Skip to main content

People with COVID-19, even mild or asymptomatic cases, have a distinct odour that these dogs can pick up. Trials show the dogs to be 94 per cent accurate, better than lateral flow tests.

Medical Detection Dogs

They’ve been used to sniff out drugs, explosives and other contraband, and now a team of scientists in Britain has trained dogs to detect COVID-19 in humans with up to 94-per-cent accuracy.

The researchers, from the London School of Hygiene & Tropical Medicine, Durham University and a British charity called Medical Detection Dogs, trained six dogs to pick out tiny bits of clothing from people who had been infected with the virus, in some cases months earlier.

The dogs were between 82-per-cent to 94-per-cent accurate in identifying the COVID-19 samples. They also achieved up to 92-per-cent specificity, which measures the ability to tell if someone doesn’t have the disease.

Coronavirus tracker: How many COVID-19 cases are there in Canada and worldwide? The latest maps and charts

“The results are extremely exciting,” said James Logan, the project leader and head of the Department of Disease Control at the London School of Hygiene & Tropical Medicine. “The dogs could detect COVID with incredible speed and accuracy, even if a person was asymptomatic.”

The study has yet to be peer-reviewed and further trials are under way, but Dr. Logan said the early results demonstrated that teams of dogs could be used at airports, sports stadiums or other large gatherings as an initial check for people who might be infected, even mildly.

He said the dogs had no trouble identifying cases of the COVID-19 variant first detected in Britain and they weren’t thrown off by samples from people who had other respiratory illnesses such as a common cold. The overall results were so impressive, he added, that the dogs performed better than lateral flow tests, a common rapid test for the virus, and they were almost as accurate as polymerase chain reaction, or PCR, tests which are considered the gold-standard for verifying COVID-19 cases.

Asher sniffs a sock sample from someone who may have had COVID-19 at the Medical Detection Dogs centre in Milton Keynes, England, in March 2020.

Marlow sniffs a sock sample.Neil Pollock /Handout

“Dogs are significantly quicker than any of the other tests,” Dr. Logan said, adding that two dogs can screen 300 people in 30 minutes. “What we are suggesting is that there could be what we call a rapid screen and test strategy. Dogs would first give the initial screen and those who were indicated as positive would then receive a confirmatory PCR test.”

Scientists have long known that body odour changes when people are sick and dogs have been trained to spot a variety of illnesses including malaria, Parkinson’s disease and bladder cancer. Recent studies have shown that COVID-19 also causes a strong and distinctive smell that dogs can pick up.

Claire Guest, the chief scientific officer at Medical Detection Dogs, said most canines have 300 million scent receptors in their noses, which gives them a remarkable sense of smell. Humans, by contrast, have roughly five million. The best breeds for detection work – Labradors, golden retrievers and cocker spaniels – have up to 350 million receptors and an inbred penchant for hunting.

“These are dogs that absolutely just love searching,” Dr. Guest said. “These are dogs that if you throw a ball into the grass a hundred times, the dog just wants to go out and find it, go out and find it, go out and find it. And they are not using their eyes, they are using their nose.” Their noses are so finely tuned, she added, that they can detect odours in concentrations as low as 1.5 parts per trillion, equivalent to finding a teaspoon of sugar in two Olympic-sized swimming pools.

Recent studies in France and Germany have also shown that dogs can be trained to find COVID-19, but the British research is the most comprehensive so far and involved 3,500 samples and hundreds of volunteers.

Dr. Guest and the others started by training the dogs to detect the virus in fragments – roughly the size of a loonie – of socks, T-shirts and masks submitted by people who had been mildly infected or who were asymptomatic. Once fully trained, the dogs were tested in a double blind, randomized study where no one in the lab knew which samples had COVID-19. A final trial involved the dogs identifying COVID-19 cases as groups of volunteers walked through a room wearing T-shirts from people who had been ill.

Tala sniffs a sock sample.

Millie sniffs a sock sample.Neil Pollock/Medical Detection Dogs

The animals had to not only identify which samples had the virus but also correctly ignore those that tested negative. “The dog understands that he must give the correct answer to get a reward, not just find the odour,” Dr. Guest said.

The next phase of the trial involves using the dogs in real-world situations and Dr. Logan believes the results could be even better.

“These dogs have been detecting a very, very small amount of odour on some samples that are several weeks old,” he said. “Sometimes the odour that will be coming off an actual person that’s infected is going to be a much, much bigger signal. So it could be that the dogs may actually improve in their accuracy when they are tested in the field.”

Even if all the trials are successful, training enough dogs to cover airports won’t be easy. Dr. Guest said it takes about two months to train a dog and only 10 are currently in training. The researchers are trying to develop a chemical that mimics the scent of COVID-19 to simplify and speed up training, but that hasn’t happened yet.

There are also questions about how the dogs will perform in densely packed crowds and whether ventilation in airports will disperse odours and reduce the dogs’ effectiveness, said Lawrence Young, a virologist at the University of Warwick.

“The big question is ‘Will this approach work in the real world on people rather than samples of socks and shirts?’” he said.

Sign up for the Coronavirus Update newsletter to read the day’s essential coronavirus news, features and explainers written by Globe reporters and editors.