This is the sixth in a series of articles on presentations at the American Diabetes Association Annual Meeting, San Diego, California, 10–14 June 2005.
Glycemia, insulin treatment, and cardiovascular disease
Many presentations at the ADA meeting addressed aspects of the relationship between diabetes and cardiovascular disease (CVD). Karin Bornfeldt (Seattle, WA) described evidence from mouse models, suggesting that dyslipidemia accounts for most of lesion formation and progression, while hyperglycemia appears to accelerate additional phases of plaque formation. A number of studies have utilized models of hyperglycemia without dyslipidemia, with inconsistent evidence of adverse effect. In contrast, animal models with both hyperglycemia and hyperlipidemia show consistent increase in atherosclerosis. In the LDL receptor–deficient mouse with autoimmune type 1 diabetes and fairly severe hyperglycemia compared with an intensively insulin-treated group, VLDL triglyceride and cholesterol levels were similar to control on a low-fat diet but elevated to levels above those in nondiabetic controls on a high-fat diet. Atherosclerotic lesion initiation was approximately twice that in both the nondiabetic and insulin-treated groups, in both the high-fat and low-fat diet groups, but the nondiabetic baseline atherosclerosis level was considerably greater on the high-fat diet. The lesions in the nonhyperlipidemic animals are macrophage-containing fatty streaks, reflecting increased recruitment of macrophages into the arterial wall, while in the high-fat diet group, there was also increased macrophage proliferation in the arterial wall, as well as increased hyaluronan deposition, with the lipid abnormality appearing to drive lesion formation. In a subsequent study, animals were fed a high-fat diet for 4 months and then changed to a low-fat diet with or without induction of diabetes, with diabetes not in itself associated with major lipid abnormality. Plaque progression was actually reduced in the diabetic model, and lesion morphology was similar to that in the nondiabetic group.
Ross Gerrity (Augusta, GA) described evidence from porcine models with hypercholesterolemia and streptozotocin-induced type 1–like diabetes. The diabetic state is associated with severe arterial calcification, which is otherwise unusual in animal models. With an insulin glargine treatment model, afternoon glucose levels were markedly reduced. The pig erythrocyte is impervious to glucose, so HbA1c (A1C) levels are not useful, but fructosamine measurement documented the effect of diabetes and the benefit of insulin treatment, with evidence of amelioration of coronary atherosclerosis. Computerized image analysis showed, on average, 87% stenosis in untreated animals; this was reduced to an average of 50% with glycemic treatment. In a porcine model of metabolic syndrome, with much less severe hyperglycemia, but with hyperinsulinemia, coronary stenosis was comparable to that in the type 1 model. The plaques in the diabetic animals are fragile and prone to rupture with formation of obstructive thrombosis. Gerrity concluded that in the type 1 diabetes model, the severity of vascular disease was associated with the degree of hyperglycemia, suggesting a direct role of glucose in progression of atherosclerosis, while in the type 2 diabetes model, with similar atherosclerosis progression despite considerably lower glucose levels, there must be additional factors.
An et al. (abstract 1984) demonstrated a rapid and sustained effect of hyperglycemia in augmenting cardiac AMP-activated protein kinase (AMPK) phosphorylation, a potential mechanism improving resistance of the diabetic heart to ischemia/reperfusion injury. Jessen et al. (abstract 236) examined LKB1, a possible upstream kinase of AMPK, showing that hearts of mice not expressing LKB1 had >95% reduction in AMPKα2 and 20% reduction in AMPKα1 activity. They also showed 60% reduction in phosphorylation of acetyl-CoA carboxylase, a physiological AMPK target, suggesting LKB1 to be an important component of normal AMPK cardiac signaling. Miller et al. (abstract 2012) studied mice expressing a dominant-negative AMPK α subunit, with absent AMPK activity. Left-ventricular diastolic pressure increased markedly, and there was decreased glycogen content following ischemia, suggesting that this may represent an important endogenous cardioprotection mechanism.
