Article Text

Download PDFPDF

Finding the real case-fatality rate of H5N1 avian influenza
  1. F C K Li1,
  2. B C K Choi2,
  3. T Sly3,
  4. A W P Pak4
  1. 1
    Centre for Infectious Disease Prevention and Control, Public Health Agency of Canada, Ottawa, Ontario, Canada
  2. 2
    Department of Epidemiology and Community Medicine, University of Ottawa, Ottawa, Ontario; Department of Public Health Sciences, University of Toronto, Toronto, Ontario, Canada
  3. 3
    School of Occupational and Public Health, Ryerson University, Toronto, Ontario, Canada
  4. 4
    Pak Consulting, Ottawa, Ontario, Canada
  1. Dr F Li, Centre for Infectious Disease Prevention and Control, Public Health Agency of Canada, 100 Colonnade Road Ottawa, Ontario, K1A 0K9, Canada; Felix_Li{at}


Background: Accurate estimation of the case-fatality (CF) rate, or the proportion of cases that die, is central to pandemic planning. While estimates of CF rates for past influenza pandemics have ranged from about 0.1% (1957 and 1968 pandemics) to 2.5% (1918 pandemic), the official World Health Organization estimate for the current outbreak of H5N1 avian influenza to date is around 60%.

Methods and results: The official estimate of the H5N1 CF rate has been described by some as an over-estimate, with little relevance to the rate that would be encountered under pandemic conditions. The reasons for such opinions are typically: (i) numerous undetected asymptomatic/mild cases, (ii) under-reporting of cases by some countries for economic or other reasons, and (iii) an expected decrease in virulence if and when the virus becomes widely transmitted in humans. Neither current data nor current literature, however, adequately supports these scenarios. While the real H5N1 CF rate could be lower than the current estimate of 60%, it is unlikely that it will be at the 0.1–0.4% level currently embraced by many pandemic plans. We suggest that, based on surveillance and seroprevalence studies conducted in several countries, the real H5N1 CF rate should be closer to 14–33%.

Conclusions: Clearly, if such a CF rate were to be sustained in a pandemic, H5N1 would present a truly dreadful scenario. A concerted and dedicated effort by the international community to avert a pandemic through combating avian influenza in animals and humans in affected countries needs to be a global priority.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

A critical parameter in planning for a pandemic is the estimate of projected mortality, usually expressed as the case-fatality (CF) rate, or the probability of a case dying from a disease before recovering or dying from another cause.1 It is traditionally estimated by the proportion of cases of a specified condition that are fatal within a specified time period.2

This paper reviews previous and current avian influenza outbreak data in an attempt to estimate the human CF rate for H5N1 with greater confidence. Certain commonly held beliefs regarding the H5N1 CF rate are also examined.

H5N1 case-fatality rate: observations to date

The CF rate of seasonal (non-pandemic) human influenza is low, generally no more than 0.1–0.2% (i.e. one to two deaths per thousand cases).3 The last two influenza pandemics (1957 and 1968) resulted in higher than normal incidence rates, but the CF rates remained in a similar range, 0.1–0.4%.4 5 However, the CF rate during the 1918 pandemic, the deadliest recorded in history, was documented at around 2.5%, 5 6 resulting in an estimated 50 million deaths worldwide.6

By comparison, official figures from the World Health Organization (WHO) reveal that the accumulative CF rate in the current outbreak of type A/H5N1 (avian) influenza among the 335 human cases at the time of writing is 61% (table 1).7 Furthermore, this accumulative CF rate has remained approximately consistent, and has not dropped below 40% since the outbreak began in 2003.

Table 1 Number of confirmed human cases and deaths of avian influenza H5N1 reported to WHO, by calendar year, as of 12 November 20077

Epidemiological differences have been observed amongst human cases of H5N1 as the outbreak has progressed,8 and genetic analysis has recently confirmed that the H5N1 avian influenza virus has evolved into at least two separate clades or strains with a non-overlapping geographic distribution. Clade 1 was isolated from humans and birds in Vietnam, Thailand, Cambodia, Laos, and Malaysia in 2004–2007; and clade 2 from birds in China, Indonesia, Japan, and South Korea during 2003–2004 with subsequent spread to the Middle East, Europe and Africa during 2005–2006.9 10 Clade 2 has been principally responsible for human infections since the latter part of 2005.11 CF rates for clade 1 (53%) and clade 2 (67%) reveal both to be alarmingly high (table 2).

