Volume 12, Number 12—December 2006
Risk Factors for Human Infection with Avian Influenza A H5N1, Vietnam, 2004
To evaluate risk factors for human infection with influenza A subtype H5N1, we performed a matched case-control study in Vietnam. We enrolled 28 case-patients who had laboratory-confirmed H5N1 infection during 2004 and 106 age-, sex-, and location-matched control-respondents. Data were analyzed by matched-pair analysis and multivariate conditional logistic regression. Factors that were independently associated with H5N1 infection were preparing sick or dead poultry for consumption <7 days before illness onset (matched odds ratio [OR] 8.99, 95% confidence interval [CI] 0.98–81.99, p = 0.05), having sick or dead poultry in the household <7 days before illness onset (matched OR 4.94, 95% CI 1.21–20.20, p = 0.03), and lack of an indoor water source (matched OR 6.46, 95% CI 1.20–34.81, p = 0.03). Factors not significantly associated with infection were raising healthy poultry, preparing healthy poultry for consumption, and exposure to persons with an acute respiratory illness.
The first indication that the current epizootic of highly pathogenic avian influenza subtype H5N1 (influenza A H5N1) would have a serious effect on human heath occurred in early 2004, when influenza H5N1 was identified in a series of patients admitted to the National Pediatric Hospital in Hanoi with severe viral pneumonia (1). Since then, large-scale and global spread of the disease in poultry has been accompanied by sporadic cases in humans. Despite many millions of avian infections and >200 human cases, knowledge of influenza H5N1 remains inadequate. Neither how these viruses are transmitted to humans nor, consequently, the most effective way to reduce the risk for infection is fully understood. Descriptive and analytic epidemiologic studies conducted in Hong Kong Special Administrative Region, People's Republic of China, during the 1997 outbreak of influenza H5N1 (2–5) identified visiting a live bird market as a risk factor. However, the current outbreak encompasses different viruses and different sociodemographic, farming, and behavioral contexts. Several seroprevalence studies of healthcare workers and a case-control study from Thailand have been published from the current outbreak (6–9), but further work is needed to develop and test hypotheses on the mechanism of transmission of influenza H5N1 to humans. To clarify the source and mode(s) of transmission of influenza H5N1 to humans and to guide the control and prevention of influenza, we conducted a case-control study of all cases of avian influenza H5N1 identified in humans in Vietnam in 2004.
All persons with laboratory-confirmed influenza A H5 cases detected in Vietnam from January 1 through December 31, 2004, were eligible for enrollment as case-patients. Case-patients were identified from persons hospitalized with an acute respiratory infection considered by clinicians, on the basis of clinical and epidemiologic findings, to have a suspected case of H5N1 infection. Clinicians did not use a systematic case definition or screening protocol to identify patients eligible for testing for H5N1 infection. Throat swabs or tracheal aspirate samples were sent to the National Institute of Hygiene and Epidemiology in Hanoi or to the Pasteur Institute in Ho Chi Minh City for reverse transcription (RT) PCR and viral isolation. When possible, samples with positive results for influenza A H5 were sent to a World Health Organization (WHO) reference laboratory for confirmatory diagnosis.
For each case-patient, 4 control-respondents, individually matched by gender, age (age difference <24 months), and place of residence (same ward or village), were selected by use of a random number table from a list of persons fitting the selection criteria provided by the community health station near each case-patient's place of residence. Potential control-respondents were excluded if they reported having suffered an illness with respiratory symptoms and fever (temperature >38°C) during their matched case-patient's period of illness (onset to recovery or death). If the selected control-respondent refused to participate or did not meet the inclusion criteria, the geographically closest eligible person was then selected from the list. All eligible control-respondents were asked to provide a throat swab and venous blood (5 mL) to confirm that they were not currently or had not previously been infected with influenza A H5. Participation of case-patients and control-respondents was voluntary and required written consent and, for those aged <18 years, signature of a parent or guardian.
Trained interviewers administered a structured questionnaire to case-patients and control-respondents. If the case-patient or control-respondent was a child or if the case-patient had died, the questionnaire was administered to a proxy, usually the child's parent or a close family member living in the same household. The questionnaire collected information about demographic characteristics; preexisting health status; smoking behavior; potential animal, human, and environmental exposures to influenza A H5; and personal and household hygienic practices. Case-patients or their proxies were asked about exposures in the 7 days before illness onset, and control-respondents were asked about exposures during the same 7-day period as their matched case-patient.
