Effects of Smoking Intervention andthe Use of an Inhaled AnticholinergicBronchodilator on the Rateof Decline of FEV1The Lung Health StudyNicholas R. Anthonisen, MD; John E. Connett, PhD; James P. Kiley, PhD; Murray D. Altose, MD; William C. Bailey, MD;A. Sonia Buist, MD; William A. Conway, Jr, MD; Paul L. Enright, MD; Richard E. Kanner, MD; Peggy O'Hara, PhD;Gregory R. Owens, MD; Paul D. Scanlon, MD; Donald P. Tashkin, MD; Robert A. Wise, MD;for the Lung Health Study Research Group

Objective.\p=m-\Todetermine whether a program incorporating smoking interven-tion and use of an inhaled bronchodilator can slow the rate of decline in forced ex-piratory volume in 1 second (FEV1) in smokers aged 35 to 60 years who have mildobstructive pulmonary disease.

Design.\p=m-\Randomizedclinical trial. Participants randomized with equal prob-ability to one of the following groups: (1) smoking intervention plus bronchodilator,(2) smoking intervention plus placebo, or (3) no intervention.

Setting.\p=m-\Tenclinical centers in the United States and Canada.Participants.\p=m-\Atotal of 5887 male and female smokers, aged 35 to 60 years,

with spirometric signs of early chronic obstructive pulmonary disease.Interventions.\p=m-\Smokingintervention: intensive 12-session smoking cessation

program combining behavior modification and use of nicotine gum, with continuing5-year maintenance program to minimize relapse. Bronchodilator: ipratropium bro-mide prescribed three times daily (two puffs per time) from a metered-dose inhaler.

Main Outcome Measures.\p=m-\Rateof change and cumulative change in FEV1over a 5-year period.

Results.\p=m-\Participantsin the two smoking intervention groups showed signifi-cantly smaller declines in FEV1 than did those in the control group. Most of this dif-ference occurred during the first year following entry into the study and was attrib-utable to smoking cessation, with those who achieved sustained smoking cessationexperiencing the largest benefit. The small noncumulative benefit associated withuse of the active bronchodilator vanished after the bronchodilator was discontinuedat the end of the study.

Conclusions.\p=m-\Anaggressive smoking intervention program significantly re-duces the age-related decline in FEV1 in middle-aged smokers with mild airwaysobstruction. Use of an inhaled anticholinergic bronchodilator results in a relativelysmall improvement in FEV1 that appears to be reversed after the drug is discontin-ued. Use of the bronchodilator did not influence the long-term decline of FEV1.

(JAMA. 1994;272:1497-1505)

From University of Manitoba, Winnipeg (Dr An-thonisen); University of Minnesota, Minneapolis (DrConnett); National Heart, Lung, and Blood Institute,Bethesda, Md (Dr Kiley); Case Western Reserve Uni-versity, Cleveland, Ohio (Dr Altose); University of Ala-bama at Birmingham (Dr Bailey); Oregon Health Sci-ences University, Portland (Dr Buist); Henry FordHospital, Detroit, Mich (Dr Conway); University of Ari-zona, Tucson (Dr Enright); University of Utah, Salt LakeCity (Dr Kanner); University of Miami (Fla) (Dr O'Hara);

University of Pittsburgh (Pa) (Dr Owens); Mayo Clinic,Rochester, Minn (Dr Scanlon); University of California\p=m-\Los Angeles (Dr Tashkin); and The Johns HopkinsUniversity, Baltimore, Md (Dr Wise).

A complete list of the members of the Lung HealthStudy Research Group appears at the end of thisarticle.

Reprint requests to Lung Health Study, 2221 Univer-sity Ave SE, Suite 200, Minneapolis, MN 55414 (DrConnett).

CHRONIC obstructive pulmonary dis¬ease (COPD) is a major cause of morbid¬ity and mortality. It occurs almost exclu¬sively in smokers, but affects only a mi¬nority of them,1 indicating that other riskfactors are important. It has a long coursewith considerable deterioration of lungfunction before symptoms develop.2 Ob¬servational studies indicate that the age-related rate of decline of lung function isaccelerated by smoking, and that spon¬taneous smoking cessation is associatedwith a slowing of the rate of decline.2·3However, spontaneous smoking cessationmight imply a powerful selection bias inthat people who decide to stop smokingmay differ in other ways from people whodo not, and these differences may influ¬ence subsequent lung function.

For editorial comment see 1539.

A second risk factor for COPD that hasreceived a great deal ofattention is bron¬chial hyperreactivity. According to the"Dutch hypothesis" of the developmentof COPD, smokers with hyperreactiveairways develop symptomatic disease.4,5This has been supported by several stud¬ies showing more rapid decline of lungfunction in smokers with airways hyper¬reactivity than in those without.6"8 In turn,this raises the question of whethertherapy aimed at reducing airways reac¬

tivity or its presumed consequence, bron-chospasm, might alter the course ofCOPD. There is evidence that the rate ofdeterioration of lung function is reducedby bronchodilator therapy in COPD pa¬tients who respond to bronchodilators.9

at University of Pennsylvania on February 7, 2011jama.ama-assn.orgDownloaded from

The Lung Health Study (LHS) is arandomized multicenter clinical trial car¬ried out from October 1986 to April 1994that was designed to test the effective¬ness of intervention—smoking cessationand bronchodilator administration—inpeople thought to be in the early stagesof COPD. This article reports on theprimary outcome of the LHS: the effectof the interventions on the decline inFEV, in the randomized groups.

METHODSDesign

A detailed description of the LHS de¬sign has been published.10 Briefly, theplan was to evaluate the effect of eachintervention over 5 years in slowing therate of decline of FEV,. Therefore, par¬ticipants were randomly assigned to oneof three groups: usual care (UC) whoreceived no intervention; smoking in¬tervention and the inhaled bronchodi¬lator ipratropium bromide (Atrovent,Boehringer Ingelheim PharmaceuticalsIne, Ridgefield, Conn) (SIA); and smok¬ing intervention and an inhaled placebo(SIP). The LHS was designed to detectdifferences between the SIA and UCgroups in the rate of decline of FEV, onthe order of 7.5 mL/y over 5 years, basedon a one-sided test with significance levela=.05. The projected treatment effectreflected anticipated changes of 15 to 30mL/y in participants who complied withtherapy, but assumed 50% compliancewith bronchodilator therapy and a 24%sustained quit rate in those assigned tosmoking intervention, compared with 7%in UC participants. The mean differ¬ence in FEV! rate of decline betweenthe SIP and UC groups was projectedto be 5 mL/y. The target sample size foreach group was set at 2000 participants.This ensured a statistical power ofabout70% to detect the projected differencebetween SIP and UC groups, and 94%to detect the projected difference be¬tween SIA and UC groups. Thus, a goalwas set of recruiting 6000 participantsfor the entire study, an average of 600at each of the 10 participating clinics.10

Entry CriteriaEntry criteria were chosen to iden¬

tify otherwise healthy smokers, aged 35to 60 years, who were thought to be athigh risk for COPD as indicated by thepresence ofmild airways obstruction andwho did not have conditions that wouldcompromise follow-up or interpretationof lung function. Smoking was definedas use of 10 cigarettes on at least 1 dayduring the 30 days preceding the firstscreening visit. Airways obstruction wasdefined as aratio of FËVj to forced vitalcapacity (FVC) of 70% or less. Partici-

pants also were required to have anFEV, between 55% and 90% ofpredictednormal. Candidates for the study wereexcluded if they gave evidence of otherserious disease that might interfere withfollow-up or influence lung function, orif they regularly used physician-pre¬scribed bronchodilators. These criteriahad the effect of excluding most physi¬cian-diagnosed asthmatics and personswith significant symptoms of COPD. In¬dividuals who reported using ß-blockermedications also were excluded.

