Adaptive Bayesian Analysis of Serum Creatinine as a Marker for Drug-Induced Renal Impairment in an Early-Phase Clinical Trial

Abstract

BACKGROUND A concern with using creatinine for the identification of drug-induced renal impairment is that small changes in serum creatinine (SCr) that frequently are perceived as measurement bias or imprecision translate into important changes in the glomerular filtration rate. Important drug-generated changes in creatinine are difficult to detect because they are frequently observed within the reference interval. The design of a crossover drug protocol is an opportunity to use study participants as their own control to identify these small but important changes. METHODS Twenty individuals participating in a phase I clinical trial were evaluated for SCr changes beyond those expected for biological variation according to individual Z scores derived from an adaptive Bayesian model. After 2 screening tests, participants were administered either drug (n = 11) or placebo (n = 9) during the first dosing interval. A washout period followed, and drug was then administered to the group that initially received placebo, and vice versa (10 visits total per participant). RESULTS Although all creatinine values fell within the reference interval, 8 participants individually showed increased concentrations (Z scores >2.33). These 8 participants were confirmed at unblinding to have received the drug in the identified dosing period, with 1 exception. CONCLUSIONS The ability to identify a drug effect on an individual-participant basis in early-phase studies permits drug developers to recognize issues early in development and rapidly engage in risk–benefit analysis. These results suggest that SCr monitoring is able to detect early kidney dysfunction when individual-based reference intervals are used.