- Email: [email protected]
Comparative analytical sensitivities of six rapid influenza A antigen detection test kits for detection of influenza A subtypes H1N1, H3N2 and H5N1
Journal of Clinical Virology 38 (2007) 169–171
Short communication
Comparative analytical sensitivities of six rapid influenza A antigen detection test kits for detection of influenza A subtypes H1N1, H3N2 and H5N1 K.H. Chan a , S.Y. Lam a , P. Puthavathana b , T.D. Nguyen c , H.T. Long d , C.M. Pang a , K.M. Chan a , C.Y. Cheung a , W.H. Seto a , J.S.M. Peiris a,∗ a
Department of Microbiology, The University of Hong Kong, University Pathology Building, Queen Mary Hospital Compound, Pokfulam, Hong Kong, SAR, China b Department of Microbiology, Sriraj Hospital, Bangkok, Thailand c National Institute of Veterinary Research, Hanoi, Vietnam d National Institute of Hygiene and Epidemiology, Hanoi, Vietnam Received 19 October 2006; accepted 27 November 2006
Abstract Background: Rapid and simple methods for diagnosing human influenza A (H5N1) disease urgently needed. The limited data so far suggest that the currently available rapid antigen detection kits have poor clinical sensitivity for diagnosis of human H5N1 disease. Objectives: To compare the analytical sensitivity of six commercially available rapid antigen detection kits for the detection of “human” (subtypes H1N1, H3N2) and “avian” (subtype H5N1) influenza A viruses. Study design: Six commercially available test kits for the detection of influenza A were investigated. Analytic sensitivity for the detection of two contemporary H1N1, two H3N2 and three H5N1 viruses was determined using virus culture as a reference method. Results and conclusions: Each test kit detected the H5N1 virus subtypes as efficiently as they detected conventional human viruses of subtypes H1N1 or H3N2. However, limits of detection of influenza viruses of all subtypes by antigen detection kits were >1000-fold lower than virus isolation. Thus, the reportedly poor clinical sensitivity of these antigen detection kits for diagnosis of patients with H5N1 disease is not due to a difference of sensitivity for detecting avian influenza H5N1 compared to human influenza viruses. © 2006 Elsevier B.V. All rights reserved. Keywords: Antigen detection; Influenza A; Human; Avian; H5N1
1. Introduction Rapid and simple diagnostic tests for confirming infection with influenza A subtype H5N1 are urgently needed. Preliminary reports suggest that existing rapid antigen detection tests for influenza A have poor clinical sensitivity for diagnosis of patients with H5N1 disease (Beigel et al., 2005). It is known that some of these antigen detection tests detect influenza viruses of diverse subtypes including H5N1 (Chan et al., 2002; Fedorko et al., 2006). It is not known, however, whether the poor clinical sensitivity for diagnosis of human H5N1 dis∗
Corresponding author. Tel.: +852 2855 4888; fax: +852 2855 1241. E-mail address: [email protected] (J.S.M. Peiris).
1386-6532/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jcv.2006.11.010
ease is related to a lower analytical sensitivity for detection of virus subtype H5N1 in comparison to human influenza A subtypes H3N2 or H1N1. We sought to compare the analytical sensitivity of six commercially available influenza A rapid antigen detection tests for the detection of cultured influenza A viruses of subtypes H5N1, H1N1 and H3N2.
2. Materials and methods 2.1. Influenza A isolates Four human clinical isolates of influenza A (two H1N1 isolates and two H3N2 isolates obtained in 2005), two human
170
K.H. Chan et al. / Journal of Clinical Virology 38 (2007) 169–171
H5N1 clinical isolates (one from Vietnam and one from Thailand) collected in 2004 and one chicken isolate collected in 2004 were used for this study.
ried out according to manufacturer’s instructions. Each kit was tested in duplicate on each virus dilution.
2.2. Viral culture
3. Results
Madin–Darby canine kidney (MDCK) cell monolayers grown in culture tubes were inoculated with 200 l of viral isolate and incubated at 35 ◦ C for 1 h. The cells were fed with 1 ml of serum-free minimum essential medium containing tosylsulfonyl phenylalanylchloromethyl ketone (TPCK)-treated trypsin (2 g/ml) (Sigma, St. Louis, MO) and antibiotics. The cultures were harvested when approximately 75% of the cell monolayer manifested cytopathic effect (CPE).
