Whether flesh and blood or plastic and silicon; the cure for late stage diabetics will be a new pancreas. How to build a mechanical version is the question addressed in a recent review by M. Hoshino and colleagues. We refer to a fully automated system as a closed loop system. The goal would be for every thing from glucose sensing to insulin administration to be automated, requiring little if any input from the user.
The first issue is detecting blood glucose and then figuring out what is going on. To do this we need a glucose sensor. At present there are two technologies being considered: enzymatic sensors and optical sensors.
Glucose sensors, used currently in glucose monitoring systems involve an enzymatic reaction. Glucose oxidase breaks down glucose and while doing so, uses up oxygen. It is actually the decrease in oxygen that is being sensed. The reaction products include glucuronic acid and hydrogen peroxide. As hydrogen peroxide is reactive and breaks down glucose oxidase, things are set up for the reaction products to rapidly disperse from the electrode. Continuous glucose monitors (CGM) use this technology to detect glucose levels in the space in between cells known as interstitial fluid. When you give yourself a shot of insulin into the abdominal region you are injecting into the interstitial region. If we think of blood vessels as rivers and creeks then the interstitial fluid is sort of like groundwater. Access to this area is great. All you have to do is stick a small needle containing the electrode into your stomach area and you are in. Interstitial fluid as a real time source for glucose has drawbacks, however, in that blood glucose levels are not always the same as interstitial glucose levels. For example, during aerobic exercise, circulation to the gut area decreases and glucose becomes undetectable as interstitial fluid moves elsewhere. This is why many runners find that their CGM stops working during their run. Also, the time needed for a change in blood glucose levels to affect the concentration of glucose in the interstitial region is simply too long. For example, in one study the lag time in human subjects was measured as 12 minutes. This may not seem like much but for hypoglycemia this is an eternity. Certainly, it would not suffice for meals or for exercise as an accurate predictor of real time insulin needs.
Implanting the sensor such that it samples directly from the blood is a possibility. Indeed a study in dogs found that stable and accurate glucose readings could be obtained for over 100 days. The problems with an intravascular probe as a clinical procedure for potentially millions of people are daunting. Simply considering the potential adverse health consequences such as an inflammatory response, clotting, or internal bleeding is enough to put a halt to this idea.
Optical sensors solve many of these problems but, in turn, generate a host of new problems. The basis of an optical sensor is that light can actually make it through the body (at least the thin parts). By shining a light through the tip of a finger and analyzing the light that emerges on the other side one can get a surprising amount of information. Since the only invasive component is light, there is no immune response or compromise of the blood vessel to contend with. This is good. So what we need is a wave length of light that conveys some unique information about glucose levels in the fingertip. Here we run into the first problem. The best region of the light spectrum for emission and detection is the near infrared (700 – 1300 nanometer wavelength of light). Visible light falls around 400 – 700 nanometers and below that, of course, is the ultraviolet region. Glucose absorbs light in the near infrared so we should be able to detect a change in the relative amount of certain wave lengths of light when glucose levels change. Unfortunately, lots of other things in the body (like hemoglobin) also absorbs in that region of the spectrum. This is not so good. People who are exceedingly clever have gotten around this problem with very complex mathematical models. The models are pretty good. One such model that uses 11 variables was able to accurately measure blood glucose between 2.5 and 27 millimoles per liter. However, these variables are SOOO sensitive that something trivial like a tiny movement of the sensing apparatus on the finger or even a change in the ambient temperature will cause the thing to lose calibration and thus give useless information. Price is also a big issue. The price of this technology will have to come down a lot before it can be used by a large number of people.
So, as far as sensing glucose goes, we are nowhere near “there” yet. Pumps are another story. I’ll talk about pumps next.
[…] Unlike glucose sensors which need quite a bit of work to take over any decision making (see my last post for details), pump technology is ready. Indeed many diabetics, both type 1 and type 2, now use an […]
I agree. That is a really good way to use the sensor. However, if you wanted it to actually control your insulin pump without your imput trouble would ensue very quickly. That being said, I just got done talking with folks at the Barbara Davis Center and getting the sensor technology to talk to insulin pumps is top priority in the clinical division.
It’s disheartening to know there are so many challenges, but I’m grateful that there are so many “exceedingly clever” people working on the problem. I personally find my CGM to be a great comfort as I go about my normal activities — it’s surprisingly accurate compared to finger sticks (I’m on Abbott’s Freestyle Navigator, which I’ve heard is the most accurate of the three available). And when I see a trend arrow heading dramatically in one direction, I take it as a cue to do an actual finger prick. In other words, my CGM gives me a sense of when… Read more »