About 1.5 years ago there was a study done in the UK by Neupane et al [1] that changed how people thought of Diabetic Alert Dogs and their training. (http://care.diabetesjournals.org/content/39/7/e97.full-text.pdf) This study found that the chemical 'isoprene' increases during hypoglycemia, so authors and trainers then claimed that dogs are smelling this particular chemical and alerting based on it. I hadn't done much research into this and it got quite a bit of publicity in the DAD realm so I went with it. However, recently I've been looking into it more and wanted to share my thoughts and skepticism. I'm not saying that I believe the study is incorrect, but I just wanted to share why I am not quite convinced that this is what the dogs are detecting and share my questions in case anyone had answers or thoughts. Some things I address are about the study itself, but most are related to the application of it - the claim that DADs are alerting to isoprene. To clarify, the claim that dogs are alerting to isoprene is not a claim of the study by Neupane et al [1], this is the claim of those in the DAD field and news articles that came about as a result of the publishing of this study.
The first concern is that this study used only 8 female type 1 diabetics. 8 is a very small number of subjects and only one gender brings up some questions as to the generalizability of the finding. Another study, by Smith et al [2], found there is a difference in isoprene levels in relation to gender - in fact males have higher isoprene levels. So the sample was not representative of the population as a whole, and in future studies many more subjects need to be included.
Since isoprene is one of the most common VOCs (Volatile Organic Compound) in exhaled breath, if the dogs are alerting based on isoprene they must be alerting to a certain amount of the isoprene (otherwise they'd be alerting all the time), but this data does not support that.
The data shows that the isoprene in the breath is significantly higher around 2.8 mmol/L (approximately 50 mg/dL). When a subject's glucose is between 3-5 mmol/L (54-90 mg/dL) the isoprene levels are much lower. Many DADs are trained to alert to levels such as 80 mg/dL and below (4.4 mmol/L), when the isoprene levels are actually quite low. So then why are the dogs alerting? Additionally, the isoprene levels are not linear, with isoprene levels increasing as glucose levels decrease, so it is not that the dogs begin to alert at a certain threshold of isoprene.
It appears that the isoprene level increases solely during hypoglycemia at 2.8mmol/L. If the dogs were only alerting during hypoglycemic events such as at 2.8 mmol/L, then it would make sense that they are responding to isoprene. But that's not the case, the dogs can begin alerting at 80mg/dL or 70 mg/dL, or 100mg/dL, depending on what the diabetic chooses to train their dog for. DADs are absolutely not perfect - they miss alerts and sometimes false alert, but a well trained DAD is capable of alerting much more accurately than what the data suggests if they were only alerting to isoprene.
For argument's sake, let's assume that the dogs were capable of alerting to the specific level of isoprene. The issue with this is that the study by Smith et al [2] claims there are other variables that affect isoprene levels such as physical activity and subject gender. In another study [3] Smith et al specifically studied isoprene in relation to age and stated, "There is a clear increase in the mean breath isoprene concentration with age for the young cohort with a doubling of the level about every 5–6 years until it reaches the age-invariant mean level of that for adult cohort." So if the isoprene levels vary between subjects based on their age, activity level, and gender, how could the dogs possibly be accurate in alerting to different people? DADs have to actually be taught to not alert to other people with out of range blood sugars - whatever substance they are alerting to must be consistent among everyone. And if they are alerting accurately to strangers, then the stranger's levels of this unknown substance must be consistent with the dog's handler's levels when the dog alerts. If the quantity of isoprene varies significantly between subjects, then the dogs would not be able to accurately alert to out of range blood sugars to people other than their handler, except in very few cases. Furthermore, trainers that use their own or another diabetic's low blood sugar saliva samples to teach the dogs to alert would have a low success rate when the dog is transitioned to its permanent handler because that person's isoprene levels could be very different from the samples the dog was trained on since the samples would probably have been from a person of a different, age, gender, etc.
My next question regards high blood sugars. Once a dog is taught to alert to low blood sugars, she then alerts to highs completely on her own. If the dogs are only alerting to isoprene, why are they alerting to highs? The data shows that when a subject's glucose is 10 mmol/L, the isoprene levels are not high, nor when it is at 11 mmol/L. So what data would the dog have to prompt them to alert? It's possible that they are alerting to a different scent for a high blood sugar (as we are even capable of smelling extreme high blood sugars) but even then, what causes them to randomly begin alerting to that scent on their own?
