New model predicts how TB drug pairs work together

Treating pulmonary tuberculosis (TB) effectively requires combinations of drugs, but most early clinical trials focus on single drugs. Phase IIa pulmonary tuberculosis (TB) trials typically assess the early bactericidal activity (EBA) of monotherapy over 14 days, and relatively few studies have looked closely at how drugs behave when given together. That matters because doses that work well by themselves may not produce the same effects when drugs are combined. To address this gap, a team led by researchers including corresponding author Rada Savic extended a translational pharmacokinetic-pharmacodynamic (PK-PD) modeling platform they had previously developed for monotherapy. Their aim was to see whether this platform could predict the EBA of two-drug combinations, using available preclinical and clinical data. By adapting their model to handle interactions between drugs, they sought to create a tool that would forecast human responses to combinations and help guide which drug pairings and doses should move forward in development.

The study characterized interactions among four TB drugs—bedaquiline, pretomanid, linezolid, and pyrazinamide—using two different modeling approaches. One was an empirical method called the SUPER method, and the other was a mechanistic approach named the General Pharmacodynamic Interaction model. Each interaction model was independently linked to the translational platform, and the combined systems were validated against mouse data and Phase IIa clinical results from the NC-001 study. Both approaches found consistent interaction patterns. Notably, antagonistic interactions emerged when bedaquiline was combined with either pretomanid or linezolid. In contrast, pyrazinamide emerged as the most effective companion for both bedaquiline and pretomanid. When used to predict 14-day clinical sputum colony-forming unit counts for multiple two-drug combinations, the platform’s forecasts were broadly accurate: most clinical observations fell within the calculated 95% prediction intervals. The platform performed well for both short- and longer-term outcomes, and its success did not depend on which interaction model was used.

These findings suggest the translational pharmacokinetic-pharmacodynamic (PK-PD) platform can play a practical role in developing combination treatments for TB. Because the platform can use preclinical data and different interaction models to predict human responses, it offers a way to assess likely outcomes of two-drug regimens before committing to large or lengthy clinical trials. The consistent identification of antagonism for certain pairings—specifically bedaquiline with pretomanid or linezolid—and the clear benefit of pyrazinamide as a companion drug provide actionable information for researchers designing regimens. By predicting 14-day EBA and aligning well with clinical data from the NC-001 study, the approach supports using modeling to inform dose selection and to pick effective companion drugs. Overall, the validated platform can help prioritize combinations and dosing choices across drug development stages, increasing confidence in which regimens should advance to further testing.