This cohort study was conducted amongst female patients manifesting lipoatrophy while receiving stavudine-containing first-line antiretroviral treatment regimens at two urban health centres in Rwanda. The objectives were to assess weight evolution after stavudine substitution and to describe any significant difference in weight evolution when zidovudine or tenofovir/abacavir was used for substitution. All adult patients on stavudine-containing first-line regimens who developed lipoatrophy (diagnosed using a lipodystrophy case definition study-based questionnaire) and whose treatment regimen was changed were included (n=114). In the most severe cases stavudine was replaced with tenofovir or abacavir (n=39), and in the remainder with zidovudine (n=75). For patients changed to zidovudine a progressive weight loss was seen, while those on tenofovir/abacavir showed a progressive weight increase from six months. The between-group difference in weight evolution was significant from nine months (difference at 12 months: 2.3kg, P=0.02). These differences were confirmed by follow-up lipoatrophy scores. In multivariate analysis, substitution with tenofovir/abacavir remained significantly associated with weight gain. This is the first study in Africa assessing weight gain as a proxy for recovery after stavudine substitution due to lipoatrophy, providing supporting evidence that tenofovir/abacavir is superior to zidovudine. The weight loss with zidovudine might justify earlier substitution and access to better alternatives like tenofovir/abacavir.