Linking Individual Natural History to Population Outcomes in Tuberculosis

Substantial individual heterogeneity exists in the clinical manifestations and duration of active tuberculosis. We sought to link the individual-level characteristics of tuberculosis disease to observed population-level outcomes.

We developed an individual-based, stochastic model of tuberculosis disease in a hypothetical cohort of patients with smear-positive tuberculosis. We conceptualized the disease process as consisting of 2 states-progression and recovery-including transitions between the 2. We then used a Bayesian process to calibrate the model to clinical data from the prechemotherapy era, thus identifying the rates of progression and recovery (and probabilities of transition) consistent with observed population-level clinical outcomes.

Observed outcomes are consistent with slow rates of disease progression (median doubling time: 84 days, 95% uncertainty range 62-104) and a low, but nonzero, probability of transition from disease progression to recovery (median 16% per year, 95% uncertainty range 11%-21%). Other individual-level dynamics were less influential in determining observed outcomes.

This simplified model identifies individual-level dynamics-including a long doubling time and low probability of immune recovery-that recapitulate population-level clinical outcomes of untreated tuberculosis patients. This framework may facilitate better understanding of the population-level impact of interventions acting at the individual host level.