Evaluating and contextualizing volatile organic compounds in dog training breath samples from a patient with type 1 diabetes

Tight control of blood glucose can reduce the risk of severe health consequences of type 1 diabetes. For individuals with impaired awareness of hypoglycemia, glycemic control can be especially challenging. Fear of hypoglycemic events can lead individuals to adjust their glucose management resulting in poor glycemic control. Glycemic alert dogs trained on breath odor have been shown to accurately alert to changes in blood glucose in individuals with type 1 diabetes, providing them with an opportunity to intervene before dangerous fluctuations occur. By analyzing the volatile organic compounds (VOCs) of breath samples used during glycemic alert dog training, the aim of this study is to see if we, too, can detect differences in odor associated with blood glucose in type 1 diabetes. Breath samples from an individual with type 1 diabetes were collected during euglycemia, hypoglycemia, and hyperglycemia and analyzed using thermal desorption gas-chromatography time-of-flight mass-spectrometry. Statistical analysis with permutational multivariate analysis of variance, canonical analysis of principal coordinates, and random forest found significant differences in VOC profiles at different blood glucose concentrations. In addition, factors such as collection material, day of sample collection, and storage length had significant effects on the VOC profiles of samples. These findings show the complexity of the samples glycemic alert dogs are presented with and can direct future sample collection, storage, and training protocols. Despite variations associated with collection methods, differences associated with blood glucose remained detectable, suggesting the odors used by glycemic alert dogs to detect hypoglycemia are preserved in training samples.