Since no SARS-specific treatment has been developed, management of this disease remains clinically based and supportive. Importantly, acute respiratory failure typically seen during the late phase of critical SARS, mainly presenting with dyspnea and hypoxia, meets the diagnostic criteria for ALI. For this reason, steroid therapy in critical SARS has become a focus of interest. After the report of clinical experience of expert panel23 from Guangzhou, there have been several reports on steroids therapy in critical SARS, and also no lack of controversy as to their outcomes and usage. Ho et al6 compared 17 cases receiving high-dose pulse methylprednisolone therapy (ie, 500 mg/d for 5 to 7 consecutive days, then tapered to 20 mg/d over a total course of 21 days) with 55 cases receiving low-dose non-pulse corticosteroid (equivalent to approximately 2 to 3 mg/kg/d of methylprednisolone) therapy. It was reported that patients receiving high-dose pulse corticosteroid had less oxygen requirement, better radiographic outcome, and less likelihood of requiring subsequent high-dose pulse corticosteroid therapy than their counterparts. There was no significant difference between the two groups in severe secondary infections. However, in the study by Ho et al,,6 there was no control (receiving no corticosteroid) group.4 Zhao and colleagues4 compared the clinical outcome of four treatment protocols in 190 cases of SARS. It was suggested that on the basis of antibiotics and antivirals (ribavirin and recombinant interferon-α), early high-dosage steroids (160 to 1,000 mg/d of methylprednisolone) resulted in best outcomes and no death as compared with low-dose steroid therapy. However, this was not a randomized controlled study. Patients admitted in different periods received different treatment protocol, and major confounders affecting the outcome of SARS were not adjusted with multivariate analysis. The study by Meng et al7 indicated that early administration of steroids was helpful for symptom relieving and pulmonary resolution. It was shown that low-dose methylprednisolone (40 to 80 mg/d) was associated with significantly shorter hospital days vs high-dose drug (320 to 640 mg/d). Sung et al9 reported the outcome a stepwise treatment protocol of steroid administration in the treatment of SARS in Hong Kong. The initial treatment was broad-spectrum antibiotics, a combination of ribavirin and low-dose corticosteroid. High-dose methylprednisolone (500 mg/d) was administered if the patient deteriorated despite initial treatment. Among the 138 patients, 25 patients (18.1%) responded to ribavirin and low-dose corticosteroid. High-dose methylprednisolone was used in 107 patients, of whom 95 patients (88.8%) responded favorably. Evidence of hemolytic anemia was observed in 49 patients (36%). Thirty-seven patients (26.8%) required admission to the ICU, 21 patients (15.2%) required invasive mechanical ventilation, and 15 patients (10.9%) died. It was suggested that the use of high-dose pulse methylprednisolone during the clinical course of a SARS outbreak was associated with clinical improvement. Nevertheless, there were also opposite opinions. Data from Li et al24showed that high-dose steroids further deteriorated the low level of CD4+, CD8+, and CD3+, giving rise to severe secondary infections. Lee et al25 reported the effect of steroid therapy on SARS-CoV load. Serial plasma SARS-CoV RNA concentrations were measured using a one-step real-time reverse transcriptase-PCR assay targeting the nucleocapsid gene. It was found that plasma SARS-CoV RNA concentrations in the second and third week of illness were significantly higher in patients who received initial hydrocortisone treatment (n = 9), as compared to those who received placebo (n = 7) [area under the curve; Mann-Whitney, p = 0.023]. An increase in the 30-day mortality associated with high-dose steroids was also implicated by Tsang et al.26 The variability of results from all the studies may lie in several reasons. First, some of the cases in the above-mentioned articles20–21 were of seronegative SARS-IgG, which were responsible for a bias due to overdiagnosis. In our study, only 401 of 1,278 registered cases in Guangzhou fulfilled the criteria of case identification instituted by World Health Organization. Second, the matching of SARS cases was not strictly controlled in these studies. As was shown in present study, noncritical SARS was not associated with mortality, but 59.0% of the patients received steroid therapy. Mixing the noncritical and critical SARS information together would interfere with the ability of statistical analysis to evaluate the effects of steroid on mortality. This might be the reason that steroid showed no efficacy on mortality in SARS patients as a whole but was significantly effective in critical SARS. Third, the adjustment of confounders was crucial in minimizing the effects of incomparability of baseline condition. In the present study, much more attention was paid to adjustment of baseline condition of important confounders. After adjusting the confounders with multivariate analysis, it was shown that proper use of steroid resulted in less fatality, lower instant mortality, and fewer hospital days.