Title : Predictors of treatment failure and development of a risk stratification tool in drug-resistant tuberculosis: A programmatic cohort analysis from Zambia

Background: Treatment outcomes among patients with drug-resistant tuberculosis (DR-TB) remain suboptimal in routine programmatic settings, with persistent gaps driven by mortality and loss to follow-up. Although several clinical and behavioral predictors of poor outcomes have been described, their application in routine care remains limited. There is a need to translate routinely collected program data into practical risk stratification tools to support targeted interventions and improve treatment outcomes.

Description: We conducted a retrospective cohort analysis of routine DR-TB program data. Treatment success was defined as cured or treatment completed, while treatment failure included death and loss to follow-up. Predictors included age, sex, HIV status, diabetes mellitus, smoking, body mass index (BMI), treatment regimen, and prior TB treatment history. Multivariable logistic regression was used to identify independent predictors of treatment failure. Model performance was assessed using receiver operating characteristic (ROC) analysis and calibration plots. A simplified point-based risk score was derived from regression coefficients and used to stratify patients into risk categories.

Lessons learnt: Among 1,503 patients evaluated, 1,029 (68.5%) achieved treatment success, while 474 (31.5%) experienced treatment failure. In adjusted analysis, factors associated with higher odd of treatment failure included increasing age (aOR 1.30 per 10-year increase), HIV infection (aOR 1.25), diabetes mellitus (aOR 1.40) smoking (aOR 1.35) and low BMI (<18.5 kg/m²). Programmatic factors included previous TB treatment history (aOR 1.60) and longer treatment regimens (aOR 1.75). A simplified risk score derived from regression coefficients showed clear separation of risk groups, with progressively increasing probabilities of treatment failure across low-, moderate-, and high-risk categories.

Conclusions: In this large programmatic cohort, DR-TB treatment outcomes were strongly influenced by a combination of demographic, clinical, and programmatic factors. The derived risk score provides a scalable high-impact tool for early identification of high-risk patients and enables targeted, patient centered interventions to reduce mortality and loss to follow-up in routine care.

To be updated shortly..