To end the tuberculosis (TB) epidemic, efficient diagnostic tools are needed. In a previous calibration study, a portable ‘point of care’ electronic nose device (Aeonose^TM) proved to be a promising tool in a hospital setting. We evaluated this technology to detect TB in an indigenous population in Paraguay.

A total of 131 participants were enrolled. eNose results were compared with anamnesis, physical examinations, chest radiography and mycobacterial cultures in individuals with signs and symptoms compatible with TB. The eNose analysis was performed in two stages: first, the training with a combination of a previous study population plus 47 participants from the new cohort (total n = 153), and second, the ‘blind prediction’ of 84 participants.

21% of all participants (n = 131) showed symptoms and/or chest radiography abnormalities suspicious of TB. No sputum samples resulted culture positive for Mycobacterium tuberculosis complex. Only one patient had a positive smell print analysis. In the training model, the specificity was 92% (95% confidence interval (CI): 85%‐96%) and the negative predictive value (NPV) was 95%. In the blind prediction model, the specificity and the NPV were 99% (95% CI: 93%‐99%) and 100%, respectively. Although the sensitivity and positive predictive value of the eNose could not be assessed in this cohort due to the small sample size, no active TB cases were found during a one year of follow‐up period.

The eNose showed promising specificity and negative predictive value and might therefore be developed as a rule‐out test for TB in vulnerable populations.