Risk-stratified treatment for drug-susceptible pulmonary tuberculosis

Abstract

The Phase 3 randomized controlled trial, TBTC Study 31/ACTG A5349 (NCT02410772) demonstrated that a 4-month rifapentine-moxifloxacin regimen for drug-susceptible pulmonary tuberculosis was safe and effective. The primary efficacy outcome was 12-month tuberculosis disease free survival, while the primary safety outcome was the proportion of grade 3 or higher adverse events during the treatment period. We conducted an analysis of demographic, clinical, microbiologic, radiographic, and pharmacokinetic data and identified risk factors for unfavorable outcomes and adverse events. Among participants receiving the rifapentine-moxifloxacin regimen, low rifapentine exposure is the strongest driver of tuberculosis-related unfavorable outcomes (HR 0.65 for every 100 µg∙h/mL increase, 95%CI 0.54-0.77). The only other risk factors identified are markers of higher baseline disease severity, namely Xpert MTB/RIF cycle threshold and extent of disease on baseline chest radiography (Xpert: HR 1.43 for every 3-cycle-threshold decrease, 95%CI 1.07-1.91; extensive disease: HR 2.02, 95%CI 1.07-3.82). From these risk factors, we developed a simple risk stratification to classify disease phenotypes as easier-, moderately-harder, or harder-to-treat TB. Notably, high rifapentine exposures are not associated with any predefined adverse safety outcomes. Our results suggest that the easier-to-treat subgroup may be eligible for further treatment shortening while the harder-to-treat subgroup may need higher doses or longer treatment.

Fig. 1. Multivariable Hazard Ratios for Tuberculosis-Related Unfavorable Outcomes.

Multivariable analysis of pharmacokinetic and baseline predictors for a rifapentine-moxifloxacin, b rifapentine, and c control regimens. Data are presented as hazard ratio estimates (point) and 95% confidence intervals (error bars). a Xpert MTB/RIF cycle threshold <18, 29/397 (7.3); Xpert MTB/RIF cycle threshold ≥ 18, 10/296 (3.4), b Rifapentine exposure <560 µg∙h/mL, 31/402 (7.7); Rifapentine exposure ≥ 560 µg∙h/mL, 14/389 (3.6), c Age <30 years, 21/354 (5.9); Age ≥ 30 years, 54/430 (12.6), d Weight <53 kg, 45/364 (12.4); Weight ≥ 53 kg, 30/419 (7.2), e Xpert MTB/RIF cycle threshold <18, 54/397 (13.6); Xpert MTB/RIF cycle threshold ≥ 18, 13/284 (7.7), f Rifapentine exposure <560 µg∙h/mL, 58/386 (15.0); Rifapentine exposure ≥ 560 µg∙h/mL, 17/398 (4.3), g Xpert MTB/RIF cycle threshold <18, 15/399 (3.7); Xpert MTB/RIF cycle threshold ≥ 18, 5/268 (1.9), h Pyrazinamide exposure <336 µg∙h/mL, 14/304 (4.6); Pyrazinamide exposure ≥ 336 µg∙h/mL, 10/462 (2.2).

Fig. 2. Xpert MTB/RIF cycle threshold and extent of disease on chest radiography stratify participants into easier-to-treat TB, moderately-harder-to-treat TB, and harder-to-treat TB disease phenotypes.

Disease phenotypes were defined by baseline Xpert MTB/RIF cycle threshold and extent of disease on chest radiography, defined as the percent involvement of the area of the thoracic cavity. Disease phenotypes were further stratified by rifamycin exposure, where Kaplan Meier estimates demonstrated that easier-to-treat TB does not need exposure optimization. Moderately-harder-to-treat TB among participants receiving the rifapentine-regimen would require dose optimization to achieve optimal outcomes. Participants with moderately-harder-to-treat TB receiving the rifapentine-moxifloxacin regimen would benefit from dose optimization, however this would not be required to achieve optimal outcomes. Participants with harder-to-treat TB and high rifamycin exposure have similar outcomes across regimens, but none of the regimens achieve <5% tuberculosis-related unfavorable outcomes regardless of rifamycin exposure levels.

Fig. 3. Risk Stratification Reveals a Low-Risk Subgroup where Further Treatment Shortening and Simplification is Likely Possible and a High-Risk Subgroup where Longer Treatment May Be Needed.

The figure shows the results of subgroup analyses of Study 31/A5349 risk groups, data are presented as percentage point differences (point) and 95% confidence intervals (error bars). Low and high rifapentine subgroups in the experimental arms were compared to low and high rifampin subgroups in the control arm. Two-tailed interaction p-values tested for interaction between regimen (experimental vs. control) and the disease phenotypes in a Cox proportional hazards model. a Analysis of the rifapentine-moxifloxacin regimen demonstrates that the high-risk group, comprising 23% of the Study 31/A5349 population, may require a longer and/or more potent regimen to achieve ≤ 5% unfavorable outcomes. b Analysis of the rifapentine-regimen demonstrates that the subpopulations at low risk regardless of rifapentine exposure, and moderate- or high-risk with high rifapentine exposure, comprising 62% of the Study 31/A5349 population in the rifapentine arm, have small differences in outcome when compared to the control. Additionally, in both rifapentine and rifapentine-moxifloxacin regimens among participants with high rifapentine exposure, the percentage point differences between experimental and control regimens are small (<1.8%) across all risk groups.

Fig. 4. Safety of the Rifapentine-Moxifloxacin regimen by Pyrazinamide and Rifapentine exposure.

(*) indicates significant by two-tailed logistic regression (P < 0.05). Among participants receiving the rifapentine-moxifloxacin regimen, higher pyrazinamide exposures were associated with increased risk of any grade 3-5 adverse events (OR 1.22 for every 100 µg∙h/mL increase in AUC_0–24h, 95% CI 1.02–1.45) and treatment-related grade 3-5 adverse events (OR 1.27 for every 100 µg∙h/mL increase in AUC_0–24h, 95% CI 1.04–1.55). There was no significant difference between quartiles of rifapentine exposure and any grade 3-5 adverse events, treatment related grade 3-5 adverse events, any serious adverse events, death, or tolerability. Participants without pharmacokinetic sampling were excluded from this figure. Percentages were calculated from the safety population for all safety outcomes except for premature discontinuation of the assigned regimen, which was calculated from the microbiologically eligible population with the exclusion of participants without PK sampling. For each quartile, the percentage of participants with safety outcomes are reported with number of events in parentheses.

Fig. 5. Kaplan Meier Estimates of RIFASHORT Stratified by Treatment Duration and Risk Phenotype.

For external validation, we applied an adjusted risk stratification algorithm to the RIFASHORT modified intention-to-treat population. The easier-to-treat TB and moderately-harder-to-treat TB phenotypes were combined for this validation. The separation in the two risk groups is very clear in both the 6 M HRZE control and the 4 M high dose rifampin regimens.