Influenza epidemiology—past, present, and future

Philippe R. S. Lagace-Wiens, MD, DTM&H, FRCPC; Ethan Rubinstein, MD, LLB; Abba Gumel, PhD

As the world was preparing fora pandemic influenza in 2008caused by H5N1—the birdflu, a highly lethal but non-

human-to-human transferrable influ-enza that raged through SoutheastAsian countries, Egypt, and other coun-tries—H1N1, the swine flu, emerged. Itappeared first in Mexico and quickly,over the course of 3 months, was re-ported from practically all countries,regardless of the season. As of October11, 2009, which was the end of the firstwave of this epidemic and the begin-ning of the second wave, close to400,000 laboratory-identified H1N1 in-fluenza cases and �4735 deaths had

been reported to the World Health Or-ganization. In addition to the humancosts, the economic burden of theH1N1 influenza is substantial and anadditional load on the present brittleworld economy. In Mexico alone, at thebeginning of the epidemic, every dayled to a loss of $57 million, and theWorld Bank estimates that the pan-demic may cost as much as $3 trillion.

Despite the progress in molecular bi-ology and vaccine research, the produc-tion of H1N1 vaccine is still occurring inchicken eggs, with a substantial delay inthe production of adequate amounts ofvaccine material for the entire world’spopulation. In addition, the research onadjuvant treatment has been slow, whichalso contributes to the paucity of ourdefense tools. Luckily, the H1N1 virus isnot highly pathogenic; therefore, thenumber of deaths and severe diseases hasbeen of low magnitude. In addition, theH1N1 is slow in acquiring resistance tooseltamivir; to date, only approximately30 strains were reported as oseltamivir-resistant, allowing for effective treatmentif therapy is started early in the course ofthe disease.

The H1N1 flu is not limited to hu-mans and has been reported in herds inCanada, Norway, and sporadically in theUnited States, defining this virus aszoonosis, or rather humano-zoonotic.In this article, we describe some of theepidemiologic features of H1N1 andmake some predictions regarding itssecond wave.

MATERIALS AND METHODS

Influenza: Virology, AntigenicEvolution, and Ecology

Influenza viruses are members of the or-thomyxovirus viruses that encode a seg-mented RNA genome. There are three groupsof influenza virus: influenza A, influenza B,and influenza C (1, 2). Influenza B and Cviruses are associated with low-level sporadicdisease and limited outbreaks and are nevercauses of pandemic influenza (2). On the otherhand, influenza A is responsible for most sea-sonal influenza and all known pandemics (2).Only influenza A is discussed here.

Influenza A virus contains an RNA genomecomprising eight RNA segments encoding 11genes. Of these, the hemagglutinin (HA),which binds to the cellular receptor sialic acid,and neuraminidase (NA), which cleaves sialicacid residues from budding viruses, are of par-ticular importance to the epidemiology of in-fluenza (1–3). Other genes, including nucleo-protein, M1 (matrix), M2 (ion pore), NS1, NS2,PA, PB1, PB1-F2, and PB2, encode for proteinscritical for structure, reproduction, and viru-lence, and may also be used as diagnostictargets by polymerase chain reaction or anti-gen detection (e.g., M1 or nucleoprotein) andmay be studied to predict virulence in certainnonhuman hosts (2–4).

HA and NA genes encode surface proteinsthat are involved in attachment to and bud-ding of the virus from host cells. They are theprimary antigenic proteins of the virus. Fif-teen serologically distinct HA and nine dis-tinct NA proteins have been identified and aresequentially named H1, H2, H3, and so on,and N1, N2, and so on (2, 3). Of these, only H1,H2, and H3 in combination with N1 and N2

From Department of Medical Microbiology and In-fectious Diseases (PRSL-W), Faculty of Medicine, Uni-versity of Manitoba and Clinical Microbiology Depart-ment, Saint-Boniface General Hospital, DiagnosticServices of Manitoba, Manitoba, Canada; Departmentof Medical Microbiology (ER), Internal Medicine and theSection of Infectious Diseases, Faculty of Medicine,University of Manitoba and Section of Infectious Dis-eases, Department of Internal Medicine, Health Sci-ences Centre, Manitoba, Canada; Institute of IndustrialMathematical Sciences (AG), Faculty of Science, Uni-versity of Manitoba, Manitoba, Canada. We acknowl-edge the expertise of C. N. Podder and O. Sharomi inhelping with simulations, and we thank S. Mahmud foruseful comments on model design.

Dr. Rubinstein holds stock options and has con-sulted for BiondVax Pharmaceuticals, and he receivedhonoraria from Merck. The other authors have notdisclosed any potential conflicts of interest.

For information regarding this article, E-mail:plagacewiens@sbgh.mb.ca

Copyright © 2010 by the Society of Critical CareMedicine and Lippincott Williams & Wilkins

DOI: 10.1097/CCM.0b013e3181cbaf34

In April 2009, Mexican, American, and Canadian authoritiesannounced that a novel influenza virus with pandemic potentialhad been identified in large segments of the population. Withinweeks, it became apparent that the world was dealing with thefirst influenza pandemic in >40 yrs. Despite the unpredictablenature of influenza severity and spread in the pandemics of the20th century, understanding the epidemiology of the past pan-demics and current influenza pandemic will help prepare physi-cians, hospitals, and governments to predict and prepare for thesubsequent waves and subsequent pandemics. We present asummary of the biology that predisposes influenza to cause

sudden pandemics, as well as a summary of the epidemiology ofthe 20th century pandemics. We also report on the epidemiology,disease severity, and risk factors for severe disease and intensivecare admission from the first wave of the current pandemic(April–August 2009). Last, we provide a mathematical modelbased on transmission dynamics of the H1N1 influenza virus thatmay provide some guidance in terms of disease incidence andhospital impact. (Crit Care Med 2010; 38[Suppl.]:S000–S000)

KEY WORDS: influenza; epidemiology; pandemic; H1N1; severity;mortality; hospitalization

01Crit Care Med 2010 Vol. 38, No. 3 (Suppl.)

