Following endorsement by the World Health Organisation, the Xpert MTB/RIF assay has been widely incorporated into algorithms for the diagnosis of adult tuberculosis (TB). However, data on its performance in children remain scarce. This prospective, multi-centre study evaluated the performance of Xpert MTB/RIF to diagnose pulmonary tuberculosis in children.

Children older than eight weeks and younger than 16 years with suspected pulmonary tuberculosis were enrolled at three TB endemic settings in Tanzania and Uganda, and assigned to five well-defined case definition categories: culture-confirmed TB, highly probable TB, probable TB, not TB, or indeterminate. The diagnostic accuracy of Xpert MTB/RIF was assessed using culture-confirmed TB cases as reference standard.

In total, 451 children were enrolled. 37 (8%) had culture-confirmed TB, 48 (11%) highly probably TB and 62 probable TB (13%). The Xpert MTB/RIF assay had a sensitivity of 68% (95% CI, 50%–82%) and specificity of 100% (95% CI, 97%–100%); detecting 1.7 times more culture-confirmed cases than smear microscopy with a similar time to detection. Xpert MTB/RIF was positive in 2% (1/48) of highly probable and in 3% (2/62) of probable TB cases.

Xpert MTB/RIF provided timely results with moderate sensitivity and excellent specificity compared to culture. Low yields in children with highly probable and probable TB remain problematic.