Aims To investigate the relationships between a known history of diabetes and ambient fasting plasma glucose (FPG) levels with death and morbidity rates in patients with severe acute respiratory syndrome (SARS). Methods In this retrospective analysis, the clinical and biochemical characteristics of 135 patients who had died from SARS, 385 Survivors of SARS and 19 patients with non-SARS pneumonia were compared. Results All patients were treated according to a predefined protocol. Before steroid treatment, the mean FPG level was significantly higher in the SARS group (deceased vs. survivors vs. non-SARS pneumonia group: 9.7 +/- 5.2 vs. 6.5 +/- 3.0 vs. 5.1 +/- 1.0 mmol/l, P < 0.01). In the SARS group, the percentage of patients with a known history of diabetes was significantly higher in the deceased patients than in the survivors (21.5% vs. 3.9%, P < 0.01). Among patients with no known history of diabetes and before commencement of steroid therapy, those who had hypoxaemia (SaO(2) < 93%) had higher FPG levels than those who did not have hypoxia in both the Survivor (8.7 +/- 4.9 vs. 6.3 +/- 2.1 mmol/l, P < 0.001) and deceased (9.8 +/- 4.8 vs. 7.2 +/- 1.5 mmol/l, P < 0.001) groups. A known history of diabetes [odds ratio (OR) 3.0, 95% confidence interval (CI) 1.4, 63; P = 0.005] and FPG >= 7.0 mmol/l before steroid treatment (OR 3.3, 95% Cl 1.4, 7.7, P = 0.006) were independent predictors of death. During the course of the illness, FPG levels were negatively associated with SaO(2) (P = -0.682 +/- 0.305, P = 0.025, general estimation equation model) in SARS patients. Survival analysis showed that FPG was independently associated with an increased hazard ratio (HR) of mortality (HR = 1.1, 95% CI 1.01 1.1, P = 0.001) and hypoxia (HR = 1.1, 95%, CI 1.0, 1.1, P = 0.002) after controlling for age and gender. Conclusions A known history of diabetes and ambient hyperglycaemia were independent predictors for death and morbidity in SARS patients. Metabolic control may improve the prognosis of SARS patients.