Recruitment for the StudyRecruitment has been described in de¬

tail previously.11·12 Screening and enroll¬ment were accomplished in three stages.Initial screening of age-eligible smokersidentified those who approximately metspirometrie criteria. Volunteers were re¬cruited in a variety of ways, includingmedia publicity, mass mailings, and work¬site and public screening. People who metfirst-screen eligibility requirements wereasked to attend a second screening ex¬amination at which lung function eligibil¬ity was definitively determined and thespirometrie response to isoproteronol wasascertained. Exclusionary criteria werechecked at the second screening exami¬nation. Individuals who qualified for thestudy were then asked to attend a thirdscreening examination at which time adetailed history ofrespiratory symptoms,environmental exposures, and smokingstatus was acquired. At that examina¬tion, participants also underwent a metha-choline challenge test and were random¬ized using a separate schedule for eachclinical center, which assured a one-thirdprobability of assignment to each treat¬ment group.

SpirometrySpirometrie methods were specific to

the LHS and have been described indetail elsewhere.13 Spirometry at the ini¬tial screening visits was not rigidly con¬

trolled, but the methods used at the sec¬ond and third screening examinationsand all subsequent clinic visits were care¬

fully standardized and monitored tomaintain quality. All centers used thesame equipment: rolling seal spirometers(Spirotech 500, Spirotech Ine, Atlanta,Ga) interfaced with computers (IBM PC/XT) with special software for qualitycontrol. (The study-specific software was

developed under a subcontract from theLHS data coordinating center by Spi¬rotech Inc.) At the second screening ex¬

amination, which determined lung func¬tion eligibility, three acceptable and tworeproducible maneuvers were requiredfrom up to eight forced expirations. Ac¬ceptable maneuvers were defined asthose with peak expiratory flow within

10% of the maximum observed, a rapidstart, absence ofmajor flow fluctuations,and adequate time of expiration.13·14 Re¬producible maneuvers agreed within10% in terms of peak flow, within thelarger of0.2 L or 5% for FVC, and withinthe larger of 0.1 L or 5% for FEV,.

The FEV, and FVC values taken tocharacterize each participant were themaximum single results from acceptablemaneuvers. The FEV, and FVC valuesbefore and afterbronchodilatorwere com¬

pared to the predicted normal values ofCrapo et al,15 with 12% downward ad¬justment applied for African Americans16and Asians.17 The participant was ineli¬gible if the second screening examinationFEV, was greater than 90% of predictedor less than 55% of predicted, the latterto exclude individuals with clinical COPD.

Bronchodilator response was mea¬sured at the second screening examina¬tion using similar spirometrie criteria.Two inhalations (200^g total dose) ofisoproterenol were administered from ametered-dose inhaler, and spirometrywas repeated after 5 minutes. Responsewas quantified in terms of percent in¬crease of prebronchodilator FEV,. Toassess airways reactivity, methacholinechallenge was carried out at the thirdscreening examination, measuringchanges in FEV, in response to aerosolscontaining diluent (saline, 0.4% phenol,and citric acid buffer) and methacholinein successive concentrations of 1, 5, 10,and 25 mg/mL. Challenges were termi¬nated when the FEV, fell by 20% or

more, or after the 25 mg/mL concentra¬tion. A detailed description ofthe metha¬choline challenge protocol used in thisstudy has been published elsewhere.18Interventions andCompliance Monitoring

The smoking cessation program wasthe same for all clinical centers and hasbeen fully described.19 It began at thetime of randomization, when all SIA andSIP participants were interviewed by a

physician who strongly recommendedthat they stop smoking, explaining in de¬tail that they were at very high risk forsymptomatic COPD. The smoking inter¬vention participants then met in groupsguided by a health educator. Each groupmet 12 times in 10 weeks, the frequencyofmeetings starting at four per week andgradually declining. Emphasis was placedon behavior modification techniques, witha quit day set early in the program. Nico¬tine replacement therapy was used ag¬gressively in the form of nicotine gum(Nicorette, Marion Merrell Dow Ine, Kan¬sas City, Mo) provided at no cost to theparticipants. Those who quit entered amaintenance program aimed at prevent¬ing relapse by teaching coping skills for

at University of Pennsylvania on February 7, 2011jama.ama-assn.orgDownloaded from

problems such as stress and increases inbody weight. Relapses were individuallytreated as soon as they occurred if theparticipant was willing, and formal groupprograms were available to smokers inthe SIA and SIP groups throughout thestudy. Spouses and significant others ofSIP and SIA participants were includedin the cessation program if they wished,and they were treated in the same way as

participants.Both SIA and SIP participants were

instructed in the use of metered-doseinhalers by a physician at the time ofrandomization. This instruction was am¬

plified and reinforced by the health edu¬cators during smoking cessation meet¬ings. Participants were instructed to usetwo puffs of medication three times a

day and to return used inhalers at follow-up visits. The bronchodilator drug cho¬sen for the study was ipratropium bro¬mide because of its low frequency ofside effects, relatively long duration ofaction, and demonstrated bronchodila¬tor effect in COPD.20·21 Ipratropium andplacebo were given to participants indouble-blind fashion, the dose of theformer being 18 µg per puff from themetered-dose inhalers.

The SIA and SIP participants wereasked to return to the clinic at 4-monthintervals to check smoking status, to en¬

courage compliance, and to exchange usedcanisters for new ones. Compliance withsmoking cessation was assessed by self-report and was objectively monitored bymeasuring expired carbon monoxide (ateach visit) and salivary cotinine levels (atannual visits) using an immunoassay tech¬nique (measurements ofsalivary cotininewere carried out by the American HealthFoundation, Naylor Dana Institute forDisease Prevention, Valhalla, NY) withparticipants being classified as smokersif their cotinine levels were greater than20 ng/mL.22 In individuals in whom theuse ofnicotine gum or other nicotine sub¬stitutes invalidated cotinine measure¬

ments, compliance with smoking cessa¬tion was checked by measuring exhaledcarbon monoxide concentrations with an

upper limit of 10 ppm23 defining success¬ful compliance.

Results of salivary cotinine assess¬ments were not revealed to participants,while exhaled carbon monoxide mea¬

surements, which were done at each fol¬low-up visit, were discussed and wereused to counsel participants regardingsmoking cessation. Participants who didnot attend follow-up visits for validationof smoking status were assumed tobe smoking. Compliance with inhalertherapy was assessed by self-report andby weighing used drug or placebo can¬isters returned by the participant. AllSIA and SIP participants were re-

quested to use their inhalers three timesdaily. Satisfactory compliance was de¬fined as reported inhaler use two to fourtimes per day and canister weightchanges consistent with this use.

Follow-upAll participants were asked to return

annually for follow-up. At these visits,FEV, and FVC were measured beforeand after two puffs of isoproteronol (as atthe second screening visit). A detailedquestionnaire regarding respiratorysymptoms and morbidity in the past 12months was administered. In participantswho could not return for clinic visits, thequestionnaire was completed by telephonewhen possible. If a participant gave a

history ofhospitalization, the relevant hos¬pital records were sought. Similar pro¬cedures were applied when participantsdied, every effort being made to docu¬ment the cause of death, which was as¬certained by a group of three expert phy¬sicians who were not otherwise involvedwith the study and who were masked totreatment assignment.

