to PM2.5 and physically demanding work.

This study shows a strong and consistent increase of HRVin association with PM2.5. The pattern of HRV-responses suggestsincreased vagal activity, that is, the involvement of the nervoussystem in the response to PM. Studies about the interactions

Riediker, Cascio, Griggs, et al.: Cardiovascular Effects of PM2.5 939

between the immune system and the brain report local inflam-matory reactions to increase the vagal sensory input to the brain,which again can result in increased vagal output (as reviewedin 36). Our results indicating mild inflammatory processes in thelungs and increased vagal output to the heart would be in agree-ment with these models. However, our HRV results are in con-trast to earlier observational studies in panels of older individuals(10, 11), in older cardiac patients (12), and in middle-aged boiler-makers (37), as well as controlled exposure experiments in oldersubjects (38), in whom a small decrease of HRV in associationwith PM2.5 exposure was observed (the boilermakers showed avery small increase in SDNN in association with concentrationsof some metallic components of the particles) (39). These oppos-ing results might be caused by differences in age, cardiovascularfitness, or prevalence of cardiovascular disease. In one study witholder subjects (10), only the group with cardiovascular conditionsshowed a decreased SDNN in response to the PM2.5 exposureof the same day. Controlled exposure to residual oil fly ashcaused decreased SDNN as an immediate reaction in rats withprior myocardial infarction, but not in healthy control rats (40).Age per se might be important as well because the subjects fromthe previously cited controlled exposure study (38) had a meanage of 67 years. Vagal function is reduced in the aged and inclinical populations such as diabetics (41–43). This can lead todifferences in the heart’s response to the vagal output becausedifferent competitive vagal pathways are involved in the auto-nomic control of the heart rate (44). The troopers were youngand physically fit and were with a well-conditioned and respon-sive cardiovascular system. Consequently, our results might re-flect a “healthy” physiologic response to PM as compared withthe older subjects, cardiac patients, and the middle-aged boiler-makers who have a reduced cardiovascular dynamism andchanged vagal function.

Premature supraventricular ectopic beats increased 23%, andventricular ectopic beats increased 19% for each 10 �g/m3 in-crease in PM2.5Lightscatter. At the same time, the markers of parasym-pathetic input to the heart suggested an increased vagal tone:the LF/HF ratio was decreased by 21%, and HF, PNN50, andSDNN were increased by over 10%. These findings implicate avagal mechanism of the premature ectopic beats. Fluctuationsin autonomic tone have been reported to be associated with thetriggering of atrial arrhythmias (45) and may thus offer insightinto the mechanism of arrhythmias associated with air pollution.The observation that the relationship was strongest with thePM2.5Lightscatter indicates that fine particles are more likely to medi-ate the cardiac responses.

In cardiac patients, increased HRV and heart rate were ob-served in the period before the onset of postoperative atrialfibrillation, which suggested vagal resurgence competing withsympathetic activity to be the primary mechanism responsiblefor the triggering of atrial fibrillation (46). The consequences ofan increased vagal tone in healthy young people are not welldefined. Apparently, the increase in (benign) ectopic beats mightbe related to the increase in vagal tone. It is unclear whetherthis poses any risk to the troopers, although it does represent aPM2.5-induced alteration of normal homeostatic control.

Conclusions

In-vehicle PM2.5 exposure of healthy, young, nonsmoking, maleNorth Carolina State Highway Patrol troopers was associatedwith significant increases in vagal activity, as reflected by thechanges in HRV, with increased number of ectopic beats andwith increases in MCV and in peripheral blood inflammatoryand coagulation markers. These effects were observed a fewhours after the exposure and suggest two general pathophysio-

logic pathways subsequent to PM2.5 deposition in the peripherallungs: (1) via proinflammatory and prothrombotic blood changesand (2) via changes of cardiac autonomic control mechanisms.The changes observed do not seem desirable and suggest thatexposure to in-vehicle PM2.5 should be minimized.

Conflict of Interest Statement : M.R. has no declared conflict of interest; W.E.C.has no declared conflict of interest; T.R.G. has no declared conflict of interest;M.C.H. has no declared conflict of interest; P.A.B. has no declared conflict ofinterest; L.N. has no declared conflict of interest; R.W.W. has no declared conflictof interest; R.B.D. has no declared conflict of interest.

References1. US-EPA. Air quality criteria for particulate matter EPA/600/P-95/001cf.

Research Triangle Park, NC: National Center for Environmental As-sessment, Office of Research and Development, U.S. EnvironmentalProtection Agency; 1996.

2. Peters A, Dockery DW, Muller JE, Mittleman MA. Increased particulateair pollution and the triggering of myocardial infarction. Circulation2001;103:2810–2815.

3. Schwartz J, Laden F, Zanobetti A. The concentration-response relationbetween PM(2.5) and daily deaths. Environ Health Perspect 2002;110:1025–1029.

4. World Health Organization. The world health report: reducing risks, pro-moting healthy life [accessed July 17, 2003]. Available from: http://www.who.int/whr/en/. Geneva, Switzerland: World Health Organiza-tion; 2002.

5. Schwartz J. Air pollution and hospital admissions for heart disease ineight US counties. Epidemiology 1999;10:17–22.

6. De Leon SF, Thurston GD, Ito K. Contribution of respiratory diseaseto nonrespiratory mortality associations with air pollution. Am J RespirCrit Care Med 2003;167:1117–1123.

7. Peters A, Liu E, Verrier RL, Schwartz J, Gold DR, Mittleman M, BaliffJ, Oh JA, Allen G, Monahan K, et al. Air pollution and incidence ofcardiac arrhythmia. Epidemiology 2000;11:11–17.

8. Peters A, Frohlich M, Doring A, Immervoll T, Wichmann HE, Hutchin-son WL, Pepys MB, Koenig W. Particulate air pollution is associatedwith an acute phase response in men: results from the MONICA-Augsburg Study. Eur Heart J 2001;22:1198–1204.

9. Lombardi F. Clinical implications of present physiological understandingof HRV components. Card Electrophysiol Rev 2002;6:245–249.

10. Liao D, Creason J, Shy C, Williams R, Watts R, Zweidinger R. Dailyvariation of particulate air pollution and poor cardiac autonomic con-trol in the elderly. Environ Health Perspect 1999;107:521–525.

11. Gold D, Litonjua A, Schwartz J, Lovett E, Larson A, Nearing B, AllenG, Verrier M, Cherry R, Verrier R. Ambient pollution and heart ratevariability. Circulation 2000;101:1267–1273.

12. Pope CA III, Verrier RL, Lovett EG, Larson AC, Raizenne ME, KannerRE, Schwartz J, Villegas GM, Gold DR, Dockery DW. Heart ratevariability associated with particulate air pollution. Am Heart J 1999;138:890–899.

13. Ghio AJ, Kim C, Devlin RB. Concentrated ambient air particles inducemild pulmonary inflammation in healthy human volunteers. Am JRespir Crit Care Med 2000;162:981–988.

14. Pekkanen J, Brunner EJ, Anderson HR, Tiittanen P, Atkinson RW.Daily concentrations of air pollution and plasma fibrinogen in London.Occup Environ Med 2000;57:818–822.

15. Bigert C, Gustavsson P, Hallqvist J, Hogstedt C, Lewne M, Plato N,Reuterwall C, Scheele P. Myocardial infarction among professionaldrivers. Epidemiology 2003;14:333–339.

16. Laden F, Neas LM, Dockery DW, Schwartz J. Association of fine particu-late matter from different sources with daily mortality in six US cities.Environ Health Perspect 2000;108:941–947.

17. Janssen NA, Schwartz J, Zanobetti A, Suh HH. Air conditioning andsource-specific particles as modifiers of the effect of PM(10) on hospitaladmissions for heart and lung disease. Environ Health Perspect 2002;110:43–49.

18. Riediker M, Williams R, Devlin R, Griggs T, Bromberg P. Exposure toparticulate matter, volatile organic compounds and other air pollutantsinside patrol cars. Environ Sci Technol 2003;37:2084–2093.

19. Riediker M, Devlin R, Griggs T, Cascio W, Herbst M, Williams R,McCorquodale S, Bromberg P. Changes in health parameters observedin NC patrol troopers exposed to PM and air toxics (COPP-study).Epidemiology 2002;13:108.

20. Riediker M, Bromberg PA, Cascio WA, Griggs T, Herbst M, WilliamsRW, Neas LM, Devlin RB. PM2.5 exposure changes heart rate vari-

940 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

ability (HRV) and blood parameters in state highway patrol troopers[abstract]. Am J Respir Crit Care Med 2003;167:A332.

21. Pinheiro JC, Bates DM. Mixed-effects models in S and S-PLUS. Heidel-berg, Germany: Springer Verlag; 2000.

22. Rea AW, Zufall MJ, Williams RW, Sheldon L, Howard-Reed C. Theinfluence of human activity patterns on personal PM exposure: a com-parative analysis of filter-based and continuous particle measurements.J Air Waste Manage Assoc 2001;51:1271–1279.

23. Conway DS, Pearce LA, Chin BS, Hart RG, Lip GY. Plasma von Wille-brand factor and soluble p-selectin as indices of endothelial damageand platelet activation in 1,321 patients with nonvalvular atrial fibrilla-tion: relationship to stroke risk factors. Circulation 2002;106:1962–1967.

24. Kim CS, Hu SC, DeWitt P, Gerrity TR. Assessment of regional depositionof inhaled particles in human lungs by serial bolus delivery method.J Appl Physiol 1996;81:2203–2213.

25. Kim CS, Jaques PA. Respiratory dose of inhaled ultrafine particles inhealthy adults. Phil Tr R Soc S-A 2000;358:2693–2705.

26. Whincup PH, Danesh J, Walker M, Lennon L, Thomson A, Appleby P,Rumley A, Lowe GD. von Willebrand factor and coronary heartdisease: prospective study and meta-analysis. Eur Heart J 2002;23:1764–1770.

27. Ridker PM. High-sensitivity C-reactive protein: potential adjunct forglobal risk assessment in the primary prevention of cardiovasculardisease. Circulation 2001;103:1813–1818.

28. Wang TJ, Larson MG, Levy D, Benjamin EJ, Kupka MJ, Manning WJ,Clouse ME, D’Agostino RB, Wilson PW, O’Donnell CJ. C-reactiveprotein is associated with subclinical epicardial coronary calcificationin men and women: the Framingham Heart Study. Circulation 2002;106:1189–1191.

29. Ridker PM, Rifai N, Rose L, Buring JE, Cook NR. Comparison ofC-reactive protein and low-density lipoprotein cholesterol levels inthe prediction of first cardiovascular events. N Engl J Med 2002;347:1557–1565.

30. Kushner I, Sehgal AR. Is high-sensitivity C-reactive protein an effectivescreening test for cardiovascular risk? Arch Intern Med 2002;162:867–869.

31. Kowluru R, Bitensky MW, Kowluru A, Dembo M, Keaton PA, BuicanT. Reversible sodium pump defect and swelling in the diabetic raterythrocyte: effects on filterability and implications for microangiopa-thy. Proc Natl Acad Sci USA 1989;86:3327–3331.

32. Seaton A, Soutar A, Crawford V, Elton R, McNerlan S, Cherrie J, WattM, Agius R, Stout R. Particulate air pollution and the blood. Thorax1999;54:1027–1032.

33. Adams SL, Roxe DM, Weiss J, Zhang F, Rosenthal JE. Ambulatoryblood pressure and Holter monitoring of emergency physicians before,during, and after a night shift. Acad Emerg Med 1998;5:71–77.

34. Furlan R, Barbic F, Piazza S, Tinelli M, Seghizzi P, Malliani A. Modifica-tions of cardiac autonomic profile associated with a shift schedule ofwork. Circulation 2000;102:1912–1916.

35. Ito H, Nozaki M, Maruyama T, Kaji Y, Tsuda Y. Shift work modifiesthe circadian patterns of heart rate variability in nurses. Int J Cardiol2001;79:231–236.

36. Maier SF, Watkins LR. Cytokines for psychologists: implications of bi-directional immune-to-brain communication for understanding behav-ior, mood and cognition. Psychol Rev 1998;105:83–107.

37. Magari SR, Hauser R, Schwartz J, Williams PL, Smith TJ, ChristianiDC. Association of heart rate variability with occupational and envi-ronmental exposure to particulate air pollution. Circulation 2001;104:986–991.

38. Devlin R, Ghio A, Kehrl H, Sanders G, Cascio W. Elderly humansexposed to concentrated air pollution particles have decreased heartrate variability. Eur Respir J Suppl 2003;40:s76–s80.

39. Magari SR, Schwartz J, Williams PL, Hauser R, Smith TJ, ChristianiDC. The association of particulate air metal concentrations with heartrate variability. Environ Health Perspect 2002;110:875–880.

40. Wellenius GA, Saldiva PH, Batalha JR, Krishna Murthy GG, CoullBA, Verrier RL, Godleski JJ. Electrocardiographic changes duringexposure to residual oil fly ash (ROFA) particles in a rat model ofmyocardial infarction. Toxicol Sci 2002;66:327–335.

41. Wieling W, van Brederode JF, de Rijk LG, Borst C, Dunning AJ. Reflexcontrol of heart rate in normal subjects in relation to age: a data basefor cardiac vagal neuropathy. Diabetologia 1982;22:163–166.

42. Weise F, Heydenreich F. Age-related changes of heart rate power spectrain a diabetic man during orthostasis. Diabetes Res Clin Pract 1991;11:23–32.

43. De Meersman RE. Aging as a modulator of respiratory sinus arrhythmia.J Gerontol 1993;48:B74–B78.

44. Porges SW. Orienting in a defensive world: mammalian modificationsof our evolutionary heritage: a polyvagal theory. Psychophysiology1995;32:301–318.

45. Zimmermann M, Kalusche D. Fluctuation in autonomic tone is a majordeterminant of sustained atrial arrhythmias in patients with focal ec-topy originating from the pulmonary veins. J Cardiovasc Electrophysiol2001;12:285–291.

46. Amar D, Zhang H, Miodownik S, Kadish AH. Competing autonomicmechanisms precede the onset of postoperative atrial fibrillation. J AmColl Cardiol 2003;42:1262–1268.

Determinants of Maximally Attained Level ofPulmonary FunctionXiaobin Wang, Tjeert T. Mensinga, Jan P. Schouten, Bert Rijcken, and Scott T. Weiss

Department of Environmental Health, Harvard School of Public Health; The Channing Laboratory, Brigham and Women’s Hospital,Harvard Medical School, Boston, Massachusetts; Department of Epidemiology, State University of Groningen, Groningen, the Netherlands

This study investigated the determinants of sex-specific maximallyattained levels of FEV1, VC, and the ratio of FEV1 to VC. Subjectswere between the ages of 15 and 35 years (1,818 males and 1,732females), participating in the Vlagtwedde/Vlaardingen study in TheNetherlands. The subjects were followed (3-year intervals) withquestionnaire, spirometry, peripheral blood eosinophil counts, andtesting for airway responsiveness to histamine. Skin tests were per-formed only at study onset. Regression splines were used to assessthe effects of these variables on levels of FEV1, VC, and the ratio ofFEV1 to VC, with adjustment for age, height, and area of residence.Current (�44 ml/pack/day) and cumulative (�85 ml/10 packs/year)cigarette smoking were significant predictors of reduced maximallevel of FEV1 in males but not in females. The presence of respiratorysymptoms (�114 ml in males, �106 ml in females), increased eosin-ophils (�128 ml [males], �53 ml [females]), and increased airwayresponsiveness (�225 ml [males], �213 ml [females]) were all signif-icant predictors of reduced level of FEV1. To the degree that thesefactors diminished plateau phase pulmonary function, they may beimportant predictors of chronic obstructive pulmonary disease inlater life.

Keywords: airway responsiveness; asthma; chronic obstructive pulmo-nary disease; cigarette smoking

Chronic obstructive pulmonary disease (COPD) is the fifth lead-ing cause of death and contributes substantially to morbidity inthe United States (1). Previous investigations on the naturalhistory of COPD have focused on the decline of lung functionmeasurements in adult subjects over the age of 40 (2). Cigarettesmoking has been thought to be the most important cause ofCOPD in these studies (3, 4). However, it has long been recog-nized that only a small fraction of smokers, about 10 to 20%, goon to develop COPD, which suggests that there is considerableindividual variation in disease susceptibility (5). Thus, investiga-tions have focused extensively on the early childhood period,particularly on the effects of passive smoking (6) and early child-hood asthma (7) on maximal level of lung function and theirrole in identifying the susceptible smoker.

Although a general picture has emerged of the overall patternof growth and decline of pulmonary function over a person’slifetime (5, 8, 9), risk factors for reduced pulmonary functionbetween the ages of 15 and 35 years have been infrequentlyexplored in the epidemiology of COPD. The reasons for thisare many and include difficulty in assembly of cohorts in thisage range due to subject mobility and lack of data on relevantexposures. Understanding the determinants of maximal growth

(Received in original form January 8, 2002; accepted in final form January 20, 2004)

Supported by the Netherlands Asthma Fund grant 32.90.52 and National Heart,Lung, and Blood Institute grant HL49460.

Correspondence and requests for reprints should be addressed to Scott T. Weiss,M.D., M.S., Channing Laboratory, Brigham and Women’s Hospital, HarvardMedical School, 181 Longwood Ave., Boston, MA 02115. E-mail: scott.weiss@channing.harvard.edu

Am J Respir Crit Care Med Vol 169. pp 941–949, 2004DOI: 10.1164/rccm.2201011Internet address: www.atsjournals.org

is critical as maximal lung growth is an important mediator ofsubsequent development of COPD. Knowledge about risk fac-tors for maximal lung growth would provide new insights intothe natural history of COPD and help guide intervention efforts.Using nonparametric curve-smoothing methods, two longitudi-nal studies (10, 11) recently examined the effects of smoking andrespiratory symptoms on pulmonary function from childhoodto late adulthood. The data suggest that reduced level and/orpremature decline of pulmonary function during the plateauphase are alternative models for increased risk of COPD.

The Vlagtwedde/Vlaardingen study is unique in having anextensive follow-up of individuals initially between the ages of15 and 54 years followed for a 24-year period, with pulmonaryfunction measurements and relevant information on cigarettesmoking, respiratory symptoms and diseases, airway respon-siveness, and allergy markers. It offers an opportunity to studylongitudinally the effects of these factors on pulmonary functionduring the plateau phase and on the subsequent decline of pul-monary function in later life. It also helps bridge the gap betweenrecent epidemiologic studies of pulmonary function growth inchildren and adolescents (12, 13) and studies of pulmonary func-tion decline in late adulthood (14–17).

The purpose of this study is to describe the sex-specific pat-terns of maximal level of FEV1, VC, and the ratio of FEV1 toVC between ages 15 and 35 years and to examine the associationsof these pulmonary function measurements with various riskfactors, including cigarette smoking, respiratory symptoms, air-way responsiveness, eosinophilia, and positive skin tests, as wellas possible interactions among these factors.

METHODS

Population

Detailed descriptions of the Vlagtwedde/Vlaardingen study have beenpublished elsewhere (18–22). Since the baseline survey, the two cohortshave been reexamined every 3 years, beginning in 1970 in Vlagtweddeand 1972 in Vlaardingen.

Respiratory Symptoms and Smoking Habits

At each survey, questionnaires on respiratory symptoms and smokinghabits were administered by trained interviewers. Chronic cough orchronic phlegm, bronchitis, persistent wheeze, dyspnea (grade � 3),and asthma were defined as in previous reports (18–22). A subject wasconsidered symptomatic if he/she had any of the six symptoms/diseasesdescribed previously.

Smoking was considered as both a categorical (never, ex, 1–4, 5–15,15–24, 25� cigarettes/day) and a continuous (packs/day) variable.Never-smokers were defined as subjects who never smoked cigarettes,and ex-smokers were those who had stopped smoking at least 1 monthbefore each examination.

Pulmonary Function

Pulmonary function was measured at each survey with a water-sealedspirometer (Lode Spirograph D53; Lode Instruments, Groningen, TheNetherlands) while the subjects were seated and wearing nose clips.An inspiratory VC was measured after a deep expiration, followed bythe measurement of FEV1.

942 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

Eosinophil Count

Eosinophil counts were performed using methods described previously(21). Eosinophilia was considered to be present at a count greater thanor equal to 275 (25 � 11) eosinophils per cubic millimeter of blood, asit provides the best correlation with both symptoms and level of lungfunction (21).

Skin Tests

In this analysis, skin tests were coded on a six-point scale after sub-tracting positive controls and were considered to be positive at a scoreof 3 or more. This definition was chosen as it provides the best correla-tion with both symptoms (22) and level of lung function (21).

Airway Responsiveness

All the challenge tests (19, 20) used the method of Tiffeneau withsome modification by De Vries and coworkers (23). This method meetsstandardization guidelines (24). In this analysis, a subject was classifiedas having increased histamine airway responsiveness if provocative con-centration of histamine sufficient to produce a 10% drop in FEV1

was 16 mg/ml or less. This definition is equivalent to a provocativeconcentration of histamine sufficient to produce a 20% drop in FEV1

of less than 8 mg/ml, the upper limit of the asthmatic range (20).

Statistical Analysis

All the observations of subjects between ages 15 and 35 years wereincluded in the analysis. Neither models proposed for pulmonary func-tion in children (12–14) nor in adults (15, 16) are adequate to describepulmonary function during this age range. Tager and coworkers (10) andSherrill and coworkers (11) used nonparametric smoothing methods todescribe lung function growth and decline from childhood throughadulthood. In this study, besides using a robust smoothing technique(25), we employed regression spline models (26, 27), which provide aflexible family of growth curve models, are fully parametric, and permitthe use of familiar regression techniques for the assessment of covari-ates.

Our modeling approach for pulmonary function level used linearregression splines with age knots (14, 16, 18, 20, 22, 24, 29, 34). Theregression coefficients are estimated assuming independence amongall observations. All variables, including the symptoms and diseases,smoking, and airway responsiveness, are considered as time-dependentcovariates, where subjects with and without the phenotypic trait arecompared. The detailed statistical methods on parameter estimationcan be found elsewhere (27). Robust variance estimates (28) werefurther calculated for the estimated regression coefficients to accommo-date repeated measures on subjects. Graphic and residual analyses wereperformed to assess modeling assumptions. All predictor variables wereevaluated individually and jointly as predictors of the level of VC,FEV1, and the ratio of FEV1 to VC. A subject’s age, height, and areaof residence were included in all models. All analyses were stratifiedby sex. Differences in the effect estimates between sexes were comparedusing two-tailed t tests. Because the stratified analyses (Table 5) sug-gested possible interactions, five of these were formally tested usingcross-product terms in the sex-stratified regression models.

RESULTS

This analysis includes a total of 1,818 male and 1,732 femalesubjects who had at least one acceptable pulmonary functiontest between ages 15 and 35 years; they contributed 4,378 and3,716 observations, respectively. Their characteristics are pre-sented in Table 1. Twenty-nine percent of males were never-smokers, compared with 46% of females. In addition, men weremore likely to be current-smokers, to smoke more heavily, andto have smoked a greater cumulative amount of tobacco thanwomen. About two-thirds of subjects resided in Vlagtwedde (arural community) and about one-third in Vlaardingen (an urbancommunity). There were significant associations between smok-ing and respiratory symptoms: 12% of male never-smokers and13% of female never-smokers reported ever having a chronicrespiratory symptom between ages 15 and 35 years. In compari-

son, the symptom-reporting rates were about three times andtwo times higher in male and female ever-smokers, respectively.Ever-smokers of both sexes were also twice as likely to be diag-nosed with bronchitis (p � 0.01), but their rates of asthma wereonly slightly higher than those of never-smokers. About 16% ofnever-smokers were classified as having eosinophilia (� 275/mm3)compared with 21% of ever-smokers. Of those tested, about 23%of never-smokers and 26% of ever-smokers were considered tobe hyperresponsive to histamine challenge test (PC10 � 16 mg/ml).Males had higher rates of positive skin tests than females amongnever-smokers (29 vs. 20%) and ever-smokers (27 vs. 24%).

Figure 1 presents the sex-specific mean FEV1, VC, and theratio of FEV1 to VC of subjects between ages 15 and 35 years(smoothed). Males had significantly higher levels of FEV1 andVC than females, even after correction for body size. However,females had a greater ratio of FEV1 to VC than that of malesover the whole age range. Consistent with previous data thatfemales begin an adolescent growth spurt about 2 years earlierthan males (13, 14), this study shows that females appear toreach the maximal adult pulmonary function level several yearsearlier than males. For example, in females, FEV1 had alreadyreached its plateau by age 15 years, whereas in males, FEV1

continued to rise until about 20 years of age. It is interestingthat the earlier maturation in females did not lead to an earlierdecline in pulmonary function. In fact, pulmonary function infemales declined at about the same age as it did in males (around25 years). In males, the ratio of FEV1 to VC increased until age17 years and then declined approximately linearly. In females,this ratio showed a uniform decline over the age range.

Sex-specific mean FEV1 residuals and their 95% confidenceintervals by personal smoking status, respiratory symptoms, eo-sinophil count, airway response, and skin tests are shown inFigure 2. The prediction models used to obtain the residualsincluded age, height, and area of residence. Current-smokershad a lower level of FEV1 than never-smokers. The data alsosuggested a dose–response relationship. The effects of smokingon FEV1 appear to be greater in males than in females. Presenceof respiratory symptoms, high eosinophil count, and airway hyp-erresponsiveness, each was a significant predictor of lower levelof FEV1. Skin tests, however, were not significantly associatedwith FEV1. Similar associations were observed for VC and theratio of FEV1 to VC, but the magnitude of effects for VC wassmaller than that for FEV1 (data not presented).

