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

Abstract

To test the accuracy of Diabetes Alert Dogs (DADs) by comparing recorded alerts to continuous glucose monitoring (CGM) device readings during waking and sleeping hours. 14 individuals (7 adults with type 1 diabetes and 7 youth with type 1 diabetes/parents) who owned DADs for ≥6 mos wore masked CGM devices over a several-week period while recording DAD alerts electronically and in paper diaries. During waking hours, sensitivity scores across participants were 35.9% for low BG events and 26.2% for high BG events. DAD accuracy was highly variable with 3/14 individual dogs performing statistically higher than chance. Sensitivity scores were lower during sleep hours of the person with diabetes (22.2% for low BG events and 8.4% for high BG events). DAD accuracy during sleeping hours was also highly variable, with 1/11 individual dogs performing statistically better than chance. Rate of change analyses indicated that DADs were responding to absolute BG level, rather than rapid shifts in glucose levels. In this study the majority of DADs did not demonstrate accurate detection of low and high BG events. However, performance varied greatly across DADs and additional studies are needed to examine factors contributing to this variability. Additionally, more research is needed to investigate the significant gap between the positive experiences and clinical outcomes reported by DAD owners and the mixed research findings on DAD accuracy.