typically cause disease in humans (1). Theremaining tend to be zoonotic, causing dis-ease in fowl and nonhuman mammals (1, 2).The human specificity is largely attributable tothe receptor specificity of hemagglutinin totarget structures of sialic acid present in hu-man cells (1). Both these surface proteins un-dergo significant variation over the course oftheir replication and during repeated epidem-ics as a result of influenza’s error-prone RNApolymerase (1, 3, 5–7). Over the course ofepidemics, insertions, deletions, and changesin the sequence of these genes result in poly-morphisms of NA and HA structures, althoughthey are of insufficient magnitude to changetheir antigenic nomenclature; these polymor-phisms are termed antigenic shift (1, 3, 7). Inother words, despite these mutations, the pri-mary antigenic properties of the NA or HAprotein in question are still maintained (e.g.,H1), but these mutations result in partial lossof host immunity within a given population(2, 8). Antigenic shift, which may result inpandemic spread because of lack of partialimmunity within the population, occurs whena sufficiently antigenically distinct HA (with orwithout a new NA) emerges (1, 6). Antigenicshift may occur either as a result of a reassort-ment event between human-adapted influenzaand animal-adapted strains within a coinfectedhost, resulting in a progeny virus capable ofsustained human transmission, or if a newhuman-adapted HA arises when an animal oravian influenza A virus is transmitted withoutreassortment from an animal reservoir to hu-mans (1, 3, 4). Neither the HA nor the NAneeds to be “numerically” different from theantigens circulating seasonally. The only pre-requisite for antigenic shift (and therefore apotential for pandemic spread) is that little orno immunity from previous influenza infec-tions or vaccination exists in the population.For example, the 2009 pandemic H1N1 strainis significantly antigenically different from theprevious circulating strains (including “sea-sonal” H1N1) (9).

The sustained survival of influenza withina population and its epidemic potential areentirely dependent on the evolution of its HAand NA genes. As population immunity in-creases as a result of natural infection or vac-cination, antigenic shifts sufficient to negatepreexisting host immunity drive continuedsporadic and epidemic infection but are insuf-ficient to generate pandemics (1, 3, 6–8, 10).Antigenic shift, however, produces a virus towhich preexisting immunity is very low orabsent, thereby exposing a very large suscep-tible population, potentially resulting in apandemic.

Influenza ecology is complicated by thepresence of avian and mammalian clades ofvirus, relaxed host specificity, and the afore-mentioned antigenic evolution. Several avian

species serve as large reservoirs for influenza Aviruses (1, 3). In general, these viruses appearto be in evolutionary stasis with their hostsand tend to cause asymptomatic infections.However, certain viruses that cause asymp-tomatic infection in shorebirds or waterfowlcan cause fatal infection in other avian species,such as domestic poultry (1). Avian species arepoorly adapted to human infection becausetheir hemagglutinins typically do not havehigh affinity for mammalian sialic acid resi-dues. As a result of this, avian-adapted virusescannot cause sustained human outbreaks orpandemics without appropriate spontaneousgenetic alterations that result in human hostadaptations (1, 3). Avian viruses can infect alimited range of other mammals, includingseals, whales, horses, and pigs (1). Pigs may bereadily infected with both avian and mamma-lian strains because of the presence of bothavian-like and mammal-like sialic acid resi-dues in their tracheal epithelium (1). Pigs,therefore, may serve as a reservoir for inter-mingling and reassortment of mammalian andavian species, potentially resulting in anti-genic shift. This potential for the pig to func-tion as a “mixing vessel” has been one of thereasons for the significant interest in Chinaand Southeast Asia as a source of past andfuture pandemics, where intermingling of hu-mans, domestic fowl, and pigs occurs at anintensity not seen anywhere else on earth. Pigsalso serve as reservoirs for swine influenza.Classic swine influenza causes mild influenza-like illness in pigs and, because the virus ismammalian, it can be transmitted to humans(1, 3). However, infections are typically lim-ited to persons with close contact to pigs anddo not result in large-scale outbreaks or pan-demics. Despite the swine HA or NA genesegments of swine influenza viruses that haveadapted to human hosts by genetic rearrange-ments, the use of the term “swine influenza”in not appropriate.

DISCUSSION

Epidemiology of Seasonal (Inter-Pandemic) Influenza

The incidence of seasonal influenzatypically increases in the late autumn andbegins to decline in mid spring. In theNorthern hemisphere, this correspondsto November through March; in theSouthern Hemisphere, this correspondsto April through September (1, 6). Intropical countries, influenza occurs spo-radically throughout the year, but moreso in the rainy periods (1, 3, 6, 8). Local-ized outbreaks of seasonal influenza alsooccur in inter-pandemic years, particu-larly when strains of virus penetrate com-munities with little or no preexisting im-

munity to the circulating virus. Thereason for seasonality remains unclear.Because the primary mode of transmis-sion is by large droplet aerosols, in-creased crowding in the colder months,the return to schools and university dor-mitories, and the start of military recruitcourses have been suggested as contrib-uting factors (1–3, 6). Fomites may serveas a secondary mode of transmission, andit has also been suggested that the higherintensity of sterilizing ultraviolet light inthe summer months may serve to reducethe environmental burden of virus. Dryenvironments, such as those that prevailduring the winter months, are alsoknown to increase transmission for un-known reasons (6). Higher serum vita-min D levels and other immunomodulat-ing factors associated with ultravioletlight levels have also been suggested asfactors influencing influenza attack rates(8). The intensity of seasonal influenzavaries from year to year and largely de-pends on the size of the susceptible pop-ulation, which in turn depends on thedegree of antigenic drift that has oc-curred in the previous seasons (6, 11).Both previous natural infections and vac-cination reduce the susceptible host pop-ulation dramatically if neutralizing anti-bodies to the virus are present, but asignificant antigenic drift results in amore susceptible population and a corre-spondingly higher incidence of disease(3). Other factors, including overcrowd-ing, crowded sleeping arrangements, andunhygienic living conditions with pooraccess to hand hygiene, can contribute tolocalized pockets of increased incidence(3, 6). This partially explains the higherincidence of seasonal influenza observedin colleges, individuals living in low so-cioeconomic conditions, daycare centers,and military settings.

