Multicountry assessment of tongue swabs for tuberculosis using a common protocol for Xpert MTB/RIF Ultra testing: a prospective diagnostic accuracy study

BACKGROUND: Despite advancements in tuberculosis diagnostics, many cases remain unconfirmed because of challenges in conventional sputum-based testing. This study aimed to evaluate the diagnostic accuracy of tongue swab sampling as a non-invasive alternative for tuberculosis diagnosis using Xpert MTB/RIF Ultra (Ultra). METHODS: We conducted a large-scale, multicountry, prospective diagnostic accuracy study of Ultra using tongue swabs in people with presumptive pulmonary tuberculosis. Participants were enrolled consecutively at primary health centres and hospitals across eight countries from June 26, 2023, to Feb 15, 2024, and the study was coordinated by three consortia. Eligible participants were individuals aged 12 years or older or 18 years or older, according to the consortium involved, with presumptive pulmonary tuberculosis or at least one risk factor for tuberculosis and a positive tuberculosis screening test at the select consortia. Standardised tongue swab collection and processing protocols were used in all countries. Sensitivity and specificity with 95% CI values were calculated against sputum liquid or solid culture (primary) and sputum Ultra (secondary) reference standards using Wilson's score method. Fisher's exact tests were used for subgroup comparisons, with p values < 0·05 considered statistically significant. FINDINGS: Among the 1844 participants included in the analysis, 389 tested positive and 1455 tested negative for pulmonary tuberculosis based on the primary sputum culture reference standard. 871 (47·2%) participants were female and 973 (52·7%) were male, with a mean age of 43 years (range 12-90). Among the 1844 participants, 399 (21·7%) were enrolled in Viet Nam, 166 (9·0%) in India, 427 (23·2%) in South Africa, 271 (14·7%) in the Philippines, 138 (7·5%) in Nigeria, 102 (5·5%) in Zambia, 175 (9·5%) in Uganda, and 166 (9·0%) in Peru. Tongue swab Ultra testing showed a sensitivity of 65·6% (95% CI 60·6-70·3) and specificity of 98·5% (95% CI 97·7-99·1) against the culture-based reference standard. Sensitivity estimates varied across collection centres and were higher in individuals without HIV than in those living with HIV (68·4% vs 50·0%; absolute difference 18·4 percentage points [95% CI 3·3-33·4]). When sputum Ultra was used as the reference standard, sensitivity was 75·4% (95% CI 69·0-78·8). Tongue swab Ultra showed higher sensitivity than sputum smear microscopy. Invalid or error result rates were variable and high at certain sites (range 0-16%). INTERPRETATION: Tongue swabs are a promising sample type for rapid diagnostic tests for tuberculosis, with moderate sensitivity and high specificity when Ultra was used as the reference standard. However, further research is needed to optimise protocols for Ultra testing and develop assays tailored to tongue swab specimens. Adoption of tongue swab-based molecular testing could expand tuberculosis diagnostics access, especially for populations unable to produce sputum, thus supporting global tuberculosis elimination goals. FUNDING: National Institute of Allergy and Infectious Diseases, United States Agency for International Development.