Performance of Xpert MTB/RIF and acid-fast bacilli smear microscopy for the diagnosis of pulmonary tuberculosis using bronchoalveolar lavage samples in negative or sputum-scarce adults in Colombia: a retrospective diagnostic accuracy study.

Background: Limited literature exists on the diagnostic yield of traditional and molecular tuberculosis (TB) tests using bronchoalveolar lavage (BAL) specimens in high burden South American countries. The aim of the study is to determine the value of acid-fast bacilli (AFB) smear microscopy and Xpert MTB/RIF using BAL samples for diagnosing pulmonary TB (PTB) in negative or sputum-scarce adults.

Methods: A cross-sectional, retrospective diagnostic accuracy study was carried out between January 1, 2015 and December 31, 2023. Demographic, clinical, imaging and microbiologic data were collected from consecutive medical records of initially negative or sputum-scarce adults, who underwent bronchoscopy due to presumptive PTB. Positive cultures were identified as Mycobacterium tuberculosis. Validity criteria was evaluated by sensitivity, specificity, positive (PPV) and negative predictive values (NPV), and Cohen's kappa with their 95% confidence intervals using Wilson's score method by OpenEpi Version 3.01.

Results: Of the 701 medical records screened, 283 adults were included. Cough (89.59%) and fever (64.31%) were the most frequent symptoms. Smear microscopy obtained a sensitivity of 39.22% (95% CI 27.03-52.92), specificity of 98.26% (95% CI 95.61-99.32), PPV of 83.33% (95% CI 64.15-93.32), NPV of 87.94% (95% CI 83.39-91.37) and Cohen's kappa of 0.472 (95% CI 0.3652-0.5788). Xpert MTB/RIF obtained a sensitivity of 91.67% (95% CI 80.45-96.71), specificity of 90.09% (95% CI 85.33-93.43), PPV of 67.69% (95% CI 55.61-77.8), NPV of 97.95% (95% CI 94.85-99.2) and Cohen's kappa of 0.7191 (95% CI 0.5998-0.8384).

Conclusions: Xpert MTB/RIF outperformed smear microscopy using BAL samples for the diagnosis of PTB in negative or sputum-scarce adults from a high burden Colombian setting. Due to rapid results, affordability and accessibility, particularly in resource-limited settings, smear microscopy should still be considered for BAL samples.