More sensitive tests are needed for the diagnosis of smear-negative and HIV-associated tuberculosis. This study compares the sensitivities and specificities of three molecular tests, namely, the Xpert MTB/RIF test, the Xpert Ultra (Ultra), and RealTime MTB (RT-MTB), in a high HIV prevalence setting. Symptomatic adults were recruited from three outpatient sites, and each provided 4 sputum specimens. The diagnostic performance of Xpert MTB/RIF, Ultra, and RT-MTB was evaluated, with culture as a reference standard. HIV infection occurred in 62% of patients, with a median CD4 count of 220 cells/l. The Ultra test had the highest sensitivity of 89.3% (95% confidence interval [CI], 78.1 to 96) compared to those of the Xpert MTB/RIF at 82.1% (95% CI, 69.6 to 91.1; P 0.12) and RT-MTB at 78.6% (95% CI, 65.6 to 88.4; P 0.68). The specificity was highest with the Xpert MTB/RIF at 100% (95% CI, 98 to 100), followed by RealTime MTB at 96.7% (95% CI, 92.9 to 98.8; P 0.03) and the Ultra at 95.6% (95% CI, 91.5 to 98.1; P 0.08). In patients with smear-negative disease, the Ultra was more sensitive than the Xpert MTB/RIF (64.7% [95% CI, 38.3 to 85.8] versus 41.2% [95% CI, 18.4 to 67.1], respectively; P 0.12), and RT-MTB performed equally to Xpert MTB/RIF. In a comparison of the Ultra and RT-MTB on the same sputum specimen pellets, the Ultra was more sensitive than RT-MTB in the overall cohort (88.9% [95% CI, 77.4 to 95.8] versus 77.8% [95% CI, 64.4 to 88], respectively; P 0.03) and among people with HIV (87.5% [95% CI, 71 to 96.5] versus 68.6% [95% CI, 50 to 83.9], respectively; P 0.03). Although these results did not reach statistical significance, they suggest that the Ultra is more sensitive than the Xpert MTB/RIF and RT-MTB, most prominently in smear-negative disease. This was accompanied by a loss of specificity.