Is isoprene found in sweat as well as exhaled breath? DADs often check ankles, knees, and crotches to confirm a low or high blood sugar prior to alerting. A well trained DAD does not jump on its handler and sniff its breath. While dogs have incredible senses of smell and are of course more than capable of smelling a person's breath from the ground, they still are sniffing at these locations where it appears the scent is most concentrated. So is isoprene released in these areas as well?
Lastly, the dogs have not been tested on this chemical yet. The researchers just found that this chemical rises during a hypoglycemic event. No one has ever presented the chemical to the dogs to determine if this is what they in fact alert to or not.
It is a possibility that the dogs are not alerting to one factor in particular, perhaps they are alerting to multiple factors. One difficulty is that it is difficult to determine how much the dogs learn on their own, without us teaching them. For example when I first am training a DAD and they start to understand what levels I want them to alert to, they tend to be quite accurate! But then they begin to alert more frequently when it appears they are not being accurate, but if you analyze the alerts you can see they are alerting to rapid rises and drops now, in addition to specific lows and highs. I don't teach them this, they just do this on their own. Are they tracking my blood sugar and anticipating that they will get a reward when it's rising or falling quickly? Or is there a chemical being released that they have figured out precedes a low or high blood sugar? Or is there a chemical being released during a rapid rise or drop that is actually the same as the chemical released during a low or high?
I am not making a claim about the validity of this study, I am just expressing my personal doubts and questions regarding the study in relation to DADs. Those in the DAD community have taken this study and assumed it to be truth, and I'm not quite as convinced. If you know of studies looking at DADs and isoprene please let me know! And if you can help me answer any of these questions or I have misinterpreted something, please also let me know because I'd love to understand this further! Feel free to comment or email.
There is a lot we do not know about Diabetic Alert Dogs. The dogs are not 100% accurate and I believe that it is not completely due to the fact that dogs are imperfect and not machines. I think we view them as not being 100% accurate because we just don't know what exactly they are alerting to. Perhaps they are much more accurate than we think, but we just don't know how to measure what they are alerting to. Further research needs to be done so that we can utilize the dog's maximum potential. Even though we don't fully understand how they work, these dogs are still very beneficial and huge assets to many type 1 diabetics to help keep them in range more frequently and stay as healthy as possible.
The first concern is that this study used only 8 female type 1 diabetics. 8 is a very small number of subjects and only one gender brings up some questions as to the generalizability of the finding. Another study, by Smith et al [2], found there is a difference in isoprene levels in relation to gender - in fact males have higher isoprene levels. So the sample was not representative of the population as a whole, and in future studies many more subjects need to be included.
Since isoprene is one of the most common VOCs (Volatile Organic Compound) in exhaled breath, if the dogs are alerting based on isoprene they must be alerting to a certain amount of the isoprene (otherwise they'd be alerting all the time), but this data does not support that.
The data shows that the isoprene in the breath is significantly higher around 2.8 mmol/L (approximately 50 mg/dL). When a subject's glucose is between 3-5 mmol/L (54-90 mg/dL) the isoprene levels are much lower. Many DADs are trained to alert to levels such as 80 mg/dL and below (4.4 mmol/L), when the isoprene levels are actually quite low. So then why are the dogs alerting? Additionally, the isoprene levels are not linear, with isoprene levels increasing as glucose levels decrease, so it is not that the dogs begin to alert at a certain threshold of isoprene.
It appears that the isoprene level increases solely during hypoglycemia at 2.8mmol/L. If the dogs were only alerting during hypoglycemic events such as at 2.8 mmol/L, then it would make sense that they are responding to isoprene. But that's not the case, the dogs can begin alerting at 80mg/dL or 70 mg/dL, or 100mg/dL, depending on what the diabetic chooses to train their dog for. DADs are absolutely not perfect - they miss alerts and sometimes false alert, but a well trained DAD is capable of alerting much more accurately than what the data suggests if they were only alerting to isoprene.
For argument's sake, let's assume that the dogs were capable of alerting to the specific level of isoprene. The issue with this is that the study by Smith et al [2] claims there are other variables that affect isoprene levels such as physical activity and subject gender. In another study [3] Smith et al specifically studied isoprene in relation to age and stated, "There is a clear increase in the mean breath isoprene concentration with age for the young cohort with a doubling of the level about every 5–6 years until it reaches the age-invariant mean level of that for adult cohort." So if the isoprene levels vary between subjects based on their age, activity level, and gender, how could the dogs possibly be accurate in alerting to different people? DADs have to actually be taught to not alert to other people with out of range blood sugars - whatever substance they are alerting to must be consistent among everyone. And if they are alerting accurately to strangers, then the stranger's levels of this unknown substance must be consistent with the dog's handler's levels when the dog alerts. If the quantity of isoprene varies significantly between subjects, then the dogs would not be able to accurately alert to out of range blood sugars to people other than their handler, except in very few cases. Furthermore, trainers that use their own or another diabetic's low blood sugar saliva samples to teach the dogs to alert would have a low success rate when the dog is transitioned to its permanent handler because that person's isoprene levels could be very different from the samples the dog was trained on since the samples would probably have been from a person of a different, age, gender, etc.