All immunologically susceptible indi-viduals exposed to seasonal influenza maybecome infected. Average attack ratesduring epidemics range from 10% to 20%but may be as high as 40% to 50% inparticularly susceptible populations (1,3). The attack rate is typically highest inschool-age children and daycare popula-tions (1–3, 7). This probably representsthe higher intensity of transmission be-haviors in this population and a relativelylow rate of immunity. In seasonal out-breaks, age-specific incidence generallyfollows a predictable course, with chil-dren being affected early in epidemics,followed by their adult caregivers, and,last, the elderly (1). Although immunity

02 Crit Care Med 2010 Vol. 38, No. 3 (Suppl.)

is the major factor in incidence of dis-ease, severity and morbidity of seasonalinfluenza are dependent on several hostfactors. The average influenza mortalityin developed countries is approximately12 in 100,000 persons, but certain popu-lations experience significantly highermorbidity and mortality from seasonalinfluenza (6). These include those at theextremes of age, with a particularly steepincrease in age-specific mortality after 64yrs of age, and those with asthma andother chronic pulmonary diseases, car-diovascular diseases, diabetes, liver cir-rhosis, immunodeficiency states, hemo-globinopathies, malignancy, and renaldysfunction (2, 3, 7, 12, 13).

Outbreaks of interpandemic influenzagenerally follow a predictable course.They begin abruptly, peak within 2 to 3wks, and have a total duration of 5 to 10wks. This pattern is also typically ob-served with pandemic influenza, al-though a much higher incidence is typi-cally noted, and second or subsequentwaves of similar duration and intensityare commonly observed (1–3, 6).

Epidemiology of PastPandemics

Three major pandemics were recordedin the 20th century: the 1918 to 1919pandemic (H1N1), the 1957 to 1958 pan-demic (H2N2), and the 1968 pandemic(H3N2) (1). Each had a unique epidemi-ologic pattern and origin.

1918 to 1919 Pandemic: H1N1“Spanish” Influenza

The geographic origin of the 1918pandemic virus remains unclear. Accord-ing to medical historians and despite thepandemic’s name, the most probable or-igin is either China or in military campsin the United States soon after the returnof soldiers from the European front (1, 4,14, 15). What is more certain is the phy-logenic origin of the virus, elucidatedthrough sequencing of the virus found inhuman tissues preserved from 1918. Us-ing drift modeling, it appears that theancestor virus first penetrated mamma-lian hosts (pigs) between 1882 and 1913,and the 1918 pandemic virus originatedfrom a human/swine H1N1 reassortmentevent that led to an effective humantransmission in approximately 1915 (16).However, the earliest reliable descriptionof the 1918 pandemic influenza was inmilitary camps during March 1918, al-

though recent reports also suggest a firstwave in New York in February to April1918 (15). Most accounts suggest thatthis “first wave” had a relatively low inci-dence of nonsevere clinical disease with alimited spread. In August 1918, the dis-ease pattern suddenly changed, withwidespread reports of severe influenzacompatible clinical disease that emergedsimultaneously in North America, Africa,and Europe (1). The global “second wave”peaked in October 1918, when school ab-senteeism was approximately 40% ac-cording to some reports (17). This wavewas followed by a smaller “third wave” inFebruary 1919 (17). Although the attackrates and age-specific incidence rates ofthis pandemic did not differ significantlyfrom subsequent pandemics, this pan-demic was characterized by its particu-larly high rates of morbidity and mortal-ity, especially among young adults (1,17). The reason for the high mortalityrates remain unclear, although recentanalysis of genetic reconstruction of the1918 strain suggest that the virus itselfwas more virulent and had a propensityto cause viral pneumonia, as well as aclinical picture compatible with a “cyto-kine storm” (1, 4, 18, 19). Controversyremains regarding whether the virus it-self, or its strong association with postvi-ral bacterial pneumonia, was the majorcause of death in 1918 (19, 20). The over-all evidence today supports that bothmechanisms, direct viral virulence andpostviral superinfection, played majorroles in causing the high mortality. Theexact molecular mechanisms behind thehypervirulence are not fully known, butin vitro genetic recombinants of H1N11918 and modern seasonal strains sup-port an important role of the genes PB1,NA, and HA in the virulence of the 1918virus in animal models (4, 18, 19). It hasalso been suggested that the NS1 proteinof the 1918 strain had the ability to in-hibit interferon activity, leading to moresevere viral infection (4).

1957 to 1958 Pandemic: H2N2Asian Influenza

This pandemic was somewhat bettercharacterized than the 1918 pandemicbecause of more reliable intercontinentalcommunication and the ability of labora-tories to isolate the virus in cell cultures.It appears the virus originated inGuizhou, China, in approximately Febru-ary 1957, and spread rapidly throughoutthat country (1). The epidemic reached

Hong Kong in April 1957, and the caus-ative virus was definitively isolated in Ja-pan in May 1957. The virus was unique inthe human population because of thepresence of different HA and NA antigens.Until then, the major circulating strainswere descendants of the 1918 H1N1strain. By November 1957, the outbreakreached pandemic proportions (1). In theNorthern Hemisphere, peak incidenceoccurred in October and a second wavewas observed in January 1958. Like otherinfluenza epidemics, age-specific attackrates were highest in children aged 5 to19 yrs and, although excess mortality wasnoted in both waves, no 1918-like excessmortality among younger adults was ob-served (1). Studies similar to those thatestablished the origin of the 1918 pan-demic virus support the notion that theH2N2 virus originated from a single re-assortment even between avian H2N2 andhuman H1N1 (1).

1968 Pandemic: H3N2 HongKong

The Hong Kong influenza pandemicwas first noted in Hong Kong in July1968 (1) and appeared to spread globallysomewhat more slowly than the previouspandemics, reaching the United States inDecember 1968/January 1969 and Europe1 yr later. This pandemic had the lowestexcess mortality of the 20th century in-fluenza pandemics, probably because ofpartial immunity to the N2 component ofthe virus from previous circulating H2N2in much of the population (1). Interest-ingly, the H2N2 virus became extinct af-ter the emergence of H3N2 in the humanpopulation, an observation further sup-porting cross-immunity. Like the H2N2virus, the H3N2 virus appears to haveoriginated from a reassortment betweenavian H3-containing viruses and the hu-man H2N2 (1).