The viral titres of A/Chicken/Vietnam/33/04, A/Vietnam/3028/04, A/Thailand/MK/04, A/Hong Kong/31988/05, A/Hong Kong/34220/05, A/Hong Kong/37039/05 and A/Hong/Kong/35048/05 were log 10 of TCID50 5.5, 6.7, 6.2, 7.3, 6.0, 7.3 and 6.5, respectively (Table 1). The limit of detection of each kit is denoted as the mean of lowest TCID50 (log 10) detectable (Table 1). The limit of detection of each kit for the H5N1 viruses was comparable to the detection limit of the human influenza viruses. There were marginal differences in overall sensitivity between test kits and the trend of sensitivity seemed to be related to the volume of specimen used for testing by each kit.
2.3. Determination of tissue culture infectious dose (50%) (TCID50 ) A 96-wells microtitre plates containing 0.1 ml of confluent MDCK cells were infected with 100 l of serial 10-fold of dilutions of each virus in serum-free medium with antibiotic and TPCK trypsin starting from 10−1 to 10−8 . The plates were incubated at 37 ◦ C for 3 to 4 days. Titrations were done in quadruplicate. Appearance of CPE was recorded daily. TCID50 was determined according to the Reed–Muench method (Reed and Muench, 1938).
4. Discussion We found that the rapid influenza A virus antigen detection kits detect both current common human (H3N2, H1N1) and “avian” H5N1 influenza A subtypes with comparable analytic sensitivity. While this is not unexpected given the fact that these tests target conserved internal viral proteins of influenza A viruses, it was clinically important that the analytic sensitivity is of these tests for human and avian influenza viruses was compared directly. Our findings imply that the reportedly poor apparent clinical sensitivity for diagnosis of influenza A H5N1 infection in patients (Beigel et al., 2005) is not related to a lower sensitivity of the H5N1 subtype virus when compared to the detection of human viruses. Alternative explanations may include inadequate specimen collection, patient age (adults have lower viral loads in upper respiratory tract than children), inherent limitations in the
2.4. Rapid kits for flu A detection The rapid antigen detection kits evaluated were QuickVue Influenza A + B (Quidel Corporation, CA, USA), BinaxNow Influenza A & B (Emergo Europe, The Netherlands), Directigen Flu A + B (Becton Dickinson and Company, MD, USA), Directigen EZ Flu A + B (Becton Dickinson and Company), Poctem Influenza A/B (Sysmex, Japan) and Rapid Testa Flu II (Genzyme Diagnostic, Japan). The procedures were car-
Table 1 The TCID50 limit of detection of each influenza A virus in different rapid antigen detection test kits TCID50 (log 10)/0.1 ml
log 10 TCID50 limit of detection Quidel Quick Vue A + B
BinaxNow
A/Chicken/Vietnam/33/04
5.5
2.5
3.0
3.0
2.7
3.3
4.0
A/Vietnam/3028/04 A/Thailand/MK2/04
6.7 6.2
3.3 3.4
4.0 4.0
3.3 3.9
4.0 3.4
4.9 4.2
5.2 4.7
Mean detection limit for H5N1 viruses
6.1
3.1
3.7
3.4
3.4
4.1
4.6
Human H1N1
A/Hong Kong/31988/05 A/Hong Kong/34220/05
7.3 6.0
4.0 2.6
4.8 3.7
4.5 2.6
4.3 2.7
4.8 3.7
5.3 4.5
Human H3N2
A/Hong Kong/37039/05 A/Hong Kong/35048/05
7.3 6.5
4.1 3.5
4.8 4.5
4.8 3.7
4.5 4.0
4.8 4.5
5.8 5.0
Mean detection limit for H1N1 or H3N2 viruses
6.8
3.5
4.4
3.9
3.9
4.4
5.1
Overall mean detection limit Equivalent volume of samples used in each test (l)
6.5
3.3 280
4.1 100
3.7 60.7
3.6 128.3
4.3 21.2
4.9 21.2
Influenza A virus source and subtype
Virus designation
Avian H5N1 Human H5N1
Directigen Flu A + B
Directigen EZ Flu A + B
Sysmx
Genzyme
K.H. Chan et al. / Journal of Clinical Virology 38 (2007) 169–171
