McBaine is a black-and-white English springer spaniel with a constantly wagging tail and a passion for sniffing out ovarian tumors. During a typical day at the University of Pennsylvania’s Vet Working Dog Center in Philadelphia, McBaine is brought into a back room that’s mostly bare except for a large metal wheel in the center of the floor. Attached to the wheel are twelve small cups, one of which contains a blood sample from a patient fighting ovarian cancer. McBaine runs around the wheel, smelling each cup one by one, the black tuft on the end of his tail waving happily throughout. Then he stops, places his paw gently on one of the cups, and sits down.
“Good boy!” his trainer yells, as she tosses the dog’s favorite rope toy across the floor. Play is McBaine’s reward for finding the cancer sample among the cups that contain healthy blood samples or nothing at all.
It was an unusually warm spring day in Philadelphia when I returned to my alma mater to learn about a completely different way dogs are helping in the war on cancer — with their noses. McBaine was one of four dogs enrolled in an innovative study with a lofty goal. After training the dogs to identify ovarian cancer by its smell, the researchers running the project planned to find out exactly what it is the dogs are detecting in the cancer and then invent a diagnostic device that could mimic the dog’s nose. That device would give physicians the power to find ovarian cancer long before its victims have any inkling that they are sick.
“We wouldn’t put dogs in every hospital or every laboratory,” said Cynthia Otto, a veterinarian, dog trainer, and executive director of the Penn Vet Working Dog Center. “Our whole goal is for the end product to be a chemical sensor — a machine with the sensitivity and specificity of the dog.”
The Penn Vet Working Dog Center is not on the university’s main campus, but two miles away in a nondescript office complex a stone’s throw from the Pennsylvania Turnpike. I walked through the main entrance into a giant room with blue foam tiles on the floor and the elements of an obstacle course scattered about — a three-step ladder, a ramp, and several old tires. A dry-erase board on the wall listed all the dogs who were being trained at the center and their progress mastering a variety of skills, from search and rescue, to agility, to impulse control.
As Otto took me on a tour, the sound of the hand-held clickers the dog trainers were using echoed through the halls. The Penn Vet Working Dog Center, she explained, was founded in 2007, but it didn’t have its own headquarters until it opened in this building five years later. The center was born from Otto’s passion for training search and rescue dogs, which she did in her spare time while working as an emergency veterinarian at Penn. Otto volunteered to provide medical care for the canine search-and-rescue teams that were deployed after Hurricane Floyd hit Florida in 1999, and again in the wake of the September 11th attacks.
After Otto returned from Ground Zero, she got word that the American Kennel Club was looking to fund a study of the 9/11 dogs. She immediately applied, and was chosen to manage the study.
“We’ve been following the dogs — their health, their behavior. We get x-rays and bloodwork every year, and when the dogs die, we examine the causes of death,” Otto said. “As I looked at the whole intersection of dogs and work, I realized there’s a lot we don’t know.”
A decade after 9/11, the Working Dog Center opened its doors with the mission of taking what was learned from the Ground Zero study and others like it to improve the breeding and training of working dogs. The center’s trainers started by teaching dogs to detect explosives and narcotics. Later they branched out into the medical realm, pairing dogs who could detect dangerous drops in blood sugar levels with diabetic patients. The cancer-detection project was the center’s newest pursuit.
The dogs at Penn’s center were named in honor of the dogs and trainers who were deployed to the 9/11 sites, or the victims who died in the act of terrorism. (McBaine’s namesake was a search-and-rescue dog who served as part of the control group for the 9/11 canine study.) McBaine and his fellow trainees lived with foster families, who brought them to the center Monday through Friday to be trained.
Otto was a businesslike clinician who, I quickly realized, turned to mush in the presence of her trained dogs. She took me to the back of the center to meet Ffoster, (spelled with a double F), a calm three-year-old yellow Lab who was the newest addition to the cancer-detection program. She was named after 9/11 victim Sandra N. Foster.
