Abstract
SETTING: Cape Town, South Africa.OBJECTIVES: We investigated the potential of breath analysis by gas chromatography-mass spectrometry (GC-MS) to discriminate between samples collected prospectively from patients with suspected tuberculosis (TB).DESIGN: Samples were obtained in a TB-endemic setting in South Africa, where 28% of culture-proven TB patients had Ziehl-Neelsen (ZN) negative sputum smear. A training set of breath samples from 50 sputum culture-proven TB patients and 50 culture-negative non-TB patients was analysed using GC-MS. We used support vector machine analysis for classification of the patient samples into TB and non-TB.RESULTS: A classification model with seven compounds had a sensitivity of 72%, a specificity of 86% and an accuracy of 79% compared with culture. The classification model was validated with breath samples from a different set of 21 TB and 50 non-TB patients from the same area, giving a sensitivity of 62%, a specificity of 84% and an accuracy of 77%.CONCLUSION: This study shows that GC-MS breath analysis is able to differentiate between TB and non-TB breath samples even among patients with a negative ZN sputum smear but a positive culture for Mycobacterium tuberculosis. We conclude that breath analysis by GC-MS merits further research.
Original language | English |
---|---|
Pages (from-to) | 777-782 |
Number of pages | 6 |
Journal | International Journal of Tuberculosis and Lung Disease |
Volume | 16 |
Issue number | 6 |
DOIs | |
Publication status | Published - Jun 2012 |
Keywords
- gas chromatography-mass spectrometry
- training set
- validation set
- volatile organic compounds
- statistical analysis
- VOLATILE ORGANIC-COMPOUNDS
- MYCOBACTERIUM-TUBERCULOSIS
- PULMONARY TUBERCULOSIS
- SPUTUM SAMPLES
- BIOMARKERS
- INFECTION
- AREA