Background We evaluated the accuracy and precision of creatinine- and cystatin C-based prediction equations for estimating glomerular filtration rate compared to measured glomerular filtration rate in an antiretroviral-naive human immunodeficiency virus population. Methods The study population consisted of 100 treatment-naive HIV patients. Glomerular filtration rate was estimated using the Cockcroft–Gault, Modification of Diet in Renal Disease (MDRD) and Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equations, as well as cystatin C-based equations (CKD-EPI_{cystatin C}, cystatin C_{van Deventer} and CKD-EPI_combined)) compared to ⁵¹Cr-EDTA plasma clearance-measured glomerular filtration rate. We calculated percentage bias, standard deviation of the differences, accuracy within 15 and 30% of measured glomerular filtration rate and sensitivity and specificity for predicting measured glomerular filtration rate <60 mL/min/1.73 m². Results Bias for all estimating glomerular filtration rate equations ranged from −9.4% to 38.4%. The CKD-EPI_combined without ethnicity correction factor equation had the least bias, 2.9% (−2.9 to 8.8). Bias was higher for the Modification of Diet in Renal Disease and CKD-EPI equation with the African–American ethnicity factor (38.4 and 33.7%) than without (14.2 and 15.3%). Standard deviation of the differences ranged from 29.2% (CKD-EPI_combined without ethnicity factor) to 54.0% (Modification of Diet in Renal Disease with ethnicity factor). Accuracy within 30% of measured glomerular filtration rate ranged from 78% for CKD-EPI_combined without ethnicity factor to 56.7% for the Cockcroft–Gault equation. Sensitivity for creatinine-based equations was less than 50% and for the CKD-EPI_{cystatin C} equation was 75%. Conclusion Sensitivity of creatinine-based equations for predicting glomerular filtration rate was poor in this group of patients. The CKD-EPI_combined equation performed better than creatinine-based equations.