My next question regards high blood sugars. Once a dog is taught to alert to low blood sugars, she then alerts to highs completely on her own. If the dogs are only alerting to isoprene, why are they alerting to highs? The data shows that when a subject's glucose is 10 mmol/L, the isoprene levels are not high, nor when it is at 11 mmol/L. So what data would the dog have to prompt them to alert? It's possible that they are alerting to a different scent for a high blood sugar (as we are even capable of smelling extreme high blood sugars) but even then, what causes them to randomly begin alerting to that scent on their own?
Is isoprene found in sweat as well as exhaled breath? DADs often check ankles, knees, and crotches to confirm a low or high blood sugar prior to alerting. A well trained DAD does not jump on its handler and sniff its breath. While dogs have incredible senses of smell and are of course more than capable of smelling a person's breath from the ground, they still are sniffing at these locations where it appears the scent is most concentrated. So is isoprene released in these areas as well?
Lastly, the dogs have not been tested on this chemical yet. The researchers just found that this chemical rises during a hypoglycemic event. No one has ever presented the chemical to the dogs to determine if this is what they in fact alert to or not.
It is a possibility that the dogs are not alerting to one factor in particular, perhaps they are alerting to multiple factors. One difficulty is that it is difficult to determine how much the dogs learn on their own, without us teaching them. For example when I first am training a DAD and they start to understand what levels I want them to alert to, they tend to be quite accurate! But then they begin to alert more frequently when it appears they are not being accurate, but if you analyze the alerts you can see they are alerting to rapid rises and drops now, in addition to specific lows and highs. I don't teach them this, they just do this on their own. Are they tracking my blood sugar and anticipating that they will get a reward when it's rising or falling quickly? Or is there a chemical being released that they have figured out precedes a low or high blood sugar? Or is there a chemical being released during a rapid rise or drop that is actually the same as the chemical released during a low or high?
I am not making a claim about the validity of this study, I am just expressing my personal doubts and questions regarding the study in relation to DADs. Those in the DAD community have taken this study and assumed it to be truth, and I'm not quite as convinced. If you know of studies looking at DADs and isoprene please let me know! And if you can help me answer any of these questions or I have misinterpreted something, please also let me know because I'd love to understand this further! Feel free to comment or email.
There is a lot we do not know about Diabetic Alert Dogs. The dogs are not 100% accurate and I believe that it is not completely due to the fact that dogs are imperfect and not machines. I think we view them as not being 100% accurate because we just don't know what exactly they are alerting to. Perhaps they are much more accurate than we think, but we just don't know how to measure what they are alerting to. Further research needs to be done so that we can utilize the dog's maximum potential. Even though we don't fully understand how they work, these dogs are still very beneficial and huge assets to many type 1 diabetics to help keep them in range more frequently and stay as healthy as possible.
Works Cited
1. Neupane, Sankalpa, et al. “Exhaled Breath Isoprene Rises During Hypoglycemia in Type 1 Diabetes.” Diabetes Care, vol. 39, no. 7, 2016, doi:10.2337/dc16-0461.
2. Smith, David, et al. “Can volatile compounds in exhaled breath be used to monitor control in diabetes mellitus?” Journal of Breath Research, vol. 5, no. 2, 2011, p. 022001., doi:10.1088/1752-7155/5/2/022001.
3. Smith, David, et al. “Isoprene Levels in the Exhaled Breath of 200 Healthy Pupils within the Age Range 7 -18 Years Studied Using SIFT-MS.” Journal of Breath Research, vol. 4, no. 1, 2009, p. 017101., doi:10.1088/1752-7155/4/1/017101.
2. Smith, David, et al. “Can volatile compounds in exhaled breath be used to monitor control in diabetes mellitus?” Journal of Breath Research, vol. 5, no. 2, 2011, p. 022001., doi:10.1088/1752-7155/5/2/022001.
3. Smith, David, et al. “Isoprene Levels in the Exhaled Breath of 200 Healthy Pupils within the Age Range 7 -18 Years Studied Using SIFT-MS.” Journal of Breath Research, vol. 4, no. 1, 2009, p. 017101., doi:10.1088/1752-7155/4/1/017101.
