16

Many people have claimed that dogs can be trained to detect hypoglycemia.

For example:

For example, we have discovered that dogs have the ability to smell chemical changes in our bodies when someone's blood sugar starts to get low. We can make the most of this amazing ability with special training designed to forewarn the onset of hypoglycemia crisis. With a simple but telling nudge, a trained dog alerts their person to a low blood sugar attack that they did not know was coming. A quick test, a little sugar, and everyone can go on with their day, no crisis, no emergency, just the wonderful feeling of security and independence.

Is there any evidence of this?

Oddthinking
  • 140,378
  • 46
  • 548
  • 638
Cthulhu
  • 263
  • 1
  • 6
  • Closely related question [regarding cancer](http://skeptics.stackexchange.com/questions/23556/can-dogs-detect-human-cancer) – Oddthinking May 05 '15 at 12:42
  • 1
    Seeing how _humans_ can easily smell a significant hypoglycemia and dogs have an olfactory sense that is a few thousand times better, why shouldn't that be possible? – Damon May 05 '15 at 14:24
  • 6
    @Damon humans can smell hypoglycemia? – Himarm May 05 '15 at 14:27
  • Yup, the famous ketoacidotic smell. Depending on your individual olfactory sense, it may be that you only smell it when someone is in coma already, or a long time before that. But dogs have _much_ better sense, so... seems entirely plausible for them to smell it early. – Damon May 05 '15 at 14:39
  • @Damon, I can tell you that someone on a low-carb diet may have the infamous "keto breath", way before they get comatose; that dog would become very nervous around such a person! (the more you know) – sleblanc May 05 '15 at 17:06
  • @Damon: Wikipedia says ketoacidosis is associated with hyperglycemia, which is the opposite direction from hypoglycemia. http://en.wikipedia.org/wiki/Ketoacidosis – Mooing Duck May 05 '15 at 18:13
  • @MooingDuck: someone will burn fat whenever their insulin level is low, producing ketone bodies. Usually, insulin level is closely related to glucose level, but in the case of a diabetic person (type I), their insulin response lags way behind their current blood glucose level (sometimes it cannot even catch up); they are burning fat instead of the excess sugar, the reason behind their hyperglycemia. It also explains why they need injections of insulin. – sleblanc May 05 '15 at 19:34
  • @sebleblanc: According to http://en.wikipedia.org/wiki/Hypoglycemia, Hypoglycemia is usually caused by _exessive_ insulin in the body, which seems like it would reduce ketone if anything. You appear to be talking about hyperglycemia, which is totally different. On the other hand, if it _does_ reduce Ketone bodies, dogs could potentially detect that. – Mooing Duck May 05 '15 at 20:25
  • I was not talking about hyperglycemia: I was talking about ketoacidosis. Ketoacidosis is often strongly linked to hypoglycemia, but in type I diabetes, *it's more often the opposite*, as ketoacidosis is only a *symptom* of hypoglycemia. A subject suffering from type I diabetes will show symptoms of hypoglycemia when in reality they are in a state of hyperglycemia. (note: at this moment, giving sugar to that person is the worst thing you could do) – sleblanc May 06 '15 at 15:50

2 Answers2

17

It is difficult to prove that it is impossible to train dogs to detect hypoglycemia. All that can be done to disprove this claim is to show systematic efforts to train dogs have failed, and that no-one has been able to demonstrate that it is possible.

In 2013, such an attempt was tried:

Note: This was just a letter, and probably did not pass a full peer-review.

They took three dogs that had already been trained by an organisation, and whose owners and trainers believed were capable of detecting hypoglycemic skin swabs.

Trained dogs were largely unable to identify skin swabs obtained from hypoglycemic T1D subjects. [...] To our knowledge, this is the first controlled study to address whether dogs can detect a hypoglycemic scent, though there are anecdotal and case reports suggesting that dogs can respond to hypoglycemia (2–4). Our results addressed only whether there is a detectable hypoglycemia scent on the skin. In future studies, it may be helpful to include behavioral elements, such as studies in the presence of human companions. It might also be helpful to obtain swabs from the usual human companions of the dogs. We found that trained dogs were unable to correctly identify skin swabs obtained during hypoglycemia in subjects with T1D. Further studies are needed to address the role of other factors that the animals might use, such as behavioral cues.

Given the lack of evidence, so far, it seems that trained dogs are not to be trusted. This provisional position might change if people can find other cues that trigger the dogs, or other ways of training dogs.

Given that blood glucose meters, while arguably not as cute as a dog, are fairly cheap and fairly accurate, it seems dog trainers will have a large hurdle to train dogs to be sufficiently sensitive and specific to warrant their use.

