Predicting resistance to fluoroquinolones among patients with rifampicin-resistant tuberculosis: A cross-country validation study
by Tianfang Shao, Mariana R. Neves, Molly Franke, Carole Mitnick, Jennifer Furin, Ted Cohen, Reza Yaesoubi Background Fluoroquinolones (FQs) are a cornerstone of most all-oral, shorter regimens endorsed by the World Health Organization for the treatment of rifampicin-resistant or multidrug-resistant tuberculosis (RR/MDR-TB). Knowledge of resistance to FQs can help guide regimen selection at the…
Fluoroquinolones (FQs) are crucial in treating rifampicin-resistant or multidrug-resistant tuberculosis (RR/MDR-TB) as recommended by the World Health Organization. Predicting FQ resistance in patients with RR-TB can aid treatment decisions, especially when rapid testing is unavailable. A study analyzed data from 5,175 RR-TB patients with FQ drug susceptibility testing (DST) results from eight countries between 2012 and 2024.
Among these patients, 1,772 (34.2%) had FQ-resistant TB. Researchers developed prediction models using logistic regression, neural networks, and XGBoost to predict FQ resistance based on patient characteristics. They evaluated three strategies: pooled models, within-country models, and cross-country models. The pooled models showed moderate optimism-corrected discrimination, while within-country models performed slightly better.
Cross-country external validation indicated that performance loss from using a model trained on external data could be minimal to 0.1 AUPROC or AUROC.
Certain predictors, including case definition and treatment-history-related variables, proved consistently informative. However, demographic, comorbidity, social-risk, education, and employment variables showed varying contributions across countries and algorithms. The study's limitation lies in the stable incidence of RR-TB and FQ resistance in the analyzed dataset. Consequently, the results might not apply to scenarios with significant changes in MDR-TB dynamics.
Overall, predicting FQ resistance using demographic and clinical characteristics can help identify FQ resistance in RR-TB patients. However, the performance and predictor patterns of these models vary across countries. Researchers caution against assuming that models developed for one or several countries will generalize to other settings without rigorous external validation.
These findings emphasize the limitations of globally trained prediction models for RR/MDR-TB and emphasize the need for locally informed models to support clinical decision-making.
Written by urgent.news from PLOS Medicine's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
