Do Cancer-Sniffing Canines Pass the Smell Test?
Despite skepticism, researchers remain doggedly determined to nosing out disease.
Since 2011, Quebec-based CancerDogs has been using four beagles to identify cancer in breath samples from over 30,000 firefighters in 100 U.S. fire departments. Their accuracy rate is remarkable—the dogs are able to identify cancer over 95 percent of the time.
They’re not the only ones. Another program, the University of Pennsylvania’s Penn Vet Working Dog Center, has employed three dogs since 2013 to sniff out ovarian cancer from blood samples with 90 percent accuracy. Before suspending their three-year-old program last year, Calgary’s Clever Canines pinpointed lung cancer using seven dogs at an average accuracy rate of 95 percent on breath samples. And since 2014, researchers at Humanitas Hospital in Milan, Italy, reached a groundbreaking 98 percent detection rate using two German shepherds to recognize prostate cancer from urine samples.
These are only a handful of the programs using the roughly 250 million olfactory cells in a dog’s snout to help detect one of humankind’s deadliest diseases. And the numbers are impressive.
But Glenn Ferguson, founder of CancerDogs, cautions those figures often don’t tell the whole story. Some researchers, he’s discovered, use as few as eight samples, and the same ones each time, in order to achieve those stellar, near perfect numbers.
“[It’s not] a large enough sample selection to provide anything meaningful,” he told The Daily Beast. “If you’re training a dog with the same eight or 10 samples for six to nine months, how can you even miss one? I didn’t begin to train dogs until we had 50 positive cancer samples [he now has 200], which took us over eight months to get. We gradually introduced new samples to ensure the dogs weren’t getting habituated to the other samples.”
That’s likely one reason many in the medical community remain skeptical of cancer-sniffing dogs. New Jersey-based science journalist Arlene Weintraub, who literally wrote the book on the subject in Heal: The Vital Role of Dogs in the Search for Cancer Cures, acknowledged the issues with using dogs to smell out cancer.
“I think there’s pretty widespread agreement that early detection is a potential cure for a lot of cancers and that we need better early detection technologies,” she said. “But by the same token, [medical professionals] want these detection methods to be very, very accurate, and I think that’s where some of the skeptism comes in. They want to see proof that they’re going to work, and that they’re not going to have a lot of false-positives.”
False-positives, or specificity rates, are a big problem. CancerDogs’ specificity rate, for example, ranges from 60 to 70 percent (meaning 30 to 40 percent will be false-positives). The dogs are hitting on something; it may be pre-cancerous tissue, it may be another disease or condition altogether.
To ensure a broad sampling base, Ferguson designed a device called a sniffing station, made up of a platform with nine holes in a rack. Underneath the rack is a slider on wheels.
“We can slide the tray one way, then another way, so we can have 13 long trays with 17 samples on each, where we hide a few cancer samples, to give the dog a different set of samples in an instant,” Ferguson said. “And I’m eye level, so I’m watching him stick his nose in every hole, making sure every sample gets a good [sniff], and that he’s not going to miss anything. Which is why we run four dogs over all — up to 112 — of the samples a day.”
At the Working Dog Center, Dr. Jennifer Essler said they work from about 40 samples containing a mixture of cancer, benign and normal. “The samples are presented to the dogs not more than once,” she said. “For example, a dog might see Sample A once, Sample A and B mixed once, and Sample A, B and C mixed once — but only once for each odor profile. The accuracy rate is based off of the correct alerts on cancer and the correct ignoring of the non-cancer samples.”
But, said Essler, “We’re not necessarily looking for accuracy rates anymore, since all the dogs have done well and continue to do well.”
Where these four-legged detectives do exceptionally well is ferret out the unique odors emitted by malignant cancer cells contained in blood plasma samples, provided by chemists at Monell Chemical Senses Center.
Essler says the dogs have made even greater strides this summer, after they aced their tests with much more diluted samples.
“The dogs have to search for less and less of the odor because the amount the chemists are extracting is much lower than the dogs had been working with previously,” she said. “Right now they’re working at 16.5 micro litres of cancer — almost nothing — in 33.5 micro litres of saline. It’s a one in two dilution, about half of a drop total liquid.”
As laser-sharp as these canine nostrils are, they do tire. So the objective, at least of the Working Dog Center, is to replicate those nostrils with a kind of electronic nose. “Once we’re able to say what part of the cancer the dogs are smelling, they can more easily make their detection system,” said Essler of researchers at both the University of Pennsylvania and Monell, who are developing a nanotechnology sensor to ultimately do the dogs’ job.
Detector dogs don’t come cheap, costing in some cases upwards of $50,000 to train each dog—making the development of a machine all the more urgent.
“We believe we could have a working prototype within three years,” A.T. Charlie Johnson Jr., professor of physics and astronomy at U. Penn, said. “The largest obstacle is to generate sufficient funding for a significant sustained effort … We might be able to accelerate this process by forming partnerships with cancer centers at other institutions. Beyond that, I see no real obstacle other than whether our preliminary conclusions hold up as we expand our data base — which I believe they will.”
Johnson said the vapor sensor he and his colleagues invented can detect early-stage ovarian cancer as well as more advanced stages. “Our goal is to move forward with larger studies of ovarian cancer to provide even more solid confirmation of the odor signature and validate our ability to detect it with very high success rate,” he said, noting his team is also exploring the system’s potential for detecting other forms of cancer, especially those for which early diagnosis is impossible.
Tracie Nielson, founder of Clever Canines and lead research facilitator for its canine cancer detection program, said her team’s goal was to have an operational screening clinic using dogs as an early screening tool for patients at high risk for developing lung cancer. “If we could take high-risk patients and, instead of them being sent back every six months for another expensive treatment through their medical system, we could say, ‘Let’s see what the dogs think,’” she said. “The dogs would smell the breath and provide their opinion, and perhaps that would change the doctor or radiologist’s opinion to say, ‘Instead of six months, I want to see you next month or every three months. With lung cancer, in six months it could go from Stage 1 to Stage 4.”
But last year her program was put on hold after running out of positive samples necessary to keep the dogs’ accuracy levels up. Although she hopes to relaunch the program at some point, Nielson conceded the cooperation and partnership she had counted on from the medical community was not there.
“Everyone thought it sounded like a great idea, but the hopes of working alongside radiologists and oncologists had logistics difficult to overcome that stemmed from the medical restrictions or processes doctors are required to follow,” she said.
Ferguson launched his doggie detection company after his own frustration with the lack of involvement by the medical community. “We shouldn’t be picky about what tools we use to fight cancer,” he said. “Dogs are the right tool; they’re used for detecting bombs and drugs for that very reason.”
He does, however, emphasize that CancerDogs offers a screening tool, not a diagnosis. He advises people to see a dermatologist and a doctor for an official diagnosis.`