ABSTRACT
BACKGROUND: Multidrug-resistant tuberculosis is a major threat to the control of tuberculosis and to public health. Whereas most conventional methods of drug susceptibility testing (DST) are precise but time consuming, pyrosequencing is a rapid, high-throughput technique. OBJECTIVE: To conduct a meta-analysis to evaluate the overall accuracy of pyrosequencing for the detection of rifampicin (RMP) resistance. METHODS: We searched PubMed, Web of Science, Elsevier and BIOSIS databases according to a written protocol and explicit study selection criteria. A summary receiver operating characteristic curve (SROC) and Cochrane (Q*) index were calculated to perform this meta-analysis using Meta-Disc software. RESULTS: Twelve studies involving 594 specimens with RMP resistance and 793 RMP-susceptible specimens met the inclusion criteria. Of these, 11 were based on Mycobacterium tuberculosis clinical isolates. The overall sensitivity and specificity were estimated at respectively 0.94 (95%CI 0.92-0.96) and 0.98 (95%CI 0.97-0.99). The area under the SROC curve was 0.99 and the Cochrane (Q*) index was 0.96. For clinical specimens, the overall sensitivity and specificity estimates were respectively 0.89 (range 0.52-1.00) and 0.99 (range 0.95-1.00). CONCLUSIONS: This meta-analysis shows that pyrosequencing is a highly sensitive and specific tool for the detection of RMP resistance in M. tuberculosis. The pyrosequencing assay is conducted in a high-throughput format, with a turnaround time of <2 h, making it substantially faster than conventional DST methods. We propose that pyrosequencing applied directly to clinical specimens instead of M. tuberculosis isolates could be of greater clinical value.