Two hundred sixty-seven patients fulfilling CDC case definition for suspected or probable SARS were admitted to the isolation wards of the Princess Margaret Hospital from February 26, 2003, to March 31, 2003. RT-PCR on NPA became available to us in mid-March. We included in our study 218 patients who had nasopharyngeal RT-PCR performed at illness onset.

Routine hematologic, biochemical, and microbiologic tests were performed for all patients. NPA samples were examined by rapid immunofluorescence antigen detection methods for viral cell culture and for common respiratory virus, including influenza viruses A and B; adenovirus; respiratory syncytial virus; and parainfluenza virus types 1, 2, and 3. Sputum samples were screened for bacterial and mycobacterial infection by conventional microscopic identification (Gram staining and acid-fast staining) and culture methods (blood, chocolate, MacConkey, and Löwenstein-Jensen media). Serologic testing for Mycoplasma pneumoniae, Chlamydia pneumoniae, and C. psittaci was performed. Urinary antigen detection tests were used to detect Legionella pneumophila and Streptococcus pneumoniae in some patients. Paired serum samples were taken 10–14 days apart to assess the serologic response to coronavirus by immunofluorescence assay (IFA). All chest radiographs were classified according to their laterality and extent of involvement.

The NPA sample collected from patients was added into a sterile vial containing 2 mL of viral transport medium and then transported on ice (4°C) to the Public Health Laboratory Centre, Government Virology Unit (GVU) of Hong Kong. Total RNA from 140 μL of each NPA sample was extracted by a QIAamp viral RNA Mini kit (QIAGEN, Valencia, CA), as instructed by the manufacturer and eluted in 60 μL of buffer. A total of 4.2 μL of eluted RNA was reverse-transcribed with use of reverse transcriptase (Applied Biosystem, Foster City, CA) in a 20-μL reaction containing 2.5 μM (final concentration) of random hexamer. The mixture was incubated at room temperature for 10 min and then at 42°C for 15 min. The reaction was stopped at 95°C for 5 min and then chilled in ice. The primers used for amplification, COR-1 and COR-2, were targeted at the coronavirus polymerase gene designed by GVU: sense 5′ CAC CGT TTC TAC AGG TTA GCT AAC GA 3′and antisense 5′ AAA TGT TTA CGC AGG TAA GCG TAA AA 3′, with expected product size of 311 bp. Five microliters of cDNA was amplified in 45 μL of master mixture containing 5 μL of 10X PCR buffer (Amersham Pharmacia Biotech, Piscataway, NJ), 1 μL of 2 5 mM extra MgCl₂, 4 μL of deoxynucleoside triphosphates (dNTPs) (2.5 mM each), 0.5 μL of each primer, 0.3 μL of Taq polymerase (5 U/mL), and 33.7 μL of molecular grade water. One positive control and one negative control were included in each PCR assay. Reactions were performed in a thermocycler (GeneAmp PCR System 9700, Applied Biosystem) with the following conditions: at 94°C for 3 min, followed by 45 cycles of 94°C for 30 s, 60°C for 30 s, 72°C for 1 min, and 72°C for 7 min. PCR products were analyzed by gel electrophoresis.

All patients received treatment according to a standard protocol. Either a β-lactam plus β-lactamase inhibitor or third-generation cephalosporin in combination with a macrolide or a fluoroquinolone was given to the patient at admission. Per the recommendation of the Hospital Authority of Hong Kong, an antiviral drug (ribavirin 24 mg/kg/day intravenously, together with hydrocortisone 10 mg/kg/day) was administered if the symptoms did not respond within 48 h (The recommendations were found available at: URL: http://www.ha.org.hk/hk/hesd/nsapi/?Mlval=ha_view_content&c_id=122711&lang=E. However, the recommendations have since changed and are available at: URL: http://www.ha.org.hk/hesd/nsapi/?MIval=ha_view_content&c_id=123510&hesd_lang=E). Methylprednisolone, in the form of two to three pulsing doses of 500 mg to 1,000 mg a day intravenously, was administered to those with a persistent fever, radiologic evidence progression of lung infiltrates, or signs of respiratory distress despite the initial antiviral-hydrocortisone combination.

Bivariate analysis was performed for epidemiologic, clinical, laboratory, radiologic data, and outcomes by using RT-PCR results as the dependent variable. Data of continuous variables were expressed as mean and standard deviation. Chi-square test was used for categorical variables, and the unpaired Student t test was performed for continuous variables. All significant factors for death with a p value <0.1 were pooled into a multivariate logistic regression model with backward stepwise analysis to identify the independent predictors for the clinical outcome. A p value <0.05 (two-tailed) was assumed to be statistically significant. All analyses were performed with the SPSS version 10.0 software (SPSS Inc, Chicago, IL).

