Blood tests predict poor TB outcomes in Brazil

Tuberculosis remains a global health challenge, and clinicians lack reliable tests that use blood rather than sputum to monitor treatment or predict who will have a poor outcome. Corresponding author Simon C. Mendelsohn and his team addressed this gap by studying adults with culture-confirmed, drug-susceptible pulmonary TB enrolled at five Brazilian sites. The researchers collected whole-blood PAXgene samples at three time points: before treatment began (baseline), at month 2 (M2), and at the end of treatment (EoT). Participants were followed for 24 months from treatment initiation to capture clinical or microbiological TB recurrence, and the study defined treatment failure as sputum culture positivity at month 5 or later. To compare outcomes, the team matched cases with unfavourable outcomes—treatment failure, death, or recurrence—approximately 1:3 to participants who achieved recurrence-free cure. By focusing on blood-based signals rather than sputum, the study aimed to find practical, non-sputum biomarkers that could help monitor how patients respond to therapy and identify those at high risk of poor outcomes early in care.

The investigators measured 22 published blood transcriptomic signatures using microfluidic RT-qPCR and benchmarked performance against the WHO Target Product Profile (TPP) criteria. In total, they matched 263 participants with recurrence-free cure to 33 with treatment failure, 24 who died (from TB or unknown cause), and 9 with recurrence. Across the three sampling points, signature scores generally declined from baseline to EoT, reflecting treatment response. Several signatures predicted recurrence when measured at baseline and at M2, with area under the curve (AUC) values ranging from 0.71 to 0.91, while performance waned at EoT (AUC range 0.42–0.89). When judged against the WHO TPP minimum criteria, 2 of 22 signatures met the standard at baseline, 13 of 22 met it at M2, and none did at EoT. Prediction of treatment failure performed poorly at all timepoints (AUC <0.70). Notably, the Thompson5 signature and others measured at baseline predicted death during treatment or follow-up with AUC ≥0.80.

These findings show that blood transcriptomic signatures can both track response to therapy and identify patients at higher risk of recurrence or death, especially when measured before treatment or at month 2. Meeting WHO TPP benchmarks at baseline and M2 suggests some signatures are strong enough to be considered for further testing in prospective trials. The authors conclude that these biomarkers could support trials designed to individualise TB care: for example, testing whether therapy can be safely shortened for early responders identified by blood signatures, or whether people flagged as high risk should receive intensified monitoring and care. While prediction of treatment failure was weak in this study, the results for recurrence and mortality justify larger, biomarker-guided studies to validate which signatures should move into clinical decision-making. The research was supported by the U.S. National Institutes of Health, CRDF Global, and DECIT–SCTIE–MS, Brazil.