OBJECTIVE: The Nutritional Support for Africans Starting Antiretroviral Therapy (NUSTART) trial was designed to determine whether nutritional support for malnourished HIV-infected adults starting antiretroviral therapy (ART) can improve early survival. Appetite is related to health outcomes in this population, but the optimal appetite metric for field use is uncertain. We evaluated two measures of appetite with the goal of improving understanding and treatment of malnutrition in HIV-infected adults. DESIGN: Longitudinal cohort study embedded in a clinical trial of vitamin and mineral-fortified, v. unfortified, lipid-based nutritional supplements. SETTING: HIV clinics in Mwanza, Tanzania and Lusaka, Zambia. SUBJECTS: Malnourished (BMI<18.5 kg/m2) HIV-infected adults starting ART. RESULTS: Appetite measurements, by short questionnaire and by weight of maize porridge consumed in a standardized test, were compared across time and correlated with changes in weight. Appetite questionnaire scores, from polychoric correlation, and porridge test results were normally distributed for Tanzanians (n 187) but clustered and unreliable for Zambians (n 297). Among Tanzanian patients, the appetite score increased rapidly from referral for ART, plateaued at the start of ART and then increased slowly during the 12-week follow-up. Change in appetite questionnaire score, but not porridge test, correlated with weight change in the corresponding two-week intervals (P=0.002) or over the whole study (P=0.05) but a point estimate of hunger did not predict weight change (P=0.4). CONCLUSIONS: In Tanzania change in appetite score correlated with weight change, but single point measurements did not. Appetite increases several weeks after the start of ART, which may be an appropriate time for nutritional interventions for malnourished HIV-infected adults.