Abstract Background Smear-negative pulmonary tuberculosis remains a major diagnostic challenge due to the limited sensitivity of conventional microscopy. Rapid molecular diagnostics, particularly Xpert MTB/RIF, provide a valuable approach for early detection. Objective To evaluate the diagnostic accuracy of Xpert MTB/RIF assay in smear-negative pulmonary tuberculosis and to assess its agreement and discordance with mycobacterial culture (reference standard) and chest radiography. Methods This prospective cross-sectional study included 50 patients with clinical suspicion of pulmonary tuberculosis and negative sputum smear microscopy. All participants underwent Xpert MTB/RIF testing, mycobacterial culture on Lowenstein–Jensen medium, and chest radiography. Diagnostic performance measures, including sensitivity, specificity, predictive values, and agreement using Cohen’s kappa coefficient, were calculated. Results Xpert MTB/RIF detected Mycobacterium tuberculosis in 30/50 patients (60%), while culture was positive in 38/50 patients (76%). Compared with culture, Xpert MTB/RIF demonstrated a sensitivity of 76.3% (95% CI: 59.8–88.6) and a specificity of 91.7% (95% CI: 61.5–99.8). The positive predictive value and negative predictive value were 96.7% and 55.0%, respectively, with an overall diagnostic accuracy of 80.0%. Agreement between Xpert MTB/RIF and culture was moderate (Cohen’s kappa = 0.50), with an overall agreement of 80%. Chest radiography showed moderate agreement with both Xpert MTB/RIF (kappa = 0.576) and culture (kappa = 0.478). Discordant results included 9 false-negative and 1 false-positive Xpert MTB/RIF case. Conclusion Xpert MTB/RIF represents a rapid adjunctive diagnostic tool for smear-negative pulmonary tuberculosis, characterized by high specificity and moderate sensitivity. While useful as a rule-in test, negative results should be interpreted cautiously, particularly in paucibacillary disease. Integration with microbiological and radiological assessment may provide complementary diagnostic information in clinically suspected cases.
Mady et al. (Fri,) studied this question.