Bio Detection dogs identify COVID-19 with up to 94% accuracy

Modelling suggests using these super-sniffers could be effective as a new rapid, non-invasive screening tool at ports of entry, as COVID-19 has a distinct smell

This is the most complete study of its kind to date, combining data collected during the first phase of the dog trial, odour analysis and modelling. 

The study, which is not yet peer-reviewed, was led by the London School of Hygiene & Tropical Medicine (LSHTM) collaboratively with the charity Medical Detection Dogs and Durham University. It was part-funded by the UK’s Department of Health & Social Care, and used over 3,500 odour samples donated by the public and NHS staff. 

It showed that COVID-19 infection has a distinct smell, which specially trained dogs can rapidly, non-invasively detect with up to 94.3% sensitivity and up to 92% specificity. The dogs were able to detect odour from individuals who were asymptomatic, as well as those with two different strains, and with both high and low viral loads.  

The researchers acknowledge the results were achieved in a trial setting with the dogs trained in a controlled environment, however, they believe they could be replicated in real–world settings. 

Accompanying mathematical modelling highlights the potential for dogs to be used at ports-of-entry or other sites, with preparatory work suggesting that two dogs could screen 300 plane passengers in around 30 minutes as part of a ‘Rapid Screen and Test’ strategy. Only individuals who are identified by the dogs would require a PCR test.  

Use of the Bio Detection dogs plus a confirmatory PCR test are estimated to detect more than twice as many cases and prevent more onward transmission than isolating symptomatic individuals only. 

Professor James Logan, Head of the Department of Disease Control at LSHTM, who led the project, said: “The results of this study far exceeded my expectations. While the mass rollout of vaccines in the UK is a great success, it will take time to achieve the coverage levels needed for our lives to return to near normal. With the threat of new variants entering the country, the need for testing means we face potential continued disruption for some time to come. 

“That’s where these amazing dogs could play a role. Further research is needed to see if the dogs can replicate these results in a real-world setting, but these findings are hugely encouraging. The advantage of using this method is being able to detect COVID-19 with incredible speed and good accuracy among large groups of people, even in asymptomatic cases. This really could help us get back to doing the things we love sooner, safely and with less disruption, such as helping to reduce queuing times at border points or sporting events. 

“This study and the exciting potential of using dogs to tackle COVID-19 would not have been possible without the volunteers from the public and NHS, and I want to thank everyone who donated valuable odour samples.” 

Dr Claire Guest, Chief Scientific Officer at Medical Detection Dogs, said: “These fantastic results are further evidence that dogs are one of the most reliable biosensors for detecting the odour of human disease. Our robust study shows the huge potential for dogs to help in the fight against COVID-19. 

“Knowing that we can harness the amazing power of a dog’s nose to detect COVID-19 quickly and non-invasively gives us hope for a return to a more normal way of life through safer travel and access to public places, so that we can again socialise with family and friends.” 

Professor Steve Lindsay, from the Department of Biosciences at Durham University, added: “This is a very exciting result showing that there is a distinct smell associated with COVID-19 and, more importantly, that trained dogs can detect this with a high degree of accuracy. Dogs could be a great way to screen a large number of people quickly and preventing COVID-19 from being re-introduced into the UK. Trained dogs could potentially act as a fast screening tool for travellers with those identified as infective confirmed with a lab test. This could make testing faster and save money.”  

This study is the first to comprehensively assess whether trained dogs can distinguish between the odour of people infected with SARS-CoV-2 and those who are uninfected, in a randomised double-blind trial, with a sufficiently high number of dogs and odour samples.  

The dogs were trained by the team at Medical Detection Dogs to identify COVID-19 using body odour samples which were sent to the research team by members of the public and NHS staff, consisting of masks, socks and t-shirts. The ARCTEC team at LSHTM collected and processed a total of 3,758 samples, and chose 325 positive and 675 negative samples for testing. 

The dogs were trained over a number of weeks by introducing them to the odour samples from individuals that had tested positive for COVID-19, as well as control samples from people who had tested negative. Samples were presented to the dogs on a stand system and the dogs were rewarded for correctly indicating a positive sample, or for correctly ignoring a negative sample.

Six dogs were then taken forward to the important ‘double-blind’ trial where the dog, technician and dog trainer were not aware of which samples were positive or negative. This removes any risk of inadvertent bias or behavioural cue that the dog could pick up on to indicate the correct response. These dogs were tested using 200 positive samples and 200 negative samples.

The highest performing dogs in the trial detected the odour of the virus in the samples with up to 94.3% sensitivity (meaning a low risk of false negative results) and up to 92% specificity (meaning a low risk of false positive results). This is a greater accuracy than recommended by the World Health Organization for COVID-19 diagnostics, with the dogs consistently outperforming lateral flow tests across sensitivities between 80-90%, which have an overall sensitivity of between 58-77%.

While PCR is the gold standard of tests, with 97.2% sensitivity and 90% specificity, the researchers emphasise the dogs have the advantage of being incredibly rapid, and non-invasive, with the potential to quickly and passively screen individuals in public places without inconvenience.

To predict the effectiveness of these dogs in real-world setting, the researchers from LSHTM’s Centre for Mathematical Modelling of Infectious Diseases and the Royal Veterinary College modelled various scenarios, which suggest the dogs could be highly effective in detecting cases and averting transmission. With the dogs rapidly identifying positive individuals, the Rapid Screen and Test strategy would mean only those indicated as positive by the dogs would require a confirmatory PCR test, resulting in far fewer individuals needing testing.

Using the dogs to identify positive individuals who then take a confirmatory PCR test – 91% of cases would be detected, resulting in 2.2 times more aversion of transmission compared to isolation of symptomatic individuals alone. In comparison, only 9% more cases were detected and 0.13 times more transmission aversion by mass screening with PCR alone. This, combined with potential for rapid screening by the dogs, could result in fast case detection which would in turn enable earlier self-isolation to prevent onward transmission. 

Previous studies show the volatile organic compounds (VOCs) released in body odour change during respiratory infections. Chemical analysis of the VOCs associated with COVID-19 infection showed a clear distinction between infected and uninfected individuals, suggesting that it has a strong and distinctive smell. These findings could be used to develop a pseudo-odour for standardised dog training.

Following these encouraging results, the researchers hope to proceed to the next phase of the trial, which will see the dogs detecting COVID-19 on real people in real world settings. The researchers also believe these dogs could serve as visual deterrents to reduce the number of passengers travelling with falsified COVID-19 negative certificates, which has anecdotally been observed with explosive and drug detection dogs at public events. 

The trial has also provided the researchers with valuable knowledge and techniques which could be applied in response to any future disease outbreaks, with the potential to quickly deploy Bio Detection dogs in a potential public health emergency. The use of dogs could be built into countries’ future pandemic planning strategies.

Professor Logan added: “COVID-19 detection dogs could play an important role in tackling the pandemic, not only here in the UK, but in other countries around the world, and even beyond COVID-19. Now we have proven proof of principle for COVID, for other disease outbreaks in the future we think dogs could be deployed quickly to screen people and help stop the outbreak when it first begins.”

The authors acknowledge limitations of the study, including that the dogs were detecting the odour of COVID-19 in samples, while their real value would be screening people in large public places.