Comparison of equations for eGFR and their predictive ability on postoperative AKI in patients undergoing CABG

Abstract

Abstract Background Estimated glomerular filtration rate (eGFR) is a recognized and available method for assessing renal functions. Owing to different design purposes and applicable objects, there are different eGFR equations in clinical practice. The aim of this study was to assess the predictive ability of different methods of eGFR for postoperative acute renal injury (AKI) in patients undergoing coronary artery bypass grafting (CABG). Methods Totally 2557 patients undergoing CABG were included. Their perioperative clinical data were collected retrospectively through the hospital information system. The predictive power of the Cockcroft-Gault equation (CG), the Berlin Initiative Study equation (BIS), the chronic kidney disease epidemiology collaboration equation (CKD-EPI), and the full age spectrum equation (FAS) was validated both in the total cohort and in the elderly (≥ 70 years) according to the applicability of each equation. The predictive power of these equations for postoperative AKI was assessed and compared using the receiver’s operating characteristic curve, the Bland-Altman plot, decision curve analysis, the net reclassification index (NRI) and integrated discrimination improvement (IDI). Results Totally 418 (16.35%) patients were diagnosed with post-operative AKI during the hospital stay, but this percentage reached 25.61% in the elderly population. The CG and CKD-EPI equations were suitable for predicting the occurrence of postoperative AKI in the total cohort, and the FAS equation performed the best (area under the curve = 0.757). The BIS, CKD-EPI, and FAS equations performed generally in the elderly population, but were more consistent and again approximately equal in terms of clinical benefit. NRI and IDI showed the FAS equation continued to perform the best in the total cohort. Conclusions The FAS equation can predict the occurrence of postoperative AKI after CABG both in the total cohort and in the elderly population. All of these equations have available predictive power in the elderly population. Clinical trial registration number: ChiCTR2200061191