Development of a template for clinical trial simulations in COPD
D. Teutonico (1), F. Musuamba (1), H.J. Maas (2), C. van Kesteren (1), A. Facius (3), S. Yang (2), M. Danhof (1), O.E. Della Pasqua (1,2)
(1) Division of Pharmacology, Leiden/Amsterdam Center for Drug Research, Leiden, The Netherlands; (2) Clinical Pharmacology Modelling & Simulation, GlaxoSmithKline, United Kingdom; (3) Dept. of Pharmacometrics, Nycomed GmbH, Konstanz, Germany.
Objectives: The ability to quickly and easily simulate clinical trial scenarios is essential in drug development [1]. However, despite the choices of software available for modelling and simulation, none offer the possibility for systematic trial simulation and consequently for the evaluation of trial performance. The objective of this exercise is to evaluate the performance of a modular template to assess scenarios (design factors), run simulations (model predictions) and evaluate output (graphical and statistical summary).
Methods: The template was conceived in a modular manner using independent scripts and MSToolkit, an R library for clinical trial simulation [2]. The template generates in silico patients, simulate response and evaluate the efficacy of the candidate design. The user can specify the parameters of study design, flag, drop out model and missing observations. The trial population can be defined from user-defined patients with a pre-specified covariate distribution matrix or from actual demographics. Given that independent scripts are used, the template may integrate functionalities to perform simulations with different software, like NONMEM, WinBUGS and R. In this case, an automated call to NONMEM VI is used to generate FEV1 responses in a COPD trial. The results are managed by the template reporting features, which includes a series of graphical and statistical summaries.
Results: The template enables the generation of relevant scenarios in an automated manner. A series of graphical and statistical summaries are generated at the end of the simulations that highlight the changes relative to baseline (ΔFEV1) for each dose level compared to placebo, among other, plots such as FEV1 vs. time, ΔFEV1 vs. time and tables such as median and percentiles are produced and saved in a simulation folder specially created. The statistical power of the study was also correlated to the number of subjects and the effect of other parameters on the outcome was evaluated (dose, study duration, drop-out).
Conclusions: The combination of MSToolkit with modular R scripts has provided key components for the creation of design scenarios and enabled the execution, analysis and reporting of clinical trial simulation results. As this is part of an ongoing effort, it is anticipated that future functionality and versatility will also facilitate exchangeability between simulation software packages. The template can be easily adapted to suit other therapeutic applications and study design types.
References:
[1] Girard P. (2005) Clinical Trial Simulation: A Tool for Understanding Study Failures and Preventing Them. Basic & Clinical Pharmacology & Toxicology. 96, 228-234.
[2] Smith M. et al. (2009) MSToolkit - An R library for simulating and evaluating clinical trial designs and scenarios. Poster at PAGE Meeting 2009.
