Background
Equations used to estimate kidney function for drug dosing rely on the assumption of homeostasis to translate a single‐point measurement of serum creatinine into clearance (CL). Our objective was to rank order the performance of alternate kidney function equations as predictors of drug CL in stable and unstable patients.
Methods
Data were extracted from medical records at a single center for all adult patients treated with vancomycin over a 5‐year period for population pharmacokinetic analysis. This analysis focused on comparison of nine kidney function equations as covariates of vancomycin CL. Both body surface area (BSA) indexed (ml/min/1.73 m2) and unindexed units (ml/min) of kidney function were tested, as time‐varying and time‐invariant covariates of vancomycin CL.
Results
The final data set consisted of 2640 patients (62% male, 81% white) with 6628 concentration measurements. The median (5th, 95th percentile) of measurements per patient, age, weight, body mass index (BMI) was 2 (1, 7) concentrations, 61.5 (28, 83) years, 90.0 (56.7, 147) kg, and 30.0 (20.7, 48.0) kg/m2. Unstable kidney function was documented in 43.6% of patients, primarily as acute kidney injury (AKI) on admission with improvement (19.4%) and AKI during the admission (17.4%). Models based on time‐varying kidney function estimates performed better than as time‐invariant. Kidney function estimated by the Chen method was ranked higher than other estimation methods.
Conclusions
A time‐varying kinetic estimated glomerular filtration rate method not indexed to BSA was identified as the highest ranked covariate model of vancomycin CL.