Persons who met any of the following criteria were considered to be laboratory-confirmed influenza A H5 case-patients: 1) influenza A H5–specific RNA detected in a single specimen by RT-PCR by using 2 different primer pairs; 2) influenza A H5 detected in a single specimen by RT-PCR identification and by sequencing or virus isolation; 3) influenza A H5–specific RNA detected by RT-PCR in 2 different specimen types (e.g., throat swab and tracheal aspirate); and 4) influenza A H5–specific RNA detected by RT-PCR in 2 samples taken on different days. Control-respondents were considered to be true control-respondents if throat swab specimens were negative for influenza A H5–specific RNA by RT-PCR and anti-H5 antibodies could not be detected in serum samples by microneutralization assay (10).
Influenza A H5 subtype–specific RNA was detected in clinical samples by RT-PCR with primers that targeted regions of the hemagglutinin gene of the influenza H5N1 virus developed by WHO, the US Centers for Disease Control and Prevention (CDC), and the Government Virus Unit in Hong Kong. Clinical specimens were injected into Madin-Darby canine kidney cells for virus isolation, and RT-PCR was used to identify influenza A H5. Specimens and cell cultures suspected of containing influenza A H5 were handled according to recognized biosafety standards.
Serum samples were immediately processed, stored at –25°C, and shipped frozen on dry ice to CDC. To measure influenza A H5–specific antibody, microneutralization assay was conducted as previously described (10) by using H5N1 viruses A/Vietnam/1194/2004 and A/Vietnam/3212/2004. Microneutralization test results were considered to be positive if an anti-H5 titer of >40 was obtained by 2 independent assays.
Data entry and analysis of individual explanatory variables was performed by using Epi-Info 6 (CDC, Atlanta, GA, USA). Mantel-Haenszel matched-pair analysis (McNemar test) was used to estimate the strength and statistical significance of associations between exposures and influenza A H5 infection. An association was considered statistically significant if 2-sided tests of significance had a p-value <0.05. To examine independence of effects, multivariate conditional logistic regression was performed by using the conditional logistical regression (CLogit) function in Stata/SE 8.0 for Windows (Stata Corp LP, College Station, TX, USA). Any variables with p<0.2 after matched analysis were included in the initial model. A backward stepwise variable- selection strategy was used to construct a final model with a significance level of >0.1 for removal and a significance level of <0.05 for re-entry into the model. Persons missing data for variables under study were excluded from any analysis involving the missing variable. Collinearity was assessed by generating a correlation coefficient matrix for all variables to be considered for inclusion in the regression model. The presence of effect-measure modification by age and sex was assessed for all variables in the final model by entering product terms. A final model was achieved by entering the variables retained in the backward selection model.
The attributable risk percent (AR%) was estimated as follows: ([odds ratio (OR) – 1]/OR) ×100. The population attributable risk percent (PAR%) was estimated as follows: PAR% = AR% × proportion of case-patients exposed.
A total of 28 laboratory-confirmed influenza A H5 cases were detected in 2004 from 15 provinces of Vietnam; 21 (75%) were fatal. All 28 cases were RT-PCR positive for influenza A subtype H5 at either the National Institute of Hygiene and Epidemiology, Hanoi, or the Pasteur Institute, Ho Chi Minh City; H5N1 virus was isolated in 12. The diagnosis of influenza A H5 infection was independently confirmed for 25: 20 at a WHO reference laboratory and 5 at the Oxford University Clinical Research Unit in Vietnam.
The interviews began on February 7, 2004, at which time 20 of the 28 case-patients had already been identified. The interval between onset of illness and interview was a mean of 35.7 days; the maximum interval was 63 days. The mean age of case-patients was 14 years (range 1–31 years, median 15 years), and 9 (32%) patients were children <10 years of age. The numbers of male and female case-patients were equal (14 each). Among confirmed case-patients were 2 family clusters (mother and daughter and 2 sisters). A total of 106 control-respondents were enrolled, 4 per case-patient, except for 3 case-patients aged <5 years for whom only 1, 2, and 3 control-respondents could be recruited per case, respectively. All control-respondents were negative for avian influenza A H5–specific RNA by RT-PCR and for anti-H5 antibodies by microneutralization assay. None of the case-patients or control-respondents worked in the commercial (industrial) poultry-raising sector.
