A Bayesian Nonparametric Approach to AUC Determination in Population Studies

AUC is an important index of exposure to a drug. Being able to estimate the population AUC without passing through the individual AUCs is of interest because the number of plasma samples that can be collected in each subject is often limited by technical, ethical and cost reasons. In the literature, ad hoc solutions have been proposed toestimate the population AUC when particular sampling schemes are adopted. However, restrictive assumptions are made in order to derive the precision of the estimated AUC. In this work, a new approach is proposed and tested on real data. AUC estimation is performed through a Markov Chain Monte Carlo algorithm.