Associate Director Certara USA Inc Norristown, Pennsylvania, United States
Disclosure(s):
Ahmed A. Abulfathi: No financial relationships to disclose
Objectives: Optimal sampling analyses are used to address important tasks such as designing an optimal sampling scheme, or estimation of the difference in informative value of different sampling schemes. NONMEM provides a subroutine called $DESIGN [1] which has powerful features to answer such questions based on an estimated model. It can be combined with other NONMEM subroutines such as $PRIOR to include parameter uncertainty in the analysis. However, executing NONMEM $DESIGN can be a cumbersome workflow, of manually modifying the estimated model, generation of input data sets and processing of the output files from NONMEM. The NMsim R package [2] offers automation of this workflow, providing a simple interface to optimal sampling analysis without need for model reimplementation, all seamlessly in R. The objectives of this work were to: • Leverage the NONMEM $DESIGN feature using NMsim [2] to avoid the need for model reimplementation. • Describe how a NONMEM model can be easily run with $DESIGN within NMsim, allowing for an automated workflow that can be executed with a single NMsim function calls in R.
Methods: The clinical trial design evaluation and optimization examples presented in this work were based on the tutorial by Bauer et al, which includes examples 1, 2, 2b, 2c, and 3b [1]. These examples were automated and easily reproduced as illustrated step-by-step in a tutorial: • First, a dataset with relevant elementary design features was generated. • Next, NONMEM was executed to perform $DESIGN. • Finally, NONMEM output was postprocessed into a practical format for decision-making.
Results: This work demonstrates examples of using NMsim to perform design evaluation and optimization with NONMEM $DESIGN, highlighting its seamless integration into pharmacokinetic/pharmacodynamic workflows. Code examples and table outputs were provided to demonstrate how automated workflows can easily be built and implemented in a single R script, requiring no reimplementation or modification of the NONMEM model outside of the R environment. The flexibility of NMsim was also illustrated in modifying a NONMEM model file to incorporate the necessary $DESIGN features. These workflows are also available with the NMsim R package, serving as useful examples for modelers seeking to streamline their optimal design analyses.
Conclusions: Performing optimal design analyses using the state-of-the art NONMEM $DESIGN routine from within R is a valuable tool for many questions related to study design in drug development. The provided solution avoids translation and reimplementation of models and is very simple to use, reducing both analysis time and risks of mistakes.
Citations: [1] Bauer RJ, Hooker AC, Mentre F. Tutorial for $DESIGN in NONMEM: Clinical trial evaluation and optimization. CPT Pharmacometrics Syst Pharmacol. 2021 Dec;10(12):1452-1465. doi: 10.1002/psp4.12713. Epub 2021 Oct 19. PMID: 34559958; PMCID: PMC8674001. [2] Delff P (2024). NMsim: Seamless 'Nonmem' Simulation Platform. https://philipdelff.github.io/NMsim/