Table 2 Number of confirmed human cases and deaths of avian influenza H5N1 reported to WHO, by clade,810 as of 12 November 20077

If these observed CF rates were to be sustained during a full pandemic, H5N1 would represent an unprecedented public health disaster.

A realistically estimated CF rate is essential when preparing for an influenza pandemic.1214 Unfounded estimates, on the other hand, can fuel either alarmist or complacent comment.12 15 A recent high-profile observation asserted that the official case-fatality rate of above 50% is “… a complete exaggeration, and not scientifically justified... much of the attention focused on H5N1 is unwarranted”.12

How reliable is the observed case-fatality rate?

A common belief is that the current H5N1 CF rate is an over-estimation, which will decrease as more information becomes available or as the epidemic progresses. Three “scenarios” are often put forward to explain this belief, paraphrased as follows:

Scenario 1. “Because the current CF rates are calculated only from the cases and deaths officially reported to and confirmed by WHO, unknown numbers of sub-clinical and milder cases probably exist. When they are taken into account, through more thorough surveillance, the estimate of the case-fatality rate will decrease.”12 16

The assumption is that asymptomatic and mild cases remain unreported, particularly in developing countries without comprehensive surveillance systems, and that these cases would increase the denominator, reducing the CF rate.

Serological surveys, however, do not support this assumption. Research conducted in Vietnam17 18 and Thailand,19 in 2004, among 168 healthcare workers exposed to confirmed H5N1 patients found none with H5N1 antibodies. Similarly, tests done on 51 household contacts of H5N1 cases and on 25 contacts with sick poultry in Vietnam in 2004 yielded similar negative serological results.20 In 2005, no H5N1 antibodies were found in 351 residents of a Cambodian village where poultry and one person had died of avian influenza,21 and serology on 45 close contacts of a girl who died of H5N1 in Cambodia in March 2006 was all found to be negative.22

By comparison, contact surveillance during the first H5N1 outbreak in Hong Kong (1997) did reveal contacts who had developed antibodies to H5N1. These included 6 of 51 (12%) household contacts, 1 of 26 (4%) tour group contacts, and 8 of 217 (4%) healthcare worker contacts, but none of 47 workplace contacts.23 24 During the same outbreak, 9 (3%) seropositives were found among 293 workers culling poultry.25 In 2004, one Indonesian (of 79 tested)20 and one Japanese26 poultry culler, both asymptomatic, showed H5N1 antibodies after exposure to infected poultry. Two household contacts of H5N1 patients in Vietnam in two separate H5N1 incidents, in 2005, were also found to have H5N1 antibodies while asymptomatic.27 In Turkey’s 2006 H5N1 outbreak, five mild/asymptomatic H5N1 cases were reported, but have not been confirmed by WHO collaboration centres.28

The above data indicate that asymptomatic or mild cases of H5N1 are probably neither common nor widespread. Yet for an observed CF rate of 50% to decrease to 2.5% (as in the 1918 pandemic, for example), an average of 19 additional undetected (mild or asymptomatic and unreported) cases would need to be found for every “reported” case; similarly, for the CF rate to decrease to 0.25% (as suggested in many current pandemic planning documents) an additional 199 undetected cases would need to be confirmed for each reported case. Serological observations thus far are not remotely close to these figures.

Scenario 2. “For economic or other reasons, some countries elect not to report cases or to delay reporting them, or to deliberately under-report the less serious cases of H5N1.”29 30

Despite WHO’s pleas for transparency,31 under-reporting of H5N1 cases represents a potential source of uncertainty and error. As in Scenario 1, any missing cases that are finally identified would decrease the CF rate through having its “total ill” denominator increased.

