BACKGROUND: Cockcroft-Gault (CG) and Modification of Diet in Renal Disease (MDRD) formulae are recommended for glomerular filtration rate (GFR) estimation, but neither has been validated or directly compared longitudinally in HIV-infected patients or in Africa. METHODS: We investigated differences between formulae in baseline GFR, GFR changes and incidence of impaired GFR after initiation of antiretroviral therapy (ART) in 3,316 HIV-infected adults in Africa, considering sex, age, body mass index and baseline laboratory parameters as predictors. RESULTS: Participants were 65% women, median age 36.8 years, median weight 56.7 kg. Baseline GFR was lower using CG (median 89 ml/min/1.73 m2, 7.4% <60 ml/min/1.73 m2) versus MDRD (103 ml/min/1.73 m2, 3.1% <60 ml/min/1.73 m2). At 36 weeks, median CG-GFR increased (92 ml/min/1.73 m2), whereas MDRD-GFR decreased (96 ml/min/1.73 m2). Weight (explicitly a factor in CG only) concurrently increased to 62.0 kg. GFR changes from weeks 36-96 (after weight stabilization) were similar across formulae. By 96 weeks, 56 patients developed severe GFR impairment (<30 ml/min/1.73 m2) using one or both formulae (both n=45, CG n=7, MDRD n=4) compared with only 24 by serum creatinine alone. Multivariate models identified different sets of predictors for each formula. CONCLUSIONS: Although severe GFR impairments are similarly classified by different formulae, moderate impairments were more frequently identified using CG-GFR versus MDRD-GFR (with Black ethnicity correction factor 1.21), and creatinine alone had low sensitivity. Given overestimation in underweight patients and sensitivity to weight changes, this MDRD formula might not necessarily be superior for monitoring ART in African HIV-infected adults.