OBJECTIVES:To compare adherence to antiretroviral therapy using drug level monitoring and self-report and to explore the relation between these 2 methods and viral load measurements. METHODS:Sixty patients received a fixed-dose combination of nevirapine, stavudine, and lamivudine in a clinical study in Cameroon. Adherence was assessed every 6 months until month 36 by nevirapine minimal plasma concentration and self-report. Plasma HIV-1 viral load was determined at the same time. Analyses included 159 complete observations. RESULTS:The proportion of patients labeled as "adherent" was significantly lower using nevirapine monitoring (88.7%, 95% confidence interval [CI]: 82.7 to 93.2) than self-report (97.5%, CI: 93.7 to 99.3; P = 0.002). Virologic failure was associated with the nevirapine concentration (adjusted odds ratio [aOR] = 4.43; P = 0.018) but not with the self-reported adherence (aOR = 0.84; P = 0.9). As compared with the virologic outcome, the sensitivity of nevirapine level monitoring for predicting inadequate adherence was 20.5%, the specificity was 91.7%, the positive predictive value was 44.4%, and the negative predictive value was 78.0%. For self-report, the respective values were 2.6%, 97.5%, 25.0%, and 75.5%. CONCLUSIONS:Drug level monitoring provided a more reliable estimate of adherence than self-report. This method could be used in research settings. Operational research is required to define how to improve the accuracy of the self-report method because it is the most feasible method in clinical practice.