The Metabolic Syndrome and 11-Year Risk of Incident Cardiovascular Disease in the Atherosclerosis Risk in Communities Study

Between 1987 and 1989, ARIC examined population-based samples of 15,792 residents aged 45–64 years in four U.S. communities in North Carolina, Mississippi, Minnesota, and Maryland (15). The study was reviewed by institutional review boards at the clinical sites, and informed consent was obtained from all participants. The study population for these analyses included all cohort members who did not meet the following exclusions: race other than black or white (n = 48); black participants residing in Minnesota or Maryland sites (n = 55); failure to fast at least 8 h before venipuncture (n = 587); missing data on metabolic syndrome components (n = 299); prevalent/missing data on coronary heart disease (CHD; prior myocardial infarction by electrocardiogram readings or self-reported physician diagnosis or cardiac revascularization procedure) or history of self-reported stroke (n = 1,286); prevalent diabetes (fasting glucose ≥126 mg/dl, self-reported physician diagnosis or use of hypoglycemic medications, n = 1,310); or missing data on LDL cholesterol or smoking (n = 118). After applying these exclusions, 12,089 individuals remained for analysis.

Detailed reporting of procedures for blood collection and centralized measurement of lipids (16) and serum glucose (17) have been published previously. Participants were asked to fast 12 h before the blood was drawn and actual fasting times were recorded. Blood was drawn from an antecubital vein of seated participants, serum was centrifuged, and frozen samples were shipped to central laboratories for analysis. Triglycerides were measured by enzymatic methods, and HDL cholesterol was measured after dextran-magnesium precipitation. LDL cholesterol was calculated using the Friedewald formula (18). Glucose was measured by the hexokinase/glucose-6 phosphate dehydrogenase method. During the baseline clinical examination, trained technicians measured waist circumference to the nearest centimeter at the umbilical level. Sitting blood pressures were measured after 5 min of rest by the use of a random-zero sphygmomanometer, and the mean of the last two of three seated systolic and diastolic blood pressure measurements was used for analysis.

The metabolic syndrome was defined according to ATP III criteria (13) and required the presence of three or more of the following: large waist circumference (>88 in women and >102 cm in men), elevated triglyceride level ≥150 mg/dl, low HDL cholesterol (men <40 and women <50 mg/dl), elevated fasting glucose (110 –125 mg/dl), and elevated systolic or diastolic blood pressure (≥130/85 mmHg) or self-reported use of antihypertension medications. In addition, a modified definition of metabolic syndrome was created by applying the American Diabetes Association’s recently recommended (19) lower range for impaired fasting glucose (100–125 mg/dl) to assess the effect of this proposed change on the estimated prevalence of the syndrome and its association with CVD.

Incident CHD and ischemic stroke events were ascertained using standard ARIC protocols (20,21). Incident CHD included fatal or nonfatal hospitalized myocardial infarction, fatal CHD, silent myocardial infarction identified by electrocardiography, or coronary revascularization. Incident ischemic stroke events (thrombotic or cardioembolic) were classified according to a computer algorithm, and an expert reviewer independently classified each eligible case using criteria adapted from the National Survey of Stroke (22); disagreements were adjudicated by a second expert physician. Follow-up time for incident CHD and stroke events was the minimum number of days between the baseline visit and either the first event, death from other causes, last contact, or 31 December 1999.

Incidence rates, produced by dividing the number of events by person-time at risk, were stratified by sex. Cox proportional hazards regression was used to model the relationship between the metabolic syndrome and time to incident CHD or ischemic stroke. Verification of the proportional hazard assumption was assessed using plots of the log(–log) survival curves and Schoenfeld residuals (23). The increment in CHD risk associated with the number of metabolic syndrome components was assessed using indicator variables for the presence of one, two, three, or four or more components of the syndrome. The excess risk associated with the metabolic syndrome, beyond that predicted by its individual components, was determined by an indicator for the syndrome in models that also included all five individual components. Survival models contained indicator variables for each race by ARIC center combination to allow for the unique contribution of each group on CVD risk.

