Bayesian individualization via sampling-based methods

We consider the situation where we wish to adjust the dosage regimen of a patient based on (in general) sparse concentration measurements taken on-line. A Bayesian decision theory approach is taken which requires the specification of an appropriate prior distribution and loss function. A simple method for obtaining samples from the posterior distribution of the pharmacokinetic parameters of the patient is described. In general, these samples are used to obtain a Monte Carlo estimate of the expected loss which is then minimized with respect to the dosage regimen. Some special cases which yield analytic solutions are described. When the prior distribution is based on a population analysis then a method of accounting for the uncertainty in the population parameters is described. Two simulation studies showing how the methods work is practice are presented.