Background Antiretroviral treatment (cART) in HIV causes lipoatrophy. We examined predictors of anthropometric outcomes over 96 weeks in HIV-infected, lipoatrophic adults receiving stable cART randomised to tenofovir-emtricitabine (TDF-FTC) or abacavir-lamivudine (ABC-3TC) fixed dose combinations. Methodology/Principal Findings The STEAL study was a prospective trial of virologically suppressed participants randomised to either TDF-FTC (n = 178) or ABC-3TC (n = 179). Anthropometric assessment was conducted at baseline, weeks 48 and 96. The analysis population included those with baseline and week 96 data remaining on randomised therapy. Distribution of limb fat change was divided into four categories (≤0%, >0-10%, >10-20%, >20%). Baseline characteristics [demographics, medical history, metabolic and cardiovascular biomarkers] were assessed as potential predictors of change in percent subcutaneous limb fat using linear regression. 303 participants (85% of STEAL population) were included. Baseline characteristics were: mean (±SD) age 45 (±8) years; thymidine analogue nucleoside reverse transcriptase inhibitor (tNRTI) duration 4 (±3) years; limb fat 5.4 (±3.0)kg; body mass index 24.7 (±3.5) kg/m2. Mean (SD) limb fat gain to week 48 and 96 was 7.6% (±22.4) and 13.2% (±27.3), respectively, with no significant difference between groups. 51.5% of all participants had >10% gain in limb fat. Predictors of greater limb fat gain at week 96 were baseline tNRTI (10.3, p = 0.001), glucose >6 mmol/L (16.1, p = 0.04), higher interleukin 6 (IL-6) (2.8, p = 0.004) and lower baseline limb fat (3.8-6.4 kg - 11.2; >6.4 kg - 15.7, p trend<0.001). Conclusions/Significance Modest peripheral fat gain occurred with both TDF-FTC and ABC-3TC. Baseline factors associated with more severe lipodystrophy (lipoatrophy, baseline tNRTI, raised IL6, and glucose) predicted greater limb fat recovery at 96 weeks.