In the CHAP randomized placebo-controlled trial of cotrimoxazole prophylaxis in HIV-infected Zambian children conducted between 2001 and 2003, cotrimoxazole was associated with significant mortality reductions. In a secondary analysis we used Cox regression models to estimate the association between adherence measured by bottle weights and caregiver report and subsequent mortality in children surviving >28 days (n = 496, 153 deaths). Adherence was high and similar in both cotrimoxazole and placebo groups; adherence from bottle weights was 100% at 71% of visits, while caregivers reported 100% adherence at 79% of visits. Every 10% lower adherence to cotrimoxazole or placebo measured by bottle weights was associated with a 10-11% increase in mortality risk. Effects remained after adjustment for baseline predictors of survival and for current and recent change in primary caregiver. Caregiver-reported adherence was not associated with survival. The association between bottle-weight adherence to placebo and survival is likely capturing unmeasured caregiver effects, whose identification will be essential for quantifying the impact of antiretroviral therapy (ART) adherence on clinical outcomes in children.