RESUMEN
Childhood tuberculosis (TB) presents significant diagnostic challenges associated with paucibacillary disease and requires a more sensitive test. We evaluated the diagnostic accuracy of Xpert MTB/RIF Ultra (Ultra) compared to other microbiological tests using respiratory samples from Ugandan children in the SHINE trial. SHINE is a randomized trial evaluating shorter treatment in 1,204 children with minimal TB disease in Africa and India. Among 352 samples and one cervical lymph node fine needle aspirate, one sample was randomly selected per patient and tested with the Xpert MTB/RIF assay (Xpert) and with Lowenstein-Jensen medium (LJ) and liquid mycobacterial growth indicator tube (MGIT) cultures. We selected only uncontaminated stored sample pellets for Ultra testing. We estimated the sensitivity of Xpert and Ultra against culture and a composite microbiological reference standard (any positive result). Of 398 children, 353 (89%) had culture, Xpert, and Ultra results. The median age was 2.8 years (interquartile range [IQR], 1.3 to 5.3); 8.5% (30/353) were HIV infected, and 54.4% (192/353) were male. Of the 353, 31 (9%) were positive by LJ and/or MGIT culture, 36 (10%) by Ultra, and 16 (5%) by Xpert. Sensitivities (95% confidence intervals [CI]) were 58% (39 to 65% [18/31]) for Ultra and 45% (27 to 64% [14/31]) for Xpert against any culture-positive result, with false positives of <1% and 5.5% for Xpert and Ultra. Against a composite microbiological reference, sensitivities were 72% (58 to 84% [36/50]) for Ultra and 32% (20 to 47% [16/50]) for Xpert. However, there were 17 samples that were positive only with Ultra (majority trace). Among children screened for minimal TB in Uganda, Ultra has higher sensitivity than Xpert. This represents an important advance for a condition which has posed a diagnostic challenge for decades.