BACKGROUND: Lactic acidosis (LA) and severe hyperlactataemia (HL) are infrequent but serious complications of antiretroviral therapy that have been associated with a high fatality rate. METHODS: In a multinational retrospective cohort study, LA was defined as arterial blood pH<7.35, bicarbonate <20 mmol/l and lactate above normal, and HL as confirmed blood lactate >5 mmol/l. Logistic regression was used to identify factors associated with fatality. Sensitivity and specificity of different case definitions as predictors of death were compared. RESULTS: The overall case-fatality rate was 19/110 (17.3%), but among acidotic patients it was 33% (16/49 cases). There were 10 asymptomatic patients and none of them died as a consequence of the event. The median lactate for fatal, non-fatal and all patients was 8.3 mmol/l (IQR 7.2-13.1), 6.4 mmol/l (IQR 5.4-7.8) and 6.7 mmol/l (IQR 5.5-8.1), respectively. After adjusting for age and current CD4(+) T-cell count, lactate >7 mmol/l (OR 6.27, 95% CI 1.13-34.93), blood bicarbonate <12 mmol/l (OR 10.02 relative to >18 mmol/l, 95% CI 1.33-75.65) and concurrent opportunistic infections (OR 8.69, 95% CI 1.45-52.22) were independently associated with case fatality. Blood lactate >7 mmol/l showed a sensitivity of 84% for fatality with a specificity of 60%, whereas bicarbonate <12 mmol/l showed a better specificity (85%) but a poorer sensitivity (42%). Bicarbonate <18 mmol/l appears to be as good as lactate <7 mmol/l at predicting death (sensitivity 90% and specificity 54%). CONCLUSIONS: Our data suggest that blood lactate >7 mmol/l and blood bicarbonate <18 mmol/l appear to predict death and might help clinicians in selecting patients who may benefit from more intense monitoring.