Sherita Golden (Baltimore, MD) reviewed epidemiological evidence of the relationship between glycemia and CVD. Sixteen-year follow-up of the 3,092 adults in the Second U.S. National Health and Nutrition Examination Survey, who underwent an oral glucose tolerance test in 1976–1980, showed that diabetes, whether diagnosed from fasting or 2-h glucose values, was associated with increased mortality (1). In the Finnish East West study of 1,373 nondiabetic and 1,059 diabetic persons, the CVD risk of persons with diabetes was similar to that of persons who had had a myocardial infarction (2). Epidemiological tools, she explained, may be used to compare glucose exposure based on A1C to clinical CVD outcomes. In distinguishing association from causation, temporal relationships, strengths of association, replication of findings, consistency with other knowledge, consideration of alternative explanations, dose-response relationships, biological plausibility, and effects of cessation of exposure need to be taken into account. Golden reviewed her meta-analysis of prospective studies with ≥1 year follow-up of the relationship between A1C and CVD end points, separately assessing peripheral, coronary, and cerebrovascular disease (3). Eighteen manuscripts were included in the final review, each following between 100 and >5,000 persons, on average for 8 years, using standard methods for CVD outcome classification. Ten studies of 7,435 persons with type 2 diabetes showed an adjusted 1.18-fold increase in CVD risk per 1% increase in A1C, ranging from a 13% increase for coronary disease to a 28% increase for peripheral arterial disease; similar effects with wider confidence intervals were seen for the three studies of 1,688 persons with type 1 diabetes. Golden noted that age, sex, race, cigarettes, lipids, adiposity, and blood pressure are associated with both CVD and diabetes, raising the possibility of confounding rather than causation. The possibility of publication bias, the small number of studies, and the heterogeneity of study results are additional concerns. Reinforcing the meta-analysis findings, an analysis of the Atherosclerosis Risk in Communities study showed that 1,638 persons with diabetes, age 45–69 years and followed for 8–10 years, had relative likelihoods of coronary disease (hospitalization for myocardial infarction, coronary artery bypass graft [CABG], or angioplasty; silent myocardial infarction on electrocardiogram; or coronary mortality) of 1.0, 1.36, 1.63, 2.21, and 2.78 in A1C quintiles <5.2, 5.2–6.7, 5.7–6.5, 6.5–9.2, and >9.2%, respectively. Adjustment for cigarette use, obesity, blood pressure, and lipids showed similar 1-, 1.28-, 1.56-, 1.99-, and 2.42-fold likelihood of coronary disease, a “dose response” relationship between glycemia and outcome supporting the finding, as well as suggesting that the A1C goal of 7% may be inadequate for CVD prevention.
David Nathan (Boston, MA) discussed clinical trial data. Persons with diabetes are subject to the general risk factors of age, blood pressure, obesity, smoking, and dyslipidemia. Twenty-five percent reductions in serious CVD outcomes are seen with statins and with blood pressure treatment and ∼15% reduction with aspirin. However, it has not been typically possible to lower the risk of persons with diabetes to that of persons without diabetes. Specific effects of hyperglycemia and of the complications of diabetes have implications for CVD, in particular those due to renal disease, autonomic neuropathy, and protein glycation. There are, furthermore, the myriad of effects of insulin resistance in type 2 diabetes. The relationship between diabetes and CVD is not as direct as that with microvascular complications such as retinopathy, but there does appear to be a continuous relationship between glycemia and both abnormalities, for CVD extending into the range of impaired glucose tolerance.
Diabetes is associated with more frequent coronary artery disease occurring at younger age, with higher mortality and excess morbidity following myocardial infarction, principally with an excess rate of development of congestive heart failure (CHF). Persons with diabetes experience worse outcome of revascularization, and there are diabetes-specific abnormalities of systolic and diastolic heart function. Nathan reviewed data from the Second National Registry of Myocardial Infarction, a 1994–1995 survey of 44,826 diabetic and 128,982 nondiabetic persons hospitalized for myocardial infarction. The 26% prevalence of diabetes at a time when diabetes was recognized in ∼6% of the population itself suggests the adverse cardiac effect of diabetes. The mean age with and without diabetes was 69 and 67 years, 47 and 36% were female, prior myocardial infarction had occurred in 31 and 23%, CHF had occurred in 22 and 11%, 17 and 30% were smokers, and 30-day mortality was 14 and 10%, respectively.