To assess the contribution of the metabolic syndrome in the context of traditional CHD risk algorithms, Framingham Risk Scores (FRSs) (13) were calculated and ARIC participants were categorized as having <10% or ≥10% risk of CHD. The third risk category outlined by ATP III (>20%) was not used because very few individuals in the study population had an FRS of this magnitude. Crude incidence rates within FRS categories were estimated among those with and without the metabolic syndrome. To determine whether the metabolic syndrome increased the prediction of CHD beyond the level predicted by the FRS alone, receiver operating characteristic (ROC) curves were generated from sex-specific logistic regression models that included 1) the FRS and 2) the FRS plus the metabolic syndrome. All analyses were performed using SAS for Windows (version 8.02; SAS Institute, Cary, NC), except for comparisons of the area under ROC curves, which were performed using Stata Statistical Software (release 8.0; StataCorp, College Station, TX).

The metabolic syndrome was present in 24% of women and 23% of men (Table 1). Among women, the prevalence was slightly higher among blacks than whites (28 vs. 23%), but black men had a lower prevalence than whites (18 vs. 24%). Approximately 60% of black participants met the criteria for elevated blood pressure, compared with ∼23% of whites. Central obesity (large waist circumference) was more prevalent in women than men (59 vs. 31%), and nearly three-quarters of black women were positive for this component. Black men had higher rates of current smoking but lower prevalence of elevated triglycerides and low-HDL components compared with other groups. The prevalence of elevated fasting glucose ranged from a low of 8% in white women to ∼15% among other race/sex subgroups.

Over an average of 11 years of follow-up, 879 incident CHD and 216 incident ischemic stroke events occurred. Crude incidence rates per 10,000 person-years for CHD and ischemic stroke were significantly greater among individuals with the metabolic syndrome than among comparison group members (Table 2). Crude hazard ratios (HRs) of CHD associated with the metabolic syndrome were 2.55 (95% CI 2.00–3.27) in women and 1.51 (1.27–1.79) in men with a statistically significant interaction by sex (P < 0.0006). No effect modification was detected by race (P = 0.57).

Although the magnitude of associations were attenuated slightly after adjustment for age, race/center, LDL cholesterol (<129, 130–159, or ≥160 mg/dl), and smoking (current, former, or never), individuals with the metabolic syndrome remained at significantly increased risk of CHD (sex interaction P < 0.03) (Table 2). Further adjustment for physical activity, education, income, use of alcohol, and family history of CHD resulted in minimal change in the HRs associated with the metabolic syndrome (results not shown). Similar associations were found between the metabolic syndrome and incident stroke, but the relatively smaller number of events decreased the precision of these estimates.

Applying the lower threshold for impaired fasting glucose (100 mg/dl) to the metabolic syndrome definition increased the overall prevalence from ∼23–32% (Table 1), but adjusted HRs for CHD and stroke were similar to those produced when applying the original ATP III definition. After adjustment for age, race/center, LDL cholesterol, and smoking, HRs of CHD were 2.11 (95% CI 1.64–2.72) among women and 1.39 (1.18–1.63) among men. Adjusted HRs of ischemic stroke were 1.90 (1.25–2.89) and 1.52 (1.05–2.18), respectively.

As shown in Fig. 1, the risk of CHD in the study population increased monotonically with increasing number of metabolic syndrome components in men and women. Within each combination (one, two, three, or four or more), the HR of CHD was larger in women than men. Compared with ARIC participants with no components of the metabolic syndrome, adjusted HRs for those with one and four or more components ranged from 1.63 (95% CI 0.97–2.72) to 5.25 (3.10–8.89) in women and from 1.34 (1.04–1.72) to 2.23 (1.64–3.04) in men.

In Cox models containing the individual components, elevated blood pressure and low HDL cholesterol were significantly associated with CHD risk after adjustment for age, race/center, LDL cholesterol, and smoking (Table 3). In separate, multivariate-adjusted models that also included an indicator for the presence of the metabolic syndrome, the incremental HR for the metabolic syndrome was close to the null value and was not statistically significant [men: HR 0.91 (95% CI 0.67–1.23); women: HR 0.71 (0.45–1.14)], indicating that the risk of CHD associated with the syndrome was not in excess of the level explained by the presence of its individual components.

