Senior Principal Scientific Product Specialist Simulations Plus, France
Disclosure(s):
Monika Twarogowska, PhD: No financial relationships to disclose
Objectives: When clinical trial data is used to fit population PK/PD models, one may design the trial to optimize the parameter estimation of the model [1]. This consists of determining sampling times, number of individuals, dose groups, etc. that enable estimating the population parameters with good confidence (i.e., with low RSEs). A common approach uses the Fisher information matrix (FIM) and a first-order linearization around the population parameters to obtain the predicted RSEs for a given design. This results in fast computation and provides RSEs similar to the simulation approach [2]. The FIM approach is implemented in several tools, the most popular ones being popED [3] and PFIM [4]. They take as input a candidate design and a proposed popPK/PD model and return the predicted RSEs. Their usage is limited by the fact that they use a model definition language which differs from the language used by nonlinear mixed-effect model estimation software such as Nonmem or Monolix. To circumvent this limitation, we have developed a new R package called mlxDesignEval, which also implements design evaluation via the FIM approach but uses a Monolix project, Simulx project or mlxtran model as input. In this work, we present a comparison of mlxDesignEval to popED and PFIM in terms of available features in the R packages and results obtained, extending the comparison presented by Nyberg et al. [2].
Methods: To compare features, we used the documentation of popED and PFIM [3,4]. To compare results, we reused the 14 examples from the popED documentation which cover a large variety of cases. Each example was implemented in both popED and mlxDesignEval and the computed RSEs were compared.
Results: The implementation of mlxDesignEval uses the lixoftConnectors R package, an API for MonolixSuite. This allows a Monolix project, Simulx project or model defined in the mlxtran language to be used as input. Standard PK models can be selected directly from the large library of models in MonolixSuite and will use an analytical solution when available. ODE-based models can easily be defined too and are efficiently solved with the C++ ODE solver available in MonolixSuite. The definition of the trial design is user-friendly and follows the same syntax as a simulation in Simulx. As in popED and PFIM, the following capabilities are available in mlxDesignEval: multiple designs, multiple administration routes, multiple outputs, continuous and categorical covariates, inter-individual and inter-occasion variability including correlation terms, fixed parameters, and uncertainty on the assumed input model parameters. mlxDesignEval does not yet support shrinkage calculation and design optimization, though this functionality is planned for a future version. The examples from the popED documentation were re-implemented using mlxDesignEval. For all examples, the predicted RSEs differed by less than 1% compared to popED.
Conclusions: mlxDesignEval offers the same functionality as popED and PFIM (with the exception of shrinkage calculation and design optimization) but with a much easier syntax and direct support of Monolix or Simulx projects. The comparison of estimated RSEs shows identical results and thus validates the mlxDesignEval implementation.
Citations: [1] Mentré, F., Chenel, M., Comets, E., Grevel, J., Hooker, a, Karlsson, M. O., Lavielle, M., & Gueorguieva, I. (2013). Current Use and Developments Needed for Optimal Design in Pharmacometrics: A Study Performed Among DDMoRe’s European Federation of Pharmaceutical Industries and Associations Members. CPT: Pharmacometrics & Systems Pharmacology, 2(March), e46. [2] Nyberg, J., Bazzoli, C., Ogungbenro, K., Aliev, A., Leonov, S., Duffull, S., Hooker, A. C., & Mentré, F. (2015). Methods and software tools for design evaluation in population pharmacokinetics-pharmacodynamics studies. British Journal of Clinical Pharmacology, 79(1), 6–17. [3] https://andrewhooker.github.io/PopED/index.html [4] https://iame-researchcenter.r-universe.dev/PFIM
