June FabianRobert KalyesubulaJoseph MkandawireChristian Holm HansenDorothea NitschEustasius MusengeWisdom P NakangaJosephine E PrynnGavin DreyerTracy SnymanBilly SsebunnyaMichele RamsayLiam SmeethStephen TollmanSaraladevi NaickerAmelia CrampinRobert NewtonJaya A GeorgeLaurie Tomlinson2024-03-252024-03-252022https://hdl.handle.net/10539/38200Background The burden of kidney disease in many African countries is unknown. Equations used to estimate kidney function from serum creatinine have limited regional validation. We sought to determine the most accurate way to measure kidney function and thus estimate the prevalence of impaired kidney function in African populations. Methods We measured serum creatinine, cystatin C, and glomerular filtration rate (GFR) using the slope-intercept method for iohexol plasma clearance (mGFR) in population cohorts from Malawi, Uganda, and South Africa. We compared performance of creatinine and cystatin C-based estimating equations to mGFR, modelled and validated a new creatinine-based equation, and developed a multiple imputation model trained on the mGFR sample using age, sex, and creatinine as the variables to predict the population prevalence of impaired kidney function in west, east, and southern Africa. Findings Of 3025 people who underwent measured GFR testing (Malawi n=1020, South Africa n=986, and Uganda n=1019), we analysed data for 2578 participants who had complete data and adequate quality measurements. Among 2578 included participants, creatinine-based equations overestimated kidney function compared with mGFR, worsened by use of ethnicity coefficients. The greatest bias occurred at low kidney function, such that the proportion with GFR of less than 60 mL/min per 1·73 m² either directly measured or estimated by cystatin C was more than double that estimated from creatinine. A new creatinine-based equation did not outperform existing equations, and no equation, including the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) 2021 race-neutral equation, estimated GFR within plus or minus 30% of mGFR for 75% or more of the participants. Using a model to impute kidney function based on mGFR, the estimated prevalence of impaired kidney function was more than two-times higher than creatinine-based estimates in populations across six countries in Africa. Interpretation Estimating GFR using serum creatinine substantially underestimates the individual and populationlevel burden of impaired kidney function in Africa with implications for understanding disease progression and complications, clinical care, and service provision. Scalable and affordable ways to accurately identify impaired kidney function in Africa are urgently needed.enArticle