Study population and setting
Eight dogs underwent seven days of training to discriminate, in real time, between samples from hospitalized, symptomatic patients who were positive for SARS-CoV-2 via RT-PCR and healthy individuals (negative for SARS-CoV-2 via RT-PCR). Saliva samples and tracheobronchial secretion samples were given by all patients and underwent chemical inactivation to protect dogs and handlers from possible exposure to infectious virus. Training and the experiment were conducted with a Detection Dog Training System, which presented one positive sample and six negative samples across seven scent holes in a randomized automated manner, without interference from trainer or observer (double masked). Dogs were exposed to 10,388 sample sets for training and 1,012 sample sets for testing their detection rates.
Summary of Main Findings
On average, the dogs’ detection rate was 94%, with a sensitivity of 82.63% (95% CI:82.02-83.24) and specificity of 96.35% (95% CI: 96.31-96.39). In other words, dogs correctly identified 82.63% of positive samples as positive and 96.35% of negative samples as negative. There was significant within-dog and between-dog variation in each dog’s ability to correctly identify positive samples (i.e., sensitivity) as compared to negative samples (i.e., specificity). No differences in detection ability between saliva and tracheobronchial secretions were observed.
Patient samples used were taken from two different hospitals, masked, and presented in a random order. Furthermore, neither the researchers observing the dogs’ reactions nor the dog trainers knew whether each sample was positive or negative. This decreases the probability that the trainers or researchers could have improved the dogs’ ability to correctly identify each sample.
Samples were obtained from a small number of total individuals (n=14). Positive samples were from hospitalized patients. Thus, this study does not provide any data on the ability for dogs to detect SARS-CoV-2 in patients with mild or no symptoms, which would be the most desirable application for such a screening test.
This was a preliminary study that analyzed trained dogs ability to discriminate between SARS-CoV-2 infected and uninfected individuals in real time. Overall diagnostic specificity was relatively high (94%), while sensitivity was 83% and displayed significant variability.
This review was posted on: 23 October 2020