Predicted 10-year risk of CHD based on the FRS algorithm was consistent with that expected from a middle-aged population initially free of diabetes, CHD, and stroke. A total of 98% of women and 53% of men had an FRS <10%. Incidence rates of CHD per 10,000 person-years were greater among those with the metabolic syndrome compared with those without in both the FRS <10% strata [women: 52.5 vs. 21.8 (difference 30.7, 95% CI 19.6–41.8); men: 90.3 vs. 65.4 (24.9, −3.6 to 53.5)] and FRS ≥10% strata [not estimable in women; men: 164.6 vs. 133.8 (30.8, 0.0–62.5)]. However, comparison of ROC curves indicated that the metabolic syndrome did not materially improve CHD risk prediction beyond the level achieved by the FRS among women (0.729 vs. 0.731) or men (0.631 vs. 0.634).

Although a recent study (8) of Native Americans found no association between the metabolic syndrome and incident CVD after adjustment for established risk factors, most published studies (4,6,7,9–12) that have applied consensus-based definitions of the metabolic syndrome have found that individuals with the syndrome are approximately two times more likely to develop CVD than individuals without the syndrome. However, limitations of several of these investigations included study populations composed exclusively of white men or individuals with a family history of type 2 diabetes or study designs that lacked CVD morbidity data (4,6,10,12).

In the current study, we found that among a population free of prevalent CHD, stroke, and diabetes, men and women with the metabolic syndrome were ∼1.5 and 2 times more likely to develop CHD after adjustment for established risk factors. The relative risk of CVD mortality associated with diabetes is greater in women than in men, and the presence of diabetes has been shown to erase the “gender gap” in CVD incidence rates between men and women (24). In a recent analysis of data from the San Antonio Heart Study, Hunt et al. (11) reported that the association between the metabolic syndrome and CVD mortality was more than twice as large in women than men in the San Antonio Heart Study population, which included individuals with prevalent diabetes and CVD, but found no sex modification in the primary prevention population, i.e., those without prevalent diabetes or CVD. The authors suggest that one possible explanation for the sex modification observed in the general population is the stronger diabetes-CVD association in women. However, our analysis indicates that the significantly greater risk of CHD associated with the metabolic syndrome in women is present among individuals without prevalent diabetes or CVD. Furthermore, in univariate and multivariate-adjusted models, the elevated fasting glucose component was not significantly associated with CHD. Therefore, in our study, it is unlikely that the sex interaction we observed is due to hyperglycemia alone.

Although we found a positive association between increasing number of ATP III metabolic syndrome components and CHD risk, results from our analysis indicate that the individual components for elevated blood pressure and low HDL cholesterol have the strongest effects on CHD risk. Using data from the ARIC study, Golden et al. (25) assessed the association between groupings of variables related to the metabolic syndrome and intimal-medial wall thickness of the carotid arteries as a measure of generalized atherosclerosis. They found that several combinations of risk factors associated with the metabolic syndrome conferred excess additive risk beyond the level predicted by the individual components. Conversely, in our analysis, multivariate proportional hazards modeling of all five components plus an indicator for the metabolic syndrome did not detect excess CHD associated with the presence of the syndrome. These results are consistent with a cross-sectional analysis by Alexander et al. (26), who reported that although elevated blood pressure, low HDL cholesterol, and diabetes were significantly associated with prevalent CHD, they found no indication of excess risk for the ATP III–defined metabolic syndrome once the individual components had been accounted for in multiplicative risk models.

