Background and importance Estimating glomerular filtration rate (GFR) in critically ill patients is challenging due to fluctuations in kidney function and creatinine production. Creatinine clearance computed from a 24-hour (CrCl_24h) urine collection cannot always be performed. Therefore, equations based on serum creatinine are commonly used to estimate GFR. However, it is still questionable which formula performs best in this setting.

Aim and objectives We aimed to assess the performance of different serum creatinine-based equations to estimate GFR in critically ill patients.

Material and methods Observational retrospective study conducted in four intensive care units of a tertiary hospital from January to September 2020, consecutive patients with a measured CrCl_24h were included. CrCl_24h was compared to the most commonly used GFR estimating equations: Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI), Modification of Diet in Renal Disease (MDRD-4) and Cockcroft–Gault (CG). Pearson coefficients were estimated to evaluate the relationship between CrCl_24h and CKD-EPI, MDRD-4 or CG. Bland and Altman plots, bias and precision were performed to contrast CrCl_24h values with estimated GFR. Data were stratified into patients with CrCl_24h between 0 and 129 mL/min/1.73m² and patients with an augmented renal clearance (ARC) (GFR ≥130 mL/min/1.73m²).

Results 261 patients were included in the study (60.2% male, with a mean±SD age of 62±15 years and a serum creatinine of 1.23±1.00 mg/dL).

For the subgroup with GFR between 0 and 129 mL/min/1.73m², Pearson coefficients estimated for CKD-EPI, MDRD-4 and CG were 0.729, 0.637 and 0.680, respectively. Bland and Altman plots showed homogenous distribution for CKD-EPI and CG but were less homogenous for MDRD-4. No statistically significant differences were found between equations in terms of bias and precision.

For the subgroup with GFR ≥130 mL/min/1.73m², Pearson coefficients estimated for CKD-EPI, MDRD-4 and CG were 0.312, 0.329 and 0.388, respectively. Bland and Altman plots showed homogenous distribution for CG and more heterogenous distribution for CKD-EPI and MDRD-4. Bias was statistically different between CKD-EPI and both CG and MDRD-4 (p=0.0032) but precision was not (Figure 1).

• Download figure
• Open in new tab
• Download powerpoint

Abstract 4CPS-123 Figure 1

Conclusion and relevance According to the data, no differences were found between formulas to estimate GFR for critically ill patients with a CrCl_24h between 0 and 129 mL/min/1.73m²; whereas for patients with ARC, CG and MDRD-4 seemed to be more appropriate for estimating GFR.

Conflict of interest No conflict of interest