“Here’s Ffoster, here’s the good girl,” Otto said, speaking to the dog in a gentle tone as she let her out of the crate where she had been napping. “Do you want to come out and say hi?”
Earlier that day, Ffoster took her turn at the cancer wheel. Ffoster was more deliberate and considerably less bouncy than McBaine, walking slowly around the wheel, checking each cup carefully before coming to a stop. She was right. She got a treat as a reward.
“You did good today,” Otto cooed. “You’re learning good stuff.”
The notion that dogs might be able to detect cancer first emerged in 1989, when the prestigious British medical journal The Lancet published a five-paragraph letter titled “Sniffer dogs in the melanoma clinic?” In the letter, two physicians at King’s College Hospital in London described the case of a forty-four-year-old woman, who came into their clinic with a lesion on her left thigh. She told them she made the appointment because her dog, a Doberman–border collie mix, was constantly sniffing a mole on her leg.
The dog, the physicians wrote, “showed no interest in the other moles on the patient’s body but frequently spent several minutes a day sniffing intently at the lesion, even through the patient’s trousers.”
One day the woman donned shorts, at which point her dog tried to bite the mole off completely. That’s when she decided she should get it examined. The object of the dog’s obsession turned out to be a malignant melanoma.
“Perhaps malignant tumours such as melanoma, with their aberrant protein synthesis, emit unique odours which, though undetectable to man, are easily detected by dogs,” the physicians wrote.
They reported that the dog had saved the patient’s life, because when she came into the clinic the tumor was so small they could easily treat it. She was cured.
Those five paragraphs were enough to mobilize dog-loving scientists all over the world to design studies testing whether dogs can smell cancer. Over the next several years, dozens of papers were published in medical journals, most of them reporting astounding success rates — often with 90 percent accuracy or better — among dogs trained to sniff cancer.
Otto read those papers avidly and watched with fascination as other dog-training organizations began teaching dogs to sniff cancer. She wanted to do the same, she just didn’t have the funding or the know-how she needed to get started.
Then one day shortly after the new center opened, Otto’s phone rang. It was a cold call from another Penn faculty member, ophthalmologist Jody R. Piltz-Seymour, a dog lover who had also been captivated by all the reports about dogs who could sniff cancer. Piltz-Seymour had lost her father and aunt to pancreatic cancer. A second aunt and her mother-in-law both died of ovarian cancer.
Piltz-Seymour was particularly interested in research done by a Swedish scientist named György Horvath. In a 2008 paper, Horvath described how he trained a four-year-old black giant schnauzer to recognize tissue samples from late-stage ovarian cancer patients by their scent. Then Horvath ran the dog through two series of tests to see if she could tell the difference between ovarian cancer and healthy tissue, even when the cancer samples were taken from women in the early stages of the disease. In both testing situations, the dog correctly identified all the cancerous samples, and when presented with 80 pieces of healthy tissue, only twice did she falsely identify them as cancerous.
Horvath’s results made Piltz-Seymour realize dogs could be the key to early detection, but she knew she would need help translating the findings into a useful clinical tool. Then the Working Dog Center opened, and she figured she had nothing to lose by giving it a shot.
“I thought we had this phenomenal new resource right in the backyard,” Piltz-Seymour said. As she reminisced about her cold call to Otto, she began to laugh. “You have to picture this — here’s an ophthalmologist calling about dogs and cancer.”
Little did she know that Otto was already keen on the idea and was looking for an ally to help get a study started.
“Jody had done a lot of research and wanted to see if she could help move it forward,” Otto said. “She spearheaded this — it was personal.”
Piltz wanted to start with ovarian cancer, not just because of her family history, but also because she believed it was where the dogs could make the most impact.
“Everybody in my family succumbed to their cancers because none of their cancers were caught early,” Piltz-Seymour told me. “The vast majority of the time, ovarian cancer is caught late. But when caught early, the success rate with treatment is brilliant. That makes it the perfect cancer to study.”