Although we are unable to assess the magnitude of this potential error, it appears that many infected countries are now reporting more comprehensive H5N1 data, including clinical observation and serologic testing results of contacts.3234 The thoroughness of reporting of H5N1 cases in Indonesia has been demonstrated in an epidemiological and clinical investigation of three clusters of H5N1 cases and their contacts.35 Also, the intense surveillance, case-finding and serologic testing conducted in the Hong Kong, Turkey and Cambodia outbreaks2325 28 33 36 still yielded CF rates between 33% and 100%.7

As before, there is no convincing evidence that substantive numbers of cases are being withheld.

Scenario 3. “Any increase in transmissibility of H5N1 in humans will be associated with a massive drop in virulence, because killing the host is not a viable evolutionary strategy for a virus.”13

This follows from the postulate that if an infectious agent is so highly virulent that it kills its host quickly, its basic reproductive number (R0, or the expected number of new infectious cases per infectious case) may reduce to less than one, and the agent will die out.1 However, the natural history of influenza and the accumulated data on human H5N1 in particular, do not support such a scenario:

Firstly, transmission of influenza from a primary case to secondary cases occurs early in the course of the illness. Influenza virus can be shed by patients before onset of symptoms,37 with peak shedding during the first 24–72 hours of illness.37 On the other hand, the median time from onset of symptoms to hospitalisation (for 150 H5N1 cases from December 2003 to April 2006) was 4 days (range: 0–18 days),38 and the median duration between onset of symptoms and death among the 113 H5N1 deaths during the same period was 9 days (range: 2–31 days).38

Secondly, over the course of four years’ continued existence in humans, H5N1 has not shown any indication of reducing virulence in terms of CF rate.7

The hypothesised decrease in mortality/virulence as transmission increases also did not receive much empirical support from the 2003 international outbreak of severe acute respiratory syndrome (SARS), when a high CF rate was maintained throughout the period of its transmission among humans.39 40

Estimating the “real” H5N1 case-fatality rate

Influenza viruses present a “moving target” due to the high mutation rate of the viral genome, and its unique ability to re-assort genetic information between human and non-human strains. As far as we know, H5 viruses have never caused a pandemic, so there is great uncertainty about their epidemiology and pathology under those circumstances. Nevertheless, we can examine useful indicators from the outbreak to date.

Among the H5N1-affected areas, Hong Kong and Turkey have arguably the most effective surveillance and reporting systems.2325 28 36 41 In the 1997 Hong Kong outbreak, out of a total of 18 patients, 6 died, yielding a CF rate of 33%.42 When the 15 patient contacts and 9 poultry workers detected through exhaustive seroprevalence studies2325 are added to the denominator, the CF rate becomes 14%. In Turkey, 4 H5N1 patients died out of 12 confirmed cases, giving a CF rate of 33%.43 These figures of 14–33%, in our opinion, provide the best estimates of the H5N1 CF rate to date.

In making such an estimation, it must be kept in mind that the CF rate of human H5N1 cases is influenced by the level of medical care available to patients with acute respiratory distress syndrome in those countries. Therefore, with the same virulence of the H5N1 virus, the CF rate is likely to be higher in countries with less well-equipped intensive care resources.

Realistic comparisons with 1918?

While most experts do not dispute that another influenza pandemic will eventually occur, there has been a great deal of debate on: (a) whether the currently occurring H5N1 virus will cause a pandemic, and (b) if it does occur, whether it will cause catastrophic consequences similar to that of 1918, or be “milder” as in 1957 or 1968.

While the first question is impossible to answer at this time, we do know that two of the three conditions for starting a pandemic have already been met,44 leaving the acquisition of efficient human-to-human transmission by the H5N1 virus, either by genetic re-assortment or mutation, as the remaining requirement for a pandemic. Taubenberger and colleagues have found that of the ten amino acid differences between the polymerase complex of human and avian influenza viruses that could account for human transmissibility, seven have already been observed individually in currently circulating H5N1 viruses recovered from birds and humans.45 Alternatively, a pandemic could be triggered by one or two E190D mutations in the HA of the H5N1 strain, which switch its binding preference from α2,3 sialic acid (the major form in the avian enteric tract) to α2,6 sialic acid (the major form in the human respiratory tract).46