The associations between personal smoking status, respira-tory symptoms, eosinophil count and airway responsiveness, andlower level of pulmonary function suggested by the residualanalyses were further evaluated using the regression splines. Asshown in Table 2, in males, each 10 cumulative pack-years wasassociated with an 85-ml (SE � 21) deficit in FEV1, and currentlysmoking one pack per day contributed an additional 44-ml (SE �22) deficit. The effects of smoking on FEV1 in females were lessthan half of those observed in males; however, the sex differencesin smoking effects were not statistically significant. Presence ofrespiratory symptoms was an independent predictor of lowerlevel of FEV1 in both sexes (�114 ml in males and �106 ml infemales), as was airway hyperresponsiveness (�225 ml in malesand �213 ml in females). Eosinophilia was also significantlyassociated with lower level of FEV1, but its effect in males (�128ml) was more than twice that in females (�53 ml). The differencewas statistically significant (p � 0.05). Positive skin tests werenot a significant predictor in either sex.

As males did not reach their pulmonary function plateau untilabout 20 years of age, additional analyses were limited to malesubjects aged 20 to 35 years so that they were comparable withfemale subjects in maturational stage. The results obtained fromthis male subgroup were similar to those from the whole malesample.

Wang, Mensinga, Schouten, et al.: Determinants of Lung Function 943

TABLE 1. CHARACTERISTICS OF SUBJECTS AGED 15 TO 35 YEARS,VLAGTWEDDE/VLAARDINGEN STUDY*

Male Female

Never-smoker Ever-smoker Ever-smoker Never-smoker

(n ) (% ) (n ) (% ) (n ) (% ) (n ) (% )

Study areasVlagtwedde 327 61.8 845 65.6 516 65.5 581 61.6Vlaardingen 202 38.2 444 34.4 272 34.5 363 38.5

No. of spirometry tests1 225 42.5 512 39.7 404 51.3 368 39.0†

2–3 177 33.5 478 37.1 310 39.3 357 37.84* 127 24.0 299 23.2 74 9.4 219 23.2

Chronic symptomsCough

Yes 26 4.9 255 19.8† 51 6.5 136 14.4†

WheezeYes 11 2.1 89 6.9† 21 2.7 63 6.7†

SputumYes 32 6.1 210 16.3† 37 4.7 76 8.1†

DyspneaYes 14 2.6 82 6.4† 45 5.7 72 7.6†

Any of aboveYes 63 11.9 385 29.9† 104 13.2 215 22.8†

Ever diagnosedAsthma

Yes 22 4.2 78 6.1 34 4.3 49 5.2Bronchitis

Yes 58 11.0 244 18.9† 83 10.5 204 21.6†

Eosinophil� 275/mm3 442 83.6 1,011 78.4‡ 667 84.6 742 78.6†

� 275/mm3 87 16.4 278 21.6 121 15.4 202 21.4Airway response

(�) 129 77.2 344 74.1 169 76.8 207 74.7(�) 38 22.8 120 25.9 51 23.2 70 25.3Not tested 362 825 568 667

Skin test(�) 302 71.2 818 72.8 528 80.1 591 76.1(�) 122 28.8 306 27.2 131 19.9 186 23.9Not tested 105 165 129 167Total subjects 529 1,289 788 944

* Based on first available observation.† 2-Square test (p � 0.01).‡ 2-Square test (p � 0.05).

Similar analyses were performed for VC and the ratio ofFEV1 to VC (Tables 3 and 4). Although the findings in generalwere similar to those observed for FEV1, there are a few pointsworthy of mention. As shown in Table 3, the effects of cigarettesmoking on VC were smaller than those for FEV1 and not sig-nificant in either sex. The same was true for respiratory symp-toms. The effects of eosinophil count and airway responsivenesson VC were smaller than those for FEV1 but remained significant.

Analyses stratified by smoking, respiratory symptoms, andeosinophil count were performed to evaluate the consistency ofthe associations across the strata (Table 5). The data suggestpossible interactions between smoking and symptoms, betweensmoking and eosinophil count, between skin tests and symptoms,and between skin tests and eosinophil count in males, and be-tween skin tests and symptoms in females. These possible inter-actions were further assessed by including their product termsin the regression models. Three of them were found statisticallysignificant: skin test � symptoms and skin test � eosinophilcount in males, and skin test � symptoms in females. Thesefindings indicate that positive skin tests could be an importantpredictor of reduced level of FEV1 in a person with respiratorysymptoms and eosinophilia.

Due to the additive nature of the effects of smoking, respira-tory symptoms, airway responsiveness, and eosinophil count,one can simply sum up a person’s total deficit in pulmonaryfunction on the basis of his/her risk factors. Figure 3 illustratessix hypothetical cases who are all 25-year-old males (and fe-males), with the same height and residence: (a) an asymptomaticnever-smoker, without airway hyperresponsiveness or eosino-philia (reference); (b) a subject who has smoked 1 pack/daysince 15 years of age but is asymptomatic, without airway hyper-responsiveness and eosinophilia; (c) a subject who is symptom-atic but has never smoked, without airway hyperresponsivenessor eosinophilia; (d) a subject who has eosinophilia but has neversmoked, is asymptomatic, and without airway hyperresponsiveness;(e) a subject who has airway hyperresponsiveness but has neversmoked, is asymptomatic, and without eosinophilia; (f) a subjectwho combines all the risk factors of subjects b, c, d, and e.The projected deficits in FEV1 are 0, 129, 114, 128, 225, and 596ml for subjects a, b, c, d, e, and f, respectively.

The relative importance of each individual symptom to pul-monary function was examined, with adjustment for personalsmoking status, eosinophil count, airway responsiveness, andskin tests. In males, wheeze and dyspnea were much stronger

944 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

Figure 1. Sex-specific smoothed mean curve of FEV1, VC, and the ratioof FEV1/VC of subjects between ages 15 and 35 years, the Vlagtwedde/Vlaardingen study. Open circles represent males and filled circles representfemales. Linear regression splines with age knots at 15, 17, 19, 21, 23,25, 30, and 35 years.

predictors of VC than cough and phlegm, whereas all four symp-toms were important predictors for FEV1. For females, all symp-toms but phlegm were significant predictors of FEV1.

The effects of physician-diagnosed asthma and bronchitis onpulmonary function during the plateau phase were also investi-gated. Asthma was a significant predictor of all three pulmonaryfunction measurements and was associated with 348 ml, 155 ml,and 5% deficits in FEV1, VC, and the ratio of FEV1 to VC,respectively, in males, after adjustment for cigarette smoking,eosinophil count, airway responsiveness, and skin tests. Similareffects were observed in females with deficits of 273, 165, and4 ml for the three spirometric measures of pulmonary function.In comparison, bronchitis was a much weaker predictor in bothsexes (data not shown).

DISCUSSION

This report focuses on factors that predict maximal level ofprebronchodilator FEV1, VC, and FEV1 to VC ratio in the agerange where pulmonary function is maximal, i.e., 15–35 years.Our results suggest that airway responsiveness, peripheral bloodeosinophil count, respiratory symptoms, and cigarette smokingare all important predictors of maximal level of lung function.An additional finding is that skin test reactivity was a significant

predictor of maximal level of FEV1 in individuals with respira-tory symptoms and/or peripheral blood eosinophilia. Finally, wefound sex differences in the length of the plateau period for thedifferent measures of pulmonary function.

The graphic plots (Figure 1) suggest substantial sex differ-ences in maximal level of lung function. In females, FEV1 hasalready peaked by age 15 years, whereas the peak in males doesnot occur until age 20 years. Decline in FEV1 occurred in bothsexes at about the same age (25 years), which suggests that theplateau phase for FEV1 is longer in females than in males. Thereasons for this are not known and may relate to differences inbiology as well as potential differences in exposures. The maxi-mal VC tended to lag behind the FEV1 for both sexes by 2–3years. Males had consistently lower FEV1 to VC ratios thanfemales throughout the age range. This finding may have prog-nostic significance in that Burrows and coworkers (29) haveshown that the presence of a low ratio is a significant predictorof accelerated decline in lung function and hence a risk factorfor COPD.

Precise data on the onset and the duration of the plateauphase are lacking. Tager and coworkers (10), using data fromthe East Boston Study, suggested a peak FEV1 for females atage 18 years and for males from ages 20–34 years. This study isnot directly comparable with the current data as it was stratifiedby symptom status and included relatively sparse data in theage range of greatest interest, e.g., (15–35 years). Our data areconsistent with the hypothesis that there are sex differences inthe plateau phase of pulmonary function. To some extent thesedifferences can be explained by differences in risk factors forless than maximal lung growth.

Our analysis suggests several important points about the riskfactors: personal cigarette smoking, skin test reactivity, eosino-philia, respiratory symptoms, and airway responsiveness. Per-sonal cigarette smoking was associated with a 5% reduction inmaximally attained FEV1. The effect was independent of theother risk factors and approximately equal in magnitude to theeffect of respiratory symptoms and eosinophilia. The magnitudeof the smoking effect is greater than the effect of smoking seenin adults older than 35 years in three ways: the absolute effectis greater, the influence is on maximal growth, and possiblypremature decline, not simply age-related decline after age 35years. These results emphasize the importance of early interven-tion on cigarette smoking to prevent the development of COPD,as this is the only time in the life cycle that there can be actualpreservation of lung function. Sex differences were observed forcigarette smoking; but, because of the small number of youngfemale smokers, these were not significant. Thus, more recentdata on this effect would be of value. Tager and coworkers foundthat cigarette smoking was associated both with a reduced leveland a premature decline in FEV1. Although we did not studythis latter phenomenon, our data are consistent with theirs onthe effect on level of FEV1.

Previous investigators have found that respiratory symptomsin childhood are associated with a reduction in maximally at-tained level of pulmonary function (7, 30). In our investigation,the effects were approximately equal for males and females. Theeffects were of comparable magnitude with that seen for activecigarette smoking and for eosinophilia and were not due toasthma alone. However, the effect of an asthma diagnosis wasgreater than the effect of individual respiratory symptoms, andwhen examined separately, an asthma diagnosis was the onlyfactor to approximate airway responsiveness in the magnitudeof its effect on maximal lung growth. A variety of studies havenoted an important effect of asthma on the maximal growth oflung function in children (7, 31). Asthma (symptoms plus airwayresponsiveness) was associated with a 10 to 15% reduction of

Wang, Mensinga, Schouten, et al.: Determinants of Lung Function 945

Figure 2. Sex-specific meanresiduals of FEV1 and their 95%confidence intervals by per-sonal smoking status, respira-tory symptoms, eosinophilcount, airway responsiveness,and skin tests of subjects be-tween ages 15 and 35 years,the Vlagtwedde/Vlaardingenstudy.

maximal FEV1. The strength of our investigation is that it con-trols for all other known risk factors and separates the effectfrom that of allergy and airway responsiveness.

Airway responsiveness was the single most important riskfactor for a reduced maximal level of lung function. In early

TABLE 2. SEX-SPECIFIC PREDICTORS OF FEV1 LEVEL BETWEEN AGES 15 AND 35 YEARS,VLAGTWEDDE/VLAARDINGEN STUDY

Male Female

Predictors Effects* (ml) SE p Value Effects* (ml) SE p Value

Cigarette smokingNever-smokers (reference)Cumulative

10 pack-years �85 21 � 0.01 �40 30 0.18Current

Pack/d �44 22 � 0.05 �17 26 0.52Respiratory symptoms†

Absence (reference)Presence �114 30 � 0.01 �106 25 � 0.01

Eosinophil count� 275/mm3 (reference)� 275/mm3 �128 28 � 0.01 �53 22 � 0.05

Airway responsivenessLow (reference)High �225 48 � 0.01 �213 39 � 0.01Unknown �123 24 � 0.01 �85 18 � 0.01

Skin testNegative (reference)Positive �31 29 0.29 3 24 0.89Unknown 5 21 0.79 59 15 � 0.01

* The spline regression models include age, height, and residence, in addition to predictors listed previously.† Respiratory symptoms: presence of any of the following chronic conditions—cough, wheeze, sputum, and dyspnea.

adulthood, the effect (5%) was similar in males and females andwas independent of all other risk factors most notably asthma,eosinophilia, and skin test reactivity. This result is consistentwith the data of Redline and coworkers (32, 33) who found suchan effect in a longitudinal study of children in East Boston,

946 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

TABLE 3. SEX-SPECIFIC PREDICTORS OF VC LEVEL BETWEEN AGES 15 AND 35 YEARS,VLAGTWEDDE/VLAARDINGEN STUDY

Male Female

Predictors Effects* (ml) SE p Value Effects* (ml) SE p Value

Cigarette smokingNever-smokers (reference)Cumulative

10 pack-years �12 24 0.63 �42 30 0.16Current

Pack/d �15 24 0.53 �7 27 0.80Respiratory symptoms†

Absence (reference)Presence �40 29 0.17 �24 28 0.38

Eosinophil count� 275/mm3 (reference) �

� 275/mm3 �77 31 0.05 �13 24 0.57Airway responsiveness

Low (reference)High �121 53 � 0.05 �122 43 � 0.01Not tested �26 25 0.30 �9 21 0.66

Skin testNegative (reference)Positive �5 30 0.86 18 27 0.51Unknown �6 22 0.80 56 18 � 0.01

* The spline regression models include age, height, and residence, in addition to predictors listed previously.† Respiratory symptoms: presence of any of the following chronic conditions—cough, wheeze, sputum, and dyspnea.

Massachusetts. It is also consistent with the concept that airwayresponsiveness is closely correlated with level of lung functionat all ages (8). The importance of isolated airway responsivenessin the absence of asthma is twofold. A number of studies inchildren and young adults suggest that airway responsivenessantedates and predicts clinical asthma (34, 35). In addition, stud-ies in older adults suggest that airway responsiveness predictsaccelerated decline in FEV1 and hence the development ofCOPD (36–39). At the present time, it is unclear whether thisrepresents subclinical inflammation and whether treatment canmodify the effect on maximal lung growth. Further studies and

TABLE 4. SEX-SPECIFIC PREDICTORS OF FEV1/VC LEVEL BETWEEN AGES 15 AND 35 YEARS,VLAGTWEDDE/VLAARDINGEN STUDY

Male Female

Predictors Effects* (% ) SE p Value Effects* (% ) SE p Value

Cigarette smokingNever-smokers (reference)Cumulative

10 pack-years �1.5 0.3 � 0.01 �0.1 0.5 0.92Current

Pack/d �0.6 0.3 � 0.05 �0.3 0.5 0.51Respiratory symptoms†

Absence (reference)Presence �1.8 0.5 � 0.01 �2.6 0.4 � 0.01

Eosinophil count� 275/mm3 (reference)� 275/mm3 �1.4 0.4 � 0.01 �1.2 0.4 � 0.01

Airway responsivenessLow (reference)High �2.6 0.7 � 0.01 �3.4 0.8 � 0.01Not tested 2.0 0.3 � 0.01 �2.1 0.3 � 0.01

Skin testNegative (reference)Positive �0.5 0.4 0.24 �0.3 0.4 0.42Unknown 0.2 0.3 0.42 0.4 0.3 0.13

* The spline regression models include age, height, and residence, in addition to factors listed previously.† Respiratory symptoms: presence of any of the following chronic conditions—cough, wheeze, sputum, and dyspnea.

confirmation of our results would add greatly to our knowledgein this area.

Although an independent effect of skin test reactivity onmaximal lung growth could not be seen in these data, therewas an independent effect of peripheral blood eosinophilia onmaximal lung growth in both males and females. The effect wasgreater in males than in females, and this sex difference wasstatistically significant. This relates at least in part to the interac-tion between skin test reactivity and eosinophils in males butmay be the result of sex differences in exposure or the biologicprocessing of antigen, differences in other exposures, or other

Wang, Mensinga, Schouten, et al.: Determinants of Lung Function 947

TABLE 5. SEX-SPECIFIC STRATIFIED ANALYSES BY SMOKING, SYMPTOMS,AND EOSINOPHIL COUNTS FOR FEV1 LEVEL BETWEEN AGES 15 AND 35 YEARS,VLAGTWEDDE/VLAARDINGEN STUDY

Male Female

Predictors Effects* (% ) SE p Value Effects* (% ) SE p Value

Stratifying by smoking statusRespiratory symptoms

Never-smokers �77 96 0.42 �151 45 � 0.01Ever-smokers �147 30 � 0.01 �99 30 � 0.01

EosinophiliaNever-smokers �175 66 � 0.01 �78 36 0.03Ever-smokers �126 31 � 0.01 �42 27 0.12

Airway hyperresponsivenessNever-smokers �290 101 � 0.01 �315 67 � 0.01Ever-smokers �220 55 � 0.01 �144 48 � 0.01

Positive skin testNever-smokers 40 56 0.48 �14 38 0.72Ever-smokers �45 34 0.18 15 31 0.63

Stratifying by respiratory symptomsCigarette smoking

Cumulative, 10 pack-yearsAsymptomatic �76 23 � 0.01 �25 29 0.40Symptomatic �126 52 � 0.05 �26 69 0.71

Current, 1 pack/dAsymptomatic �66 23 � 0.01 �33 28 0.24Symptomatic 45 50 0.37 3 57 0.96

Eosinophil � 275/mm3

Asymptomatic �132 28 � 0.01 �47 23 � 0.05Symptomatic �86 74 0.25 �59 60 0.33

Airway hyperresponsivenessAsymptomatic �206 54 � 0.01 �217 42 � 0.01Symptomatic �293 94 � 0.01 �183 106 0.08

Positive skin testAsymptomatic 22 29 0.44† 26 26 0.32†

Symptomatic �323 91 � 0.01 �148 63 � 0.02Stratifying by eosinophil count

Cigarette smokingCumulative, 10 pack-years

� 275/mm3 �78 22 � 0.01 �41 28 0.15� 275/mm3 �142 57 � 0.05 28 84 0.74

Current, 1 pack/d� 275/mm3 �61 23 � 0.01 �23 28 0.40� 275/mm3 78 51 0.13 23 62 0.71

Respiratory symptoms� 275/mm3 �107 31 � 0.01 �99 27 � 0.01� 275/mm3 �144 79 0.07 �108 60 0.07

Airway hyperresponsiveness� 275/mm3 �235 52 � 0.01 �220 42 � 0.01� 275/mm3 �168 121 0.17 �197 91 0.03

Positive skin test� 275/mm3 8 31 0.79† 11 26 0.67� 275/mm3 �248 64 � 0.01 �30 56 0.59

* The spline regression models include age, height, and residence, in addition to above-listed predictors (except for stratifyingfactor).

† Test for interaction (p � 0.05).

factors requiring further investigation. Skin test reactivity wasalso a significant predictor of maximal lung growth in those withrespiratory symptoms regardless of sex.

Using a stratified analysis, Robbins and coworkers (40) haveshown that there is substantial individual variability in growthtrajectories for FEV1 within sexes. Smokers were much morelikely to begin to decline early in the plateau phase. These find-ings are consistent with our results. The host and environmentalfactors we assessed contribute to explaining this variation, butwe cannot rule out the role of other as yet unidentified sourcesof variation. Although our analysis is the most comprehensiveperformed to date on the quantitation of risk factors influencingmaximally attained level of lung function, certain limitations on

exposures must be acknowledged. First, we have no data on indoorand outdoor air pollution, specifically we have no data on allergen-exposure levels, exposure to farm animals, and passive smoking.We did choose to include area of residence as a covariate in ourmodels as we wanted to control for sampling differences in thetwo communities and any potential urban/rural differences inexposure. Second, we have no data on birth-related events, early-life respiratory illness, and passive-smoke exposure. Third, wehave no data on total and specific IgE, and only two measuresof skin test reactivity in comparison with the multiple measureof exposure are available for other risk factors. An additionalconcern is that maximum prebronchodilator FEV1 occurred infemales before age 15 years at a time in the life cycle when we

948 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

Figure 3. Sex-specific predicted deficits in FEV1 by risk statusfor six hypothetical cases who are all 25-year-old male sub-jects, with the same height and residence. all � a combinationof all the risk factors; eos � eosinophilia; hist � airway hyper-responsiveness; smk � smoking, 10 pack-years; symp � pres-ence of chronic respiratory symptoms.

had no other data. Finally, by measuring pulmonary function at3-year intervals we may have missed the maximum value. Thislatter concern is minimized as the majority of subjects had twoor more values of pulmonary function in the interval 16–35 years.

Despite these deficiencies, these data confirm the indepen-dent and additive effects of respiratory symptoms and cigarettesmoking and identify that asthma and airway responsiveness arethe greatest risk factors for a reduced maximal level of lung func-tion in early adult life. It remains to be seen whether treatmenttrials currently underway will be able to modify the deleteriouseffects of childhood asthma on lung growth. PrebronchodilatorFEV1 has been the standard outcome variable in epidemiologicstudies. To the extent that resting bronchometer tone due toairways responsiveness is an important predictor of maximalattained level of prebronchodilator FEV1, the effects observedhere would tend to minimize what might be seen if postbronchod-ilator FEV1 was the outcome variable.

Maximally attained level of pulmonary function is importantbecause it is the most important risk factor for the developmentof COPD potentially. This suggests that an understanding of thedeterminants of maximal lung growth is essential in designingintervention programs to prevent COPD.

Our investigation has several unique features. First, we haveinformation on a large number of factors thought to be importantin determining maximal lung growth: respiratory symptoms, skintest reactivity, airway responsiveness, eosinophilia, and cigarettesmoking. Second, we used longitudinal data to predict maximallyattained level of prebronchodilator FEV1 and VC. More than50% of subjects had at least two values in the age interval. Third,our mathematical modeling of maximally attained level of lungfunction was not constrained to be linear.

Although differences in risk factors and body size can clearlycontribute to sex differences in maximally attained level of pul-monary function, it is also worthwhile to consider the importanceof absolute versus proportional differences in rate of decline inFEV1. Not only do women have lower absolute levels of FEV1

and VC, but proportionally, as a percent of VC, effects forsymptoms and airway responsiveness seem to be greater, sug-gesting that both absolute and proportional differences maycontribute to sex-specific differences in risk. This concept needsfurther exploration.

In summary, we have explored the relationship of cigarettesmoking, respiratory symptoms, airway responsiveness, skin testreactivity, and eosinophilia on maximal levels of lung growth in

a population-based cohort. Our results suggest that smoking,respiratory symptoms, and especially airway responsiveness areimportant determinants of maximal lung growth as measuredby FEV and VC. Given the importance of maximally attainedlevel of prebronchodilator FEV1 as a predictor of COPD, thesefindings target subjects at potentially high risk in whom interven-tion is justified.

Conflict of Interest Statement : X.W. has no declared conflict of interest; T.T.M.has no declared conflict of interest; J.P.S. has no declared conflict of interest; B.R.has no declared conflict of interest; S.T.W. received a grant for $900,065, AsthmaPolicy Modeling Study, from AstraZeneca from 1997–2003, was a Co-Investigatoron a grant from Millennium Pharmaceuticals to pursue asthma genetics in Chinafrom 1996–2001, received a grant from Pfizer to examine diabetes mellitus and itsrelationship to lung function between 2000–2003, was a consultant for Schering-Plough and received $5000 from 1999–2000, was Co-Investigator on a grantfrom Boehringer Ingelheim to investigate a COPD natural history model whichbegan in 2003 and received no funds for his involvement in this project, has beena consultant for Variagenics on human subject issues and received $5000 in 2003,has been a consultant to Genome Therapeutics in 2003 and received $1500, wasa consultant for Merck Frost on asthma genetics in 2002 and received $2000,has been an advisor to the TENOR Study for Genentech and has received $5000for 2002–2003, received a grant from Glaxo-Wellcome for $500,000 for genomicequipment for 2000–2003, and was consultant for Roche Pharmaceuticals in 2000and received no financial remuneration for this consultancy.

References1. U.S. Department of Health and Human Services. The Health Conse-

quences of Smoking: Chronic Obstructive Lung Disease. Washington,DC: U.S. Government Printing Office; DHHS Publication No. 84-50205.

2. Fletcher C, Peto R, Tinker C, Speizer FE. The natural history of chronicbronchitis and emphysema. New York: Oxford University Press; 1976.

3. Xu X, Dockery DW, Ware JH, Speizer FE, Ferris BG Jr. Effects ofcigarette smoking on rate of loss of pulmonary function in adults: alongitudinal assessment. Am Rev Respir Dis 1992;146:1345–1348.

4. Xu X, Weiss ST, Rijcken B, Schouten JP. Association of smoking andchanges in smoking habits with rate of loss of FEV1: a new insightinto gender differences. Eur Respir J 1994;7:1056–1061.

5. Speizer FE, Tager IB. Epidemiology of chronic mucus hypersecretionand obstructive airway disease. Epidemiol Rev 1979;1:124–142.

6. Tager IB, Hanrahan JP, Tosteson TD, Castile RB, Brown RW, WeissST, Speizer FE. Lung function, pre- and post-natal smoke exposure, andwheezing in the first year of life. Am Rev Respir Dis 1993;147:811–817.