Epidemiology of the 2009 H1N1Pandemic

In April 2009, near-simultaneous re-ports surfaced of an epidemic of severeinfluenza-like illness in various parts ofMexico and a report of influenza infectionin three children in the southwest UnitedStates caused by an H1N1 strain of influ-enza closely related to domestic swineinfluenza (21). None of the children hadexposure to swine, and sequencing of theisolated virus later confirmed that all theMexican cases were caused by the same

03Crit Care Med 2010 Vol. 38, No. 3 (Suppl.)

novel H1N1 strain. Although the misno-mer “swine flu” was initially used, thevirus was clearly capable of human-to-human transmission and was later moreappropriately named swine-origin H1N1influenza virus, human-adapted swineH1N1 influenza virus, novel H1N1, or2009 H1N1 (contrasting it from 1918H1N1).

Origins

Despite modern epidemiologic tech-niques and communication, the true geo-graphic origin of the novel H1N1 virusremains a mystery and is of limited rele-vance. Although the virus appears to haveemerged first in Mexico (more specificallyin San Luis Potosi) in late February 2009(personal communication, Celia Alpuche,Mexican Institute of Diagnostic and Epi-demical Reference, Mexico City, Mexico),this likely represents the first large-scaletransmission of this virus, which isreadily supported by the nature of subse-quent global outbreaks, with initial caseshaving exposure or links to Mexico. How-ever, like the 1918 pandemic, the novelH1N1 strain or a similar swine-adaptedvirus may have been circulating at lowlevels among humans for months, or lesslikely years, before February 2009; it alsomay have been circulating among mam-mals (pigs) for years earlier. Molecularanalysis suggests that the common an-cestor to the human 2009 pandemicH1N1 probably emerged in humans be-tween August 2008 and January 2009,whereas the nearest common ancestor inpigs has probably been circulating in pigsfor at least a decade (22–24). Becauseworldwide pandemic influenza surveil-lance has focused on Asian avian strains,little work was performed to identify andtype strains circulating in pigs, leading tothe likelihood that a closely related an-cestor swine H1N1 virus existed long be-fore its emergence as a human-adaptedstrain in February 2009. Unlike other in-fluenza viruses, the genetic origin of thisvirus is better understood. Complete se-quence analysis reveals that the virus is atriple-reassortment virus, with elementsoriginating from multiple establishedswine viruses (22, 25, 26). Although Eur-asian swine, North American swine,avian, and human components are in-cluded in the genome, these componentswere already present in ancestor triplereassortment swine viruses. Therefore,the virus certainly originated from pigsand has likely emerged as a result of

reassortment between swine-adapted vi-ruses containing human and avian ele-ments (23). Lack of systematic swine sur-veillance allowed for the undetectedpersistence and evolution of this poten-tially pandemic strain to evolve for manyyears.

2009 H1N1 Pandemic:Descriptive Epidemiology of the“First Wave”

As noted, the epidemic was first de-scribed in Mexico in April 2009 (26–28).The first wave of the epidemic followedthe course typical to influenza, with ex-ponentially increasing incidence fromApril 9 to April 24, peaking on April 26,and declining rapidly to effectively zeroby May 25 (29). The duration of the firstwave, therefore, was approximately 12wks. The confirmed rate of infection wasapproximately five per 100,000 but is pre-sumably much higher in reality, becausemany individuals had subclinical diseaseand therefore were not tested (29). Age-specific incidence of confirmed cases washighest among children aged 0 to 14 yrs(29). Age-specific confirmed case mortal-ity of 55.7% suggested higher mortalityrates among adults aged 30 to 59 yrs,although this case fatality rate is likely tohave been grossly inflated by biased sam-pling (29).

After World Health Organization notifi-cation of the increased incidence of influ-enza cases in Mexico, countries worldwideimplemented enhanced surveillance. Sub-sequently, outbreaks of 2009 H1N1 werereported in Canada, the United States, theMiddle East, East Asia, and Europe (30–32). The first appearance of cases outsideof Mexico corresponded with travel toMexico, supporting the notion that Mex-ico was the initial site of sustained hu-man-to-human transmission (26, 33–35).In Canada, the epidemic followed acourse similar to that in Mexico, begin-ning in mid April, peaking on June 10,and effectively disappearing by early Julyfor a total duration of approximately 11wks (30, 36). In the United States, theepidemic reached a peak approximatelyJune 20 and a nadir approximately Au-gust 25 before increasing again (31). Thesituation in Europe mirrored that inNorth America with the outbreak startingslightly later (late April) and reaching anadir in 10 to 12 wks (32). The globalspread of influenza was very rapid, touch-ing every country by early July.

Contrary to countries in the NorthernHemisphere, some countries in theSouthern hemisphere reported somewhatlater appearance of the virus (early to midJune in New Zealand and Australia; lateJune in South Africa), with somewhatlonger duration in the Southern autumn.The wave lasted approximately 18 wks inAustralia and 13 wks in New Zealand andSouth Africa (37–40).

Early during the enhanced surveil-lance for influenza, patient characteris-tics could be accurately pinpointed.These data indicated that patients atgreatest risk for infection were younger,primarily younger than 20 yrs of age,with no sex predilection, and with rela-tively low rates of morbidity and mortal-ity (29–31, 35, 41–43).

Attack Rates and Estimates ofTrue Incidence

Worldwide, confirmed attack rateswere highest in younger age groups, andthe average age of cases increased overthe course of the epidemic, as was typicalof previous influenza outbreaks (33, 42,44, 45). It became clear during the earlypart of the pandemic that the case sever-ity, hospitalization, and mortality rateswere low, lower than seasonal influenzain some countries, but excess hospitaliza-tions were noted throughout the affectedareas because of the very high incidenceof disease (30–32). Although diagnostictesting was used extensively in developedcountries, attack rates and true incidenceare difficult to ascertain because of un-der-testing of less symptomatic cases andthe lack of a reliable method for the ret-rospective serological diagnosis of H1N1infection. Therefore, the incidence of lab-oratory-confirmed cases or even clinicalcases represent a gross underestimationof the total cases because most were min-imally symptomatic, and serologicalstudies were severely hampered by sero-logical cross-reactivity with vaccine andseasonal strains. Although reference levelserologic tests can differentiate betweenpandemic H1N1 and infection or vaccina-tion with other H1N1 viruses, these arenot practical or cost-effective to use out-side of research settings. Estimates of at-tack rate and incidence can be made us-ing epidemiologic techniques andmathematical modeling on the basis ofepidemic curves. These are discussedlater.

