Prediction Models for Fluoroquinolone Resistance in Tuberculosis Falter

A PLOS Medicine study warns that fluoroquinolone-resistance prediction models for drug-resistant tuberculosis may not generalize well across countries. An international team evaluated whether statistical models trained in one epidemiological setting retain accuracy when deployed elsewhere, using TB Portals data from 5,175 patients with rifampicin-resistant tuberculosis. The study’s central finding is a cautious “no”: models lose meaningful precision in new locations, and the patient characteristics that signal resistance shift by place. The stakes are high because fluoroquinolones such as levofloxacin and moxifloxacin underpin shorter, all-oral regimens endorsed by the World Health Organization when rifampicin fails to respond.