Abstract
BACKGROUND: COVID-19 continues to be a public health concern and the demand for fast and reliable screening tests remains. SARS-CoV-2 infection in humans generates a specific volatile organic compound signature; this 'volatilome' could be used to deploy highly trained canine scent detection teams if they could reliably detect odours from infected individuals. METHODS: Two dogs were trained over 19 weeks to discriminate between the odours produced by breath, sweat, and gargle specimens collected from SARS-CoV-2 infected and uninfected individuals. Third party validation was conducted in a randomized double-blinded controlled manner using fresh odours obtained from different patients within 10 days of their first positive SARS-CoV-2 molecular result. RESULTS: Cumulatively, the dogs completed 299 training sessions on odours from 108 unique participants. Validation was conducted over 2 days with 120 new odours. Twenty-four were odours collected from SARS-CoV-2 positive individuals (8 gargle, 8 sweat, and 8 breath); 21 were from SARS-CoV-2 negative individuals (5 gargle, 8 sweat, and 8 breath) and the remaining 75 were odours that the dogs could have associated with the target odour during training. The dogs were able to identify odours from positive specimens with an overall sensitivity of 100% and a specificity of 87.5%. Considering a community prevalence of 10%, the combined negative predictive value of the dogs was 100% and the positive predictive value was 47.1%. CONCLUSIONS: Multiple dogs can be trained to accurately detect SARS-CoV-2 positive individuals. Future research is required to determine how and when canine scent detection teams should be deployed.