Predictive performance of glomerular filtration rate equations based on cystatin C, creatinine and their combination in critically ill patients.

Abstract

Objective: 24-hour urine creatinine clearance (ClCr 24 hours) remains the gold standard for estimating glomerular filtration rate (GFR) in critically ill patients; however, simpler methods are commonly used in clinical practice. Serum creatinine (SCr) is the most frequently used biomarker to estimate GFR; and cystatin C, another biomarker, has been shown to reflect GFR changes earlier than SCr. We assess the performance of equations based on SCr, cystatin C and their combination (SCr-Cyst C) for estimating GFR in critically ill patients.

Methods: Observational unicentric study in a tertiary care hospital. Patients with cystatin C, SCr and ClCr 24 hours measurements in ±2 days admitted to an intensive care unit were included. ClCr 24 hours was considered the reference method. GFR was estimated using SCr-based equations: Chronic Kidney Disease Epidemiology Collaboration based on creatinine (CKD-EPI-Cr) and Cockcroft-Gault (CG); cystatin C-based equations: CKD-EPI-CystC and CAPA; and Cr-CystC-based equations: CKD-EPI-Cr-CystC. Performance of each equation was assessed by calculating bias and precision, and Bland-Altman plots were built. Further analysis was performed with stratified data into CrCl 24 hours <60, 60-130 and ≥130 mL/min/1.73 m².

Results: We included 275 measurements, corresponding to 186 patients. In the overall population, the CKD-EPI-Cr equation showed the lowest bias (2.6) and best precision (33.1). In patients with CrCl 24 hours <60 mL/min/1.73 m², cystatin-C-based equations showed the lowest bias (<3.0) and CKD-EPI-Cr-CystC was the most accurate (13.6). In the subgroup of 60≤ CrCl 24 hours <130mL/min/1.73 m², CKD-EPI-Cr-CystC was the most precise (20.9). However, in patients with CrCl 24 hours ≥130mL/min/1.73 m², cystatin C-based equations underestimated GFR, while CG overestimated it (22.7).

Conclusions: Our study showed no evidence of superiority of any equation over the others for all evaluated parameters: bias, precision and Lin's concordance correlation coefficient. Cystatin C-based equations were less biased in individuals with impaired renal function (GFR <60 mL/min/1.73 m²). CKD-EPI-Cr-CystC performed properly in patients with GFR from 60-130 mL/min/1.73 m² and none of them were accurate enough in patients ≥130 mL/min/1.73 m².