7. Weiss ST, Segal MR, Tager IB, Tosteson T, Redline S, Speizer FE.Effects of asthma on pulmonary function in children: a longitudinalpopulation-based study. Am Rev Respir Dis 1992;145:58–64.

8. O’Connor G, Sparrow D, Weiss ST. The role of atopy and airwaysresponsiveness in the pathogenesis of chronic airflow obstruction. AmRev Respir Dis 1989;140:225–252.

9. Sparrow D, Weiss ST. Background In: Weiss ST, Sparrow D, editors. Therelationship of airways responsiveness and atopy to the development ofthe obstructive airways diseases. New York: Raven Press; 1989. p. 1–21.

Wang, Mensinga, Schouten, et al.: Determinants of Lung Function 949

10. Tager IB, Segal MR, Speizer FE, Weiss ST. The natural history of forcedexpiratory volumes. Am Rev Respir Dis 1988;138:837–849.

11. Sherrill DL, Lebowitz MD, Knudson RJ, Burrows B. Smoking and symp-toms effects on the curves of lung function and decline. Am Rev RespirDis 1991;144:17–22.

12. Wang X, Dockery DW, Wypij D, Fay ME, Ferris BG Jr. Pulmonaryfuncion between 6 and 18 years of age. Pediatr Pulmonol 1993;15:75–88.

13. Wang X, Wypij D, Gold DR, Speizer FE, Ware JH, Ferris BG Jr, DockeryDW. A longitudinal study of the effects of parental smoking on pulmo-nary function in children 6–18 years. Am J Respir Crit Care Med 1994;149:1420–1425.

14. Dockery DW, Ware JH, Ferris BG Jr, Glicksberg DS, Fay ME, Spiro A3rd, Speizer FE. Distribution of forced expiratory volume in one sec-ond and forced vital capacity in healthy white adult never-smokers insix US cities. Am Rev Respir Dis 1985;131:511–520.

15. Ware JH, Dockery DW, Louis TA, Xu X, Ferris BG, Speizer FE. Longitu-dinal and cross-sectional estimates of pulmonary function decline innever-smoking adults. Am J Epidemiol 1990;132:685–700.

16. Burrows B, Lebowitz MD, Camilli AE, Knudson RJ. Longitudinalchanges in forced expiratory volume in one second in adults: method-ological considerations and findings in healthy nonsmokers. Am RevRespir Dis 1986;133:974–980.

17. Xu X, Christiani DC, Dockery DW, Wang LH. Exposure-response rela-tionships between occupational exposures and chronic respiratory ill-ness: a community based study. Am Rev Respir Dis 1992;146:413–418.

18. Van Der Lende R, Kok TJ, Peset Reig R, Quanjer PhH, Schouten JP,Orie NGM. Decreases in VC and FEV1 with time: indicators for effectsof smoking and air pollution. Bull Eur Physiopathol Respir 1981;17:775–792.

19. Rijcken B, Schouten JP, Weiss ST, Speizer FE, van der Lende R. Therelationship of nonspecific bronchial responsiveness to respiratorysymptoms in a random population sample. Am Rev Respir Dis 1987;136:62–68.

20. Rijcken B, Schouten JP, Weiss ST, Meinesz AF, de Vries K, van derLende R. The distribution of bronchial responsiveness to histaminein symptomatic and in asymptomatic subjects. Am Rev Respir Dis 1989;140:615–623.

21. Mensinga TT, Schouten JP, Weiss ST, Van Der Lende R. Relationshipof skin test reactivity and eosinophilia to level of pulmonary functionin a community-based population study. Am Rev Respir Dis 1992;146:638–643.

22. Mensinga TT, Schouten JP, Weiss ST, Speizer FE. The relationship ofeosinophilia and positive skin test score to respiratory symptom preva-lence in a community based population study. J Allergy Clin Immunol1990;86:99–107.

23. De Vries K, Goei JT, Booy-Noord H, Orie NGM. Changes during 24hours in the lung function and histamine hyperreactivity of the bron-chial tree in asthmatic and bronchitic patients. Int Arch Allergy 1962;20:93–101.

24. Eiser NM, Kerrebijn KF, Quanjer PhH. Guidelines for standardization

of bronchial challenges with (non-specific) bronchoconstricting agents.Bull Eur Physiopathol Respir 1983;19:495–514.

25. Cleveland WS. Robust locally weighted regression and smoothing scat-terplots. J Am Stat Assoc 1979;74:829–836.

26. Wegman EJ, Wright IW. Splines in statistics. J Am Stat Assoc 1983;78:351–365.

27. Wypij D, Pugh M, Ware JH. Modeling pulmonary function growth withregression splines. Statist Sinica 1993;3:329–350.

28. Liang KY, Zeger SL. Longitudinal data analysis using generalized linearmodels. Biometrika 1986;73:13–22.

29. Burrows B, Lebowitz MD, Camilli AE, Knudson RJ. Longitudinalchanges in forced expiratory volume in one second in adults: methodo-logic considerations and findings in healthy nonsmokers. Am Rev Re-spir Dis 1986;133:974–980.

30. Gold D, Tager IB, Weiss ST, Tosteson T, Speizer FE. The relationshipof acute respiratory illness to level and growth of lung function. AmRev Respir Dis 1989;140:877–884.

31. Gold DR, Wypij D, Wang X, Speizer FE, Pugh M, Ware JH, Ferris BGJr, Dockery DW. Gender- and race-specific effects of asthma andwheeze on level and growth of lung function in children in six UScities. Am J Respir Crit Care Med 1994;149:1198–1208.

32. Redline S, Tager IB, Segal MR, Gold DR, Speizer FE, Weiss ST. Therelationship between longitudinal change in pulmonary function andnonspecific airway responsiveness in a population based sample ofchildren and young adults. Am Rev Respir Dis 1989;140:179–184.

33. Redline S, Tager IB, Rosner B, Speizer FE, Weiss ST. Longitudinalvariability in airway responsiveness in a population based sample ofchildren and young adults. Am Rev Respir Dis 1989;140:172–178.

34. Carey VJ, Weiss ST, Tager IB, Leeder SR, Speizer FE. Airway respon-siveness, wheeze onset and recurrent asthma episodes in young adoles-cents: the East Boston childhood respiratory disease cohort. Am JRespir Crit Care Med 1996;153:356–361.

35. Zhong NS, Chen RC, Yang MO, Wu ZY, Zheng JP, Li YF. Is asymptom-atic bronchial hyperresponsiveness an indication of potential asthma?A two-year follow-up of young students with bronchial hyperrespon-siveness. Chest 1992;102:1104–1109.

36. Frew AJ, Kennedy SM, Chan-Yeung M. Methacholine responsiveness,smoking and atopy as risk factors for accelerated FEV1 decline in maleworking populations. Am Rev Respir Dis 1992;146:878–883.

37. Villar MT, Dow L, Coggon D, Lampe FC, Holgate ST. The influence ofincreased bronchial responsiveness, atopy, and serum IgE on declinein FEV1: a longitudinal study in the elderly. Am J Respir Crit CareMed 1995;151:656–662.

38. Rijcken B, Xu X, Schouten JP, Rosner B, Weiss ST. Longitudinal analysesof the relationship of bronchial responsiveness and pulmonary functiondecline. Am J Respir Crit Care Med 1995;151:1377–1382.

39. O’Connor GT, Sparrow D, Weiss ST. A prospective longitudinal studyof methacholine airway responsiveness as a predictor of pulmonary-function decline: the Normative Aging Study. Am J Respir Crit CareMed 1995;152:87–92.

40. Robbins DR, Enright PL, Sherrill DL. Lung function development inyoung adults: is there a plateau phase? Eur Respir J 1995;8:768–772.

Twenty-four–hour Ambulatory Blood Pressure inChildren with Sleep-disordered BreathingRaouf S. Amin, John L. Carroll, Jenny L. Jeffries, Charles Grone, Judy A. Bean, Barbara Chini, Ronald Bokulic,and Stephen R. Daniels

Sleep Disorder Center, Divisions of Pulmonary Medicine, Epidemiology and Biostatistics, and Cardiology, The University of Cincinnati Collegeof Medicine, Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio; and Department of Pediatric Pulmonary Medicine, Universityof Arkansas for Medical Sciences, Little Rock, Arkansas

Obstructive sleep apnea causes intermittent elevation of systemicblood pressure (BP) during sleep. To determine whether obstructiveapnea in children has a tonic effect on diurnal BP, 24-hour ambula-tory blood pressure was obtained from 60 children with mean ageof 10.8 � 3.5 years. Thirty-nine children had obstructive apneaand 21 had primary snoring. Children with obstructive apnea hadsignificantly greater mean BP variability during wakefulness andsleep, a higher night-to-day systolic BP, and a smaller nocturnaldipping of mean BP. Variability of mean arterial pressure duringwakefulness was predicted by the desaturation, body mass, andarousal indices, whereas variability during sleep was predicted byapnea–hypopnea and body mass indices. Nocturnal BP dipping waspredicted by the desaturation index. There were no significant dif-ferences in systolic, diastolic, or mean arterial BP during sleep be-tween the groups. Diastolic BP during wakefulness was significantlydifferent between the groups and correlated negatively withapnea–hypopnea index. We conclude that obstructive apnea inchildren is associated with 24-hour BP dysregulation and that, inde-pendent of obesity, the frequency of obstructive apnea, oxygen de-saturation, and arousal contributes to abnormal BP control.

Keywords: obstructive apnea; ambulatory blood pressure; child; cardio-vascular

Obstructive sleep apnea (OSA) in adults increases the risk forheart failure, coronary artery disease, and stroke (1–3). Themechanisms underlying the link between OSA and cardiovascu-lar diseases are not completely understood; however, systemichypertension is thought to be one pathway leading to end-organdamage and cardiovascular morbidity (4–6). The early stages ofabnormal blood pressure (BP) control may present with auto-nomic dysfunction in the form of increased sympathetic activityand/or decreased vagal tone. These autonomic changes alter thediurnal control and variability of BP (7). The adverse effect ofabnormal BP control on the cardiovascular system is not onlydue to hypertension but also to earlier stages of abnormal BPcontrol, such as increased BP variability and decreased nocturnaldipping (8–16). Because children with OSA, in comparison withchildren with primary snoring, are at greater risk for developingend-organ damage in the form of left ventricular hypertrophy(17), we hypothesized that children with OSA will demonstratethe earlier stages of abnormal BP control in the form of elevation

(Received in original form September 23, 2003; accepted in final form February 4, 2004)

Supported by the American Heart Association.

Correspondence and requests for reprints should be addressed to Raouf S. Amin,M.D., Division of Pulmonary Medicine, Cincinnati Children’s Hospital MedicalCenter, 3333 Burnet Avenue, Cincinnati, OH 45229. E-mail: aminr0@chmcc.org

Am J Respir Crit Care Med Vol 169. pp 950–956, 2004Originally Published in Press as DOI: 10.1164/rccm.200309-1305OC on February 5, 2004Internet address: www.atsjournals.org

and decreased dipping of nocturnal BP and an increase in BPvariability. Some of the results of these studies have been re-ported previously in the form of an abstract (18).

METHODS

Study Design

Pediatric subjects who were referred for evaluation of obstructivebreathing disorder during sleep underwent polysomnography (PSG)with continuous BP recording, followed by a 24-hour recording ofambulatory blood pressure (AMBP). Evaluation at the time of enroll-ment consisted of a history and physical examination, BP, and bodymass index (BMI). All subjects presented with a history of snoring7 nights per week. PSG results were used to divide subjects into thosewith OSA and those with primary snoring. Subjects were classifiedas having primary snoring when they had no evidence of nocturnalhypoventilation and had an obstructive apnea–hypopnea index (AHI)between 0 and 1 per hour of sleep (Group 1). Subjects with AHI greaterthan 1 per hour of sleep were classified as having OSA and were furthersubdivided into two groups: those with AHI from 1 to 5 (Group 2) andthose with AHI greater than 5 per hour of sleep (Group 3). Informedconsent was obtained from the parents/legal guardian of each child,and assent was obtained from children older than 11 years of age. TheInstitutional Review Board of Cincinnati Children’s Hospital MedicalCenter, Ohio, approved the study.

Study Group

Subjects aged 5 to 17 years, who were referred to the pediatric SleepDisorder Clinic at Cincinnati Children’s Hospital Medical Center forevaluation of obstructive breathing during sleep, were recruited sequen-tially. Children with genetic syndromes, children with chronic medicalconditions including psychiatric conditions and attention deficit/hyper-activity disorders, children with conditions that might alter BP, andchildren who receive daily medications were excluded from the study.

PSG

PSG studies were performed overnight according to the AmericanThoracic Society standards (19, 20) using computerized systems (Grass;Telefactor, West Warwick, RI). The following parameters were re-corded during the study: (1 ) EEG, (2 ) right and left electrooculogram,(3 ) submental and tibial EMG, (4 ) ECG, (5 ) nasal/oral airflow mea-sured by thermocouple, (6 ) end-tidal Pco2 measured at the nose byinfrared capnometry and SaO2 using the Nelcor N1000 (Nelcor, VanNuys, CA) and oximeter pulse waveform, (7 ) snoring microphone, (8 )video monitoring using an infrared video camera, and (9 ) chest andabdominal wall motion by computer-assisted respiratory inductanceplethysmograph (Somnostar; Noninvasive Monitoring System Inc., Mi-ami Beach, FL).

The following parameters were measured.Sleep staging was scored according to published standards (21).

Arousals were defined as recommended by the American Sleep Disor-ders Association (22). Awakening from sleep was defined as an increasein the frequency of the EEG for more than 15 seconds.

Amin, Carroll, Jeffries, et al.: Blood Pressure in Children with Sleep Apnea 951

Obstructive apnea was defined as the presence of chest/abdominal wallmotion in the absence or decrease of airflow and/or the sum channelfrom the inductive plethysmography by more than 80% of the pre-ceding breath. All obstructive events greater than or equal to twobreaths duration were counted.

Obstructive hypopnea was defined as a reduction in airflow and/or thesum channel from the inductive plethysmography between 20 to50% in the presence of chest/abdominal wall motion, associatedwith oxyhemoglobin desaturation greater than/or equal to 4% andor followed by an arousal.

AHI was defined as the number of obstructive apneas and obstructivehypopneas per hour of sleep.

Oxyhemoglobin desaturation index (DI) was defined as the number ofevents per hour of sleep when SaO2 decreased by 4% or greater.

The average end-tidal Pco2 was calculated as an average of the maxi-mum values per 30-second epoch. Maximum carbon dioxide (CO2)level and time spent during sleep with an end-tidal CO2 level above45 and 50 mm Hg were calculated. The diagnosis of alveolar hypo-ventilation during sleep was made when two-thirds of the sleep timewas spent with an end-tidal CO2 higher than 45 mm Hg and/or 10%of the time was spent with level above 50 mm Hg.

BP Recording

BP was recorded with a SpaceLab monitor (Spacelabs Medical, Red-mond, WA). The reproducibility and validity of Spacelabs monitorshave been extensively studied (23–28). Subjects were asked to keep asleep diary and report their wake and sleep time. We ensured thatsleep–wake time was clearly documented before an AMBP study wasconsidered acceptable. When in doubt, the research coordinator con-tacted the child’s parent to confirm the reported data. The monitor wasprogrammed to cycle every 15 minutes, giving a maximum number ofmeasurements of 96 per 24 hours. The readings were divided into BPduring wakefulness and BP during sleep on the basis of the informationacquired from the sleep diary. The study was considered adequate whena minimum of 70% of the measurements were obtained without errorsboth for sleep and/or wake BP. The effect of BP recording on sleepwas derived from the BP data acquired during PSG from 41 subjects.The percentage of BP measurements associated with an arousal orawakening in the 30-second epoch that preceded the measurement andthe 30 seconds during the measurement was calculated. The BP cuffwas placed around the same arm as the pulse oximeter. The timing ofBP measurement could therefore be accurately determined becausecuff inflation was associated with gradual loss of the pulse amplitudewaveform obtained from the oximeter. The following parameters wereobtained and/or calculated from the continuous BP recording.

Measures of BP variability. BP variability was derived from the 24-hour AMBP by calculating the average SD of awake and sleep systolic,diastolic, and mean BP.

Measures of nocturnal BP dipping. The degree of systolic, diastolic,and mean BP dipping during sleep was derived by calculating thedifference between awake and sleep BP obtained from 24-hourAMBP and expressed as a percentage of BP during wakefulness.To determine whether the change in the degree of BP dipping isstage specific, the difference between mean awake BP derived fromthe 24-hour AMBP and BP obtained during the PSG while awake inbed and during each stage of sleep was also calculated. A comparisonbetween groups for night-to-day systolic BP ratio was performed,given its prognostic value in predicting future cardiovascular events(14, 15).

Average BP during wakefulness and sleep. Average systolic, diastolic,and mean arterial BP during wakefulness and sleep was calculated.Pressure load was measured by calculating the percentage of systolicand diastolic BP measurements above the 95th percentile for awakeBP. Published age-appropriate values for the 95th percentile for BPaccording to age and sex were used to determine the percentage ofBP greater than the 95th percentile (29).

Statistical Analysis

All results are expressed as mean � SD. Log transformation of BMI,AHI, BP SDs and night-to-day systolic BP ratio was performed toachieve normal distribution. BP measurements were expressed as a

mean of awake and sleep BP. To control for the effect of the wide agerange on BP, we calculated the difference between measured BP andthe published systolic and diastolic 95th percentile for each subject (30).We indexed this value to the 95th percentile using the following formula:

BP index �(Measured BP � BP at 95th percentile)

BP at 95th percentile� 100

The average of this index was compared between groups. BP resultswere also expressed as a percentage of the 95th percentile and comparedbetween groups. The BMI was converted into a Z score according tothe standards published by the CDC (Z BMI) (31). For comparison ofmeans, a one-factor analysis of variance was performed. An orthogonalcontrast for a linear trend was used to determine if a linear trend existedacross the mean of the groups: primary snorers, children with an AHIbetween 1 and 5, and children with an AHI greater than 5. Pearsoncorrelation was performed between log-transformed BP measurementsand log-transformed polysomnographic and demographic variables.Multiple regression analysis was performed to identify demographicand polysomnographic factors that might predict measures of BP vari-ability and nocturnal BP dipping. The following independent variableswere entered in a backward elimination regression analysis: age, sex,race, Z BMI, AHI, arousal index, DI, lowest SaO2, and maximum end-tidal CO2. Variables with a p value less than or equal to 0.05 were keptin the model. To control for possible effect of BMI on nocturnal BPdipping, this variable was forced into the model despite a p value greaterthan 0.05. Race was also forced into the model to control for ethnicity.

RESULTS

Study Population

Seventy-two subjects consented to participate in the study.Twelve subjects requested the discontinuation of BP recordingduring the night and refused to complete the 24-hour AMBPrecording. Subjects who withdrew from the study were: 41%females, 11.6 � 2.6 years old with a BMI of 28.7 � 9.7. Sixtysubjects completed the 24-hour AMBP monitoring. Forty-ninesubjects completed the BP recording during the PSG and 24-hour AMBP monitoring. Eleven subjects refused to completeBP monitoring during PSG but adequately completed the 24-hour study. The demographic and polysomnographic character-istics are shown in Tables 1 and 2.

Effect of BP Recording on Sleep

The effect of BP recording on sleep efficiency and frequency ofarousals was determined by calculating the percentage of BP

TABLE 1. DEMOGRAPHIC AND POLYSOMNOGRAPHICCHARACTERISTICS OF 60 SUBJECTS WHO COMPLETED24-HOUR AMBULATORY BLOOD PRESSURE RECORDING

24-h AMBP (n � 60)

Group 1 Group 2 Group 3 p Value

n 21 17 22Age, yr 10.1 � 3 10.7 � 4 11.6 � 3 NSMale, % 57 53 68 NSWhite, % 90 53 64 0.03BMI 23 � 8 24 � 8.7 31 � 9 0.01AHI 0.17 � 0.2 2.7 � 0.9 26.8 � 28 � 0.001DI 0.75 � 0.7 1.7 � 1.6 19.9 � 25 0.002Lowest saturation 91 � 2.3 90 � 2.9 80 � 10 � 0.001Maximum

carbon dioxide 49 � 1.9 51 � 3.5 54 � 7.8 0.003Arousal index 11 � 7 10 � 4 25 � 25 0.003

Definition of abbreviations: AHI � apnea–hypopnea index; AMBP � ambulatoryblood pressure; BMI � body mass index; DI � desaturation index; NS � notsignificant.

Group 1 � AHI less than 1 per hour of sleep; Group 2 � AHI 1 to 5 per hourof sleep; Group 3 � AHI greater than 5 per hour of sleep.

952 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

TABLE 2. DEMOGRAPHIC AND POLYSOMNOGRAPHICCHARACTERISTICS OF 49 SUBJECTS WHO COMPLETEDBLOOD PRESSURE RECORDING DURINGPOLYSOMNOGRAPHY FOLLOWED BY 24-HOURAMBULATORY BLOOD PRESSURE MONITORING

24-h AMBP and BP-PSG (n � 49)

Group 1 Group 2 Group 3 p Value

n 14 17 18Age 10.6 � 3 10.7 � 4 12 � 3 NSMale, % 57 58 66 NSWhite, % 86 47 61 NSBMI 23.9 � 8.4 24.1 � 8.7 31.7 � 9 0.03AHI 0.15 � 0.16 2.7 � 0.9 34 � 34 � 0.001DI 0.5 � 0.5 1.7 � 1.6 21 � 27 0.005Lowest saturation 91 � 2.1 90 � 2.9 79 � 11 � 0.001Maximum

carbon dioxide 49 � 2 51 � 3.5 56 � 8 0.01Arousal index 12 � 9 10 � 4 28 � 27 0.004

Definition of abbreviations: AHI � apnea–hypopnea index; AMBP � ambulatoryblood pressure; BMI � body mass index; BP � blood pressure; DI � desaturationindex; NS � not significant; PSG � polysomnography.

Group 1 � AHI less than 1 per hour of sleep; Group 2 � AHI 1 to 5 per hourof sleep; Group 3 � AHI greater than 5 per hour of sleep.

measurements obtained during the PSG that were associatedwith arousals and those associated with full awakening. Forty-one of the 49 subjects who completed BP recording during PSGhad data that could be accurately analyzed. The remaining eightsubjects requested that the BP monitor be placed on the oppositearm from the oximeter; therefore, the exact timing of the mea-surement could not be estimated and the relationship betweencuff inflation and arousals could not be determined. The percent-age of measurements associated with arousals was 18% � 6,16% � 10, and 20% � 11 for Groups 1, 2, and 3, respectively(p � not significant). The percentage of measurements associ-ated with arousals that led to awakenings was 5% � 5, 4% �4, and 4% � 5, respectively. These results show that more than80% of the measurements do not affect sleep and that a verysmall number of measurements lead to full awakening. Totalsleep time was similar for all three groups. Sleep time was 390 �55, 381 � 49, and 391 � 77 minutes for Groups 1, 2, and 3,respectively. Sleep efficiency after sleep onset as a percentageof total sleep time was 89% � 7, 89% � 6, and 85% � 10 forGroups 1, 2, and 3, respectively.

For subjects who completed 24-hour AMBP recording andBP monitoring during PSG, there was no significant differencefound between nocturnal BP obtained in the sleep laboratoryand nocturnal BP obtained at home. Mean sleep systolic BP athome was 104 � 6, 106 � 10, and 107 � 10 mm Hg, whereasmean systolic BP during PSG was 101 � 8, 104 � 10, and 110 �11 mm Hg for Groups 1, 2, and 3, respectively. Mean sleepdiastolic BP at home was 58 � 5, 60 � 6, and 59 � 6 mm Hg,whereas mean sleep diastolic BP during PSG was 58 � 5, 59 �5, and 60 � 8 for Groups 1, 2, and 3, respectively.

Systolic, Diastolic, and Mean Arterial BP

For daytime BP, a significant difference in diastolic BP wasobserved among the three groups with the lowest pressure seenin Group 3. There was no difference measured among groups forsystolic or mean arterial BP. Similarly, there was no significantdifference detected among groups for systolic, diastolic, andmean arterial BP during sleep. A trend toward a higher systolicBP during sleep was observed with increased severity of OSA

TABLE 3. TWENTY-FOUR–HOUR AMBULATORY SYSTOLIC,DIASTOLIC, AND MEAN ARTERIAL BLOOD PRESSURE(mm Hg) OBTAINED FROM 60 SUBJECTS DURINGWAKEFULNESS AND DURING SLEEP*

Group 1 Group 2 Group 3 p Value

Mean wake SBP 115 � 7 116 � 10 116 � 8 NSMean sleep SBP 102 � 6 106 � 10 107 � 9 NSMean wake DBP 70 � 5 70 � 7 66 � 5 0.014Mean sleep DBP 58 � 4 60 � 6 58 � 6 NSMean wake MAP 86 � 4 85 � 6 83 � 4 NSMean sleep MAP 74 � 4 76 � 6 76 � 6 NS% SBP � 95th 22. � 15 21 � 25 26 � 22 NS% DBP � 95th 13 � 10 13 � 15 11 � 14 NS

Definition of abbreviations: AHI � apnea–hypopnea index; DBP � diastolic bloodpressure; % DBP � 95th � percentage of diastolic blood pressure measurementsabove the 95th percentile; MAP � mean arterial pressure; NS � not significant;SBP � systolic blood pressure; % SBP � 95th � percentage of systolic bloodpressure measurements above the 95th percentile.

