Abstract
Objective
There is a growing market for diabetes alert dogs but little published regarding their ability to reliably detect hypoglycemia. We aimed to determine whether two dogs could detect hypoglycemic breath samples from people with type 1 diabetes (T1D) and then transfer detection to novel hypoglycemic breath samples.
Methods
Breath samples were collected from individuals with T1D during times of normo-, hypo- and hyperglycemia. Two dogs, previously trained (3 alternative forced choice) with breath samples from three different individuals with T1D, were presented with three breath samples from the same individual: one hypoglycemic, one normoglycemic, and one hyperglycemic, and trained to identify the hypoglycemic sample using a Yes/No procedure. The dogs’ ability to transfer detection was then tested by presenting them with a novel sample set from the same individual. Then we tested whether one dog could transfer detection of the odour of hypoglycemia by presenting new samples from a different individual.
Results
One dog was able to transfer detection of the odour of hypoglycemia to samples from the same individual (Specificity 89%; Sensitivity 62%), but a second dog was not. Results were inconclusive regarding the ability of one dog to transfer detection of the odour of hypoglycemia across two individuals.
Conclusions
The results suggest that some dogs can be trained to detect hypoglycemic breath of an individual with T1D but detection may not transfer to novel samples from other individuals. Results should be interpreted cautiously since the dogs were trained with a small number of breath samples before testing.
There is a growing market for diabetes alert dogs but little published regarding their ability to reliably detect hypoglycemia. We aimed to determine whether two dogs could detect hypoglycemic breath samples from people with type 1 diabetes (T1D) and then transfer detection to novel hypoglycemic breath samples.
Methods
Breath samples were collected from individuals with T1D during times of normo-, hypo- and hyperglycemia. Two dogs, previously trained (3 alternative forced choice) with breath samples from three different individuals with T1D, were presented with three breath samples from the same individual: one hypoglycemic, one normoglycemic, and one hyperglycemic, and trained to identify the hypoglycemic sample using a Yes/No procedure. The dogs’ ability to transfer detection was then tested by presenting them with a novel sample set from the same individual. Then we tested whether one dog could transfer detection of the odour of hypoglycemia by presenting new samples from a different individual.
Results
One dog was able to transfer detection of the odour of hypoglycemia to samples from the same individual (Specificity 89%; Sensitivity 62%), but a second dog was not. Results were inconclusive regarding the ability of one dog to transfer detection of the odour of hypoglycemia across two individuals.
Conclusions
The results suggest that some dogs can be trained to detect hypoglycemic breath of an individual with T1D but detection may not transfer to novel samples from other individuals. Results should be interpreted cautiously since the dogs were trained with a small number of breath samples before testing.
| Original language | English |
|---|---|
| Journal | Canadian Journal of Diabetes |
| Early online date | 08 May 2019 |
| DOIs | |
| Publication status | Early online date - 08 May 2019 |
Access to Document
- Reeve et al. (2019) An idiographic investigation of diabetic alert dogs' ability to learn from a small sample set
© 2019 Canadian Diabetes Association. This work is made available online in accordance with the publisher’s policies. Please refer to any applicable terms of use of the publisher.