Six studies (584 patients, range 36 to 156) were included in the review. The results of methodological quality assessment were not reported, although the authors noted that none of the included studies reported blinding of the investigators.
No studies on the BACTEC 9000MB system or the Difco ESP tests met the inclusion criteria.
MB/BacT test (two studies, 202 patients): For the detection of rifampicin-resistant tuberculosis, the pooled estimate of sensitivity was 100% (95% CI 81 to 100) and the pooled estimate of specificity was 99% (95% CI 97 to 100). The pooled positive likelihood ratio was 65.12 (95% CI 16.42 to 258.19) and the pooled negative likelihood ratio was 0.06 (95% CI 0.01 to 0.4).
BACTEC MGIT 960 test (five studies, 464 patients): The pooled estimate of sensitivity was 100% (95% CI 96 to 100) and the pooled estimate of specificity was 96% (95% CI 93 to 98). The pooled positive likelihood ratio was 29.84 (95% CI 5.24 to 169.87) and the pooled negative likelihood ratio was 0.03 (95% CI 0.01 to 0.11). The summary receiver operating characteristic curve gave an area under the curve of 0.9943 and a Q* of 0.9704.
Manual MGIT test (one study, 36 patients): The reported sensitivity was 100% (95% CI 84 to 100) and the specificity was 96% (95% CI 87.3 to 98). The positive likelihood ratio was 51.55 (95% CI 3.30 to 805.55) and the negative likelihood ratio was 0.05 (95% CI 0.00 to 0.70).