sensitivity of these tests for detecting patients with influenza A, or that patients with H5N1 disease have lower levels of virus antigen in the upper respiratory tract in comparison to those with human H3N2 or H1N1 disease. It is interesting to note that a similar antigen detection test (Directigen Flu A) had a clinical diagnostic sensitivity of 82% in patients with H5N1 in Hong Kong in 1997 (Yuen et al., 1998). It is notable, however, that most of these specimens were nasopharyngeal aspirates rather than throat or nose swabs which are used in patients investigated in 2004 onwards. It is possible that the contemporary H5N1 virus has lower viral load profiles in the upper respiratory tract. However, viral load in patients with H5N1 disease is reportedly even higher than that observed in human influenza (de Jong et al., 2006). It is clear, however, that all these antigen detection tests are far less sensitive than virus culture for detection of virus of all three subtypes and have limits of detection of >103 TCID50 /0.1 ml. This is comparable to previous data (Ryan-Poirier et al., 1992; Fedorko et al., 2006). Human volunteers infected with human influenza viruses who had positive Directigen Flu A test results had mean viruses titres of 2.9 × 104 TCID50 /ml and those with negative results had mean titres of 2.5 × 102 TCID50 /ml (Kaiser et al., 1999). These results are in line with our results of the analytic sensitivities of these tests in general. This limitation in sensitivity in direct antigen detection tests is likely to be relevant in the poor sensitivity in detecting patients with H5N1 disease. The marginal differences in analytical sensitivity between tests may be related to the volume of sample recommended for use in the assay methods, those more sensitive methods using larger test volumes. These antigen detection tests detect influenza type A viruses but do not discriminate between virus subtypes. Therefore, a positive result in any of these antigen detection tests in a patient suspected of “avian flu” must be further subtyped by RT-PCR methods or virus culture in order to confirm or exclude a diagnosis of H5N1 infection. Ideal rapid antigen detection tests will have improved sensitivity for detection of H5N1 virus and will also be able to distinguish between
171
subtypes. Until such tests become available, the “gold standard” diagnostic test for clinical diagnosis of H5N1 disease in humans remains virus culture or RT-PCR methods.
Acknowledgements This study was funded by the Research Fund for the Control of Infectious Diseases of the Health, Welfare and Food Bureau of the Hong Kong SAR Government.
References Beigel JH, Farrar J, Han AM, Hayden FG, Hyer R, de Jong MD, Lochindarat S, Nguyen TK, Nguyen TH, Tran TH, Nicoll A, Touch S, Yuen KY. Writing Committee of the World Health Organization (WHO) Consultation on human influenza A/H5. Avian influenza A (H5N1) infection in humans. N Engl J Med 2005;353:1374–85. Chan KH, Maldeis N, Pope W, Yup A, Ozinskas A, Gill J, Seto WH, Shortridge KF, Peiris JS. Evaluation of the Directigen FluA + B test for rapid diagnosis of influenza virus type A and B infections. J Clin Microbiol 2002;40:1675–80. de Jong MD, Simmons CP, Thanh TT, Hien VM, Smith GJ, Chau TN, Hoang DM, Van Vinh Chau N, Khanh TH, Dong VC, Qui PT, Van Cam B, Ha DQ, Guan Y, Peiris JS, Chinh NT, Hien TT, Farrar J. Fatal outcome of human influenza A (H5N1) is associated with high viral load and hypercytokinemia. Nat Med 2006;12:1203–7. Fedorko DP, Nelson NA, McAuliffe JM, Subbarao K. Performance of rapid tests for detection of avian influenza A virus types H5N1 and H9N2. J Clin Microbiol 2006;44:1596–7. Kaiser L, Briones MS, Hayden FG. Performance of virus isolation and Directigen Flu A to detect influenza A virus in experimental human infection. J Clin Virol 1999;14:191–7. Reed LJ, Muench H. A simple method of estimating fifty per cent endpoints. Am J Hyg 1938;27:493–7. Ryan-Poirier KA, Katz JM, Webster RG, Kawaoka Y. Application of Directigen FLU-A for the detection of influenza A virus in human and nonhuman specimens. J Clin Microbiol 1992;30:1072–5. Yuen KY, Chan PK, Peiris M, Tsang DN, Que TL, Shortridge KF, Cheung PT, To WK, Ho ET, Sung R, Cheng AF. Clinical features and rapid viral diagnosis of human disease associated with avian influenza A H5N1 virus. Lancet 1998;351:467–71.
