Glomerular filtration rate correlation and agreement between common predictive equations and standard 24-hour urinary creatinine clearance in medical critically ill patients

Abstract

Background Determining kidney function in critically ill patients is paramount for the dose adjustment of several medications. When assessing kidney function, the glomerular filtration rate (GFR) is generally estimated either by calculating urine creatinine clearance (UCrCl) or using a predictive equation. Unfortunately, all predictive equations have been derived for medical outpatients. Therefore, the validity of predictive equations is of concern when compared with that of the UCrCl method, particularly in medical critically ill patients. Therefore, we conducted this study to assess the agreement of the estimated GFR (eGFR) using common predictive equations and UCrCl in medical critical care setting. Methods This was the secondary analysis of a nutrition therapy study. Urine was collected from participating patients over 24 h for urine creatinine, urine nitrogen, urine volume, and serum creatinine measurements on days 1, 3, 5, and 14 of the study. Subsequently, we calculated UCrCl and eGFR using four predictive equations, the Cockcroft–Gault (CG) formula, the four and six-variable Modification of Diet in Renal Disease Study (MDRD-4 and MDRD-6) equations, and the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation. The correlation and agreement between eGFR and UCrCl were determined using the Spearman rank correlation coefficient and Bland–Altman plot with multiple measurements per subject, respectively. The performance of each predictive equation for estimating GFR was reported as bias, precision, and absolute percentage error (APE). Results A total of 49 patients with 170 urine samples were included in the final analysis. Of 49 patients, the median age was 74 (21–92) years-old and 49% was male. All patients were hemodynamically stable with mean arterial blood pressure of 82 (65–108) mmHg. Baseline serum creatinine was 0.93 (0.3–4.84) mg/dL and baseline UCrCl was 46.69 (3.40–165.53) mL/min. The eGFR from all the predictive equations showed modest correlation with UCrCl (r: 0.692 to 0.759). However, the performance of all the predictive equations in estimating GFR compared to that of UCrCl was poor, demonstrating bias ranged from −8.36 to −31.95 mL/min, precision ranged from 92.02 to 166.43 mL/min, and an unacceptable APE (23.01% to 47.18%). Nevertheless, the CG formula showed the best performance in estimating GFR, with a small bias (−2.30 (−9.46 to 4.86) mL/min) and an acceptable APE (14.72% (10.87% to 23.80%)), especially in patients with normal UCrCl. Conclusion From our finding, CG formula was the best eGFR formula in the medical critically ill patients, which demonstrated the least bias and acceptable APE, especially in normal UCrCl patients. However, the predictive equation commonly used to estimate GFR in critically ill patients must be cautiously applied due to its large bias, wide precision, and unacceptable error, particularly in renal function impairment.