Using an adapted mathematical model of TB and HIV for South Africa, the study projected adult TB outcomes under uncertainty to 2040. Median model projections indicated a 46% decline in adult TB incidence by 2030 relative to 2015 (95% uncertainty interval [UI]: 17–69%). TB mortality was projected to fall by 54% by 2030 relative to 2015 (95% UI: 21–84%). These projections summarize central tendencies across parameter uncertainty but include wide uncertainty ranges.
The investigators adapted an existing TB–HIV transmission model calibrated to South Africa. They identified 27 model parameters considered both highly uncertain and potentially influential for future TB dynamics. Prior distributions were specified for these parameters to capture plausible uncertainty. Latin Hypercube Sampling was then used to draw 1,000 parameter combinations from those priors, and the model was simulated forward to 2040 for each sampled parameter set. To quantify association between each uncertain parameter and outcomes, partial rank correlation coefficients (PRCCs) were calculated between parameters and average adult TB incidence and mortality over 2025–2040.
The analysis found several parameters with strong correlations to future adult TB incidence (average over 2025–2040). The largest single correlate was the magnitude of increased microbiological testing in symptomatic individuals enabled by near-point-of-care/tongue swab (NPOC/TS) testing, which had a PRCC of -0.67. This indicates that larger increases in microbiological testing among symptomatic patients are associated with substantially lower projected incidence.
A second major correlate was reductions in social contact rates following the COVID-19 pandemic, with PRCC = -0.61. This suggests that sustained lower contact rates contribute meaningfully to incidence declines.
Other notable correlates included the base probability that symptomatic individuals receive sputum testing in the absence of NPOC/TS (PRCC = -0.39) and the efficacy of TB preventive therapy (PRCC = -0.35). These associations indicate that both baseline testing practices and preventive therapy effectiveness are important, though less influential than increases in testing through new diagnostics and contact reductions.
Predictors of future TB mortality mirrored those for incidence. The study reports that the same set of parameters—expanded microbiological testing via NPOC/TS, reductions in social contact post-COVID, baseline sputum testing probability, and preventive therapy efficacy—were among the strongest correlates of lower TB mortality. The similarity of predictors for incidence and mortality implies that interventions reducing transmission and improving diagnosis will also reduce deaths.
The authors highlight that increasing microbiological testing rates among people with TB symptoms is likely to have the largest impact on progress toward WHO targets in South Africa. The analysis specifically identifies the potential population-level benefit of the 2026 WHO guidance recommending near-point-of-care and tongue swab testing in symptomatic patients. Because baseline microbiological testing rates are reported to be very low in many resource-limited settings, the modelling suggests that substantial improvements in testing coverage—facilitated by new diagnostic approaches—could be transformational for incidence and mortality trajectories.
The WHO End TB strategy sets targets of 80% and 90% reductions in TB incidence and mortality, respectively, between 2015 and 2030. Under the model and sampled uncertainty ranges used in this analysis, the projected median declines by 2030 (46% incidence, 54% mortality) fall well short of those End TB targets. The study therefore concludes that while expanding microbiological testing (including via NPOC/TS) and sustaining post-COVID contact reductions would have the largest influence on future declines, attainment of the WHO End TB targets by 2030 is unlikely based on these projections.
The report describes an uncertainty analysis based on an adapted existing model and priors for 27 uncertain parameters. Specific numerical priors, details of model structure beyond being an adapted TB–HIV transmission model, and the full set of 27 parameter definitions were not included in the abstract. The analysis summarizes associations (PRCCs) that reflect correlation under sampled uncertainty rather than causal inference from intervention trials. Readers should consult the full article for detailed methods, parameter definitions, and supplemental material.