Can Trained Dogs Detect a Hypoglycemic Scent in Patients With Type 1 Diabetes?

An Idiographic Investigation of Diabetic Alert Dogs' Ability to Learn From a Small Sample of Breath Samples From People With Type 1 Diabetes

“…The current study, combined with the findings by Dehlinger et al (24) and Hardin et al (25), contributes important findings to the literature pertaining to DADs and has notable implications. Dogs carefully selected and assessed for their motivation, trainability and olfactory acuity (29) can be trained to detect VOCs associated with hypoglycemia from individuals with T1D, but, as our findings illustrate, dogs may have difficulties generalizing the odour of hypoglycemia, suggesting that the collective VOC profile of hypoglycemia may vary enough between individuals that learning to identify a universal "hypoglycemia profile" is a much harder task than identifying 1 sample.…”

“…Although different species require different numbers of exemplars for generalization (33), we are not aware of any research regarding the number of exemplars required for dogs to generalize stimuli (olfactory or otherwise). The complexity of the putative hypoglycemic odour, combined with a small number of training samples, could explain Nutella's performance here and may have also contributed to the poor performance observed by Dehlinger et al (24), during which 3 DADs were unable to detect hypoglycemia in samples provided by individuals that were unknown to them. Furthermore, it is worth noting that, during this experiment, Nutella completed 80 consecutive trials in 1 workday.…”

“…In 2013, Dehlinger et al (24) tested whether 3 previously trained DADs could identify hypoglycemia in sweat samples from 3 individuals unknown to the dogs. They found that none of the dogs could indicate the hypoglycemic samples at above-chance levels.…”

Diabetes Alert Dogs (DADs): An assessment of accuracy and implications

“…In the first study, a total of 12 hypoglycemic and 12 euglycemic samples (taken from three individuals who were not the DADs’ owners) were presented separately to three trained dogs. 3 Results suggested poor DAD accuracy, with all three dogs showing chance level performance with approximately 50% sensitivity and 50% specificity. In contrast, Hardin et al tested six DADs over eight trials using a total of 56 samples (8 of which were hypoglycemic), and found higher sensitivity scores ranging from 50.0% to 87.5% (77.6% across dogs), with all DADs performing above statistical chance levels.…”

“…Recent studies attempting to test DAD ability to detect extreme BG levels have yielded mixed findings. For example, two studies tested DAD accuracy in highly controlled experimental settings using skin swab samples, 3 or skin swab and breath samples, 4 taken from individuals with type 1 diabetes when BG level was hypoglycemic or euglycemic. The hypo- and euglycemic samples, as well as “blank” samples in one study, 4 were placed in containers in a room where the DAD alert test was conducted with experimenters observing DAD behavior from a separate room.…”