Background Tuberculosis is a leading cause of infectious disease-related death and is one of the top 10 causes of death worldwide. The World Health Organization (WHO) recommends the use of specific rapid molecular tests, including Xpert MTB/RIF or Xpert Ultra, as initial diagnostic tests for the detection of tuberculosis and rifampicin resistance in people with signs and symptoms of tuberculosis. However, the WHO estimates that nearly one-third of all active tuberculosis cases go undiagnosed and unreported. We were interested in whether a single test, Xpert MTB/RIF or Xpert Ultra, could be useful as a screening test to close this diagnostic gap and improve tuberculosis case detection. Objectives To estimate the accuracy of Xpert MTB/RIF and Xpert Ultra for screening for pulmonary tuberculosis in adults, irrespective of signs or symptoms of pulmonary tuberculosis in high-risk groups and in the general population. Screening "irrespective of signs or symptoms" refers to screening of people who have not been assessed for the presence of tuberculosis symptoms (e.g. cough). To estimate the accuracy of Xpert MTB/RIF and Xpert Ultra for detecting rifampicin resistance in adults screened for tuberculosis, irrespective of signs and symptoms of pulmonary tuberculosis in high-risk groups and in the general population. Search methods We searched 12 databases including the Cochrane Infectious Diseases Group Specialized Register, MEDLINE and Embase, on 19 March 2020 without language restrictions. We also reviewed reference lists of included articles and related Cochrane Reviews, and contacted researchers in the field to identify additional studies. Selection criteria Cross-sectional and cohort studies in which adults (15 years and older) in high-risk groups (e.g. people living with HIV, household contacts of people with tuberculosis) or in the general population were screened for pulmonary tuberculosis using Xpert MTB/RIF or Xpert Ultra. For tuberculosis detection, the reference standard was culture. For rifampicin resistance detection, the reference standards were culture-based drug susceptibility testing and line probe assays. Data collection and analysis Two review authors independently extracted data using a standardized form and assessed risk of bias and applicability using QUADAS-2. We used a bivariate random-effects model to estimate pooled sensitivity and specificity with 95% credible intervals (CrIs) separately for tuberculosis detection and rifampicin resistance detection. We estimated all models using a Bayesian approach. For tuberculosis detection, we first estimated screening accuracy in distinct high-risk groups, including people living with HIV, household contacts, people residing in prisons, and miners, and then in several high-risk groups combined. Main results We included a total of 21 studies: 18 studies (13,114 participants) evaluated Xpert MTB/RIF as a screening test for pulmonary tuberculosis and one study (571 participants) evaluated both Xpert MTB/RIF and Xpert Ultra. Three studies (159 participants) evaluated Xpert MTB/RIF for rifampicin resistance. Fifteen studies (75%) were conducted in high tuberculosis burden and 16 (80%) in high TB/HIV-burden countries. We judged most studies to have low risk of bias in all four QUADAS-2 domains and low concern for applicability. Xpert MTB/RIF and Xpert Ultra as screening tests for pulmonary tuberculosis In people living with HIV (12 studies), Xpert MTB/RIF pooled sensitivity and specificity (95% CrI) were 61.8% (53.6 to 69.9) (602 participants; moderate-certainty evidence) and 98.8% (98.0 to 99.4) (4173 participants; high-certainty evidence). Of 1000 people where 50 have tuberculosis on culture, 40 would be Xpert MTB/RIF-positive; of these, 9 (22%) would not have tuberculosis (false-positives); and 960 would be Xpert MTB/RIF-negative; of these, 19 (2%) would have tuberculosis (false-negatives). In people living with HIV (1 study), Xpert Ultra sensitivity and specificity (95% CI) were 69% (57 to 80) (68 participants; very low-certainty evidence) and 98% (97 to 99) (503 participants; moderate-certainty evidence). Of 1000 people where 50 have tuberculosis on culture, 53 would be Xpert Ultra-positive; of these, 19 (36%) would not have tuberculosis (false-positives); and 947 would be Xpert Ultra-negative; of these, 16 (2%) would have tuberculosis (false-negatives). In non-hospitalized people in high-risk groups (5 studies), Xpert MTB/RIF pooled sensitivity and specificity were 69.4% (47.7 to 86.2) (337 participants, low-certainty evidence) and 98.8% (97.2 to 99.5) (8619 participants, moderate-certainty evidence). Of 1000 people where 10 have tuberculosis on culture, 19 would be Xpert MTB/RIF-positive; of these, 12 (63%) would not have tuberculosis (false-positives); and 981 would be Xpert MTB/RIF-negative; of these, 3 (0%) would have tuberculosis (false-negatives). We did not identify any studies using Xpert MTB/RIF or Xpert Ultra for screening in the general population. Xpert MTB/RIF as a screening test for rifampicin resistance Xpert MTB/RIF sensitivity was 81% and 100% (2 studies, 20 participants; very low-certainty evidence), and specificity was 94% to 100%, (3 studies, 139 participants; moderate-certainty evidence). Authors' conclusions Of the high-risks groups evaluated, Xpert MTB/RIF applied as a screening test was accurate for tuberculosis in high tuberculosis burden settings. Sensitivity and specificity were similar in people living with HIV and non-hospitalized people in high-risk groups. In people living with HIV, Xpert Ultra sensitivity was slightly higher than that of Xpert MTB/RIF and specificity similar. As there was only one study of Xpert Ultra in this analysis, results should be interpreted with caution. There were no studies that evaluated the tests in people with diabetes mellitus and other groups considered at high-risk for tuberculosis, or in the general population.