04 Crit Care Med 2010 Vol. 38, No. 3 (Suppl.)

Co-Circulation of OtherInfluenza Viruses

Co-circulation of other influenza vi-ruses is important for two reasons. When asingle strain of influenza is circulating,anti-viral susceptibility is usually predict-able. For example, seasonal H3N2 strainsare widely susceptible to neuraminidase in-hibitors, whereas seasonal H1N1 strains areresistant (31). However, if co-circulation isoccurring, it may be impossible to providedefinitive therapy before genotyping. Sec-ond, when co-circulation is occurring, re-assortment between viruses is possible, po-tentially leading to new, more virulent, ormore resistant viruses.

During the first wave of the H1N1pandemic, the presence of co-circulatingvirus was variable from country to coun-try. In much of the temperate NorthernHemisphere, minimal or no co-circula-tion occurred with H1N1 2009 (30, 31).In the United States, where a second wavehas begun, �99% of circulating virusesare H1N1 2009 (31), and in Canada�97% of influenza viruses circulating areH1N1 2009 as of mid-October (36). In theSouthern Hemisphere, in countries out-side of the Americas, reports of co-circulation were more common duringthe first wave. Australia (40, 46), NewZealand (46, 47), and South Africa (46,48) all reported significant co-circulation

with H3N2 for significant parts of thefirst wave. However, as the epidemic un-folded, H1N1 2009 virtually replaced allother influenza A types and became thedominant strain circulating globally (32).

Epidemiology of Resistance toAdamantanes andNeuraminidase Inhibitors

Two classes of drugs are available for thetreatment of influenza A. Adamantanes(amantadine and rimantadine) function byinhibiting the ion pump M2, thus interfer-ing with viral cell invasion, and neuramin-idase inhibitors (oseltamivir, zanamivir,peramivir), which prevent neuraminidaseactivity, thereby preventing release of infec-tious virus particles from infected cells (2).Well defined mutations in the M2 gene orNA gene confer resistance to these agents.Resistance to adamantanes is class-specific,whereas resistance to oseltamivir does nottypically confer resistance to zanamivir(49). Surveillance is ongoing worldwide todetect the early development of resistancein the H1N1 2009 pandemic strain. To date,virtually all H1N1 2009 pandemic strainshave been resistant to adamantanes (31).More than 10,000 viruses have been testedfor resistance to oseltamivir worldwide and�1% have been found to be resistant. As ofOctober 14, 2009, 31 cases of oseltamivir-resistant H1N1 2009 infections have

been reported, and all had the H275Ymutation conferring resistance (30,31). No strain to date has been resistantto zanamivir (30 –32).

Second Wave in the NorthernHemisphere

Whereas pandemic activity is cur-rently subsiding in the Southern Hemi-sphere, a second wave of H1N1 influenzahas begun in some areas of the NorthernHemisphere. In particular, the UnitedStates and parts of Europe are reportingincreasing incidence since early Septem-ber 2009 (30). Mexico reports a secondepidemic wave started in early October(30). Population demographics and sever-ity have been similar to those observed inthe first wave, with the virus primarilyaffecting children and young adults in theearly phase of the epidemic; low morbid-ity and mortality rates continue to bereported in most populations (31). Be-tween August 30 and October 3, 2009, theUnited States reported a total of 3874laboratory-confirmed, influenza-associ-ated hospitalizations, with 240 laborato-ry-confirmed H1N1 2009-associateddeaths (31). In Canada, where activity waslower than in the United States, 50 hos-pitalizations and six deaths were reportedin the same period, although some deathsare likely attributable to illness duringthe first wave (30). Influenza activity dur-ing this second wave is �99% attribut-able to H1N1 2009 and remains generallysusceptible to oseltamivir (31).

Severe Respiratory Infection

Severe respiratory infection is variablydefined in the literature. Tangible defini-tions include case hospitalization rates,case mortality rates, and case intensivecare unit (ICU) admission rates. Becausetotal cases cannot be determined reliably,and testing practices change over thecourse of the epidemic, it is impossible todetermine true rates. However, mostcases resulting in hospitalization will beinvestigated and total laboratory-con-firmed influenza hospitalizations, ICU ad-missions, and deaths provide an epidemi-ologic surrogate for severe infections. Bymid-September, Canada had recorded1459 hospitalizations, of which 288(19.7%) were ICU admissions and 76(5.2%) resulted in death (30). Both ad-mission and, to a lesser extent, deathswere strongly associated with the influ-enza epidemic curve, accelerating in mid-

Figure 1. Total hospitalizations (left axis) and deaths (right axis) by epidemic week, starting April 26,2009 (week 17), in Canada. Data from FluWatch (30).

Table 1. Select characteristics of patients with laboratory confirmed H1N1 2009 influenza in Canada

Total Cases,n � 7107

Hospitalized,n � 1504

Admitted to ICU,n � 295

Deaths,n � 76

Females, % 51.9 51.4 56.3 60.5Median age 18 23 37 50Aboriginal status, % 12.5 17.5 15.3 11.8Comorbidities, % 39.2 (728/1859) 58.5 (576/988) 68.6 (151/220) 78.9 (45/57)Pregnancy, % 5.0 (87/1724) 28.1 (77/274) 19.7 (15/76) 28.6 (4/14)

Data from FluWatch (30).

05Crit Care Med 2010 Vol. 38, No. 3 (Suppl.)

June. Deaths were somewhat spread outacross the epidemic, mostly as a result ofintensive care practices and prolongedaggressive care before death (Fig. 1).Similar observations were noted in theUnited States, with 9079 hospital admis-sions by September 3, of which 593(6.5%) resulted in death (31).

