New model sharpens view of tuberculosis spread

Tuberculosis continues to spread in many parts of the world, and stopping its spread depends on understanding how the disease moves between people. To tackle this challenge, Anne N. Shapiro and colleagues developed a new statistical approach that uses genetic similarity between tuberculosis bacteria as a stand‑in, or proxy, for transmission. Rather than taking every genetically linked pair at face value, the team improved an existing iterative model that adjusts genetically linked tuberculosis case data so it better reflects who actually infected whom. The goal was to translate those adjusted links into numbers that describe how different characteristics — like age, sex, or incarceration history — are associated with the chance that two cases are part of the same transmission chain. By doing so, researchers can produce adjusted odds ratios (ORs) that account for the messy reality that the same person can appear in multiple potential transmission pairs, which can otherwise distort standard analyses.

The method combines the iterative adjustment of genetically linked cases with bootstrapped logistic regression to calculate adjusted ORs and confidence intervals that account for correlated data. Specifically, the team incorporated a bootstrapping procedure into logistic regression so the uncertainty estimates reflect the fact that individuals often contribute to more than one pairwise link. They tested the approach using simulation studies to see how well it recovered known relationships, and then applied it to a cohort from Lima, Peru. The iterative algorithm produced estimates similar to conventional logistic regression, but with larger confidence intervals; the wider intervals reflect the appropriate adjustment for correlation among pairs. In the Lima data, transmission pairs including at least one person older than 34 years showed decreased odds of transmission, while pairs that included at least one incarcerated person or at least one male had increased adjusted odds of transmission.

By producing adjusted ORs that explicitly account for the correlation inherent in pairwise genetic relatedness data, this approach offers a more cautious and realistic measure of which factors are linked to tuberculosis spread. The authors conclude these adjusted ORs serve as an accurate proxy for the association between covariates and transmission, improving our understanding of the social and demographic factors that matter. Because the method was validated with simulations and demonstrated on real cohort data, it provides a statistical framework researchers and public health teams can use when they rely on genetic links to infer transmission patterns. That improved clarity can help researchers prioritize which risk factors to study further and can shape how surveillance data are interpreted when trying to stop chains of tuberculosis transmission.