Abstract
BACKGROUND: Obtaining respiratory samples to diagnose tuberculosis in young children is challenging. Oral swabs are alternative noninvasive specimens for microbiology. METHODS: We conducted an interim prospective diagnostic accuracy evaluation of Xpert MTB/RIF Ultra (Ultra) on oral swabs for pulmonary tuberculosis in children aged <5 years in Uganda. Most children had 2 consecutive swabs collected in a single cryovial (double swabs). Reference tests consisted of Ultra and culture on 2 nasopharyngeal aspirates and 1 gastric aspirate and Ultra on 1 stool. Children were classified as having confirmed tuberculosis, unconfirmed tuberculosis, or unlikely tuberculosis per the National Institutes of Health. Diagnostic accuracy was determined against a microbiological reference standard and a composite reference standard. RESULTS: From August 2021 to February 2024, 444 children were enrolled, of whom 399 had complete classifications: 33 had confirmed tuberculosis, 269 had unconfirmed tuberculosis, 70 had unlikely tuberculosis, and 27 were unclassifiable. The median age was 16 months and 17% had HIV. Most children (398/399) had oral swabs collected, all with conclusive Ultra results. The sensitivity of double swabs was 6.9% with a microbiological reference standard (95% CI, 1.9%-22.0%) and 1.8% with a composite reference standard (95% CI, .8%-4.1%). Specificity was at least 99%. Swabs detected tuberculosis in 4 children with negative reference test results, of whom 3 had unconfirmed tuberculosis. CONCLUSIONS: The low sensitivity of Ultra on double swabs precludes its role as a principal diagnostic approach in young children. However, detection of tuberculosis in children who were not otherwise microbiologically diagnosed suggests the utility of oral swabs as add-on samples to increase yield.