Q&A | Howard Wolpert on CGM Research and the Artificial Pancreas
Monday, September 20, 2010
A senior physician in the Joslin adult clinic and assistant professor at Harvard Medical School, Howard Wolpert, M.D., analyzes the best ways to adopt technology for type 1 diabetes management. Here, Dr. Wolpert answers six questions about his lab’s Continuous Glucose Monitoring (CGM) studies and the quest for an artificial pancreas.
What’s an example of your current projects using CGM closed-loop systems?
One is cognitive dysfunction from low blood glucose. When people get low blood glucose, they become cognitively impaired since glucose is the fuel the brain depends on. It’s been known for some time that blood fingerstick readings are not a good measure of people’s brain glucose and their actual cognitive status. The glucose concentration in the interstitial fluid that’s under the skin, which is what the CGM sensors measure, is actually a better measure of what the brain sees.
We’re doing a study in which we take people (initially non-diabetic people), give them an insulin infusion to carefully lower their glucose level, and do cognitive testing. At the same time we’ll measure their interstitial glucose and blood glucose level. The goal is to test our hypothesis that interstitial measurements will be a better measure of brain glucose and cognitive ability.
If so, that will shift the thinking in the field about therapeutic decision-making. We may have to change the nomenclature, to say “glucopenia” (low glucose) rather than “hypoglycemic” (low glucose in the blood).
What else are you studying with closed-loop systems?
When people with diabetes have a high-fat meal, pizza for example, typically the glucoses go up quite dramatically because fat makes you less responsive to insulin. But there are essentially no scientific guidelines on how to adjust insulin to compensate for that.
We are doing a study in which we’ll bring people into the clinical research center and control their dietary intake very closely. We’ll have two situations in which people have dinners with the same carbohydrate and protein content, but one dinner will be low fat and the other dinner will be high fat. We’ll use CGM, an insulin pump and closed-loop software to regulate their insulin delivery.
So we’ll be able to define exactly how much extra insulin people need when they have a high-fat meal. We’ll also know how it needs to be delivered, because we know from experience that it won’t just be one bolus; the insulin will need to be extended out.
We also can examine how results vary among individuals, the role of animal fats versus other kinds of fats, and other research topics.
Overall, we’re aiming to give people much more scientifically based recommendations on how to adjust their insulin for changes in fat, and this could cause a shift in nutrition guidelines. The work also could be helpful in focusing people on the importance of controlling their fat intake as a way of optimizing their glucose control.
This is another instance in which closed-loop research may give us insights for open-loop glucose control as well.
Is your closed-loop system a bit like an artificial pancreas?
Yes, when people talk about an artificial pancreas, that’s essentially what they’re talking about.
Do some artificial-pancreas approaches go beyond insulin administration?
In addition to an insulin pump, several groups are looking at administering glucagon, a hormone the body produces to counteract low blood glucose levels.
There’s also work going on elsewhere about administering pramlintide, which is a peptide that delays stomach emptying. Because one of the issues that one finds with CGM is that people’s glucoses often spike up very rapidly after meals and the insulin action is too slow to right that. In another approach to that problem, other groups are investigating faster-acting insulins.
What steps do you see toward an artificial pancreas?
At the moment, a person actually has to interpret the CGM data to adjust the insulin pump. Their brain is the pancreas that makes that decision. Things are evolving toward software that adjusts insulin delivery based on the CGM data.
But there will be huge regulatory steps required for an artificial pancreas and frankly I don’t think it’s ever going to be fully feasible, because if the system ever goes into overdrive you’ll be in real trouble. The closest one will see is a system where the insulin delivery is suspended if a person has a low blood glucose. That would be protection against low blood glucoses overnight, for example. But you could end up with the opposite problem where the system inappropriately suspends and you get real high glucose levels.
There are other challenges as well. One issue, for example, with mealtimes is that it takes too long for the increase in glucose to register by the sensor. And when you’re delivering insulin subcutaneously with a pump, the delivery is much slower than with the pancreas. So there’s a delay in sensing the increase in glucose and a delay in delivering the insulin to cover that. The end result is that the system can’t turn on and off quickly enough.
Also, how would the system know you just ate two pieces of pizza or jogged for an hour?
Absolutely, you’ll have to provide some input. At best what you’re looking at is a hybrid system.
But while the pace of developing an artificial pancreas is slower than people want, it’s still very important to take advantage of today’s incremental advances in diabetes management such as CGM.
The risk of long-term complications for people with type 1 diabetes can be quite different than it was in previous eras. A follow-up to the Diabetes Control and Complications Trial published last year, for instance, showed that only about 1% of the 1,441 participants required kidney dialysis or transplantation. We can substantially reduce the risk of complications with best use of today’s diabetes tools.