The first Diabetes Mellitus, Insulin Glucose Infusion in Acute Myocardial Infarction (DIGAMI) study enrolled diabetic persons treated within 24 h of acute myocardial infarction (AMI) randomized to intravenous insulin and glucose for at least 24 h with goal glucose 126–196 mg/dl or to standard therapy. Participants in the intensively treated group had a 30% 1-year and 28% 3.4-year decrease in mortality (4). In a study carried out in Leuven, Belgium, mainly following CABG surgery, of 1,548 patients admitted to an intensive care unit (ICU) requiring respiratory support and receiving parenteral glucose (5). Participants were randomized to glucose targets of 110 vs. 180–200 mg/dl, with 43% lower mortality in the intensively treated group, from 8 to 4.6%. Mortality was particularly reduced for those in the ICU >5 days, at 10.6 vs. 20.2%, and reduction was also seen in multiorgan failure, renal failure, bacteremia, critical illness polyneuropathy, and the need for prolonged stay in the ICU. Taken together, Nathan commented, the DIGAMI and Leuven studies suggest that short-term intensive therapy appears to improve outcomes in critically ill diabetic and nondiabetic patients, although he pointed out that the benefit might be due either to glycemia or to insulin treatment itself. Nathan also noted the noncontrolled studies suggesting that sternal wound infections following thoracotomy are decreased by intensive glycemic treatment (6,7).
For chronic glycemic therapy, the Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications (EDIC) data showed that, comparing 611 persons who had been randomly assigned to receive conventional diabetes treatment during the DCCT with 618 who had been assigned to receive intensive treatment, and excluding persons with preexisting CVD, hypertension, or dyslipidemia, carotid artery intima-media thickness worsened less rapidly in those persons previously assigned to intensive treatment, independent of other conventional CVD risk factors (8). Coronary artery calcification scores >100 and 200 units at years 8–9 were found in ∼14 vs. 10% and ∼7 vs. 4%, respectively. Furthermore, analysis reported at the ADA meeting showed 46 vs. 98 aggregate CVD events in 31 and 52 persons from the former intensive versus conventional care groups by year 11 of EDIC, giving 6 vs. 11% event rates overall: a 42% reduction. Severe events occurred in 11 and 29 persons, respectively, a 57% reduction in events. The presence of microalbuminuria appeared to explain ∼45% of the treatment effect, but A1C differences during the DCCT continued to explain the effect after adjustment for the presence of increase in urinary albumin (9).
Nathan raised the concern that drug toxicity may play a role in CVD among persons with diabetes, noting that sulfonylureas have been implicated as causing the 1%/year excess CVD mortality seen in the University Group Diabetes Program (UGDP) (10). Metformin, when added to sulfonylurea therapy in the U.K. Prospective Diabetes Study (UKPDS), was associated with a 96 and 60% increase in diabetes-related and all-cause mortality rates, respectively, compared with that seen in persons treated with sulfonylureas alone (11). There is also evidence of drug benefit, with metformin monotherapy associated with reduction in CVD mortality in the UKPDS. Thiazolidinediones improve endothelial function and decrease restenosis after angioplasty, although Nathan suggested that “at this point, I think it’s premature to recommend [thiazolidinediones] for widespread use.”
Intensive insulin treatment for AMI.
These and additional studies were further scrutinized in a debate on the role of intensive insulin therapy in the setting of AMI. Silvio Inzucchi (New Haven, CT) presented an argument against this treatment, suggesting that pertinent trials are limited. In the Cooperative Cardiovascular Project, >142,000 Medicare beneficiaries with AMI were studied (12). The admission glucose was strongly associated with 30-day mortality, but only among persons not known to have diabetes, perhaps suggesting that in this group, glucose is a marker of the degree of illness or of lack of access to care. Inzucchi warned against the error of “post hoc ergo propter hoc” reasoning (literally, “that which comes after must be the result of”), stressing that it is just as likely that elevated glucose will be associated with higher risk rather than directly causing this risk, with those patients with the greatest degree of acute illness and/or the more severe diabetes tending to have the most hyperglycemia. Thus, elevation in proinflammatory cytokines and counterregulatory hormones might cause both hyperglycemia and adverse cardiac effects, although it is certainly possible that hyperglycemia in the setting of AMI causes adverse effects such as osmotic diuresis, electrolyte changes, abnormality of immune function and wound healing, and elevation in free fatty acids (FFAs).
