Purpose: The aim of this study was to evaluate the performance of the NONMEM prior functionality compared to a full Bayesian method when applied to population physiological models using diazepam as a case study.
Methods: Whole-body physiologically based pharmacokinetic (WBPBPK) models for diazepam were initially developed, tested and calibrated for rats and man using a full Bayesian analysis as implemented in WINBUGS: . The final models were implemented in NONMEM and the results from the two analyses compared in terms of parameter estimates, measures of parameter precision and run times.
Results: NONMEM population parameter estimates were in close agreement with those produced by the Bayesian analysis although there was a substantial shortening of run time for both the animal WBPBPK model (4.5 vs. 21 h) and human WBPBPK models (2 vs. 167 h). The adequacy of the model and the final parameter estimates were judged to be sufficient by the model's ability to describe individual tissue concentration-time profiles. The model provided a good overall description of the plasma concentration-time data in both rat and man with comparable parameter precision. A limited nonparametric bootstrap (n = 50) was performed to assess parameter sensitivity, bias and imprecision. No systematic bias was seen when comparing bootstrap means to final parameter estimates.
Conclusions: The ease of implementation and reductions in run time hopefully provide a further step forward in allowing the wider use of these complex and information-rich models together with clinical data in the future.