In China, all suspected H5N1 cases are reported to the Chinese Center for Disease Control and Prevention (China CDC, Beijing, China) through a national surveillance system, which is based upon reporting of hospitalized cases of pneumonia of unknown origin, and by enhanced 1-month surveillance for cases of influenza-like illness at all health-care facilities within a 3-km radius after the occurrence of a suspected or confirmed H5N1 poultry outbreak with high bird mortality.

A case of pneumonia of unknown origin was defined as a patient with all of the following criteria without specific laboratory diagnosis: fever (temperature ≥38°C); radiological evidence of pneumonia or acute respiratory distress syndrome (ARDS); normal white blood cell count (WBC; range 4–10×10⁹ cells per L), leukopenia (WBC <4×10⁹ cells per L), or lymphopenia (lymphocyte count <0.8×10⁹ cells per L) at clinical presentation; and absence of clinical improvement after treatment with broad-spectrum antibiotics. A case of influenza-like illness was defined as a patient with fever (temperature ≥38°C) and cough or sore throat, in the absence of any other confirmed diagnosis.

A confirmed case of H5N1 was defined as a patient with pneumonia or influenza-like illness and laboratory evidence of H5N1 virus infection diagnosed by viral isolation or reverse transcriptase (RT) PCR by testing respiratory specimens, or a four-fold or greater increase in H5N1 antibody titre in paired acute and convalescent sera.

All suspected H5N1 case-patients were interviewed by staff of the local CDC, and respiratory specimens, and acute- and convalescent-phase sera were obtained if available for laboratory investigations following the WHO protocol [21]. Respiratory specimens were tested by conventional [22] and real-time RT-PCR [23] to detect H5-specific viral RNA in biosafety level (BSL) 2 facilities at the National Influenza Center (NIC) of China CDC, and were inoculated into amniotic and/or allantoic cavities of specific pathogen free (SPF) embryonated chicken eggs for viral isolation [24] in enhanced BSL 3 facilities at the NIC. H5N1 antibody testing was performed on sera at the NIC by microneutralization (MN) assay [25] in a BSL-3 laboratory, and modified hemagglutination-inhibition (HI) assay using horse red blood cells [26] in BSL-2 conditions.

In mainland China, 30 confirmed human H5N1 cases have been identified to date. We included 26 laboratory-confirmed H5N1 cases identified by surveillance in 12 provinces in China between October 2005 and April 2008. Our analyses included limited data from 2 case reports [27], [28] and 6 urban cases reported previously in a brief epidemiological dispatch [29]. Of the 26 cases, H5N1 virus infection was confirmed by both virus isolation and RT-PCR in 20 (77%) cases, one (4%) case by virus isolation only, three (11%) by RT-PCR and serology, and two (8%) by serology only. Twenty-four cases in southern China were infected with clade 2.3.4 H5N1 viruses, and two cases from northern China had clade 2.2 H5N1 virus infections [2]. We excluded four H5N1 cases, including 2 military cases with unavailable clinical data and 2 cases in a cluster with limited person-to-person transmission reported elsewhere [5].

A trained team from the China CDC interviewed all confirmed H5N1 cases or their proxies, and collected clinical data through review of hospital medical records. A standardized form was used to collect information on demographic characteristics and clinical data, including clinical findings, blood chemistry testing and chest radiograph results performed during clinical management, complications, treatments, and outcomes. Data were collected during field investigations by China CDC staff, and was part of a continuing public-health outbreak investigation and determined by the Ministry of Health to be exempt from institutional review board assessment in China.

We used the following definitions: cardiac failure was defined as requiring use of inotropic agents; respiratory failure was defined as the need for assisted ventilatiory support; ARDS was defined as clinical deterioration with severe arterial hypoxaemia and diffuse bilateral infiltrates on chest radiograph; disseminated intravenous coagulation (DIC) was defined as elevated prothrombin time (PT) with elevated activated partial thromboplastin time (APTT), and decreased fibrinogen (FIB) level with thrombocytopenia; liver function impairment was defined as aminotransferase (ALT or AST) levels ≥ 2× upper range of normal values; renal dysfunction was defined as creatinine level >178mmol/L for adults or ≥ 2× upper limit of normal for age. High-dose corticosteroid use was defined as ≥250 mg hydrocortisone or equivalent intravenous (IV) administration daily. For children <13 years old, high-dose corticosteroid use was defined as ≥5 mg hydrocortisone or equivalent IV/kg/day.