At annual follow-up visits, cross-sec¬tional smoking status was assessed inall participants, self-report being vali¬dated by cotinine or carbon monoxidemeasurements. Nonsmoking was definedas abstinence from all smoked tobaccoproducts (cigarettes, cigars, pipes, andcigarillos) validated by the cotinine/car-bon monoxide data. Sustained nonsmok¬ing was defined as abstinence at each ofannual visits 1 through 5, also validatedby cotinine or carbon monoxide mea¬

surements; participants who did not at¬tend a given annual visit were countedas smokers at that visit. Cross-sectionaland sustained rates of smoking cessa¬tion at annual visits were computed us¬

ing these definitions. Inhaler compliancealso was checked and recorded in theSIA and SIP groups.

For the fifth anniversary, each partici¬pant was asked to return twice to theclinic. At the first ofthese visits, the usualfollow-up protocol was completed as out¬lined herein. At this time, SIA and SIPparticipants discontinued inhaler use. Atleast 40 hours after the initial fifth annualvisit, most participants returned for a re¬

peat methacholine challenge, conductedin the same way as that done just prior torandomization. Individuals with seriousnonrespiratory disease and those whoseFEV, was less than 50% ofpredicted nor¬mal at the first fifth annual visit wereexcused from the second.

Quality ControlConsiderable effort was expended to

ensure high-quality data and strict ad¬herence to protocol by the clinical cen¬ters. Physicians, interviewers, lung fune-

tion technicians, and health educatorsfrom all centers underwent common

training and worked in the LHS onlyafter their performance was shown tobe satisfactory. Each clinical center wasvisited at least twice during the courseof the study by an expert team thatdirectly observed study procedures.Lung function technician performancewas assessed and rated throughout thestudy based on monthly review of spi¬rometry data. Quality control ofspirom¬etry itself was quite rigorous, as hasbeen detailed elsewhere.13Data Collection, Management,and Monitoring

Baseline and follow-up data on smok¬ing history, demographic characteristics,medical history, exposure to dust andfumes, and other information were col¬lected by trained interviewers on stan¬dard study forms. Data on respiratorysymptoms were collected on a version ofthe ATS-DLD [American Thoracic So¬ciety—Division ofLung Diseases] Symp¬toms Questionnaire.24 All forms weremailed to the data coordinating centerfor keying, error checking, and updat¬ing ofcomputer master files. NOMAD225was used as the central data manage¬ment and query language. Spirometrydata were collected by an automatedsystem developed under a contract withSpirotech Inc.13 Data were transmittedregularly from the clinics to the datacoordinating center, using either auto¬mated telephone-transfer methods or

mailing of diskettes.Reports and statistical analyses were

generated using NOMAD2,25 SAS,26BMDP,27 and FORTRAN. Monthly re¬

ports on recruitment, follow-up, smok¬ing cessation, and inhaler use were dis¬tributed to the clinics throughout thestudy.

An independent seven-member safetyand data monitoring board (SDMB) wasestablished to periodically review dataon recruitment, follow-up, adverse con¬

ditions, and the end point data on pul¬monary function changes, mortality, andmorbidity. The SDMB met at 6-monthintervals throughout the operational pe¬riod of the trial. A standing charge tothe SDMB at each of its meetings wasto determine whether the trial shouldcontinue. Possible reasons for early ter¬mination would have been any of thefollowing: (1) adverse conditions, mor¬

tality, or morbidity associated with in¬tervention, (2) a strong treatment ef¬fect, or (3) evidence that no significanttreatment effect would be observed. Sta¬tistical guidelines for early terminationwere established.10 At no time did theSDMB recommend early termination ofthe trial.

at University of Pennsylvania on February 7, 2011jama.ama-assn.orgDownloaded from

Table 1.—Selected Baseline Characteristics by Study Group*Study Group

Baseline CharacteristicsSIA

(n=1961)SIP

(n=1962)UC

(n=1964)Mean (SD) age, y 48.4 (6.8) 48.6 (6.8) 48.4 (6.9)Male, % 60.Í 64.0 63.8Mean (SD) pack-years of cigarette smoking 40.4(19.7) 40.4(18.8)Mean (SD) cigarettes per day 31.2(13.2) 31.5(12.6)Mean (SD) age started smoking, y 17.5(3.9) 17.4(3.9) 17.6(3.8)History of exposure to dust/fumes, % 47.7 47.2 47.3Mean (SD) FEV,, L (prebronchodilator) 2.62(0.61) 2.64 (0.59)Mean (SD) FEV,, L (postbronchodilator) 2.73 (0.64) 2.75 (0.62)Mean (SD) FEV,, predicted % 75.1 (8.8) 75.2 (8.8) 75.1 (8.8)Mean (SD) FEV,/FVC, % 62.9 (5.6) 63.0 (5.5) 62.9 (5.5)Mean (SD) bronchodilator response (percentage of FEV,) 4.2 (5.2) 4.4(5.1) 4.2(5.1)Methacholine reactivity

Percentage with 20% drop in FEV, at 5 mg/mL 34.8 32.7 33.6

Percentage with 20% drop in FEV, at 25 mg/mL 70.8 71.4

Percentage with test at year 5 (of those eligible) 88.9

Physician-confirmed, %Asthma 7.6 7.2 6.8Bronchitis 29.6 28.4

Emphysema 3.4 3.3

*SIA indicates smoking intervention and ¡pratropium bromide (Atrovent); SIP, smoking intervention and placebo;UC, usual care; and FEV,, forced expiratory volume in 1 second.

Statistical MethodsData on the primary end point (change

in FEVi from baseline to fifth annualvisit) were analyzed in several ways.Random-effects models as described byLaird and Ware28 were used to estimatemean rates of decline in FEV! and totest for between-group differences. Es¬timation and testing for the generalLaird-Ware model were implementedusing the procedures BMDP5V27·29 andSAS PROC MIXED.26 The basic ran¬dom-effects model for the primary analy¬sis of LHS data was of the formFEV1(g,ij)=agi + ßgi-tj + Ygi..Xj + 6gy

where g denoted treatment group, i de¬noted the i-th person within the group,j denoted the j-th measurement, t¡ de¬noted time in years from baseline to thej-th annual visit, and Xj=0 for j=0, andXj=l for j>0. Here oc^, ß^, and ^ wereunknown coefficients, unique to each par¬ticipant. The i-th participant's true base¬line FEVi was . The ßgi denoted theparticipant's FEV] slope (milliliters peryear) from annual visit 1 to annual visit5. The inclusion of the indicator variableXj and its coefficient ^ enabled model¬ing FEV! decline for each participant asa broken line, with a one-time changebetween baseline and year 1 representedby Ygj, and linear decline from year 1 toyear 5. (Thus, the pattern of FEVichanges implied by this model may re¬semble an inverted hockey stick. Themodel implies that the angle betweenthe handle and the "business end" of thehockey stick is allowed to vary betweengroups. This angle and the slope of thehandle are estimated from the observeddata.)

A simpler model, linear in time withYgi omitted, did not provide a good fit tothe observed data. The coefficients o:^,ßgi, and Ygi were assumed to follow amultivariate normal distribution withpopulation mean values <xg, ßg, and yg.The objectives of analysis based on thismodel were to estimate the populationmeans ßg and yg for each group and totest for differences between groups inag and yg. As described in the study'soriginal protocol and in the publishedarticle on study design,10 group differ¬ences were assessed using one-sidedtests of significance with oc=.05.

