Model predicts why TB sometimes returns after treatment

Relapse after tuberculosis treatment remains a troubling and poorly predicted outcome. Predictors of which patients will return to active disease are limited. Denise E. Kirschner and colleagues set out to explore why some people who finish therapy later develop recurrent tuberculosis, and whether different biological processes lead to relapse. The team focused on two broad mechanisms that could cause disease to come back: a threshold mechanism, in which remaining bacteria exceed a level that triggers renewed disease, and a persistence mechanism, in which bacteria survive in a non-replicating state and later reactivate. To study these ideas in a controlled way, the researchers used a computational model that captures whole-host Mtb infection dynamics. By running simulations of infection, treatment, and post-treatment follow-up, they mimicked both the completion of antibiotic regimens and the point at which a person might be diagnosed as cured. This modeling approach let them compare the effects of different treatments and diagnostic outcomes on the chance of relapse without relying on any single clinical trial or patient cohort.

Using simulations, the study examined how diagnostic tests and treatment regimens interact with underlying bacterial behavior to produce relapse. The model generated rates of erroneous TB-negative diagnoses after treatment completion—so-called false cure—and found these rates depend on the regimen used. Historically standard HRZE was more likely to result in false cure than the contemporary regimens RMZE or BPaL. The simulations also revealed how the two relapse mechanisms relate to measurable factors: threshold-driven or persistence-driven relapse correlated with both pre-treatment bacterial burden and the diagnostic tests used at treatment completion. Importantly, post-cure relapse—relapse that occurs after a person is judged cured—was almost exclusively persistence driven in the model, whereas threshold-driven relapse was most common when there was no "cured" inclusion criterion. Taken together, the results suggest that for patients who have negative bacteriological diagnostic results at treatment completion, subsequent relapse risk may be best addressed by targeting non-replicating Mtb.

These findings have several practical implications. First, they suggest that the choice of regimen matters not only for initial cure but also for the likelihood of false cures and the mechanism by which relapse happens. A regimen like HRZE may leave more patients in a state that looks bacteriologically negative yet still at risk, whereas RMZE or BPaL appeared less likely to produce that false-negative outcome in the simulations. Second, understanding whether relapse is threshold-driven or persistence-driven changes the preventive strategy: threshold-driven relapse points to remaining replicating bacteria above a critical level, while persistence-driven relapse points to surviving non-replicating Mtb that can later reactivate. For patients who test negative at the end of therapy, the study suggests personalizing follow-up and interventions toward the biology of non-replicating bacteria. Finally, using whole-host computational models offers a way to explore these interactions between regimens, diagnostics, and bacterial states to inform clinical decision making and future research priorities.