Glucose-insulin-potassium (GIK) infusion has been proposed as being advantageous during AMI, with putative benefits including increase in intracellular cardiac potassium and ATP levels, with maintenance of citric acid cycle intermediate levels, increase in coronary blood flow, decrease in cardiomyocyte apoptosis, inhibition of peripheral lipolysis, and decreasing FFA oxidation and myocardial oxygen demand, leading to reduction in cardiomyocyte necrosis and infarct size. Although meta-analysis suggests benefit of such an approach (13), the CREATE-ECLA study of GIK versus standard care in 20,201 persons showed no benefit (14). Inzucchi suggested that the first DIGAMI study “tried to do too much,” leaving it unclear which of the inpatient and outpatient components to care of the intervention group was responsible. Furthermore, a second DIGAMI study included 1,274 persons with type 2 diabetes having AMI, comparing intensive inpatient, intensive inpatient plus outpatient, or neither form of glycemic treatment. Although Inzucchi acknowledged that recruitment was poor, that the glycemic difference was minimal (mean glucose 164 mg/dl for groups 1 and 2 vs. 180 mg/dl in group 3; A1C slightly improved at 3 months with subsequent similar levels in all groups), and that overall mortality was below that expected at 21%, making it unlikely that the study would become positive, he pointed out that there actually was a trend to improved outcome in the control group receiving neither inpatient nor outpatient intensive glycemic treatment (15). The Leuven ICU study showed decreased mortality, sepsis, dialysis, transfusion, and requirement for ventilatory support among patients receiving aggressive insulin treatment, but Inzucchi noted that the benefit was seen in long-stay patients, quite a different group from persons with AMI.
The potential adverse effects of insulin include hypoglycemia, sympathetic stimulation, increased stroke volume, shifts in intracellular potassium and phosphate levels, and the mitogenic actions of insulin, which might cause proliferation of abnormal blood vessels. Inzucchi reminded listeners that even mild hypoglycemia (50 mg/dl) increases epinephrine levels 100-fold, noting that hypoglycemia occurred in 7% of patients in the Leuven study, in 4–13% of aggressively treated persons in a study at his institution, and in 10–13% of persons in the second DIGAMI study. Inzucchi reviewed a recent study of 713 persons with diabetes and acute coronary syndrome, with a 2.66-fold increase in mortality for the highest versus lowest quartile of admission blood glucose but a 1.77-fold increase in mortality for those experiencing any hypoglycemia during the hospitalization (16), suggesting the potential for adverse effect of aggressive insulin treatment.
Furthermore, there are multiple competing clinical priorities for the treatment of myocardial infarction, with stents, aspirin, fibrinolysis, ACE inhibitors, β blockers, statins, and many other important modalities to be considered. Not all critically ill patients are the same, complicating the situation even more, with completely different diagnoses in the medical, surgical, and coronary ICUs, and there is danger that emphasizing aggressive diabetes treatment might lead to less attention being paid to other treatment approaches. Inzucchi proposed a “rational evidence-based recommendation” that glucose should be maintained above 140 mg/dl, with levels exceeding 200 mg/dl to be avoided, and suggested a goal of 180 mg/dl based on the second DIGAMI study findings. He further implied that there may be little benefit of parenteral insulin and urged that hypoglycemia be assiduously avoided.
Hertzel Gerstein (Hamilton, Canada) countered with a presentation in favor of intensive insulin therapy, commenting on the difficulty of interpretation of the limited data. He reviewed the particularly adverse prognosis of AMI in persons with diabetes. In the VALIANT (VALsartan In Acute myocardial iNfarcTion) trial of 14,703 persons after myocardial infarction (70% Q-wave), diabetes, regardless of whether newly diagnosed, increased risk of death, heart failure, and recurrent myocardial infarction ∼1.5-fold (17). For persons with acute coronary syndrome, in the OASIS (Organization to Assess Strategies for Ischemic Syndromes) registry of 8,103 persons, those with diabetes had 1.57-fold increase in mortality, with similar increase in coronary death, myocardial infarction, CHF, and stroke (18).
Gerstein reviewed data showing that abnormal glucose is common in persons with AMI, with one study showing abnormal glucose in 71% of AMI patients and in 65% of those seen for outpatient cardiology evaluation (19). Furthermore, any degree of glucose abnormality conveys increased risk, with glucose >137 mg/dl associated with fourfold increase in mortality in a study of 662 persons with AMI (20). For acute coronary syndrome, there is a linear relationship between admission glucose and left-ventricular failure (21). Gerstein showed data from a study of fasting glucose levels the day following myocardial infarction. Persons with levels 110–121 mg/dl had 4.6-fold, persons with 122–138 mg/dl had 6.4-fold, and persons with ≥139 mg/dl had 11.5-fold increases in mortality (22). Another study of 168 persons after myocardial infarction showed an association of abnormal glucose with more than a fourfold worsening of prognosis (23).
