Detection of metabolic profile in urine for diagnosing pediatric tuberculosis

Abstract

Children’s paucibacillary tuberculosis (TB) and their difficulty to expectorate leads to low diagnostic sensitivity. Most TB in children are diagnosed by clinical scoring systems limited by the TB clinical presentation. Pediatric TB diagnosis research should be focused on new biomarkers from non-invasive and non-sputum-based samples. By metabolomics we can obtain a “fingerprint” of the metabolites presents in a biological sample, allowing the study of sets of metabolites affected by host-pathogen interactions and the identification of diagnostic markers. This study aims to identify urine metabolite biomarkers with pediatric TB diagnostic potential. The urine spectra from High-resolution 1H Nuclear Magnetic Resonance (NMR) spectroscopy were obtained in a cross-sectional study of 73 children (0-14 years) screened for suspected TB in a pediatric hospital of Haiti. Enrolled children had a positive Tuberculin Skin Test and were classified as follows: 23 TB, 27 probable TB, 13 unlikely TB, and 10 latent TB infection (LTBI). Among all NMR tested samples, an algorithm was developed with 33 spectra from core groups (23 TB, 10 LTBI). We identified eight metabolites with significant changes between groups and when comparing TB and LTBI, we achieved a specificity and sensitivity of 95.49% and 89.90%, respectively. Moreover, when including probable TB and unlikely TB, we classified 92.65% and 53.85%, respectively as TB. We suggest a combination of metabolites that discriminate TB from LTBI, improving the diagnostic accuracy. This urine-based metabolomic algorithm may improve the pediatric TB diagnosis identifying those children with unconfirmed TB diagnostic and therefore, reducing the undiagnosed pediatric TB.