On admission, nasopharyngeal RT-PCR was performed on 90 male and 128 female patients (mean age 39.6 ± 14 years). All patients, except six, were Chinese; two were Indonesian, and four were Filipino. Twenty-one of the patients (10%) were healthcare workers, including 5 clinicians, 9 nurses, 5 ward assistants, and 2 allied health workers who worked in the SARS wards. Forty-one patients (19%) reported having recently traveled to SARS-endemic areas in the 2 weeks before admission; the most common areas visited were in the southern part of China. Our cohort consisted of patients (46.8%) from a local housing estate, the Amoy Gardens. Ten patients (4.6%) had one or more coexisting medical problems: diabetes mellitus (3 cases), a history of cerebrovascular disease (4 cases), ischemic heart disease (3 cases), chronic rheumatic heart disease (1 case), hypertrophic obstructive cardiomyopathy (1 case), sick sinus syndrome (1 case), cirrhosis of the liver secondary to chronic hepatitis B (1 case), bronchiectasis (1 case), end-stage renal disease (1 case), Sjögren syndrome (1 case), and nasopharyngeal carcinoma (1 case). The proportions of patients from Amoy Gardens that were RT-PCR positive (48%) and negative (45.7%) were not significantly different (p = 0.76). Likewise, the proportions of healthcare workers (p = 0.28) or patients with coexisting conditions (p = 0.83) did not differ significantly.

NPA samples for RT-PCR were taken from all patients at admission; samples from 156 patients (71.6%) were positive. The mean time from disease onset to sample collection was 4.4 ± 2.3 days. No significant difference in the mean sampling time was found between RT-PCR–positive or –negative patients. The optimal time for sample collection was day 8–10 when 13 of 14 patients (92.9%) were positive (Figure).

The most common clinical features for both RT-PCR–positive and –negative cases included fever, chill, malaise, myalgia, cough, rigor, and headache (Table 1). Shortness of breath and dizziness were significantly higher in RT-PCR–positive patients in bivariate analysis. Vital signs taken on admission (temperature, heart rate, and systolic and diastolic blood pressure) were similar between the two groups. Common laboratory findings included anemia with hemoglobin level <12 g/dL (14.7%), lymphopenia with leukocyte count <4 x 10⁹/L (72%), thrombocytopenia with platelet count <150 x 10⁹/L (52.3%), hypokalemia with plasma potassium level <3.5 mmol/L (41.3%), hyponatremia with plasma sodium level <135 mmol/L (61.5%), and elevated levels of lactate dehydrogenase >230 U/L (46.6), alanine transaminase > 40 U/L (30.8%), and C-reactive protein (77.8%). By bivariate analysis, lymphopenia and elevated lactate dehydrogenase level on admission were significantly different between RT-PCR–positive and –negative patients (Table 2).

Eighty-seven NPA RT-PCR–positive patients and 33 RT-PCR–negative patients had serologic tests on their paired serum samples 10–14 days apart. Of the positive RT-PCR patients, 74 patients (85.1%) had total antibodies detected by IFA, while serologic tests for 25 patients (75.8%) in the RT-PCR–negative group were positive. Results for 13 patients in the RT-PCR–positive group and 8 patients in the RT-PCR–negative group were negative.

Initial chest radiographs for 210 patients (96.3%) were abnormal. Sixty-five (41.7%) RT-PCR–positive patients and 13 (21%) RT-PCR–negative patients had bilateral chest involvement shown by radiograph. Multifocal radiologic involvement was found in 74 (47.4%) RT-PCR–positive patients and 15 (24.2%) RT-PCR–negative patients. By bivariate analysis, RT-PCR–negative patients were less likely to have abnormal bilateral (p = 0.01) and multifocal (p = 0.003) radiographs.

The overall 30-day mortality rate was 10.1% (22 patients). Fifty-two (23.9%) patients required intensive-care unit (ICU) admission, and 43 patients (19.7%) needed mechanical ventilation. In nine (4.1%) patients, acute renal failure further complicated SARS. When compared to the RT-PCR–negative patients, the RT-PCR–positive patients were more likely to need treatment in the ICU (p = 0.002), require mechanical ventilation (p = 0.008), and die (p = 0.044) (Table 3).

Admission parameters, including epidemiologic data, vital signs, and laboratory and chest radiographic findings, were analyzed separately. By bivariate analysis, factors associated with death were age >60 (p = 0.037), male sex (p = 0.007), major coexisting medical conditions (p = 0.001), shortness of breath (p = 0.005), total leukocyte count >4.0 x 10⁹/L (p = 0.041), bilateral chest radiographic involvement (p = 0.046), RT-PCR positivity on NPA samples (p = 0.034), and pulsing doses of steroid (p = 0.001). By multivariate analysis, independent predictors of 30-day mortality were RT-PCR positivity on NPA samples (odds ratio [OR] 6.4; 95% confidence interval [CI] 1.1 to 38.0; p = 0.038), shortness of breath on admission (OR 3.9; 95% CI 1.2 to 12.3; p = 0.02), presence of important coexisting condition (OR 13.4; 95% CI 3.1 to 58.2; p = 0.001), total leukocyte count >4.0 x 10⁹/L (OR, 6.94; 95% CI 1.18 to 41.6; p = 0.033), and pulsing doses of methylprednisolone (OR 26.0; 95% CI 4.4 to 154.8; p = 0.001) (Table 4).