The question, then, is whether the association of elevated glucose levels with worse prognosis implies that insulin treatment would be of benefit. Gerstein noted the particular importance of the exponential rise in myocardial FFA exposure after myocardial infarction, as a result of elevation in catecholamines and cortisol and of falls in insulin, effects of heparin treatment, and alterations in albumin binding. Consequences of FFA elevations include reductions in uptake of pyruvate and in glycolysis and anaerobic metabolism, as well as increase in lactate with development of acidosis, reduction in ATP production, reducing contractility and coronary flow. Insulin is either absolutely or relatively deficient in persons with elevated glucose. Insulin secretion and action is suppressed by AMI. Insulin lowers FFA levels and has anti-inflammatory, vasodilatory, and antithrombotic effects. Insulin has no contraindications, is easily titrated, is safe in acute settings, and can be added to any other treatment. The only side effect is hypoglycemia, and that occurs only, Gerstein said, “when the wrong amount of insulin is given.” Insulin-mediated normoglycemia may protect against ischemic myocardium, leading to adverse outcome.
Gerstein emphasized that the first DIGAMI study did show benefit, in association with a 38-mg/dl lowering of glucose levels, while the second DIGAMI study did not lead to sustained differences in glucose and had no benefit. Similarly, the CREATE-ECLA study of GIK failed to show benefit, but he pointed out that insulin was given after reperfusion, and glucose levels increased to 180 vs. 145 mg/dl in the control group, leading to his wry comment that “if you’re going to give insulin you’ve got to be sure you do more than raise the glucose level.” In further analysis of the Leuven study, both higher insulin dose and higher glucose levels were associated with worse outcome (24), leading Gerstein to emphasize that giving insulin “is not enough, you have to lower the blood glucose.” Furthermore, there is now evidence that persons with diabetes undergoing CABG with aggressive glucose management have reduced atrial fibrillation, shorter postoperative length of stay, survival advantage, and decrease in recurrent ischemia and wound infections (25). A recent meta-analysis of the use of insulin in critical illness showed that it leads to a 15% risk reduction (26), which may particularly show the benefit of having a glucose goal. “The purpose of evidence-based medicine,” Gerstein summarized, “is to inform clinical decision making,” although “we have to deal with uncertainty.” There is growing evidence that the majority of persons with myocardial infarction have increased glucose; that the higher the glucose, the worse the prognosis; and that clinical trials of interventions that do in fact lower glucose suggest that insulin has benefit.
Inpatient diabetes treatment.
A number of presentations at the ADA meeting addressed approaches to optimizing glycemia among hospitalized persons with diabetes. Falciglia et al. (abstract 1043) analyzed 28,48 patients hospitalized at 34 ICUs in 17 geographically diverse hospitals from 1996 to 1997; 27% had a diagnosis of diabetes and 51% a diagnosis of CVD. Ninety percent of diabetic patients with CVD were hyperglycemic (42% severely) during ICU admission, as were 66% of nondiabetic patients with CVD, suggesting an important opportunity for intervention.
Juneja et al. (abstract 419) used a program based on calculation of insulin sensitivity to recommend intravenous insulin doses in 34 persons in the ICU. A total of 4,137 blood glucose measurements were obtained with an average of 104 mg/dl and with 61% of glucose readings 80–110 mg/dl, and 92% 60–150 mg/dl, with 0.7% of levels <50 mg/dl, suggesting the feasibility of an automated insulin dose calculator. Dinapoli (abstract 422) reported that use of insulin infusion in 518 persons following CABG with goal glucose 110–160 mg/dl was associated with reduction in mean postoperative glucose from 220 to 140 mg/dl, in association with deep sternal infection rates decreasing from 1.5 to 0%, inpatient mortality rates from 5.0 to 1.2%, and length of stay from 9.9 to 8.1 days. Balkin et al. (abstract 506) reported a community hospital ICU insulin infusion protocol with changes in insulin infusion rate given in a table of previous versus current blood glucose levels, with 3,562 h of treatment in 40 persons leading to mean glucose 134 mg/dl. Garnica et al. (abstract 548) compared nurse- versus physician-controlled titration of intravenous insulin infusions in a general ward, showing mean glucose 168 vs. 217 mg/dl. In a meta-analysis of four studies of cost effectiveness of intensive glycemic treatment in the ICU, Hilleman et al. (abstract 528) reported that the total of 6,162 patients showed 6.7 vs. 10.7% mortality, and mean excess cost of $248 per patient, suggesting a highly cost-effective intervention with cost per death avoided $6,250. Gandhi et al. (abstract 690) reported a retrospective analysis of 409 persons having had cardiac surgery, showing that each 50-mg/dl increase in mean glucose was associated with more than a doubling of the risk of adverse outcome. Rizza et al. (abstract 694) studied 92 persons without known diabetes at least 3 months following myocardial infarction, all with coronary angiography. Glucose tolerance testing showed that 20% had diabetes and 38% impaired glucose tolerance, with the number of angiographic stenotic lesions correlating with postload glycemia.
