Rapid diagnosis of pulmonary tuberculosis on induced sputum in children in a high burden setting for HIV and malnutrition.

Microbiological diagnosis of childhood pulmonary tuberculosis can be challenging. We investigated the accuracy of Xpert MTB/RIF Ultra on the induced sputum, focused on high-risk groups (infants, malnourished children, and those living with HIV), where data is limited. Children (under 15 years) investigated for pulmonary tuberculosis at Dora Nginza Hospital, South Africa, were prospectively enrolled between January 2018 and January 2024, and followed up for 12 months. Induced sputa were tested with liquid culture and Xpert MTB/RIF Ultra (Ultra). Children were categorized with confirmed, unconfirmed, or unlikely tuberculosis, according to NIH consensus definitions. Ultra diagnostic accuracy was calculated against culture, with analysis by high-risk subgroups. Of 636 children (median age, 27.4 (IQR 11.8-61.6) months), 173 (27.2%) were classified with confirmed TB, 384 (60.4%) with unconfirmed TB, and 79 (12.4%) with unlikely TB. There were 223 (37.8%) children underweight-for-age, 168 (26.5%) living with HIV, and 163 (25.6%) infants. Of 557 children with TB, 145 (26.0%) were confirmed on induced sputa; with 137/557 (24.6%) confirmed on Ultra (46.0%; 63/137, of which were "trace positive"), and 88/557 (15.8%), on culture. Of those confirmed, 57/145 (39.3%) were Ultra positive, culture negative - contributing 10.2% to the overall yield. Ultra performed better in children over 5 years and with weight-for-height Z-scores < -3. Overall, Ultra sensitivity was 90.9% (95%CI 82.9-96.0) and specificity, 89.6% (95%CI 86.7-92.0), with high accuracy across high-risk groups. Mortality amongst children treated for PTB was 4.9% (27/557) with aRR highest for stunted children (7.3; 95% CI, 2.2-24.2), followed by children with HIV (3.5; 95% CI, 1.6-7.5), and infants (3.3; 95% CI, 1.6-6.7). Ultra on induced sputum was feasible, with high yield and diagnostic accuracy for high-risk children.