Background: High-level adherence to antiretroviral therapy (ART) is associated with favourable patient outcomes. In resource-constrained settings, however, there are few validated measures. We examined the correlation between clinical outcomes and the medication possession ratio (MPR), a pharmacy-based measure of adherence. Methods: We analysed data from a large programmatic cohort across 18 primary care centres providing ART in Lusaka, Zambia. Patients were stratified into three categories based on MPR-calculated adherence over the first 12 months: optimal (≥95%), suboptimal (80-94%) and poor (<80%). Results Overall, 27 115 treatment-naïve adults initiated and continued ART for ≥ 12 months: 17 060 (62.9%) demonstrated optimal adherence, 7682 (28.3%) had suboptimal adherence and 2373 (8.8%) had poor adherence. When compared with those with optimal adherence, post-12-month mortality risk was similar among patients with sub-optimal adherence [adjusted hazard ratio (AHR) = 1.0; 95% CI: 0.9-1.2] but higher in patients with poor adherence (AHR = 1.7; 95% CI: 1.4-2.2). Those <80%; MPR also appeared to have an attenuated CD4 response at 18 months (185 cells/μl vs 217 cells/μl; P < 0.001), 24 months (213 cells/μl vs 246 cells/μl; P < 0.001), 30 months (226 cells/ μl vs 261 cells/μl; P < 0.001) and 36 months (245 cells/μl vs 275 cells/μl; P < 0.01) when compared with those above this threshold. Conclusions: MPR was predictive of clinical outcomes and immunologic response in this large public sector antiretroviral treatment program. This marker may have a role in guiding programmatic monitoring and clinical care in resource-constrained settings. © Published by Oxford University Press on behalf of the International Epidemiological Association © The Author 2009; all rights reserved.