Dogs Sniff Out Prostate Cancer With 98 Percent Accuracy

By Carl Engelking | May 19, 2014 4:40 pm


The power of dogs’ noses is well documented, and that reputation continues to improve. Researchers have discovered that our canine companions’ snouts may be more accurate than advanced laboratory procedures when it comes to detecting certain forms of cancer.

Researchers at the Istituto Clinico Humanitas in Italy have trained two dogs that can sniff out the scent of prostate cancer in urine samples with a success rate of 98 percent, a new study reports. The pool of over 600 subjects makes this the largest study ever conducted using cancer-sniffing dogs.

Smelling Out Cancer

Researchers used two professionally trained dogs to test their ability to detect prostate cancer from a pool 677 people. One group of participants was cancer-free; the other group ranged from individuals with low-risk tumors to those whose cancer had metastasized to other tissues.

The two dogs sniffed urine samples, and identified signs of prostate cancer with a combined 98 percent accuracy. In a few cases, the dogs identified cancer when it wasn’t there — called a false positive — accounting for the remaining 2 percent of cases. That success rate represents a vast improvement over the standard Prostate-Specific Antigen test, which has a false positive rate as high as 80 percent, Bloomberg reports.

The results were presented Saturday at the American Urological Association in Boston.

Cancer’s Scent

Dogs’ noses have four times the number of olfactory cells as humans, making them sensitive enough to detect the volatile organic compounds emitted from tumors.

Several smaller-scale studies have demonstrated dogs’ diagnostic abilities. Dogs have successfully sniffed out lung tumors, colon cancer, melanoma and ovarian cancer. However, despite these findings, using dogs at the doctor’s office has drawn skepticism — mostly regarding the logistics of using dogs in a clinical setting. Questions still remain about the type of dog that’s ideal for testing, how to systematize canine testing, and the financial feasibility of using dogs, Bloomberg reports.

Next Steps

A nonprofit called the InSitu Foundation hopes put the dogs’ potentially life-saving noses to work immediately, and they are looking to gain FDA approval for a canine medical scent detection kit that uses breath samples to detect lung cancer. The foundation is also funding new studies exploring canine detection of early-stage stomach and pancreatic cancers, which both lack screening methods.

Other researchers are focused on identifying the biological mechanisms dogs use to sense cancer. Using this information, researchers could build robotic sniffers that mimic dogs’ disease-sensing power.


Photo credit: benokky1972/Shutterstock

  • Dan Johnson

    If the dog sniffs you and wags “HE’S GOT IT”, then what is the actual probability that the patient has it? Unless the rate in the group being sampled is very high, then the dog’s diagnosis is quite a bit lower than “98% accuracy” would imply. They don’t tell us how many of the men (assuming they were all men) in the study actually had it, or how old they were. We can work out the actual chance of a positive having it for difference options. Someone can check my figures. Suppose we had 677 men and the rate was 10%, a bit like the normal population of old men (you can try different rates). So 67.7 actually have it (we could use fractions, or round them). The story said the false positive rate is 2%. So of the 677, 13.54 should have a false positive. (You could do the same calculation for a larger population.) The total number cases of those who have it and those who wrongly seem to have it would be 81.24. This means that if the dog diagnosis is YES, and if they don’t miss it when it is there, as the article implies, then the actual probability of having it is only 83% (67.7/(13.54+67.7)). If the true rate in the 677 is 5% with prostate cancer, then we still have a 2% false positive rate, and the number who really have it is 33.85, so if the dog says YES, then the probability that the patient actually has it is only 71% (33.85/(13.54+33.85)).

    The proportion actually with prostate cancer has to be 20% or more before the probability of the dog being right is over 90%. If the group getting tested only has 1% ‘true’ prostate cancer, then if the dog says YES, the actual probability of having it is only 33%.

    • timothyhood

      You are attributing only 10% of the sample as actually having prostate cancer! yet applying the 2% false positive to the entire set. The 2% false positive applies to the number that actually had prostate cancer. Therefore, with a 10% cancer sample! the 2% false positive would be 67.7 x 0.02, or 1.35 false positives.

      • Dan Johnson

        That is how ‘false positive’ is defined and used in epidemiology. If a test has a false positive rate of 5%, and 1000 people are tested, then 50 would be wrongly indicated as having the disease. That is Health 101.

        • John Z

          If a total of test diagnosis results in four types of possibilities (a +b +c +d) where a = positive test with the disease, b = positive test without the disease, c = negative test with the disease, and d = negative test without the disease then false positive rates used in epidemiology would be b/(a+b). You’re describing a 5% margin of error which would affect the entire population of the test or b+c/(a+b+c+d). False positives can only affect tests that are positive and not the entire sample group as it is impossible for you to be a false positive with a negative diagnosis and a positive disease (type c) as that would make you a false negative.