The results of matched-pair analysis are shown in Table 1. Direct handling of sick or dead poultry in the 7 days before onset of illness had the strongest point estimate of effect (matched OR 31) and high statistical significance (p<0.001) despite wide confidence limits (95% confidence interval [CI] 3.4–1150). The presence of sick or dying poultry in the household (matched OR 7.4, 95% CI 2.7–59) or neighborhood (matched OR 3.9, 95% CI 1.0–55.7) was also statistically associated with infection as was the absence of an indoor water source in the household (matched OR 5.0, 95% CI 1.3–77.0) and education to high school level or higher (matched OR 16.0, 95% CI 1.2–594.1).
Eight variables with p<0.2 were considered for inclusion in the conditional logistic regression model to estimate independence of effects. Although significantly associated with infection in the single-variable analysis, the presence of sick or dead poultry in the neighborhood was excluded from the final regression model because missing data for this variable led to the exclusion of 36 participants (6 case-patients and 30 control-respondents). Educational level was excluded because it was not a relevant variable for the 13 case-patients <15 years of age. Because of the 2 family clusters, each comprising 2 case-patients, the influence of clustering of household-level factors on the regression model was investigated by running the regression model first with all cases and then again including only 1 case from each of these 2 households. All 4 variations of 1 case from each household were run. Because the outcomes of these different approaches did not differ, all cases were included in the final model.
The final conditional logistic regression model included 3 variables as independent risk factors for H5N1 infection (Table 2). Of the 28 case-patients, 16 (57%) had either sick or dead poultry in their household or had directly prepared sick or dead poultry for consumption; another 6 reported sick or dead poultry in the neighborhood. Of the 28 case-patients, 22 (79%) did not have an indoor water source. No statistically significant effect-measure modification was detected.
Among persons who prepared sick or dead poultry for consumption, the proportion of H5N1 cases attributable to this practice (AR%) is estimated in this study to be 89% ([(8.99 – 1)/8.99] × 100). However, because only 32% of all case-patients reported this practice, stopping this practice would prevent only an estimated 28% of H5N1 cases (PAR% = 0.89 × 0.32).
Source of Infection
This study identified the presence in the household and the handling of dead or sick poultry in an H5N1-affected area as risk factors for human H5N1 infection. Although not surprising, these findings reinforce the hypothesis that close contact with infected domestic poultry is the primary source of transmission of influenza H5N1 to humans. The absence of a statistical association between infection and contact with other animals such as pigs, cats, or dogs is reassuring. Replication and excretion of H5N1 by asymptomatic domestic waterfowl has been demonstrated and is a plausible source of infection for humans (11); however, although a 1997 case-control study found that visiting live poultry markets was a risk factor for human influenza A H5N1 infection (2), our study and the study from Thailand (9) did not identify contact with healthy poultry as a risk factor. However, viral titers in asymptomatic waterfowl may be much lower than in diseased poultry (11) and may therefore pose a low risk to humans, which our study was underpowered to identify.
Despite evidence that limited human-to-human transmission of H5N1 has occurred (3,5,12), we observed no significant differences between case-patients and control-respondents in terms of exposure to persons who might be a source of H5N1 infection, e.g., patients with an acute respiratory infection. This finding is consistent with that of the case-control study in Thailand (9).
Our study found an association between infection and direct and household contact with diseased poultry. This association, if true, could operate by 2 mechanisms. First, transmission may be by inhalation or conjunctival deposition of large infectious droplets, which may travel only short distances (13); second, the presence of infected poultry in the home and preparation of infected poultry for consumption may result in exposure to higher virus concentrations than other types of exposure. An alternative hypothesis is that the consumption, rather than the preparation, of infected poultry is the route of infection. Domestic cats and possibly tigers have been infected by the oral route (14,15), and the observed association between preparation and infection could be the result of a confounding association between preparing an infected bird (apparent risk factor) and consuming an infected bird (true risk factor). Unfortunately, the study did not ask about consumption of sick or dead poultry because the gastrointestinal route of transmission was not considered plausible at the time the study was designed.
