There is no consensus on a gold standard for monitoring adherence to antiretroviral therapy (ART). We compared different adherence monitoring tools in predicting virologic failure as part of a clinical trial. HIV-infected Thai and Cambodian children aged 1-12 years (N= 207) were randomized to immediate-ART or deferred-ART until CD4% < 15%. Virologic failure (VF) was defined as HIV-RNA > 1000 copies/mL after ≥ 6 months of ART. Adherence monitoring tools were: (1) announced pill count, (2) PACTG adherence questionnaire (form completed by caregivers), and (3) child self-report (self-reporting from children or caregivers to direct questioning by investigators during the clinic visit) of any missed doses in the last 3 days and in the period since the last visit. The Kappa statistic was used to describe agreement between each tool. The median age at ART initiation was 7 years with median CD4% 17% and HIV-RNA 5.0 log10copies/mL and 92% received zidovudine/lamivudine/nevirapine. Over 144 weeks, 13% had VF. Mean adherence by announced pill count before VF in VF children was 92% compared to 98% in children without VF (p = 0.03). Kappa statistics indicated slight to fair agreement between tools. In multivariate analysis adjusting for gender, treatment arm ethnicity and caregiver education, significant predictors of VF were poor adherence by announced pill count (OR 4.56; 95%CI 1.78-11.69), reporting any barrier to adherence in the PACTG adherence questionnaire (OR 7.08; 95%CI 2.42-20.73), and reporting a missed dose in the 24 weeks since the last HIV-RNA assessment (OR 8.64; 95%CI 1.96-38.04). In conclusion, we recommend the child self-report of any missed doses since last visit for use in HIV research and in routine care settings, because it is easy and quick to administer and a strong association with development of VF.