The rationale of the study is that people with type 2 diabetes who have a heart attack can have quite poor diabetes control, and this can be quite difficult to lower their glucose safely without causing hypoglycemia. So what we wanted to do is to see whether a modern glucose testing technologies can improve understanding of glucose profile and whether that will help reduce the glucose levels safely in people with type 2 diabetes and recent heart attacks. So this was the main idea of this study. We wanted to make sure that we can reduce the levels to safe levels, and we defined it as time in a range between 3.9 and 10 millimole per liter without increasing the risk of hypoglycemia.
You recruited people who were taking sulfonylureas or insulin. Was this so you would get people who were likely to be experiencing hypoglycemia?
That's correct. So the population studied, we took-- by definition, they needed to be on insulin or sulfonylurea. There were also some minority who were on both. But generally, we recruited people who were on insulin or sulfonylurea. And the reason for that is that these are agents that can cause hypoglycemia, whereas somebody who's on metformin or diet control, they're highly unlikely to have any significant hypoglycemia. So this is why we chose this particular population of patients.
If you're focusing down on this subgroup of patients, how many will that be in clinical practice? What's the scale of the problem?
So if you think about it, the type 2 diabetes patients who have a heart attack, around a fifth will be on insulin. The sulfonylurea use will differ according to which part of the world you're in. But overall, it's around at least 30% of the population of type 2 diabetes. Patients who had a heart attack will be on a sulfonylurea or insulin therapy. So it's still a significant number of individuals.
With your focus being on hypoglycemia, why did you choose to make time in range your primary endpoint, rather than hypoglycemia?
Yeah, I mean, that's a very good question, actually, because what we wanted to do is to see whether we can improve, globally improve glycemic profile in these patients. So what we wanted them to have is to get their glucose in the normal range. So for instance, if you focus on hypoglycemia, you're ignoring high glucose levels, whereas when you take time in range, that will hopefully address both.
And particularly if you're combining it with hypoglycemia, you're getting a pretty global picture of their glucose profile. And the other reason is that the time in range is becoming quite an important measure. We know that it correlates with outcome in people with diabetes. So that was another reason to use time in range as the primary endpoint. Of course, the hypoglycemia was the pre-specified secondary endpoint.
And again, with the focus being on hypoglycemia, why did you opt to use the FreeStyle Libre rather than, say, a continuous glucose monitor, like Dexcom where you could set a low glucose alert.
Yeah, it's very simple. It was about convenience. So the fact that it doesn't need calibrating relatively easy for the patient. And remember, we are dealing with an older age group that can find technology sometimes challenging. And in addition, if we are going to implement this in routine clinical work, this is a cheaper option than anything else that's on the market. So there was sort of a number of reasons why we went for this particular glucose monitoring in terms of, it's convenient, it's quite easy to understand. The health care professionals can deal with it relatively easily, does not require calibration, and then even the costs are very reasonable with this particular device.
What do you feel you've learned about the glucose profiles in this subgroup of patients?
Yeah, so we learned a lot. I mean, there was-- some areas were unexpected, and we need to look into that in a bit more detail. So one was the fact that some people in sulfonylurea were having very significant hypoglycemia that they did not know about.
We did not expect it to be such a problem. The other thing that we noticed-- and again, we need a little bit more data analysis to be absolutely sure-- is that the people who were on sulfonylurea at baseline benefited even more in terms of reduction in hypoglycemia. And I hypothesize that the reason for that is that when you escalate therapy in this group, and you don't have their full glucose profile, they can have a lot more hypos that people don't know about. And this is why they should have bigger-- potentially bigger benefit. As I said, we need to analyze the data in a little bit more detail to be absolutely sure that that's correct. But we should be able to do so.
Can you recap the main study findings, please?
So the main findings of this study are, we have done two different kind of statistical analysis. One is a Bayesian and one is a frequentist analysis. Now, the reason we've done Bayesian is because you get so much data with continuous glucose monitoring that it is important to look at sort of day-to-to variability. Now, we should look at the Bayesian analysis.
What we had is a modest increase in time in range, depending what sort of cut-off that you use for the data. You get a modest 18 to 29 minutes increasing in time in range per day. That actually failed to reach our cut-off -- our posterior probability cut-off of 0.80. It's only reached 0.67.
But the thing is with the Bayesian analysis, you can do what is called a skeptical analysis, enthusiastic analysis, all. So the analysis that we've done is very, very robust. We assume nothing. We did not say that, actually, this intervention will increase or decrease. We did not know. And we had a really wide data distribution. And this is where these numbers came from.
If you learn to be more enthusiastic-- and what I mean by that is that if you have a hypothesis that actually, intervention is going to be positive, which is what we had from our preliminary data, and the data are not going to be very widely distributed, then actually the analysis crosses 0.80 point, the pre-specified point. The take-home message from this Bayesian analysis, because a lot of people not familiar with this, is that no matter what your preconception is, what you use for the analysis, the data are always moving in favor of the intervention, which is very encouraging.
Now, if you do a frequent dialysis, you get the 48-minute increase a day in time in range. But the confidence intervals or the confidence limits were quite wide, and they cross zero. And this is why-- once they cross zero, you can't say significant there increase. So the time in range-- so to cut a long story short, there was a beneficial effect in time in range, but it was modest in this study.
In contrast, when you analyze the hypoglycemia exposure, that was very interesting because you get an 80 minutes, that's 8-0, 80-minute decrease in hypoglycemic exposure, defined as less than 3.9 millimole per liter. Now, all these time points are at three months. When you analyze it at an earlier stage, i.e., at one month, you actually get an increase in time in range of three hours in insulin-treated patients at baseline.
And the confidence interval do not cross that zero, so that looks significant. So in other words, what I can tell you is that if you use flash glucose monitoring in people with recent myocardial infarction type 2 diabetes on insulin or sulfonylurea, you get a decrease in hypoglycemic exposure early and late. You get an increase in time in range early in the insulin users, and you get a modest-- well, I should call it non-significant-- increase in time in range at end of the study.
So in a nutshell, these are the main study findings. Now, in addition to the glucose measurements, you get an improvement in treatment satisfaction-- actually, you get an improvement in both groups, in the control and the flash glucose monitoring. But in the flash glucose monitoring, the improvement is more pronounced. So this is sort of, in a nutshell, what the study showed so far.
What is the way forward after these results? Would you like to see a phase 3 trial?
Yes. So it depends what your question is. So this is a discussion I was having with a cardiologist yesterday. So from the diabetologist point of view, you can see clear-cut glycemic benefits. Yes, the time range, you can argue, is not significant. It's going in the right direction. But you're having a massive reduction in hypoglycemia, and we know that hypoglycemia for people with myocardial infarction and type 2 people, is not good news at all.
So from the diabetologist point of view, you're having a positive effect. From the cardiologist point of view, cardiologist, they want hard endpoint data. So what they want, they want to see, does this reduce mortality? Does it reduce reinfarction? Now, we have looked at major adverse cardiac events, but the numbers are way too small to draw any meaningful conclusions.
And actually, the data's still getting analyzed. When it comes to deaths, we only had three deaths in the control and two in the intervention. So numbers are very small, which is very encouraging, just shows how well we are treating these patients these days. And so this is why, if you talk to a cardiologist, cardiologist would say, yeah, data are interesting, but I want a definitive trial that looks at hard outcome points. And I think that's what we should be doing next if we can get the funding for that.