(M-029) Bayesian Analysis of Nonclinical Studies to Inform Regimen Selection for Tuberculosis
Monday, October 20, 2025
7:00 AM - 5:00 PM MDT
Location: Colorado A
James Clary – Allucent (US), LLC; Shayne Watson – Gates Medical Research Institute; Khisi Mdluli – Gates Medical Research Institute; Micha Levi – Gates Medical Research Institute; Debra Flood – Gates Foundation; David Hermann – Gates Foundation; Alexander Berg – IceBerg Consulting, LLC
Associate Director, Pharmacometrics Allucent, LLC, United States
Disclosure(s):
James Clary: No financial relationships to disclose
Objectives: Tuberculosis (TB) remains a leading cause of infectious-disease related mortality worldwide. To address this critical global health challenge, new and novel shorter-duration treatment regimens must be identified. Evaluation using relevant animal models of TB such as the relapsing mouse model (RMM) is an important step in prioritizing regimens for potential clinical utility. To appropriately compare regimen performance in the RMM, incorporation of prior data and information across studies is important. The objective of this work was to improve upon our previous model-based methodology [1] by adopting a Bayesian estimation approach that incorporates prior information, adjusts for influential study-level covariates and quantifies inter-study variability to generate robust estimates of regimen efficacy to guide regimen prioritization.
Methods: Individual (mouse) level relapse events from multiple RMM studies were combined in a comprehensive dataset including relevant study level covariates for analysis. The pooled dataset was analyzed using the Bayesian modeling software, Stan (as implemented via Rstan), to obtain model-based estimates describing regimen-specific relapse probabilities at each treatment duration. The resulting parameters were used to generate cure probability vs. treatment duration profiles for all regimens in the dataset and derive metrics of interest such as time to 95% cure probability (T95).
Results: An inverse Emax (Hill) model sufficiently described the relapse probability vs. treatment duration profiles for all regimens included in the analysis. Due to differences in dosing and data availability, several related regimens (i.e., similar dosing / components) were combined which improved model stability. Credible intervals for T95, adjusted for relevant study-level covariates including mycobacterial inoculum, were obtained from the final model for comparison and relative rank-ordering assessment for regimens of interest.
Conclusions: A Bayesian modeling approach was applied for analysis of RMM study data, which improved upon the mixed-effects logistic regression model previously reported1 in both performance and flexibility in incorporation of prior information. Importantly, this serves as a framework from which performance metrics may be obtained for new and novel anti-TB regimens in the context of regimens with existing data, even in cases where within-study comparisons were not possible or historical animal-level data is not fully available.
Citations: [1] Berg A, Clary J, Hanna D, Nuermberger E, Lenaerts A, Ammerman N, Ramey M, Hartley D, Hermann D. Model-Based Meta-Analysis of Relapsing Mouse Model Studies from the Critical Path to Tuberculosis Drug Regimens Initiative Database. Antimicrob Agents Chemother. 2022 Mar 15;66(3):e0179321.