Although the ATP III report recommends screening for the metabolic syndrome, specific recommendations for therapeutic lifestyle change and/or pharmacologic therapy are based on the number of CHD risk factors present at the time of screening and category of 10-year risk of CHD estimated by the FRS algorithm (<10%, 10–20%, and >20%). In the present analysis, inclusion of the metabolic syndrome did not add to CHD prediction models that included the FRS. However, as outlined in the ATP III report (13), the FRS was developed to identify individuals primarily for short-term risk reduction (≤10 years), whereas identification of individuals with metabolic syndrome is aimed at long-term risk reduction primarily through lifestyle modification.

Nevertheless, patients falling in the category of <20% CHD risk who met the criteria for the metabolic syndrome had a higher absolute rate of CHD than those without the syndrome. These findings are similar to those reported by Girman et al. (27), who categorized individuals from two large lipid-lowering clinical trials by FRSs of either >20 or ≤20% and found that individuals with the ATP III metabolic syndrome had a higher risk of major coronary events than comparison subjects within FRS categories. In a joint analysis of the Framingham Offspring and San Antonio Heart Studies (28), those with the metabolic syndrome had a greater average FRS than those without, regardless of ethnicity. Taken together, these findings indicate that individuals within the same FRS category who have the metabolic syndrome are at increased risk of CHD relative to subjects without the syndrome. This suggests that identification of metabolic syndrome may have prognostic value to medical practitioners for patients in both the upper (>20%) and lower (≤20%) FRS categories.

The present investigation has several strengths. The ARIC study is one of the largest and longest followed biracial samples of U.S. adults with validated CHD and stroke outcomes and standardized collection of covariate information. However, several limitations of the data should be kept in mind. The ARIC cohort was selected from four U.S. communities but is not representative of the U.S. in general. Because the prevalence of the metabolic syndrome can vary widely by demographic characteristics of study participants (29), caution should be used in extrapolating these results to other populations. Participants excluded due to missing information (7.7%) were more likely to be black and hypertensive but otherwise were similar to the analysis population. Sensitivity analyses performed by varying the proportion of these individuals classified as having CHD or metabolic syndrome at baseline and refitting Cox models resulted in negligible changes from reported HRs.

Finally, in Cox models that included all five components of the metabolic syndrome, elevated blood pressure, and low HDL cholesterol, but not elevated triglycerides, had statistically significant associations with CHD. Although it is not uncommon to include both HDL cholesterol and triglycerides as independent variables in CHD prediction models, such inclusion can result in biased estimates of effects because of the generally strong correlation between the two lipid measures and the relatively greater intraindividual variability of fasting triglycerides (30). Under such conditions, HDL cholesterol may be more likely to show an independent association with CHD as a function of its measurement qualities, rather than through any more direct causal link with CHD. A previously published assessment (31) of the reliability of fasting-state lipoprotein measurements in ARIC indicated that the reliability coefficient was higher for HDL cholesterol than triglycerides (0.94 vs. 0.85, respectively). As such, we cannot rule out this phenomenon in our results.

According to the American Diabetes Association (32), CVD is the cause of death in as many as 75–80% of individuals with diabetes. The “common soil” (33) hypothesis posits that diabetes is not a direct cause of CVD but rather one of several atherosclerotic sequelae arising from shared precursors. The metabolic syndrome may be a clinically recognizable and, in many cases, an early manifestation of such shared antecedents that increase the risk of diabetes and CVD among individuals. Identifying individuals with the metabolic syndrome and investigating effective treatment options may provide opportunities for intervention early in the disease pathways of both diseases. However, additional research is needed to refine the metabolic syndrome definition and to determine whether treatment of underlying conditions such as insulin resistance results in improved CHD prevention for patients with the metabolic syndrome beyond the level achieved by interventions currently targeted at its specific components.

This research was funded by National Institutes of Health Contracts N01-HC-55015, N01-HC-55016, N01-HC-55018, N01-HC-55019, N01-HC-55020, N01-HC-55021, and N01-HC-55022. A.M.M. received an NHLBI Research Service Award grant from the Cardiovascular Disease Epidemiology Training Program (5-T32-HL-07055) at the University of North Carolina at Chapel Hill, Chapel Hill, North Carolina.

We thank the staff and participants of the ARIC study for their important contributions.