Dog Sniffs Out Cancer In Human Urine

randomErr writes: University of Arkansas for Medical Sciences found out that a scent-trained dog can identify thyroid cancer in human urine samples 88.2 percent of the time. Frankie, a male German shepherd mix identified the presence of cancerous cells in 30 out of 34 samples. The shepherd was only slightly less accurate than a standard thyroid biopsy. This offers the possibility of a cheaper, less invasive approach to diagnosis of the illness said Donald Bodenner, M.D., PhD, the study's senior investigator.

Comments

the 11.8%?

[Anonymous Coward]

are they false positives or failure to detect?

if it's false positives, that'll get found later....... not a big deal.

if it's a complete miss-- ouch...

Re:the 11.8%?

[ganjadude]

good point. but I think it would make more sense to do things like this

dog sniffs - finds nothing, get a second opinion. I think when it comes to cancer and other bad diseases a second opinion is always warranted.

Re:the 11.8%?

[ShanghaiBill]

Another thing that would make sense would be to try other dogs, including other breeds. Then give them more training. Then mate the best cancer detectors. Within a few generations (2 years/generation for dogs) we could likely get the error rate below 5%. That is better than the biopsy, andlthough the dogs may have false positives, so can the biopsy.

Re:the 11.8%?

[CrimsonAvenger]

Before running any test, ask yourself what you're going to do if the test succeeds. Then ask yourself what you're going to do if the test fails.

If they're the same thing, then don't run the test.

In other words, if you're going to ignore the dog, don't bother with the dog.

What would the combined accuracy be?

[Anonymous Coward]

I did not RTFA but I'm curious as to whether both the dog and biopsy tend to fail on the same samples, or if we could approach near perfect accuracy by using both?

Re:What would the combined accuracy be?

[Jane Q. Public]

I did not RTFA but I'm curious as to whether both the dog and biopsy tend to fail on the same samples, or if we could approach near perfect accuracy by using both?

The problem with scent dogs -- as we have found to our dismay with drug-sniffing dogs -- is that while they CAN distinguish and react to the odors they are trained to detect, instead, in independent tests, they did not. Instead they reacted to subtle cues from their handlers, in preference to the odors they were trained to detect.

The cues were so subtle that the handlers themselves often did not realize they were sending signals.

Having a tool that CAN distinguish these things easily is not the same as having one that WILL.

Re:What would the combined accuracy be?

[pr0fessor]

When I was in highs school I had a friend who is allergic to dogs that was stopped by an officer. They searched him and his car and called in a dog. He protested because of allergies so they were all the more determined to search him and the car with dog because they thought he was trying to hide something. Nope they ended up taking him to the hospital and since he was a minor his parents filed a complaint that resulted in the officers' terminations. Afterward they worked as security guards they would scowl and follow us around the mall.

The Talisman came out when?

[rmdingler]

The canine olfactory organ is thousands of orders of magnitude more sensitive than ours... identifying drugs, bombs, and cancers is rudimentary to our best friend.

Yet, despite this superpower, they choose roll around in the foulest smelling dead shit they can find.

Are there smells we cannot appreciate in the same vein that there are sounds we cannot hear?

Re:The Talisman came out when?

[DigiShaman]

Yet, despite this superpower, they choose roll around in the foulest smelling dead shit they can find

Obviously your primitive sense of smell can't appreciate the wonderful bouquet of such odors that a dog can.

Why not multiple dogs?

[Irate Engineer]

Why just one dog (and why Frankie)?

Why don't they use 10 dogs and run the sample by each? If 9 out of 10 dogs agree that the sample indicates cancer, wouldn't that reduce the potential for missed diagnoses or false positives? No risk of adverse affects unlike biopsies (unless you're allergic to dogs).

Re:Why not multiple dogs?

[BennyB2k4]

That theory doesn't hold if all 10 dogs are missing the same 20% due to a common failure mode.

Re:Why not multiple dogs?

[jblues]

You mean like a Beau-woof cluster?

Re:Why not multiple dogs?

[Firethorn]

Why don't they use 10 dogs and run the sample by each?

Because research funding is limited, and it's difficult to train the proper behavior. They might of tried training 10 dogs, and only Frankie gave good results.

As for missed diagnoses or false positive, you're expecting the tests to be independent(statistics term, means each dog has the same chance to be right/wrong no matter what the other dogs are). They're probably not.

For example, it might be that Frankie is sniffing for a specific set of chemical markers that just aren't present when he ends up with a false negative. The markers might not be strong enough, they might be there tomorrow, there might be some bit of diet that brings it out. Without further research, we just don't know.

So you could have a raging tumor and the dog's won't trigger on it because, for whatever reason, your cancer doesn't give out the necessary 'smell'.

Now, yes, 3-5 dogs 'cooperating' can help moderate other factors, but you still have to be careful.

Re:You have the choice

[fahrbot-bot]

and a life time of being on the Pre-Existing Conditions black list. I don't think the ER covers much of the cancer stuff. If you get layed off at 60 just do some time in jail till you can get on medicare

If you're in the US, you can no longer can be denied medical insurance based on pre-existing conditions nor can your premiums be different because of those conditions - unless, of course, the Republicans succeed in abolishing the ACA.

Woof Woof Rooar Woof

[Registered Coward v2]

What Lassie? Dad has thyroid cancer? I better go call an oncologist.

Re:You have the choice

[fahrbot-bot]

If you're in the US, you can no longer can be denied medical insurance based on pre-existing conditions nor can your premiums be different because of those conditions - unless, of course, the Republicans succeed in abolishing the ACA.

That is simply not true. It is a State matter, and it does differ among the 13 States that created ACA exchanges. I know of at least one state that adopted the ACA except the pre-existing condition inclusion.

You are flatly incorrect. The ACA is a Federal Law and the only thing the states can opt-opt of is the Medicaid expansion - as per the Supreme Court ruling.

Perhaps you're thinking of the Pre Existing Condition Insurance Plan which was a *temporary* measure (that states could choose to participate in) that expired in 2014:

The temporary program covers a broad range of health benefits and is designed as a bridge for people with pre-existing conditions who cannot obtain health insurance coverage in today’s private insurance market.

In 2014, all Americans – regardless of their health status – will have access to affordable coverage either through their employer or through Health Insurance Marketplaces, and insurers will be prohibited from charging more or denying coverage to anyone based on the state of their health.

Re:what are we missing?

[forty-2]

Because you can't cram them into the body tricorder and they don't always make the right noise when you waive them over patients.

Missing important info

[aepervius]

The specificity of the test is not shown, at least the false positive is not. ~85% chance of true positive is interresting and warrant further research. But I see conspicuously absent the false positive rate. See 88% true positive does not matter if you get 20% false positive. If you test 200 peoples , 100 with thyroid cancer 100 without, 12 you will not detect the cancer, 88 will be detected and 20 will be false detected as having cancer. And that's a good case where you already have a good idea this could be thyroid cancer with a high probability. As a screening test it would be terrible.

So the question is what is the false negative rate compared to the existing test.