Jérémy Seurat, Yuxin Tang, Thu Thuy Nguyen, France Mentré, Hervé Le Nagard, on behalf of the PFIM group (1)
(1) IAME, INSERM, UMR 1137, University Paris Diderot, Paris, France
Introduction: Nonlinear mixed effect models are increasingly used for the analysis of longitudinal studies in drug development. Using the Fisher Information Matrix (FIM) to optimize the design of these studies is an efficient alternative to clinical trial simulation. PFIM 4.0 [1] is one of the R program devoted to the design evaluation and optimisation. S4 is an object-oriented programming language which can offer several advantages compared to the traditional programming [2].
Objectives: To program in R S4 language, a new version of PFIM in order to increase its simplicity of use, its comprehensibility and its modularity.
Methods: PFIM was re-programmed from scratch in R S4 language, according to a top-down approach. The conception of the new PFIM is based on multiple classes and inheritances, which can be represented as a class diagram, with PFIM as a central object. The different programmed classes and objects are conceived to be easily used or modified for programmers and users of PFIM. The FIM is evaluated by first order linearisation of the model [3], as in PFIM 4.0. Under given design constraints and based on the D-criterion, the design is optimised using a multiplicative algorithm [4] which is a new feature as compared to PFIM 4.0. Several tests and examples were performed during the new PFIM programming process. The examples were composed of models with one or two responses (e.g. PK/PD model), expressed as analytical solutions or as ordinary differential equations. The results were compared to those obtained with PFIM 4.0.
Results: The new PFIM, by its conception, is different from PFIM 4.0. First, the use of the program is closer to most of R packages than PFIM 4.0. For the different tested examples, the evaluated FIM is consistent with the one obtained with PFIM 4.0. The different elements of a project as the model, the design (evaluated or optimized), the evaluated FIM and predicted standard errors (SE) of parameters can be stored as objects. Moreover, the project can be easily saved and reloaded. Design optimisation through the multiplicative algorithm allows to optimize the number of arms, measuring times and doses simultaneously. After performing design evaluation and/or optimisation, the results are displayed in a summary, with the different elements that could be manipulated in R
Conclusions: There is a need to increase the use of model based optimal design approaches, as it can anticipate ‘fatal’ studies. The new version of PFIM fulfill some needs by its usability. Nevertheless, this PFIM version is not final: some features implemented in PFIM 4.0. have to be also implemented in the new PFIM (e.g. Fedorov-Wynn and Simplex algorithms for design optimisation, Bayesian FIM to give shrinkage predictions [5], discrete covariates and Wald test power predictions [6]). The perspectives are also to implement new features such as alternative methods to evaluate the FIM (e.g. MC/AGQ [7]) for discrete response models. It should also be of interest to increase the interoperability with estimation parameter software tools, as aimed by the DDMoRe project.
References:
[1] Dumont C, Lestini G, Le Nagard H, Mentré F, Comets E, Nguyen TT, et al. PFIM 4.0, an extended R program for design evaluation and optimization in nonlinear mixed-effect models. Comput Methods Programs Biomed. 2018;156:217–29.
[2] Chambers JM. Object-Oriented Programming, Functional Programming and R. Stat Sci. 2014;29:167–80.
[3] Mentré F, Mallet A, Baccar D. Optimal Design in Random-Effects Regression Models. Biometrika. 1997;84:429–42.
[4] Yu Y. Monotonic convergence of a general algorithm for computing optimal designs. Ann Stat. 2010;38:1593–606.
[5] Combes FP, Retout S, Frey N, Mentré F. Prediction of shrinkage of individual parameters using the bayesian information matrix in non-linear mixed effect models with evaluation in pharmacokinetics. Pharm Res. 2013;30:2355–67.
[6] Retout S, Comets E, Samson A, Mentré F. Design in nonlinear mixed effects models: optimization using the Fedorov-Wynn algorithm and power of the Wald test for binary covariates. Stat Med. 2007;26:5162–79.
[7] Ueckert S, Mentré F. A new method for evaluation of the Fisher information matrix for discrete mixed effect models using Monte Carlo sampling and adaptive Gaussian quadrature. Comput Stat Data Anal. 2017;111:203–19.
Reference: PAGE 28 (2019) Abstr 9003 [www.page-meeting.org/?abstract=9003]
Poster: Study Design