Géraldine Ayral (1), Monika Twarogowska (1), Pauline Traynard (1), Jonathan Chauvin (1)
(1) Lixoft, Antony, France
Objectives: During drug development, the elaboration of population PK/PD models allows to better characterize the behavior of a drug candidate. Developed models can also be used to simulate the expected outcome in new situations such as future clinical trials. Simulating from a population PK/PD model is a simple task, but designing a flexible and user-friendly tool is challenging. Several simulation tools are currently available (Berkeley Madonna, mrgsolve, NONMEM, PKPDsim, RxODE, Simulo), but they usually have one of the following disadvantages [1]:
- They use a model language different from the model languages that can be used for estimation. The translation of the model to another language before simulation is time-consuming and error prone.
- They are script-based which requires coding skills for the user.
In this poster, we present Simulx-GUI, a tool for advanced simulations integrated within the MonolixSuite. Models estimated with Monolix (population PK/PD) or PKanalix (compartmental analysis) can be exported to Simulx in a single step. Not only will the model itself be exported, but also the estimated population parameters, the EBEs, the treatments and covariates defined in the Monolix data set, etc. These elements can be reused for the simulations or ignored. The clear graphical user interface allows for the definition of new simulations inputs, the setup of the simulation itself and its post-processing in a straightforward way. Simulx-GUI uses the same calculation engine in efficient C++ code and same model language as the other applications of the suite.
Methods: The functionalities of Simulx-GUI are illustrated using a PK/PD model for the analgesic Remifentanil and its effect on EEG. The goal is to simulate a clinical trial with 2 arms corresponding to two different infusion rates. As this model has been developed in Monolix, the Monolix run is directly exported to Simulx-GUI. This creates in the Simulx-GUI interface the following elements: the model, the population parameters, the design (treatments and measurement times) of the data set used to estimate the parameters as well as the covariates. To setup the simulation, new elements are created. We create two treatment elements by giving the start time, the dose amount and the infusion duration. The first corresponds to a 25ug/min infusion rate for 2 hours, and the second to a 50ug/min rate. Multiple doses can be defined easily via time vectors and it is also possible to define dose amount based on covariates (e.g weight-based dosing) and cycles (e.g dose on the 5 first days of 3-weeks cycles). The model includes covariates, which can be resampled from the original data set, sampled from an external file or defined via a distribution. Two model outputs are defined: the smooth PD prediction over a fine grid and PD measurements (with residual error) at a few time points. The simulation is setup by setting the numbers of arms and individuals and assigning the desired elements to each arm. After running the simulation, plots are automatically generated for all outputs.
To analyze the clinical trial simulation results, we define an endpoint: the percentage of individuals reaching a target PD interval after 20 minutes. The uncertainty of the endpoint is then assessed via simulation replicates, for different number of individuals, and plotted.
Results: Only a few clicks are necessary to setup, run and plot the simulation. The simulation of the predicted EEG for a large number of individuals shows that 68% of the individuals are within the target EEG interval of 5-15Hz for the higher infusion rate and 53% for the lower infusion rate. When simulating a clinical trial with a limited number of individuals, it appears that at least 50 individuals per arm are required to have an at least 85% chance to demonstrate that a significantly larger proportion of individuals will reach the EEG target in the 50ug/min arm compared to the other arm.
Conclusions: Simulx-GUI is a flexible, fast and integrated application for simulations. It can be used to simulate a wide variety of models: population PK/PD (including count, categorical and time-to-event), QSP and PBPK with or without inter-individual and inter-occasion variability. Clinical trial simulations can be easily setup via the definition of arms, outcomes and endpoints. Replicates of the simulations allow to assess the uncertainty of the endpoints and can also incorporate the uncertainty of the population parameters.
References:
[1] Craig Fancourt, Pratik Bhagunde, Wei Gao. A Comparison of Pharmacometrics Simulation Platforms (abstract W-091 ACoP9). J Pharmacokinet Pharmacodyn 45, 3–134 (2018).
Reference: PAGE () Abstr 9359 [www.page-meeting.org/?abstract=9359]
Poster: Methodology - Other topics