Abstract
The microbiological diagnosis of tuberculosis (TB) in children is challenging, as it relies on the collection of relatively invasive specimens by trained health care workers, which is not feasible in many settings. Mycobacterium tuberculosis is detectable from the stools of children using molecular methods, but processing stool specimens is resource intensive. We evaluated a novel, simple, centrifugation-free processing method for stool specimens for use on the Xpert MTB/RIF assay (Xpert), using two different stool masses: 0.6 g and a swab sample. Two hundred eighty children (median age, 15.5 months; 35 [12.5%] HIV infected) with suspected intrathoracic TB were enrolled from two sites in South Africa. Compared to a single Xpert test on respiratory specimens, the sensitivity of Xpert on stools using the 0.6-g and swab samples was 44.4% (95% confidence interval [CI], 13.7 to 78.8%) for both methods, with a specificity of >99%. The combined sensitivities of two stool tests versus the first respiratory Xpert were 70.0% (95% CI, 34.8 to 93.3) and 50.0% (95% CI, 18.7 to 81.3) for the 0.6-g and swab sample, respectively. Retesting stool specimens with nondeterminate Xpert results improved nondeterminate rates from 9.3% to 3.9% and from 8.6% to 4.3% for 0.6-g and swab samples, respectively. Overall, stool Xpert detected 14/94 (14.9%) children who initiated antituberculosis treatment, while respiratory specimens detected 23/94 (24.5%). This stool processing method is well suited for settings with low capacity for respiratory specimen collection. However, the overall sensitivity to detect confirmed and clinical TB was lower than that of respiratory specimens. More sensitive rapid molecular assays are needed to improve the utility of stools for the diagnosis of intrathoracic TB in children from resource-limited settings.