Abstract

Background Oral samples show promise for tuberculosis (TB) diagnosis, but data from different sample types and sputum-scarce individuals remain limited. Methods We evaluated Xpert MTB/RIF Ultra (Ultra) in symptomatic clinic attendees (cohort A, n = 891) and people initiating antiretroviral therapy without symptom screening (cohort B, n = 258). In cohort A, we collected oral washes (OWs) and, separately, tongue swabs (flocked or foam with heat). In cohort B, we collected OWs, 3 flocked tongue swabs (1 heated, 2 pooled), and, separately, buccal swabs and periodontal brushes. Sputum induction was offered, and different culture methods were applied to a subset of cohort B tongue swabs. Results In cohort A, Ultra sensitivity was 80% (95% confidence interval [CI], 56%–94%) for OWs, 59% (95% CI, 53%–65%) for flocked swabs, and 65% (95% CI, 58%–72%) for foam swabs, with high specificity. Foam swabs detected more people with Ultra sputum semi-quantitation categories of low or less than flocked swabs (53% [95% CI, 41%–64%] vs 37% [95% CI, 29%–46%]). In cohort B, OWs and single heated swabs had sensitivities of 71% (95% CI, 42%–92%) and 64% (95% CI, 35%–87%), respectively. Pooled tongue swabs, buccal swabs, and brushes had lower sensitivity. MGIT960 showed the highest sensitivity (64% [95% CI, 35%–87%]) among culture methods. Oral sampling identified TB in sputum-scarce people: 25% (7/28) positive by flocked or foam swabs (cohort A); 18% (10/56) were OW and 23% (13/56) single swab positive (cohort B). In cohort B, this could double Ultra positivity if induction were unavailable. Conclusions Ultra on OWs or foam swabs offers higher sensitivity than other oral methods and effectively detects TB in sputum-scarce individuals.