Science

Studies suggesting dogs can detect cancer may lead to innovative early detection technologies. By Wendy Zukerman.

Using sniffer dogs to diagnose prostate and other cancers

A cocker spaniel during tests at Medical Detection Dogs in Britain.
Credit: CORBIS

The dog was constantly sniffing at a mole on her owner’s left thigh. It was uncanny. She wasn’t interested in other spots, but for several minutes a day intently sniffed this one. Things got ridiculous when the border collie-Doberman cross tried biting off the blemish. After months, the owner visited Dr Hywel Williams at the dermatology department in King’s College London. The curious mole, now as thick as a 10-cent piece, was a malignant melanoma. 

“The dog may have saved her owner’s life,” wrote Williams, documenting the case in The Lancet in 1989. The canine prompted its owner to get treatment when the cancer was still curable, and from this remarkable tale Williams had a tantalising thought. Perhaps, he wrote, malignant tumours “emit unique odours” that are imperceptible to us but “easily detected by dogs”.  

“It was high-risk stuff to dare suggest that dogs could smell skin cancer,” says Williams, now a professor at the University of Nottingham. But decades later, evidence is emerging that dogs – with their remarkable ability to pick up smells – may be trained to diagnose cancer.

Schooled canines have detected lung, breast, prostate, ovarian and melanoma cancers by smelling skin as well as urine, exhaled breath and extracted tumours. “Imagine a future in which the term PET takes on a new meaning,” wrote Dr Michael McCulloch at Pine Street Foundation in California, referring to PET medical imaging. His editorial was titled: “Will there be a lab in the lab?”

Earlier this year, a British team announced it had received ethics approval to begin Britain’s first prostate, kidney and bladder cancer trial using dogs. The plan is to take samples from 3000 participants, making it the largest trial of its kind. Meanwhile, Central South University in China is recruiting patients for a trial into whether dogs can sniff tumours. And in Paris, researchers at Tenon Hospital are launching a new program with two dogs trained to detect prostate cancer. 

Williams’ anecdote at King’s College was the first report in modern literature of a dog detecting a cancer. And it took a decade for the doctor to see the curiosity again. This time the hero was Parker, a pet labrador, who “persistently” sniffed a small patch of eczema-looking skin on his owner’s thigh, which was later found to be a basal cell carcinoma. Soon, a handful of academics around the world started seriously investigating whether dogs could smell cancer.

The ultimate goal is to improve our blunt methods of diagnosing some cancers. High-resolution CT, MRI and PET scans can be accurate, but they’re expensive, and can confuse benign lesions for pathological tumours, leading to unnecessary surgery and anxiety for healthy patients. 

Particularly troublesome is prostate cancer, which is partly diagnosed with a prostate-specific antigen (PSA) test. This test detects the level of a protein in the blood, which tends to rise in people with prostate cancer. But PSA levels spike for other reasons. “We’ve been desperate for years for a better test than the PSA,” says Dr Elizabeth Hovey, an oncologist at the Prince of Wales Hospital in Sydney. So can hounds really help? 

McCulloch says yes. He describes a dog’s sense of smell as “extraordinary”, noting that pooches can pick up scents in concentrations “as low as parts per trillion”, even discriminating complex chemical mixtures, such as those found in human breath. 

And to test these skills, McCulloch and colleagues asked cancer patients to exhale into a tube that was quickly sealed. Five dogs, specifically chosen for their “eagerness to sniff objects and respond to commands”, were trained for several weeks to detect the breath of a cancer patient. Dogs were rewarded with snacks if they sat down, or lay, in front of a cancerous sample. A mild rebuke of “no” was given for resting in front of an empty tube. 

In the final exam, tubes filled with the breath of cancer patients or healthy whiffs were shown to the dogs. Pups picked out lung cancer patients with 99 per cent accuracy, and breast cancer was detected 88 per cent of the time. Publishing in Integrative Cancer Therapies in 2006, McCulloch wrote that the dogs compared “favourably” with conventional methods of cancer detection. 

“This approach has surprised and challenged the medical community,” says Professor Olivier Cussenot at Tenon Hospital. “We don’t know if all cancers could be diagnosed by dogs, but we suppose that it is the case.” He says that when cells turn cancerous their metabolism changes, triggering the production of particular volatile organic compounds that are detectable by dogs. 

A golden retriever and schnauzer have been trained to detect melanoma on patients, while a schooled riesenschnauzer could spot extracted ovarian cancer samples with almost perfect accuracy. In a review paper, McCulloch described that canine as “remarkable”. 

More recently, Cussenot’s team in France taught a Belgian malinois shepherd to accurately detect prostate cancer by smelling urine samples. In European Urology the authors cautiously warned, “We obtained these powerful results with one dog, and this may not be reproducible with other dogs.” But, last year, success was struck once more, this time with an Italian team and two German shepherds. 

While results are promising, Cussenot still calls his work a “proof of concept”, as no studies have been replicated using the same training procedures and dog breed. This makes it difficult to know if the results are generalisable. Sydney-based oncologist Hovey, who is not involved in dog trials, adds that the power of pooches will be unclear until studies directly compare dogs with conventional detection methods, such as combinations of PSA tests and MRI scans.

Some studies have also been criticised for using the same samples to train and test the dogs, which means it’s possible canines are merely remembering a particular scent rather than truly detecting a “cancerous” signature. Hywel Williams, who first daringly suggested that dogs could sniff cancer, is acutely aware of this problem, publishing a paper in BMC Urology last year noting that “it is very easy to draw misleading conclusions”. He says it took four years to publish his critical work, while poor-quality studies “with hyped-up positive results” quickly get the stamp of approval and “hit the media”. “The public just love the idea of cuddly dogs being able to detect cancer,” says Williams. “What we need are good studies.” Still, McCulloch defends his work, and the canines’ competence, citing well-designed studies that produce impressive results.

It’s still unclear how many dogs could be schooled for cancer detection. While the few published studies are overwhelmingly positive, recently, one German trial was terminated because of the “inconsistent training status of sniffer dogs”. And Williams’ recent paper attempted to train 10 dogs to detect prostate cancer. Only two made the cut; the others were “too excitable” or “insufficiently motivated”. 

“It costs a lot of money and takes a lot of time to train a dog,” says Dr James Covington of Warwick University. Dogs also get tired, so can “only work for short periods at a time”, he says. Covington wants to exploit the scent of cancers, without canines. He’s developing so-called “electronic noses”, devices that detect chemical fingerprints using sensors, combined with algorithms that recognise patterns. Even with his promising dog work, Cussenot agrees that e-noses are the real endgame here. Unlike sleepy, excitable dogs, electronic contraptions “could be easily used in routine clinical practice”, he says.

But in his 2012 editorial, McCulloch noted that “dogs still appear to be ahead in the race” when compared with e-noses. This is largely because it’s unknown precisely what the dogs might be smelling, making it impossible to create a device mimicking their potential skills. Still, some chemical patterns have been spotted within the breath and urine of cancer patients, and Cussenot’s team in France is now embarking on a project to pit dog against e-nose. 

While critical and cautious, Williams still hopes his “original ideas will bear fruit” and dogs may be useful for cancer diagnosis. But he’s looking forward to future studies that have more bite than bark.

This article was first published in the print edition of The Saturday Paper on Oct 10, 2015 as "Pet scans". Subscribe here.

Wendy Zukerman
is a science journalist and host of the Science Vs. podcast.