Group 1 � AHI less than 1 per hour of sleep; Group 2 � AHI 1 to 5 per hourof sleep; Group 3 � AHI greater than 5 per hour of sleep.

* Also shown are the systolic and diastolic BP load expressed as the percentageof BP measurements above the 95th percentile across the three groups.

but did not reach statistical significance (Table 3). When BPindex was used to normalize for the effect of age and growthon BP, a trend similar to the one seen with absolute BP wasobserved (Figure 1). A significant trend for diastolic BP indexduring wakefulness was demonstrated. There was no change inthe trend of systolic and diastolic BP across groups when BPwas expressed as a percentage of the 95th percentile (data notshown). There was also no significant difference in pressure loadfor systolic or diastolic BP among the three groups (Table 3).

BP Variability

A dose-dependent increase in BP variability was observed acrossthe three groups with the largest SD seen in subjects with AHIgreater than 5 per hour of sleep. A significant linear trend wasfound for the variability of awake systolic and mean arterial BPand for sleep systolic, diastolic, and mean BP (Figures 1 and 2).This observation demonstrates that increased BP variability with

Figure 1. Trend for blood pressure (BP) index for systolic and diastolicBP during wakefulness and during sleep across the three groups. Dia-stolic BP index showed a significant trend across the three groups withthe larger difference seen in Group 3 (*p � 0.005). Group 1 � apnea–hypopnea index (AHI) less than 1 per hour of sleep; Group 2 � AHI 1to 5 per hour of sleep; Group 3 � AHI greater than 5 per hour of sleep.SDBP � sleep diastolic blood pressure; SSBP � sleep systolic bloodpressure; WDBP � awake diastolic blood pressure; WSBP � awake sys-tolic blood pressure.

Amin, Carroll, Jeffries, et al.: Blood Pressure in Children with Sleep Apnea 953

Figure 2. BP variability during wakefulness and sleep.(A) Log-transformed SD across the three groups of sys-tolic (**p � 0.01), diastolic, and mean arterial BP (*p �

0.014) during wakefulness. (B ) Log-transformed SDacross the three groups of systolic (**p � 0.01), dia-stolic (**p � 0.01), and mean arterial BP (**p � 0.01)during sleep. Group 1 � AHI less than 1 per hour ofsleep; Group 2 � AHI 1 to 5 per hour of sleep; Group3 � AHI greater than 5 per hour of sleep. WMAP �

awake mean arterial blood pressure; SMAP � sleepmean arterial blood pressure.

an increased severity of OSA is not limited to nighttime BP butpersists during wakefulness.

Nocturnal BP Dipping

A significant linear trend for the difference between awake andsleep BP expressed as a percentage of awake BP was observedfor systolic, diastolic, and mean arterial BP from Groups 1 to 3.The trend from Group 1 to Group 3 was 11.5% � 4.2, 8.6% � 4.9,7.4% � 7.3 for systolic BP (p � 0.01); 17.5% � 5, 14% � 8,11% � 9 for diastolic BP (p � 0.01); and 13.4% � 4.3, 10.6% �5.9, 8.4% � 6.7 for mean arterial pressure (p � 0.002). The ratioof night-to-day systolic BP was 0.88 � 0.04, 0.90 � 0.04, and0.93 � 0.07 for Groups 1, 2, and 3, respectively (p � 0.02).These results indicate that children with OSA have a night-to-day systolic BP ratio that surpasses the cutoff ratio of 0.899 forfemales and 0.9009 for males that is known to increase the riskfor cardiovascular morbidity in adults (14, 15). The differencebetween awake BP and BP during different stages of sleep dem-onstrates that BP dipping during sleep is not stage specific andthat the effect of OSA on nocturnal BP is similar during REMand non–REM sleep (Figure 3).

Bivariate Correlations

Diastolic BP during wakefulness correlated negatively with AHIand DI and positively with Z BMI. BP variability during wake-fulness correlated positively with log AHI, DI, and Z BMI,whereas BP variability during sleep correlated positively with

Figure 3. Difference between mean arterial BP obtained from 24-hourambulatory BP recording during daytime wakefulness, and BP duringwakefulness in bed, and BP during different stages of sleep. A significantlinear trend was observed for the difference in mean arterial BP acrossthe three groups, with the largest difference seen in Group 1 and thesmallest in Group 3 (*p � � 0.05). Group 1 � AHI less than 1 per hourof sleep; Group 2 � AHI 1 to 5 per hour of sleep; Group 3 � AHI greaterthan 5 per hour of sleep. A�W � difference in mean arterial BP whenactive awake and awake in bed; A�2 � difference in mean BP when ac-tive awake and Stage 2 sleep; A�SWS � difference in mean BP whenactive awake and slow wave sleep (SWS); A�REM � difference in meanBP when active awake and REM sleep.

log AHI, maximum end-tidal CO2, and Z BMI. The degree ofnocturnal BP dipping correlated negatively with measures ofseverity of OSA, namely, log AHI, DI, and maximum end-tidalCO2 (Table 4). These results show that there is an associationbetween the degree of BP dysregulation and increased severityof OSA in children with sleep-disordered breathing.

Multiple Regression Analysis

BP variability. Variability of mean arterial pressure duringwakefulness was predicted by a model (p � � 0.0001) thatincluded DI (p � 0.002), Z BMI (p � 0.005), and arousal index(p � 0.006). Variability of mean arterial pressure during sleepwas predicted by a model (p � � 0.0001) that included AHI(p � 0.001) and Z BMI (p � 0.003). Models for BP variability andthe dependent variables remained significant after controlling forethnicity.

Nocturnal dipping of BP. Nocturnal BP dipping was pre-dicted by a model (p � 0.015) that only included DI. The modeland the dependent variable remained significant after controllingfor Z BMI and ethnicity.

These results suggest that sleep-related factors that affect BPvariability during sleep differ from factors affecting variabilityduring wakefulness and that intermittent oxygen desaturationcontributes to BP dysregulation in children with OSA both dur-ing wakefulness and sleep.

Diastolic BP. Diastolic BP index during wakefulness waspredicted by a model (p � � 0.001) that included age(p � � 0.001), AHI (p � 0.014), race (p � 0.03), and Z BMI(p � 0.04).

DISCUSSION

This study demonstrates that children with OSA show evidenceof dysregulation of systemic BP in the form of increased BPvariability and a decreased degree of nocturnal dipping and thatBP dysregulation correlates with measures of severity of OSA.

Previous studies have demonstrated that children with OSAtend to have higher systolic and/or diastolic BP compared withcontrol subjects; however, the prevalence of systemic hyperten-sion remained insignificant between the two groups (27, 32–35).In our study, we have shown that the early stages of BP dysregu-lation in children do not present with sustained elevation of BPbut rather with an alteration in the circadian rhythm of BPprofile and an increase in BP variability. Although this studyconcurs with previous observations as to the lack of significantdifference in the prevalence of hypertension in children withOSA compared with control subjects, it differs regarding thetrend of diastolic BP during sleep and during wakefulness. Mar-cus and coworkers (34) have shown that children with OSAtend to have a higher diastolic BP during sleep compared withchildren with primary snoring. We, on the contrary, showedno difference among groups in diastolic BP during sleep. Thediscrepancy between the two studies raises an important ques-

954 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

TABLE 4. PEARSON CORRELATION BETWEEN BLOOD PRESSURE PARAMETERS ANDPOLYSOMNOGRAPHIC VARIABLES*

Wake Diastolic BP index Wake–Sleep MAP Log SD WMAP Log SD SMAP

Log AHI �0.37 (p � 0.0037) �0.3 (p � 0.019) 0.29 (p � 0.02) 0.47 (p � 0.0002)DI �0.38 (p � 0.0023) �0.31 (p � 0.013) 0.30 (p � 0.01) NSMaximum carbon dioxide NS �0.29 (p � 0.024) NS 0.31 (p � 0.013)Z BMI 0.28 (p � 0.03) NS �0.45 (p � 0.0001) �0.46 (p � 0.0002)

Definition of abbreviations: AHI � apnea–hypopnea index; BMI � body mass index; BP � blood pressure; DI � desaturationindex; NS � not significant; SMAP � sleep mean arterial pressure; Wake–sleep MAP � difference between wake to sleep meanarterial pressure; WMAP � wake mean arterial pressure; Z BMI � Z score for BMI.

* Four BP parameters, namely, diastolic BP index, difference between wake-to-sleep mean arterial BP as a measure of nocturnalBP dipping, and log-transformed SD of awake and asleep BP as a measure of variability were correlated with polysomnographicvariables and Z BMI.

tion as to whether the cardiovascular morbidity of OSA in chil-dren is age dependent, given the younger age of the populationin Marcus’s study. Future studies are needed to determine ifyoung age increases the susceptibility to cardiovascular morbid-ity from OSA. This study is the first to characterize the diurnaland nocturnal BP profile in children with OSA. We have ob-served in this study a lower diastolic BP during wakefulness inchildren with severe OSA compared with children with a milderdegree of sleep apnea and children with simple snoring. Lowdiastolic BP during wakefulness seen in children with more se-vere OSA suggests abnormal elastic recoil of blood vessels dur-ing diastole. This finding, in addition to increased BP variability,could represent the early stages of autonomic and/or endothelialdysfunction in children with OSA (36–38).

In hypertensive and normotensive adults, increased BP vari-ability and decreased nocturnal BP dipping are associated withend-organ damage and increased risk for cardiovascular diseases(1, 2, 11, 12, 14–16, 22). In a prospective study that examinedprognostic significance of BP variability in 1,542 subjects 40 yearsor older, Kikuya and coworkers (39) showed a significant linearrelationship between daytime systolic AMBP variability and rel-ative risk for cardiovascular mortality. In a study of 1,433 sub-jects, the variance of systolic BP was significantly associatedwith coronary artery disease after adjustment for mean systolicpressure (40). The relationship between BP variability and cereb-rovascular diseases has been examined in several studies. Havlikand coworkers (10) demonstrated in a prospective study thatsystolic BP variability in midlife is associated with increased riskfor white matter lesions in the brain later in life and that subjectswho were in the highest category of systolic BP variability hadsignificantly more brain atrophy. The importance of this observa-tion stems from the knowledge that the presence of periventricu-lar white matter lesions, which are thought to be caused bysmall-vessel disease, strongly correlates with progressive loss ofcognitive function (18, 41, 42). Knowledge about the associationbetween BP variability and adverse cardiovascular outcomes inchildren is limited. Because evidence is growing that childrenwith symptoms of obstructive breathing during sleep are at anincreased risk for low academic performance (43–46), furtherstudies are needed to determine whether OSA-associated neuro-cognitive morbidity in children is related to BP dysregulationand small-vessel disease.

Several studies have examined the relationship between 24-hour BP profile and end-organ damage and have demonstratedthat AMBP is superior prognostically to office BP (47, 48). Ver-decchia and coworkers (14, 15) studied a group of 1,187 hyperten-sive subjects and 205 normotensive control subjects and foundthat a night-to-day systolic BP ratio greater than 0.899 for menand greater than 0.9009 for women was associated with signifi-cantly higher risk for cardiovascular events after adjusting for

risk factors and 24-hour systolic BP. In our study, we demonstratethat children with OSA have a higher night-to-day systolic BPratio compared with children with simple snoring. More recently,Hoshide and coworkers (49) examined the association betweennondipping of nocturnal BP and left ventricular geometry innormotensive subjects. They demonstrated that left ventricularmass index and relative wall thickness were greater in nondippersthan in dippers, suggesting that even in the absence of sustaineddiurnal or nocturnal hypertension, nondipping of BP increases therisk for myocardial damage. It is therefore plausible that the BPprofile we described in children with sleep-disordered breathingcould lead to future cardiovascular morbidity.

We observed in this study that increased BP variability insubjects with OSA is present both during sleep and wakefulness,suggesting that the hemodynamic and blood gas abnormalitiesassociated with OSA have a tonic effect on the cardiovascularsystem. In subjects with OSA, the repeated fall in intrathoracicpressure and the marked increase in left ventricular transmuralpressure and afterload contribute largely to BP variability duringsleep. In addition to the mechanical effects of OSA on the sys-temic circulation, there is also repeated activation of the centraland peripheral chemoreceptors, which elicit both hyperventila-tion and increased sympathetic traffic to peripheral blood vessels.Simultaneously, the enhanced chemoreceptor response elicitsan inhibitory response from the baroreceptors and pulmonarystretch receptors. Therefore, the final BP response and variabil-ity are the product of this intricate interaction among the chemo-receptors, baroreceptors, pulmonary stretch receptors, and themechanical effect of OSA on cardiac dynamics and function (50,51). This study has shown that the AHI is a strong predictor ofBP variability during sleep, whereas the frequency of oxygendesaturation is a strong predictor for daytime variability. A plau-sible explanation of these findings is that during wakefulness,when the mechanical effect of OSA is not affecting BP variabil-ity, there is a residual perturbing effect of intermittent hypoxiaon the balance between chemoreceptor and baroreceptor func-tion. We have also demonstrated that the DI is the best predictorof the degree of BP dipping. This suggests that intermittenthypoxia plays an important role in BP dysregulation in childrenwith OSA.

This study shows the independent effects of obesity and OSAon BP variability. This study has shown that BMI contributesto increased BP variability both during wakefulness and sleep.Studies that examined the changes in the autonomic nervoussystem among obese subjects showed that obesity is associatedwith decreased sympathetic and parasympathetic tone, alteredratio of sympathetic to parasympathetic tone, and decreasedbaroreceptor sensitivity. Both the duration of obesity and thefat distribution contribute to these autonomic changes (52–56).It is likely that obesity and OSA contribute to dysregulation of

Amin, Carroll, Jeffries, et al.: Blood Pressure in Children with Sleep Apnea 955

BP by different mechanisms and independently contribute tocardiovascular morbidity.

There are several limitations to this study. The control groupconsisted of children with primary snoring. It is possible thatwith normal control subjects a significant trend for mean BPacross the three groups would have been identified. It is alsolikely that the small sample size provided limited power to detecta significant difference in nocturnal BP among the three groups.Although this study showed an independent effect of OSA onBP regulation, it does not address the possibility that OSA mighthave a differential effect on BP control in lean versus obesechildren.

In summary, this study shows that children with OSA haveevidence of BP dysregulation that manifests during sleep andwakefulness. These findings could represent the early stagesof BP dysregulation, which could ultimately lead to sustainedhypertension if OSA is left untreated or if additional risk factorsfor hypertension develop over time. Lack of nocturnal dippingand increased BP variability may also be useful in determininga subset of patients for whom more aggressive management ofOSA is indicated.

Conflict of Interest Statement : R.S.A. has no declared conflict of interest; J.L.C. hasno declared conflict of interest; J.L.J. has no declared conflict of interest; C.G. has nodeclared conflict of interest; J.A.B. has no declared conflict of interest; B.C. hasno declared conflict of interest; R.B. has given lectures for Astra-Zeneca over thelast three years with payment �$5,000; S.R.D. has no declared conflict of interest.

References

1. Shahar E, Whitney CW, Redline S, Lee ET, Newman AB, Javier Nieto F,O’Connor GT, Boland LL, Schwartz JE, Samet JM. Sleep-disorderedbreathing and cardiovascular disease: cross-sectional results of theSleep Heart Health Study. Am J Respir Crit Care Med 2001;163:19–25.

2. Silvestrini M, Rizzato B, Placidi F, Baruffaldi R, Bianconi A, DiomediM. Carotid artery wall thickness in patients with obstructive sleepapnea syndrome. Stroke 2002;33:1782–1785.

3. Schafer H, Koehler U, Ewig S, Hasper E, Tasci S, Luderitz B. Obstructivesleep apnea as a risk marker in coronary artery disease. Cardiology1999;92:79–84.

4. Hla KM, Skatrud JB, Finn L, Palta M, Young T. The effect of correctionof sleep-disordered breathing on BP in untreated hypertension. Chest2002;122:1125–1132.

5. Peppard PE, Young T, Palta M, Skatrud J. Prospective study of theassociation between sleep-disordered breathing and hypertension.N Engl J Med 2000;342:1378–1384.

6. Nieto FJ, Young TB, Lind BK, Shahar E, Samet JM, Redline S, D’Agos-tino RB, Newman AB, Lebowitz MD, Pickering TG. Association ofsleep-disordered breathing, sleep apnea, and hypertension in a largecommunity-based study: Sleep Heart Health Study. JAMA 2000;283:1829–1836.

7. Davrath LR, Goren Y, Pinhas I, Toledo E, Akselrod S. Early autonomicmalfunction in normotensive individuals with a genetic predispositionto essential hypertension. Am J Physiol Heart Circ Physiol 2003;285:H1697–H1704.

8. Lucini D, Mela GS, Malliani A, Pagani M. Impairment in cardiac auto-nomic regulation preceding arterial hypertension in humans: insightsfrom spectral analysis of beat-by-beat cardiovascular variability. Circu-lation 2002;106:2673–2679.

9. Liu M, Takahashi H, Morita Y, Maruyama S, Mizuno M, Yuzawa Y,Watanabe M, Toriyama T, Kawahara H, Matsuo S. Non-dipping is apotent predictor of cardiovascular mortality and is associated withautonomic dysfunction in haemodialysis patients. Nephrol Dial Trans-plant 2003;18:563–569.

10. Havlik RJ, Foley DJ, Sayer B, Masaki K, White L, Launer LJ. Variabilityin midlife systolic blood pressure is related to late-life brain whitematter lesions: the Honolulu-Asia aging study. Stroke 2002;33:26–30.

11. Staessen JA, Asmar R, De Buyzere M, Imai Y, Parati G, Shimada K,Stergiou G, Redon J, Verdecchia P. Task force II: blood pressuremeasurement and cardiovascular outcome. Blood Press Monit 2001;6:355–370.

12. Tozawa M, Iseki K, Yoshi S, Fukiyama K. Blood pressure variability asan adverse prognostic risk factor in end-stage renal disease. NephrolDial Transplant 1999;14:1976–1981.

13. Goldstein IB, Bartzokis G, Hance DB, Shapiro D. Relationship betweenblood pressure and subcortical lesions in healthy elderly people. Stroke1998;29:765–772.

14. Verdecchia P, Clement D, Fagard R, Palatini P, Parati G. Blood pressuremonitoring: Task force III: target-organ damage, morbidity and mor-tality. Blood Press Monit 1999;4:303–317.

15. Verdecchia P, Schillaci G, Borgioni C, Ciucci A, Gattobigio R, GuerrieriM, Comparato E, Benemio G, Porcellati C. Altered circadian bloodpressure profile and prognosis. Blood Press Monit 1997;2:347–352.

16. Sega R, Corrao G, Bombelli M, Beltrame L, Facchetti R, Grassi G,Ferrario M, Mancia G. Blood pressure variability and organ damagein a general population: results from the PAMELA study (PressioniArteriose Monitorate E Loro Associazioni). Hypertension 2002;39:710–714.

17. Amin RS, Kimball TR, Bean JA, Jeffries JL, Willging JP, Cotton RT,Witt SA, Glascock BJ, Daniels SR. Left ventricular hypertrophy andabnormal ventricular geometry in children and adolescents with ob-structive sleep apnea. Am J Respir Crit Care Med 2002;165:1395–1399.

18. De Groot JC, De Leeuw FE, Oudkerk M, Van Gijn J, Hofman A, JollesJ, Breteler MM. Periventricular cerebral white matter lesions predictrate of cognitive decline. Ann Neurol 2002;52:335–341.

19. Standards and indications for cardiopulmonary sleep studies in chil-dren: American Thoracic Society. Am J Respir Crit Care Med 1996;153:866–878.

20. Cardiorespiratory sleep studies in children: establishment of normativedata and polysomnographic predictors of morbidity: American Tho-racic Society. Am J Respir Crit Care Med 1999;160:1381–1387.

21. Rechtschaffen A, Kales A. editors. A manual of standardized terminol-ogy: techniques and scoring systems for sleep stages of human subjects.UCLA Brain Information Service. Los Angeles: Brain Research Insti-tute; 1968.

22. Sleep Disorders Atlas Task Force: C. Guilleminault, Chairman. Sleep1992;15:173–184.

23. Sun M, Tien J, Jones R, Ward R. A new approach to reproducibilityassessment: clinical evaluation of SpaceLabs medical oscillometricblood pressure monitor. Biomed Instrum Technol 1996;30:439–448.

24. Sulbaran TA, Silva Rondon E. Normal values during ambulatory bloodpressure monitoring in male adolescents. Invest Clin 1997;38:55–63.

25. Schillaci G, Verdecchia P, Zampi I, Battistelli M, Bartoccini C, PorcellatiC. Non-invasive ambulatory BP monitoring during the night: randomi-sed comparison of different reading intervals. J Hum Hypertens 1994;8:23–27.

26. Hietanen E, Wendelin-Saarenhovi M. Ambulatory blood pressure repro-ducibility and application of the method in a healthy Finnish cohort.Scand J Clin Lab Invest 1996;56:471–480.

27. Enright PL, Goodwin JL, Sherrill DL, Quan JR, Quan SF. Blood pressureelevation associated with sleep-related breathing disorder in a commu-nity sample of white and Hispanic children: the Tucson Children’sAssessment of Sleep Apnea Study. Arch Pediatr Adolesc Med 2003;157:901–904.

28. Dimsdale JE, von Kanel R, Profant J, Nelesen R, Ancoli-Israel S, ZieglerM. Reliability of nocturnal blood pressure dipping. Blood Press Monit2000;5:217–221.

29. National High Blood Pressure Education Program Working Group onHypertension Control in Children and Adolescents. Update on the1987 task force report on high blood pressure in children and adoles-cents: a working group report from the National High Blood PressureEducation Program. Pediatrics 1996;98:649–658.

30. Rosner B, Prineas RJ, Loggie JM, Daniels SR. Blood pressure nomo-grams for children and adolescents, by height, sex, and age, in the UnitedStates. J Pediatr 1993;123:871–886.

31. Ogden CL, Kuczmarski RJ, Flegal KM, Mei Z, Guo S, Wei R, Grummer-Strawn LM, Curtin LR, Roche AF, Johnson CL. Centers for DiseaseControl and Prevention 2000 growth charts for the United States:improvements to the 1977 National Center for Health Statistics ver-sion. Pediatrics 2002;109:45–60.

32. Guilleminault C, Eldridge FL, Simmons FB, Dement WC. Sleep apneain eight children. Pediatrics 1976;58:23–30.

33. Shiomi T, Guilleminault C, Stoohs R, Schnittger I. Obstructed breathingin children during sleep monitored by echocardiography. Acta Paediatr1993;82:863–871.

34. Marcus CL, Greene MG, Carroll JL. Blood pressure in children with ob-structive sleep apnea. Am J Respir Crit Care Med 1998;157:1098–1103.

35. Kohyama J, Ohinata JS, Hasegawa T. Blood pressure in sleep disorderedbreathing. Arch Dis Child 2003;88:139–142.

36. Imadojemu VA, Sinoway LI, Leuenberger UA. Vascular dysfunction in

956 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

sleep apnea: a reversible link to cardiovascular disease? Am J RespirCrit Care Med 2004;169:328–329.

37. Imadojemu VA, Gleeson K, Quraishi SA, Kunselman AR, Sinoway LI,Leuenberger UA. Impaired vasodilator responses in obstructive sleepapnea are improved with continuous positive airway pressure therapy.Am J Respir Crit Care Med 2002;165:950–953.

38. Imadojemu VA, Gleeson K, Gray KS, Sinoway LI, Leuenberger UA.Obstructive apnea during sleep is associated with peripheral vasocon-striction. Am J Respir Crit Care Med 2002;165:61–66.

39. Kikuya M, Hozawa A, Ohokubo T, Tsuji I, Michimata M, MatsubaraM, Ota M, Nagai K, Araki T, Satoh H, et al. Prognostic significanceof blood pressure and heart rate variabilities: the Ohasama study.Hypertension 2000;36:901–906.

40. Grove JS, Reed DM, Yano K, Hwang LJ. Variability in systolic bloodpressure: a risk factor for coronary heart disease? Am J Epidemiol1997;145:771–776.

41. de Leeuw FE, de Groot JC, Oudkerk M, Witteman JC, Hofman A, vanGijn J, Breteler MM. Hypertension and cerebral white matter lesionsin a prospective cohort study. Brain 2002;125:765–772.

42. de Groot JC, de Leeuw FE, Oudkerk M, Hofman A, Jolles J, BretelerMM. Cerebral white matter lesions and subjective cognitive dysfunc-tion: the Rotterdam Scan Study. Neurology 2001;56:1539–1545.

43. Gozal D. Sleep-disordered breathing and school performance in children.Pediatrics 1998;102:616–620.

44. Gozal D, Pope DW Jr. Snoring during early childhood and academicperformance at ages thirteen to fourteen years. Pediatrics 2001;107:1394–1399.

45. Peker Y, Hedner J, Norum J, Kraiczi H, Carlson J. Increased incidenceof cardiovascular disease in middle-aged men with obstructive sleepapnea: a 7-year follow-up. Am J Respir Crit Care Med 2002;166:159–165.

46. Urschitz MS, Guenther A, Eggebrecht E, Wolff J, Urschitz-Duprat PM,Schlaud M, Poets CF. Snoring, intermittent hypoxia and academicperformance in primary school children. Am J Respir Crit Care Med2003;168:464–468.