Demographics and Risk Factorsfor Severe Infections

The demographics of severe infectiondiffer somewhat from those of uncompli-

cated infection. Overall illness with 2009H1N1 seems to be in keeping with sea-sonal influenza, affecting primarily chil-dren and young adults, affecting malesand females equally, and showing mini-mal predilection for people with underly-ing comorbidities. Contrary to that, se-vere illness shows a unique age, gender,and risk-factor predisposition (50 –53).The proportion of female patients af-fected, the median age, and the propor-tion of cases with underlying medicalconditions increased with severity of ill-ness (Table 1). In a recent publication,

98.2% of critically ill patients with H1N12009 had at least one comorbidity (52).Those comorbidities most frequently re-ported were chronic lung disease(41.1%), obesity (33.3%), hypertension(24.4%), diabetes (20.8%), immunosup-pression (19.6%), neurologic disease(15.5%), cardiac disease other than hy-pertension (14.9%), and pregnancy(7.7%) (52). In addition, certain popula-tions, including aboriginals, were at highrisk for severe disease in several countries(30, 39, 52). Age distribution of death alsowas not typical to interpandemic influ-enza. The age distribution of population-adjusted mortality in laboratory-con-firmed influenza H1N1 2009 is skewed tomiddle-aged individuals, rather than thevery young and very old, creating a moreW-shaped mortality curve reminiscent ofthe 1918 H1N1 pandemic (Fig. 2).

Modeling the Next Wave

To provide information on the impactof the second wave of influenza, modelscan be constructed using known charac-teristics of transmission (e.g., reproduc-tive number) and estimated rates of se-vere illness, ICU admissions, and deaths.One such model is presented here.

The modeling component of this studyis based on the design and use of a newdeterministic compartmental model forthe transmission dynamics of H1N1 inthe population. The total population ofindividuals in the province of Manitoba issplit into a number of mutually exclusivecompartments depending on their infec-tion (or risk of infection) status. We con-sider 15 compartments of the following:susceptible individuals, vaccinated indi-viduals, latently infected individuals, in-fectious individuals without diseasesymptoms, high-risk symptomatic indi-viduals in the early stage (first 2 days) ofinfection, low-risk symptomatic individu-als in the early stage (first 2 days) ofinfection, high-risk symptomatic individ-uals in the later stage of infection, low-risk symptomatic individuals in the laterstage of infection, high-risk treated in-fected individuals, low-risk treated in-fected individuals, high-risk hospitalizedindividuals not in the ICU, low-risk hos-pitalized individuals not in the ICU, high-risk hospitalized individuals in the ICU,low-risk hospitalized individuals in theICU, and recovered individuals. Themodel takes the form of the deterministicsystem of nonlinear differential equa-

Figure 2. Total deaths (left axis) and age-adjusted mortality (AAM, right axis) by age category in theUnited States. The highest age-adjusted mortality was in the 50- to 64-yr group, whereas the highestabsolute mortality was in the 25- to 49-yr group. Data from Centers for Disease Control andPrevention, Atlanta, Georgia (31).

Figure 3. Number of hospitalized individuals in Manitoba (population 1.1 million) according to variousvaccine effect start times. Numbers assume a 10% seroprevalence before the wave.

06 Crit Care Med 2010 Vol. 38, No. 3 (Suppl.)

tions, similar to those for pandemic avianinfluenza (54–56).

The model is parameterized using ep-idemiologic and demographic data for theprovince of Manitoba, Canada, but shouldallow for approximate estimates to be ex-trapolated to other regions. For compar-

ison, the population of Manitoba is ap-proximately 1.1 million.

Main modeling assumptions includesecond wave begins first week of October2009; high-risk individuals more likely tohave severe illness develop, require hos-pitalization, and ICU admission in com-

parison to low-risk individuals; mass vac-cination commences October 26, 2009(vaccine efficacy �80%); at least 10% to20% of the total population have previousimmunity (attributable to first-wave in-fection); first wave reproduction numberis approximately 1.3, as reported by oth-ers (57–60); and the R0 of the secondwave is higher (1.9 before vaccine effect).

Figure 3 depicts the time series of thenumber of hospitalized individuals; thescenario is that 10% of the total popula-tion are assumed to have infection-acquired immunity for various timeswhen the vaccine impact takes effect. It isevident that the peak, which is expectedto occur at the end of November or earlyDecember, increases with increasing du-ration of time before the vaccine impactis felt. Furthermore, the figure showsthat the pandemic would run until lateJanuary or early February 2010. It alsodemonstrates dramatically the impactthat timely vaccination can have on thecourse of the pandemic. Similar plots aredepicted for the ICU admissions (Fig. 4)and H1N1-induced mortality (Fig. 5).This study estimates that where 10% ofthe total population has previous immu-nity, the province of Manitoba could havebetween 946 and 2223 hospitalizations,194 and 459 ICU cases, and 45 and 108H1N1-induced mortalities, depending onwhen the vaccine impact occurs. How-ever, simulating a similar scenario inwhich population immunity resultingfrom the first wave is higher at 20% pre-dicts a lower disease burden: 436 to 849hospitalizations, 90 to 175 ICU admis-sions, and 21 to 41 deaths. In summary,this study projects that the burden of thesecond wave of the H1N1 pandemic couldbe at least three times that of the firstwave, and that the second wave may lastuntil early 2010. In the absence of sero-logical surveys, it is difficult to predictthe seroprevalence of disease after thefirst wave, but studies in New Zealandhave predicted a seroprevalence after thefirst wave to be between 10% and 20%(39). The simulation results obtained aresensitive to changes in the parametersand initial values used, and although thesimulations provide some idea of the out-come of various scenarios, cautionshould be used in interpreting the re-sults.

CONCLUSION

The first influenza pandemic of the21st century has provided us with insight

Figure 4. Number of intensive care unit (ICU) admissions in Manitoba (population 1.1 million)according to various vaccine effect start times. Numbers assume a 10% seroprevalence before the wave.

Figure 5. Modeled mortality in Manitoba (population 1.1 million) according to various vaccine effectstart times. Numbers assume a 10% seroprevalence before the wave.

