Accuracy of Xpert and alternative sampling methods to diagnose childhood pulmonary tuberculosis, a prospective cohort study.

Abstract

Objectives: The aim of the study was to estimate sensitivity and specificity of childhood pulmonary tuberculosis (CPTB) diagnostic tests, including Xpert and alternative samples such as nasopharyngeal aspirate and stool. We used latent class analysis to overcome the lack of gold standard.

Methods: We included 1165 children suspected of having CPTB in a prospective cohort study conducted in Ivory Coast, Madagascar, and Cameroon. We used Bayesian latent class analysis to evaluate the performance of multiple diagnostic tests for CPTB: smear microscopy, mycobacterial culture, and Xpert MTB/RIF carried out on different types of samples (sputum after 10 years of age, gastric aspirate before 10 of age, nasopharyngeal aspirate, stool), tuberculin skin test (TST), and chest X-ray.

Results: Median age was 3.5 years (95% credible interval [1.3-8.2]). Smear microscopy was highly specific in all types of samples but lacked sensitivity (sputum 31% (95% credible interval [18-46]), gastric aspirate 36% [27-45], nasopharyngeal aspirate 24% [17-32], stool 24% [17-32]). Culture sensitivity was slightly higher in gastric (75% [65-83]) than nasopharyngeal aspirate (64% [54-72]). Xpert sensitivity was similar in gastric (69% [59-78]) and nasopharyngeal aspirate (66% [57-74]) but lower in sputum (58% [43-74]) and stools (53% [44-62]). Xpert was highly specific in all respiratory samples (sputum 96% [96-100], gastric aspirate 100% [99-100], nasopharyngeal aspirate 100% [99-100]).

Conclusions: Xpert performed on nasopharyngeal aspirate shows similar accuracy than culture. It was also similar to that in gastric aspirate. It highlights it's an interesting CPTB diagnostic combination.