Okeke et al. (abstract 276) reported glycemic outcome of nonrandomized treatment of 33 persons with type 2 diabetes on general medical hospital units receiving subcutaneous insulin according to a “sliding scale” of glucose levels versus 27 persons treated with intravenous insulin, showing mean glucose on the 2nd and 3rd days of treatment of 239 vs. 195 mg/dl, with 24 vs. 9% of glucose levels exceeding 300 mg/dl and with 3 vs. 0% having glucose <70 mg/dl. Baldwin et al. (abstract 275) randomized 81 persons following gastric bypass surgery whose first postop glucose exceeded 144 mg/dl to insulin glargine once daily, initially at 0.3 units/kg daily (34 were first admitted to the ICU and given glargine at a daily dose of 20 times the last hourly insulin infusion rate, averaging 0.4 units/kg daily, which the authors consider a more optimal starting dose), subsequently increased or decreased by 20% for glucose >180 or <80 mg/dl, resulting in mean postoperative glucose 134 mg/dl, as compared with regular insulin based on a sliding scale of 4–12 units of regular insulin every 6 h for blood glucose >144 mg/dl, with mean postoperative glucose significantly higher at 153 mg/dl. Yeldandi et al. (abstract 512) compared glargine at a daily dose of 20 times the last hourly insulin infusion rate versus twice daily treatment with NPH at eight times plus regular insulin at four times the last insulin infusion rate in persons leaving the cardiovascular ICU. The protocols led to similar outcome with 60 vs. 63% of glucose levels between 80 and 140 mg/dl, mean glucose 131 vs. 124 mg/dl, and 0.5 vs. 2% of glucose levels <60 mg/dl.
CVD risk prediction.
Coleman et al. (abstract 697) reported that, using two separate datasets of persons with diabetes, the UKPDS risk engine (http://www.dtu.ox.ac.uk/index.html?maindoc=/riskengine/) came closer to predicting CVD event rates than the Framingham equations. Al-Saraj et al. (abstract 698) used the UKPDS predictive model to estimate that the 10-year risk of a cohort of 78 recently referred persons with type 2 diabetes was 25 vs. 9% in males versus females, with considerable variability, suggesting this to be a useful approach for determining intervention requirements for type 2 diabetic individuals. Donnan et al. (abstract 705) presented a diabetes–coronary heart disease (CHD) risk equation based on the population database of Tayside, Scotland, including 4,259 persons with diabetes, calculating risk of a first CHD event based on age at diagnosis, duration of diabetes, smoking, sex, systolic blood pressure, total cholesterol, and A1C.
Kuller et al. (abstract 1036) compared lipids in 214 participants in the Multiple Risk Factor Intervention Trial dying of coronary disease versus 214 control subjects matched for number of metabolic syndrome components, prior CVD status, and age. HDL was 37 vs. 38 mg/dl and LDL cholesterol 161 vs. 153 mg/dl, with lower HDL particle concentration based on nuclear magnetic resonance (NMR) spectroscopy, affecting both large and medium HDL particles. Lemieux et al. (abstract 743) followed 1,890 nondiabetic middle-aged men from the Québec Cardiovascular Study for 13 years, finding that among those with low Framingham score, hyperinsulinemia increased the likelihood of CVD events 1.8-fold, with further analysis showing insulin to be a risk factor only in the presence of elevated apolipoprotein B, suggesting small LDL particles.
Burrows et al. (abstract 1009) reported 6 and 19% prevalences of CHD among American Indians and Alaska Natives aged 35–54 and ≥55 years, respectively, in 2002, with considerable state-to-state variation. Howard et al. (abstract 730) reported that among 2,341 nondiabetic and 2,124 diabetic 45- to 74-year-old American Indians with and without baseline CVD, the 10-year risk of a CVD event was 10.1% with neither diabetes nor baseline CVD, 19.9% with diabetes alone, 57.5% with baseline CVD without diabetes, and 61.2% with both, suggesting diabetes alone not to be as strong a risk determinant in this population. Saley et al. (abstract 742) followed persons with and without diabetes who had coronary angiography. Persons with neither diabetes nor angiographic abnormality had a 9% 4-year event rate and those with diabetes without angiographic abnormality a 10% 4-year event rate, while those with angiographic abnormality alone had a 24% 4-year event rate and those with both diabetes and angiographic abnormality a 40% 4-year event rate. Tierney et al. (abstract 996) analyzed a North Dakota dataset to determine heart disease mortality rates in 1992–1998 and 1999–2003 among men and women with and without diabetes. Rates decreased from 8.6 to 6.7 per 1,000 women with diabetes and from 13 to 10 per 1,000 men with diabetes, while decreasing from 2.5 to 2.3 per 1,000 women and from 4.4 to 3.5 per 1,000 men without diabetes.