07Crit Care Med 2010 Vol. 38, No. 3 (Suppl.)

into the fundamental epidemiologic char-acteristics of influenza and has affordedus a much greater understanding of theimpact of pandemic influenza on thehealthcare system and on individuals. Al-though the reasons for the enormousvariability of influenza pandemic severityand spread remain elusive, it is certainthat anunderstanding of the basic epide-miologic principles, pathophysiology, andnatural history of this pandemic, as wellas those of history, will serve to help allphysicians—from the ICU to the hallwaysof public health—provide the best possi-ble individual and population-level inter-ventions to reduce the impact of thistruly formidable pathogen. In addition topresenting data on the impact of previouspandemics and the first wave of the currentpandemic, we also have used a determinis-tic compartmental model for the transmis-sion dynamics of H1N1 in the population topredict the impact of the next wave. Ourpredictions suggest that vaccination andexisting natural immunity from the firstwave will have a significant impact on thedisease and severe disease incidence andthat the peak is likely to occur approxi-mately in early December 2009. The waveis expected to last into early 2010.

REFERENCES

1. Cox NJ, Subbarao K: Global epidemiology ofinfluenza: Past and present. Annu Rev Med2000; 51:407–421

2. Treanor JJ: Influenza virus. In: Principlesand Practice of Infectious Diseases. Vol. 2.Mandell GL, Bennett JE, Dolin R (Eds). Phil-adelphia, PA. Elsevier, 2005, pp 2060–2085

3. LaForce FM, Nichol KL, Cox NJ: Influenza:Virology, epidemiology, disease, and preven-tion. Am J Prev Med 1994; 10(Suppl):31–44

4. Taubenberger JK: The origin and virulenceof the 1918 “Spanish” influenza virus. ProcAm Philos Soc 2006; 150:86–112

5. The macro-epidemiology of influenza vacci-nation in 56 countries, 1997–2003. Vaccine2005; 23:5133–5143

6. Monto AS: Epidemiology of influenza. Vac-cine 2008; 26(Suppl 4):D45–D48

7. Stephenson I, Zambon M: The epidemiologyof influenza. Occup Med (Lond) 2002; 52:241–247

8. Cannell JJ, Zasloff M, Garland CF, et al: Onthe epidemiology of influenza. Virol J 2008;5:29

9. Garten RJ, Davis CT, Russell CA, et al: Anti-genic and genetic characteristics of swine-origin 2009 A(H1N1) influenza viruses circu-lating in humans. Science 2009; 325:197–201

10. Nguyen-Van-Tam JS, Hampson AW: The ep-idemiology and clinical impact of pandemicinfluenza. Vaccine 2003; 21:1762–1768

11. Monto AS, Comanor L, Shay DK, et al: Epi-demiology of pandemic influenza: use of sur-veillance and modeling for pandemic pre-paredness. J Infect Dis 2006; 194(Suppl 2):S92–S97

12. Barker WH, Mullooly JP: Impact of epidemictype A influenza in a defined adult popula-tion. Am J Epidemiol 1980; 112:798–811

13. Sabin AB: Mortality from pneumonia andrisk conditions during influenza epidemics.High influenza morbidity during nonepi-demic years. JAMA 1977; 237:2823–2828

14. Reid AH, Taubenberger JK: The origin of the1918 pandemic influenza virus: a continuingenigma. J Gen Virol 2003; 84(Pt 9):2285–2292

15. Olson DR, Simonsen L, Edelson PJ, MorseSS: Epidemiological evidence of an earlywave of the 1918 influenza pandemic in NewYork City. Proc Natl Acad Sci U S A 2005;102:11059–11063

16. Dos Reis M, Hay AJ, Goldstein RA: Using non-homogeneous models of nucleotide substitu-tion to identify host shift events: Application tothe origin of the 1918 “Spanish” influenza pan-demic virus. J Mol Evol 2009; ●●●

17. Frost WH: The epidemiology of influenza.1919. Public Health Rep 2006; 121(Suppl1):149–159

18. Pappas C, Aguilar PV, Basler CF, et al: Singlegene reassortants identify a critical role forPB1, HA, and NA in the high virulence of the1918 pandemic influenza virus. Proc NatlAcad Sci U S A 2008; 105:3064–3069

19. McAuley JL, Hornung F, Boyd KL, et al:Expression of the 1918 influenza A virusPB1–F2 enhances the pathogenesis of viraland secondary bacterial pneumonia. CellHost Microbe 2007; 2:240–249

20. Brundage JF, Shanks GD: Deaths from bac-terial pneumonia during 1918–19 influenzapandemic. Emerg Infect Dis 2008; 14:1193–1199

21. Update: swine influenza A (H1N1) infec-tions—California and Texas, April 2009.MMWR Morb Mortal Wkly Rep 2009; 58:435–437

22. Trifonov V, Khiabanian H, Rabadan R: Geo-graphic dependence, surveillance, and ori-gins of the 2009 influenza A (H1N1) virus.N Engl J Med 2009; 361:115–119

23. Smith GJ, Vijaykrishna D, Bahl J, et al. Ori-gins and evolutionary genomics of the 2009swine-origin H1N1 influenza A epidemic.Nature 2009; 459:1122–1125

24. Schnitzler SU, Schnitzler P: An update onswine-origin influenza virus A/H1N1: A re-view. Virus Genes 2009; ●●●

25. Itoh Y, Shinya K, Kiso M, et al: In vitro andin vivo characterization of new swine-originH1N1 influenza viruses. Nature 2009; 460:1021–1025

26. Dawood FS, Jain S, Finelli L, et al: Emer-gence of a novel swine-origin influenza A(H1N1) virus in humans. N Engl J Med 2009;360:2605–2615

27. Lopez-Cervantes M, Venado A, Moreno A, et

al: On the spread of the novel influenza A(H1N1) virus in Mexico. J Infect Dev Ctries2009; 3:327–330

28. Dominguez-Cherit G, Lapinsky SE, MaciasAE, et al: Critically ill patients with 2009influenza A(H1N1) in Mexico. JAMA 2009;●●●

