Nicolas Luyckx1, Emilie Hénin-Devessiere1, Andreas Lindauer1
1Calvagone
Introduction: Campsis (www.campsis.org) is an open-source PK/PD simulation suite running R, freely available on CRAN [1] and GitHub [2], designed to facilitate the simulation of population PK/PD models, and by extension, the simulation of clinical trials. This abstract presents a revisited way of simulating with parameter uncertainty, including random sampling from the scaled inverse Chi-squared or Wishart distributions for the OMEGA and SIGMA matrix. Objectives: •Revisit the previous process of simulating with parameter uncertainty in Campsis. •Sample parameters from various sources based on user choice: oVariance-covariance matrix (THETAs, OMEGAs and SIGMAs) oScaled inverse Chi-squared distribution (for single OMEGA or SIGMA) oScaled inverse Wishard distribution (for block(s) of OMEGAs or SIGMAs) oExternal source (e.g. manual data frame, CIR file, etc.) •Allow the users to enter min and max values for each parameter and check these limits during parameter sampling •When parameters are sampled from the variance-covariance matrix, check the OMEGA or SIGMA matrix for positive definiteness Methods: Design The sampling of parameter uncertainty was implemented in the package Campsismod as a generic iterative process. First, the model parameters are sampled. Second, the model parameters are checked based on the minimal and maximal values provided by the user and based on the positive definiteness of the OMEGA or SIGMA matrix. If the parameters are not valid, the process is repeated until valid parameters are found. This iterative process is called once to sample the parameters from the variance-covariance matrix (THETAs, and OMEGAs and SIGMAs on user’s request) and is called a second time to sample the OMEGAs and SIGMAs from the scaled inverse Chi-squared or Wishart distributions. Before sampling the OMEGAs/SIGMAs, the structural blocks of the OMEGA/SIGMA matrix are automatically detected. Different degrees of freedom corresponding to these blocks may be entered by the user. Inputs •Campsis model: it includes all model parameters and the variance-covariance matrix. All parameters include min and max values. •Settings for random sampling: oBy default, parameters are randomly sampled from the variance-covariance matrix oWishart argument: OMEGAs/SIGMAs are sampled from the scaled inverse Chi-squared or Wishart distribution. In that case, degrees of freedom are required. oodf argument: degrees of freedom for the OMEGAs, single value (same for all) or a vector (one specific value per OMEGA block) osdf argument: degrees of freedom for the SIGMAs, single value (same for all) or a vector (one specific value per SIGMA block) oCheck min/max values: TRUE/FALSE oCheck OMEGA/SIGMA matrix for positive definiteness: TRUE/FALSE •Import from data frame: alternatively, model parameters can be provided in the form of an external data frame from a different source (e.g. from a bootstrap or SIR run). Output The output of the procedure is an object containing the original Campsis model and the set of sampled parameters, in the form of a data frame, ready for simulations with uncertainty. Results: Distributions of the parameter replicates can be easily visualized and checked prior to running the simulation thanks to utility functions. The replicated model object is then passed to Campsis to run the simulations. Campsis will automatically simulate each trial replicate based on the set of sampled parameters in the model object. The revisited sampling of parameter uncertainty is available in Campsis v1.6.x / Campsismod v1.2.x and will be integrated in e-Campsis [3] in the near future. Conclusion: A new, robust procedure of generating random samples representing parameter uncertainty was successfully implemented in Campsis/campsismod. The user can now choose between different sources of parameter uncertainty. The sampling process is iterative and ensures that parameters (THETAs) are within the boundaries specified by the user and that the OMEGA and SIGMA matrices are positive definite. The user can also import pre-sampled parameters from an external file. This makes Campsis a great tool for running clinical trial simulations with uncertainty.
[1] http://www.calvagone.com/ [2] https://github.com/Calvagone/campsis [3] https://calvagone.github.io/ecampsis.html
Reference: PAGE 33 (2025) Abstr 11552 [www.page-meeting.org/?abstract=11552]
Poster: Methodology - Other topics