BACKGROUND: Many HIV care and treatment programs in resource-limited settings rely on clinical and immunologic monitoring of antiretroviral therapy (ART), but accuracy of this strategy to detect virologic failure (VF) among children has not been evaluated. METHODS: A cross-sectional sample of HIV-infected children aged 1-16 years on ART >or=6 months receiving care at a Tanzanian referral center underwent clinical staging, CD4 lymphocyte measurement, plasma HIV-1 RNA level, and complete blood count. Associations with VF (HIV-1 RNA >or=400 copies/mL) were determined utilizing bivariable and multivariate analyses; accuracy of current clinical and immunologic guidelines in identifying children with VF was assessed. FINDINGS: Of 206 children (median age 8.7 years, ART duration 2.4 years), 65 (31.6%) demonstrated VF at enrollment. Clinical and immunological criteria identified 2 (3.5%) of 57 children with VF on first-line therapy, exhibiting 3.5% sensitivity and 100% specificity. VF was associated with younger age, receipt of nevirapine vs. efavirenz-based regimen, CD4% < 25%, and physician documentation of maladherence (P < 0.05 on bivariable analysis); the latter 2 factors remained significant on multivariate logistic regression. INTERPRETATION: This study demonstrates poor performance of clinical and immunologic criteria in identifying children with virologic failure. Affordable techniques for measuring HIV-1 RNA level applicable in resource-limited settings are urgently needed.