Mathematical modeling to guide programmatic decisions about the use of new tuberculosis drugs, regimens, and diagnostics
Abstract
Statement of the Problem: Approximately ten million people develop tuberculosis (TB) each year, and this global incidence is declining at a rate of only 2% annually. New diagnostic assays and treatment regimens could improve TB control in high-burden settings, but decisions about whether and how to use these tools often involve multiple competing considerations. For example, the new Xpert MTB/RIF Ultra assay improves sensitivity for TB diagnosis but at the cost of lower specificity. New TB treatment regimens that shorten treatment duration too aggressively could reduce treatment efficacy. A shorter regimen for multidrug-resistant (MDR) TB may increase treatment access, but the non-inferiority of its treatment efficacy remains uncertain. Given such trade-offs, international and country-level decision-makers need to anticipate how new tools are likely to affect various local patient populations and epidemics.
Methods: In this dissertation, dynamic transmission models of TB epidemics are used to (1) project the impact on MDR-TB incidence in Southeast Asia of a new 9-month treatment regimen, and (2) to compare the dependence of novel regimens’ impact on regimen characteristics such as efficacy, duration, tolerability, and barrier to resistance. In addition, a Markov model is used to simulate clinical outcomes (unnecessary treatments, deaths averted) of adopting the Xpert MTB/RIF Ultra cartridge, in several hypothetical settings emblematic of global TB epidemiology and in a specific study community in urban Uganda.
Results: The 9-month MDR-TB regimen could lower MDR-TB incidence by 23% (95% uncertainty range 10–38%), but this projection depended on raising both treatment effectiveness and treatment availability and on limiting detrimental effects of second-line drug resistance. For novel regimens for both rifampin-susceptible and rifampin-resistant TB, treatment efficacy was the most critical of the regimen characteristics modeled and was responsible for approximately half of a regimen’s maximal epidemiologic impact. The clinical impact of switching to the Xpert MTB/RIF Ultra cartridge varied widely between settings but was most favorable in clinical contexts with high prevalence of both TB and HIV, such as in sub-Saharan Africa.
Conclusions: These mathematical models provide useful, context-specific guidance for deciding whether and how to implement new TB treatment regimens or diagnostic tools.