The change-point model just describedis a special case of the Laird-Ware ran¬dom-effects model. Both BMDP5V27 andPROC MIXED in SAS26 permittedanalysis ofdata sets that are incomplete(because of missed visits) using thismodel. However, this analysis gave more

weight to data from participants withcomplete data than to those who mayhave missed visits, resulting in biasedestimates of coefficients in certain cir¬cumstances. For this reason, the pri¬mary end point data also were exam¬ined using an unweighted means analy¬sis: for each participant with at least onevalid follow-up pulmonary function mea¬

surement, a least-squares slope ofFEV,was computed, and mean slopes for thethree treatment groups were comparedusing a one-way analysis of variance(ANOVA). This method may in somecases be less subject to bias than theLaird-Ware analysis.30 As specified inthe study's protocol,10 for the primarybetween-group comparisons, one-sidedtests of significance are reported. For

Table 2.—Follow-up Rates (%) by Study GroupFollow-up, y

I IStudy Group_12 3 4 5

Usual careInterview 95 95 94 93 96Lung function 89 90 89 88 94

Smoking interventionand placebo

Interview 95 94 94 93 97Lung function 92 91 89 88 94

Smoking intervention andipratropium bromide

Interview 96 95 95 94 97Lung function 92 92 90 88 95

other comparisons, two-sided tests and values are reported.

Pairwise differences between groupsin cumulative (baseline to year 5)changes in FEVi were compared usingt tests.

RESULTSOver approximately 2 years, 5887 par¬

ticipants were recruited and random¬ized. Selected characteristics at base¬line are shown in Table 1; baseline dataare considered in detail elsewhere.31 Theaverage age was 48 years, and age dis¬tribution was fairly even between 40and 60 years, with participants aged 35to 39 years comprising only 12% of thestudy. Approximately 37% of the par¬ticipants were women. Relatively mildairways obstruction was present, withan average FEV/FVC ratio of63%, andan FEVi near the lower limit of pre¬dicted normal (average, 75% of pre¬dicted). The average increase in FEViin response to the ß-adrenergic bron-chodilator at the second screening ex¬amination was approximately 0.10 L.Most participants had acquired theirsmoking habit in adolescence and were

smoking on average 31 cigarettes perday at the time of study entry. Nearlyhalf gave a history of exposure to dustand/or fumes. About 30% gave a historyofphysician-diagnosed bronchitis, whichwas much more common than asthma or

emphysema. About a third of the par¬ticipants developed a 20% decrease inFEVi after inhaling methacholine con¬centrations of5 mg/mL, while about 70%showed a similar decline at the highestdose administered (25 mg/mL). Propor¬tionally, more women were randomizedto the SIA group than to the SIP andUC groups, and correspondingly themean baseline FEVi for this group was

slightly lower. There were no signifi¬cant differences among groups for thecharacteristics noted in Table 1. Of thoseparticipants who completed the first partofthe fifth annual visit and were eligiblefor repeat methacholine testing, nearly89% repeated the test.

Table 2 indicates follow-up rates. Atannual visits 1 through 4, spirometrie

at University of Pennsylvania on February 7, 2011jama.ama-assn.orgDownloaded from

»< CO

«Oo.2E <°5? (

50

40

30-

20

10

0

SI Cross-sectional

UC Sustained

12 3 4 5

Follow-up, y

Figure 1.—Cotinine/carbon monoxide-validatedsmoking cessation rate (ordinate) as a function ofyears of follow-up. Usual care (UC) participants(squares) are compared with smoking intervention(SI) participants (circles). Both sustained (closedsymbols) and cross-sectional (open symbols) ces¬sation rates are shown. Nonattenders of follow-upvisits are counted as smokers.

2 3 4

Follow-up, y

Figure 2.—Inhaler compliance as a function of yearsof follow-up. Smoking intervention and placebo(circles) and smoking intervention and ipratropiumbromide (triangles) are shown, with self-reportedcompliance indicated by open symbols and compli¬ance as verified by canister weights indicated byclosed symbols. Participants not attending the visitwere classified as noncompliant.

measurements were done in approxi¬mately 90% of the participants; morethan 94% in all three of the treatmentgroups (SIP, SIA, and UC) had spiro¬metrie measurements done at the fifthannual visit. A somewhat larger frac¬tion of participants completed inter¬views. At the end of the study, thewhereabouts and vital status of only 21participants (0.4%) were unknown.

Figure 1 shows smoking cessation ratesas measured and validated by cotinineand/or carbon monoxide at annual visitsthroughout the study. For the sake ofsimplicity, the SIA and SIP groups arecombined in Figure 1; they were virtu¬ally identical in terms of quit rates. Sus¬tained quit rates denote individuals whohad stopped smoking at the time of theinitial cessation program and maintainedthis status at subsequent annual visits.Inevitably, the number of sustained non-smokers declined during the 5 years of

Table 3.—Deaths Within 5 Years of Randomization

Study GroupI I

Cause of Death SIA SIP UCCardiovascular disease 18 7 12Lung cancer 18 20 19Other 18 17 20Total 54 44 51

*SIA Indicates smoking intervention and ipratropiumbromide (Atrovent); SIP, smoking intervention and pla¬cebo; and UC, usual care.

the study, but still approximated 22% atthe fifth annual visit in the smoking in¬tervention groups as compared with about5% in UC participants. Cross-sectionalquit rates are based on the number ofparticipants who were validated as notsmoking at a given annual visit, irrespec¬tive of their previous status. Cross-sec¬tional quit rates at annual visits wereabout 35% in SIA and SIP participantsand increased only slightly during thestudy, while cross-sectional quit rates inUC participants approximately doubledto about 20% at the fifth annual visit.

Figure 2 shows inhaler complianceduring the study. Self-reported compli¬ance indicates individuals who said theywere using their inhalers at least twicea day and is probably an overestimate oftrue compliance. Conversely, complianceas estimated from canister weights maybe an underestimate of true compliance,because 20% to 30% of participants whodid return for follow-up did not bringtheir canisters to be weighed and 20% to30% of them reported adequate compli¬ance. By either measure, inhaler com¬

pliance did not differ between SIA andSIP groups and declined over time inboth groups. As noted in a previous re¬

port,32 a study of inhaler compliance us¬

ing an electronic monitoring device(Nebulizer Chronolog, Forefront Engi¬neering Corp, Denver, Colo) was car¬ried out with a subset of smoking in¬tervention participants at two of theLHS clinics. The results indicated thatself-report and canister weight tend tooverestimate true levels of compliance.

Serious side effects thought to be pos¬sibly ascribable to inhaler use were un¬

common, occurring in 1.2% of SIA par¬ticipants and 0.8% of SIP participants,not significantly different. These side ef¬fects included cardiac symptoms, hyper¬tension, skin rashes, and urinary reten¬tion.

Of the 5887 participants, 149 died dur¬ing the 5-year follow-up period of thestudy: 54 in the SIA group, 44 in the SIPgroup, and 51 in the UC group (Table 3).These death rates do not differ betweentreatment groups: 22=1.05; P=.59. Morethan half the deaths were due to lungcancer and cardiovascular disease. A pro¬portional hazards model analysis of thetime to event for cardiovascular disease

death indicated a value for comparisonofall three groups of .08. However, therewas a difference in cardiovascular deathsbetween the SIA and SIP groups (Table3) that was nominally significant (P=.03,based on a proportional hazards modelanalysis without adjustment for multipletesting). Cardiovascular death was unre¬lated to inhaler use: of the 18 SIA par¬ticipants who died of cardiovascular dis¬ease, only eight had reported compliancewith inhaler use in the period prior to thefatal event; in the SIP group, five of theseven participants who died of cardio¬vascular disease were compliant. The dif¬ference in cardiovascular disease mortal¬ity also was not due to a clear prepon¬derance of a specific diagnosis in the SIAgroup. The SIA group had 12 deaths as¬cribed to coronary artery disease and sixdeaths secondary to a variety of othercardiovascular diseases, while the SIPgroup had five deaths due to coronaryartery disease and two deaths due to othercardiovascular disease.

