About 500 people have sent me this article over the past week, so I guess I should get around to making some comments. The article entitled "Using a dog’s superior olfactory sensitivity to identify Clostridium difficile in stools and patients: proof of principle study" (Bomers et al 2012) is in the Christmas edition of BMJ, an edition in which they typically publish something fun or light. The study describes the use of a dog trained to sniff out C. difficile, an important cause of diarrhea people, especially hospital patients. The premise is that dogs could be used as a rapid and cheap way to diagnose C. difficile infection, and therefore allow for earlier treatment and implementation of enhanced infection control measures.
There were 3 components of the study:
1) Detection of C. difficile on a culture plate
The dog was trained using culture plates containing C. difficile. That’s how they have to start, but detecting C. diff on a culture plates is pretty easy. I can do that, and my nose is nowhere near as good as a dog’s. Clostridium difficile has a very characteristic odour on a culture plate and odour is one of the methods that’s commonly used to determine whether C. difficile might be present on the plate.
2) Detection of C. difficile in stool
For this, the researchers set the bar pretty low. A positive fecal sample was considered one that was culture-positive positive on a test to detect the C. diff toxins. We know toxin tests are pretty insensitive (they give a lot of false-negatives), which is why there’s a major movement to replace them with molecular tests. By requiring the sample be positive on both culture and toxin test, it means that the samples had to have been quite positive to be considered (i.e. they didn’t test the dog with "weaker" positive samples that may have had less C. diff and C. diff toxin in them). That weakens the results a bit, but they’re still interesting.
They presented the dog with 50 positive and 50 negative samples. The dog gave a positive response to all 50 positive and a negative response to 47/50 negative samples.
3) Detection of C. difficile in patients
Here’s where it gets more interesting and potentially more relevant, since the real value in a sniffer dog would be to detect C. diff directly from patients, as a rapid and cheap screening tool.
For this part, they enrolled 30 patients with C. difficile infection and 270 controls. One problem I have is that 94% of their controls were non-diarrheic. It raises questions about whether the dog is detecting C. difficile or just diarrhea, since the groups don’t just differ by their C. diff status, as would be most appropriate for a control group. The more differences there are between the groups, the greater the potential that a difference other than the one of interest (i.e. C. diff status) is actually the thing that’s being detected. There’ a big difference between a dog that can detect C. difficile and a dog that just detects diarrhea.
Another issue is that some C. difficile strains don’t produce toxins and are not able to cause disease, but they’d presumably be detected the in the same way based on odour, in contrast to tests that are based on detection of the bacterial toxin or genes that encode toxin production.
Anyway, the dog correctly identified 25/30 (83%) cases and 265 (98%) controls. Not as good as current molecular tests but pretty remarkable, nonetheless.
Overall, it’s an interesting story and shows how good a dog’s nose can be, how smart (some) dogs are, and how thinking outside the box can result in some interesting ideas. Though I don’t think diagnostic testing companies have much to worry about at this time in terms of competition from sniffer dogs.
Cool concept. Fun paper. Not coming to a hospital near you in the near future, but not something to completely dismiss.