Dr. Haffner is Professor of Internal Medicine at the University of Texas Health Science Center in San Antonio.

This presentation will review some of the literature looking at insulin resistance and cardiovascular disease and some mechanisms. It is important to say that the data are not completely definitive, but in middle-aged subjects, at least, there is really quite a strong relationship. The newer studies over the last 5 or 6 years have generally confirmed this.

If we try to make an estimate of the relative importance of the effects of glucose and insulin, let's review what the data look like. These are data from the UKPDS in the BMJ, and it's a set of really nice papers looking at epidemiological models, not the intention-to-treat clinical trial data that Dr. Goldstein presented.

The first of these papers is by Amanda Adler, and this is a paper looking at updated or average systolic blood pressure in relationship to diabetic complications. The yellow squares are myocardial infarctions (MIs), and the blue dots are microvascular endpoints. There are two interesting points here. One is you can see at any given level of systolic blood pressure in new diabetics, there are twice as many MIs as microvascular events, so MIs are relatively more common early on in type 2 diabetes.

But also important here is that the slope of the relationship between systolic blood pressure and MI and microvascular events looks relatively similar.

Let us now contrast that to the average effects of hemoglobin A_1c. Same schemata. If we look from hemoglobin A_1c of about 5.5% to 11%, and this is a paper, again, in the BMJ in August of last year, this time by Irene Stratton, you double the risk of MIs as the hemoglobin A_1c goes from 5.5% to 11%. The first point is hemoglobin A_1c is clearly related to the development of MI. This now has been definitely shown. These data are also adjusted for other cardiovascular risk factors, such as lipids, blood pressure, and so on.

However, if we look at the effect of microvascular complications here, there's a 10-fold increased effect. Thus, the relative strength of the association between hemoglobin A_1c and MIs is relatively weak. You all know how difficult it would be to reduce hemoglobin A_1cs from 11% to 5.5%. Now what we're going to do is explore a little bit more about why this might be.

This is from another paper in the UKPDS 23, a baseline paper by the late Bob Turner. This analysis looks at baseline predictors of cardiovascular disease and the order of entry in the Cox proportional hazard model. In this model, LDL enters first as the strongest predictor, HDL enters second, and hemoglobin A_1c enters third. You can see the clearly significant relationship with systolic blood pressure and smoking.

Our first lesson in this sense is both hemoglobin A_1c and conventional cardiovascular risk factors are very important. Therefore, we are likely to need several drugs if we are going to wipe out this epidemic of cardiovascular disease in patients with diabetes.

Why isn't hemoglobin A_1c at the top of the list? There are a number of hypotheses one can raise. One that has been of recent attention is the possible differential role of postprandial glycemia, or actually postglucose load glycemia.

The hypothesis our group has been most actively investigating is the possibility that there is increased risk of cardiovascular disease prior to the onset of type 2 diabetes. There are actually seven population studies that have looked at this, and they have all found fairly similar results. Now I will show you data mainly from the San Antonio Heart Study. This is from a paper we wrote about 10 years ago in which we looked at people who started with normal glucose tolerance test at baseline and then either later developed type 2 diabetes or remained normal. Those people who later developed type 2 diabetes had higher triglycerides, lower HDLs, higher systolic blood pressures, slightly higher glucose levels but statistically significant, and higher insulin concentrations.

Ten years ago we suggested that hyperinsulinemia and insulin resistance drove the possible increase in risk prior to the onset of type 2 diabetes. The limitation that we did not fully appreciate is that the glucose levels were also slightly higher, and there have been a number of nice reviews and probably the most coherent of it has been presented by Hertzel Gerstein, who wrote an opinion piece in the Lancet on dysglycemia a couple of years ago, suggesting even small increases in glucose may be associated with cardiovascular disease.

The question is, how can we separate out the glucose from the insulin effects? To do that, we looked at another cohort of the San Antonio Heart Study in which we had information both at baseline, and we used a surrogate of insulin resistance, HOMA insulin resistance, though, in nondiabetics, it is extremely highly correlative with fasting insulin. We had a measure of insulin secretory ability, change in insulin to change in glucose levels over the first 30 minutes of an oral glucose tolerance test.

