Abstract
SETTING
In early studies, Xpert® MTB/RIF accurately detected culture-proven pulmonary tuberculosis (TB). Recent reports have, however, found a lower than expected specificity in previously treated TB patients.
OBJECTIVE
To investigate the diagnostic accuracy of Xpert in presumptive pulmonary TB patients in Southwestern Uganda.
DESIGN
We obtained demographic and clinical information and collected three sputum samples from each patient for smear microscopy, Xpert and culture. We estimated Xpert sensitivity and specificity against culture, and stratified the analysis by previous treatment and sputum quality status.
RESULTS
We analyzed results from 860 presumptive TB patients, including 109 (13%) with a previous history of anti-tuberculosis treatment; 205 (24%) were culture-positive. Xpert specificity was lower (91.8%, 95%CI 84.9–96.2) in previously treated than in new TB patients (97.5%, 95%CI 96.1–98.5; P = 0.01). In an adjusted analysis, patients with culture−, Xpert+ results were more likely to have been previously treated for TB (OR 8.3, 95%CI 2.1–32.0; P = 0.002), and to have mucosalivary sputum (OR 4.1, 95%CI 1.1–14.6; P = 0.03), but were less likely to self-report fever (OR 0.23, 95%CI 0.1–0.7; P = 0.008) than patients with concordant positive results.
CONCLUSION
Xpert specificity was lower in previously treated patients with suspected TB. The clinical and programmatic impact of culture−, Xpert+ results requires evaluation in future studies.
