Aziz Chaouch (1,2), Thierry Buclin (2) and Valentin Rousson (1)
(1) Statistics Unit, Institute for Social and Preventive Medicine, University Hospital, Lausanne, Switzerland, (2) Division of Clinical Pharmacology and Toxicology, University Hospital, Lausanne, Switzerland
Objectives: Prediction percentiles for non-linear mixed effect models are commonly calculated using a large number of Monte Carlo simulations. In some applications (e.g. implementation of population pharmacokinetic models on miniaturized systems or calculation of confidence intervals for predicted quantiles), the computation time is critical and it is desirable to develop approximation methods to obtain such quantiles.
Methods: We present a general methodology to approximate the conditional predictive distribution (i.e. given covariate values) of the continuous response in a non-linear mixed effects model at a given time point by considering the distribution of a second order Taylor expansion of the model. We derive analytically its first four moments and match them with the corresponding moments of a flexible family of parametric distributions [1] which can accommodate different levels of skewness and kurtosis. The proposed methodology is applied to the computation of steady-state prediction percentile curves for a published population pharmacokinetic model of an anti-fungal agent [2].
Results: The proposed approach performs reasonably well and preliminary results suggest that it is either faster (for the same degree of accuracy) or more accurate (for the same amount of computation time) when compared to Monte Carlo simulations.
Conclusions: The proposed approximation method provides an alternative to Monte Carlo simulations for the computation of prediction percentiles for non-linear mixed effects models. Further validation on different parameter sets is still warranted, using a sufficiently large number of MC simulations as gold standard.
References:
[1] Jones, M.C. and Pewsey, A. Sinh-arcsinh distributions. Biometrika (2009), 96(4):761-780[2] Pascual, A. et al. Challenging recommended oral and intravenous voriconazole doses for improved efficacy and safety: population pharmacokinetics-based analysis of adult patients with invasive fungal infections. Clin Infect Dis (2012), 55(3):381-390
Reference: PAGE 24 () Abstr 3363 [www.page-meeting.org/?abstract=3363]
Poster: Methodology - Other topics