Medians and interquartile ranges (IQRs) were calculated for continuous variables, and compared between fatal and nonfatal cases using Wilcoxon rank sum test. For categorical variables, percentages of case-patients in each category were compared using Fisher's exact test. Fatal cases were compared to nonfatal cases by demographic characteristics, H5N1 virus clade, underlying medical conditions, medical care practices, haematological and biochemical markers at admission or during hospitalization, clinical complications, and treatments, in the bivariate analyses using logistic regression. All statistical tests were two-sided with a significance level set at α = 0.05. Data were analyzed with SPSS (version 13.0, SPSS Inc, Chicago, IL, USA).

The earliest reported symptoms and signs of 26 patients at illness onset and noted at hospital admission are shown in Table 1. Many patients reported fever (92%) or cough (58%) initially, but very few reported upper respiratory symptoms such as rhinorrhea or sore throat. All patients developed cough a median of 1 day (IQR 1–3) from illness onset, and 85% had sputum production a median of 3 days (IQR 1–5.3) after illness onset. Lower respiratory tract signs and symptoms such as tachypnea and dyspnea increased substantially from illness onset to hospital admission. Most patients (88%) had tachypnea a median of 5 days (IQR 4–7) from illness onset and 46% reported dyspnea a median of 6.5 days (IQR 4.5–8.5) from illness onset. Diarrhea was reported in only one adult case at illness onset, and in two cases at hospital admission, but developed in six patients (one child and five adults) after hospitalization. The duration of diarrhea in these nine (35%) cases was a median of 1 day (IQR 1–4).

All case-patients had abnormal chest radiographs at admission; unilateral or bilateral infiltrates were observed in 10 (38%) and 16 (62%) case-patients, at a median of 6.5 days (IQR 4–7.3) and 7.5 days (IQR 6.3–9) from illness onset, respectively. The 10 case-patients with unilateral infiltrates at admission all developed bilateral pneumonia (Table 2). Chest radiographs showing rapid progression from unilateral to bilateral pulmonary infiltrates and ARDS in adult and paediatric cases are shown in Figure 1. Radiographic findings included patchy or diffuse infiltrates or consolidation with air bronchograms in multi-segmental or lobular distribution.

Laboratory findings on initial testing, at admission and during hospitalization are shown in Table 3. The prevalence of patients with abnormal haematological findings at admission [leukopenia (46%), lymphopenia (62%), and moderate thrombocytopenia (50%)] increased to 92%, 89% and 73%, respectively, during hospitalization. At admission, the median leukocyte count was 3.5×10⁹ cells/L (IQR 2.3–4.5) and median lymphocyte count was 0.6×10⁹ cells/L (IQR 0.4–1.0). These declined during hospitalization to a median leukocyte count of 2.3×10⁹ cells/L (IQR 1.5–2.8) and median lymphocyte count of 0.3×10⁹ cells/L (IQR 0.3–0.5), after a median of 8.0 days.

Abnormal percentage, peak measurement and median days from illness onset of biochemical markers on initial testing, at hospital admission, and during hospitalization are shown in Table 3. Elevated ALT, AST, creatine kinase (CK), creatine phosphokinase isoenzymes (CPK), lactic dehydrogenase (LDH), and plasma glucose concentration, and decreased albumin levels were observed in more than half of cases at admission, and developed in nearly all cases during hospitalization. Elevated creatine level was observed in 25% of cases during hospitalization. Seventeen (77%) cases developed proteinuria at a median of 9.0 days (IQR 7.0–11) after illness onset.

All cases received empiric treatment with broad-spectrum antibiotics during hospitalization, including ceftriaxone (n = 6), moxifloxacin (n = 8) and azithromycin (n = 15). Corticosteroids (median methylprednisolone dosage 1.6 [1]–[5] mg/kg per day IV) were initiated at a median of 6.5 days (IQR 6.0–8.0) from illness onset and administered to 24 (92%) cases for a median of 6 days (IQR 3–13). Of these, 21 (88%) received high-dose corticosteroids.

