Nabil Semmar, Nicolas Simon
Department of Medical and Clinical Pharmacology, Medical School of Marseille, EA3784, 27 Bd Jean Moulin, F-13385 Marseille
Objectives: A high covariate number could become problematic when one needs to determine the most significant variable combination, in order to reduce the inter-individual variability (IIV). Alternatively to multiple introductions of single variables, we propose a single introduction of one multivariate variable: hierarchical cluster analysis (HCA) is a stratification method combining the initial single covariates to build a multivariate categorical covariate.
Methods: HCA stratifies the population into homogeneous and non-overlapping groups (clusters) according to similarities from the set of covariates. This uses a distance measure and a linkage algorithm. HCA approach is illustrated by a database of plasmatic cortisol in 82 patients after intravenous bolus administration of synacthen. Using NONMEM, a basic infusion model was initially achieved, and then a classical covariate selection was applied to improve IIV. Alternatively, a categorical covariate was built by HCA from the initial database containing 6 continuous covariates. Such categorization was carried out using Euclidean distance and complete-linkage algorithm. The new covariate was then included into a population PK model, and the PK results were compared to those of the basic and classical covariate models.
Results: With a classical covariate selection, the best fit was between the elimination rate constant k and the body mass index (BMI), which improved IIV of k compared with the basic model. HCA stratified the 82 patients into 5 dissimilar clusters that differed by increasing BMI, obesity duration, and waist-hip ratio. The dispersion of k according to the 5 clusters showed 3 distinct variation ranges a priori, which corresponded a posteriori (after NONMEM modeling) to 3 sub-populations of k. After grouping the overlapped clusters for k, we obtained three final clusters representing non obese, intermediate, and extreme obese sub-populations. The PK model based on 3 clusters was better than the basic model, similar to the classical covariate model, but had a stronger interpretability: it showed that the stimulation and elimination of cortisol were higher in the extreme obese followed by intermediate then non obese subjects.
Conclusions: HCA is proposed as an alternative approach for covariate selection in population pharmacokinetics, particularly when the initial covariates are numerous.
Reference: PAGE 14 (2005) Abstr 819 [www.page-meeting.org/?abstract=819]
Poster: poster