Sunday, May 6, 2012

A Boy and His Rat

It seems like every few months we’re treated to another heart-warming story about scent-detecting service dogs. Lately it’s been dogs for diabetics. You know the drill: Fido alerts one way when Johnnie’s blood sugar is too low, and another way when it’s too high.

The latest entry is Ben Ownby, a middle-schooler from San Antonio, Texas, and his Labradoodle Dakota. Before them it was another Texan, Katie Jane Brashier, and her Labrador Retriever mix named Shots.

This is the adorable public face of the scientific enthusiasm for training dogs to detect various human diseases by odor. I was skeptical of it in What the Nose Knows:
So, yes, dogs can smell odors associated with bladder-cancer. But this is a far cry from “What’s that Lassie? Timmy has bladder cancer?” To make use of this canine talent, your local hospital would have to maintain a half dozen dogs and their trainers, supply copious medically-certified human urine samples, and provide ongoing statistical support and chemical analysis. At which point 6 out of 10 bladder cancers would go undetected.
Recently published studies report more impressive success rates. A French group used a Belgian Malinois shepard to sniff urine sample from patients with prostate cancer. The dog correctly chose cancer samples in 30 of 33 cases, for a detection sensitivity of 91%. Detection specificity was also 91%. (In other words the dog correctly identified 91% of cancer samples, and correctly rejected 91% of non-cancer samples.)

This is a pretty impressive performance, but keep in mind that it came after 16 months of training, and that “training was a full-time job for the team, who worked with the dog 5 d/wk over the study period.” No wonder the authors caution that
the present work is a proof-of-principle study, and the use of these dogs is not supposed to be generalized. We tested a limited number of subjects in a costly, long study that makes it difficult to conceive of an extended use for this test in clinical practice.
Meanwhile, a German research group claims 71% sensitivity and 93% specificity for dogs trained to detect odors associated with lung cancer.

If you’re going to use animals for routine disease detection in the clinic, you need a species that is less expensive and less demanding that the dog. And now, thanks to a multi-national research team led by Georgies F. Mgode, we have a candidate—a rodent, less. Allow us to present Cricetomys gambianus, the African giant pouched rat. (It weighs in slightly more than one of Paris Hilton’s dogs.)

Mgode et al. have previously trained the giant rats to detect odor associated with tuberculosis. Now they use lab culture samples to ask whether the animals can discriminate TB samples from non-TB bacterial samples using artificial laboratory cultures. Which brings us to this classic line from their methods section:
Negative sputum samples from TB clinics in Dar es Salaam, Tanzanai, were used for spiking test microorganisms.
Mmmmmm . . . make mine a Rhodococcus!

Having obtained the necessary sputum, the next step was to train the animals.
during training sessions, rats were rewarded with food (mashed banana mixed with crushed commercial rat food) when they paused for 5 s at known TB-positive sputum samples. They did not receive food for pausing at known TB-negative samples. With extensive training the rats learnt to consistently pause at TB-positive samples but not at TB-negative samples.
Here’s a photo of co-author Maureen Jubitana, two trainers, and the study animals.

Mgode and colleagues tested the trained rats on sputum samples containing either TB culture or cultures of other, nontuberculosis mycobacteria. The rats responded only to the TB-spiked samples, indicating that there is an odor profile specific to TB, and not to other pulmonary bacterial infections. Interestingly, the detection rate for the TB-spiked sputum samples was lower than that for naturally occurring TB-positive sputum. And the spiked samples were more detectable when the spiking dose was taken from certain growth phases of the bacterial culture. Evidently, the exact character and intensity of the TB-associated scent depends on the biological context and growing condition of the bacteria.

Dogs, African giant pouched rats, whatever. What I want to know is, why don’t any of the labs just straight-out run a sniff test with human odor judges?

The articles discussed here are “Olfactory detection of prostate cancer by dogs sniffing urine: a step forward in early diagnosis,” by Jean-Nicolas Cornu, Géraldine Cancel-Tassin, Valérie Ondet, Caroline Girardet, and Olivier Cussenot, which appeared in European Urology 59:197-201, 2011, and “Ability of Cricetomys rats to detect Mycobacterium tuberculosis and discriminate it from other microorganisms,” by Georgies F. Mgode, Bart J. Weetjens, Christophe Cox, Maureen Jubitana, Robert S. Machang’u, Doris Lazar, January Weiner, Jean-Pierre Van Geertruyden, and Stefan H.E. Kaufmann, which appeared in Tuberculosis (Edinb), 92:182-186, 2012.


EdC said...

Has anyone ever seriously tried to train doctors to make these detections? You've mentioned many times that human scent detection is only quantitatively - not qualitatively - less than that of dogs. And I've heard that doctors typically do recognize a smell of impending death in very ill patients.

Avery Gilbert said...


I recall once reading about someone creating a sniff kit for MDs to learn the typical odor of certain diseases. Sort of Le Nez du Vin for physicians. And from time to time there are retrospective columns in medical journals about "the old days" of diagnosis by smell. But I'm not aware of any active efforts currently underway.

But it sure looks like the diabetes dog market is a hot one.