BACKGROUND: Severe hyperlactataemia and lactic acidosis are rare serious complications of antiretroviral therapy (ART). METHODS: Lactic acidosis was defined as pH < 7.35, bicarbonate < 20 mmol/l and raised lactate; hyperlactataemia as two consecutive lactates > 5 mmol/l. The case-control study of 110 cases and 220 controls(two randomly selected from treated patients by centre and calendar year) from centres in 10 countries included 40 (36.4%) female cases and 40 female controls (18.2%) (P < 0.001). Median age was 42.4 years [interquartile range (IQR, 36.0-52.5] for cases and 40 (IQR, 35.0-47.1) for controls (P = 0.013). More cases were nonwhite (41.9%) than controls (31.2%) (P = 0.032). Cases had a shorter duration of exposure to dideoxynucleosides. RESULTS: After adjusting for age, gender and current CD4 cell count, hyperlactataemia/lactic acidosis remained associated with exposure to didanosine in every category of exposure duration but was most strongly associated with exposure < 12 months. In a separate multivariable model, apart from exposure to stavudine, didanosine, or even more strongly both, age above 40 years [odds ratio (OR), 2.6; 95% confidence interval (CI), 1.08-6.29], female gender (OR, 5.97; 95% CI, 1.92-18.5) and advanced immunosuppression were independent associations (CD4 cell count 200-349, 100-199 and < 100 cells/mul: OR, 3.89, 7.58 and 8.11, respectively). INTERPRETATION: Hyperlactataemia/lactic acidosis was associated with exposure to dideoxynucleosides, female gender, advanced immunosuppression and possibly ethnicity. This has important consequences for choice of ART in resource-limited settings. The association with shorter duration of exposure may support the hypothesis of susceptibility in a small proportion of patients.