NMsim: Simulate Nonmem models seamlessly in R without any model reimplementation

Authors

Philip Delff, PhD - Vertex Pharmaceuticals

Abstract

Objectives: While Nonmem is widely used for its strengths and flexibility in estimation of pharmacometrics
models, post-processing including simulation is often preferably handled in R. The new R package
NMsim (available on CRAN) fills this gap by allowing simulation of Nonmem models directly
from R. In contrast to other powerful simulation packages, NMsim relies on Nonmem to perform
the simulation, hence eliminates the need for translating to other software.

NMsim aims at providing a seamless and close to model-independent R interface to simulation of
Nonmem models only dependent on estimation control stream and a simulation data set.

NMsim works on both ADVAN subroutines and $PRED models and can currently perform the following
types of simulations:
o New subjects (default)
o Typical subject (ETAs equal 0)
o Subjects already estimated in Nonmem model (EBE's)
o Simulation with parameter uncertainty

Moreover, NMsim provides an interface to modify control streams. This can among other things be
used to modify parameter values.

Methods: NMsim does not simulate, translate or otherwise interpret a Nonmem model. Instead, it automates
the Nonmem simulation workflow (including execution of Nonmem). In the example given above,
NMsim will do the following:
o Save the simulation input data in a csv file for Nonmem
o Create a simulation input control stream based on file.mod matching the saved simulation
data set
o Update and fix initial values based on estimate (from file.ext)
o Run Nonmem on the generated simulation control stream
o Collect output data tables, combine them, and merge with the simulation input data
o Return the collected data in R

NMsim is written in R and data.table leveraging functionality from the NMdata R package.

Results: NMsim provides a simulation interface that does not require reimplementation of a Nonmem
model. This reduces amount of work and risk of errors. It allows for simulation of all intermediate
model steps simplifying simulation-based model qualification, and for automation of a wide range
of simulation based analyses.

In this poster presentation, we showcase the capabilities of NMsim through examples of different
types of simulations of a popPK model. From the information provided on the poster and the freely
available R package, the audience will be able to perform simulations on their own models without
any further model coding.

Conclusions: By bridging the gap between Nonmem's powerful modeling capabilities and the readily availability
of a simulation interface in R, NMsim expedites model development and evaluation. We believe
that our package will facilitate collaboration and innovation in the pharmacometrics community,
ultimately advancing drug development efforts.

Citations: [1] 2024. NMdata: Preparation, Checking and Post-Processing Data for PK/PD Modeling. https://cran.rproject.
org/package=NMdata
[2] 2024. NMsim: Simulate Nonmem models from R. https://philipdelff.github.io/...
[3] 2024. NMsim: Seamless 'Nonmem' Simulation Platform. https://cran.r-project.org/web...

Keywords

Nonmem, Simulation, Automation

Date of Conference

November 10-13, 2024

Conference Location

Phoenix, Arizona, USA

DOI

10.70534/PGJU4426