Development of a pharmacokinetic and Bayesian optimal sampling model for individualization of oral busulfan in hematopoietic stem cell transplantation

Therapeutic drug monitoring is used to minimize toxicity and maximize the therapeutic efficacy of busulfan, which shows high intra- and interpatient pharmacokinetic variability and erratic oral absorption. This study was designed to develop a pharmacokinetic model that could accommodate the erratic oral absorption of busulfan and to use this model to develop an optimal sparse pharmacokinetic sampling strategy to improve the precision and efficiency of therapeutic drug monitoring. Twenty-one pharmacokinetic profiles were collected from 12 patients receiving oral busulfan before hematopoietic stem cell transplantation. Each pharmacokinetic profile was defined by 5 to 9 plasma concentrations. Candidate pharmacokinetic models were initially fit to the data by maximum likelihood, with model discrimination by Akaike's Information Criterion. Maximum likelihood results were used to derive Bayesian previous parameter estimates, and D-optimal design was used to determine optimal sparse sampling strategies. Each candidate sampling strategy was tested in each patient by comparing the resultant Css obtained from the sparse strategy to the actual Css derived from each patient's full pharmacokinetic dataset. The final model was a 1-compartment model, with oral busulfan absorbed in 1 to 3 phases, and fit the data well. All limited sampling models tested were unbiased in their results, and a 4-sample scheme proved to adequately characterize busulfan pharmacokinetics, and should allow for a reduced sampling frequency for therapeutic drug monitoring.