RESUMEN
BACKGROUND: Xpert MTB/Rif, a molecular test to detect tuberculosis (TB), has been proven to have high sensitivity and specificity when compared with liquid culture in clinical settings. However, little is known about its performance in community TB screening. METHODS: In Vietnam, a national TB prevalence survey was conducted in 2017. Survey participants who screened positive by chest X-ray, cough symptoms and/or recent history of tuberculosis were requested to provide at least two sputum samples that were tested for Mycobacterium tuberculosis by Xpert MTB/Rif G4 (Xpert) and BACTEC MGIT960 culture (MGIT). RESULTS: There were 4,649 eligible participants provided both samples for testing. Among them, 236 (5.1%) participants tested positive for TB by Xpert, 244 (5.3%) tested positive by MGIT and 317 tested positive by at least one test; 163 (51.4%) had discordant test results. Of the positive Xpert, 162 (68.6%) showed a low or very low bacterial load. In multivariate logistic regression comparing discordant with Xpert-MGIT concordant positive results, discordant Xpert-positive results occurred more often among participants who had low sputum bacterial load, male sex, a history of TB treatment, or night sweats. The associated factors were male sex, abnormal chest X-ray and having night sweats when the logistic model was against those with both Xpert and MGIT negative. CONCLUSIONS: We found high rates of discordance in the performance of Xpert and MGIT for community-based TB case finding. In situations where the majority of TB cases are expected to have a low bacterial load, multiple diagnostic tests and/or multiple samples are required to reach sufficient sensitivity.