Date of Award
2009
Degree Type
Dissertation
Abstract
Linear and non-linear mixed effects modeling techniques are of great importance for the analysis of repeated measure data collected from various pharmacokinetic (PK)-pharmacodynamic (PD) and efficacy clinical trial during drug development. Monte-Carlo simulations were used to evaluate and compare the properties of mixed-model for repeated measure (MMRM) to fixed-effects ANOVA. The methods were compared using three different response profiles, four types of correlation structures (unstructured, compound symmetry, autoregressive and Toeplitz) and three types of missing mechanisms (MCAR (missing completely at random), MAR (missing at random) and MNAR (missing not at random). The MMRM methods adequately controlled type-I error and provided unbiased estimates of treatment effects in almost all simulated scenarios. When data were analyzed using fixed-effects ANOVA approach, the type-I errors were inflated (up to 40%) and the treatment effect was either underestimated or overestimated. A population pharmacokinetic model was developed for atorvastatin acid (parent drug) and its lactone metabolite using non-linear mixed-effects modeling approach. A total of 26 subjects (N=13, healthy and N=13, patients) received 10 mg of atorvastatin once daily for a week and samples were collected after last dose. The combined parent-metabolite model that best described the data consisted of a 2-compartment PK model for atorvastatin acid and 1-compartment PK model for atorvastatin lactone. The AST was found to influence the volume of distribution of atorvastatin acid and the clearance of atorvastatin lactone. The model was validated using predictive checks. In conclusion, MMRM is based on robust statistical methodology, consistent with intent-to-treat analysis and all its features can be easily pre-specified in protocols. It is recommended that MMRM be used for primary analysis of data from phase-III trials. The PK of atorvastatin acid and its metabolite (atorvastatin lactone) were evaluated and a covariate population PK model (parent-metabolite model) for atorvastatin has been developed and validated.
Recommended Citation
Narwal, Rajesh, "Evaluation of mixed-effects modeling techniques for the analysis of longitudinal data collected from PK-PD and efficacy clinical trials" (2009). Open Access Dissertations. Paper 2272.
https://digitalcommons.uri.edu/oa_diss/2272
Terms of Use
All rights reserved under copyright.