The unexpected finding that the absence of an indoor water source is associated with infection may point toward a role for self-inoculation into conjunctival, nasal, or oral mucosa by contaminated hands or possibly foodstuffs. Certainly, the environmental stability of avian influenza viruses is sufficient for this to occur (16), and handwashing has been shown to decrease risk for respiratory infections (17). However, we asked 2 questions specifically about handwashing behavior, and neither was significantly associated with infection (Table 1). An alternative, and more controversial, explanation is that people without access to an indoor water source may acquire infection by drinking or washing in outdoor water sources contaminated with feces from infected poultry. This hypothesis is plausible, given the extended survival of avian influenza viruses in water (18,19) and the demonstration of oral infectivity in cats (14). Although an indoor water source might simply be a proxy indicator of socioeconomic status or the priority the household gives to hygiene, all other hygiene factors and a composite hygiene index did not show a statistically significant association with infection.
That 5 case-patients did not report any exposure to sick poultry in the 7 days before illness onset has several possible explanations: recall bias by case-patients or their proxies, infection acquired from infected but asymptomatic animals such as ducks, an incubation period >7 days, or infection from a contaminated environment as discussed above.
A major source of potential bias in this and other studies of risk factors for human H5N1 infection is the use of self-reported prior exposure to sick poultry as a screening tool for identifying potential case-patients. Use of this tool introduces a selection bias that favors finding greater exposure to sick poultry among case-patients than other groups, regardless whether the relationship is causal. In our study, clinicians were not using a systematic screening tool to identify possible H5N1 case-patients, but knowledge of poultry as the source of human H5N1 infection was ubiquitous. Possibly, H5N1 case-patients who did not have exposure to sick poultry may have been less likely to be identified than case-patients who did report this exposure. However, 5 of the 28 case-patients (18%) did not report exposure to sick poultry, indicating that exposure to sick poultry was not a prerequisite for identification as a case-patient. In light of this conflict between clinical necessity and study purity, estimates of the size of the association between exposure to sick poultry and H5N1 infection could be interpreted as maximums that are likely to have been inflated by this selection bias.
The relatively small number of case-patients means that the study may be underpowered to detect factors posing only a moderate risk for infection and to detect effect modification. A standardized questionnaire and trained interviewing staff were used to try to minimize interviewer bias, but masking the interviewers as to the case or control status of the respondents was not possible. Recall bias was likely to have occurred, especially because the high case-fatality rate meant that a larger proportion of interviews in the case group (26/28) than the control group (35/106) were completed by proxies. The substantial delay between onset of illness and interviews (mean 35.7 days) is also a potential source of recall bias.
The finding of a significant positive association between level of education and risk for infection was unexpected and is difficult to explain. It may be the consequence of a bias introduced by proxy respondents for deceased case-patients reporting higher levels of education than case-patients had actually achieved.
Misclassification of case-patients and control-respondents was unlikely. All control-respondents were demonstrated to have no detectable antibodies to H5N1, and all case-patients had a clinically compatible illness with laboratory evidence of H5N1 infection, which was independently verified for 25 (89%) of the 28 cases.
Public Health and Research Implications
Preparing sick or dead poultry for consumption in an H5N1-affected area is a risky practice. Although this study cannot estimate the absolute risk, among those who prepared sick or dead poultry for consumption, a high proportion of infections could be attributed to this practice. However, as the practice was not that widespread in our study participants, stopping it would prevent only an estimated 28% of H5N1 cases. Less risky but more widespread practices probably account for a greater proportion of H5N1 cases; these practices must also be identified and tackled. Regardless whether consumption of infected poultry is itself a risk factor, preparation and consumption of sick or dead poultry in infected areas must stop. That all 106 persons selected as control-respondents from communities with at least 1 confirmed human H5N1 case were negative for H5N1 antibodies adds further evidence to the belief that widespread subclinical H5N1 infection has not yet occurred in Southeast Asia (20).
The finding of an association between lack of access to an indoor water source and H5N1 infection provides an interesting basis for formulating new hypotheses, but it is not sufficiently strong evidence for concluding that H5N1 transmission is occurring by water or as a result of inadequate hygiene. Despite 2 reports of exposure to potentially contaminated water in Vietnamese H5N1 case-patients (; pers. comm., Ministry of Health, Vietnam), no human cases of H5N1 infection have been directly attributed to exposure to contaminated water. Nevertheless, hygiene practices and access to safe water have collateral benefits regardless of H5N1 and should be encouraged and pursued. Environmental investigations are needed to sample water sources in and around the households of incident H5N1 case-patients and compare the findings to water sources sampled in and around unaffected households.
