Fig. 1: A priori characterization of the three routes to disease in the placebo arm of a Phase 2b clinical trial conducted on IGRA+ participants.

A Compartmental model used to describe TB dynamics in the placebo arm of a trial conducted on IGRA+ individuals without past or present evidence of active TB. According to this model, trial participants can be divided in fast (F) vs. slow (L) progressors, each of which show different risks of progression to disease (D) per unit time that can be further divided into three routes to disease. B Compartmental model used to describe TB transmission at country-level. Individual states are: susceptible S, infected (either fast \({L}_{F}\) or slow \({L}_{S}\) progressors, analogous to F and L reservoirs in A); active disease D, disease under treatment T (pulmonary smear-positive TB (p+), pulmonary smear-negative TB (p−) and non-pulmonary TB), disease recovery R (Natural (_N), successful after treatment (_S), treatment default (_D)) and treatment failure (F). The model is used first to obtain estimates of parameters to inform clinical trial simulations, and later to evaluate vaccine impact (for further details see Supplementary text S1 and ref. ²³). C The country-level model sketched in B is calibrated to reproduce TB incidence trends (also mortality, see Supplementary Fig. S1) in each country. Error bars represent the reported uncertainty of incidence estimates in the WHO tuberculosis database, shaded areas capture the 95% CI in all the trajectories forecasted by the model (N = 500). D From the calibrated simulations conducted country-wise, we retrieve estimates for the relative fraction of fast progressors over the total population of IGRA+ individuals without a past of active TB \({L}_{F}\)/(\({L}_{F}\) + \({L}_{S}\)) that is expected in each country. E From the same simulations, we obtain estimates of the force of infection per country and age group (fraction of susceptible individuals infected per year and age group). D, E bars represent the median, boxes capture the inter-quartile range, and error bars represent the 95% CI from a set of N = 500 simulations. F With those items at hand, along with literature-based estimates for the epidemiological parameters \(r\), \({r}_{L}\), \(q\), and \(p\), the placebo arm of the M72/ASO1_E study can be simulated in-silico, from which we can estimate the expected fraction of incident TB cases associated to each of the routes to disease. G Weighting the contributions estimated in F, according to the age and country-wise distributions of participants in the M72/AS01_E trial⁹, we obtain an overall estimate of the relative contribution of each route to disease to the total incidence observed in the global placebo arm of the trial. Source data are provided as a Source Data file.