Devin Pastoor
University of Maryland, Baltimore
Objectives: In pharmacometrics, modelers are exposed to heterogeneous data analytic projects, including data preparation, analysis, simulation, and visualization. Though each project presents unique challenges, many components of the workflow, including dealing with the behavior and quirks of R, are consistent across projects. Packages such as Xpose [1] and metrumrg [2] have been developed, their focus does not encompass specific components of the analytic process as they were developed primarily to ease interaction with NONMEM [3]. PKPDmisc aims to standardize and accelerate workflows by providing flexible and robust functions that address tasks common to many projects.
Methods: PKPDmisc is written in R, with a focus on code quality and performance. In many cases, by leveraging modern packages such as dplyr, and even writing custom C++ functions internally. As such, PKPDmisc can run 100’s or even 1000’s of times faster than traditional R implementations. PKDmisc also aims to easily address common issues with R, such as converting columns between data types, splitting data for plotting with ggplot2, and performing multi-step summary statistic calculations, where previously users may have resorted to unwieldy scripts or hacks.
Results: PKPDmisc offers functionality in the following areas:
- Quickly reading data, including non-standard nonmem simulation tables and Phoenix [4] tables with units.
- Convenient wrappers on common data manipulation operations
- Fast noncompartmental analysis
- Data preparation tasks such as handling BQL, missing data, and creating flags
- Summary statistic calculations for plots such as quantile plots
- Evaluating NONMEM run status and visualizing gradients during long runs
- Templates analysis folder structure and creating reports using Rmarkdown
- Curve stripping for initial estimates for two-compartment models
- Performing Wald’s approximation for covariate model selection with NONMEM.
- Robust resampling functions for bootstrap or clinical trial simulations.
- Professional quality default themes for plots created with ggplot2.
Conclusions: PKPDmisc is open-source, free, and transparently developed on github (github.com/dpastoor/PKDPmisc), allowing users to easily ask questions, report issues, view documentation, and contribute.
References:
[1] Jonsson, E.N. & Karlsson, M.O. (1999) Xpose–an S-PLUS based population pharmacokinetic/pharmacodynamic model building aid for NONMEM. Computer Methods and Programs in Biomedicine. 58(1):51-64.
[2] Bergsma, Timothy T., William Knebel, Jeannine Fisher, William R. Gillespie, Matthew M. Riggs, Leonid Gibiansky, and Marc R. Gastonguay. 2013. “Facilitating Pharmacometric Workflow with the Metrumrg Package for R.” Computer Methods and Programs in Biomedicine 109 (1): 77–85. doi:10.1016/j.cmpb.2012.08.009.
[3] Beal, S., Sheiner, L.B., Boeckmann, A., & Bauer, R.J., NONMEM User’s Guides. (1989-2009), Icon Development Solutions, Ellicott City, MD, USA, 2009.
[4] http://www.certara.com/products/pkpd/phx-nlmec
Reference: PAGE 24 () Abstr 3483 [www.page-meeting.org/?abstract=3483]
Poster: Methodology - Other topics