Reliability of serum creatinine-based formulae estimating renal function in non-critically ill surgery patients:...

Formulae estimating glomerular filtration rate (GFR) are frequently used to guide drug dosing. The objectives of this prospective single-center study were to evaluate agreement between these equations and measured creatinine clearance (CrCl) in non-critically ill surgery patients with normal kidney function and augmented renal clearance (ARC, CrCl ≥ 130 mL/min/1.73 m²), to determine predictors for disagreement, define a GFR estimator cut-off value identifying ARC and determine the ARC prevalence and duration in non-critically ill surgical patients.

Hospitalized adult non-critically ill abdominal and trauma surgery patients were eligible for inclusion. Measured CrCl based on an 8-hour urinary collection (CrCl_8h ) was used as the primary method for determining kidney function. Agreement between equations and measured CrCl_8h was assessed in terms of precision, defined as a bias within ±10 mL/min/1.73 m². Predictors for disagreement were identified for the most precise estimator using an ordinal logistic regression model with negative bias, agreement and positive bias as outcome variables. A receiver operating characteristic (ROC) analysis was performed to identify an estimator cut-off predicting ARC, which was subsequently applied for the daily proportion of patients displaying ARC and ARC duration.

During the study period (14/11/2013 - 13/05/2014), in 232 adult non-critically ill abdominal and trauma surgery patients, all estimators tend to underestimate CrCl_8h (mean bias ranging from 17 to 22 mL/min/1.73 m²), especially in patients displaying ARC (mean bias ranging from 44 to 56 mL/min/1.73 m²). eGFR_CKD-EPI performed the best. Younger age and low ASA score independently predicted underestimation of CrCl_8h . Three different eGFR_CKD-EPI cut-offs with decreasing sensitivity and increasing specificity (84, 95 and 112 mL/min/1.73 m²) identified, respectively, 65%, 44% and 14% patients displaying ARC. The median ARC duration was 4, 4 and 3 days, respectively.

In surgical patients, eGFR frequently underestimates measured CrCl, especially in young patients with low ASA score. eGFR cut-offs predicting ARC were identified.