Dogs can smell cancer because tumors produce distinct chemical signatures that enter a person’s breath, urine, blood, and sweat. These chemicals, called volatile organic compounds (VOCs), are released in tiny concentrations that human noses can’t detect but dogs can pick up in the parts-per-trillion range. In controlled studies, trained dogs have identified lung cancer from breath samples with 99% sensitivity and 99% specificity, meaning they almost never miss a case and almost never give a false alarm.
What Cancer Smells Like to a Dog
Cancer cells have altered metabolism compared to healthy cells. They burn fuel differently, produce different waste products, and shed different molecules into the bloodstream. Some of those molecules are small enough to evaporate into the air, which is what makes them “volatile.” These airborne chemicals escape the body through exhaled breath, skin secretions, and urine, creating a scent profile that differs from a healthy person’s.
The specific compounds vary by cancer type. Lung cancer patients exhale elevated levels of chemicals like benzene, styrene, and hexanal. For breast cancer, a prediction model using five VOCs in exhaled breath achieved 93.8% sensitivity and 84.6% specificity in identifying the disease. Ovarian cancer also produces a detectable odor signature in blood plasma, though researchers are still working to isolate exactly which compounds dogs are responding to. In each case, the cancer doesn’t produce a single telltale molecule. It creates a pattern of dozens of compounds at unusual concentrations, forming a kind of chemical fingerprint.
Why Dogs Are So Good at This
A dog’s nose contains roughly 300 million olfactory receptors, compared to about 6 million in a human nose. But the advantage goes beyond raw receptor count. Dogs use what scientists describe as “broadly selective receptors and combinatorial signal processing,” meaning each receptor responds to multiple chemicals and the brain assembles the full picture by combining signals from many receptors at once. This lets dogs detect and distinguish complex mixtures at vanishingly small concentrations.
The result is a biological detection system that can pick out one or two molecules among trillions of others. When a trained detection dog sniffs a urine sample or a person’s breath, it’s not just noticing “something different.” It’s recognizing a specific pattern it has been taught to associate with cancer. In prostate cancer screening using urine samples, trained dogs achieved a 99% success rate.
How Dogs Are Trained to Detect Cancer
Training follows the same basic principles used for bomb-sniffing or search-and-rescue dogs. A dog is presented with multiple samples, only one of which contains the target scent. When it correctly identifies the cancer sample, it gets a reward, usually a toy or treat. Over many repetitions, the dog learns to freeze, sit, or stare when it encounters the target odor.
At the University of Pennsylvania’s Working Dog Center, dogs are trained on a stainless-steel scent wheel outfitted with ports containing malignant samples, benign samples, normal samples, and various distractors. A dog named Osa, for example, learned to methodically sniff each port and freeze when she detected ovarian cancer in plasma. Not every dog succeeds. Trainers select candidates based on breed, temperament, and drive. Factors like arousal level, the relationship between the handler and dog, early life experiences, and even the need to pant in hot environments can all affect performance.
Accuracy Across Different Cancers
The numbers vary by cancer type and study design, but the best results are striking. In a study published in Integrative Cancer Therapies, dogs detecting lung cancer from breath samples achieved 99% sensitivity and 99% specificity when compared against biopsy-confirmed diagnoses. For breast cancer in the same study, sensitivity was 88% and specificity was 98%. These figures held across both early-stage and late-stage cancers, which is notable because early detection is where conventional screening often falls short.
Ovarian cancer research is still in earlier stages, but the fact that dogs can reliably distinguish malignant plasma samples from benign ones confirms that ovarian tumors produce a detectable scent. Researchers are now working backward from the dogs’ responses, using the animals as a kind of living quality check: they isolate chemical fractions from patient samples and present them to trained dogs to see which components “smell like cancer.”
Why Dogs Aren’t in Clinics Yet
Despite the impressive accuracy, dogs haven’t replaced screening tests for a practical reason: they don’t scale. Training a reliable cancer-detection dog takes months, and each dog can only process a limited number of samples per day before fatigue sets in. Clinically certified samples cost roughly $1,000 each to collect, document, and ship for analysis. And performance depends on variables that are hard to standardize, including handler influence, the dog’s mood, and environmental conditions.
Public acceptance is another barrier. In a large survey of concertgoers who experienced detection dogs screening for COVID-19, only about 40% supported using detection dogs in hospitals, while over 57% opposed the idea. People are comfortable with dogs in airports but less so in medical settings.
A feasibility study using detection dogs to screen for SARS-CoV-2 among thousands of concertgoers showed performance comparable to PCR tests, which demonstrated that dogs can work reliably in real-world, high-volume scenarios. Efforts are now underway to establish EU-level quality standards for medical detection dogs, similar to the existing standards for explosive detection teams.
Electronic Noses and the Future of Scent-Based Screening
The most promising direction may not be dogs themselves but the technology they’re inspiring. Researchers at MIT have developed a miniaturized detection system that uses mammalian olfactory receptors as sensors, with data processed by a smartphone. In controlled tests, this device proved 200 times more sensitive than a dog’s nose at detecting trace molecules. The catch: it’s “100 percent dumber,” as the lead researcher put it. The device can detect the chemicals but struggles to interpret what they mean as a pattern.
When both the artificial system and trained dogs were tested on the same prostate cancer urine samples, both scored above 70% accuracy. The gap between the device’s raw sensitivity and a dog’s interpretive ability highlights how much of the magic lies not in the nose itself but in the brain behind it. Closing that gap with machine learning is an active area of work, and the goal is a portable, inexpensive device that could screen for cancer the way a breathalyzer screens for alcohol.
For now, dogs remain the gold standard for scent-based cancer detection. They can’t be deployed at the scale of a blood test, but they’ve proven something that matters enormously: cancer has a smell, and that smell is detectable long before many tumors would show up on a scan.