Short communication
Comparative analytical sensitivities of six rapid influenza A antigen detection test kits for detection of influenza A subtypes H1N1, H3N2 and H5N1 K.H. Chan a , S.Y. Lam a , P. Puthavathana b , T.D. Nguyen c , H.T. Long d , C.M. Pang a , K.M. Chan a , C.Y. Cheung a , W.H. Seto a , J.S.M. Peiris a,∗ a
Department of Microbiology, The University of Hong Kong, University Pathology Building, Queen Mary Hospital Compound, Pokfulam, Hong Kong, SAR, China b Department of Microbiology, Sriraj Hospital, Bangkok, Thailand c National Institute of Veterinary Research, Hanoi, Vietnam d National Institute of Hygiene and Epidemiology, Hanoi, Vietnam Received 19 October 2006; accepted 27 November 2006
Abstract Background: Rapid and simple methods for diagnosing human influenza A (H5N1) disease urgently needed. The limited data so far suggest that the currently available rapid antigen detection kits have poor clinical sensitivity for diagnosis of human H5N1 disease. Objectives: To compare the analytical sensitivity of six commercially available rapid antigen detection kits for the detection of “human” (subtypes H1N1, H3N2) and “avian” (subtype H5N1) influenza A viruses. Study design: Six commercially available test kits for the detection of influenza A were investigated. Analytic sensitivity for the detection of two contemporary H1N1, two H3N2 and three H5N1 viruses was determined using virus culture as a reference method. Results and conclusions: Each test kit detected the H5N1 virus subtypes as efficiently as they detected conventional human viruses of subtypes H1N1 or H3N2. However, limits of detection of influenza viruses of all subtypes by antigen detection kits were >1000-fold lower than virus isolation. Thus, the reportedly poor clinical sensitivity of these antigen detection kits for diagnosis of patients with H5N1 disease is not due to a difference of sensitivity for detecting avian influenza H5N1 compared to human influenza viruses. © 2006 Elsevier B.V. All rights reserved. Keywords: Antigen detection; Influenza A; Human; Avian; H5N1
1. Introduction Rapid and simple diagnostic tests for confirming infection with influenza A subtype H5N1 are urgently needed. Preliminary reports suggest that existing rapid antigen detection tests for influenza A have poor clinical sensitivity for diagnosis of patients with H5N1 disease (Beigel et al., 2005). It is known that some of these antigen detection tests detect influenza viruses of diverse subtypes including H5N1 (Chan et al., 2002; Fedorko et al., 2006). It is not known, however, whether the poor clinical sensitivity for diagnosis of human H5N1 dis∗
Corresponding author. Tel.: +852 2855 4888; fax: +852 2855 1241. E-mail address: [email protected] (J.S.M. Peiris).
1386-6532/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jcv.2006.11.010
ease is related to a lower analytical sensitivity for detection of virus subtype H5N1 in comparison to human influenza A subtypes H3N2 or H1N1. We sought to compare the analytical sensitivity of six commercially available influenza A rapid antigen detection tests for the detection of cultured influenza A viruses of subtypes H5N1, H1N1 and H3N2.
2. Materials and methods 2.1. Influenza A isolates Four human clinical isolates of influenza A (two H1N1 isolates and two H3N2 isolates obtained in 2005), two human
170
K.H. Chan et al. / Journal of Clinical Virology 38 (2007) 169–171
H5N1 clinical isolates (one from Vietnam and one from Thailand) collected in 2004 and one chicken isolate collected in 2004 were used for this study.
ried out according to manufacturer’s instructions. Each kit was tested in duplicate on each virus dilution.
2.2. Viral culture
3. Results
Madin–Darby canine kidney (MDCK) cell monolayers grown in culture tubes were inoculated with 200 l of viral isolate and incubated at 35 ◦ C for 1 h. The cells were fed with 1 ml of serum-free minimum essential medium containing tosylsulfonyl phenylalanylchloromethyl ketone (TPCK)-treated trypsin (2 g/ml) (Sigma, St. Louis, MO) and antibiotics. The cultures were harvested when approximately 75% of the cell monolayer manifested cytopathic effect (CPE).
