Can Dogs Smell Cancer?

Partly true, and the part that is true is genuinely interesting.

Dogs smell far better than people do. In laboratory studies, some trained dogs have picked out breath, urine, or tissue samples from people with cancer more often than chance would predict. That is a real result, published repeatedly, and it points at something real about the biology.

What has not been shown is that this works as a test. Across the published research, accuracy swings from barely better than guessing to near-perfect, and the studies with the tightest controls tend to produce the least impressive numbers. A 2022 systematic review of the whole field concluded that scent detection by animals "lacks evidence and performances so far to be applied in the clinic."

So: a fascinating line of research, not a thing you can currently be tested with.

Cancer changes cell metabolism, and metabolism produces chemicals. Some of those chemicals are volatile — they evaporate and end up in breath, urine, sweat and on the skin. The theory is that a tumor shifts the mix enough that a sensitive enough nose could notice.

Dogs have that nose. This is not a fringe idea, and it is not the same category of claim as, say, a cure hidden by pharmaceutical companies. It is an ordinary scientific hypothesis that researchers have been testing for over twenty years.

The most thorough look at the field reviewed 62 published studies from 2000 to 2021. Roughly two-thirds were about cancer, covering lung, breast, prostate, ovarian, colorectal, cervical and liver cancers and melanoma.

The headline problem is inconsistency. Sensitivity ranged from 0.17 to 1.0, and specificity from 0.29 to 1.0. In plain terms, some studies had dogs finding essentially every cancer sample and rejecting every healthy one; others had dogs missing more than eight cases in ten. That is not the profile of a stable, reproducible test. It is the profile of a result that depends heavily on how the experiment was set up.

Several specific problems recur:

• Almost nothing was tested under screening-like conditions. Only 6 of the 62 studies used a double-blind, unforced-choice design. Most used a forced-choice setup, where a dog is shown a row of samples and knows one of them is the positive. Real screening does not tell you in advance that a cancer is present somewhere in the room.
• Samples were reused. When the same sample comes back repeatedly, a dog can learn to recognize that individual sample rather than the disease. Reviewers flagged contamination and "memory effects" as widespread.
• Control samples often were not confirmed healthy. If some of the supposedly cancer-free samples came from people who had undiagnosed cancer, the dog gets marked wrong for being right.
• The evidence rests on very few animals. Fewer than 200 dogs worldwide have been studied, with no standard for how they are selected and high rates of training failure. Results from a handful of exceptional dogs do not generalize to dogs in general.

The reviewers rated the body of evidence as low quality with a high risk of performance bias.

No dog is an approved cancer test. Not in the United States, not anywhere. If you encounter a service selling cancer screening by dog, there is no regulatory approval behind it and no published evidence that it performs reliably.

Do not let a dog change your medical decisions in either direction. This is the part that matters most. If your dog has been sniffing or nudging at one spot, that is not a diagnosis — dogs investigate people constantly, for reasons ranging from a wound to a change in soap. And if your dog shows no interest at all, that is not reassurance either. A change in your body that has lasted a few weeks deserves a look because of the change, and the recommended screening tests for your age and history are worth doing regardless of what any animal does.

The risk with a story like this one is not that it is silly. It is that it is charming enough to be believed, and someone delays getting a lump checked because the dog seems unbothered.

The more promising direction may not involve dogs at all. If tumors really do change the mix of volatile compounds a person gives off, the useful outcome is identifying which compounds those are, and then building an instrument that measures them the same way every time. A machine does not have off days, does not become attached to its handler, and does not need retraining.

Dogs may end up being the proof that the signal exists rather than the tool that reads it. That would still be a real contribution.

This page is general education and is not a diagnosis. Decisions about screening and about symptoms belong with your own clinician.