I was floored by some recent statements in my local newspaper from one trainer of diabetes alert dogs: "once completely trained, the dog likely could smell a change in blood sugar from up to five miles away" and “a diabetic alert dog could smell a teaspoon of sugar in an Olympic-size swimming pool.” I wrote to his organization, and got an odd reply: “We have a significant number of personal accounts of our dogs alerting from good distances, but it is just that; personal anecdotal accounts. There is no guarantee that all dogs are able to do it and no scientific evidence to support the claims yet.” He sold a dog (for $19,000) to a local family which (according to the story) “will be trained to sense and smell abnormal blood sugar levels, and to alert her and her parents before [the 7-year-old child] experiences symptoms.”
Other trainers are equally enthusiastic: “in most cases the dog prompts the person to test outside of a particular range of values. Unlike what is reported in popular media outlets, a real service dog for Diabetes work can let the individual know whether they are high or low, or about to swing.” (personal communication).
And it’s not just dogs: a cat saved her owner “from the grip of diabetic seizure mere hours after she adopted him from a local animal shelter.” Turns out the cat was new to the household, hence had no training; and the "seizure" was not a true epileptic episode as the woman "came to her senses" -- no mention of whether she was high or low sugar, or for that matter, normal! But anyway, that's a cat, not a dog.
I recently wrote to a number of websites that train diabetes service dogs, and asked if they were aware of any published peer-reviewed studies to support the claims that the dogs are useful: there are none. There are anecdotal reports, mainly in the media, but nothing more. The closest is a survey by researchers at Queen's University Belfast, which found that 65 percent of 212 people with insulin-dependent diabetes reported that when they had a hypoglycemic episode their pets had reacted by whining, barking, licking or some other display.
Forgive me if I am not persuaded.
One dog-training-organization was a fraud pure-and-simple: The Attorney General of the state of Missouri received numerous complaints from consumers, alleging that one company (Heaven Scent Paws, HSP):
I do not mean to imply that any other trainers are frauds, but are they being overenthusiastic? No one seems to agree on what these dogs are sensing; options that have been discussed by pediatric endocrinologists include:
One peds endo commented "I just met my first diabetic service dog. The dog came with the child and parents, and they reported amazing stories of the dog's ability to sense his hypoglycemia and alert them, day and night. He has been trained to recognize his scent from socks removed after documented hypoglycemic events. The dog has no false positives." Let me pour a little cold water on this claim. Although the dog might have no false positives (that is, would never react when the sugar level is normal), that isn’t the problem: the issue really is how often the sugar might go low and the dog would not respond (in medical jargon, a false negative). And the dog was from... HSP!
And if the trainers and physicians don’t know, the public sure doesn’t: there are many on-line breathless stories about families spending huge sums of money for dogs that they are assured will accurately, reliably, and predictably alert the parents to their child’s hypoglycemia.
What do we know for sure that dogs can do?
1) They can occasionally react in various ways when owners with diabetes have abnormally low blood sugar levels -- but that doesn’t mean they react predictably to lows, nor does it mean that expensive training programs can teach dogs to react reliably to lows.
2) They have reduced parental anxiety about hypoglycemia.
3) They (or more correctly, the trainers) have persuaded parents (and occasionally, insurance companies) to part with many thousands of dollars.
What is needed is more science and less amazing anecdotal evidence in news reports. Until a well-designed study is carried out, presumably with the patients wearing continuous glucose monitors (CGM) so glucose levels can be continuously monitored, we won’t know. With CGM, one would easily be able to determine the percentage of lows that caused the animals to react, and whether they sense and react to changes in blood sugar. Another benefit of CGM is that the number of times the dogs react when the sugar is normal and stable (false positives) can be ascertained. Until we know such information from a well-designed and published study, it’s risky to assume that dogs are more reliable than CGM.