This slide shows predictors of type 2 diabetes. Those individuals who have normal insulin sensitivity at baseline and had good insulin secretion, only about 2% of them developed type 2 diabetes over 7 years, a very low rate. Those who just had low insulin secretion had about a 3-fold increased risk of developing type 2 diabetes. Those individuals who were just insulin resistant had a 5-fold increased risk, and those people who had both defects had a 20-fold increased risk. Both low insulin secretion, increased insulin resistance predicted development of type 2 diabetes. This has been shown by other people as well--Steve Lillioja in the Pima Indians in the New England Journal in 1993.

Our focus was how does this relate to cardiovascular disease? As our first attempt, we tried to characterize - and this is a mixed group of high-risk Mexican-Americans and low-risk non-Hispanic whites - what they looked at metabolically prior to the onset of type 2 diabetes. Fifty-four percent of the 195 prediabetics, people who would later develop type 2 diabetes, were both insulin resistant and had low insulin secretion. Twenty-nine percent were insulin resistant, but had good insulin secretion. In total, about 82% of prediabetics already had insulin resistance, defined by the median in nondiabetics.

A little less than 2% had neither defect. Some of these people later developed type 2. In some cases it cannot be explained, and 16% just had low insulin secretion. We thought they were insulin sensitive.

What was important for what I'm going to tell you is that all four of these groups who eventually developed type 2 diabetes had similar glucose levels but different metabolic abnormalities.

We are going to now review what the cardiovascular risk factors are in the predominantly insulin resistant group, as opposed to the group that's predominantly insulin sensitive.

In this particular figure, we are going to look at triglyceride levels. The two bars on the left are people who have converted to type 2 diabetes. The insulin resistant converters to type 2 diabetes have much higher triglycerides than the insulin sensitive converters to type 2 diabetes, in spite of the fact they have the same glucose levels. They do differ markedly by weight, but if we stratify by weight or adjust for body mass index or waist circumference, very little of this difference is attenuated. In fact, the insulin-sensitive converters have triglycerides that are very close to people who do not develop type 2 diabetes.

A second lesson is not only do prediabetics have increased cardiovascular risk factors as a class, but there is heterogeneity among prediabetics, and it appears to be related to whether they are insulin resistant or not.

The same pattern is evident with HDL cholesterol. Only insulin-resistant prediabetics have low HDL. Insulin-sensitive prediabetics are normal for HDL.

This pattern is also similar for systolic blood pressure. One of the limitations in this analysis is that it is based on cardiovascular risk factors. Until recently, there were no data on cardiovascular events. New data have been presented that is very parallel to this, looking at nonfatal cardiovascular events. Why is it nonfatal? Well, it is because if people die of heart disease, we can't tell if they had diabetes. We showed that, comparing these two groups, this group had about 4-fold the risk of developing nonfatal cardiovascular disease as nondiabetics. Additionally, Frank Hu presented an abstract that shows that in the Nurses' Health Study, prediabetics have an increased risk of cardiovascular disease, almost exactly intermediate between prevalent diabetics and people who never develop type 2 diabetes.

There has been a lot of interest in prevention. We believe that the same sort of interventions to prevent type 2 diabetes - drug treatment and lifestyle changes - are likely to work across a group of ethnic groups.

The way we conceptualized this issue is by a "ticking clock" hypothesis. We suggested that for microvascular complications of type 2 diabetes, for retinopathy or renal disease, their period of increased risk starts at the onset of type 2 diabetes. To prevent diabetic eye and kidney disease, you don't actually have to prevent type 2 diabetes. All you need is early screening, aggressive treatment. But for macrovascular complications of type 2 diabetes, because their risk can start years or even decades earlier, prevention of type 2 diabetes, aggressive treatment, the risk factors and so on, we need a much more multifactorial approach.

We will briefly review a couple of these issues. This is from "Does Insulin Predict Cardiovascular Disease?" or in this case, cardiovascular risk factors. This, again, are data from the San Antonio Heart Study. We are looking at quartiles of fasting insulin. We are comparing quartile 4 to quartile 1. You can see that high insulin levels predict hypertension and high triglycerides. Even stronger, high insulin levels predict low HDL. They do not predict the absolute concentration of LDL cholesterol, but it is related to the composition of LDL with small, dense LDL. And lastly, type 2 diabetes is the strongest of these associations. Also in this paper, we showed that high fasting insulin predicted combinations of these metabolic disorders. In other words, people with higher insulin levels are more likely to have multiple disorders than a single disorder.

