Background.Maintaining treatment adherence among the growing number of patients receiving antiretroviral treatment in Africa is a dramatic challenge. The objective of our study was to explore the results of a computerized pill count method and to test the validity, sensitivity, and specificity of this method with respect to viral load measurement in an African setting.

Methods.We performed a prospective, observational study involving patients who received first‐line highly active antiretroviral therapy in Mozambique from 1 April 2005 through 31 March 2006. Enrolled patients had received treatment for at least 3 months before the study. For defining treatment adherence levels, pill counts were used, and the results were analyzed with viral load measurements at the end of the observation period.

Results.The study involved 531 participants. During the 12 months of observation, 137 patients left the program or discontinued first‐line therapy. Of the remaining 394 patients, 284 (72.1%) had >95% treatment adherence; of those 284 patients, 274 (96.5%) had a final viral load <1000 copies/mL. A Cox proportional hazards analysis revealed that the relationship between >95% treatment adherence and the final viral load was closer than that between >90% treatment adherence and viral load.

Conclusions.Treatment adherence >95% maximizes the results of the nonnucleoside reverse‐transcriptase inhibitor–based regimen. The pill count method appears to be a reliable and economic tool for monitoring treatment adherence in resource‐limited settings.