          • Michael Keener

            article is about dogs finding cancer, not a frickin math lesson…

          • VarshaMoretoniss

            my buddy’s sister makes $87 every hour on the internet
            . She has been unemployed for 6 months but last month her payment was $19402
            just working on the internet for a few hours. go right here M­o­n­e­y­d­u­t­i­e­s­.­C­O­M­

      • Michael Keener

        you tell’em.. they are all really screwed up…

    • Michael Keener

      you tell’em.. they are all screwed up…

    • Terry Kunick

      Since the article said 1 group of participants was cancer free and the other group had various stages of the disease, wouldn’t you think the two groups were similar in size. Would they run the test using only 67 men (yes, that assumption has to be correct-women don’t get prostate cancer) who they knew to have the disease, or would they use an equal number of cancer-free and therefore a little more than 300 with the disease? Seems to me the participants should be divided equally to run a good test.

  • Donna Schultz

    Bad news for dogs. I can see labs lined up with slave dogs sniffing away.

    • maneauleau

      Stray dogs could be trained and given food and shelter for their work

      • Olivia

        That’d be awesome

    • Bearpants42

      Here I was thinking, “Hey cool news this is going to save a lot of lives in the future. The comments are going to be supportive.” Way to find a way be a pessimist about it. Why you would assume the dogs would be kept as ‘slaves’ is beyond me. If anything, this is an opportunity for cottage industry.

      • Bill

        If the dogs were sniffing breasts and detecting cancer there, I’m sure Donna would have been all for it!!

        • Olivia

          Breast cancer has nothing to do with it. Cancer is cancer. You can get it on any part of your body. Males can get breast cancer too.

          • William Malmstrom

            So are you saying that women are not at MUCH greater risk for breast cancer than men are?!

          • Guest

            No. But men are also at a greater risk for prostate cancer, so… 😉

          • Olivia

            Never said that. But women can get cancer anywhere, just like men can (except for the prostate… we don’t have one of those. But we can get ovarian or cervical cancer)

      • Donna Schultz

        Cause I work in animal rescue and witness what man does to animals daily. Just a bit more intune with the problem than the average bear…uh pants.

    • Bill

      If the dogs were sniffing breasts and detecting cancer there, I’m sure you would have been all for it!!

      • Donna Schultz

        Not really. I only let my boyfriend sniff my breast.

        • Bill

          I said nothing about your breasts, but it is sad that he thinks that
          they are to be sniffed. If he gets prostate cancer, you may wish that
          the dogs are available to sniff him. But given his prowess with
          breasts, you may not care.

  • Guest

    That’s the definition of false positive in epidemiology. If a test has a false positive rate of 5%, and 1000 people are tested, then 50 would be wrongly indicated as having the disease. That is Health 101.

    • sleepvark

      Does this explain why so many dogs love to sniff people’s crotches?

  • Kelly Frier

    I just hope they get this all figured out before my prostate kills me.

  • garedawg

    Next time a dog comes up and sniff my rear end, I’m getting those PSA levels checked.

    • Michael Keener

      psa’s only 20% accurate…dog is 98%…

      • Jexiah8bit


  • RT

    If you look up information about this study, you can find the actual numbers of folks in two categories — ones with and ones without. There were close to 50% in each category. If these are published in reputable journals, I can assure you that the referees have “sniffed out” any problems with the statistics in the published papers (one can not say the same about quotes in the media of course). So, lets take the numbers at face value unless you can get access to the raw data and do a better analysis.

    Also a quick search of the web show that there are numerous studies showing dogs can detect a variety of cancers and other diseases. And, this study used urine samples, not body odors and the dogs are trained for specific odors, so your random untrained dog is not sniffing you for this purpose.

  • bwana

    “In a few cases, the dogs identified cancer when it wasn’t there — called a false positive — accounting for the remaining 2 percent of cases.”
    Since our tests are so inaccurate who’s to say the false positives were really “false”!?

  • Chris Hardy

    funny how people who supposedly luv dogs are trying to deride their abilities does it make you jealous that we are not so endowed dogs are truly man’s best friend and i do not and never would own one and all you haha people i believe the article says that they were sniffing urine samples

  • A. Chavez

    There should be more of these dogs trained and a list of where people can be checked by them. I trust dogs more than cancer doctors, because a dog does not care about money. Also they are amazing!!!!

  • Peter Apps

    Dogs are unlikely ever to be the basis of large scale disease surveillance programs, and so there needs to be a serious effort to identify the chemical compounds that the dogs are using to discriminate the odours of positive and negative urines. Once the compounds are identified the next step would be to develop tests based on repeatable and reproducible chemical reactions.

    Matching a dog’s sense of smell, and its ability to process its sensory input is a huge challenge, but it is tantalizingly close to being within reach.


Discover's Newsletter

Sign up to get the latest science news delivered weekly right to your inbox!


See More

Collapse bottom bar