Mark Lovern (1), Matteo Gazzin (1), Daniel Röshammar (1), Nathan Teuscher (1,2)
(1) InSilicoTrials, Italy (2) Teuscher Solutions LLC, USA
Objectives: We are working to develop cloud based trial simulation tools that reduce the effort required to leverage existing pharmacometrics models and permit the incorporation of these models into more complex simulation frameworks.
Introduction: The InSilicoTrials NONMEM simulator is a cloud-based application that endows users with the power to easily and quickly simulate from NONMEM models without having to modify the original model file or create a simulation data set. To run a simulation, the user simply identifies a NONMEM model file and final parameter estimate file, and inputs key simulation features through an intuitive graphical interface. Current options include dosing regimen(s), the number of subjects, desired observation time points, covariate distributions, and the number of simulation replications. Upon simulation execution, the NONMEM automatically generates the requisite simulation input files for NONMEM execution. Once the simulations have completed, a secure output page is created in which simulation outputs may be viewed and downloaded. This NONMEM simulator is the first member of a growing suite of cloud-based simulation applications which are being developed specifically for the purpose of simplifying simulation for modelers and putting the power of modeling and simulation directly in the hands of non-modellers. Forthcoming constituents of our suite will support additional modeling platforms, linking multiple models together to form complex simulation workflows, and clinical trial simulations.
Methods: The NONMEM Simulator is integrated within the InSilicoTrials platform, a modern, cloud-native, and modular web platform that utilizes a broad array of Azure Cloud Services, including Azure Storage, Azure Batch, Azure B2C, and Azure Database, among others.
Upon receiving a simulation execution request via the InSilicoTrials web interface, the system triggers an asynchronous operation that dynamically launches a dedicated NONMEM engine through Azure Batch. This engine operates exclusively for the duration of the simulation, ensuring efficiency by automatically terminating post-execution, with scalability facilitated by Azure Batch’s capacity to simultaneously run multiple dedicated engines in parallel.
Results: The InSilicoTrials NONMEM simulator was used to simulate from 5 different NONMEM models. For each model the model (.mod) file and parameter estimate (.ext) file were used with standard values for the number of subjects, desired observation time points, covariate distributions, and number of simulations. The simulation output were generated for each of the 5 models and available within the platform for download by the user.
Conclusions: Simulations from NONMEM models is a key activity required in most pharmacometrics projects; however, the process to perform simulations can often be time-consuming. Simulation using NONMEM requires updates to the model file by adding the final parameter estimates, and creating of a specific input file that includes both the desired time points and the number of subjects, before the simulation can be performed. The use of another platform (like R) requires re-coding the model in a new language and verifying that the model code conversion is accurate. The InSilicoTrials NONMEM simulator simplifies the process of creating NONMEM simulations for the user. Furthermore, simulation inputs and outputs are stored on a secure platform with a complete audit trail on all files to ensure compliance with company procedures and regulatory agency expectations. This NONMEM simulator is the first product that can automate NONMEM simulation in a cloud-based platform and is an initial step toward a comprehensive suite of cloud-based simulation solutions.
References:
[1] Text for reference 1.
[2] Text for reference 2, etc etc
Reference: PAGE 32 (2024) Abstr 10771 [www.page-meeting.org/?abstract=10771]
Poster: Methodology - New Modelling Approaches