Familial clusters of cases have been a significant feature of the epidemiology of H5N1 infection since 2004 in that numerous clusters have occurred in Vietnam, Thailand, Cambodia, Indonesia, and Turkey (22,23). Although common exposures and behavior may be one explanation for the marked clustering, most clusters have involved blood relatives such as sibling pairs or parent-child groups rather than unrelated pairs such as husbands and wives. This finding suggests that inherited biologic factors, such as sialic acid receptor phenotype or immune response, may be determinants of infection and disease. Studying intrinsic determinants of susceptibility will require pooling of data and samples from affected families across affected countries. If intrinsic susceptibility were a risk determinant, it might dilute associations between certain behavior and infection unless the analyses were undertaken within subgroups that are homogeneous with respect to their intrinsic susceptibility. In this respect, intrafamilial studies that combine measures of biologic susceptibility with data about behavioral patterns, including food consumption and hygiene practices, may be particularly enlightening.
Dr Dinh is associate professor of epidemiology and vice-director of the National Institute of Hygiene and Epidemiology in Vietnam. His research interests focus on the epidemiology, etiology, and control of communicable diseases of public health importance, especially emerging communicable diseases such as severe acute respiratory syndrome, avian influenza, and viral encephalitides.
The National Institute of Hygiene and Epidemiology thanks the World Health Organization office in Vietnam for identifying the financial resources and providing technical support and the influenza laboratory of US Centers for Disease Control and Prevention for conducting the microneutralization assays. We respectfully thank the Ministry of Health, Vietnam Administration of Preventive Medicine, Preventive Medicine Centers of 15 provinces where the H5N1 cases occurred, National Pediatric Hospital, National Institute for Infectious and Tropical Diseases, Pediatric Hospital No 1, Pediatric Hospital No 2, and the Hospital of Tropical Diseases-Ho Chi Minh City for providing the conditions and human resources needed to implement this study. Particularly, the authors of this study are indebted to the 28 influenza A H5N1 patients in Vietnam (2004) and their relatives. Without their great cooperation and help, this study would not have been completed.
The members of the WHO/Global Outbreak Alert and Response Network Avian Influenza Investigation Team in Vietnam are Bach Huy Anh (Hanoi Medical University, Vietnam); Philippe Barboza (Institut de Veille Sanitaire, France); Niranjan Bhat, Aaron Curns, Nguyen Cong Doan, Jackie Katz, Keiji Fukuda, Amy Funk, Taronna Maines, Mark Simmerman, Terry Tumpey, Timothy Uyeki (CDC, United States); Arnold Bosman, Mirna Du Ry van Beest Holle (National Institute for Public Health and the Environment, the Netherlands); Sofia Boqvist (Smittskyddsinstitutet, Sweden); Rick Brown, Futoshi Hasabe (Asian Development Bank, the Phillipines); Pascale Brudon, Philippe Calain, Robert Dietz, Rodger Doran, Tom Grein, Peter Horby, Hitoshi Oshitani (WHO); Valerie Delpech (Health Protection Agency, United Kingdom); Patrice Gautier (Vétérinaires sans Frontières, Vietnam); Shigeyuki Itamura, Takehiko Saito (National Institute for Infectious Diseases, Japan); Donna Mak (Curtin University of Technology, Australia); Noel Miranda (Association of Southeast Asian Nations Secretariat, Indonesia); Reiko Saito (Niigata University, Japan); and Joel Francart.
This study was generously supported by a donation from the Italian Development Cooperation Fund (Cooperazione Italiana) through the WHO office in Vietnam.
- World Health Organization. Avian influenza A (H5N1) in humans and poultry in Viet Nam, 2004 January 13 [cited 2006 May 18]. Available from http://www.who.int/csr/don/2004_01_13/en/index.html
- Mounts AW, Kwong H, Izurieta HS, Ho Y, Au T, Lee M, Case-control study of risk factors for avian influenza A (H5N1) disease, Hong Kong, 1997. J Infect Dis. 1999;180:505–8.
- Buxton Bridges C, Katz JM, Seto WH, Chan PK, Tsang D, Ho W, Risk of influenza A (H5N1) infection among health care workers exposed to patients with influenza A (H5N1), Hong Kong. J Infect Dis. 2000;181:344–8.
- Bridges CB, Lim W, Hu-Primmer J, Sims L, Fukuda K, Mak KH, Risk of influenza A (H5N1) infection among poultry workers, Hong Kong, 1997–1998. J Infect Dis. 2002;185:1005–10.
