Mario González-Sales (1,2), Olivier Barriere (3), Pierre Olivier Tremblay (1), Guillaume Bonnefois (1), Julie Desrochers (1), Fahima Nekka (4)
(1) Syneos Health, Quebec, Canada, (2) Modeling Great Solutions, Escaldes-Engordany, Andorra, (3) Certara, Quebec, Canada, (4) Université de Montreal, Quebec, Canada.
Introduction: The pharmacometrics workflow has routine steps: 1) assemble the dataset, 2) explore, 3) model the data, 4) evaluate, 5) validate the model, and 6) communicate the findings. The automation of these steps saves time and money, reduces the risk of errors, and increases reproducibility. Currently, there are a number of excellent tools available to enhance steps 2-6).[1-11] However, to the best of our knowledge, there is no open-source tool to support step 1). Because of the core or ‘heart’ of each pharmacometric analysis is a dataset, and the time required to construct a pharmacometrics dataset can sometimes be higher than the effort required for the modeling per se, Syneos Health’s pharmacometrics team has created puzzle, an open-source R package that is freely available on Github (https://github.com/syneoshealth/puzzle).
Objective: The objectives of this work have been:
- To develop an R package to simplify the time consuming and error prone task of assembling pharmacometrics datasets in order to speed up the pharmacometrics workflow.
- To increase the reproducibility of pharmacometric analysis by decreasing the probability of errors during the data assembling step and facilitating the quality control task.
Methods: Puzzle consists of a series of functions written in R. These functions create, from tabulated files, datasets that are compatible with the formatting requirements of the NONMEM® software.[12] In order to facilitate its use and to decrease the slope of the learning curve, users are only required to learn the behavior and syntax of one function, puzzle(). Nevertheless, the puzzle package involves additional functions intentionally working “under the hood” to enhance the user experience. Furthermore, a web interface is also available and was developed as a shiny application that can be used for those users with little or no experience using R, or for those pharmacometricians not feeling comfortable with the R syntax.
Results: With only one function, puzzle(), complex pharmacometrics databases can easily be assembled. Users are able to select from different absorption processes such as zero- and first-order, or a combination of both. Furthermore, datasets containing data from one or more analytes, and/or one or more responses, and/or time in- and/or dependent covariates, and/or urine data can be simultaneously assembled. The output of puzzle() is always a “.csv” file that can be read by NONMEM® because it is fully compatible with its formatting rules. The puzzle package can be easily installed using the following R syntax: devtools::install_github(“syneoshealth/puzzle”).
Conclusions: The puzzle package is a powerful and efficient tool that helps modelers, programmers and pharmacometricians through the difficult and complex process of assembling pharmacometrics datasets. In particular, it is the first open-source tool supporting pharmacometricians during the challenging, error prone, and time consuming process of data assembling.
References:
[1] Keizer RJ, Van Benten M, Beijnen JH, Schellens JHM, Huitema ADR. Piraña and PCluster: a modeling environment and cluster infrastructure for NONMEM. Comput Methods Programs Biomed 2011: 101; 72-9.
[2] Lindbom L, Pihlgren P, Jonsson EN. PsN-Toolkit–a collection of computer intensive statistical methods for non-linear mixed effect modeling using NONMEM. Comput Methods Programs Biomed. 2005: 79; 241-57.
[3] Jonsson EN, Karlsson MO. Xpose–an S-PLUS based population pharmacokinetic/pharmacodynamics model building aid for NONMEM. Comput Methods Programs Biomed. 1999: 58; 51-64.
[4] Mouksassi S. Ggplot and summary statistics quick exploration of data. Ggquickeda package. Available at: https://github.com/smouksassi/ggquickeda. Accessed 22 February 2019.
[5] Mouksassi S. Rshiny app as interface to ggplot2. Ggplotwithyourdata package. Available at: https://github.com/smouksassi/ggplotwithyourdata. Accessed 22 February 2019.
[6] Keizer RJ. R library to create visual predictive checks. The vpc package. Available at: https://github.com/ronkeizer/vpc. Accessed 22 February 2019.
[7] Metrum Research Group. Mrgsolve package. Available at: https://mrgsolve.github.io. Accessed Accessed 22 February 2019
[8] Keizer RJ. R library for simulation of PKPD models defined as ODE systems. The PKPDsim package. Available at: https://github.com/ronkeizer/PKPDsim. Accessed 22 February 2019
[9] Wang W, Hallow KM, James DA.A Tutorial on RxODE: Simulating Differential Equation Pharmacometric Models in R. CPT Pharmacometrics Syst Pharmacol. 2006: 5; 3-10.
[10] Lavielle M. mlxR package. Available at: https://github.com/MarcLavielle/mlxR.
[11] RStudio team. Rmarkdown: dynamic documents for R. The rmardown package. Available at: https://rmarkdown.rstudio.com/ Accessed 22 February 2019
[12] Beal S, Sheiner LB, Boeckmann A, Bauer RJ. NONMEM User’s guide. (1989-2009). Icon development Solutions, Ellicott City, MD, USA, 2009.
Reference: PAGE 28 (2019) Abstr 8934 [www.page-meeting.org/?abstract=8934]
Poster: Methodology - Other topics