I've written previously about my skepticism that dogs can accurately warn humans that blood sugar levels are low (Diabetes Alert Dogs Are Still an Unproven Concept). Now there's scientific evidence, from a medical and veterinary school collaborating with DAD trainers, about whether these dogs can do what is claimed that they are able to do: reliably detect a scent that reflects hypoglycemia.
On August 24, 2012, there was a posting at Facebook from a representative of a dog-training organization as follows:
"On Saturday, I will be conducting a study with OHSU (Oregon Health Sciences University) where we will investigate the dog's ability to smell the metabolic change in a diabetic's body. It is amazing to have some of the dogs I trained in a clinical research study Send me and my dogs some good energy Saturday morning!"
There has been no followup at Facebook, nor at the organization's website, about the outcome of the study, but the July 2013 issue of Diabetes Care describes the study and its results, in a letter titled Can Trained Dogs Detect a Hypoglycemic Scent in Patients With Type 1 Diabetes?
The authors included (among others) a veterinarian, several physicians, and trainers from a dog-training organization. The three dogs in the study had previously been trained by the organization, and the procedure used to test the dogs' abilities "was chosen because the dog-training organization affiliated with one of the authors used this method to train dogs to respond to hypoglycemia in their human companions. The three adult dogs used in this study had been trained to respond to hypoglycemia by pressing a bell after sniffing the open-capped container with the hypoglycemic swab. Each of these dogs had been placed in the home of a person with T1D. The owners and trainer believed that the dogs chosen for this study were consistently able to detect hypoglycemia in the home."
The results must have been very disappointing to both the dogs' owners and their trainers: the dogs were unable to identify hypoglycemia using the method that the organization had depended upon. The percent of the time that the three dogs could detect low or normal blood glucoses ranged from 50 to 58.3%. Of the 24 samples presented to the dogs (12 were between 47 and 59 mg/dL, and 12 between 101 and 130mg/dL), only two hypoglycemic samples were detected by all three dogs. One normal sample (112mg/dL) caused all three dogs to alert. Several "normal" samples was alerted by two of the three dogs, and several hypoglycemic samples were not detected by two of the three. And these were dogs that both the owners and trainer believed were "consistently able to detect hypoglycemia in the home."
Despite these clearly negative results, the organization's website continues to claim that their dogs can detect both low and high blood sugar levels: they "are trained using positive reinforcement to detect the chemical change a person's body goes through when experiencing a high or a low blood sugar level. It is not known for certain what the dogs are smelling, but the process for training works, nonetheless" Once trained, a diabetic alert service dog is able to detect both a high and a low blood sugar level.
Sorry, but I don't believe this claim. The evidence is now clear that DADs can't detect hypoglycemic vs normoglycemic scent reliably, and if they can detect something else, there's no evidence of what it might be.