Rapid and Accurate Detection of Mycobacterium tuberculosis in Sputum Samples by Cepheid Xpert MTB/RIF Assay—A Clinical Validation Study

BACKGROUND: A crucial impediment to global tuberculosis control is the lack of an accurate, rapid diagnostic test for detection of patients with active TB. A new, rapid diagnostic method, (Cepheid) Xpert MTB/RIF Assay, is an automated sample preparation and real-time PCR instrument, which was shown to have good potential as an alternative to current reference standard sputum microscopy and culture. METHODS: We performed a clinical validation study on diagnostic accuracy of the Xpert MTB/RIF Assay in a TB and HIV endemic setting. Sputum samples from 292 consecutively enrolled adults from Mbeya, Tanzania, with suspected TB were subject to analysis by the Xpert MTB/RIF Assay. The diagnostic performance of Xpert MTB/RIF Assay was compared to standard sputum smear microscopy and culture. Confirmed Mycobacterium tuberculosis in a positive culture was used as a reference standard for TB diagnosis. RESULTS: Xpert MTB/RIF Assay achieved 88.4% (95%CI = 78.4% to 94.9%) sensitivity among patients with a positive culture and 99% (95%CI = 94.7% to 100.0%) specificity in patients who had no TB. HIV status did not affect test performance in 172 HIV-infected patients (58.9% of all participants). Seven additional cases (9.1% of 77) were detected by Xpert MTB/RIF Assay among the group of patients with clinical TB who were culture negative. Within 45 sputum samples which grew non-tuberculous mycobacteria the assay's specificity was 97.8% (95%CI = 88.2% to 99.9%). CONCLUSIONS: The Xpert MTB/RIF Assay is a highly sensitive, specific and rapid method for diagnosing TB which has potential to complement the current reference standard of TB diagnostics and increase its overall sensitivity. Its usefulness in detecting sputum smear and culture negative patients needs further study. Further evaluation in high burden TB and HIV areas under programmatic health care settings to ascertain applicability, cost-effectiveness, robustness and local acceptance are required.