47. Perloff D, Sokolow M, Cowan R. The prognostic value of ambulatoryblood pressures. JAMA 1983;249:2792–2798.

48. Perloff D, Sokolow M, Cowan RM, Juster RP. Prognostic value of ambu-latory blood pressure measurements: further analyses. J HypertensSuppl 1989;7:S3–S10.

49. Hoshide S, Kario K, Hoshide Y, Umeda Y, Hashimoto T, Kunii O, OjimaT, Shimada K. Associations between nondipping of nocturnal bloodpressure decrease and cardiovascular target organ damage in strictlyselected community-dwelling normotensives. Am J Hypertens 2003;16:434–438.

50. Narkiewicz K, Somers VK. Sympathetic nerve activity in obstructivesleep apnoea. Acta Physiol Scand 2003;177:385–390.

51. Kara T, Narkiewicz K, Somers VK. Chemoreflexes: physiology and clini-cal implications. Acta Physiol Scand 2003;177:377–384.

52. Martini G, Riva P, Rabbia F, Molini V, Ferrero GB, Cerutti F, CarraR, Veglio F. Heart rate variability in childhood obesity. Clin AutonRes 2001;11:87–91.

53. Nagai N, Matsumoto T, Kita H, Moritani T. Autonomic nervous systemactivity and the state and development of obesity in Japanese schoolchildren. Obes Res 2003;11:25–32.

54. Laederach-Hofmann K, Mussgay L, Ruddel H. Autonomic cardiovascu-lar regulation in obesity. J Endocrinol 2000;164:59–66.

55. Beske SD, Alvarez GE, Ballard TP, Davy KP. Reduced cardiovagalbaroreflex gain in visceral obesity: implications for the metabolic syn-drome. Am J Physiol Heart Circ Physiol 2002;282:H630–H635.

56. Rabbia F, Silke B, Conterno A, Grosso T, De Vito B, Rabbone I, Chian-dussi L, Veglio F. Assessment of cardiac autonomic modulation duringadolescent obesity. Obes Res 2003;11:541–548.

Pressure–Volume Curve Does Not Predict Steady-StateLung Volume in Canine Lavage Lung InjuryJohn M. Downie, Arthur J. Nam, and Brett A. Simon

Department of Anesthesiology and Critical Care Medicine, Johns Hopkins Medical Institutions, Baltimore, Maryland

To better understand strategies for recruiting and maintaininglung volume in acute lung injury, we examined relationships be-tween steady-state lung volume and cumulative cyclic recruitment/derecruitment volume history and the quasi-static pressure–volumecurve, in an animal saline lavage lung injury model. Small-volumetidal pressure–volume loops performed after inflation from func-tional residual capacity demonstrated incremental, cyclic recruit-ment only if the peak pressure achieved exceeded the pressure atwhich the compliance increased (Pflex) on the pressure–volumecurve, whereas loops performed after deflation from total lungcapacity remained close to the envelope deflation curve. Recruit-ment continued to occur up to and beyond a peak inspiratory airwaypressure of 40 cm H2O, as demonstrated by both the tidal loopsand by computed tomography-derived lung volume data. Tidal-specific compliance was relatively constant across positive end-expiratory pressure levels after inflation from functional residualcapacity, but peaked at moderate positive end-expiratory pressureafter deflation from total lung capacity, further demonstrating theeffects of volume history and providing experimental validation ofthe recruitment models of Hickling (AJRCCM 2001;163:69–78). Theseresults support the interpretation of Pflex as pressure threshold forrecruitment, but otherwise do not suggest a role for the pressure–volume curve in predicting steady-state lung volume.

Keywords: acute lung injury; computed tomography; mechanical venti-lation; surfactant

Recruitment and maintenance of ventilated lung volume areessential for improving oxygenation in acute lung injury andthe acute respiratory distress syndrome (ARDS) (1). The mostcommonly employed strategy to achieve this end is the use ofpositive end-expiratory pressure (PEEP), with or without theaddition of sighs or periodic high inflation pressure recruitmentmaneuvers. Significant controversy persists over the optimalmethod for determining the amount of PEEP to apply. Somehave suggested that PEEP be set on the basis of parametersderived from an inspiratory pressure–volume (P–V) curve andthe presence of a slope change (Pflex) therein (2, 3), whereasthe scheme used in the National Institutes of Health ARDSNetwork low Vt ventilation trial was a simple PEEP–FiO2 (frac-tional inspired oxygen concentration) ladder (4). Others haveadvocated finding optimal PEEP by reducing PEEP in stepsfrom a high level until the arterial partial pressure of oxygen

(Received in original form May 6, 2003; accepted in final form February 2, 2004)

Supported by National Institutes of Health grant HL58504.

Present affiliation of J.M.D.: Department of Pediatrics, University of Chicago, Chi-cago, Illinois.

Correspondence and requests for reprints should be addressed to Brett A. Simon,M.D., Ph.D., Department of Anesthesia, Tower 711, Johns Hopkins Hospital, Balti-more, MD 21287-8711. E-mail: bsimon@jhmi.edu

This article has an online supplement, which is accessible from this issue’s tableof contents online at http://www.atsjournals.org

Am J Respir Crit Care Med Vol 169. pp 957–962, 2004Originally Published in Press as DOI: 10.1164/rccm.200305-614OC on February 5, 2004Internet address: www.atsjournals.org

(PaO2) falls below a prescribed level, and then rerecruiting thelung and adjusting PEEP to a level just above that at which thefall occurred (5), or finding the PEEP that yields the maximaltidal compliance after a recruiting maneuver (6).

Originally, the rationale behind the use of the inspiratoryP–V curve to guide ventilator settings was that the shape of thecurve, in particular the increasing slope or compliance abovePflex, reflected the recruitment of poorly ventilated lung regionsand that the flattening of the curve at higher pressures reflectedoverdistension (2, 3). More recent data suggest that recruitmentactually occurs progressively and continuously along the curveabove Pflex (5, 7–10). However, animal studies and models ofacute lung injury have demonstrated that volume history is animportant determinant of end-expiratory volume and, further,that steady-state lung volume during tidal ventilation may notreflect that seen in a static or quasi-static P–V curve (5, 6).Therefore, we sought to examine the relationship betweensteady-state lung volume and cumulative cyclic recruitment orderecruitment to the quasi-static P–V curve, using the salinelavage model of acute lung injury in anesthetized dogs. Welooked at changes in end-expiratory volume over short timeperiods, as functions of volume history and end-expiratory pres-sure, by monitoring the cumulative changes in lung volume ofsmall tidal loops. We also compared lung volume during steady-state mechanical ventilation, measured by computed tomogra-phy (CT) imaging, with the volume predicted by the P–V curve.Some of the results of these studies have been previously re-ported in abstract form (11, 12).

METHODS

Procedures were approved by the Institutional Animal Care and UseCommittee (see the online supplement for additional details). Six dogswere anesthetized, intubated, and volume ventilated at an FiO2 of 1.0and a rate of 15 breaths/minute. Vts (10–12 ml/kg) were adjusted toan end-tidal Pco2 (PetCO2) of 30–35 mm Hg and maintained thereafter.Airway pressure (Paw), esophageal pressure (Pes), arterial blood pres-sure (Pa), PetCO2, and oxygen saturation (SaO2) were continuously mea-sured and recorded.

Lung Injury

Acute lung injury was induced by repeated lavage with warmed saline(60 ml/kg), repeated every 10 minutes while shifting each dog from theprone to the supine position, until the PaO2 fell below 90 mm Hg for30 minutes.

P–V Curves

A computer-controlled system (see the online supplement) generatedquasi-static P–V curves at constant flow (3 L/minute) over a prepro-grammed range of pressures. Paw, Pes, and flow signals were digitizedand stored, allowing measurement of cumulative lung volume changefrom FRC over several cycles. The same P–V curve series were obtainedbefore and after lavage injury. Vital capacity P–V curves were measuredfrom 0 to 40 to 0 cm H2O Paw for three cycles. Curves were fit to themodel of Venegas and coworkers (13), using the point of maximalcompliance increase to define Pflex. Two series of tidal P–V loops withdifferent volume histories were performed. Series 1 recorded three10-cm H2O amplitude tidal P–V loops over different pressure rangesafter inflation from FRC, whereas Series 2 recorded the same series of

958 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

loops after deflation from total lung capacity (TLC, defined as 40 cmH2O Paw). For both, loops were performed over incrementally increas-ing pressures from 0 to 40 cm H2O (i.e., 0–10, 5–15, 10–20…30–40 cmH2O). Five to 10 minutes of tidal ventilation was resumed betweenmeasurements, until PetCO2 returned to normal.

CT Imaging

Five dogs were prepared as described above and underwent supinewhole-lung CT imaging before and after lavage for measurement ofsteady-state air volumes over a range of PEEP values. Contiguous10-mm images were obtained with a GE 9800 scanner (General Electric,Fairfield, CT) during steady-state mechanical ventilation gated to endexpiration, one image per breath, with incremental table movementbetween images. Ventilation was not held or interrupted for imaging.Imaging was repeated at 0 (FRC), 5, 10, 15, and 20 cm H2O PEEP,with a 10-minute equilibration period after each PEEP change. Vitalcapacity quasi-static P–V curves were obtained as described abovebefore imaging.

Images were analyzed using NIH Image (http://rsb.info.nih.gov/nih-image) on a Macintosh computer. Images were calibrated against themeasured density of air and tissue from each image set to quantifydensity as “percent air.” The lung tissue in each slice was manuallyoutlined and the air and tissue volumes were calculated from the productof lung area, slice thickness, and density. The individual slice volumeswere summed to give total, tissue, and air volume, which was plottedagainst PEEP to generate “CT P–V curves” representing the end-expiratory volumes of the lung during steady-state ventilation. The airCT P–V curve was plotted alongside the quasi-static curve, assumingthat the CT measured lung volume at FRC corresponded to the initiallung volume for the quasi-static P–V curve.

Statistical Analysis

Values are reported as means � SEM. A paired Student’s t-test wasused to assess differences in �V and specific compliance (sC) (StatView4.5; SAS Institute, Cary, NC). Statistical significance was accepted atp � 0.05.

RESULTS

Lung lavage resulted in a significant decrease in oxygenation,with arterial Po2 falling from 496 � 24 mm Hg in control animalsto 85 � 5 mm Hg 30 minutes after lavage. The control vitalcapacity P–V curves exhibited the classic sigmoid shape withminimal hysteresis. Whereas the first inspiratory curve fre-quently showed lower volumes and initial compliance, particu-larly at lower pressures, than subsequent loops, the descendingcurve and the second and third loops were generally superimpos-able (Figure 1, top). After lavage, the curves exhibited increasedhysteresis and a change in compliance in the midinflation limb,allowing identification of a clear Pflex (Figure 1, bottom). Pflexwas evident whether the Paw (19.9 � 2.7 cm H2O) or transpulmo-nary pressure (Ptp) (17.8 � 2.6 cm H2O) pressure was used.

Figure 1. Vital capacity pressure–volume(P–V) curves from representative animalbefore and after saline lavage injury. Totallung capacity was defined at an airwaypressure of 40 cm H2O, which resulted ina slightly lower maximum transpulmonarypressure (Ptp).

Frequently, the first inflation loop was different from the subse-quent two, suggesting some recruitment attained in the first loopwas maintained for subsequent loops. Vital capacity, defined asthe volume of the P–V curve from 0 to 40 cm H2O Paw, fell17 � 4% after lavage (mean control VC was 1.39 � 0.16 L).

In the control series, tidal loops from FRC were narrow andremained close to the envelope P–V curve except over the upperthird of the pressure range, where they became more parallelto the pressure axis. On deflation from TLC the tidal loops wereclose to the deflation limb P–V curve over the entire pressurerange (Figure 2). Tidal loops after lavage changed considerablyfrom control. Loops from FRC tended to remain close to theascending P–V curve at low pressures with a small amount oftidal recruitment. However, when the peak tidal loop pressureexceeded Pflex, the tidal recruitment per loop greatly increased(Figure 3). Loops performed after deflation from TLC hadgreater hysteresis area than controls and were generally parallelto the deflation P–V curve, although at the lowest pressure rangesthe slope of the loops dropped off (Figure 3).

Tidal recruitment (�V) during tidal P–V loops was definedas half the increase in end-expiratory volume over the secondand third tidal loops (Figure 4). This definition avoids includingvolume changes from deflation hysteresis. Loop compliance wasdetermined from the difference between the starting (end-expir-atory) and peak (end-inspiratory) lung volumes during the sec-ond tidal loop, divided by the 10-cm H2O pressure difference.Loop compliance was normalized to control vital capacity foreach animal to give specific compliance (sC) to facilitate combin-ing results from different-sized animals for statistical analysis.

Quantification of tidal recruitment (Figure 5) shows that,after inflation from FRC, the volume recruited in the lavagedlungs becomes significantly greater than in the control lungs(p � 0.05) when the peak loop airway pressure exceeds Pflex(Figure 5). In fact, at the lowest pressure range control recruit-ment was significantly greater than lavage recruitment, reflectingthe higher opening pressures needed to recruit the lavaged lung.Deflation limb recruitment was less than inflation limb recruit-ment, and was significantly different between control and lavageat the lowest pressure range (control greater) and the highestpressure range (lavage greater) (Figure 5).

Specific compliance of control tidal loops was maximal at lowpressure ranges and decreased as pressure increased for bothinflation and deflation series (Figure 6). In contrast, sC in thelavage lung after inflation from FRC was low and changed onlyslightly over the entire range. After deflation from TLC, how-ever, sC changed twofold over the range of pressures, reachinga maximum at a PEEP of 10 cm H2O (Figure 6). Note thesedata are plotted against minimum loop pressure (PEEP) forcomparison with the literature (see below); midcycle Paw andpeak Paw are 5 and 10 cm H2O greater, respectively.

Figure 2. Composite tidal P–V loops from a representative control con-dition after inflation from FRC (A ) or deflation from TLC (B ). Individualloops for each separate tidal P–V run with their volume histories aresuperimposed on the same axes.

Downie, Nam, and Simon: Tidal Pressure–Volume Curves 959

Figure 3. Composite tidal P–V loops for six animals postlavage lung injury after inflation from FRC or deflation from TLC. Individual loops for eachseparate tidal P–V run with their volume histories are superimposed on the same axes.

Quasi-static and CT P–V curves for five dogs were normalizedto each lung’s quasi-static curve TLC, averaged, and plotted onthe same axes with the assumption that the lung volumes wereequal at FRC (Figure 7). Pflex, obtainable for four of the fivequasi-static P–V curves, was equal to 15.3 � 2.4 cm H2O. Peakinspiratory airway pressures during CT imaging averaged 19 �1.8, 20 � 2.3, 24 � 1.5, 33 � 3.2, and 44 � 2.9 at 0, 5, 10, 15,and 20 cm H2O PEEP, respectively, at an average Vt of 260 �14.1 ml. The steady-state end-expiratory lung volume, as definedby the CT data, was not related in any obvious way to the quasi-static P–V curve. Lung volume was underestimated at all PEEPlevels by the inflation P–V curve. The deflation P–V curve over-

Figure 4. Example of three tidal P–V loops over the airway pressurerange of 15 to 25 cm H2O after inflation from FRC. �V � tidal recruitmentper breath measured over second and third tidal loops; PEEP � positiveend-expiratory pressure; Ppeak � peak airway pressure.

estimated lung volume at low PEEP and underestimated it athigh PEEP. It is important to note that at the 10- to 12-ml/kgVts used for these CT studies, peak inspiratory pressures weregreater than the 10-cm H2O increments used in the tidal loopstudies, exceeding 40 cm H2O at the highest PEEP level. Themean end-inspiratory pressures and volumes during CT imagingat each PEEP, estimated by adding the peak inspiratory pressureto the PEEP and the set Vt to the CT lung air volume, arepresented as open circles in the same figure (Figure 7). Thesetidal P–V trajectories appear to bear little relationship to thequasi-static P–V curve.

To present the air/tissue-partitioned CT P–V curves, the data

Figure 5. Tidal recruitment perbreath (�V) data from all six ani-mals before (solid columns) andafter (gray columns) lavage lung in-jury for tidal loops performed afterinflation from FRC or deflationfrom TLC. Data are plotted againstPpeak for the cycle to highlight therole of Pflex as a pressure thresholdfor recruitment. Pflex was deter-mined by curve fitting each ani-mal’s vital capacity P–V curve, usingthe method of Venegas and co-workers (13). Data representmeans � SEM; *p � 0.05, injuredversus control.

960 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

Figure 6. Specific compliance (sC)data for tidal loops from all animalsbefore (control) and after lavage lunginjury (lavage) for tidal loops per-formed after inflation from FRC ordeflation from TLC. These data areplotted as functions of minimumloop pressure (PEEP) for comparisonwith the model data of Hickling (6).Data represent means � SEM.

for each animal were normalized to the total (air plus tissue)volume at 20 cm H2O PEEP in the control condition, and thenaveraged (Figure 8). Because this total lung volume is greaterthan the TLC air volume used to normalize the data in Figure 7,the resulting normalized air volumes in Figure 8 are lower butallow direct comparison of how the air and tissue volume compo-nents change with injury and incremental PEEP. Tissue volumeat FRC increased from 12.4 � 1.0% under control conditionsto 22.5 � 1.5% after lavage, and the tissue volume changed lessthan 1% over the range of PEEPs, indicating minimal changein intrapulmonary blood volume and/or clearance of edema fluidfrom the lung with increasing PEEP. Total lung volume wasunchanged after lavage, suggesting that the loss of air volumewith injury was due to flooding rather than collapse or atelectasisbecause the reduction in air volume was offset by an equalincrease in tissue volume.

DISCUSSION

The optimal management of mechanical ventilation during acutelung injury depends on the recruitment and maintenance ofventilated lung volume, balancing end-inspiratory recruitmentwith overdistension while also balancing end-expiratory derecruit-ment with the reduced Vts and hemodynamic consequences ofPEEP. The changed shape of the inspiratory P–V curve in lunginjury, with an increased and linear compliance between a lowerand upper Pflex, was interpreted to represent a range of pressuresover which the lung could be ventilated without derecruitmentor overdistension (2, 3). Many subsequent animal (5) and human(8–10) studies have suggested that recruitment occurs progres-sively and continuously above Pflex and, further, that steady-state lung volume during tidal ventilation may not reflect thatseen in the P–V curve. Hickling used a simple mathematicalmodel of the ARDS lung to elegantly show how this could be

Figure 7. Quasi-static (solid line)and CT-derived (solid circles) P–Vcurves for five lavage-injureddogs, superimposed with the as-sumption that volumes are equalat FRC. Open circles estimate end-inspiratory pressure and volumeduring CT imagingbased on peakinspiratory pressure and set VT.Data represent means � SEM.

Figure 8. Whole-lung CTP–V curves partitionedinto total, air, and “tis-sue” volumes for the fiveanimals of Figure 7 beforeand after lavage injury.Data represent means �

SEM.

so (7), and then went on to predict that tidal compliance andend-expiratory volume will depend on both the volume historyand Vt applied (6). Our data provide experimental confirmationof these model predictions of Hickling, emphasizing the impor-tance of short-term volume history in the recruitment behaviorof the lung and the interpretation of Pflex as a threshold forend-inspiratory tidal recruitment. These data further indicatethat the steady-state end-expiratory lung volume is not predictedby the quasi-static P–V curve.

Experimental Considerations

We chose the saline lavage model of acute lung injury for thesestudies, a commonly employed model of surfactant depletion andalveolar edema known for its consistency and stability (14, 15).Whole lung P–V curves performed before and after our 1- to2-hour tidal curve protocol were minimally changed, demonstra-ting the stability of the preparation despite the large Vt ventila-tion and repeated vital capacity inflations. This stability is evidentin the overlapping inflation and deflation P–V trajectories seenas superimposed volume histories for the repeated individualP–V loop measurements in Figure 3. Compared with other lunginjury models such as oleic acid, the lavage model may be consid-ered to be highly recruitable, achieving close to full preinjurylung volume at TLC (12, 16). Although surfactant inactivationis an important component of lung mechanical dysfunction inclinical acute lung injury (17), the relevance of these results to thehuman condition may vary depending on the degree of mechanicalsimilarity of an individual’s lung injury to this model.

Obtaining safe, consistent and repeatable P–V curves in in-jured lungs is difficult (18), and many attempts have been madeto automate this process to improve speed, safety, and reliability(19–22). Although resistive losses may cause small differencesbetween results obtained by static methods, in which the lungis permitted to reach an equilibrium pressure at each volumestep, and quasi-static methods, in which pressure and volume aremeasured continuously during a steady inflation and deflation ofthe lung, these differences are minimal at low flow rates and aremore than offset by the technical difficulties in determining sta-ble plateau pressures in the injured lung. At a steady flow of3 L/minute, neither expiratory flow limitation nor auto-PEEPshould have been significant (23–25). Our computer-controlledsystem (see the online supplement) has the advantage of con-tinuously tracking cumulative changes in volume over severalprogrammed inflation/deflation cycles, allowing measurement ofincremental recruitment. The system was tested for leaks byconfirming conservation of volume with repeated cycling of arubber bag (see Figure E1 in the online supplement) and byclamping the endotracheal tube at end inspiration and checkingfor maintenance of airway pressure for 20 seconds. We did notcorrect for differences between O2 consumption and CO2 pro-duction, which we estimate to be less than 2.3 ml for a tidal loopand 20 ml for a vital capacity curve (see the online supplement for

Downie, Nam, and Simon: Tidal Pressure–Volume Curves 961

details). We did, however, correct for gas compression, becausemeasurements are made over a wide range of airway pressures.Finally, we chose to use the average recruitment over the secondand third tidal loops as our measure of recruitment, avoidingattributing the difference in inflation and deflation limb hystere-sis volume to Vt recruitment. We limited the tidal P–V dataacquisition to three cycles as a trade-off between reasonabletime for each protocol, potential hypoxemia and hypercarbiawith reduced ventilation in the injured lung, and additional infor-mation available from added cycles. In pilot studies we foundthat although inspiratory tidal recruitment frequently continuedbeyond 3 cycles, it generally reached a plateau after 6–10 cycles.

Vital Capacity P–V Curves

Vital capacity P–V curves showed the expected change in shapeafter lung injury, with moderate reduction in vital capacity, in-creased hysteresis, and a change in slope in the midinflation limbidentified as Pflex (Figure 1). Of note, after disconnecting thecircuit and deflation to FRC there was commonly a differencebetween the first and subsequent inflation curves under controlconditions, demonstrating that even in the normal lung therecan be some lung derecruitment, units with elevated openingpressures that recruit and remain open with a subsequent defla-tion to zero pressure. This dependence on volume history, evenin uninjured lungs, underscores the importance of standardizedconditions in interpreting P–V data in the clinical setting. Weinflated the lungs to a maximum Paw of 40 cm H2O, correspond-ing to a Ptp of 33–36 cm H2O, and although there was frequentlya slight decrease in slope at the top of the inflation curve(Figure 3) no upper Pflex or plateau was identified. Although aplateau may have been identified with inflation to higher pres-sures, we choose a maximum Paw of 40 cm H2O as a compromiseto preserve the stability of the preparation.

Tidal P–V Loops

Control tidal P–V loops from FRC remained parallel to theinflation P–V curve until a relatively high peak lung volumewas attained (Figure 2). Because there appeared to be minimalrecruitment with each cycle, the change in slope (which wassimilar for inflation and deflation of the loops) may reflect re-duced surface tension from the increased presence of surfaceactive molecules at the air–liquid interface of alveoli and smallairways after an increase in surface area (26), the normal hystere-sis behavior of the lung. After lavage injury, however, there wasclearly volume recruitment with each cycle in which the peakpressure exceeds Pflex (Figure 3). The first cycle of the inflationlimb data (Figure 3) closely resembles the predictions of Hickling(6 [Figures 2 and 3]), in which the compliance change is entirelyattributable to recruitment phenomena. Subsequent tidal in-creases in volume reflect additional recruitment not included inHickling’s model. As predicted by the model (6) and as demon-strated by our data and several studies (9, 10, 27–29), recruitmentoccurred all the way up to TLC. Pflex may thus be interpretedas a pressure threshold for recruitment: when peak inspiratorypressure exceeds Pflex, incremental volume recruitment occurs(Figure 5).

Deflation limb tidal loops lie close to the P–V envelope, withlittle tidal recruitment (Figures 3 and 5). A statistically significantdifference in deflation limb recruited volume between controland injured lungs occurred at the lowest tidal range, with theinjured lung exhibiting less recruitment, and the highest tworanges, with the injured lung recruiting more volume. This differ-ence likely reflects the rapid derecruitment of the lavaged lungand inability to rerecruit with peak inspiratory pressure less thanPflex. Similar to the predictions of Hickling (6), the slope of thetidal loop became less than the deflation P–V limb only at lowPEEP values, again reflecting end-expiratory alveolar collapsewith resultant lower ventilated volume and lower compliance.