The viral titres of A/Chicken/Vietnam/33/04, A/Vietnam/3028/04, A/Thailand/MK/04, A/Hong Kong/31988/05, A/Hong Kong/34220/05, A/Hong Kong/37039/05 and A/Hong/Kong/35048/05 were log 10 of TCID50 5.5, 6.7, 6.2, 7.3, 6.0, 7.3 and 6.5, respectively (Table 1). The limit of detection of each kit is denoted as the mean of lowest TCID50 (log 10) detectable (Table 1). The limit of detection of each kit for the H5N1 viruses was comparable to the detection limit of the human influenza viruses. There were marginal differences in overall sensitivity between test kits and the trend of sensitivity seemed to be related to the volume of specimen used for testing by each kit.
2.3. Determination of tissue culture infectious dose (50%) (TCID50 ) A 96-wells microtitre plates containing 0.1 ml of confluent MDCK cells were infected with 100 l of serial 10-fold of dilutions of each virus in serum-free medium with antibiotic and TPCK trypsin starting from 10−1 to 10−8 . The plates were incubated at 37 ◦ C for 3 to 4 days. Titrations were done in quadruplicate. Appearance of CPE was recorded daily. TCID50 was determined according to the Reed–Muench method (Reed and Muench, 1938).
4. Discussion We found that the rapid influenza A virus antigen detection kits detect both current common human (H3N2, H1N1) and “avian” H5N1 influenza A subtypes with comparable analytic sensitivity. While this is not unexpected given the fact that these tests target conserved internal viral proteins of influenza A viruses, it was clinically important that the analytic sensitivity is of these tests for human and avian influenza viruses was compared directly. Our findings imply that the reportedly poor apparent clinical sensitivity for diagnosis of influenza A H5N1 infection in patients (Beigel et al., 2005) is not related to a lower sensitivity of the H5N1 subtype virus when compared to the detection of human viruses. Alternative explanations may include inadequate specimen collection, patient age (adults have lower viral loads in upper respiratory tract than children), inherent limitations in the
2.4. Rapid kits for flu A detection The rapid antigen detection kits evaluated were QuickVue Influenza A + B (Quidel Corporation, CA, USA), BinaxNow Influenza A & B (Emergo Europe, The Netherlands), Directigen Flu A + B (Becton Dickinson and Company, MD, USA), Directigen EZ Flu A + B (Becton Dickinson and Company), Poctem Influenza A/B (Sysmex, Japan) and Rapid Testa Flu II (Genzyme Diagnostic, Japan). The procedures were car-
Table 1 The TCID50 limit of detection of each influenza A virus in different rapid antigen detection test kits TCID50 (log 10)/0.1 ml
log 10 TCID50 limit of detection Quidel Quick Vue A + B
BinaxNow
A/Chicken/Vietnam/33/04
5.5
2.5
3.0
3.0
2.7
3.3
4.0
A/Vietnam/3028/04 A/Thailand/MK2/04
6.7 6.2
3.3 3.4
4.0 4.0
3.3 3.9
4.0 3.4
4.9 4.2
5.2 4.7
Mean detection limit for H5N1 viruses
6.1
3.1
3.7
3.4
3.4
4.1
4.6
Human H1N1
A/Hong Kong/31988/05 A/Hong Kong/34220/05
7.3 6.0
4.0 2.6
4.8 3.7
4.5 2.6
4.3 2.7
4.8 3.7
5.3 4.5
Human H3N2
A/Hong Kong/37039/05 A/Hong Kong/35048/05
7.3 6.5
4.1 3.5
4.8 4.5
4.8 3.7
4.5 4.0
4.8 4.5
5.8 5.0
Mean detection limit for H1N1 or H3N2 viruses
6.8
3.5
4.4
3.9
3.9
4.4
5.1
Overall mean detection limit Equivalent volume of samples used in each test (l)
6.5
3.3 280
4.1 100
3.7 60.7
3.6 128.3
4.3 21.2
4.9 21.2
Influenza A virus source and subtype
Virus designation
Avian H5N1 Human H5N1
Directigen Flu A + B
Directigen EZ Flu A + B
Sysmx
Genzyme
K.H. Chan et al. / Journal of Clinical Virology 38 (2007) 169–171