Four children received late antiviral treatment. One was treated with amantadine (100 mg per os (po) twice daily (BID) on illness days 10–15) and ribavirin (200 mg IV/d on illness days 9–16), and one received rimantadine (100 mg po each day (qD) on illness days 9–11); both cases survived. One child received oseltamivir (37.5 mg po BID) on illness days 12–14, and one child was treated with oseltamivir (40 mg po qD) on illness day 10; both died. Eight adults received late oseltamivir treatment, including two fatal cases – one received 75 mg po BID on illness days 8–11, and one received both oseltamivir (75mg po BID on illness days 11–20) and rimantadine (200mg po qD on illness day 11). Six adults treated with oseltamivir survived: one was treated with 75 mg/day on illness days 8–12, four received 75 mg BID on illness days 4–11, illness days 8–14, illness days 10–14, and illness days 8–12, respectively, and one was treated with 75mg BID and amantadine (100 mg po BID) on illness days 8–12.

Two critically ill adult H5N1 cases (31-year-old male, 44-year-old female) with ARDS were treated with convalescent plasma obtained from one of two fully recovered H5N1 adult donor cases. Plasma was obtained 129 days after illness onset from an adult female case and 81 days after illness onset from an adult male case. Both donors' convalescent plasma tested negative for hepatitis B, hepatitis C, and HIV, and were separated and heat-inactivated at 56°C for 10 h before transfusion. The male ARDS case received three units (200 mL/unit) of transfused convalescent plasma from the female donor for 2 days, beginning on illness day 13. His H5N1 viral titre in bronchial-alveolar lavage fluid declined substantially and was undetectable for the next 3 consecutive days after receipt of the third convalescent plasma dose. The female ARDS case, who had a history of bronchiectasis, received one unit (200 mL) of transfused convalescent plasma from the male donor once daily for 3 days, starting on illness day 13. Further virological testing has not been done for this case. Both cases also received oseltamivir (75 mg po BID) on illness days 10–14 and days 8–12, respectively. Both cases recovered fully and were discharged home.

Twenty-three (88%) cases required ventilatory support for respiratory failure. ARDS developed in 21 (81%) cases at a median of 8 days (IQR 7–9) after illness onset. Liver function impairment, renal dysfunction and cardiac failure occurred in 9 (43%), 4 (17%) and 13 (50%) patients.

Seventeen (65%) cases died (2 children, 1 adolescent and 14 adults), including one pregnant woman at 4 months' gestation [28] after a median of 10 days (IQR 8–20.5). Nine (35%) nonfatal cases were discharged at a median of 41 days (IQR 31.5–64.0) after illness onset. Five (24%) of the 21 cases with ARDS survived, including one pregnant woman, two adults who received convalescent H5N1 plasma, and two other previously healthy adults. The pregnant woman survived after developing ARDS and experiencing a spontaneous abortion during mechanical ventilation. Her pulmonary status subsequently improved and her temperature normalised quickly; the patient was extubated and recovered completely. All 17 fatal cases had multi-organ failure, including respiratory failure (94%), cardiac failure (71%), renal failure (27%) and 24% had disseminated intravenous coagulation (Table 4).

In the bivariate analyses, demographic characteristics, year of illness onset, clade of H5N1 virus infection, and underlying medical conditions were similar between fatal and nonfatal cases (Table 4). Fatal cases had significantly lower median nadir platelet count during hospitalization (64.5×10⁹ cells/L vs 93.0×10⁹ cells/L, p = 0.02), higher median peak LDH level during hospitalization (1982.5 U/L vs 1230.0 U/L, p = 0.001), higher frequency of ARDS (94% [n = 16] vs 56% [5], p = 0.034), more frequent cardiac failure (71% [n = 12] vs 11% [1], p = 0.011), and shorter median duration of corticosteroid therapy (4.0 days vs 12.0 days, p = 0.025) compared to cases that survived. A higher proportion of cases survived that received any antiviral treatment compared to those that did not receive antivirals (67% [8/12 patients] vs 7% [1/14 patients], p = 0.003), with a positive correlation between antiviral therapy and disease outcome (Gamma coefficient = 0.664, p = 0.005).