Jousilahti et al. (abstract 985) analyzed a dataset of 49,680 Finnish men and women 25–74 years of age without baseline history of stroke, CHD, or type 1 diabetes, having 2,564 stroke events, 812 fatal, over 17.2 years. The stroke risks associated with untreated hypertension with blood pressure <160/95 mmHg (I), treated hypertension or blood pressure ≥160/95 mmHg (II), diabetes, or both diabetes and hypertension I and II were 1.28, 1.90, 2.52, 3.30, and 4.06, respectively, and the stroke mortality risks were 1.59, 2.75, 2.77, 5.49, and 8.64, respectively, with the authors commenting, “A significant proportion of the risk of stroke assumed to be related to hypertension is actually due to concomitant diabetes.” Yuan et al. (abstract 1013) compared stroke determinants among 980 persons with and 12,951 without diabetes in the National Health and Nutrition Examination Survey (NHANES) III (1988–1994). Controlling for age, sex, race, smoking, obesity, and hypertension, LDL cholesterol was not significant, with best lipid prediction from categorizing the triglyceride–to–HDL cholesterol (mg/dl) ratio based on the diabetes sample as <1.4, <25th percentile; 1.4–3.9, 25–75th percentile; and >3.9, >75th percentile. Persons with and without diabetes with high versus low triglyceride–to–HDL cholesterol ratio had 7.5- and 2.7-fold respective increases in stroke, suggesting this to be a particularly important risk factor for those with diabetes.
James Shepherd (Glasgow, U.K.) presented an analysis of 753 vs. 748 diabetic participants in the Treating to New Targets (TNT) 5-year study of 5,006 and 4,995 persons with clinical evidence of CHD (previous myocardial infarction, previous or current angina with objective evidence of atherosclerotic CHD, and/or a history of coronary revascularization) randomized to 10 vs. 80 mg atorvastatin daily (27). Baseline A1C was 7.4%, with 81 and 85% having angina, 71 and 70% hypertension, and 55 and 54% having had CABG. On treatment, LDL was 99 and 77 mg/dl, triglyceride 178 and 145 mg/dl, and HDL cholesterol 44 mg/dl in both. There were a total of 135 and 103 coronary events in the two groups, 18 and 14% incidences, compared with 10 and 8% incidences among persons without diabetes, with 31 and 23 coronary deaths, 61 and 49 nonfatal myocardial infarctions, and 75 and 52 strokes or transient ischemic attacks in the diabetic subgroup randomized to 10 and 80 mg atorvastatin daily, respectively. No diabetic patient in either group had rhabdomyolysis, and myalgia occurred in 27 and 18 persons, respectively. Thus, the incidence of cardiovascular events was higher in patients with than in those without diabetes, as expected, and the aggressive lipid-lowering strategy conveyed somewhat greater relative benefit in the diabetic subgroup, as well as considerably greater absolute benefit.
Haffner et al. (abstract 4) analyzed lipid parameters from 1,486 participants in the Insulin Resistance Atherosclerosis Study. LDL particle number, measured by NMR, showed a stronger relationship than apolipoprotein B to HDL cholesterol, triglyceride, 2-h glucose, waist circumference, and insulin sensitivity, and the NMR measure of LDL size was more strongly associated with these parameters than was LDL size measured by gradient gel electrophoresis. Soda et al. (abstract 953) measured LDL and small LDL concentrations before and following each meal and at bedtime in 12 persons with type 2 diabetes, finding that both parameters were highest in the morning, LDL decreasing up to 13%, and small LDL decreasing up to 37%, suggesting an important diurnal change that may be due to increased catabolism during the daytime. Thus, it may be particularly important to measure fasting lipid levels in persons with diabetes. Itoh et al. (abstract 675) gave a standardized meal to 68 type 2 diabetic persons, finding that a 2-h triglyceride of 200 mg/dl predicted an increase in carotid intima-media thickness.