There were approximately 400 hos-pitalizations ofparticipants in each yearof follow-up, or about nine hospitaliza-tions per 100 person-years of exposure.Differences among treatment groupswere small and not significant, rangingfrom 8.7 per 100 person-years in the SIPgroup to 9.4 in the UC group.

Figure 3 shows the observed mean

postbronchodilator FEV, for each treat¬ment group at each annual visit. Theinitial postbronchodilator FEV, wasabout 0.02 L lower in the SIA groupthan in the other two groups, reflectingthe slightly larger fraction of women inthe SIA group (Table 1). Mean post¬bronchodilator FEV, in the UC groupdeclined in a linear fashion, while in theSIA and SIP groups, mean FEV, in¬creased in the first year of the study andthereafter declined at a rate close tothat of the UC group.

The data of Figure 3 did not fit a linearmodel as well as a change-point longitu¬dinal model (see "Methods") character¬ized by independent linear changes inFEV, before and after the first year. Theapplication of this model to the observedstudy data is shown in Figure 3. Thechange-point model, as is clear from Fig¬ures 3 and 4, provided a better fit to theobserved data than considering FEV, asa simple linear function of time. The 2statistic for this improvement in fit washighly significant: 26=1131.4; P<.001. Anomnibus 2 test based on the change-pointmodel for differences between the treat¬ment groups in the ensemble of change-point and slope coefficients was also highlysignificant: 24=167.5; P<.001.

Application ofthe change-point modelto the observed data resulted in the fol¬lowing estimated mean changes in post-

at University of Pennsylvania on February 7, 2011jama.ama-assn.orgDownloaded from

2.

2.- 2

¿LL

S 2co S 2

2

2Screen 2,1

Follow-up, y

Figure 3.—Mean postbronchodilator forced expira¬tory volume at 1 second (FEV,) over the course ofthe study in all participants in whom the measure¬ment was made. Circles/dotted line represent thesmoking intervention and placebo group, triangles/solid lines represent the smoking intervention andipratropium bromide group, and squares/dashedline represent the usual care group.

Follow-up, y

Figure 4.—Best-fit model of changes in postbron-chodilator forced expiratory volume at 1 second(FEV,) over the course of the study. FEV, under¬goes a one-time change during the first year offollow-up and then declines linearly, the initialchange being independent of the subsequent de¬cline. Circles are observed means for the smokingintervention and placebo (SIP) group, triangles forthe smoking intervention and ipratropium bromide(Atrovent) (SIA) group, and squares for the usualcare (UC) group. Predicted values for the SIA groupare represented by a solid line, for the SIP group bya dotted line, and for the UC group by a dashed line.

bronchodilator FEV, during the firstyear of follow-up: a 34.3-mL decrease(SE, 4.4 mL) in the UC group; a 11.2-mLincrease (SE, 4.3 mL) in the SIP group;and a 38.8-mL increase (SE, 4.3 mL) inthe SIA group. These estimated changesall differed significantly (P<.005 for eachcomparison). After the first year, post¬bronchodilator FEV, declined in all threegroups at similar rates (Figure 4). Inthe ÜC group, the rate of decline be¬tween the first and fifth annual visit was56.2 mL/y (SE, 1.3 mL/y), while that inthe SIP group was 52.3 mL/y (SE, 1.3mL/y), and that in the SIA group was52.7 mL/y (SE, 1.3 mL/y). An omnibustest for differences between groups inyear 1 to year 5 FEV, rates of changeyielded a value of .07. Pairwise com-

2.9

2.8-

2.7

2.6

2.5

2.4Screen 2 ,1

Follow-up, y

Figure 5.—Mean postbronchodilator forced expira¬tory volume at 1 second (FEV,) for participants inthe smoking intervention and placebo group whowere sustained quitters (open circles) and continu¬ous smokers (closed circles). The two curves di¬verge sharply after baseline.

parisons between groups of this vari¬able gave the following one-sided sig¬nificance levels: UC vs SIP, P=.02; UCvs SIA, P=.03; and SIP vs SIA, P=.40.

The unweighted means analysis ofFEVi rates of decline after annual visit1 yielded similar results: 51.8 mL/y forSIA; 52.2 mL/y for SIP; and 55.3 mL/yfor UC (omnibus comparison from one¬

way ANOVA, P=.25; for all pairwisecomparisons, P>.10).

For participants who completed thefifth annual visit, the mean cumulativechanges in postbronchodilator FEVi dif¬fered sharply between the two inter¬vention groups and the UC group. Av¬erage decreases from baseline to annualvisit 5 were the following: UC, 267 mL;SIP, 209 mL; and SIA, 184 mL. All pair-wise comparisons between groups were

significant (P<.002).COMMENT

Distinct and significant differences be¬tween treatment groups were noted forthe main outcome variable (change inFEVi). These differences appear to bedirectly attributable to the interventionswe used. The treatment groups were wellmatched at baseline (Table 1), and follow-up rates were quite high and similar inthe three groups (Table 2). Relativelygood compliance with intervention, par¬ticularly smoking intervention, was ob¬tained in both the SIP and SIA groups.The rates of 5-year sustained smokingcessation in these groups were amongthe highest ever reported for a majorstudy and far exceeded the rates of sus¬tained cessation in the UC group. Fur¬ther, inhaler use was the same in the SIAand SIP groups, indicating that partici¬pants were effectively masked. As origi¬nally planned, we examined postbroncho¬dilator FEVi as the main outcome vari¬able10 because it is related more stronglyto prognosis in COPD than is prebron-

chodilator FEV,,33·34 but the latter fol¬lowed patterns very similar to those de¬picted in Figures 3 and 4.

If smoking cessation were the cause ofthe difference in longitudinal change inFEV, between the UC group and theSIP group, and if the latter had a greatercessation rate than the former, then simi¬lar but larger changes should have beenapparent when subsets ofindividuals whodid or did not stop were compared. Thiswas in fact the case. About 35% of theSIP group had stopped smoking at thefirst annual visit, and in these individualsthe mean FEV, increased by 57 mL, whilein SIP participants who did not stop smok¬ing the FEV, declined by 38 mL. Afterthe first year, the FEV, declined slightlyfaster in the UC group than in the SIPgroup, and the likelihood that this alsowas due to different rates of smokingcessation was supported by the fact thatcontinuous smokers in the SIP group de¬clined at a rate of 63 mL/y from the firstto the fifth annual visit while sustainedquitters in the SIP group declined at 34mL/y. The cumulative 5-year average de¬cline in FEV, was 267 mL in the UCgroup and 209 mL in the SIP group. Inthe SIP group, sustained quitters had anoverall 5-year decline of 72 mL while con¬

tinuing smokers in this group declined by301 mL (Figure 5). Therefore, there seemslittle question that the differences ob¬served between UC and SIP groups (Fig¬ure 3) were due to differing rates ofsmok¬ing cessation and underestimated the ef¬fects ofsmoking cessation on the averageindividual.