Oddthinking
  • 140,378
  • 46
  • 548
  • 638
  • its weird that their are many sites that actively sell/ train these dogs for use, http://www.eenp.org/main/diabetic as example. – Himarm May 05 '15 at 13:07
  • 4
    @Himarm: This [blog article](http://www.healthcentral.com/diabetes/c/110/162286/diabetes-detect-hypoglycemic/) monitors one such site that provided dogs for the experiment, and noted that their claims did not change in the face of evidence. Would I be verging into cynicism if I wondered aloud if claims about dogs are covered by the FDA? – Oddthinking May 05 '15 at 13:13
  • 1
    heres some earlier research, suggesting that dogs do have the ability to detect, but that the way in which they detect hypoglycemia is still unknown. http://www.ncbi.nlm.nih.gov/pubmed/19040375 – Himarm May 05 '15 at 13:27
  • 4
    Re @Himarm, that's exactly what I was going to suggest. The question says 'detect', you are answering 'smell'. Obviously dogs are mostly know for their sense of smell, but maybe there are other subtle triggers that they notice? – Benjol May 05 '15 at 13:29
  • 2
    For future reference: https://clinicaltrials.gov/ct2/show/NCT02126605 – Benjol May 05 '15 at 13:33
  • 1
    @Benjol i believe i found some results from that study, http://integrateddiabetes.com/attd-2015-the-latest-research-reports-in-diabetes-technology/ "18 DAD owners, all of whom obtained a DAD from the same training organization in Virginia, participated in the study. Individual DADs varied greatly in accuracy. Overall, low BG (<90 mg/dl) was caught by DADs 66% of the time and high BG (>200 mg/dl) was caught 52% of the time. Half of DADs caught highs and lows more than 65% of the time." – Himarm May 05 '15 at 13:41
  • overall this is a very complex issue with the actual hypoglycemia, my wife is a type one diabetic and she can get low or high in a few different ways, typically her body just slowly gets either lower or higher depending on what she ate insilin usage, and this is a gradual lowering over an hour or 2, however she has spiked low in a mater of 15-30 min, and the symptoms still may take the usual amount of time to show, so its a question too with the dogs, can they detect the spikes, or the gradual. – Himarm May 05 '15 at 13:48
  • 1
    Not a good test: only 3 dogs; only smelling a skin swab; and of patients who the dogs didn't know. Your conclusion might go too far, too: meters are all very well but you need to be smart enough to use it ... and I'm not saying people aren't smart, and don't already use their meters often, but one of the symptoms of hypoglycemia (if it happens) is confusion etc. – ChrisW May 05 '15 at 13:51
  • @Benjol: The trainers claimed they had trained on smell, and that they had had success. Once there is evidence against that, arguing it is behavioural is moving the goal posts. Nonetheless, I quoted the section of the letter that said behavioral cues would require further studies. – Oddthinking May 05 '15 at 13:53
  • 1
    @ChrisW: I carefully don't argue it can't be done. I argue that the closest thing we have to serious evidence says "Nope. The claimants can't do what they said they could do." The required size of the study depends on the magnitude of the claim. – Oddthinking May 05 '15 at 13:55
  • The issue of confusion applies to dogs as well as meters. – Oddthinking May 05 '15 at 13:56
  • @Himarm: J. Altern Complement Med? Survey about the past? Reminds me too much of [this](http://skeptics.stackexchange.com/questions/2578/do-dogs-know-when-their-owner-is-coming-home-through-some-form-of-telepathic-lin) to take seriously. – Oddthinking May 05 '15 at 14:00
  • Here's an interesting article stating that dogs actually react to hypoglycemic episodes: http://online.liebertpub.com/doi/abs/10.1089/acm.2008.0288 But, as they conclude, `Further research is now needed to elucidate the mechanism(s) that dogs use to perform this feat.` – Cthulhu May 05 '15 at 14:02
  • 1
    @Cthulhu: Yes, that's the first article that Himarm found above, and that I dismissed. They didn't watch the dogs as part of the study - they asked people to remember how their dogs behaved. That's unreliable, and their conclusion is too strong. – Oddthinking May 05 '15 at 14:05
  • 1
    @Himarm: The integrateddiabetes summary has too little information to tell. It describes a low sensitivity but doesn't mention specificity. We can't tell if it is any different to random chance with the data provided. – Oddthinking May 05 '15 at 14:07
  • This may be a bit cheeky, but http://area51.stackexchange.com/proposals/76206/living-with-diabetes – Benjol May 06 '15 at 04:31
  • I'm working on a new version of this answer. – Oddthinking May 06 '15 at 05:20
  • @Oddthinking It's not moving the goal posts to dismiss smell and ask the actual question posed: *"Is it possible to train dogs to detect hypoglycemia?"* with the body *"Many people have claimed that dogs can be trained to detect hypoglycemia... Is there any evidence of this?"* Yes, a specific example was provided, but the question is larger than the example of a smell claim. – Adam Davis Sep 15 '17 at 10:18
3

Is it possible to train dogs to detect hypoglycemia?

Two studies suggest that dogs can detect hypoglycemia. The mechanism whereby they detect this state is unknown - the example linking scent, for instance, may be a baseless supposition for marketing purposes.

In the 2008 paper Canine responses to hypoglycemia in patients with type 1 diabetes the conclusion states, "The findings suggest that behavioral reactions to hypoglycemic episodes in pet owners with type 1 diabetes commonly occur in untrained dogs. Further research is now needed to elucidate the mechanism(s) that dogs use to perform this feat."

Note that the training component of your question was not evaluated in this study, but a later 2013 paper Investigation into the value of trained glycaemia alert dogs to clients with type I diabetes concluded that "Based on owner-reported data we have shown, for the first time, that trained detection dogs perform above chance level." Further it showed that "dogs alerted their owners, with significant, though variable, accuracy at times of low and high blood sugar"

It does appear possible to train dogs to detect hypoglycemia conditions.

Adam Davis
  • 1,460
  • 10
  • 15