The authors developed a deterministic compartmental model with 1,296 compartments to estimate population-level effectiveness and economics of VPM1002 for tuberculosis prevention in India. The compartments arise from stratification by four age groups, two HIV statuses, three body-mass-index strata, three socioeconomic strata, and 18 states, explicitly representing demographic and epidemic heterogeneity. The model separates pulmonary TB (PTB) and extrapulmonary TB (EPTB) outcomes to translate disaggregated trial hazard ratios into population impact, and it compares static and dynamically baselined no-vaccine scenarios.
Vaccine-efficacy posteriors for PTB and EPTB were obtained by Bayesian evidence synthesis of the PreVenTB Phase III trial's per-protocol and modified intention-to-treat (mITT) analyses. The synthesis used a power-prior-adjusted conjugate normal-normal update with α = 0.082. Separate posteriors for PTB and EPTB were produced and carried forward into population simulations. Uncertainty was propagated via Monte Carlo simulation with 1,000 iterations per scenario.
A dynamic no-vaccine baseline was constructed by fitting time-varying logistic curves for case-detection-rate CDR(t) and treatment-success-rate TSR(t) to WHO and India NTEP data from 2015 to 2024. The fitted time-varying baseline projected PTB and EPTB incidence to 2050. Validation statistics for the incidence projection versus observed data were reported as R2 = 0.896 and RMSE = 4.90 per 100,000, indicating the calibration captured historical trends while allowing for ongoing NTEP-driven incidence decline.
The analysis evaluated multiple scenarios that varied the duration of vaccine protection (examples reported include 3-year and 10-year protection) and whether the baseline was static or dynamic. For each scenario, Monte Carlo propagation of the vaccine-efficacy posteriors yielded distributions of cases averted, deaths averted, and disability-adjusted life years (DALYs) averted per 1,000,000 vaccinated. Economic analyses were performed from both societal and health-system perspectives with costs and benefits discounted at 3% annually.
Posterior vaccine effectiveness estimates were substantially larger for EPTB than for PTB: EPTB effectiveness 40.0% (95% CI −5.6 to 69.1%), PTB effectiveness 12.8% (95% CI −19.0 to 37.9%), and all-TB 16.2% (95% CI −11.7 to 38.6%). Across scenarios, EPTB protection drove the majority of population health gains. In the 10-year/dynamic baseline scenario, EPTB accounted for 74% of deaths averted (381 of 513) and 73% of DALYs averted (5,347 of 7,308). In every scenario the number of EPTB cases averted exceeded the combined total of PTB and concurrent disease cases averted.
Reported mean cases averted per 1,000,000 vaccinated ranged from 2,047 (3-year protection, dynamic baseline) to 3,394 (10-year protection, static baseline). Mean DALYs averted per 1,000,000 vaccinated ranged from 4,792 to 8,125 across the examined scenarios.
Economic evaluation used two perspectives. The societal perspective included a value-of-statistical-life-inclusive benefit-cost ratio (BCR); the health-system perspective produced cost-effectiveness acceptability curves and incremental cost-effectiveness ratios (ICERs). Across scenarios the societal BCR ranged from 3.4 (3-year protection, dynamic baseline) to 8.2 (10-year protection, static baseline) and exceeded break-even in every scenario reported. Median gross ICERs ranged from US$647 (10-year static) to US$1,568 (3-year dynamic) per DALY averted. All reported median ICERs were below India’s 3× gross domestic product per capita threshold (~US$8,084), as used by the authors for context.
The model demonstrates that VPM1002’s EPTB protection is proportionally larger and a more statistically reliable signal than its PTB effect, and that EPTB protection drives most averted cases, deaths, and DALYs in population projections. The authors emphasize that policy assessments anchored only to composite or predominantly pulmonary endpoints risk systematically underestimating VPM1002’s population value. By incorporating a dynamic baseline that reflects anticipated NTEP-driven declines in incidence, the analysis provides a more conservative and context-sensitive estimate of vaccine impact than static baselines alone.
The study was partially funded by Serum Life Science Europe GmbH; the funder had no role in model design, data analysis, interpretation, or writing, according to the authors. Several authors are affiliated with or are founders of Aseesa Inc.; these competing interests are declared. No individual-level participant data were used in the modelling study. Epidemiological inputs came from the WHO Global Tuberculosis Report and India NTEP; vaccine efficacy inputs derived from the PreVenTB Phase III trial (Singh et al., BMJ 2026). The model's full parameter values, calibration targets, epidemiological data sources, and health-economic inputs are reported in the supplementary appendix. The underlying differential-equation implementation code is proprietary to Aseesa Inc. and is not publicly released, but the calibrated model is accessible as a runnable interface at Hugging Face upon request to the corresponding author.