In comparing the UC and SIP groupdata of Figure 3, two features of interestare apparent. First, the annual decline ofFEV, in smokers and nonsmokers hasbeen characterized as approximately lin¬ear,2 and while the mean FEV, in the UCgroup did pursue a course that could bedescribed as linear, that was not the casein the SIP group. Second, after the firstyear, the differences between the groupsin terms of rate of decline were verysmall in spite of the considerably largernumber ofnonsmokers in the SIP group.The reason that the mean FEV, did notdecrease in the SIP group in the firstyear of the study has been alluded toherein. On average, the FEV, increasedslightly in those individuals who stoppedsmoking. This has been observed by oth¬ers,35,36 but since the change is small, it isonly noted in longitudinal data when ces¬sation occurs synchronously, as was thecase in the present study. Rates of de¬cline after the first year were relativelyclose probably because smoking cessa¬tion gradually increased in the UC group(Figure 2), tending to decrease its rate ofdecline, while in the SIP group, the num¬ber of nonsmokers did not increase and

at University of Pennsylvania on February 7, 2011jama.ama-assn.orgDownloaded from

Table 4.—Mean Cumulative Declines (±SD) in Prebronchodilator FEV, (mL)*

Time IntervalUC

(n=1280)

Study GrouptSIP

(n=1448)SIA

(n=1470)Entryt to year 5 (first part) 249 (±236) 188 (±246) 172 (±263)Entn/ to year 5 (second part) 263 (±248) 196 (±257) 208(±270)Year 1 to year 5 (first part) 196 (±220) 169 (±208) 178(±230)Year 1 to year 5 (second part) 210 (±236) 177 (±217) 214(±248)Year 5, first part to second part 14(±141) 9 (±135) 36 (±173)

*Restricted to participants who attended annual visit 1 and both parts of annual visit 5.tUC indicates usual care; SIP, smoking intervention and placebo; and SIA, smoking intervention and ipratropium

bromide (Atrovent).t-"Entry" refers to screen 3.

new cessation approximately equalled re¬

lapse. In the SIP group, rates of declineafter the first year were considerablyless in sustained quitters than in con¬tinuous smokers (Figure 5), which is inagreement with conventional wisdom.

The effects of bronchodilator therapywere evaluated by comparing the SIPgroup with the SIA group. Results in thelatter group differed even more from alinear pattern than did those in the SIPgroup: FEVi increased by an average of40 mL in the SIA group during the firstyear. This does not appear to be accountedfor entirely by smoking cessation, sincecessation rates were very similar in thetwo groups. The obvious conclusion wasthat bronchodilator therapy significantlyincreased the FEVi during the first year.Thereafter, this effect was maintained anddid not increase because, as noted herein,the rate of decline in the SIA group fromannual visit 1 to annual visit 5 was es¬

sentiallythe same as that inthe SIPgroup.The hypothesis that the early increase inthe SIA group was due to the broncho¬dilator agent was supported by the factthat it was particularly notable (47 mL)in SIA participants who claimed to usethe inhaler regularly, compared with SIPparticipants with similar inhaler compli¬ance whose mean FEVi increased by only10 mL during the first year. On the otherhand, in SIP and SIA participants whowere poorly compliant with the inhaler,FEVj declined 20 to 30 mL in the firstyear; these participants also had low ratesof smoking cessation. Beyond the firstyear, the rate of decline of FEV] was 52mL/y in the SIP and SIA groups. Fur¬ther, it was not influenced by inhaler com¬

pliance; in both groups, the rate of declinewas similar whether or not the inhalerswere used by the participants. The mecha¬nism of the difference between SIA andSIP groups was not entirely clear. It ap¬pears unlikely that bronchodilator therapyaltered the underlying disease processbecause there was no progressive differ¬ence between the two groups; they be¬haved very similarly after the first year.Further, after the first year, compliancewith bronchodilator therapy conferred no

apparent advantage in terms of rate ofdecline. The bronchodilator appeared toinduce an early increase in FEV, thatwas maintained at subsequent annual vis¬its. One potential cause of such a changecould be the presence of a residual bron¬chodilator effect at the time of the annuallung function tests. Ifsuch an effect were

present in the SIA group, the submaxi¬mum dose of isoproterenol used duringthe test would have been additive andthe postbronchodilator FEV, would belarger than that measured in the absenceof such a residual effect, as in the SIPgroup. Such a residual bronchodilator ef¬fect would, however, have to be presentin most cases for at least 12 hours be¬cause participants were instructed not touse their inhalers on the day of testing.This is considerably longer than the statedduration of action of ipratropium bro¬mide.20·21 On the other hand, studies ofthe drug's duration of action involvedmany fewerparticipants thanwere in theSIA group and relatively small increasesof FEV, such as those observed herewould have been difficult to discern withreliability.

Whatever the cause, the importantquestion regarding the additional earlyincrease in FEV, seen in the SIA groupis whether it was a permanent change.That is, if the SIA group stopped takingipratropium, would their mean FEV, re¬main higher than that of the SIP group?Analysis ofspirometry performed at base¬line for the final (year 5) methacholinechallenge test (Table 4) afforded the op¬portunity to examine this question, be¬cause to prevent ipratropium from inter¬fering with the challenge, all participantswere required to discontinue ipratropiumat the first year 5 visit, at least 40 hoursprior to the second year 5 visit. The meaninterval between the two year 5 visitsranged from 35 days in the UC group to39 days in the SIA group. At the secondyear 5 visit, postbronchodilator spirom¬etry was not performed, so it was nec¬

essary to compare results from this visitwith previous prebronchodilator data.Mean prebronchodilator FEV, declinedbetween the first and second year 5 visits

in all three groups: 14 mL in the UCgroup, 9 mL in the SIP group, and 36 mLin the SIA group. The mean decline of 36mL in the SIA group was significantlygreater than that observed in either ofthe other two groups (P<.001 for bothcomparisons); SIP and UC did not differsignificantly (P=.25).

The impact of these changes on thecumulative decline in prebronchodilatorFEVi is shown in Table 4, which con¬trasts results acquired about 12 hoursafter being off ipratropium (first visit)with those obtained after at least 40 hoursof abstinence (second visit). The changesin prebronchodilator FEVi from baselineor year 1 to the first year 5 visit were

very close to those previously cited forpostbronchodilator measurements. Thedecrease in FEVi from entry to the firstyear 5 visit was largest in the UC group(249 mL) and larger in the SIP group (187mL) than in the SIA group (172 mL).Change from year 1 to the first year 5visit was similar in the SIA and SIPgroups and was slightly less than that inthe UC group. In the UC and SIP groups,the cumulative change in FEVi from ei¬ther baseline or year 1 to the second year5 visit differed little from those to thefirst year 5 visit. However, for the SIAgroup, there were substantial increasesin the cumulative FEVi changes mea¬sured to the second, as opposed to thefirst, year 5 visit. These results indicatethat the advantage ipratropium bestowedon the SIA group (an increase ofabout 30mL in FEVi in the first year) was re¬versed when SIA participants stoppedthe drug for 40 hours or more. That is, thefirst-year increase in FEVi observed withthe start of ipratropium therapy was fol¬lowed by a comparable decrease whenipratropium was discontinued 5 yearslater.

It should be noted that the above con¬clusion is based on a somewhat incom¬plete data set. About 82% of the partici¬pants who completed the first year 5 visitcompleted the second. Of the remainder,some were ineligible for the final metha¬choline challenge because of poor pulmo¬nary function; this was more common inUC participants (9.0%) than in the SIA(7.0%) or SIP (6.9%) groups. Of those whowere eligible for the second methacholinechallenge, approximately 88% of eachgroup were studied (Table 1). There waslittle evidence that treatmentgroups were

sampled in a biased way; therefore, webelieve that the results of Table 4 are

representative of the group as a whole.Also it should also be noted that a similarmechanism (unexpected persistence ofanacute drug effect) might have accountedfor results appearing to indicate long-termbenefits of bronchodilator therapy inCOPD in a previous study.9

at University of Pennsylvania on February 7, 2011jama.ama-assn.orgDownloaded from

