BACKGROUND: The prevalence of potentially stigmatizing lipoatrophy in children receiving antiretroviral therapy in Southern Africa is high, affecting around a third of children. Early diagnosis of lipoatrophy is essential for effective intervention to arrest progression. METHODS: Prepubertal children receiving antiretroviral therapy were recruited from a hospital-based family HIV clinic in Cape Town and followed up prospectively. Lipoatrophy was identified and graded by consensus between 2 HIV pediatricians. A dietician performed anthropometric measurements of trunk and limb fat. Anthropometric measurements in children with and without lipoatrophy were compared using multivariable linear regression adjusting for age and gender. The most discerning anthropometric indicators of lipoatrophy underwent receiver operating characteristic curve analysis. The precision of anthropometric measurements performed by an inexperienced healthcare worker was compared with that of a research dietician. RESULTS: Of 100 recruits, 36 had lipoatrophy at baseline and a further 9 developed lipoatrophy by 15-month follow-up. Annual incidence of lipoatrophy was 12% (confidence interval [CI]: 5-20%) per person-year of follow-up. A biceps skin-fold thickness <5 mm at baseline had a sensitivity of 89% (CI: 67-100%) and a specificity of 60% (CI: 46-75%) for predicting development of lipoatrophy by 15-month follow-up. Negative and positive predictive values were 97% (CI: 91-100%) and 32% (CI: 14-50%). CONCLUSION: Biceps skin-fold thickness <5 mm in prepubertal children exposed to thymidine analogue-based antiretroviral therapy may be a useful screening tool to identify children who are likely to develop lipoatrophy. The variation in precision of measurements performed by an inexperienced healthcare worker only marginally impacted performance.