- Katz JM, Lim W, Bridges CB, Rowe T, Hu-Primmer J, Lu X, Antibody response in individuals infected with avian influenza A (H5N1) viruses and detection of anti-H5 antibody among household and social contacts. J Infect Dis. 1999;180:1763–70.
- Liem NT, Lim W; World Health Organization International Avian Influenza Investigation Team, Vietnam. Lack of H5N1 avian influenza transmission to hospital employees, Hanoi, 2004. Emerg Infect Dis. 2005;11:210–5.
- Apisarnthanarak A, Erb S, Stephenson I, Katz JM, Chittaganpitch M, Sangkitporn S, Seroprevalence of anti-H5 antibody among Thai health care workers after exposure to Avian influenza (H5N1) in a tertiary care center. Clin Infect Dis. 2005;40:e16–8.
- Schultsz C, Dong VC, Chau NV, Le NT, Lim W, Thanh TT, Avian influenza H5N1 and healthcare workers. Emerg Infect Dis. 2005;11:1158–9.
- Centers for Disease Control and Prevention. Investigation of avian influenza (H5N1) outbreak in humans–Thailand, 2004. MMWR Morb Mortal Wkly Rep. 2006;55(Suppl 1):3–6.
- Rowe T, Abernathy RA, Hu-Primmer J, Thompson WW, Lu X, Lim W, Detection of antibody to avian influenza A (H5N1) virus in human serum by using a combination of serologic assays. J Clin Microbiol. 1999;37:937–43.
- Sturm-Ramirez KM, Hulse-Post DJ, Govorkova EA, Humberd J, Seiler P, Puthavathana P, Are ducks contributing to the endemicity of highly pathogenic H5N1 influenza virus in Asia? J Virol. 2005;79:11269–79.
- Ungchusak K, Auewarakul P, Dowell SF, Kitphati R, Auwanit W, Puthavathana P, Probable person-to-person transmission of avian influenza A (H5N1). N Engl J Med. 2005;352:333–40.
- Bridges CB, Kuehnert MJ, Hall CB. Transmission of influenza: implications for control in health care settings. Clin Infect Dis. 2003;37:1094–101.
- Rimmelzwaan GF, van Riel D, Baars M, Bestebroer TM, van Amerongen G, Fouchier RA, Influenza A virus (H5N1) infection in cats causes systemic disease with potential novel routes of virus spread within and between hosts. Am J Pathol. 2006;168:176–83.
- Keawcharoen J, Oraveerakul K, Kuiken T, Fouchier RA, Amonsin A, Payungporn S, Avian influenza H5N1 in tigers and leopards. Emerg Infect Dis. 2004;10:2189–91.
- Shortridge KF, Zhou NN, Guan Y, Gao P, Ito T, Kawaoka Y, Characterization of avian H5N1 influenza viruses from poultry in Hong Kong. Virology. 1998;252:331–42.
- Luby SP, Agboatwalla M, Feikin DR, Painter J, Billhimer W, Altaf A, Effect of handwashing on child health: a randomised controlled trial. Lancet. 2005;366:225–33.
- Ito T, Okazaki K, Kawaoka Y, Takada A, Webster RG, Kida H. Perpetuation of influenza A viruses in Alaskan waterfowl reservoirs. Arch Virol. 1995;140:1163–72.
- Stallknecht DE, Shane SM, Kearney MT, Zwank PJ. Persistence of avian influenza viruses in water. Avian Dis. 1990;34:406–11.
- Vong S, Coghlan B, Mardy S, Holl D, Seng H, Ly S, Low frequency of poultry-to-human H5N1 virus transmission, southern Cambodia, 2005. Emerg Infect Dis. 2006;12:1542–7.
- de Jong MD, Bach VC, Phan TQ, Vo MH, Tran TT, Nguyen BH, Fatal avian influenza A (H5N1) in a child presenting with diarrhea followed by coma. N Engl J Med. 2005;352:686–91.
- World Health Organization. Situation updates–avian Influenza [cited 2006 May 18]. Available from http://www.who.int/csr/disease/avian_influenza/updates/en/index.html
- Olsen SJ, Ungchusak K, Sovann L, Uyeki TM, Dowell SF, Cox NJ, Family clustering of avian influenza A (H5N1). Emerg Infect Dis. 2005;11:1799–801.