Eighteen studies were included in the review. There were 10,224 specimens with suspected TB: 2,983 had microbiologically confirmed TB and 335 had clinically/radiologically diagnosed TB; 9,166 specimens were included in the meta-analysis (range 25 to 4,742 specimens). The median QUADAS score was 12 (range 9 to 14).
When Xpert MTB/RIF was used to diagnose pulmonary TB (15 studies) the overall pooled estimates were: sensitivity 90% (95% CI 89% to 91%; Ι²=89%), specificity 98% (95% CI 98% to 99%; Ι²=76%) and DOR 328.3 (95% CI 154.3 to 698.3). Results seemed similar when derived from the SROC analysis. When used to diagnose RIF pulmonary TB (seven studies) the overall pooled estimates were: sensitivity 94% (95% CI 92% to 96%; Ι²=91%), specificity 97% (95% CI 96% to 98%; Ι²=79%) and DOR 177.8 (95% CI 41.7 to 757.7). Sensitivity was lower (approximately 80%) and specificity similar when pooled estimates were derived from the SROC analysis.
Results were provided for HIV coinfection and diagnosing extrapulmonary TB. Full results of the meta-regression analysis were reported.