Aeonose breath analysis on TB screening South Africa
This year, the Union World Conference on Lung Health was held in Hyderabad, India. At the TBScience pre-conference, an official event entirely devoted to basic and translational tuberculosis (TB) research, study results were presented on the Aeonose exhaled-breath tests in Cape Town.
It shows the Aeonose may be useful as a triage test for tuberculosis in HIV-infected and uninfected persons as it meets the FIND/WHO minimum Target Product Profile (TPP) for a rule-out TB test.
Abstract:
An exhaled-breath screening test for the detection of active pulmonary tuberculosis in patients presenting to primary health care clinics in Cape Town, South Africa.
Introduction: GeneXpert Ultra is unsuitable for community-based mass screening and in up to one-third of patients an adequate sputum sample is unavailable. By contrast, the analysis of exhaled breath obtained by the Aeonose-TB device has the potential to provide an easy, onsite, rapid, and non-invasive diagnosis of tuberculosis. Thus, it may serve as a useful community-based triage tool (a major unmet need in the TB field).
Methods: We enrolled 1,143 participants (31% HIV- infected) with suspected TB (243 microbiologically-con- firmed TB, 130 probable TB, and 770 non-TB) from primary care clinics in Cape Town, South Africa. Sputum GeneXpert MTB/RIF and/ or culture served as the reference standard. Volatile organic compounds in exhaled-breath were detected using an electronic nose containing 3 metal-oxide sensors (AeonoseTM, Zutphen, The Netherlands). Data were analysed by machine learning using artificial neural networks (ANN) in a ‘leave-10%-out’ cross-validation training set (n=756; 189 TB and 567 non-TB) and the findings ratified in a test set (n=257).
Results: In the training set the sensitivity, specificity, PPV, NPV of Aeonose-TB (95% CI) was 81% (74-86) and 60% (55-64), 40% (35-45), and 90% (87-93), respectively. However, in HIV uninfected patients, using a ANN- derived rule-in threshold, performance characteristics were 59% (50-69), 91% (87-93), 68% (58-77), and 87% (83-90), respectively. In the same group, using an ANN- derived rule-out threshold, the performance characteristics were 90% (83-95), 59% (54-65), 43% (36-49) and 95% (91-97), respectively. Results in the validation set, and in HIV-infected persons, showed comparable performance characteristics.
Conclusion: Aeonose-TB may be useful as a triage test for tuberculosis in HIV-infected and uninfected persons as it meets the FIND/WHO minimum Target Product Profile (TPP) for a rule-out TB test. However, it may also have utility to rule in TB in sputum scarce or smear-negative persons. Further studies are now required to clarify these findings.