29. Update: novel influenza A (H1N1) virus in-fection—Mexico, March-May, 2009. MMWRMorb Mortal Wkly Rep 2009; 58:585–589

30. Flu Watch Weekly Reports: http://www.phac-aspc.gc.ca/fluwatch/08-09/index-eng.php. Ac-cessed October 14, 2009

31. Flu activity and surveillance: http://www.cdc.gov/flu/weekly/fluactivity.htm. Accessed Oc-tober 14, 2009

32. Pandemic H1N1 2009: http://www.who.int/csr/disease/swineflu/en/. Accessed October14, 2009

33. Epidemiology of new influenza A (H1N1) vi-rus infection, United Kingdom, April–June2009. Euro Surveill 2009; 14:●●●

34. Swine-origin influenza A (H1N1) virus infec-tions in a school—New York City, April 2009.MMWR Morb Mortal Wkly Rep 2009; 58:470–472

35. Update: infections with a swine-origin influ-enza A (H1N1) virus–United States and othercountries, April 28, 2009. MMWR Morb Mor-tal Wkly Rep 2009; 58:431–433

36. Surveillance—H1N1 influenza virus. http://www.phac-aspc.gc.ca/alert-alerte/h1n1/surveillance-eng.php. Accessed October 15,2009

37. National Institute for Communicable DiseasesCommunique: http://www.nicd.ac.za/pubs/communique/2009/NICDCommSept09Vol08_09.pdf. Accessed October 15, 2009

38. Lopez L: New Zealand Weekly InfluenzaReport: http://www.surv.esr.cri.nz/PDF_surveillance/Virology/FluWeekRpt/2009/FluWeekRpt200941.pdf. Accessed October15, 2009

39. Baker MG, Wilson N, Huang QS, et al: Pan-demic influenza A(H1N1)v in New Zealand:the experience from April to August 2009.Euro Surveill 2009; 14:●●●

40. Australian influenza surveillance summaryreport: http://www.healthemergency.gov.au/internet/healthemergency/publishing.nsf/Content/ozflu2009.htm/$File/ozflu-no21-2009.pdf. Accessed October 15, 2009

41. Peiris JS, Poon LL, Guan Y: Emergence of anovel swine-origin influenza A virus (S-OIV)H1N1 virus in humans. J Clin Virol 2009;45:169–173

42. Kelly H, Grant K: Interim analysis of pan-demic influenza (H1N1) 2009 in Australia:surveillance trends, age of infection and ef-fectiveness of seasonal vaccination. EuroSurveill 2009; 14:●●●

43. Human infection with new influenza A(H1N1) virus: Mexico, update, March–May2009. Wkly Epidemiol Rec 2009; 84:213–219

44. Shimada T, Gu Y, Kamiya H, et al: Epidemi-ology of influenza A(H1N1)v virus infectionin Japan, May–June 2009. Euro Surveill2009; 14:●●●

08 Crit Care Med 2010 Vol. 38, No. 3 (Suppl.)

45. Kelly HA, Grant KA, Williams S, et al: Epi-demiological characteristics of pandemic in-fluenza H1N1 2009 and seasonal influenzainfection. Med J Aust 2009; 191:146–149

46. Mott J: Novel influenza A (H1N1) epidemiologyupdate—International. Advisory Committee onImmunization Practices Meeting. Centers forDiseases Control and Prevention; 2009

47. Surveillance for the 2009 pandemic influ-enza A (H1N1) virus and seasonal influenzaviruses—New Zealand, 2009. MMWR MorbMortal Wkly Rep 2009; 58:918–921

48. National Institute of Communicable Dis-eases: www.nicd.ac.za. Accessed October 15,2009

49. Wang MZ, Tai CY, Mendel DB: Mechanism bywhich mutations at his274 alter sensitivity ofinfluenza a virus n1 neuraminidase to osel-tamivir carboxylate and zanamivir. Antimi-crob Agents Chemother 2002; 46:3809–3816

50. Rello J, Rodriguez A, Ibanez P, et al: Inten-sive care adult patients with severe respira-

tory failure caused by Influenza A (H1N1)v inSpain. Crit Care 2009; 13:R148

51. Perez-Padilla R, de la Rosa-Zamboni D,Ponce de Leon S, et al: Pneumonia and re-spiratory failure from swine-origin influenzaA (H1N1) in Mexico. N Engl J Med 2009;361:680–689

52. Kumar A, Zarychanski R, Pinto R, et al: Crit-ically ill patients with 2009 influenzaA(H1N1) infection in Canada. JAMA 2009;●●●

53. Jain S, Kamimoto L, Bramley AM, et al: Hos-pitalized Patients with 2009 H1N1 Influenzain the United States, April–June 2009. N EnglJ Med 2009; ●●●

54. Nuno M, Reichert TA, Chowell G, et al: Pro-tecting residential care facilities from pan-demic influenza. Proc Natl Acad Sci U S A2008; 105:10625–10630

55. Nuno M, Chowell G, Gumel AB: Assessingthe role of basic control measures, antivi-rals and vaccine in curtailing pandemic

influenza: Scenarios for the US, UK and theNetherlands. J R Soc Interface 2007;4:505–521

56. Gumel AB, Nuno M, Chowell G: Mathemati-cal assessment of Canada’s pandemic influ-enza preparedness plan. Can J Infect Dis MedMicrobiol 2008; 19:185–192

57. Gurevich KG: Formal kinetics of H1N1 ep-idemic. Theor Biol Med Model 2009; 6:23

58. Boelle PY, Bernillon P, Desenclos JC: A pre-liminary estimation of the reproduction ratiofor new influenza A(H1N1) from the out-break in Mexico, March–April 2009. EuroSurveill 2009; 14:●●●

59. Balcan D, Hu H, Goncalves B, et al: Seasonaltransmission potential and activity peaks ofthe new influenza A(H1N1): A Monte Carlolikelihood analysis based on human mobility.BMC Med 2009; 7:45

60. Mathematical modelling of the pandemicH1N1 2009. Wkly Epidemiol Rec 2009; 84:341–348

09Crit Care Med 2010 Vol. 38, No. 3 (Suppl.)