| Studies n | Sensitivity# (95% CI) % | I2 % | p-value | Specificity# (95% CI) % | I2 % | p-value | |
| Analysis A¶ | 7 | 60 (38–79) | 97.0 | <0.0001 | 93 (88–96) | 84.5 | <0.0001 |
| HIV positive | 4 | 51 (32–69) | 90.7 | <0.0001 | 94 (79–98) | 77.6 | 0.004 |
| HIV negative | 4 | 14 (7–24) | 53.3 | 0.09 | 97 (90–99) | 63.1 | 0.04 |
| Analysis B+ | 7 | 47 (26–69) | 97.3 | <0.0001 | 93 (83–97) | 94.8 | <0.0001 |
| HIV positive | 4 | 56 (40–71) | 24.1 | <0.0001 | 95 (77–99) | 92.0 | <0.0001 |
| HIV negative§ | 3 | 18 (10–29) | 0.0 | 0.88 | 90 (85–93) | 69.3 | 0.04 |
| Analysis Cƒ | 5 | 34 (14–62) | 98.4 | <0.0001 | 94 (87–98) | 85.5 | <0.0001 |
| HIV positive§ | 2 | 49 (43–54) | 54.4 | 0.14 | 91 (85–96) | 69.5 | 0.07 |
| HIV negative§ | 2 | 16 (8–28) | 0.0 | 0.4 | 94 (86–98) | 70.8 | 0.06 |
I2: inconsistency. #: due to imprecision of estimates, sensitivity and specificity are rounded to the nearest whole number; ¶: analysis excludes clinical cases (patients that were microbiologically negative, but had strong clinical/radiological suspicion for tuberculosis); +: analysis groups clinical cases with all other patients that were microbiologically negative to be reference negative; §: pooled estimates calculated with univariate random effects models (confidence intervals are expected to be narrower); ƒ: analysis groups clinical cases with patients that were microbiologically positive as reference positive.