Background

Xpert MTB/RIF Ultra (Xpert Ultra) has improved the sensitivity to detect pulmonary tuberculosis (TB) in adults. However, there have been limited prospective evaluations of its diagnostic accuracy in children.

Methods

We enrolled children undergoing assessment for pulmonary TB in Kampala, Uganda, over a 12-month period. Children received a complete TB evaluation and were classified as Confirmed, Unconfirmed, or Unlikely TB. We calculated the sensitivity and specificity of Xpert Ultra among children with Confirmed vs Unlikely TB. We also determined the diagnostic accuracy with clinical, microbiological, and extended microbiological reference standards (MRSs).

Results

Of the 213 children included, 23 (10.8%) had Confirmed TB, 88 (41.3%) had Unconfirmed TB, and 102 (47.9%) had Unlikely TB. The median age was 3.9 years, 13% were HIV-positive, and 61.5% were underweight. Xpert Ultra sensitivity was 69.6% (95% confidence interval [CI]: 47.1-86.8) among children with Confirmed TB and decreased to 23.4% (95% CI: 15.9-32.4) with the clinical reference standard. Specificity was 100% (95% CI: 96.4-100) among children with Unlikely TB and decreased to 94.7% (95% CI: 90.5-97.4) with a MRS. Sensitivity was 52.9% (95% CI: 35.1-70.2) and specificity 95.5% (95% CI: 91.4-98.1) with the extended MRS. Of the 26 positive Xpert Ultra results, 6 (23.1%) were "Trace-positive," with most (5/6) occurring in children with Unconfirmed TB.

Conclusions

Xpert Ultra is a useful tool for diagnosing pulmonary TB in children, but there remains a need for more sensitive tests to detect culture-negative TB.