In control lungs, specific compliance (sC) of tidal loops fellwith increasing PEEP (Figure 6), indicating the lungs becamestiffer as they were more inflated. In addition, sC was higherafter deflation from TLC than on the inflation limb (Figure 7).As discussed above, the higher deflation sC values may reflectrecruitment and larger lung volumes or, more likely in thesenormal lungs, the effect of surfactant physiology (26). Note thatthese data were normalized to control condition vital capacityto facilitate combining data from different animals. Thus, thisspecific compliance parameter is not a measure of intrinsic lungmechanical properties because it is not normalized to the actualventilated volume at the time of each measurement. sC behaviorof the injured lung was considerably different. sC was relativelyuniform along the inflation limb, whereas it varied more thantwofold on the deflation limb, peaking at moderate PEEP wellbelow Pflex (Figure 7). Again, these data provide strong experi-mental support of the models of Hickling (6), implemented atlow Vt, and suggest that the whole lung compliance in acutelung injury is significantly influenced by tidal recruitment andderecruitment.

CT Image Data

The lung volumes measured by CT imaging reflect a steady-stateequilibrium between end-inspiratory recruitment and end-expir-atory derecruitment obtained after incremental PEEP increases,in contrast to the transient or nonequilibrium volumes obtainedduring a P–V curve. These steady-state lung volumes exceededthose on the inspiratory P–V curve for all pressures, and alsoexceeded the volumes on the deflation limb at higher PEEPlevels. At the relatively large tidal volumes used (chosen to reducehypercapnia during imaging because of the slow maximum im-aging rate of the CT scanner), the peak inflation pressures weregreater than Pflex for all PEEPs. Thus, the elevation of end-expiratory volume above the inspiratory P–V curve would be ex-pected because of the effect of recruitment, as modeled by Hickling(6), with an additional contribution from the cumulative tidalrecruitment demonstrated by the tidal P–V loops. At the highestPEEP levels the peak inflation pressures slightly exceeded the40-cm H2O limit of the P–V curve, so the additional recruitmentthus obtained caused the end-expiratory volumes to break outof the P–V envelope, an effect also predicted by Hickling’s model(6). It is important to underscore that these CT data do not lookat the time course of changes after a recruitment maneuver, butrather the net effect of recruitment and derecruitment after anincrement in PEEP in a relatively short (10- to 20-minute) timeframe. Similar results, in which the CT-determined lung volumewas monitored after a recruitment maneuver in lavage injureddogs, were reported by Lim and coworkers, with total lung airvolumes continuing to increase for 30 minutes after a PEEPincrease compared with only small additional increases over timeafter a recruitment maneuver (30).

Air/Tissue Partitioned CT P–V Curves

Partitioning of the CT P–V curves into air, “tissue” (which in-cludes all components of approximately water density such asblood, edema, and cells), and total (air plus tissue) volumes addsan additional perspective on the nature of steady-state lungrecruitment in the lavage model. First, total lung volumes beforeand after lavage are approximately equal, suggesting that theloss of air volume with injury is primarily due to flooding becauseof the offsetting changes in air and tissue volumes, consistentwith observations made in oleic acid-injured lungs using theparenchymal marker technique (31). Atelectasis or airspace col-lapse would show up as a loss of total lung volume compared withcontrol in this analysis, although one cannot rule out regionalatelectasis with equally offsetting air trapping at FRC as analternative explanation. Second, the tissue volume remainedconstant as PEEP increased, suggesting that the total amountof edema (including residual lavage fluid) was constant over

962 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

time and, further, that intrapulmonary blood volume did notchange with increasing PEEP. Thus, as the lung is recruited thisfluid volume must be redistributed and relocated within the lung.Further studies of the changes in regional lung volume and densitywith inflation and recruitment, combined with whole lung analysissuch as this, are needed to better understand the local phenomenaof lung recruitment in different injury models and patients.

Conclusions

Small-volume tidal P–V loops performed after inflation fromFRC in the lavage-injured lung demonstrated incremental, cyclicrecruitment if the peak pressure achieved exceeded Pflex on thevital capacity quasi-static P–V curve, whereas loops performedafter deflation from TLC remained close to the envelope deflationP–V curve. Recruitment continued to occur up to and beyond apeak inspiratory airway pressure of 40 cm H2O, as demonstratedby both the tidal loops and the steady-state large Vt CT-derivedlung volume data. Tidal-specific compliance was relatively con-stant across PEEP levels after inflation from FRC, but peaked atmoderate PEEP after deflation from TLC, further demonstratingthe effects of volume history and providing experimental valida-tion of the recruitment models of Hickling (6, 7). These resultssupport the interpretation of Pflex as pressure threshold forrecruitment, but otherwise do not suggest a role for the pressure–volume curve in predicting steady-state lung volume.

Conflict of Interest Statement : J.M.D. has no declared conflict of interest; A.J.N.has no declared conflict of interest; B.A.S. has no declared conflict of interest.

Acknowledgment : The authors thank Vince Lerie for technical support with theCT scanner, Matt Piper for assistance in the laboratory, and Respironics (Murrays-ville, PA) for providing the modified PLV-102 ventilator.

References

1. Lachmann B. Open up the lung and keep the lung open [editorial; com-ment]. Intensive Care Med 1992;18:319–321.

2. Roupie E, Dambrosio M, Servillo G, Mentec H, el Atrous S, Beydon L,Brun-Buisson C, Lemaire F, Brochard L. Titration of tidal volumeand induced hypercapnia in acute respiratory distress syndrome. AmJ Respir Crit Care Med 1995;152:121–128.

3. Amato MB, Barbas CS, Medeiros DM, Magaldi RB, Schettino GP, Lo-renzi-Filho G, Kairalla RA, Deheinzelin D, Munoz C, Oliveira R,et al. Effect of a protective-ventilation strategy on mortality in theacute respiratory distress syndrome. N Engl J Med 1998;338:347–354.

4. Brower RG, Matthay MA, Morris A, Schoenfeld D, Thompson BT,ARDS Network. Ventilation with lower tidal volumes as comparedwith traditional tidal volumes for acute lung injury and the acuterespiratory distress syndrome. N Engl J Med 2000;342:1301–1308.

5. Rimensberger PC, Cox PN, Frndova H, Bryan AC. The open lung duringsmall tidal volume ventilation: concepts of recruitment and “optimal”positive end-expiratory pressure. Crit Care Med 1999;27:1946–1952.

6. Hickling KG. Best compliance during a decremental, but not incremental,positive end-expiratory pressure trial is related to open-lung positiveend-expiratory pressure: a mathematical model of acute respiratorydistress syndrome lungs. Am J Respir Crit Care Med 2001;163:69–78.

7. Hickling KG. The pressure–volume curve is greatly modified by recruit-ment: a mathematical model of ARDS lungs. Am J Respir Crit CareMed 1998;158:194–202.

8. Svantesson C, Sigurdsson S, Larsson A, Jonson B. Effects of recruitmentof collapsed lung units on the elastic pressure–volume relationship inanaesthetised healthy adults. Acta Anaesthesiol Scand 1998;42:1149–1156.

9. Richard JC, Brochard L, Vandelet P, Breton L, Maggiore SM, JonsonB, Clabault K, Leroy J, Bonmarchand G. Respective effects of end-expiratory and end-inspiratory pressures on alveolar recruitment inacute lung injury. Crit Care Med 2003;31:89–92.

10. Jonson B, Richard JC, Straus C, Mancebo J, Lemaire F, Brochard L.

Pressure–volume curves and compliance in acute lung injury: evidenceof recruitment above the lower inflection point. Am J Respir Crit CareMed 1999;159:1172–1178.

11. Downie JM, Simon BA. Pressure threshold for volume recruitment dur-ing tidal ventilation after lung lavage [abstract]. Am J Respir Crit CareMed 1999;161:A76.

12. Simon BA, Marcucci C, Piper MG, Downie JM. CT P–V analysis differen-tiates flooding vs collapse in oleic acid (OA) and lavage injury models[abstract]. Am J Respir Crit Care Med 2000;161:A486.

13. Venegas JG, Harris RS, Simon BA. A comprehensive equation for thepulmonary pressure–volume curve. J Appl Physiol 1998;84:389–395.

14. Lichtwarck-Aschoff M, Hedlund AJ, Nordgren KA, Wegenius GA, Mar-kstrom AM, Guttmann J, Sjostrand UH. Variables used to set PEEPin the lung lavage model are poorly related. Br J Anaesth 1999;83:890–897.

15. Lewis J, McCaig L, Hafner D, Spragg R, Veldhuizen R, Kerr C. Dosingand delivery of a recombinant surfactant in lung-injured adult sheep.Am J Respir Crit Care Med 1999;159:741–747.

16. Kloot TE, Blanch L, Melynne Youngblood A, Weinert C, Adams AB,Marini JJ, Shapiro RS, Nahum A. Recruitment maneuvers in threeexperimental models of acute lung injury: effect on lung volume andgas exchange. Am J Respir Crit Care Med 2000;161:1485–1494.

17. Lewis JF, Jobe AH. Surfactant and the adult respiratory distress syn-drome. Am Rev Respir Dis 1993;147:218–233.

18. Hudson LD. Protective ventilation for patients with acute respiratorydistress syndrome. N Engl J Med 1998;338:385–387.

19. Servillo G, De Robertis E, Maggiore S, Lemaire F, Brochard L, TufanoR. The upper inflection point of the pressure–volume curve: influenceof methodology and of different modes of ventilation. Intensive CareMed 2002;28:842–849.

20. Servillo G, Svantesson C, Beydon L, Roupie E, Brochard L, Lemaire F,Jonson B. Pressure–volume curves in acute respiratory failure: auto-mated low flow inflation versus occlusion. Am J Respir Crit Care Med1997;155:1629–1636.

21. Lu Q, Vieira SR, Richecoeur J, Puybasset L, Kalfon P, Coriat P, RoubyJJ. A simple automated method for measuring pressure–volume curvesduring mechanical ventilation. Am J Respir Crit Care Med 1999;159:275–282.

22. Kondili E, Prinianakis G, Hoeing S, Chatzakis G, Georgopoulos D. Lowflow inflation pressure–time curve in patients with acute respiratorydistress syndrome. Intensive Care Med 2000;26:1756–1763.

23. Greville HW, Arnup ME, Mink SN. Density dependence of maximalflow is lung volume dependent during bronchoconstriction. J ApplPhysiol 1987;62:691–705.

24. Macklem PT, Mead J. Factors determining maximum expiratory flow indogs. J Appl Physiol 1968;25:159–169.

25. Vieillard-Baron A, Prin S, Schmitt JM, Augarde R, Page B, Beauchet A,Jardin F. Pressure–volume curves in acute respiratory distress syndrome:clinical demonstration of the influence of expiratory flow limitation onthe initial slope. Am J Respir Crit Care Med 2002;165:1107–1112.

26. Nunn JF. Nunn’s applied respiratory physiology, 4th ed. Boston: Butter-worths; 1993.

27. Crotti S, Mascheroni D, Caironi P, Pelosi P, Ronzoni G, Mondino M,Marini JJ, Gattinoni L. Recruitment and derecruitment during acuterespiratory failure: a clinical study. Am J Respir Crit Care Med 2001;164:131–140.

28. Chelucci GL, Dall’Ava-Santucci J, Dhainaut JF, Chelucci A, Allegra A,Lockhart A, Zin WA, Milic-Emili J. Association of PEEP with two dif-ferent inflation volumes in ARDS patients: effects on passive lung defla-tion and alveolar recruitment. Intensive Care Med 2000;26:870–877.

29. Richard JC, Maggiore SM, Jonson B, Mancebo J, Lemaire F, BrochardL. Influence of tidal volume on alveolar recruitment: respective roleof PEEP and a recruitment maneuver. Am J Respir Crit Care Med2001;163:1609–1613.

30. Lim CM, Soon Lee S, Seoung Lee J, Koh Y, Sun Shim T, Do Lee S,Sung Kim W, Kim DS, Dong Kim W. Morphometric effects of therecruitment maneuver on saline-lavaged canine lungs: a computedtomographic analysis. Anesthesiology 2003;99:71–80.

31. Martynowicz MA, Minor TA, Walters BJ, Hubmayr RD. Regionalexpansion of oleic acid-injured lungs. Am J Respir Crit Care Med 1999;160:250–258.

NHLBI Workshop

Obesity and AsthmaDirections for Research

Scott T. Weiss and Stephanie Shore

Channing Laboratory, Brigham and Women’s Hospital; and Harvard School of Public Health, Boston, Massachusetts

The Lung Division of the NHLBI convened a workshop onobesity and asthma on July 15 and 16, 2003. Drs. Patricia Noeland Virginia Taggart, the workshop organizers, sought to exam-ine current research on the role of obesity in the onset andpersistence of asthma, as well as the mechanistic basis for thisassociation. An additional goal of the workshop was to describebarriers in conducting research in this area and to identify themost promising opportunities for future research.

EPIDEMIOLOGY OF ASTHMA AND OBESITY

Asthma is defined by episodic airflow obstruction, increasedairways responsiveness, and airway inflammation characterizedby infiltration with eosinophils and T lymphocytes, particularlyCD4� T lymphocytes that express T helper (Th) cell type 2cytokines such as interleukin (IL)-4, IL-5, and IL-13. The histo-pathologic appearance of the airways includes denudation ofthe airway epithelium, thickening of the basement membrane,mucus production, and airway smooth muscle hypertrophy. Al-though asthma is a chronic, often lifelong disease that affectshumans of all ages, the onset of the disease occurs primarilyin early childhood. Fifty percent of all male asthma cases arediagnosed by age three, and 50% of all female cases are diag-nosed by age eight (1). From 1980 to 1996, data from the NationalCenter for Health Statistics demonstrate an increase in self-reported asthma prevalence of 74% (2). Although this increasehas occurred in children and young adults of all ages, it has beenmost pronounced in children 5 years or younger. This increasehas been most marked in minority populations, particularly Afri-can Americans and Puerto Rican Hispanics. There has beensubstantial morbidity and cost to this epidemic as asthma is thenumber one cause of hospitalizations in children as well as themost common chronic condition for days lost from school (3).At the present time, medications and healthcare utilization forchildhood asthma costs approximately $10 billion per year.

In addition to the asthma epidemic, there is an obesity epi-demic in the United States. One-third of all 16-year-old childrenin the United States are overweight, and 15% are obese. Cross-sectional studies indicate an increased prevalence of asthma inthe obese (4–7). Although it is possible that this association isthe result of reduced exercise by subjects with asthma leading toobesity, longitudinal studies (8–10), which indicate that obesity

(Received in original form March 20, 2003; accepted in final form January 21, 2004)

National Heart, Lung, and Blood Institute Workshop, Bethesda, Maryland, July15 and 16, 2002.

Correspondence and requests for reprints should be addressed to Scott T. Weiss,M.D., M.S., Channing Laboratory, Brigham and Women’s Hospital, 181 Long-wood Avenue, Boston, MA 02110. E-mail: Scott.weiss@channing.harvard.edu

Am J Respir Crit Care Med Vol 169. pp 963–968, 2004Originally Published in Press as DOI: 10.1164/rccm.200303-403WS on January 23, 2004Internet address: www.atsjournals.org

antedates asthma, suggest that this is not the case. An importantaspect of these epidemiologic data is that the impact of obesityon asthma is much stronger in females than males. For example,the incidence of asthma after the age of 11 years is five- tosevenfold higher in female children who become obese versusthose who remain lean, whereas no such relationship exists formales (8). Further understanding of this sex difference may helpelucidate the basis for the relationship between these two healthconditions. This relationship between asthma and obesity hasbeen reviewed recently (11).

OBESITY AND AIRWAY RESPONSIVENESS

Although the relationship between obesity and asthma is reason-ably clear, the relationship between obesity and airway respon-siveness is less so. In an initial report, Schachter and coworkers(12) found a relationship between obesity and asthma occur-rence; however, they found no relationship between obesity andan increase in airway responsiveness. Data on white adults be-tween the ages of 17 and 73 years from three large epidemiologicstudies in New South Wales, Australia, from 1971 were exam-ined. In this population, asthma risk was significantly increased(odds ratio 2.04, p � 0.048), but there was no increase in airwayresponsiveness to histamine (odds ratio 2.87, p � 0.78). In alarge cross-sectional, population-based study in Anqing, China,extremes of body mass index (BMI), either very low or veryhigh, were associated with a 2.5-fold increase in symptomaticairway hyperresponsiveness (6). The effects were seen in bothmales and females and in both extremes of the BMI distribution.In longitudinal data from the Normative Aging Study, there wasalso an association between BMI and increased airway hyper-responsiveness with a reported odds ratio of 7 (13). Thus, thedata in the literature is conflicting as to whether airway hyper-responsiveness is increased by obesity.

OBESITY AND ALLERGY

There is at least one article examining data from the NationalHealth and Nutrition Examination Survey that suggests thatobesity is associated with an increase in skin test reactivity butnot peripheral blood eosinophils. IgE levels were not measuredin this study (7). There is only one other report of cross-sectionaldata showing a relationship between BMI and an increase inskin test reactivity (14). There has been only one attempt tolink the increase in obesity to the increase in asthma wheezingprevalence. Figueroa-Munoz and coworkers (15) used data froma National Study of Health and Growth in the United Kingdomto examine if obesity preceded the increase in asthma and wheez-ing. In their initial report on data from this population, the investi-gators show a strong relationship between BMI and asthmaoccurrence cross-sectionally. In a subsequent report, the investi-gators suggest that this cross-sectional association may not havebeen temporally related (16). This latter report can be criticized

964 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

because it is primarily a repeated cross-sectional qualitativeanalysis rather than a true longitudinal analysis, it did not useall the available data, and it did not clearly show that BMI valuesactually antedated the asthma wheeze determinations. Issues ofselection in the data make this article a less than optimal testof this particular hypothesis. As is commonly seen in complextraits like asthma, there are multiple obesity phenotypes. Forexample, central adiposity is linked with insulin resistance,whereas centrifugal or gluteal adiposity is not. These phenotypesdiffer with respect to their relationship to birth weight and toactivation of the sympathetic nervous system (SNS). To date,no attention has been given to elucidating the obesity phenotypesthat are linked to asthma or the asthma phenotypes that arelinked to obesity.

Given that asthma is a disease of young children, it is possiblethat in utero events are contributing to the relationship betweenobesity and asthma. For example, birth weight appears to impactthe development of both asthma (5) and obesity (17). Low birthweight is associated with a higher subsequent incidence ofasthma, whereas high birth weight is associated with a highersubsequent incidence of obesity. There are data indicating thatmaternal caloric intake during pregnancy in humans, animal-rearing temperature (i.e., the cage temperature that mother andneonate are kept at), and human fetal hyperinsulinemia impactthe development of obesity. However, there are few data as tothe impact of these factors on asthma and no data as to theirimpact on the relationship between obesity and asthma.

OBESITY AND LEPTIN

Leptin is a hormone produced by adipocytes that acts in thehypothalamus to signal satiety and to increase basal metabolicrate. Serum leptin is increased in obesity (18). The role of leptinin the relationship between obesity and asthma is unknown.Leptin receptors are expressed outside the hypothalamus, andperipheral effects of leptin on many organs and tissues, includingthe lung and hematopoietic cells, have been reported. For exam-ple, leptin stimulates surfactant synthesis in fetal lung cells(19, 20), as well as proliferation of tracheal epithelial cells (21),and mice that lack dysanapsis leptin have markedly reducedlung size (22). Given that larger lungs and smaller airways arepotentially important in the etiology of asthma (23), understand-ing the role of leptin in fetal lung development may prove tobe very important (10).

Leptin is a member of the IL-6 family of cytokines, andhematopoietic cells respond to leptin with proliferative re-sponses as well as altered cytokine production. In general, theoverall response to leptin is to stimulate a Th1 cytokine profile:peritoneal macrophages pretreated with leptin generate in-creased amounts of IL-12 in response to endotoxin (24), whereasin CD4� T cells, leptin increases Th1 and suppresses Th2 cyto-kine production (25). Starvation and malnutrition are associatedwith immune dysfunction, and leptin administration reverses theimmunosuppressive effects of acute starvation in animal models(25). Leptin also reverses the impairments in T cell proliferationand cytokine release that are observed in humans with congenitalleptin deficiency (26). In addition to its effects on immune cellfunction, leptin also has effects on inflammation such as promo-tion of the release of tumor necrosis factor-� and IL-6 fromendotoxin-treated macrophages and lymphocytes. Consistentwith these effects, leptin enhances ozone-induced airway hyper-responsiveness and increases ozone-induced neutrophil influxand eotaxin release into bronchoalveolar lavage fluid in mice(27). In contrast, fasting, which reduces serum leptin, attenuatesozone-induced inflammation. Finally, obesity per se is associatedwith systemic inflammation including elevations in peripheral

blood leukocytes, which causes an increase in serum levels ofC-reactive protein, proinflammatory cytokines, such as tumor ne-crosis factor-� and IL-6, and cell adhesion markers, and markersof lipid peroxidation (28–31). Greater understanding of the simi-larities and differences between the inflammatory characteristicof obesity and asthma might aid in understanding the impact ofobesity on asthma.

An additional effect of leptin that could have important impli-cations for asthma is its ability to activate the SNS, an effectthat appears to occur at the level of the hypothalamus in animalmodels (32). Leptin increases activity in sympathetic nerves inthe kidney, brown adipose tissue, hind limbs, and adrenal me-dulla (33). The impact of leptin on the activity of sympatheticnerves in lung is not known, but increased sympathetic activationof the adrenal medulla could lead to release of catecholaminesthat would be expected to impact lung function.

Although serum leptin is increased in obesity, there is resis-tance to the effects of peripherally administered leptin on eatingbehavior and on sympathetic activity to brown fat in rodentmodels of obesity (34, 35). In contrast, renal sympathetic activityis preserved in obese mice (36), suggesting the possibility ofselective leptin resistance. Whether there is also resistance tothe effects of leptin in the lung or on immune function is un-known, but such resistance might be expected to polarize to aTh2 response and to reduce lung size in developing children,conditions that could be expected to impact the developmentof asthma. In the brain, the mechanism of leptin resistance is viaupregulation of suppressor of cytokine signaling 3, a phosphatasethat limits leptin receptor signaling (37). Suppressor of cytokinesignaling 3 limits IFN-� signaling (38), which might enhance Th2cytokine signaling.

MECHANICAL FACTORS

In obese individuals, airway smooth muscles are likely to beunloaded (have less tension) in part because obese individualshave a decreased FRC and in part because obese individualsassume a breathing pattern with higher frequencies and lowerVt that lean control individuals. Decreased FRC shortens musclelength and hence decreases tension. Tidal stretch of airwaysmooth muscle is an extremely potent bronchodilator: stretch ofairway smooth muscle detaches cross bridges, leading to reducedforce generation and reduced shortening (39). The reduction incross bridges also makes the muscle less stiff. Thus, in the obeseindividual with a reduced FRC and a decreased Vt, it is easy tocreate a downward spiral, wherein less stretch leads to greaterstiffness making the muscle even harder to stretch and resultingin increased airway shortening and airway hyperresponsiveness.These mechanical effects have been observed in the massivelyobese. The extent to which these changes occur in subjects overa wide range of body weight and age is an important researchquestion. Understanding whether there are age- or sex-relateddifferences in these mechanical factors and how different pat-terns of obesity and breathing produce these effects would beworthwhile.

FEMALE SEX STEROID HORMONES

The clear-cut sex differences in the epidemiology of obesity andits impact on asthma suggest that female sex hormones maybe contributing to the increased risk of asthma in obesity.Aromatase, the enzyme responsible for converting androgensto estrogens, is found in adipose tissue. Therefore, it is reasonableto hypothesize that obesity increases estrogen and is associatedwith early menarche (40) and delayed puberty in males (41).The risk of developing asthma in girls who gain weight is also

NHLBI Workshop 965

particularly strong in those with early menarche (8). The twodifferent estrogen receptors (ERs), ER� and ER�, are expressedin adipose tissue. In general, estrogen leads to an increase inbasal metabolic rate as well as increased ambulatory activity anddecreased activity of lipoprotein lipase in laboratory animals.Mice genetically deficient in ER� (42) as well as those lackingin aromatase (43) are obese; however, ovariectomized ER� micehave much less fat than intact ER� mice, suggesting that theestrogen signaling through the ER� may promote fat deposition.Although there is some literature on the effect of estrogen onairway responsiveness in animal models (44), it is unclear howestrogen might impact the development of asthma, but bothestrogen and progesterone have been shown to increase IL-4and IL-13 in peripheral blood mononuclear cells (45), and theremay be other effects on immune or inflammatory cells. It is alsopossible that there are ERs on airway smooth muscle or otherairway cells and that estrogen directly impacts airway function.If so, it will be important to understand how ER expression andsignal transduction are regulated in these cells.

SNS ACTIVITY

Although human airway smooth muscle is not sympatheticallyinnervated (46), other cells in the airways that may impact airwayfunction are goblet cells and mucus glands (47). Moreover,smooth muscle cells do express �2-adrenergic receptors, respondto �-agonists with increases in cAMP formation and relaxation,and are likely to be influenced by catecholamines of medullaryorigin. There are changes in sympathetic activity with obesitythat appear to be organ specific. In animal models of obesity,baseline renal sympathetic activity is increased, whereas activa-tion in brown fat is reduced (48). This should be contrasted withthe leptin-induced increase in SNS activity in brown fat notedearlier. In contrast, 24-hour urinary epinephrine levels indicatedecreased adrenal medullary function associated with obesity ora high-fat diet in humans and animals (49, 50). The observationsthat adrenal medullary function is reduced in obesity and thatSNS activity is decreased in asthma suggests that changes insympathetic activity may be the link between these two syn-dromes but much remains to be established. For example, thereare no data on the impact of obesity on sympathetic outflow tothe lung. Furthermore, little is known about 24-hour urinarynorepinephrine and epinephrine levels in asthma relative to nor-mal levels. Greater attention to the physiologic changes seen inobesity as applied to the subject with asthma might provide aninsight into the mechanistic interrelationship between these twoconditions.