Does insulin predict cardiovascular disease? The first thing I should say is that there are very little data on insulin resistance. We just do not have the data, so we are going to use insulin concentrations. This may or may not be the right way to do it.

The second issue I'm going to talk about is almost all of these data have been generated in nondiabetics. It is very hard to do this study in patients with diabetes, both because diabetics have a truncated distribution of insulin resistance and surrogates like HOMA insulin resistance do not work as well in diabetics as they do in nondiabetics.

This is from a review we published a couple of years ago, and you can see that the "Y's" mean there is a relationship, the "N's" say there is not a relationship. The answer is, there is heterogeneity in the results. Most of the results, including newer papers, do show a significant relationship, particularly if they're middle-aged. The elderly subjects are generally negative, though that may reflect a survival bias, whereas people with high insulin levels who live to an old age may be relatively protected from heart disease, and it is no longer important if you are 75 or 80.

Let's look at one of the classic early studies. This was the Finnish Policemen's Study, although this is now the 20-year follow-up, and this was a paper by Pyörälä and colleagues. People in the most insulin sensitive 20% have much better survival than people in the most insulin resistant 20%. There is a nice stepwise relationship that continues for many years. Remember, if you carry the thing out 50 years, they eventually all come together because everybody dies. So it is when you develop your death that is important, not the fact that you eventually die.

Let's look at combinations of some of the cardiovascular risk factors and insulin concentrations. The best study that has looked at this has been the Quebec Cardiovascular Study. This is from one of Jean Pierre Després early papers in the New England Journal of Medicine in 1996. In this particular study, they looked independently at the effect of insulin tertiles, fasting insulin. This is using a specific insulin assay that does not recognize proinsulin, therefore, it is not an issue. The yellow (orange) bars represent people with low levels of plasma apolipoprotein B and the blue bars are people with high plasma apolipoprotein B. You can see that insulin levels predict CHD in people with low apolipoprotein B, predict CHD in people with high apolipoprotein B. But high apolipoprotein B also predicts CHD. This is an informative figure for talking about therapies in diabetic subjects. You might improve insulin sensitivity. You might use lipid-lowering agents toreduce apolipoprotein B levels. The most commonly lipid-lowering therapy used in the United States is statin therapy.

This is another paper from the same group. This appeared in Circulation, by Benoit Lamarche and colleagues in January of 1997. In this case, they looked at the interaction between apolipoprotein B levels, again, low apolipoprotein B in yellow (orange), high apolipoprotein B in pink (blue). On the right are the people with smaller, denser LDL, and presumably more atherogenic, and the people with larger LDL are represented on the left. Again, you see high apolipoprotein B levels predict CHD, and among people with high apolipoprotein B levels or high concentrations, small, dense LDL is an important predictor of cardiovascular disease.

Lastly, I want to talk about one more paper from the Quebec Cardiovascular Study, and this is a paper that some people have referred to as the Nontraditional Cardiovascular Risk Factor paper. This, again, is by Benoit Lamarche and in JAMA in June of 1998. The first slide deals with two issues. One is, how common are these nontraditional risk factors? And what the group described it as, is high levels of fasting insulin, high levels of apolipoprotein B, small, dense LDL. And second, how strongly related they are.

If we look first at the nontraditional risk factors, they range among people with coronary heart disease at 69% to 81%. They are really, really common. The traditional risk factors -- LDL, triglycerides, and HDL -- a little bit less common but still common. You might note the most common of the traditional and nontraditional risk factors in this group, in fact, is high fasting insulin, with 81% of the cases.

If we now look at the odds ratios here to look at the strength of the association, they're all pretty strong as well, but, again, high fasting insulin is the strongest of the predictors of cardiovascular disease in this study.

The next figure now looks at clustering of traditional risk factors--LDL, HDL, triglycerides. You can have from 0 to 3 of these. The blue are the cases; the yellow (orange) are the controls, and on average, controls have one of these risk factors; cases have three. If you look at the ratio of people among the cases with three risk factors to zero risk factors, it's about 6 to 1. It is a quite strong relationship.