Indeed, the five-year survival rate for ovarian cancer patients who are diagnosed and treated before their cancer spreads is 92 percent. Yet only 15 percent of ovarian cancers are found that early. These tumors rarely produce symptoms in their early stages, and when they do, their victims commonly mistake what they’re feeling for something far more benign, like constipation.
As I listened to Piltz-Seymour talk about this conundrum, I thought about Beth, who didn’t know she was gravely ill until her cancer had spread beyond her stomach.
“Gastric cancer is similar to ovarian cancer in that it doesn’t produce symptoms until it’s too late,” I told Piltz-Seymour.
The five-year survival rate for gastric cancer patients who are diagnosed at the earliest stages is as high as 71 percent. That’s because surgery and chemo can be quite effective when the cancer hasn’t spread. The problem is, the only good way to find stomach tumors is with an endoscopy, an invasive and sometimes risky procedure that involves sticking a tiny camera down the patient’s throat while he or she is sedated.
In most countries, healthy people aren’t given an endoscopy unless they have a strong family history of gastric cancer. Because we had no history of the disease in our family, the doctors who saw Beth for her checkups and minor complaints never would have imagined they should look in her stomach for signs of trouble. So they didn’t.
In Japan, where the prevalence of gastric cancer is high, mass screening is routine, and now half of all cases of the disease are found in the early stages. Hence the five-year survival rate following surgery for Japanese patients is a staggering 96 to 99 percent.
Whenever I see dogs like McBaine sticking their noses into human tissue samples and instantly identifying the cancerous ones or I read about them in the press, I have a fantasy: I imagine everyone in the world going to the doctor for a routine physical each year and breathing into a tube, or leaving a blood or urine sample, and learning on the spot if there’s any sign of cancer in their stomachs, lungs, ovaries, or anywhere else in the body where tumors can easily hide. If dogs can detect cancer, then it seems only logical that scientists can unlock the mysteries of the canine sense of smell and replicate it with a diagnostic device — an electronic nose that could be put to work in every medical clinic in the world, but that wouldn’t need to be fed and walked twice a day.
Piltz-Seymour told me she had exactly the same fantasy.
“I kept seeing these very interesting articles about dogs detecting cancer, but no one seemed to follow through to develop anything that could be used in clinical practice,” she said.
When she set out to develop an early-detection system for ovarian cancer, Piltz-Seymour knew she needed to put together a team that included more than just dog trainers. She began calling and e-mailing oncologists and doctors, most of whom ignored her. Then she got a return call from a gynecologist at Penn named Janos L. Tanyi. He offered to join the team and referred Piltz-Seymour to two other Penn faculty members he believed would be interested: molecular scientist George Preti and physics professor A.T. Charlie Johnson. Preti and Johnson had already done some research to try to identify the odors associated with ovarian cancer. They both jumped at the opportunity to add dogs to their team.
“Our initial hypothesis that ovarian cancer would emanate a unique odor signature was based on the work done in Europe with dogs,” said Preti, who in addition to his appointment at Penn was a scientist at the Monell Chemical Senses Center, a Philadelphia research house devoted entirely to unlocking the mysteries of smell and taste. “I knew dogs would be a very good biological detector.”
* * *
The dog’s nose is among the most powerful sensory organs the world has ever known. Dogs have 200 million olfactory receptors (ors), which are proteins on the surface of the neurons inside their noses that help their brains to perceive and process odors. Humans, by contrast, have just five million ors. The mucous membranes inside dogs’ nostrils are extraordinarily dense, and their noses exquisitely structured for the efficient processing of smells. That allows them to detect odors that are present in quantities as low as parts per trillion — an amount so minuscule it’s almost impossible to imagine. On top of that, many experts believe the proportion of the dog’s brain that’s dedicated to analyzing scents is 40 times larger than that of humans. Suffice it to say, it’s enough for scientists to estimate that the dog’s ability to recognize minute traces of particular odors is one million times better than that of people.