Cardiovascular toxicity of sulfonylureas
Harold Lebovitz (Brooklyn, NY) further discussed the potential cardiovascular toxicity of sulfonylureas. “[Sulfonylureas] go back an awfully long time,” he noted, “and we’ve never really resolved this issue about whether sulfonylureas are really dangerous … Since 1970 the question about whether sulfonylureas may have [cardiovascular] toxicity has plagued the medial community.” In 1970, the UGDP reported increased cardiovascular mortality in tolbutamide versus placebo or insulin-treated patients. The finding that ATP-sensitive K+ channels (KATP channels) are present in cardiac myocytes in 1983, with subsequent realization that sulfonylureas act at the KATP channels of β-cells and understanding of their role in many tissues in adapting to stress has led to a theoretical mechanism for sulfonylurea cardiovascular toxicity (28). The KATP channel is composed of 4 units, each made up of a separate potassium inward rectifier (Kir) and sulfonylurea receptor (SUR) subunits, the latter carrying regulatory binding sites for benzamido (C6H5·CO·NH−) groups, sulfonylurea, and ATP, all closing the channel. Separate repaglinide and tolbutamide binding sites have been identified. Physiologically, in the β-cell, glucose increases intracellular ATP levels, closing the channel, resulting in membrane depolarization, opening the adjacent calcium channel with consequent increase in intracellular calcium concentration. There are KATP channels in myocytes and in neurons. Kir 6.1 and 6.2 and SUR1 are present in the β-cell and neuron, SUR2A in cardiac and skeletal muscle, and SUR2B in vascular smooth muscle. The β-cell channel is open in the basal state because of low intracellular ATP levels, while in the heart, the channel is closed under basal conditions because of high ATP levels, but under circumstances of hypoxia, ATP levels decrease and the channel opens, explaining the phenomenon of ischemic preconditioning described below. Glyburide affects the cardiac as well as the β-cell channel, as does repaglinide, while nateglinide affects the β-cell channel while not having effect on the cardiac channel. The β-cell–to–cardiac ratio is highest for nateglinide and lowest for glyburide and repaglinide.
Ischemic preconditioning is the phenomenon whereby exposure of the myocardium to a brief episode of ischemia and reperfusion markedly reduces tissue necrosis induced by a subsequent episode of prolonged ischemia, initially described almost 2 decades ago. Ischemia causes release of adenosine, bradykinin, and oxygen radicals; activates G-protein, cell phospholipases, protein kinase Cε, and p38 MAPK; and opens KATP channels by reducing intracellular ATP levels. Electrocardiographic ST elevation is mediated in part by the KATP channel, so that channel closers may mask the typical findings of myocardial infarction. Cardiomyocyte mitochondrial KATP channels may also play a role in this phenomenon. Given this information, it is noteworthy that glyburide inhibits the beneficial effects of ischemic preconditioning in animal models of myocardial infarction (29), as well as blocking the effects of KATP channel openers. In humans, glyburide also can be shown to prevent ischemic preconditioning (30,31).
Lebovitz asked whether these findings have clinical applicability. Comparing outcomes after angioplasty for AMI in 67 diabetic persons receiving sulfonylureas vs. 118 not taking these drugs, hospital mortality was 24 vs. 11%, with a 2.8-fold increase in early mortality associated with sulfonylurea treatment in multivariate logistic regression, not explained by an increase in ventricular arrhythmias, and potentially due to adverse sulfonylurea effects on myocardial tolerance for ischemia and reperfusion (32). There is also evidence of echocardiographic abnormality with glyburide treatment. Although some authors have not reported increased mortality with glyburide treatment (33), investigators of this topic have recommended that “discontinuation of sulfonylurea treatment should be considered” in persons with repeated coronary ischemia and during and after emergency or elective angioplasty and CABG (34). As glyburide is the only agent shown to block ischemic preconditioning, Lebovitz recommended that it no longer be used and noted that nateglinide may be the least likely to cause these effects. He noted, however, that the UGDP results appeared to be “a fluke,” at least insofar as reflecting a sulfonylurea effect on KATP channels, as tolbutamide does not exhibit the strong cardiac KATP channel binding seen with glyburide.
Zachary T. Bloomgarden, MD, is a practicing endocrinologist in New York, New York, and is affiliated with the Division of Endocrinology, Mount Sinai School of Medicine, New York, New York.