RESUMO
Facing the COVID-19 pandemic, testing individuals in order to promptly isolate positive people is one of the key actions. One approach to rapid testing might be to consider the olfactory capacities of trained detection dogs in order to develop a non-invasive, rapid and cheap mass detection approach, through the Volatile Organic Compounds (VOCs) signature of SARS-CoV-2 infection. The goal of this study was to determine the individual values of sensitivity and specificity of trained dogs when performing olfactory detection of COVID-19 on axillary sweat samples. Seven dogs were used to sniff a total of 218 samples (62 COVID-19 positive and 156 COVID-19 negative samples), based on a randomised and double-blinded protocol carried out on olfaction cone line-ups. To ensure a wide olfactory range as close as possible to operational conditions, the samples were retrieved from 13 different sites. Sensitivities varied from 87% to 94% for 6 dogs, and were above 90% for 3 of them. Only one dog, whose sensitivity was 60%, was not selected to subsequently enter the operational stage. Specificities varied from 78% to 92%, with 6 dogs over 85% and 4 dogs over 90%. Based on the calculated sensitivity and specificity for each dog, positive and negative predictive values (PPV and NPV) were calculated according to several prevalence rates of SARS-CoV-2 scenarii, and were compared to PPV and NPV of an "almost perfect" diagnostic tool. For 6 out of the 7 dogs, and for a prevalence rate of SARS-CoV-2 lower or equal than 40%, the NPV of the dogs were virtually the same as the one of the "almost perfect" tool. Along with other studies on olfactory detection of COVID-19 by dogs these positive and encouraging results suggest that olfactory dogs may play an important part in mass COVID-19 pre-testing situations.
