Biomarker for Individualized Therapy Durations in Patients With TB

European researchers developed a biomarker to guide individualized treatment durations for patients with both drug-susceptible TB and MDR-TB.

In patients with tuberculosis (TB), a host 22-gene RNA-based model has been identified that may be predictive of individual treatment durations with antimicrobial therapy, potentially leading to a substantially shorter treatment durations for many patients. This was among the results of an analysis recently published in the European Respiratory Journal.

Researchers conducted a prospective study of adult patients with TB who were enrolled into 5 independent cohorts in Germany and Romania. Researchers collected clinical and microbiologic data, along with whole blood for RNA transcriptome analysis, at predefined times throughout the course of the patient’s treatment. Tuberculosis Network European Trials group (TBNET) criteria (ie, 6-month culture statistics and 1-year follow-up) were used to establish treatment outcomes. Investigators then developed a whole-blood RNA therapy end model using data collected on end-of-treatment timepoints.

Data from 2 German identification cohorts (drug-susceptible [DS]-TB cohort, 50 patients; MDR-TB cohort, 30 patients) were used to create a 22-gene RNA model that defined cross-associated end-of-therapy time points. The model yielded individual probabilities for cure-associated end-of-therapy time points at any given moment over the duration of therapy. This model was subsequently applied to 3 independent validation cohorts: 2 in Germany (DS-TB cohort, 28 patients; MDR-TB cohort, 32 patients) and 1 in Romania with 52 patients with MDR-TB.  

The RNA model identified end-of-therapy time points with high accuracy for patients in the DS-TB German validation cohort (area under the curve [AUC], 0.94; 95% CI, 0.90-0.98). After 15 months of therapy, the proportion of patients in the 3 MDR-TB cohorts reaching a cure according to the model predictions were 84.6% in the German identification cohort, 40% in the German validation cohort, and 88.5% in the Romanian validation cohort. Results further suggested that in the 3 MDR-TB cohorts, a cure might be attained with shorter treatment durations among patients (German identification cohort: mean reduction of 218.0 days; P <.001; German validation cohort: mean reduction of 211.0 days; P <.001; Romanian validation cohort: mean reduction of 161.0 days; P =.001).

According to researchers, the RNA model was also superior to other published signatures for the accurate prediction of end-of-therapy timepoints in TB. The investigators concluded that the translation of the current RNA model into clinical practice requires additional clinical assessment in large studies, along with the development of an implementation platform to support practicality in settings with limited resources.

Disclosure: One of the study authors has declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of authors’ disclosures. 


Heyckendorf J, Marwitz S, Reimann M, et al. Prediction of anti-tuberculosis treatment duration based on a 22-gene transcriptomic model. Eur Respir J. 2021;58(3):2003492. doi:10.1183/13993003.03492-2020