OBJECTIVE: The aim of the study was to assess whether a simple, routinely available measure of antiretroviral therapy (ART) adherence predicts viral rebound at the next HIV viral load (VL) measurement in virally suppressed patients. METHODS: The analysis was performed on the Royal Free HIV Cohort, London, UK. Each 'drug coverage-viral load episode' (DCVL episode) was defined as a 6-month period immediately prior to a VL < or =50 HIV-1 RNA copies/mL (time-zero), during which the patient had been continuously on HAART, with all measured VLs < or =50 copies/mL. The next VL after time-zero was used to assess whether VL rebound (defined as >200 copies/mL) had occurred. Drug coverage, our measure of adherence, was calculated as the proportion of days in the 6-month period covered by a valid prescription for at least three antiretroviral drugs. RESULTS: A total of 376 (2.4%) VL rebounds occurred in 15 660 DCVL episodes among 1632 patients. Drug coverage was 100% for 32% of episodes, 95-99% for 16% of episodes and < or =60% for 10% of episodes. The risk ratio of rebound associated with a 10% increase in drug coverage, adjusted for potential confounding variables, was 0.93 (95% confidence interval 0.88-0.98). CONCLUSIONS: Antiretroviral drug coverage assessed at the time of VL measurement in patients with undetectable VL is potentially clinically useful for predicting VL rebound at the next VL measurement.