Predicting tuberculosis treatment outcome using metabolomics

Abstract

Aim: Predicting a poor treatment outcome would offer significant benefits for patient care and for new drug development. Materials, methods & results: Urine samples from tuberculosis-positive patients with a successful and unsuccessful treatment outcome were collected at baseline and analyzed. The identified metabolites were used in a forward logistic regression model, which achieved a receiver operating characteristic area under the curve of 0.94 (95% CI: 0.84–1) and cross-validated well in a leave-one-out context, with an area under the curve of 0.89 (95% CI: 0.7–1). Two possible predictors were identified, which are associated with a gut microbiota imbalance. Discussion & conclusion: Our findings show the capacity of metabolomics to predict treatment failure at the time of diagnosis, which potentially offers significant benefits for the use in new drug development clinical trials and individualized patient care