In rodents, sympathetic activation of brown fat appears toplay an important role in increasing thermogenesis and basalmetabolism via activation of uncoupling proteins. All three typesof �-adrenergic receptors, �1, �2, and �3, are expressed in adiposetissue. Moreover, in rodent models of obesity, there are deficitsin �1- and �3-receptor expression in brown and white adiposetissues, as well as deficits in �1 and �3 coupling, as indicated bycAMP formation in response to receptor-specific agonists (51).In contrast, �2-receptors are unaffected. The �2-receptor is theprimary �-receptor expressed in the lung, but there are tissue-specific differences in �-receptor regulation, and the impact ofobesity on �-receptor expression in the lung is unknown.

The classical pathway by which �-receptors activate cells isactivation of G proteins leading to adenyl cyclase activation,cAMP formation, and protein kinase A activation. However,�-receptors can also activate the mitogen-activated proteinkinase pathway. Activation of the extracellular regulated kinasemitogen-activated protein kinases by �2-receptor activation re-quires Src family kinase (c-Src) and appears to occur as a result

of G protein–coupled receptor kinase phosphorylation of thereceptor, consequent �-arrestin binding, and recruitment ofcytoplasmic activator of the c-Src to �-arrestin. In contrast, the�3-receptor directly recruits c-Src (52). In adipocytes, p38 mito-gen-activated protein kinase is also activated by �2-agonists andappears to be required for the ability of these agents to induceuncoupling protein 1 (53). The role of these nonclassical path-ways of �2-receptor activation in asthma and in the relation-ship between obesity and asthma is unexplored but warrantsexploration.

Environmental factors such as maternal diet, stress, physicalactivity, and cage temperature play a major role in SNS develop-ment in utero and in the ability of model organisms such as themouse to respond to environmental stress. These factors play arole in SNS development in utero (54). As discussed previously,intrauterine growth impacts the subsequent development of obe-sity. Similarly, SNS activity tends to be impaired in infants within utero growth retardation, and it remains unclear what effectprematurity has on SNS development. A better understandingof diet, child diet, maternal and child exercise, and their impacton SNS function might help elucidate joint or overlapping mech-anisms (54, 55).

GENETICS

Linkage analysis has identified several linkage peaks with chro-mosomal regions that are shared for obesity and asthma pheno-types (11). 2P, 5Q, 6P, 7P, and 12Q all contain regions with locicommon to both complex phenotypes. The testing of candidategenes and their role in both the SNS and inflammatory pathwaysprovides unique opportunity to look for shared genes as well asshared environmental exposures that may be common to thesetwo complex traits. In addition to linkage, specific candidategenes deserve careful attention. For example, the �2-adrenergicreceptor polymorphisms may be important in understanding air-way smooth muscle cell function as well as fat cell and immuneand inflammatory cell functions.

SLEEP-DISORDERED BREATHING

Sleep-disordered breathing is characterized by recurrent epi-sodes of upper airway obstruction (apneas and hypopneas) andis often associated with abnormal gas exchange, snoring, andsleep disruption. The prevalence of this condition is 1 to 3% inchildren, 2 to 10% in the middle aged, and greater than 25% inthe elderly. Importantly, the prevalence of snoring and sleep-disordered breathing is higher among subjects with asthma thanamong those without asthma (56), and asthma is a risk factorfor sleep-disordered breathing in children (57). Moreover, insubjects with both asthma and sleep-disordered breathing, therecan be an improvement in asthma symptoms on treatment withnasal continuous positive airway pressure (58). Sleep-disorderedbreathing and asthma share some common risk factors, includingAfrican American race, prematurity, and obesity: 50% of theobese have sleep apnea syndrome, and among those with sleepapnea, 40% are obese (59, 60). Other common risk factors areallergy and lower respiratory infections. The mechanistic basisfor the association between asthma and sleep-disordered breath-ing is unclear. The two phenotypes may overlap because of thecommon occurrence of the two complex traits. It may also bethat hormonal changes, changes in neural tone to the airways,gastroesophageal reflux, alterations in breathing pattern leadingto altered smooth muscle tone, or hypoxemia contribute.

CONCLUSIONS

Clearly, both asthma and obesity are common conditions, andboth are major public health problems. Moreover, obesity appears

966 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

to increase the risk of asthma. Both disorders may share commongenetic and environmental causes. There are mechanical, develop-mental, hormonal, signal transduction, and immunologic reasonsfor their effects. The workshop developed a total of 10 complexrecommendations to further research in this area.

RESEARCH RECOMMENDATIONS FROM THE OBESITYAND ASTHMA WORKSHOP, NHLBI

1. More information, including better phenotyping and morelongitudinal cohort studies, is required to determine whatis the attributable risk for developing asthma given obesityand in sorting out cause–effect relationships between asthmaand obesity. For example, do treatments for asthma, such as�-agonists or inhaled glucocorticoids, predispose to obesity?Obesity, like asthma, is very heterogeneous, and it will beimportant to understand which of the various obesity pheno-types are linked to asthma. Ethnic differences in obesity andasthma phenotypes may help strengthen such associations.Given the apparent importance of obesity in the develop-ment of asthma, can ongoing studies of cardiovasculardisease or weight reduction (Cardia, Lookahead) be usedto determine the impact on asthma by collecting asthmaoutcomes measurements? Can more use be made of gas-tric reduction surgery patients in terms of evaluating howweight reduction improves asthma?

2. It is clear that the impact of obesity in asthma is morepronounced in females than in males. Understanding thissex difference may help elucidate the mechanistic basisfor the relationship between these two syndromes. In thiscontext, it is important to understand what role estrogenis playing. What are the effects of estrogen on eatingcenters in the brain? Which cells in the airways expressERs and how is ER expression and signal transductionregulated? Does ER signaling interact with other signalingpathways of relevance to asthma? Does estrogen influenceairway inflammation? Are there interactions between lep-tin and estrogen in the brain, in the adipose tissue, or inthe airways? Are there interactions between estrogen andthe SNS?

3. There is decreased adrenal medullary activity in obesity,as evidenced by 24-hour urine epinephrine levels. How-ever, the decreased SNS activation does not affect allinnervated organs. Renal SNS is increased in obesity. Howis sympathetic outflow to the lung affected in obesity? Howdoes this impact the development of asthma symptoms? Isthere a potential link between obesity-related changes inSNS activation and asthma? How is sympathetic innerva-tion of the lung regulated developmentally, and does obe-sity impact innervation? How do polymorphisms of the�-adrenergic receptors influence the relationship betweenobesity and asthma? Is the impact of �-receptor polymor-phisms in airway smooth muscle, fat cells, and immune/inflammatory cells the same? Do exercise, diet, and poly-morphisms of the �-receptor interact to predict whichobese individuals get asthma? Are there differences inresponse to sympathetic activation in the lungs of obeseversus nonobese? Are the signal transduction pathways(protein kinase A, extracellular regulated kinase, p38)activated by �-agonists affected by obesity? What is therelationship of 24-hour urinary norepinephrine and epi-nephrine levels to asthma and its phenotypes? Can thisphenotype for SNS activity be used to assess environmen-tal factors and genetic factors influencing obesity andasthma?

Leptin is increased in obesity. Thus, it is important tounderstand the mechanistic basis for effects of leptin onthe immune system, on lung growth, and on the SNS andto determine how these effects influence the developmentof asthma. For example, what cells in the lung expressleptin, and how does leptin influence their function? Howdoes leptin influence sympathetic outflow to the lung orrelease of catecholamines from the adrenal medulla? Howdoes leptin impact immune system responses that occurduring allergen sensitization and challenge? Resistance tosome but not all the effects of leptin is evident in animalmodels of obesity. It will be important to understand themechanistic basis for leptin resistance and to determineif and how obesity impacts resistance to effects of leptinin the lungs, on T cells and other immune/inflammatorycells, and in the SNS. Development of more animal modelsmay facilitate such studies, including tissue-specific knock-outs of leptin receptors and regulated transgenic miceoverexpressing leptin or leptin receptors in target organs.Regulated knockouts of leptin may also assist in determin-ing when, developmentally, leptin may be important forthe development of asthma.

4. Besides leptin, there are other hormones produced fromfat cells, including resistin, adiponectin, and tumor necro-sis factor-�, as well as other hormones that are affectedby obesity such as the renin–angiotensin system. It may beimportant to determine what effect, if any, these hormoneshave on the lung or on inflammatory/immune responsesand to determine if these are relevant to asthma.

5. Evidence suggests that obesity is an inflammatory state.Are there similarities in the inflammation produced byobesity and asthma? Do they have a common etiology?

6. Little is known about how intrauterine and early life relateto development of obesity and asthma. What dietary, envi-ronmental, or genetic factors influence the developmentof these two conditions? If so, what is the mechanisticbasis for this relationship? Which phenotypes, such asbirth weight corrected for gestational age and birth length,growth in the first year of life, serial ultrasounds duringfetal development, best predict the development ofasthma and/or obesity in early childhood? Given that pre-term birth appears to influence the development of bothasthma and obesity, are there genes linked to preterm birththat may influence the development of these conditions?

7. Genome-wide scans have identified regions linked toasthma and to obesity, some of which overlap. There arealso common candidate genes, including, but not limitedto, the �2-receptor, the glucocorticoid receptor, trans-forming growth factor-�, and peroxisome proliferator–activated receptor–�. Functional genomics of mediators,genes, and pathways that are likely to participate in bothobesity and asthma and interactions between these path-ways may help elucidate the mechanistic basis for therelationship between these conditions. Understandinghow diet, exercise, and other environmental factors inter-act with these genes may also help determine risk.

8. Obesity provides an important load to the respiratorysystem and alters lung volumes and the pattern of breathing,both of which can affect airway smooth muscle shortening.How does obesity impact the loading of airway smoothmuscle including the frequency of deep inspirations? Arethere age- or sex-related differences in deep inspirationsor in the impact of deep inspirations on airway function? Dothese vary with obesity? How does obesity impact airway

NHLBI Workshop 967

development and are these effects sex specific? Doesbreathing at low lung volumes over long periods of time,as occurs in obesity, cause remodeling of the airways?

9. There appear to be important interactions not only be-tween asthma and obesity but also between asthma andsleep and obesity and sleep. Measurement of sleep pheno-types (polysomnograms, sleep questionnaires, diet, exer-cise, sleep deprivation) in ongoing longitudinal asthmacohort studies and measurement of asthma phenotypes(asthma questionnaires, lung function testing, skin testing,IgE levels) in ongoing sleep studies may help elucidatethese relationships. Further understanding of the mecha-nistic basis for the impact of sleep on asthma may alsohelp clarify the relationship between obesity and asthma.Can one separate the effects of obesity and sleep on respi-ratory mechanics? Are the same mechanisms operatingin obese patients with sleep apnea and obese subjects withasthma in terms of airway mechanics?

10. Many participants emphasized that cross-discipline collab-oration between investigators interested in asthma andthose interested in obesity would likely accelerate progressin this area.

Conflict of Interest Statement : S.T.W. received a grant from Glaxo-Welcome($300,000) in 2000 for the purchase of genotyping and sequencing equipmentand an additional grant from Glaxo-Welcome ($60,000) in 2003 for genotypingand also served on the mortality committee for the SMART study for Glaxo-Welcome from 2000–2003 and received a grant from AstraZeneca for $300,000each year from 1999–2001 for the development of an asthma policy modeland he was a coinvestigator on a grant from Millennium Pharmaceuticals from1996–1999 to study the genetics of asthma in China and received 10% of hissalary from that grant for these years and in 2001 received a grant from Pfizerfor $65,000 to study the effects of insulin and diabetes on lung function andreceives $10,000 per year as member and then chair of the scientific advisorycommittee for the Tenor study sponsored by Genetech and served for two years2001–2003, and he served as an advisor (�$5,000 each) to the following pharma-ceutical companies for the last five years: Roche Pharmaceuticals, Schering-Plough,Boehringer Ingelheim, Variagenics, Genome Therapeutics, and Merck Frost; S.S.has no declared conflict of interest.

Acknowledgment : The authors would like to thank the National Heart, Lung,and Blood Institute Workshop Group, especially Patricia Noel, Ph.D. and VirginiaTaggart, M.P.H., for their assistance in organizing the workshop. Members of theNHLBI Workshop Group are: Sheila Collins, Ph.D.; Jeffrey J. Fredberg, Ph.D.; Wil-liam Haynes, M.D.; Marilyn Halonen, Ph.D.; Tricia Heine, D.V.M., Ph.D.; GrahamLord, M.D., Ph.D.; Fernando D. Martinez, M.D.; Susan S. Redline, M.D., M.P.H.;Kelan Tantisira, M.D.; and James B. Young, M.D.

References

1. Yunginger JW, Reed CE, O’Connell EJ, Melton LJ III, O’Fallon WM,Silverstein MD. A community-based study of the epidemiology ofasthma. Am Rev Respir Dis 1992;146:888–894.

2. Mannino DM, Homa DM, Pertowski CA, Ashizawa A, Nixon LL, John-son CA, Ball LB, Jack E, Kang DS. Surveillance for asthma: UnitedStates, 1960–1995. MMWR CDC Surveill Summ 1998;47:1–27.

3. Mannino DM, Homa DM, Akinbami LJ, Moorman JE, Gwynn C, ReddSC. Surveillance for asthma: United States, 1980–1999. MMWR Sur-veill Summ 2002;51:1–13.

4. Chen Y, Dales R, Krewski D, Breithaupt K. Increased effects of smokingand obesity on asthma among female Canadians: the National Popula-tion Health Survey, 1994–1995. Am J Epidemiol 1999;150:255–262.

5. Shaheen SO, Sterne JA, Montgomery SM, Azima H. Birth weight, bodymass index and asthma in young adults. Thorax 1999;54:396–402.

6. Celedon JC, Palmer LJ, Litonjua AA, Weiss ST, Wang B, Fang Z, XuX. Body mass index and asthma in adults in families of subjects withasthma in Anqing, China. Am J Respir Crit Care Med 2001;164:1835–1840.

7. von Mutius E, Schwartz J, Neas LM, Dockery D, Weiss ST. Relation ofbody mass index to asthma and atopy in children: the National Healthand Nutrition Examination Study III. Thorax 2001;56:835–838.

8. Castro-Rodriguez JA, Holberg CJ, Morgan WJ, Wright AL, MartinezFD. Increased incidence of asthmalike symptoms in girls who become

overweight or obese during the school years. Am J Respir Crit CareMed 2001;163:1344–1349.

9. Carmargo CA, Weiss ST, Zhang S, Willett WC, Speizer FE. Prospectivestudy of body mass index, weight change, and risk of adult-onset asthmain women. Arch Intern Med 1999;159:2582–2588.

10. Beckett WS, Jacobs DR Jr, Yu X, Iribarren C, Williams OD. Asthma isassociated with weight gain in females but not males, independent ofphysical activity. Am J Respir Crit Care Med 2001;164:2045–2050.

11. Tantisira KG, Weiss ST. Complex interactions in complex traits: obesityand asthma. Thorax 2001;56:ii64–ii73.

12. Schachter LM, Salome CM, Peat JK, Woolcock AJ. Obesity is a riskfor asthma and wheeze but not airway hyperresponsiveness. Thorax2001;56:4–8.

13. Litonjua AA, Sparrow D, Weiss ST. The FEF25-75/FVC ratio is associ-ated with methacholine airway responsiveness: the Normative AgingStudy. Am J Respir Crit Care Med 1999;159:1574–1579.

14. Jedrychowski W, Maugeri U, Flak E, Mroz E, Bianchi I. Predispositionto acute respiratory infections among overweight preadolescent chil-dren: an epidemiologic study in Poland. Public Health 1998;112:189–195.

15. Figueroa-Munoz JI, Chinn S, Rona RJ. Association between obesity andasthma in 4–11 year old children in the UK. Thorax 2001;56:133–137.

16. Chinn S, Rona RJ. Can the increase in body mass index explain therising trend in asthma in children? Thorax 2001;56(11):845–850.

17. Law CM, Barker DJ, Osmond C, Fall CH, Simmonds SJ. Early growthand abdominal fatness in adult life. J Epidemiol Community Health1992;46:184–186.

18. Maffei M, Halaas J, Ravussin E, Pratley RE, Lee GH, Zhang Y, Fei H,Kim S, Lallone R, Ranganathan S, et al. Leptin levels in human androdent: measurement of plasma leptin and ob RNA in obese andweight-reduced subjects. Nat Med 1995;1:1155–1161.

19. Bergen HT, Cherlet TC, Manuel P, Scott JE. Identification of leptinreceptors in lung and isolated fetal type II cells. Am J Respir Cell MolBiol 2002;27:71–77.

20. Torday JS, Sun H, Wang L, Torres E, Sunday ME, Rubin LP. Leptinmediates the parathyroid hormone-related protein paracrine stimula-tion of fetal lung maturation. Am J Physiol Lung Cell Mol Physiol2002;282:L405–L410.

21. Tsuchiya T, Shimizu H, Horie T, Mori M. Expression of leptin receptorin lung: leptin as a growth factor. Eur J Pharmacol 1999;365:273–279.

22. Tankersley CG, O’Donnell C, Daood MJ, Watchko JF, Mitzner W,Schwartz A, Smith P. Leptin attenuates respiratory complications asso-ciated with the obese phenotype. J Appl Physiol 1998;85:2261–2269.

23. Weiss ST, Tosteson TD, Segal MR, Tager IB, Redline S, Speizer FE.Effects of asthma on pulmonary function in children: a longitudinalpopulation-based study. Am Rev Respir Dis 1992;145:58–64.

24. Loffreda S, Yang SQ, Lin HZ, Karp CL, Brengman ML, Wang DJ,Klein AS, Bulkley GB, Bao C, Noble PW, et al. Leptin regulatesproinflammatory immune responses. FASEB J 1998;12:57–65.

25. Lord GM, Matarese G, Howard JK, Baker RJ, Bloom SR, Lechler RI.Leptin modulates the T-cell immune response and reverses starvation-induced immunosuppression. Nature 1998;394:897–901.

26. Farooqi IS, Matarese G, Lord GM, Keogh JM, Lawrence E, Agwu C,Sanna V, Jebb SA, Perna F, Fontana S, et al. Beneficial effects ofleptin on obesity, T cell hyporesponsiveness, and neuroendocrine/metabolic dysfunction of human congenital leptin deficiency. J ClinInvest 2002;110:1093–1103.

27. Johnston RA, Schwartzman IN, Krishna Murthy GG, Shore SA. Effectof leptin on ozone induced airway hyperreactivity and injury in C57BL/J6 mice [abstract]. Am J Respir Crit Care Med 2002;165:A776.

28. Rajala MW, Scherer PE. Minireview: the adipocyte: at the crossroads ofenergy homeostasis, inflammation, and atherosclerosis. Endocrinology2003;144:3765–3773.

29. Collins S, Kuhn CM, Petro AE, Swick AG, Chrunyk BA, Surwit RS.Role of leptin in fat regulation. Nature 1996;380:677.

30. Ziccardi P, Nappo F, Giugliano G, Esposito K, Marfella R, Cioffi M,D’Andrea F, Molinari AM, Giugliano D. Reduction of inflammatorycytokine concentrations and improvement of endothelial functions inobese women after weight loss over one year. Circulation 2002;105:804–809.

31. Davi G, Guagnano MT, Ciabattoni G, Basili S, Falco A, MarinopiccoliM, Nutini M, Sensi S, Patrono C. Platelet activation in obese women:role of inflammation and oxidant stress. JAMA 2002;288:2008–2014.

32. Haynes WG, Morgan DA, Djalali A, Sivitz WI, Mark AL. Interactionsbetween the melanocortin system and leptin in control of sympatheticnerve traffic. Hypertension 1999;33:542–547.

33. Haynes WG, Morgan DA, Walsh SA, Mark AL, Sivitz WI. Receptor-

968 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

mediated regional sympathetic nerve activation by leptin. J Clin Invest1997;100:270–278.

34. Van Heek M, Compton DS, France CF, Tedesco RP, Fawzi AB, GrazianoMP, Sybertz EJ, Strader CD, Davis HR Jr. Diet-induced obese micedevelop peripheral, but not central, resistance to leptin. J Clin Invest1997;99:385–390.

35. Hausberg M, Morgan DA, Mitchell JL, Sivitz WI, Mark AL, HaynesWG. Leptin potentiates thermogenic sympathetic responses to hypo-thermia: a receptor-mediated effect. Diabetes 2002;51:2434–2440.

36. Correia ML, Haynes WG, Rahmouni K, Morgan DA, Sivitz WI, MarkAL. The concept of selective leptin resistance: evidence from agoutiyellow obese mice. Diabetes 2002;51:439–442.

37. Bjorbaek C, Elmquist JK, Frantz JD, Shoelson SE, Flier JS. Identificationof SOCS-3 as a potential mediator of central leptin resistance. MolCell 1998;1:619–625.

38. Crespo A, Filla MB, Murphy WJ. Low responsiveness to IFN-gamma,after pretreatment of mouse macrophages with lipopolysaccharides,develops via diverse regulatory pathways. Eur J Immunol 2002;32:710–719.

39. Fredberg JJ, Inouye DS, Mijailovich SM, Butler JP. Perturbed equilib-rium of myosin binding in airway smooth muscle and its implicationsin bronchospasm. Am J Respir Crit Care Med 1999;159:959–967.

40. Cooper C, Kuh D, Egger P, Wadsworth M, Barker D. Childhood growthand age at menarche. Br J Obstet Gynaecol 1996;103:814–817.

41. Kaplowitz P. Delayed puberty in obese boys: comparison with constitu-tional delayed puberty and response to testosterone therapy. J Pediatr1998;133:745–749.

42. Heine PA, Taylor JA, Iwamoto GA, Lubahn DB, Cooke PS. Increasedadipose tissue in male and female estrogen receptor-alpha knockoutmice. Proc Natl Acad Sci U S A 2000;97:12729–12734.

43. Jones ME, Thorburn AW, Britt KL, Hewitt KN, Wreford NG, ProiettoJ, Oz OK, Leury BJ, Robertson KM, Yao S, et al. Aromatase-deficient(ArKO) mice have a phenotype of increased adiposity. Proc Natl AcadSci U S A 2000;97:12735–12740.

44. Degano B, Prevost MC, Berger P, Molimard M, Pontier S, Rami J,Escamilla R. Estradiol decreases the acetylcholine-elicited airway re-activity in ovariectomized rats through an increase in epithelial acetyl-cholinesterase activity. Am J Respir Crit Care Med 2001;164:1849–1854.

45. Hamano N, Terada N, Maesako K, Hohki G, Ito T, Yamashita T, KonnoA. Effect of female hormones on the production of IL-4 and IL-13 fromperipheral blood mononuclear cells. Acta Otolaryngol Suppl 1998;537:27–31.

46. Richardson J, Beland J. Nonadrenergic inhibitory nervous system inhuman airways. J Appl Physiol 1976;41:764–771.

47. Rogers, D. F. Motor control of airway goblet cells and glands. RespirPhysiol 2001;125:129–144.

48. Morgan DA, Anderson EA, Mark AL. Renal sympathetic nerve activityis increased in obese Zucker rats. Hypertension 1995;25:834–838.

49. Tataranni PA, Young JB, Bogardus C, Ravussin E. A low sympathoa-drenal activity is associated with body weight gain and developmentof central adiposity in Pima Indian men. Obes Res 1997;5:341–347.

50. Uemura K, Young JB. Effects of fat feeding on epinephrine secretionin the rat. Am J Physiol 1994;267:R1329–R1335.

51. Collins S, Daniel KW, Rohlfs EM, Ramkumar V, Taylor IL, Gettys TW.Impaired expression and functional activity of the beta 3- and beta 1-adrenergic receptors in adipose tissue of congenitally obese (C57BL/6J ob/ob) mice. Mol Endocrinol 1994;8:518–527.

52. Cao W, Luttrell LM, Medvedev AV, Pierce KL, Daniel KW, Dixon TM,Lefkowitz RJ, Collins S. Direct binding of activated c-Src to the beta3-adrenergic receptor is required for MAP kinase activation. J BiolChem 2000;275:38131–38134.

53. Cao W, Medvedev AV, Daniel KW, Collins S. Beta-adrenergic activationof p38 MAP kinase in adipocytes: cAMP induction of the uncouplingprotein 1 (UCP1) gene requires p38 MAP kinase. J Biol Chem 2001;276:27077–27082.

54. Young JB, Morrison SF. Effects of fetal and neonatal environment onsympathetic nervous system development. Diabetes Care 1998;21:B156–B160.

55. Young JB. Programming of sympathoadrenal function. Trends Endocri-nol Metab 2002;13:381–385.

56. Larsson LG, Lindberg A, Franklin KA, Lundback B. Symptoms relatedto obstructive sleep apnoea are common in subjects with asthma,chronic bronchitis and rhinitis in a general population. Respir Med2001;95:423–429.

57. Redline S, Tishler PV, Schluchter M, Aylor J, Clark K, Graham G. Riskfactors for sleep-disordered breathing in children: associations withobesity, race, and respiratory problems. Am J Respir Crit Care Med1999;159:1527–1532.

