Performance of creatinine-based equations to estimate glomerular filtration rate with a methodology adapted to the context of drug dosage adjustment

Aim The Cockcroft-Gault (CG) creatinine-based equation is still used to estimate glomerular filtration rate (eGFR) for drug dosage adjustment. Incorrect eGFR may lead to hazardous over- or underdosing. Methods In a cross-sectional analysis, CG was validated against measured GFR (mGFR) in 14 804 participants and compared with the Modification-of-Diet-in-Renal-Diseases (MDRD), Chronic-Kidney-Disease-Epidemiology (CKD-EPI), Lund-Malmo-Revised (LMR) and European-Kidney-Function-Consortium (EKFC) equations. Validation focused on bias, imprecision and accuracy (percentage of estimates within +/- 30% of mGFR, P30), overall and stratified for mGFR, age and body mass index at mGFR <60 mL/min, as well as classification in mGFR stages. Results The CG equation performed worse than the other equations, overall and in mGFR, age and BMI subgroups in terms of bias (systematic overestimation), imprecision and accuracy except for patients >= 65 years where bias and P30 were similar to MDRD and CKD-EPI, but worse than LMR and EKFC. In subjects with mGFR <60 mL/min and at BMI 18.5-25 kg/m(2), all equations performed similarly, and for BMI < 18.5 kg/m(2) CG and LMR had the best results though all equations had poor P30-accuracy. At BMI >= 25 kg/m(2) the bias of the CG increased with increasing BMI (+17.2 mL/min at BMI >= 40 kg/m(2)). The four more recent equations also classified mGFR stages better than CG. Conclusions The CG equation showed poor ability to estimate GFR overall and in analyses stratified for mGFR, age and BMI. CG was inferior to correctly classify the patients in the mGFR staging compared to more recent creatinine-based equations.