Purpose: To quantify the error in evaluating recovery from acute kidney injury (AKI) with estimated GFR (eGFR) in relation to ICU stay.
Methods: Secondary analysis performed on the database of the EPaNIC trial. In a cohort of patients who developed AKI during ICU stay we compared eGFR with measured creatinine clearance (Clcr) at ICU discharge. Recovery of kidney function was assessed by comparison with baseline eGFR and the accuracy of eGFR to detect "potential CKD status" defined by Clcr was quantified. The same analysis was performed in subgroups with different ICU stay. Multivariate regression was performed to determine independent predictors of the eGFR-Clcr difference.
Results: A total of 757 patients were included. The bias (limits of agreement (LOA)) between eGFR and Clcr at ICU discharge related to ICU stay, increasing from +1.3 (-37.4/+40) ml/min/1.73 m(2) in patients with short stay to +34.7 (-54.4/+123.8) ml/min/1.73 m(2) in patients with ICU stay of more than 14 days. This resulted in a significantly different incidence of complete recovery with the two evaluation methods and reduced sensitivity to detect "potential CKD status" with eGFR in patients with prolonged ICU stay. Independent predictors of the bias included creatinine excretion on the last day in ICU, baseline eGFR, ICU stay, gender, and age.
Conclusion: Compared to Clcr, discharge eGFR results in overestimation of renal recovery in patients with prolonged ICU stay and in reduced accuracy of "CKD staging". Since age, gender and race do not change during ICU stay the same conclusion can be drawn with regard to plasma creatinine.