Model-Based Meta-Analysis of Relapsing Mouse Model Studies from the Critical Path to Tuberculosis Drug Regimens Initiative Database

Abstract

Tuberculosis (TB), the disease caused by Mycobacterium tuberculosis (Mtb), remains a leading infectious disease-related cause of death worldwide, necessitating the development of new and improved treatment regimens. Nonclinical evaluation of candidate drug combinations via the relapsing mouse model (RMM) is an important step in regimen development, through which candidate regimens that provide the greatest decrease in the probability of relapse following treatment in mice may be identified for further development. Although RMM studies are a critical tool to evaluate regimen efficacy, making comprehensive "apples to apples" comparisons of regimen performance in the RMM has been a challenge in large part due to the need to evaluate and adjust for variability across studies arising from differences in design and execution. To address this knowledge gap, we performed a model-based meta-analysis on data for 17 unique regimens obtained from a total of 1592 mice across 28 RMM studies. Specifically, a mixed-effects logistic regression model was developed that described the treatment duration-dependent probability of relapse for each regimen and identified relevant covariates contributing to interstudy variability. Using the model, covariate-normalized metrics of interest, namely, treatment duration required to reach 50% and 10% relapse probability, were derived and used to compare relative regimen performance. Overall, the model-based meta-analysis approach presented herein enabled cross-study comparison of efficacy in the RMM and provided a framework whereby data from emerging studies may be analyzed in the context of historical data to aid in selecting candidate drug combinations for clinical evaluation as TB drug regimens.

Keywords: Mycobacterium; model-based meta-analysis; modeling and simulation; relapsing mouse model; tuberculosis.

Conflict of interest statement

The authors declare no conflict of interest.

Figures

FIG 1

Relapse proportion by treatment duration across stratified by regimen and study. Each study is grouped with lines representing the relapse-time course in a particular study. Points represent individual time points in a particular study. “RMZ/RM_BIDM” denotes a version of the RMZ/RM regimen where moxifloxacin was administered at 100 mg/kg twice daily for a total daily dose of 200 mg/kg. M, moxifloxacin; R, rifampin; Z, pyrazinamide.

FIG 2

Visual predictive check for the final model stratified by regimen. Black dots and solid lines represent the observed relapse proportion. Blue lines represent the median prediction from the final model. The red shaded area represents the 90% prediction interval. “RMZ/RM_BIDM” denotes a version of the RMZ/RM regimen where moxifloxacin was administered at 100 mg/kg twice daily for a total daily dose of 200 mg/kg. M, moxifloxacin; R, rifampin; Z, pyrazinamide.

FIG 3

Relapse probability versus treatment duration by regimen. Blue lines and areas represent the median and 95% confidence interval, respectively, of relapse probability versus treatment duration profiles for each regimen as derived by bootstrap (N = 500 runs). Black lines and areas represent HRZE/HR as the clinical standard of care regimen. For comparative purposes, all regimens are presented as normalized to the median covariate values. “RMZ/RM_BIDM” denotes a version of the RMZ/RM regimen where moxifloxacin was administered at 100 mg/kg twice daily for a total daily dose of 200 mg/kg. E, ethambutol; H, isoniazid; M, moxifloxacin; R, rifampin; Z, pyrazinamide.

FIG 4

Forest plots of T50 and T10 relapse probability by regimen. Median estimates and 95% confidence intervals from bootstrapped data sets (n = 500 runs) for (A) T50 and (B) T10 metrics. Regimens are in descending rank orders for metrics based on median value, with red and blue coloring indicating regimens for which the confidence intervals are completely above or below the median value for HRZE/HR (included as the clinical standard of care regimen). For comparative purposes, all regimens are presented as normalized to the median covariate values. “RMZ/RM_BIDM” denotes a version of the RMZ/RM regimen where moxifloxacin was administered at 100 mg/kg twice daily for a total daily dose of 200 mg/kg. E, ethambutol; H, isoniazid; M, moxifloxacin; R, rifampin; Z, pyrazinamide.

FIG 5

Simulations of the model with a range of covariate effects. Simulations of covariate effects from range of data for (A) inoculum (log10 CFU/mL) and (B) baseline bacterial burden (log10 CFU/mL).