Trained dogs sniff out blood sugar abnormalities in type 1 diabetes

Hypoglycemia can lead to serious consequences, especially in the absence of warning symptoms. A study has now shown that specially trained dogs are able to detect hypoglycemia and hyperglycemia in their owners. However, reliability varies considerably from animal to animal.

Reliability of detection may vary from animal to animal

Hypoglycemia is a frequent complication of intensified insulin therapy, which can lead to serious consequences, especially in the absence of warning symptoms. A study has now shown that specially trained dogs are able to detect hypoglycemia and hyperglycemia in their owners and to sound the alarm if necessary. However, reliability varies considerably from animal to animal.

About a quarter of type 1 diabetics do not develop early warning symptoms of hypoglycemia. They are therefore particularly at risk under insulin therapy - the risk of particularly severe hypoglycemia with serious neurological and cardiovascular consequences is increased by a factor of six to eight. In addition, there is the risk that patients may reduce the dose of insulin, which in turn increases the risk of hyperglycemia with its negative consequences.

Specially trained diabetes warning dogs are on the rise

The Medical Detection Dogs Foundation, based in Milton Keynes, England, specializes, among other things, in training so-called "diabetes warning dogs" who are trained to recognize hypo- and hyperglycemia of their type 1 diabetic owners by their specific odor, and sound the alarm accordingly. In individual cases, this works well, but the question remains whether the patients can really rely on it.

Veterinarian Nicola Rooney from the University of Bristol and colleagues wanted to investigate this in a study. 27 diabetes warning dogs trained by the Medical Detection Dogs Foundation were used. To qualify for participation, animals had to be able to detect 75 percent of hypoglycemia over a three-month period and produce less than 15 percent false alarms.

Subsequently, the patients were asked to determine their blood sugar over 6 to 12 weeks whenever their dogs struck. Of course, the patients should also perform their normal measurements. In total, more than 4,000 hypo- and hyperglycaemic events were evaluated (between three and 411 in individual patients). 

Over 80 percent of hypoglycemia was "sniffed"

The hypoglycaemic episodes were detected with a sensitivity of 83 percent (66 to 94 percent), the hyperglycaemic episodes with a sensitivity of 67 percent (17 to 91 percent). Now and then there was also a false alarm with normal blood sugar values. The positive predictive value - i.e. the proportion of warnings for which blood glucose was actually outside the target range - was 81 percent on average.

As expected in living organisms, not all study dogs were equally good at sniffing out glucose abnormalities. Four of the 27 dogs actually only alerted when there was a blood sugar derailment - they thus achieved a positive predictive value of 100 percent. 

Reliability not the same for all dogs

Otherwise, the researchers were able to identify some factors that influenced the reliability of the dogs. For example, the diabetes warning dogs were better able to sniff out hypoglycemia than hyperglycemia immediately after completing the previous training. If the training had occurred long ago, the animals reacted better to hyperglycemia while the reaction to hypoglycemia decreased somewhat. Also, the willingness of the owners to reward the dogs and confidence in the ability of the four-legged companions played a role. Last but not least, there were also differences in character: Some dogs were able to do their job reliably and obviously with joy and motivation, others seemed to like to try a new behavior and did not react to the trained alarm situations in between.

Of course, this study also has some weaknesses, the authors write. Some hypo- and hyperglycemia may not have been detected because there was no continuous glucose measurement. Also, it can not be completely excluded that the patients of some false alarms of their dogs did not report, in order to let their darlings stand in a better light.

Conclusions

Even though diabetes warning dogs in this study achieved some amazing results in detecting hypo- and hyperglycemia, they will always be only a part of diabetes management. The reliability varies greatly from animal to animal and is not sufficient to dispense with regular blood sugar self-management. Nevertheless, they can be a valuable enrichment for the individual patient, which can contribute to a better quality of life. 

Source: 
Nicola J. Rooney et al; How effective are trained dogs at alerting their owners to changes in blood glycaemic levels?: Variations in the performance of glycemia alert dogs; PLOS One (2019) DOI: https://doi.org/10.1371/journal.pone.0210092