NEW ORLEANS -- Dogs may be good at sniffing out some types of cancer, but that skill doesn't appear to transfer to detecting hypoglycemic events, according to research presented here.
www red DiabetologNytt
Diabetic dog alerts had a low sensitivity and positive predictive value (PPV), reported Evan Los, MD, of Oregon Health & Science University in Portland, and colleagues.
When the trained dog issued an alert to a hypoglycemic event in a patient with type 1 diabetes, these alerts were only "timely" -- defined as within 10 minutes before or up to 30 minutes after a diabetic event -- a little over a third of the time (36%).
When comparing dog alerts with continuous glucose monitoring (CGM), of the 30 hypoglycemic events detected by both the CGM and the dog, the CGM alerted to these events by a clinically significant median of 22 minutes ahead of the dog. The CGM detected the threshold for hypoglycemia 73% of the time, they said in a presentation at the American Diabetes Association annual meeting.
In addition, because of inappropriate alerts, the PPV of a dog alert for hypoglycemia was 12%. The false-positive rate was high -- dogs that alerted 16-20 times a week led to a mean false positive rate of 14.5 times per week, the authors reported.
"Published literature regarding diabetes alert dogs is limited to case reports, retrospective surveys, and in vitro evaluation of dogs not reflecting real-life use," Los told. "Still, patients are utilizing them and clinicians are without much guidance when asked the question: 'What about diabetes alert dogs?'"
Diabetes alert dogs undergo rigorous training, starting with obedience and socialization, as well as scent training. For the latter, they are trained based on a cotton swab of sweat from a human companion during a hypoglycemic event. The training takes 6-24 months to complete, but there is no universal competence test for dogs.
This is the first controlled study of the reliability of diabetes alert dogs to hypoglycemia in their diabetic companions under real-life conditions, Los said. His group examined eight patients (the youngest was 4-years-old) who had both a diabetes alert dog and a blinded CGM. Dog alerts were recorded in a diary and those were compared with capillary blood glucose (CBG) and CGM downloads.
Hypoglycemia was defined as CBG and/or CGM <70 mg/dL. The patients reported impressions of dog reliability and reasons for obtaining a diabetes alert dog.
While the dogs' performance compared with CGM was lackluster, their human companions reported being happy with their performance. Dog owners self-reported as very satisfied (8.9 of 10 on a Likert scale) and largely confident (7.9 of 10) in their dog's ability to detect hypoglycemia. The latter was cited by the study participants as the reason for having a diabetes alert dog.
The rate of dog alerts during hypoglycemia were 3.2 times greater (95% CI 2.0-5.2) than during euglycemia. However, the first sign of hypoglycemia was the CGM (70%), followed by the dog (19%), followed by a patient's symptoms (12%).
Study limitations included the small sample size, short duration of the study, and the fact that the dogs were different breeds and different ages, and from different trainers. The most reliable dog in the sample had completed 24 months of dog training, suggesting that dog skills diminish over time and may require re-training.
Los said that if someone has an alert dog and finds it beneficial, he wouldn't tell them to get rid of the animal because it is identifying more hypoglycemia than not having a dog at all.
"This is not the final word on whether trained dogs might be helpful for patients with diabetes," Los told. "There may be other benefits not assessed by this study that are important such as having a positive partner in the daily management of a chronic disease."
This study was supported by the Helmsley Trust.
Los disclosed no relevant relationships with industry.