But what exactly are the dogs sniffing in cancer that they find so alarming? Shortly after The Lancet publication, a popular theory emerged suggesting that tumors must be emitting volatile organic compounds (vocs), which are carbon-based chemicals that naturally occur in the body. Since the late 1980s, scientists have identified a variety of vocs that are overly abundant in cancerous tissues, including some types of benzene.
Because vocs have a low boiling point, they easily evaporate and travel to places far from the original tumor, including through urine, blood, and exhaled breath. As the scientific world became aware of the dog’s ability to sniff cancer, some researchers set out to prove that tumors have unique voc signatures, or specific odor profiles that distinguish them from healthy tissue and that can be easily detected and analyzed — if not by dogs than by some sort of non-invasive test.
Among those researchers were scientists at the Cleveland Clinic, who in the early 2000s collected breath samples from lung cancer patients, healthy volunteers, and patients with other lung diseases. They analyzed the exact chemical makeup of the cancerous breath and used it to develop a device containing thirty-two tiny sensors.
There are two criteria used to judge the value of a diagnostic test in oncology: sensitivity and specificity. A sensitive test can pick up the presence of cancer. A specific test can do so with a high rate of accuracy — meaning it rarely detects the disease in people who don’t have it.
Back when the Cleveland Clinic was doing its study, pet scanning in lung cancer was 97 percent specific and 78 percent sensitive. The Cleveland Clinic’s electronic nose stacked up pretty well, scoring 92 percent specificity and 71 percent sensitivity. That means the electronic nose missed the abnormality about a third of the time, but when it found something fishy, it was nearly flawless at figuring out when it was cancer and when it was not.
The dogs also performed well. In a 2012 trial of breath samples from 220 people, sniffer dogs scored 90 percent on sensitivity and 72 percent on specificity for lung cancer, even when the scientists tried to confuse them with samples from patients with the non-cancerous lung condition known as chronic obstructive pulmonary disorder (copd). Other dog studies that had trained dogs with urine, blood, or tissue samples achieved comparable or better results in detecting melanoma, as well as ovarian, breast, bladder, and colorectal cancer.
As it became more and more clear that dogs could smell cancer, some researchers wondered if it really made sense to spend time and money creating electronic noses. Some thought a smarter strategy would be to employ the dogs themselves in medical clinics as a sort of living warning system — an advance guard to alert doctors and patients that further testing would be a good idea.
One such researcher was Robert Gordon, an obstetrician and gynecologist at Scripps Health in La Jolla, California. Gordon envisioned equipping volunteer church groups, for example, with urine-collection kits and a few dogs who were trained to sniff them.
“If anybody came up positive then they could be sent to a center where they could get more sophisticated testing,” he said. “I thought in areas that were not developed, in the backwoods, you could use a dog.”
Gordon, who is hearing impaired, was inspired by his own dog, a Chihuahua-beagle mix named Ginger, whom he had trained to be a certified hearing-assistance dog.
“She’s a fabulous animal,” Gordon said. “She goes everywhere with me — restaurants, movies — helping me as a hearing dog. I put a towel down so her hair doesn’t get on anything. She snores through the loud movies.”
In 2004, Gordon set out to train Ginger and nine other dogs to discriminate between patients who had prostate cancer, those who had breast cancer, and those who were healthy, based on urine samples. The dogs made a lively crew that included a miniature goldendoodle, an Italian greyhound, and a Pembroke Welsh corgi. But few of them showed much talent for sniffing cancer — only two performed better than chance, or luck, at specificity in detecting each of the two tumor types. None were better than chance at sensitivity, not even Ginger.
After thinking about it, Gordon concluded that the imperfections were not in the dogs themselves but rather in the way he had designed the study. He could only gather a few urine samples, which he froze, thawed, and re-used several times during the training process. He wondered if all that freezing and thawing might have destroyed the scents he was training the dogs to recognize. He had also discovered that the dogs would get fatigued after less than two hours of work. It all made Gordon realize that developing an electronic nose might indeed be far wiser than using dogs at medical clinics or in the backwoods.