As anticipated, deaths were uncom¬mon during follow-up, but there was an

unanticipated difference in cardiovas¬cular deaths between the SIA and theSIP groups (Table 3). If this differencewere interpreted literally, it would ap¬pear to indicate that ipratropium useincreased cardiovascular mortality. Wedo not believe that this interpretation iscorrect. No relationship was observedbetween reported or canister-weight as¬sessment of inhaler compliance and car¬diovascular mortality, as would havebeen expected if ipratropium were in¬volved. In fact, there were lower levelsof compliance observed in the SIA par¬ticipants who died ofcardiovascular dis¬ease than in their counterparts in theSIP group (eight of 18 vs five of seven;P>.50). The apparent difference in car¬diovascular mortality was statisticallysignificant only when no correctionswere made for multiple testing, which isinappropriate for this study in whichthe purpose of the analysis was to com¬

pare mortality from a variety of causes

among three treatment groups. Analy¬ses of data on cardiovascular morbidity(ie, hospitalizations attributable to car¬diovascular disease) indicated no signifi¬cant difference between the SIA andSIP groups (P>.20), which does not lendsupport to the hypothesis that ipratro¬pium might cause or exacerbate cardio¬vascular disease. Neither data on self-reported cardiovascular events nor ad¬verse effects data provided evidence thatipratropium is associated with increasedrisk of cardiovascular disease. Finally,it is difficult to postulate a crediblemechanism for the mortality difference,since ipratropium is essentially not ab¬sorbed from either the respiratory tractor the gastrointestinal tract.37 There¬fore, we believe that the differences incardiovascular mortality of Table 3 aremost likely attributable to chance, al¬though a causative relationship cannotbe ruled out.

The fact that there were nearly 50%more deaths from lung cancer than fromcardiovascular disease (57 vs 38) in studyparticipants indicates that their baselinelevel of pulmonary obstruction requiredfor study eligibility was an important riskfactor for lung cancer mortality, as is con¬sistent with other reports.38"40

In summary, the LHS recruited a

large group ofsmokers with mild to mod¬erate airways obstruction and followedthem for 5 years. Very high rates offollow-up were achieved, as well as un¬

usually high rates of sustained smokingcessation in the intervention groups. Acomparison of the randomly assignedgroups gives the strongest evidence todate that smoking cessation results insubstantial benefit to lung function. The

benefit, while most evident in the firstyear after cessation, continued to in¬crease over the 5 years of the study(Figure 5). These results strongly sup¬port smoking cessation as the first andmost important clinical intervention insmokers with mild airways obstruction.Bronchodilator therapy produced an in¬crease in FEVi that was evident at theend of the first year and maintained forthe succeeding 4 years, but was not cu¬mulative in that the improvement in lungfunction did not increase with time. Fur¬ther, it appeared that when bronchodi¬lator use was discontinued, the improve¬ment in lung function associated withactive bronchodilator was reversed.These data are best explained by a rela¬tively short-term, pharmacologically me¬diated change in lung function attribut¬able to the bronchodilator. Such an ef¬fect is presumably available at any timeduring the course of disease and sup¬ports the use of bronchodilators forsymptomatic benefit. On the other hand,these results do not support the use ofbronchodilator medications to change thelong-term course of asymptomatic pa¬tients with airways obstruction.

The principal investigators and senior staff of theclinical and coordinating centers, the National Heart,Lung, and Blood Institute, and members of theSafety and Data Monitoring Board are as follows:Case Western Reserve University, Cleveland, Ohio:M. D. Altose, MD (principal investigator), A. F.Connors, MD (co-principal investigator), S. Redline,MD (co-principal investigator), C. Deitz, PhD, andR. F. Rakos, PhD; Henry Ford Hospital, Detroit,Mich: W. A. Conway, Jr, MD (principal investigator),A. DeHorn, PhD (co-principalinvestigator), J. C.Ward, MD (former co-principal investigator), C. S.Hoppe-Ryan, CSW, R. L. Jentons, MA, J. A. Red-dick, RN, and C. Sawicki, RN, MPH; Johns HopkinsUniversity School ofMedicine, Baltimore, Md: R. A.Wise, MD (principal investigator), S. Permutt, MD(co-principal investigator), and C. S. Rand, PhD(co-principal investigator); Mayo Clinic, Rochester,Minn: P. D. Scanlon, MD (principal investigator),L. J. Davis, PhD (co-principal investigator), R. D.Hurt, MD (co-principal investigator), R. D. Miller,MD (co-principal investigator), D. E. Williams, MD(co-principal investigator), G. M. Caron, G. G.Lauger, MS, and S. M. Toogood (Pulmonary Func-tion Quality Control Manager); Oregon Health Sci-ences University, Portland: A. S. Buist, MD (princi-pal investigator), W. M. Bjornson, MPH (co\x=req-\principal investigator), and L. R. Johnson, PhD(LHS Pulmonary Function Coordinator); Universityof Alabama at Birmingham: W. C. Bailey, MD(principal investigator), C. M. Brooks, EdD (co\x=req-\principal investigator), J. J. Dolce, PhD, D. M. Hig-gins, M. A. Johnson, C. D. Lorish, PhD, and B. A.Martin; University of California\p=m-\LosAngeles: D. P.Tashkin, MD (principal investigator), A. H. Coulson(co-principal investigator), H. Gong, MD (formerco-principal investigator), P. I. Harber, MD (co\x=req-\principal investigator), V. C. Li, PhD, MPH (co\x=req-\principal investigator), M. Roth, MD (co-principalinvestigator), M. A. Nides, PhD, M. S. Simmons, andI. Zuniga; University ofManitoba, Winnipeg: N. R.Anthonisen, MD (principal investigator, SteeringCommittee Chair), J. Manfreda, MD (co-principalinvestigator), R. P. Murray, PhD (co-principal in-vestigator), S. C. Rempel-Rossum, and J. M. Stoyko;University ofMinnesota Data Coordinating Center,Minneapolis: J. E. Connett, PhD (principal investi-

gator), M. O. Kjelsberg, PhD (co-principal investi-gator), M. K. Cowles, MS, D. A. Durkin, P. L.Enright, MD, K. J. Kurnow, MS, W. W. Lee, MS,P. G. Lindgren, MS, S. J. Mongin, MS, P. O'Hara,PhD (LHS Intervention Coordinator), H. T. Voelker,and L. A. Waller, PhD; University of Pittsburgh(Pa): G. R. Owens, MD (principal investigator), R. M.Rogers, MD (co-principal investigator), J. J. Johns-ton, PhD, F. P. Pope, MSW, and F. M. Vitale, MA;University of Utah, Salt Lake City: R. E. Kanner,MD (principal investigator), M. A. Rigdon, PhD (co\x=req-\principal investigator), K. C. Benton, and P. M.Grant; Safety and Data Monitoring Board: M. Beck-lake, MD, B. Burrows, MD, P. Cleary, PhD, P. Kim-bel, MD (chairperson; deceased October 27,1990), L.Nett, RN, RRT (former member), J. K. Ockene,PhD, R. M. Senior, MD (chairperson), G. L. Snider,MD, W. Spitzer, MD (former member), and O. D.Williams, PhD; National Heart, Lung, and BloodInstitute staff, Bethesda, Md: S. S. Hurd, PhD(director, Division of Lung Diseases), J. P. Kiley,PhD (project officer), and M. C. Wu, PhD (Divisionof Epidemiology and Clinical Applications); Mortal-ity and Morbidity Review Board: S. M. Ayres, MD,R. E. Hyatt, MD, and B. A. Mason, MD.

This study was supported by contract N01-HR-46002 from the Division of Lung Diseases of theNational Heart, Lung, and Blood Institute. TheSalt Lake City Center has been assisted by theClinical Research Center, Public Health Researchgrant M01-RR00064 from the National Center forResearch Resources.

The following pharmaceutical companies sup¬plied drugs used in this study: Boehringer Ingel-heim Pharmaceuticals Ine, Ridgefield, Conn (Atro-vent and placebo inhalers); Marion Merrell DowIne, Kansas City, Mo (Nicorette, 2 mg).

