Approximately 10 million people globally developed tuberculosis disease in 2018. Of these, nearly 500,000 were sick with isolates resistant to at least isoniazid and/or rifampin, referred to as multidrug-resistant TB (MDR-TB). Conventional treatment for MDR-TB is long, difficult, and toxic. Regulatory approval of two new TB drugs, bedaquiline and delamanid, for the treatment of MDR-TB has offered the potential for more effective and less toxic regimens. This study will use observational data from the endTB Project. The observational study of the endTB Project is the largest study cohort of MDR-TB patients receiving treatment with bedaquiline and delamanid to date. We will implement robust causal inference methods to generate evidence on the effectiveness and safety of regimens containing bedaquiline and delamanid for the treatment of MDR-TB. We will explore the following Specific Aims: (1) identify the optimal adverse events management strategy that maximizes safety and effectiveness for patients receiving linezolid; (2) examine whether the effect of delamanid on successful treatment outcome when added to an MDR-TB regimen varies according to the number of drugs in the regimen that are likely to be effective; (3) investigate the magnitude of bias due to time-dependent confounding affected by previous exposure in analyses of MDR-TB treatment. The results of these Specific Aims will inform clinical practice, treatment guidance, and future approaches to analyzing MDR-TB treatment cohort data. Causal inference theory and frameworks will be applied; Robins' generalized methods (?g-methods?), such as inverse probability of treatment weighting in marginal structural models and the parametric g-formula, will be used in analyses.
Conventional treatment for multidrug-resistant tuberculosis (MDR-TB) is long in duration, often results in debilitating side effects and cures only half of patients. In the absence of randomized controlled trials to study every clinical question of interest in MDR-TB treatment, analyses of observational data using appropriate statistical methods is required in order to inform policy and timely decision-making. The proposed project will apply novel causal inference methods to answer two important clinical questions in the comparative effectiveness and safety of strategies for the treatment and management of MDR-TB and explore the magnitude of bias due to time-dependent confounding in a large, observational cohort of patients treated with regimens containing bedaquiline and delamanid.
