Advancing Mathematical Models of Mycobacterium tuberculosis Transmission to Support Vaccine Introduction

INTRODUCTION: Tuberculosis causes substantial morbidity and mortality, with 10 million new cases and 1.5 million deaths per year worldwide. We may acquire a new tuberculosis prevention tool in the foreseeable future as the anti-tuberculosis vaccine candidate M72/AS01E is poised to enter phase III trials. If phase III trials demonstrate efficacy, we must ensure that models and model evidence are rapidly and reliably available to support decision making around whether to introduce the new vaccine. Prior mathematical models of tuberculosis vaccination have aimed to inform vaccine development, for example, by investigating the impact of varying vaccine durability and efficacy, or the host-infection status required for vaccine efficacy in low and high burden settings. These studies have identified vaccine characteristics most likely to achieve global tuberculosis control goals, providing core evidence for the World Health Organization (WHO) Preferred Product Characteristics for New Tuberculosis Vaccines. Collectively, this evidence has provided direction for vaccine research and development efforts, including the identification of indications and clinical trial endpoints. However, work to substantiate vaccine introduction, bridging the gap between development and broad-scale adoption, is limited. I address two aims in this thesis, corresponding to two research needs that we should meet to advance models to support vaccine introduction: 1. Estimate the epidemiologic impact, cost-effectiveness and affordability of new tuberculosis vaccines in India and China, incorporating drug-resistance transmission and acquisition. This aim reflects the need to adapt models to include locally important features of, and uncertainty in, tuberculosis epidemiology and health systems. 2. Describe how different assumptions of adapting social contact structures to long-term demographic trends in India—as a country undergoing the demographic transition—might affect vaccine impact estimates. This aim reflects the need to establish whether vaccine impact estimates are robust to structural decisions in model design. METHODS: I constructed an age-, drug-resistance, and treatment-history stratified difference equation-based dynamic transmission model of Mycobacterium tuberculosis, set in India and China, calibrated to epidemiologic data over 2000–2017. To this, I attached a country-specific cost model of programmatic tuberculosis control—including multidrug-resistant or rifampicin-resistant TB (MDR/RR-TB) diagnosis and treatment—and vaccine delivery. The calibrated model was used to estimate the epidemiologic impact and cost-effectiveness of a prevention-of-disease vaccine modelled on M72/AS01E, with 50% efficacy, conferring 10-years of protection over 2027–2050. To assess the affordability of vaccine deployment, I estimated the total cost of untargeted mass vaccination of all adults and adolescents (ages ≥ 10y) compared to 10-year wide age bands in India and China, while valuing the health benefits of vaccination at cost-effectiveness thresholds based on country-specific healthcare opportunity costs. Finally, I developed a second transmission model of tuberculosis to investigate whether adapting age-specific social contact patterns to evolving demography—using multiple update methods, each preserving different properties of social mixing matrices—affected model-based estimates of vaccine impact in an India-like scenario. Results Vaccination was found to substantially reduce future MDR/RR-TB in China and India, reducing incidence rate by 73% (uncertainty interval: 66–76) and 72% (UI: 65–77) in 2050, with a similar impact on drug-susceptible tuberculosis. In both countries, vaccination was found likely to be cost-effective at country-specific willingness to pay thresholds. Untargeted yet cost-effective large scale adult mass vaccination was estimated to require $21 billion (uncertainty interval: 16–27) and $15 billion (UI: 12–29) by 2050 in India and China, respectively. In India and China, targeting 50–59-year-olds and 60–69-year-olds, respectively, was found to avert the most disability-adjusted life-years per vaccine course delivered. Targeted yet cost-effective mass vaccination of these age groups was estimated to require $5 billion (UI: 4–6) and $6 billion (UI: 4–7) in India and China, respectively. Vaccine epidemiologic impact estimates remained robust to different methods of updating age-specific social contact structures to match secular trends in demography. Across a range of methods that spanned no updates to match demography, to methods that preserved both contact reciprocity (balanced total contacts between age groups) and assortativity (inherent preference for contact by age-group with another), the maximum difference in vaccine-mediated incidence rate reduction in 2050 was found to be 7%. CONCLUSIONS: In this thesis, I develop mathematical models that provide evidence to support decision making around tuberculosis vaccine introduction. This thesis makes three unique contributions to the tuberculosis vaccine modelling literature: estimating the impact of new tuberculosis vaccines on MDR/RR-TB, incorporating both direct and transmission effects of a vaccine, estimating the total maximum cost of large scale adult tuberculosis vaccination at country-specific healthcare opportunity cost-based thresholds, and investigating whether structural assumptions around how social contact patterns change with evolving demography affect estimates of vaccine impact. Vaccines are predicted to substantially and cost-effectively reduce the future burden of drug-resistant (and drug-susceptible) tuberculosis in India and China and could be an integral tool in MDR/RR-TB control efforts. The expected total cost of cost-effective untargeted mass adult vaccination for tuberculosis is likely to be substantial at current willingness-to-pay thresholds, but age-targeting may improve affordability. Funding for tuberculosis vaccines will require a careful situating within that for wider tuberculosis control efforts. Vaccine impact estimates may be reasonably robust to different methods of updating social contact patterns to evolving demography. This finding improves confidence in existing estimates of vaccine impact from long time-horizon models and increases the utility of model results in vaccine decision making. Overall, this thesis adds evidence in favour of tuberculosis vaccine introduction, contributing to the initial knowledge base that decision-makers may build on to address further context-specific questions regarding vaccine feasibility and implementation.