Introduction Sputum smear-negative pulmonary tuberculosis (TB) is increasingly prevalent with bronchoalveolar lavage (BAL) frequently used for diagnostic sampling. Direct molecular testing has reported higher sensitivities compared to smear microscopy. This study aims to assess the predictive accuracy and clinical impact of Xpert MTB/RIF; a PCR-based cartridge assay used to identify M.tb in BAL fluid samples.

Methods A retrospective evaluation of adult patients (n = 293) with suspected pulmonary TB who underwent BAL in a tertiary centre in London between January 2011 and December 2014 were collected. MTB/RIF, smear microscopy, and liquid culture were performed on all sets of BAL fluid. The impact of MTB/RIF on time to TB diagnosis and anti-TB treatment initiation were recorded as markers of clinical impact.

Results 57/293 (19.5%) patients had BAL culture-positive TB for which a significantly higher proportion had positive MTB/RIF results compared to smear microscopy (77.2%, 95% CI 63.8%–86.8% vs. 38.6%, 95% CI 26.3%–52.4%; p < 0.001). The specificity of MTB/RIF was 95.7% (92.1%–97.8%) with a negative predictive value (NPV) of 94.6% (90.7%–97.0%). 22/57 (38.6%) culture-positive patients had negative smear microscopy results but positive MTB/RIF results.

90/293 (30.7%) patients were clinically-diagnosed and treated for pulmonary TB. In this subgroup, MTB/RIF again outperformed smear microscopy in terms of sensitivity (54.2%; 95% CI 43.7%–64.3% vs. 27.1%, 95% CI 18.8%–37.3%; p < 0.001). The specificity of MTB/RIF was 81.6% (80.0%–86.2%).

The use of MTB/RIF provided a significant advantage in time to TB diagnosis in culture-positive patients as compared to smear microscopy (1 days; IQR 0–2 days versus 10 days; IQR 0–15 days; p < 0.05). In a specific cohort of smear-negative culture-confirmed TB patients, MTB/RIF reduced the time to TB diagnosis from an average of 14 days (95% CI 12–16 days) to 1 day (95% CI 0–2 days; p < 0.05).

There was no statistical difference in time to anti-TB treatment initiation between those who were either smear microscopy positive and/or MTB/RIF positive in culture-positive patients (1 day; IQR 0–3 days versus 1 day; IQR 0–2 days; p = 0.164).

Conclusion MTB/RIF used in BAL samples had a higher and more rapid diagnostic accuracy compared to smear microscopy and could replace routine smear microscopy in pulmonary TB diagnosis.