Now we will look at the nontraditional risk factors. Here this looks superficially very, very similar. But if you look at the ratio among the cases with people with three risk factors, it's 46%, and no risk factors is only 2%. It is actually 20 to 1. The reason why the nontraditional risk factors are in a sense much better predictors is because very few people who develop ischemic heart disease, at least in Quebec, have low levels of apolipoprotein B, have normal LDL composition, and have normal insulin sensitivity. One study does not make the case. You need more studies. But there are data looking at this, though in some cases not as sophisticated, and show similar results.

One of the other issues that has been troubling, and unfortunately is not resolved, is whether there is a different relationship between insulin resistance and atherosclerosis across different ethnic groups. There is a study called the Insulin Resistance Atherosclerosis Study. It used a frequently sampled intravenous glucose tolerance test, and basically what we showed is the relationship between Hispanics and non-Hispanic whites was relatively similar. It was not much attenuated--this is looking at carotid wall thickness as a primary endpoint by adjusting for conventional risk factors. However, there was no significant relationship in African Americans, and in fact, it was in the opposite direction. We need further confirmation to see whether this is actually true.

I want to talk now a little bit about diabetes. How common, first of all, is insulin resistance in type 2 diabetes? If we look at data from the Insulin Resistance Atherosclerosis Study (IRAS) in the United States, combinations of Mexican Americans, non-Hispanic whites, and African Americans, with an average BMI of about 31, 92% of all these diabetics are insulin resistant. So if you are thinking about whether you should use a thiazolidinedione or not use a thiazolidinedione, is this person insulin resistant? If they are a type 2 diabetic, they have a BMI (body mass index) of greater than 27, you don't have to measure insulin resistance. The chances are overwhelming.

If we, however, compare the insulin resistant to the insulin sensitive diabetics, and there are not a lot, one thing that characterizes insulin resistance among type 2 diabetes is obesity. That clearly is important. This is not shown on the slide. This is based on cardiovascular risk factors. In the IRAS data, in diabetics we don't see a difference in blood pressure. We do, however, see higher fibrinogens in insulin resistant subjects who have diabetes, and much, much higher PAI-1 (plasminogen activator inhibitor type 1) levels.

If we look at the lipid abnormalities, we see the same LDL levels in insulin sensitive and insulin resistant subjects with type 2 diabetes. We see lower HDLs, higher triglycerides, and smaller, denser LDL, in spite of the fact that these diabetics have the same glucose levels.

I want to end by talking a little bit about some of the newer factors. PAI-1 is one, plasminogen activator inhibitor type 1. This is from the IRAS data by Andres Festa showing that there's a stepwise increase in PAI-1 levels as you go from normal to impaired to type 2 diabetes.

In the overall population, insulin sensitivity (S_I) is strongly inversely associated to PAI-1 levels, and significantly, but more weakly, associated to fibrinogen levels. It's not only related to glucose tolerance, it's also related to insulin sensitivity.

If we now look at C-reactive protein (CRP), which is now becoming a hot marker looking at subclinical inflammation--a number of excellent papers by Paul Ridker and others have been showing it's related to heart disease--we see strong correlations to CRP levels in nondiabetics with obesity, upper body obesity, weaker but still significant with systolic blood pressure, with fasting glucose, and a strong inverse association with insulin sensitivity.

The question is, do insulin sensitizers have an effect? Clearly weight loss has an effect on decreasing PAI-1 and CRP, but reasonable data are becoming available that suggest thiazolidinediones have this effect as well.

Preliminary data suggest that adding rosiglitazone to the regimen of patients already taking a sulfonylurea results in significant reductions in PAI-1 antigen and PAI-1 activity.

This are some data looking at CRP. Of interest here, and we don't fully understand it, is the reduction in CRP. While it is significant with both rosiglitazone 4 and 8 mg/day, it is not dose-related. We are still struggling with these data.

In conclusion, if we are going to think about the epidemic of cardiovascular disease in type 2 diabetes, probably no single intervention is going to wipe this out. Clearly improved glycemic control and the prevention of type 2 diabetes are important. As I mentioned, there was a lifestyle intervention study from Finland that showed a 58% reduction in type 2 diabetes with it. There are currently studies under way with alpha-glucosidase inhibitors, with metformin, with thiazolidinediones, looking at this issue. Intensive control of risk factors is also important. Many people believe that type 2 diabetes should be treated just as aggressively as people with CAD for both blood pressure and lipids. And last, I think a lot of attention needs to be placed on insulin sensitizing interventions.