As for the second question, there are already good indications that, in terms of its clinical picture, pathology, genetics, and also the demographic characteristics of its human victims, H5N1 resembles the 1918 pandemic virus (H1N1) far more than the 1957 (H2N2) or 1968 (H3N2) varieties.47 Clinically, both the H5N1 and 1918 pandemic virus infections are characterised by rapidly progressive, primary viral pneumonia leading to acute respiratory distress and death.48 Both infections are accompanied pathologically by a reactive haemophagocytic syndrome that is probably the result of a “cytokine storm”,48 49 a deadly over-reaction of the victim’s own immune system. Similarly, in the laboratory, the vigorous release of cytokines in mice injected with the 1918 influenza virus was accompanied by rapid onset of pulmonary disease and death.47 On a molecular basis, recent studies have found remarkable similarities between the polymerase genes, which contribute to the virulence of the H5N1 and 1918 influenza viruses.45 46 Demographically, unlike the 1957/1968 pandemics and inter-pandemic influenza that causes excess mortality mostly among the elderly and those with chronic illnesses, both the H5N1 and the 1918 influenza virus affected young healthy adults more severely,38 50 a characteristic that may be related to the cytokine storm phenomenon.

Pandemic prevention and intervention

The increasing spread of H5N1 globally and its increasing occurrence among poultry, migratory birds, other animal species, and humans, particularly over the past two years, combined with the potential for a very high CF rate in humans, clearly represents a most serious threat to global human health that should not be countered with complacency.

Kilbourne contended that, with the occurrence of an influenza pandemic, “no amount of hand washing, hand wringing, public education, or gauze masks will do the trick. The keystone of influenza prevention is vaccination”.51 He further stressed the importance of making a vaccine available, even without an exact match to the antigens of an eventual pandemic strain. This is supported by the apparent ameliorating effect of N2 immunity provided by the circulating H2N2 strain from 1957 to 1968 in many populations, and by the protective effect of a H2N2 vaccine among United States Air Force cadets during the 1968 (H3N2) pandemic.51 In the same context, we think that the low incidence rate (5%) and CF rate (2%) for H5N1 recorded in people 50 years and over, as shown in recent statistics,52 may also reflect some cross-protection by N1 antibodies conferred among people exposed to the predominantly circulating H1N1 influenza strains from 1918 up until 1957.51

While development and deployment of effective vaccination would be the most useful intervention in a pandemic prepared-ness plan, its aim is to reduce the impact of the catastrophe. The prevention of an influenza pandemic in the first place, through vigorous, dedicated and concerted efforts toward eliminating and/or controlling avian influenza virus in poultry/animals and humans should be of equal, if not higher, priority.

Although it has been generally conceded that the global eradication of H5N1, at least in the short term, does not appear feasible at this stage, experience to date in Hong Kong, Turkey, Thailand, Vietnam and other countries has clearly shown that H5N1, both in animals and humans, could be successfully controlled or eliminated through the committed and dedicated effort of governments and the international community. Unfortunately, many of the H5N1 outbreaks are occurring in countries without optimal public health infrastructure, resources, and/or expertise. In today’s highly interlinked global village, no individual country can remain isolated from the impact of devastating infectious diseases occurring in other parts of the world; and combating emerging infectious diseases, such as H5N1, in affected countries has become a de facto international responsibility and obligation. The world simply cannot afford to allow the smouldering H5N1 outbreaks in a few countries to develop into a global wildfire. A pandemic arising from uncontrolled avian influenza outbreaks in any country would fittingly be viewed as not just the failure of one country, but rather that of the global community.

What this study adds

  • The official WHO estimate of the H5N1 case-fatality rate of around 60% is considered by some as an over-estimate due to several problems.

  • It is unlikely that the real H5N1 case-fatality rate will be at the 0.1–0.4% level.

  • This paper reviews potential errors in estimation based on observations to date and suggests that the real H5N1 case-fatality rate should be around 14–33%.

Policy implications

This paper discusses current potential challenges in estimating the case-fatality rate of H5N1 avian influenza. Accurate estimation of the H5N1 case-fatality rate is central to pandemic planning.


View Abstract


  • Disclaimer: Opinions expressed in this paper are solely those of the authors and do not necessarily represent the views of any agencies, organisations or universities.

  • Competing interests: None declared.

Linked Articles

  • In this issue
    Mauricio L Barreto