References

1. O'Connor GT, Sparrow D, Weiss ST. The role ofallergy and nonspecific airway hyperresponsivenessin the pathogenesis of chronic obstructive pulmonarydisease. Am Rev Respir Dis. 1989;140:225-252.2. Fletcher C, Peto R, Tinker C, Speizer FE. TheNatural History of Bronchitis and Emphysema.Oxford, England: Oxford University Press; 1976.3. Xu X, Dockery DW, Ware JH, Speizer FE, FerrisBG Jr. Effects of cigarette smoking on rate of loss ofpulmonary function in adults: a longitudinal assess-ment. Am Rev Respir Dis. 1992;146:1345-1348.4. de Vries K, Booiji H, Goei JT, et al. Hyperre\x=req-\activity of the bronchial tree to drugs, chemicalsand physical agents. In: Orie NJM, Sluiter HJ, eds.Bronchitis II. Assen, the Netherlands: Royal VanGorcum; 1964.5. Sluiter HJ, Ko\l=e"\terGH, de Monchy JGR, PostmaDS, de Vries K, Orie NGM. The Dutch Hypothesis(chronic nonspecific lung disease) revisited. EurRespir J. 1991;4:479-489.6. Barter CE, Campbell AH, Tandon MK. Factorsaffecting the decline in FEV1 in chronic bronchitis.Aust N Z J Med. 1974;4:339-345.7. Campbell AH, Barter CE, O'Connell JM, HugginsR. Factors affecting the decline of ventilatory func-tion in chronic bronchitis. Thorax. 1985;40:741-748.8. Vollmer WM, Johnson LR, Buist AS. Relation-ship of response to a bronchodilator and decline inforced expiratory volume in one second in populationstudies. Am Rev Respir Dis. 1985;132:1186-1193.9. Anthonisen NR, Wright EC, and the IPPB TrialGroup. Bronchodilator response in chronic obstruc-tive pulmonary disease. Am Rev Respir Dis. 1986;133:814-819.10. Connett JE, Kusek JW, Bailey WC, O'Hara P,Wu M, for the Lung Health Study Research Group.Design of the Lung Health Study: a randomizedclinical trial of early intervention for chronic ob-structive pulmonary disease. Control Clin Trials.1993;14:3S-19S.11. Durkin DA, Kjelsberg MO, Buist AS, ConnettJE, Owens GR, and Recruitment Directors of theLung Health Study for the Lung Health StudyResearch Group. Recruitment of participants inthe Lung Health Study, I: description of methods.

at University of Pennsylvania on February 7, 2011jama.ama-assn.orgDownloaded from

Control Clin Trials. 1993;14:20S-37S12. Connett JE, Bjornson-Benson WM, Daniels K,for the Lung Health Study Research Group. Re-cruitment ofparticipants in the Lung Health Study,II: assessment ofrecruiting strategies. Control ClinTrials. 1993;14:38S-51S.13. Enright PL, Johnson LR, Connett JE, VoelkerH, Buist AS. Spirometry in the Lung Health Study,1: methods and quality control. Am Rev Respir Dis.1991;143:1215-1223.14. American Thoracic Society. Statement: stan-dardization of spirometry, 1987 update. Am RevRespir Dis. 1987;136:1285-1298.15. Crapo RO, Morris AH, Gardner RM. Referencespirometric values using techniques and equipmentthat meet ATS recommendations. Am Rev RespirDis. 1981;123:659-664.16. American Thoracic Society. Lung function test-ing: selection of reference values and interpretivestrategies. Am Rev Respir Dis. 1991;144:1202-1218.17. Oscherwitz O, Edlavitch S, Baker TR, JarboeT. Differences in pulmonary functions in variousracial groups. Am J Epidemiol. 1972;96:319-327.18. Tashkin DP, Altose MD, Bleecker ER, et al.The Lung Health Study: airway responsiveness toinhaled methacholine in smokers with mild to mod-erate airflow limitation. Am Rev Respir Dis. 1992;145:301-310.19. O'Hara P, Grill J, Rigdon MA, et al. Design andresults of the initial intervention program for theLung Health Study. Prev Med. 1993;22:304-315.20. Gross NJ. Sch 1000: a new anticholinergic bron-

chodilator. Am Rev Respir Dis. 1975;112:823-828.21. Tashkin DP, Ashutosh K, Bleecker ER, BrittEJ, et al. Comparison of the anticholinergic bron-chodilator ipratropium bromide with metaproter-enol in chronic obstructive pulmonary disease: a

90-day multi-center study. Am JMed. 1986;81(suppl5A):81-89.22. Haley NJ, Axelrad CM, Tilton KA. Validationof self-reported smoking behavior: biochemicalanalyses of cotinine and thiocyanate. Am J PublicHealth. 1983;73:1204-1207.23. Ruth KJ, Neaton JD. Evaluation of two bio-logical markers of tobacco exposure. Prev Med.1991;20:574-589.24. Ferris BG. Epidemiologic standardization project.Am Rev Respir Dis. 1978;118(6, pt 2):1-88.25. NOMAD2 Reference Manual, Version 4.50.Norwalk, Conn: Must Software International; 1988.26. SAS/STAT User's Guide, Version 6. 4th ed.Cary, NC: SAS Institute Inc; 1993.27. BMDP Statistical Software Manual. Berkeley:University of California Press; 1990;2:1207-1243.28. Laird NM, Ware JH. Random effects modelsfor longitudinal data. Biometrics. 1982;38:963-974.29. Jennrich RI, Schluchter MD. Unbalanced re-peated-measures models with structured covari-ance matrices. Biometrics. 1986;42:805-820.30. Wu MC, Bailey K. Analysing changes in thepresence of informative right censoring caused bydeath and withdrawal. Stat Med. 1988;7:337-346.31. Buist AS, Connett JE, Miller RD, et al. ChronicObstructive Pulmonary Disease Early Interven-

tion Trial: baseline characteristics of randomizedparticipants. Chest. 1993;103:1863-1872.32. Rand CS, Wise RA, Nides M, et al. Metered-dose inhaler adherence in a clinical trial. Am RevRespir Dis. 1992;146:1559-1564.33. Burrows B, Earle RH. Prediction of survival inpatients with chronic airways obstruction. Am RevRespir Dis. 1969;99:865-871.34. Anthonisen NR, Wright EC, Hodgkin JE, IPPBTrial Group. Prognosis in chronic obstructive pul-monary disease. Am Rev Respir Dis. 1986;133:14-20.35. McCarthy DS, Craig DB, Cherniack RM. Ef-fect of modification of the smoking habit on lungfunction. Am Rev Respir Dis. 1976;104:103-113.36. Buist AS, Nagy JM, Sexton GJ. The effect ofsmoking cessation on pulmonary function: a 30\x=req-\month follow-up of two smoking cessation clinics.Am Rev Respir Dis. 1979;120:953-957.37. Massey KL, Gotz VP. Ipratropium bromide.Drug Intell Clin Pharmacol. 1985;19:5-12.38. Kuller LH, OckeneJ, Meilahn E, Svendsen KH.Relation of forced expiratory volume in one second(FEV1) to lung cancer mortality in the MultipleRisk Factor Intervention Trial (MRFIT). Am JEpidemiol. 1990;132:265-274.39. Skillrud DM, Offord KP, Miller RD. Higherrisk of lung cancer in chronic obstructive pulmo-nary disease. Ann Intern Med. 1986;105:503-507.40. Tockman MS, Anthonisen NR, Wright EC, Don-ithan MG. Airways obstruction and the risk forlung cancer. Ann Intern Med. 1987;106:512-518.

at University of Pennsylvania on February 7, 2011jama.ama-assn.orgDownloaded from