ObjectiveTo assess the validity of Xpert Tuberculosis Fingerstick score for monitoring treatment response and analyze factors influencing its performance. Methods122 adults with pulmonary tuberculosis were recruited and stratified into three cohorts: Diabetic-drug-susceptible-TB (DM-TB), Non-diabetic-drug-susceptible-TB (NDM-TB) and Non-diabetic Multidrug-resistant TB (MDR-TB). Fingerstick blood specimens were tested at treatment initiation (M0) and the end of the first (M1), second (M2), and sixth month (M6) to generate a TB-score. ResultsThe TB-score in all participants yielded an AUC of 0.707 (95% CI: 0.579–0.834) at M2 when its performance was evaluated against sputum culture conversion. In all non-diabetes patients, the AUC reached 0.88 (95% CI: 0.756–1.000) with an optimal cut-off value of 1.95 at which sensitivity was 90.0% (95% CI: 59.6–98.2%) and specificity was 81.3% (95% CI: 70.0–88.9%). The mean TB score was higher in patients with low bacterial loads (n = 31) than those with high bacterial loads (n = 91) at M0, M1, M2, and M6, and was higher in non-cavitary patients (n = 71) than those with cavitary lesions (n = 51) at M0, M1, and M2. ConclusionXpert TB-score shows promising predictive value for culture conversion in non-diabetic TB patients. Sputum bacterial load and lung cavitation status have an influence on the value of TB score.