58. Guilleminault C, Quera-Salva MA, Powell N, Riley R, Romaker A,Partinen M, Baldwin R, Nino-Murcia G. Nocturnal asthma: snoring,small pharynx and nasal CPAP. Eur Respir J 1988;1:902–907.

59. Lean ME. Obesity: burdens of illness and strategies for prevention ormanagement. Drugs Today (Barc) 2000;36:773–784.

60. Netzer NC, Hoegel JJ, Loube D, Netzer CM, Hay B, Alvarez-Sala R,Strohl KP. Prevalence of symptoms and risk of sleep apnea in primarycare. Chest 2003;124:1406–1414.

CorrespondenceAll Roses Are Flowers, But Not All Flowers Are Roses

To the Editor:

We read the editorial by Drs. Cornfield and Haddad (1) onour article (2) with considerable interest. The editorial raisesimportant issues, including the importance of accurate defini-tions for disease states, but it has several misconceptions anderrors that we wish to clarify and correct. Drs. Cornfield andHaddad (1) refer to our study as a trial. Although the informationwas collected as part of a weaning trial, this article is not a reportof a trial. It is a descriptive study across nine centers for 6consecutive months of patients who were mechanically venti-lated for more than 24 hours (2).

Drs. Cornfield and Haddad (1) have used the term respira-tory failure to refer specifically to acute respiratory distress syn-drome (ARDS), which is incorrect. The editorial also incorrectlystates that we were not specific in defining ARDS and that theincidence of ARDS was probably higher than reported. Wedefined ARDS as a PaO2/FiO2 ratio of 200 or less with bilateralinfiltrates and no evidence of heart failure (2). Although 52%of our patients had a PaO2/FiO2 ratio of 200 or less during theirstay, only 7.6% met the criteria for ARDS at 24 hours afteradmission. Timmons and colleagues (3) found that ARDS wasfound in only 39% of patients with acute hypoxemic respiratoryfailure, but they did not include any patient with acute ventilatoryfailure without hypoxia.

Drs. Cornfield and Haddad stated: “Implicit in the objectiveof the trial is a goal of determining the normative incidence andnatural history of respiratory failure in children” (1). We clearlystated that our objective was to describe those children receivingmechanical ventilation for greater than 24 hours across largepediatric referral centers in North America to determine thefeasibility of conducting clinical trials to improve the healthoutcomes of this population (2). The editorial (1) also criticizesour exclusion criteria. We excluded these patients for the expresspurpose of defining a population in which clinical trials may beundertaken. Patients with pulmonary hypoplasia, bone marrowtransplantation, and abnormal vascular tone may not be suitablecandidates for a clinical trial of a therapy to improve outcomesfrom acute respiratory failure. There have been at least threemajor clinical trials in children with respiratory failure (4–6).Our exclusion criteria are consistent with published studies.

Drs. Cornfield and Haddad (1) have criticized us for notincluding trauma patients. That is not correct. Trauma was re-ported in 30 patients (9.6%), and 25 had traumatic brain injury(2). In addition, they point out the difference in mortality ratesbetween our study and that published in the literature (4–6).The studies on high-frequency oscillation and inhaled nitric ox-ide only included patients whose oxygenation index was greaterthan 13 and 15, respectively (4, 6). Clearly, these patients hadsevere disease. The surfactant study had an overall mortality of11.9% for patients with an oxygenation index greater than 7 (5).Our lower mortality rate might have been secondary to inclusioncriteria with a broad range of severity and to improved manage-ment strategies in an era of lung-protective ventilation.

We disagree with the conclusion that the significance of thestudy is diminished. In fact, we would state that it is exactly theopposite for the following reasons. First, it defines a group ofchildren who can be potentially enrolled in clinical trials. Second,it provides an estimate of the population size that can be studied.For example, it would not be feasible to conduct a randomizedcontrolled trial in all children with respiratory failure using mor-

tality as the primary outcome because the mortality is low, espe-cially in infants with bronchiolitis, the single most common diag-nosis in this population of children. If we focus on a subset ofpatients with a higher mortality rate, such as those requiringmechanical ventilation with a clinical diagnosis of ARDS after24 hours of mechanical ventilation, such a study would have toenroll patients from multiple centers to achieve sufficient power.Third, the editorial (1) points out the heterogeneity in the popu-lation requiring acute mechanical ventilator support for morethan 24 hours and shows that half of these children are youngerthan 1 year of age. We believe that our study (2) did meet its in-tended goal: to provide future clinical investigators an estimatefor conducting clinical trials in children with acute respiratoryfailure. In addition, our study provides estimates of individualsubsets of patients with acute respiratory failure who may beeligible for different clinical trials.

Conflict of Interest Statement : S.V., A.R., J.H., P.F., I.C., R.G., and P.L. have nodeclared conflict of interest.

Shekhar VenkataramanAdrienne RandolphJames HansonPeter ForbesIra CheifetzRainer GedeitPeter Luckettfor the Palisi Network

References

1. Cornfield DN, Haddad IY. A rose by any other name is yet a rose: acuterespiratory failure in children. Am J Respir Crit Care Med 2003;168:268–269.

2. Randolph AG, Meert KL, O’Neil ME, Hanson JH, Luckett PM, ArnoldJH, Gedeit RG, Cox PN, Roberts JS, Venkataraman ST, et al. Thefeasibility of conducting clinical trials in infants and children with acuterespiratory failure. Am J Respir Crit Care Med 2003;167:1334–1340.

3. Timmons OD, Havens PL, Fackler JC. Predicting death in pediatric pa-tients with acute respiratory failure. Pediatric Critical Care Study Group.Extracorporeal Life Support Organization. Chest 1995;108:789–797.

4. Arnold JH, Hanson JH, Toro-Figuero LO, Gutierrez J, Berens RJ, AnglinDL. Prospective, randomized comparison of high-frequency oscillatoryventilation and conventional mechanical ventilation in pediatric respira-tory failure. Crit Care Med 1994;22:1530–1539.

5. Willson DF, Zaritsky A, Bauman LA, Dockery K, James RL, Conrad D,Craft H, Novotny WE, Egan EA, Dalton H. Members of the Mid-Atlantic Pediatric Critical Care Network. Instillation of calf lung surfac-tant extract (calfactant) is beneficial in pediatric acute hypoxemic respi-ratory failure. Crit Care Med 1999;27:188–195.

6. Dobyns EL, Cornfield DN, Anas NG, Fortenberry JD, Tasker RC, LynchA, Liu P, Eells PL, Griebel J, Baier M, et al. Multicenter randomizedcontrolled trial of the effects of inhaled nitric oxide therapy on gas ex-change in children with acute hypoxemic respiratory failure. J Pediatr1999;134:406–412.

Fungal Exposure and Lower Respiratory Illnessin Children

To the Editor:

The limitation of data concerning clinically relevant exposuresto agents, including fungi, in the indoor environment has beenidentified as an important problem by scientists studying build-

970 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

ing-related illness (1, 2). Despite the attempts to objectivelycharacterize fungal exposure, the recent study of lower respira-tory illness (LRI) among children by Stark and colleagues (3)appears to have important limitations in exposure assessment.

The method of defining fungal exposure used in that studyplaces emphasis on specific fungal types, but does not allow forany characterization of the total fungal count in the areas beingevaluated. As described in the study by Stark and colleagues(3), a home could have a “high fungal level” when persons inother homes could easily be exposed to greater concentrationsof total fungi. For example, a home with an airborne Aspergilluslevel of 100 cfu/m3 but no other fungi would be considered tohave a “high fungal level,” whereas a home with an Aspergilluslevel of 37 cfu/m3, Cladosporium of 177 cfu/m3, and Penicilliumof 130 cfu/m3 (mean airborne concentrations from the study, yettotaling 240 cfu/m3) would not be “high” because none of thoseconcentrations exceed the 90th percentile. Given the uncertaint-ies concerning mechanisms of illness related to fungi, misclassifi-cation bias would be an important consideration for this typeof exposure assessment.

We agree with the authors that another important limitationof their exposure assessment involves fungal samples being takenonly once at the beginning of the survey, whereas informationon the outcome measures were collected every 2 months for ayear. It remains unclear whether there is any clinical relevanceof measures of fungal exposure taken up to 12 months beforeLRI (or any other health effect).

In the article by Stark and colleagues (3), there appears not tohave been any assessment of exposure to environmental tobaccosmoke (ETS), whether at home or in daycare settings. BecauseETS is known to be associated with increased morbidity in chil-dren (4), some assessment of ETS among the study participantswould be important.

Finally, the authors draw conclusions concerning the relation-ship of measured fungal concentrations in houses with the pres-ence or absence of water damage or visible mold in those houses.The methods of assessment of water damage/visible mold are notdescribed, but the assessment appears to be based on parentalself-report. We question the usefulness of such occupant self-report and suggest that an objective assessment of the indoor en-vironment for moisture (perhaps using a moisture meter) wouldbe more appropriate.

In summary, problems with assessment of exposure to fungiin the indoor environment, such as those pointed out above, arecertainly not unique to the study in question. However, it isimportant to point out these limitations because they call intoquestion the ability of the authors to draw conclusions concern-ing (1) appropriate types of evaluation techniques for indoorenvironments (the need for fungal sampling) and (2) the rela-tionship of illness to exposure to fungi in the indoor environment.

Conflict of Interest Statement : D.B.T. and E.H.P. have no declared conflict ofinterest.

Douglas B. TroutElena H. PageNational Institute for Occupational Safety

and HealthCenters for Disease Control and PreventionCincinnati, Ohio

References

1. Institute of Medicine. Committee on the Assessment of Asthma and In-door Air. Clearing the air: asthma and indoor air exposures. Washing-ton, DC: National Academy Press; 2000.

2. Centers for Disease Control and Prevention. State of the science on moldsand human health: statement of Stephen C. Redd, M.D. Atlanta, GA:

CDC; July 18, 2002. Available from: http://www.cdc.gov/nceh/airpollu-tion/images/moldsci.pdf (accessed August 19, 2003).

3. Stark PC, Burge HA, Ryan LM, Milton DK, Gold DR. Fungal levels inthe home and lower respiratory tract illnesses in the first year of life.Am J Respir Crit Care Med 2003;168:232–237.

4. Mannino DM, Siegel M, Husten C, Rose D, Etzel R. Environmentaltobacco smoke exposure and health effects in children: results from the1991 National Health Interview Survey. Tob Control 1996;5:13–18.

From the Authors:

There is no consensus on a correct manner of characterizingpotential fungal exposure in the home. Each method of estimat-ing exposure through fungal measurement has significant limita-tions that would generally result in an underestimate of healtheffects by attenuating associations (1) (bias to the null). Becauseour starting point was examining the prospective associationsbetween lower respiratory illness (LRI) and the presence ofhome levels of individual taxa that were relatively high for thecohort (�90th percentile for the cohort), it was sensible to followthese analyses by testing the a priori hypothesis that having arelatively high level of any one of the taxa might increase therisk for LRI. Drs. Trout and Page present an equally valid butdifferent hypothesis: that even if one does not have a relatively“high” level of any one taxon, having an absolutely high totalfungal count in the home may be associated with health ef-fects. By examining “total” fungal counts exclusively, one wouldplace emphasis only on the most abundant taxa—taxa that growwell on the plate or are abundant in ambient air or in the carpet/floor coverings. For example, homes with relatively high dustAlternaria levels (90th percentile, 8,333 cfu/g) but low yeasts(90th percentile, 58,000 cfu/g) would not be considered to havehigh fungal levels. We tested the alternative hypothesis thatbeing in the 90th percentile of total fungi (for the taxa presentedin the article [2]) predicted LRI, and found that being in homeswith greater than 90th percentile for total airborne fungi pre-dicted marginally higher risk for LRI (relative risk � 1.46; 95%confidence interval: 1.00–2.15), but being in homes with greaterthan 90th percentile for total dust-borne fungi did not. Thissuggests that in our cohort, having relatively high levels of knowntaxa, many of which have documented irritant or allergenic prop-erties, is more predictive of infant LRI than having absolutelyhigher total levels of fungi, regardless of taxon composition. Anadditional limitation of using total culturable fungi is that manyfungi that could be involved in exposure may be intrinsicallyunculturable, and thus are not included in the total count, consti-tuting a significant confounder for total counts, but not for countsof specific culturable taxa.

As we stated (2), we were limited in that air and dust sampleswere taken only once in the first 2 to 3 months of each child’s lifeand may not represent integrated exposure over the entire firstyear of life. However, it is a reasonable hypothesis that exposureto fungi in the first few months of life may influence the immunesystem or the propensity to respiratory symptoms, or both, overthe first year of life. Moreover, it is possible that dust fungi repre-sent, perhaps more than air fungi, the fungal characteristics ofthe home over longer periods.

We did consider the potential independent or confoundingeffects of environmental tobacco smoke (ETS) (2). Only 7% ofchildren were exposed to ETS at home and 6% in daycare set-tings. Thus, whereas other studies with higher smoking rateshave found reproducible significant effects of ETS on LRI ininfancy, we did not. In univariate analyses, smoking in daycaresettings was marginally (p � 0.12) associated with LRI; but inour final models, the association was less significant and did notchange the magnitude or precision of the estimate of the effectof fungi on risk for LRI.

Correspondence 971

Moisture meters and point measurements of relative humidityare not gold standards for evaluating home dampness (3, 4).With their acknowledged limitations, questions regarding homedampness are a well established and accepted tool used to evalu-ate the reproducible associations between dampness and LRI ininfancy (4). The combination of exposures that home dampnessrepresents is not fully understood. Dampness likely representsother exposures in addition to fungi, and may represent condi-tions leading to fungal exposure not detected through culturablefungal methods (5). In cross-sectional studies, parents who havechildren with symptoms may tend to be biased toward answeringpositively to home dampness questions, but our study was longi-tudinal and less prone to this bias.

In summary, although we recognize the limitations of all studiesof associations of home fungal levels with LRI, we believe thatit is unlikely that the associations that we found between fungiand LRI risk are caused by bias or unmeasured confounders.Confidence in these findings will come if they are reproduciblein other studies (4), and more work is needed to develop feasiblemethods of improving the characterization of early childhoodfungal exposure in large-scale birth cohort studies.

Conflict of Interest Statement : D.G. and P.C.S. have no declared conflict of inter-est. H.A.B. consults for the Sharper Image, which markets an air cleaner designedto address exposure to fungal spores and other airborne contaminants, and re-ceives approximately $6000 per month for these services.

Diane R. GoldChanning Laboratory, Brigham and

Women’s HospitalHarvard Medical SchoolBoston, MassachusettsPaul C. StarkTufts–New England Medical CenterBoston, MassachusettsHarriet A. BurgeHarvard School of Public HealthBoston, Massachusetts

References1. Douwes J, Pearce N. Invited commentary: is indoor mold exposure a risk

factor for asthma? Am J Epidemiol 2003;158:203–206.2. Stark PC, Burge HA, Ryan LM, Milton DK, Gold DR. Fungal levels in

the home and lower respiratory tract illnesses in the first year of life.Am J Respir Crit Care Med 2003;168:232–237.

3. Macher J, Milton DK, Burge HA, Morey P. Bioaerosols: assessment andcontrol. Cincinnati, OH: American Conference of Governmental Indus-trial Hygienists; 1999.

4. Institute of Medicine. Committee on the Assessment of Asthma and In-door Air. Clearing the air: asthma and indoor air exposures. Washing-ton, DC: National Academy Press; 2000.

5. Chew GL, Rogers C, Burge HA, Muilenberg ML, Gold DR. Dustborneand airborne fungal propagules represent a different spectrum of fungiwith differing relations to home characteristics. Allergy 2003;58:13–20.

Toll-like Receptors and Allograft Rejection

To the Editor:

We read with interest the article from Palmer and colleagues(1) demonstrating that recipients heterozygous for a mutationin the toll-like receptor (TLR) 4 gene had reduced acute allograftrejection after lung transplantation. The authors also demon-strated that mutations in the allograft had no impact. It wouldbe of interest to investigate whether transplant recipients withthe mutated TLR4 gene (Asp299Gly or Thr399Ile) had defectsin adaptive immune function after transplantation in additionto reduced rejection rates. Do lymphocytes from these patientsdemonstrate reduced interferon-� production in response to do-

nor antigen? Is there a reduction in the number of donor-specificantibodies in these recipients? These are important questionsas work in experimental models has demonstrated that micethat are deficient in an important TLR signal adaptor protein,MyD88, have impaired priming and production of helper T celltype 1 immune responses in infectious and transplant models(2, 3). In our work using TLR2-, TLR4-, and MyD88-deficientanimals, we demonstrated that allograft rejection of HY incom-patible skin allografts was largely abrogated in the absence ofMyD88 (2). TLR2-deficient animals had a marginal delay, butTLR4 recipients did not manifest reduced acute rejection, incontrast to the work by Palmer and colleagues (1). Furthermore,rejection could be reestablished by restoring MyD88 signalingin either the donor or the recipient. Clearly, there are likely to belarge differences between an experimental model and a clinicalstudy. It would be of interest to investigate whether the TLR4mutation (or any other TLR mutation) impacts other types ofsolid organ transplantation.

Conflict of Interest Statement : D.R.G. and B.M.T. have no declared conflict ofinterest.

Daniel R. GoldsteinBethany M. TesarYale University School of MedicineNew Haven, Connecticut

References

1. Palmer SM, Burch LH, Davis RD, Herczyk WF, Howell DN, ReinsmoenNL, Schwartz DA. The role of innate immunity in acute allograft rejec-tion after lung transplantation. Am J Respir Crit Care Med 2003;168:628–632.

2. Goldstein DR, Tesar BM, Akira S, Lakkis FG. Critical role of the toll-like receptor signal adaptor protein MyD88 in acute allograft rejection.J Clin Invest 2003;111:1571–1578.

3. Schnare M, Barton GM, Holt AC, Takeda K, Akira S, Medzhitov R. Toll-like receptors control activation of adaptive immune responses. NatImmunol 2001;2:947–950.

From the Authors:

We thank Drs. Goldstein and Tesar for their comments regardingour study (1). Goldstein and colleagues recently observed thatskin allograft rejection did not occur in mice with targeted disrup-tion of MyD88, but did occur in those with disruption of toll-like receptor (TLR) 2 or TLR4 (2). In contrast, we found de-creased acute rejection in lung transplant recipients heterozy-gous for either of two mutations in TLR4 previously associatedwith endotoxin hyporesponsiveness (3). There are likely severalimportant explanations for the differences between their animalmodel and our clinical study. First, environmental exposure di-rectly into the allograft makes clinical lung transplant unique.Inhalational exposure to air pollution (including endotoxin), in-fectious agents (such as gram-negative bacteria), and other nox-ious toxins occurs on a regular basis after lung transplantation.As a result, TLR4 may be of particular importance in the initia-tion of innate and adaptive immune responses after lung trans-plantation. Genetic differences in TLR4 signaling, therefore, mightexert a greater influence on posttransplant rejection in lung trans-plant as compared with other organs. Second, in contrast to themurine model used by Dr. Goldstein, in which mice were identi-cal at the major histocompatibility (MHC) loci, almost all humanlung allograft recipients have multiple MHC mismatches withthe donor (4). MHC matching is not performed because of shortcold ischemic times tolerated by lung allografts. The absence ofa significant effect with the TLR4 disruption in the murine modeldoes not address the impact of impaired TLR4 signaling in the

972 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 169 2004

setting of multiple MHC differences. Finally, we were interestedto see that in the study by Goldstein and colleagues (2) therewas a trend toward decreased skin allograft rejection in micewith disruption of TLR4 (p � 0.13), with one TLR4�/�-recipientmouse demonstrating indefinite skin graft survival, providingsome experimental support for our results.

We agree that allograft rejection is a complex biological re-sponse, and further study of the cellular and humoral responseto the allograft in lung transplant recipients with the 299/399polymorphisms is critical. We look forward to pursuing additionalstudies that elucidate the mechanisms by which innate and adap-tive immunity interact in the setting of human organ transplan-tation. Further investigation into the immunogenetics of the allo-immune response may greatly enhance our ability to prevent andtreat clinical rejection. Ultimately, both our clinical study andthe animal work suggest an important and previously unrecog-nized role for innate immunity in the development of allograftrejection.

Conflict of Interest Statement : S.M.P. and L.H.B. have no declared conflict ofinterest. D.A.S. has a patent pending on toll-4 assay and has no other declaredconflict of interest.

Scott M. PalmerLauranell H. BurchDavid A. SchwartzDuke University Medical CenterDurham, North Carolina

References

1. Palmer SM, Burch LH, Davis RD, Herczyk WF, Howell DN, ReinsmoenNL, Schwartz DA. The role of innate immunity in acute allograft rejec-tion after lung transplantation. Am J Respir Crit Care Med 2003;168:628–632.

2. Goldstein DR, Tesar BM, Akira S, Lakkis FG. Critical role of the toll-like receptor signal adaptor protein MyD88 in acute allograft rejection.J Clin Invest 2003;111:1571–1578.

3. Arbour NC, Lorenz E, Schutte BC, Zabner J, Kline JN, Jones M, Frees K,Watt JL, Schwartz DA. TLR4 mutations are associated with endotoxinhyporesponsiveness in humans. Nat Genet 2000;25:187–191.

4. Quantz MA, Bennett LE, Meyer DM, Novick RJ. Does human leukocyteantigen matching influence the outcome of lung transplantation? Ananalysis of 3,549 lung transplantations. J Heart Lung Transplant 2000;19:473–479.

Forces in Emphysema:Newtonian v Quantum Mechanics

To the Editor:

We read with great interest the recent review of the pathobiologyof emphysema by Suki and colleagues (1). In it the authors ad-dressed classic hypotheses regarding the etiologic factors of em-physema—for example, protease–antiprotease, inflammation, me-chanical forces, and collagen deposition. We were particularlyinterested in the authors’ position on the role of mechanicalforces as merely having blunt physical effects on the lung paren-chyma, which overlooks a growing body of evidence that showshow physical forces can have specific molecular physiologic orpathophysiologic effects on the developing and adult lung (2).Workers in our laboratory (3) and others (2) are studying the

influence of mechanical forces on normal lung morphogenesisand its possible role in lung pathology. Suki and colleagues (1)have taken the more conservative position that physical forcesmerely dissect the tissue as if it were a hank of rope under ten-sion (their analogy). We prefer to think of physical forces “mold-ing” the lung during normal morphogenesis, and perhaps explain-ing how varying tension can explain both ventilation–perfusionmatching and the cause of emphysema. The bases for such specu-lation are myriad: lung morphogenesis is highly plastic, literallybeing “molded” by amniotic fluid distension in utero (4); postna-tally, lung structure/function are influenced by physical forcesthat allow the alveolar wall of the lung to remodel under theinfluences of distension and contraction, through a common par-acrine mechanism; and perhaps even wedge resection for thetreatment of emphysema represents a means by which the re-maining healthy, but restricted parenchyma, can be salvaged byallowing it to “stretch.”

Parathyroid hormone–related protein (PTHrP) is a stretch-regulated paracrine hormone that is necessary for normal lungdevelopment; knocking out this gene causes stage-specific failureof lung morphogenesis (5). Recent studies from our laboratoryhave demonstrated the importance of PTHrP signaling betweenepithelium and interstitium for alveolar homeostasis (6); in theabsence of PTHrP signaling, both the interstitial and epithelialcells readapt in a manner mimicking fibrosis. We acknowledgethat this is a minority position, but one that offers the possibilitythat chronic lung disease is part of the continuum of develop-ment, homeostasis, and repair (7), and that as such it may be areversible (i.e., treatable) process.

Conflict of Interest Statement : J.S.T. and V.K.R. have no declared conflict of in-terest.

John S. TordayVirender K. RehanHarbor-UCLA Medical Center Research

and Education InstituteTorrance, California

References1. Suki B, Lutchen KR, Ingenito EP. On the progressive nature of emphy-

sema: roles of proteases, inflammation, and mechanical forces. Am JRespir Crit Care Med 2003;168:516–521.

2. Vlahakis NE, Hubmayr RD. Response of alveolar cells to mechanicalstress. Curr Opin Crit Care 2003;9:2–8.

3. Torday JS, Rehan VK. Mechanotransduction determines the structureand function of lung and bone: a theoretical model for the pathophysiol-ogy of chronic disease. Cell Biochem Biophys 2003;37:235–246.

4. Hooper SB, Harding R. Fetal lung liquid: a major determinant of thegrowth and functional development of the fetal lung. Clin Exp Pharma-col Physiol 1995;22:235–247.

5. Rubin LP, Torday JS. Parathyroid hormone–related protein (PTHrP)biology in fetal lung development. In: Mendelson CR, editor. Endocri-nology of the lung: development and surfactant synthesis. Totowa, NJ:Humana Press; 2000. p. 269–297.

6. Torday JS, Torres E, Rehan VK. The role of fibroblast transdifferentiationin lung epithelial cell proliferation, differentiation, and repair in vitro.Pediatr Pathol Mol Med 2003;22:189–207.

7. Demayo F, Minoo P, Plopper CG, Schuger L, Shannon J, Torday JS.Mesenchymal–epithelial interactions in lung development and repair:are modeling and remodeling the same process? Am J Physiol LungCell Mol Physiol 2002;283:L510–L517.

Dr. Suki was given the opportunity to respond to this letter but declined todo so.