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

Abstract

Objectives: To estimate the sensitivity and specificity of childhood pulmonary tuberculosis (CPTB) diagnostic tests, including Xpert MTB/RIF and alternative samples such as nasopharyngeal aspirate and stool. We used latent class analysis to overcome the lack of reference standards.

Methods: We included 1165 children suspected of having CPTB in a prospective cohort study conducted in the 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 carried out on different types of samples (sputum after 10 years of age, gastric aspirate before 10 years of age, nasopharyngeal aspirate, and stool), tuberculin skin test, 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], and stool 24% [17-32]). Culture sensitivity was slightly higher in gastric (75% [65-83]) than in 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], and nasopharyngeal aspirate 100% [99-100]).

Discussion: The Xpert performed on the nasopharyngeal aspirate shows similar accuracy to the culture. It was also similar to that in gastric aspirate. It highlights an interesting CPTB diagnostic combination.

Keywords: Bayesian analysis; Diagnosis; Paediatrics; Pulmonary tuberculosis; Sensitivity; Specificity.