Since the 90's, the importance of evaluating and optimising population designs was highlighted. Indeed, pharmacologists and statisticians must specify a population design before running a study and performing estimation. This choice is crucial for an unbiased and efficient estimation of the parameters of the nonlinear mixed effects model used to analysed the data. Population designs consist of a set of elementary designs to be performed in group(s) of patients. Simulation studies confirmed that the balance between the number of groups, the number of subjects per group and the number and the allocation of the sampling times in each group strongly influence the precision of the parameter estimates, leading sometimes to unreliable results datasets. The classical approach to evaluate a design was based on clinical trial simulation; however, this approach is very time consuming and does not allow testing several designs in a reasonable delay. PFIM was the first software tool proposed in 2001 to circumvent this problem.

PFIM evaluates and/or optimises population designs based on the expression of the Fisher information matrix (FIM) in multi-response nonlinear mixed effects models.

Until 2022, PFIM was a set or R functions. Two main versions were implemented which are still maintained: a graphical user interface (GUI) package using the R software (PFIM Interface) and an R script version (PFIM). The GUI version, PFIM Interface 4.0, implements most features of the R script version of PFIM 4.0. More details on options available in PFIM Interface 4 and PFIM 4 are given in the news section. The documentations for both these versions are available in the relevant section and include detailed explanations and examples as to how to use these versions of PFIM.

Since June 2022, PFIM 5.0 is relased as an R package available in CRAN.

This new version of PFIM is implemented using the formal object oriented system S4 that defines objects having clear object oriented programming characteristics including class and argument definitions, inheritance, as well as argument checking, instantiation and implementation methods. More details on options available in PFIM 5.0 are given in the news section.

See references in the reference page.

Contact information

PFIM is maintained by the PFIM group, IAME, UMR1137, INSERM, Université Paris Cité, Paris, France.

Members of the PFIM Group:

For questions, please use the PFIM group mailing list:

For the package PFIM 5.0, if you encounter a clear bug, please file the issue with a reproducible example on GitHub:


PFIM is distributed under the terms of the GNU GENERAL PUBLIC LICENSE (GNU GPL) Version 2, June 1991. The terms of this license are in a file called COPYING found in the library. PFIM Interface and PFIM have been registered at the Agency for the Protection of Programs (APP). Please read the "Conditions of use" found in the download page.


When using PFIM 5.0 package in a publication, please cite the following reference:
Leroux, R Seurat J, Nagard H, Mentré F, and PFIM group (2022). New features in PFIM for optimal design in nonlinear mixed effects model using the R Package PFIM5.0. PAGE 30 (2022) Abstr 10183

When using PFIM 4.0 in a scientific publication, please cite the following publication:
Dumont C, Lestini G, Le Nagard H, Mentré F, Comets E, Nguyen TT, PFIM group. PFIM 4.0, an extended R program for design evaluation and optimization in nonlinear mixed-effect models. Computer Methods and Programs in Biomedicine, 2018;156:217-29

When using PFIM for Bayesian design of individual parameters and prediction of shrinkage in a scientific publication, please cite the following publication:
Combes FP, Retout S, Frey N, Mentré F. Power of the likelihood ratio test and of the correlation test using empirical Bayes estimates for various shrinkages in population pharmacokinetics. CPT: Pharmacometrics & Systems Pharmacology, 2014, 3: e109

When using the multiplicative algorithm with PFIM in a scientific publication, please cite the following publication:
Seurat J, Tang Y, Mentré F, Nguyen TT. Finding optimal design in nonlinear mixed effect models using multiplicative algorithms. Computer Methods and Programs in Biomedicine, 2021;207: 106-126