“It’s impractical to spend the money on dogs,” Gordon said. “That money would be best served finding what the chemicals are that the cancer puts off — the chemical signature that makes them so destructive.”
Other researchers had already discovered that finding the vocs was the easy part. The hard part was figuring out how to train a machine to recognize cancer vocs amid the thousands of other scents emitted by the human body. In dogs, that process is known as imprinting, and it involves teaching the animals to recognize the scent of cancer simply by exposing them to it. Much like a bloodhound who’s directed to find a missing child might be presented with a t-shirt infused with that child’s scent, a cancer-detection dog is asked to sniff cancerous tissue, or samples of urine, blood or breath from sick patients, and then taught to find that scent while ignoring everything else.
As the research papers proving the dog’s ability to sniff cancer continued to pile up, a handful of scientists took on the challenge of developing an electronic nose that could be imprinted with the scent of cancer as effectively as a dog’s could. Among the most passionate scientists in that pursuit was someone in my sister’s backyard: Hossam Haick, a professor at the Technion-Israel Institute of Technology in Haifa, Israel.
In 2007, Haick began developing a device he dubbed Na-Nose, short for nanotechnology nose. He used tiny gold particles to create an array of electronic sensors, which were connected to a tube. The result was a device that operated much like a breathalyzer, but was far more sensitive because it could analyze up to one thousand different gases in exhaled breath. Haick set out to prove that the Na-Nose could detect much more than the mere presence of cancer — that it could also distinguish one form of the disease from another.
Haick’s main challenge, he told me, was figuring out how to deal with all the confounding molecules, as he called them, that appear in breath, like mouthwash ingredients, or residual odors from the garlic pizza a patient ate two days before taking the breath test (and failed to eliminate with the mouthwash). Haick, a chemical engineer, used artificial intelligence to train the Na-Nose to recognize cancer vocs and ignore everything else. Over the next few years, Haick and his colleagues tested their device in clinics around the world, recruiting four thousand patients with a variety of different cancers to try it.
The results were often remarkable. For instance, in 2012, Haick teamed up with scientists in China and Latvia to test his technology in 130 patients with stomach disorders. Thirty-seven of the patients had gastric cancer, while the rest had ulcers or a benign condition. They found that the Na-Nose technology could distinguish between the cancerous and non-cancerous diseases with 83 percent sensitivity and 96 percent specificity. Even better, it could pick out the patients who had early-stage gastric cancer from those who were in the late stages of the disease with 89 percent sensitivity and 94 percent specificity. The Na-Nose’s accuracy was consistent regardless of the patients’ tobacco and alcohol consumption — habits that had the potential to produce confounding odors.
I told Haick about my sister and asked if he thought his device might someday be used to screen healthy people, allowing stomach tumors to be detected even when they’re completely silent.
“Absolutely,” he replied. “That is our intention. We want to catch the cancer at the early stages before anybody has symptoms. Then we can increase the survival rate by using currently available treatments. It’s true with other cancers, too.”
In a later study, presented at a medical meeting on stomach disorders in the summer of 2014, the Na-Nose performed even better, distinguishing gastric cancer from ulcers and pre-cancerous lesions with 98 percent specificity.
Haick’s technology proved equally adept at discerning lung cancer, often performing better than sniffer dogs or the Cleveland Clinic’s early rendition of an electronic nose. In a study published in 2012, for example, the Na-Nose could determine whether the nodules found in patients with lung disorders were benign or cancerous with 86 percent sensitivity and 96 percent specificity.
“If you detect the lung cancer at the early stages you can increase the survival rate from 10 percent to more than 70 percent,” Haick said. “But only by early detection.”
Excerpted from "Heal: The Vital Role of Dogs in the Search for Cancer Cures" by Arlene Weintraub. Copyright © 2015 by Arlene Weintraub. All rights reserved. Published by ECW Press Ltd.