sensitivity of these tests for detecting patients with influenza A, or that patients with H5N1 disease have lower levels of virus antigen in the upper respiratory tract in comparison to those with human H3N2 or H1N1 disease. It is interesting to note that a similar antigen detection test (Directigen Flu A) had a clinical diagnostic sensitivity of 82% in patients with H5N1 in Hong Kong in 1997 (Yuen et al., 1998). It is notable, however, that most of these specimens were nasopharyngeal aspirates rather than throat or nose swabs which are used in patients investigated in 2004 onwards. It is possible that the contemporary H5N1 virus has lower viral load profiles in the upper respiratory tract. However, viral load in patients with H5N1 disease is reportedly even higher than that observed in human influenza (de Jong et al., 2006). It is clear, however, that all these antigen detection tests are far less sensitive than virus culture for detection of virus of all three subtypes and have limits of detection of >103 TCID50 /0.1 ml. This is comparable to previous data (Ryan-Poirier et al., 1992; Fedorko et al., 2006). Human volunteers infected with human influenza viruses who had positive Directigen Flu A test results had mean viruses titres of 2.9 × 104 TCID50 /ml and those with negative results had mean titres of 2.5 × 102 TCID50 /ml (Kaiser et al., 1999). These results are in line with our results of the analytic sensitivities of these tests in general. This limitation in sensitivity in direct antigen detection tests is likely to be relevant in the poor sensitivity in detecting patients with H5N1 disease. The marginal differences in analytical sensitivity between tests may be related to the volume of sample recommended for use in the assay methods, those more sensitive methods using larger test volumes. These antigen detection tests detect influenza type A viruses but do not discriminate between virus subtypes. Therefore, a positive result in any of these antigen detection tests in a patient suspected of “avian flu” must be further subtyped by RT-PCR methods or virus culture in order to confirm or exclude a diagnosis of H5N1 infection. Ideal rapid antigen detection tests will have improved sensitivity for detection of H5N1 virus and will also be able to distinguish between
171
subtypes. Until such tests become available, the “gold standard” diagnostic test for clinical diagnosis of H5N1 disease in humans remains virus culture or RT-PCR methods.
Acknowledgements This study was funded by the Research Fund for the Control of Infectious Diseases of the Health, Welfare and Food Bureau of the Hong Kong SAR Government.
References Beigel JH, Farrar J, Han AM, Hayden FG, Hyer R, de Jong MD, Lochindarat S, Nguyen TK, Nguyen TH, Tran TH, Nicoll A, Touch S, Yuen KY. Writing Committee of the World Health Organization (WHO) Consultation on human influenza A/H5. Avian influenza A (H5N1) infection in humans. N Engl J Med 2005;353:1374–85. Chan KH, Maldeis N, Pope W, Yup A, Ozinskas A, Gill J, Seto WH, Shortridge KF, Peiris JS. Evaluation of the Directigen FluA + B test for rapid diagnosis of influenza virus type A and B infections. J Clin Microbiol 2002;40:1675–80. de Jong MD, Simmons CP, Thanh TT, Hien VM, Smith GJ, Chau TN, Hoang DM, Van Vinh Chau N, Khanh TH, Dong VC, Qui PT, Van Cam B, Ha DQ, Guan Y, Peiris JS, Chinh NT, Hien TT, Farrar J. Fatal outcome of human influenza A (H5N1) is associated with high viral load and hypercytokinemia. Nat Med 2006;12:1203–7. Fedorko DP, Nelson NA, McAuliffe JM, Subbarao K. Performance of rapid tests for detection of avian influenza A virus types H5N1 and H9N2. J Clin Microbiol 2006;44:1596–7. Kaiser L, Briones MS, Hayden FG. Performance of virus isolation and Directigen Flu A to detect influenza A virus in experimental human infection. J Clin Virol 1999;14:191–7. Reed LJ, Muench H. A simple method of estimating fifty per cent endpoints. Am J Hyg 1938;27:493–7. Ryan-Poirier KA, Katz JM, Webster RG, Kawaoka Y. Application of Directigen FLU-A for the detection of influenza A virus in human and nonhuman specimens. J Clin Microbiol 1992;30:1072–5. Yuen KY, Chan PK, Peiris M, Tsang DN, Que TL, Shortridge KF, Cheung PT, To WK, Ho ET, Sung R, Cheng AF. Clinical features and rapid viral diagnosis of human disease associated with avian influenza A H5N1 virus. Lancet 1998;351:467–71.