Breath can discriminate tuberculosis from other lower respiratory illness in children

Area of Science:

• Pulmonology
• Infectious Diseases
• Biotechnology

Background:

• Pediatric tuberculosis (TB) presents significant diagnostic challenges, with many cases lacking bacterial confirmation.
• Current diagnostic methods for childhood TB are often invasive or lack sensitivity, contributing to delayed treatment.
• Lower respiratory tract infections (LRTRIs) share symptoms with TB, complicating differential diagnosis in children.

Purpose of the Study:

• To identify a volatile compound profile (breathprint) in children with TB.
• To develop a machine learning model for diagnosing pediatric TB using breath analysis.
• To assess the potential of breath analysis as a non-invasive diagnostic tool for childhood TB.

Main Methods:

• A pilot study involving 31 children, including those with confirmed TB and LRTI.
• Gas chromatography-mass spectrometry (GC-MS) to analyze breath samples.
• Machine learning algorithms to classify patients based on identified breathprints.

Main Results:

• A 4-compound breathprint was identified that accurately classified confirmed TB cases (10/10) and LRTI cases (10/10) with 80% sensitivity and 100% specificity.
• The breathprint also identified 9 out of 11 children with clinically suspected but unconfirmed TB, whose symptoms improved with TB treatment.
• Cross-validation demonstrated robust performance of the machine learning model.

Conclusions:

• Breath analysis, using a specific 4-compound breathprint and machine learning, shows potential for diagnosing pediatric TB.
• This non-invasive method could serve as a valuable triage tool or part of a larger diagnostic strategy.
• Further validation studies are required to confirm the clinical utility of breath analysis for pediatric TB diagnosis.