Objective: Several methods have been developed for the prediction of carboplatin exposure to facilitate pharmacokinetic guided dosing. The aim of this study was to develop and validate sparse data Bayesian methods for the estimation of carboplatin exposure and to validate other commonly applied techniques, such as the Chatelut formula, the Sorensen limited sampling model, and the Calvert formula, in which glomerular filtration rate was estimated with the Cockcroft-Gault, the Jelliffe, and the recently proposed Wright formulas.
Methods: Complete concentration-time curves were available for a total of 43 patients (45 courses) receiving carboplatin (265 or 400 mg/m2/day) in a 1-hour infusion for 4 consecutive days in combination with thiotepa and cyclophosphamide. A population two-compartment model was developed on an index set of 12 courses. The other 33 courses served as validation set. Bayesian estimates were generated with the population parameters by use of either one or two randomly timed samples or two samples at optimal time points determined with the D-optimality theory.
Results: The Bayesian methods provided an accurate and precise prediction of the area under the concentration-time curve (bias <4% and precision <18%). The other formulas (Sorensen model, Chatelut, and Calvert with Jelliffe, Cockcroft-Gault, and Wright) resulted in a precision >18%, whereas the Jelliffe formula and the Sorensen model resulted in a bias >12%.
Conclusion: The applicability of a Bayesian method for the prediction of the carboplatin exposure by use of one or two samples without the necessity for exact timing of infusion duration and sampling was demonstrated. The Bayesian method may be very instrumental to execute pharmacokinetic guided dosing for carboplatin.