select search filters
roundups & rapid reactions
before the headlines
Fiona fox's blog

expert reaction to preprint looking at the ability of trained bio-detection dogs to identify SARS-CoV-2 infections on clothing samples

A preprint, an unpublished non-peer reviewed study, looks at using trained dogs and organic semi-conducting (OSC) sensors to identify asymptomatic and mild SARS-CoV-2 infections.

This Roundup accompanied an SMC Briefing.


Prof James Wood, Head of Department of Veterinary Medicine, University of Cambridge, said:

This is an interesting and carefully conducted, high quality study that quantifies the utility of ‘sniffer dogs’ in detecting individuals (or their clothes) infected with SARS-COV-2.  The best trained dogs had a good sensitivity of detecting clothing from infected individuals and the authors considered that they could screen 300 people getting off a plane, within 30 minutes.  However, given the values of sensitivity and specificity that the authors provide (94.3% sensitivity and up to 92% specificity), if there were 5 infected people on the plane, while the dogs would likely be able to detect them, they might also falsely classify a further 20-25 individuals as positive and in need of further testing – thus, even at this high prevalence, the likelihood of a positive result being correct, would be no more than around 25%.”


Prof Mick Bailey, Professor of Comparative Immunology, Bristol Veterinary School at the University of Bristol, said:

Current rapid screening tests like the lateral flow tests have been criticised for the potential for false positives and false negatives, and for the contribution to disposable plastic waste, so there is an urgent need for rapid, accurate approaches to screen large numbers of people for infection with SARS-CoV-2, given that many people may have only very mild disease or even be asymptomatic and feel OK to go out in public.

“The use of detection systems for complex organic signatures (like the organic semi-conducting (OSC) sensors, or ‘sniffer’ dogs) has credible scientific support and is well established.

“The methods used here (open training followed by double-blinded testing using facemasks and socks) are appropriate for a laboratory-based validation study, but one concern would be exposing lab workers and dogs to breath and clothes from people known to be infected with SARS-CoV-2.  The study was approved by LSHTM Ethics and Animal Welfare Ethics committees, but I couldn’t find any description of methods to inactivate the virus in the samples. Infection of dogs with SARS-CoV-2 has been reported, both asymptomatic and associated with variable levels of disease as in humans, so using dogs deliberately to sniff breath and clothes from potentially infected people seems a bit of a risk to the dogs, let alone their handlers.

“My other concern would be the relatively small number of people tested.  The results are good in the patient sample they use, but breath signatures vary a lot across the range of gender, age, ethnicity and, importantly, with other diseases.  Their sensitivity and specificity look good with their test set of patients, but I would be more cautious about how well it will roll out to large populations in less controlled environments.  It could be, for example, that the signature that the dogs are detecting might be a general one associated with respiratory viruses like influenza, rather than a specific one associated with SARS-CoV-2.  Although they suggest that many of their negative patients had symptoms of colds or flu, you wouldn’t really know that until you test them on patients with laboratory confirmed seasonal flu, colds etc.  The same would be the case for bronchitis, and even lung cancer.

“So, I think this is a really important start and could lead to a useful, usable system, but there’s a lot more validation needs to be done before we could be confident that the dogs can reliably and specifically detect asymptomatic SARS-CoV-2 infection in people in airports, train stations etc.  And even if they could, I would have a concern about deliberately exposing the dogs to potential sources of infection – you’d need to have some way round that, like vaccinating the sniffer dogs.  Currently, no SARS-CoV-2 vaccine is licensed for use in dogs although there’s no reason to think that the ones approved for humans wouldn’t work, but the low reported incidence of infections in dogs would make a phase 3 trial difficult.”


Prof Lawrence Young, Virologist and Professor of Molecular Oncology, Warwick Medical School, University of Warwick, said:

“This study is not to be sniffed at!  Dogs have long been known to be able to sniff out various diseases ranging from cancer to Parkinson’s disease and influenza – sometimes even before humans know they are sick.  Dogs have smell receptors that are at least 10,000 times more sensitive than humans and can detect down to a few parts of a substance per trillion.

“This pre-print confirms previous studies showing that dogs can be trained to detect SARS-CoV-2 in samples from infected individuals.

“This proof of concept study suggests that trained detection dogs could be used in places like airports, sports stadiums and concert venues to rapidly identify positive individuals who would then take a confirmatory PCR test.  It’s a carefully performed study with a large number of samples incorporating a double-blind study and mathematical modelling.

“The big question is will this approach work in the real world on people rather than samples of socks and shirts?  While the numbers given for sensitivity and specificity of SARS-CoV-2 detection appear to be higher than that of lateral flow devices, there is variation from dog-to-dog particularly in specificity which could result in increased numbers of false positives.  The data suggests that the detection dogs are able to identify asymptomatic individuals with very low levels of virus in their swab samples.  More thorough analysis of the sensitivity of dog screening is warranted.

“It will also be important to study larger groups of infected individuals to determine the efficacy of this approach in rapidly screening crowds of people.  This could be compromised by the density of individuals in crowded spaces and whether well-ventilated external spaces, where odours are rapidly dispersed, compromise the ability of the dogs to detect individuals with low levels of infection.  Differential detection of SARS-CoV-2 from other respiratory virus infections has not been studied and will become important as we head towards the winter months.”



Preprint (not a paper): ‘Using trained dogs and organic semi-conducting sensors to identify asymptomatic and mild SARS-CoV-2 infections’ by Claire Guest et al.

This work is not peer-reviewed.



All our previous output on this subject can be seen at this weblink:



Declared interests

Prof Lawrence Young: “No conflicts.”

None others received.




in this section

filter RoundUps by year

search by tag