We used baseline data from the Multi-Ethnic Study of Atherosclerosis, a multi-center study of four US race/ethnicities (White, Chinese, Black and Hispanic). Of the 6814 men and women free of clinical cardiovascular disease at baseline, a random sample of 1000 individuals had measurements of circulating biomarkers. This analysis included 997 individuals who had measurements available for oxidized low density lipoprotein (ox-LDL, a marker for oxidative stress) or for the following markers of endothelial damage: soluble intercellular adhesion molecule-1 (sICAM-1), CD40 ligand (CD40L), soluble thrombomodulin (sTM), soluble E-selectin (sESEL) and von Willebrand factor (vWF).
MetS was defined according to the modified NCEP-ATP-III criteria (Grundy et al., 2005
) using medical history, anthropometric measurements, seated blood pressure, and fasting glucose and lipid panels. Persons having ≥3 of the following abnormalities met the criteria for MetS: (1) waist circumference ≥0.88 m in women, ≥1.02 m in men, (2) serum triglycerides ≥1.695 mmol/L, (3) HDL-C ≤1.295 mmol/L in women, ≤1.036 mmol/L in men, (4) systolic blood pressure ≥135 and/or diastolic blood pressure ≥85 mmHg or the use of antihypertensive medications, (5) fasting glucose ≥ 5.55 mmol/L, or taking anti-diabetic medications.
Assessment of demographic and clinical variables
Personal history including current smoking during the past 30 days, education, exercise, demographic data, and medical history were collected using interviewer-administered forms. Medication use was confirmed by examination of medication containers, if available. Height and weight were measured using a stadiometer and platform balance, respectively. Waist girth was measured horizontally at the level of the umbilicus in the standing position using an anthopometric tape. Serum lipid profile and plasma glucose were measured from fasting blood draws and analyzed at the Core Laboratory at University of Vermont, Burlington, VT.
Levels of oxLDL were measured by Dr. Holvoet's laboratory, with an mAb-4E6-based competition ELISA (analytical CV 7.4-8.3%).(Holvoet et al., 1998
) All other biomarkers were assayed at the Laboratory for Clinical Biochemistry Research (University of Vermont, Burlington, VT) as follows: sICAM-1 by ELISA (Parameter Human sICAM-1 Immunoassay; R&D Systems, Minneapolis, MN, Laboratory; Analytical CV 5.0%), CD40L by ultra-sensitive quantitative sandwich enzyme immunoassay (Quantikine Human soluble CD40 Ligand Immunoassay; R&D Systems, Minneapolis, MN, CV 4.5-6.4%), sTM by ELISA (Asserachrom Thrombomodulin, Diagnostica Stago; Asnières-sur-Seine, France, CV 12%), soluble E-selectin by high sensitivity quantitative sandwich enzyme immunoassay (Parameter Human sE-Selectin Immunoassay; R&D Systems, Minneapolis, MN, CV 4.7-8.8%), vWf by immunoturbidimetric assay on the STAR analyzer (Liatest vWF; Diagnostica Stago, Parsippany, NJ, intra- and inter-assay CV 3.7-4.5%).
Assessment of subclinical atherosclerosis
Coronary artery calcium was measured using computerized tomography using either electron beam tomography or helical tomography (Carr et al., 2005
). The Agatston score (Agatston et al., 1990
) was used to quantify coronary artery calcification (CAC). The intimal-medial thickness (IMT) of the common and internal carotid arteries was assessed using B-mode ultrasound (O'Leary et al., 1991
). Multiple views were obtained of the left and right sides, and the maximal IMT in each view was averaged for the common (C-IMT) and internal (I-IMT) carotid arteries.
The difference in demographic variables by MetS status was tabulated. Differences were evaluated using t-tests for continuous variables and chi-squared tests for categorical variables.
The medians and interquartile ranges of biomarker variables were tabulated by MetS status, and age, sex, and race-adjusted differences in log-transformed biomarker levels were tested using linear regression. For multivariable regression analyses, the biomarker levels were log-transformed because they had a right-skewed distribution. A large number of individuals had no detectable CAC, and those with detectable CAC had a right-skewed variable distribution. Thus CAC was analyzed in two stages: in the first stage, the relative prevalence of detectable CAC was modeled as a dichotomous variable; in the second stage, log-transformed CAC was used in regression analysis only among those with detectable CAC. The standard error of association statistics for IMT may be incorrectly estimated because of non-normality of the IMT distributions. We have thus estimated bootstrapped standard errors for these regression analyses using 1000 resampled datasets.
Adjusted models assessing the association of subclinical atherosclerosis variables with MetS and the biomarkers were performed using general linear model methods. Presence of dichotomous detectable CAC was modeled with a logarithmic link function, and Gaussian error, to obtain prevalence ratios associated with the independent variables. All other dependent variables were continuous, and modeled using ordinary least squared linear regression. Choice of covariates for adjustment included the demographic covariates age, sex and race, as well as those cardiovascular risk factors not represented among the metabolic abnormalities included in the metabolic syndrome definition. BMI was not included as a covariate because it is strongly correlated with waist girth (Spearman correlation = 0.85 within the sample).
In analysis of MetS (as defined by NCEP) dichotomizing each MetS component for classification and making a simple count of abnormalities may lose information regarding independent associations. Thus, we repeated all models above including the individual components of MetS, in place of MetS presence or absence.
The sICAM-1 assay immunoreactivity depends on the K56M polymorphism (rs5491) (Register et al., 2004
), which is frequent in African Americans but not in Caucasian, East Asian or Hispanic populations (NCBI, 2009
). Thus sICAM-1 analyses were also performed excluding the African-American sample, as done by others (Tang et al., 2007
Because many biomarkers are tested, there are issues regarding inflation of false positive hypothesis tests due to multiple testing. Thus age, sex and race adjusted associations of 6 biomarkers is considered primary analysis, with the significance level for p-values being set at 0.05/6 = 0.0083. All other covariate models are exploratory and explanatory.
Sensitivity analyses: In a multi-site multiethnic study, there is a possibility that study site and socioeconomic status may confound associations. Furthermore habitual physical activity may confound the associations. We performed all primary association analyses including study site, educational status (as a proxy for socioeconomic status), and habitual intentional weekly exercise (METS/week) as covariates to assess whether the associations remain after these adjustments. Analyses using fasting glucose and insulin measures instead of the metabolic syndrome to estimate insulin resistance (HOMA-IR, Matthews et al., 1985
) or its inverse transformation (QUICKI, Katz et al., 2000
) are presented in the online appendix.
The demographic and cardiovascular risk factor characteristics are shown by MetS status in . MetS was present in a third of the sample. The individuals with MetS were somewhat older, had a lower percentage of men, and were more likely to be Black or Hispanic than those without MetS. There was no difference in prevalence of current smoking or high cholesterol levels. The predominant components of the MetS were a large waist and hypertension.
Characteristics of the Multi-Ethnic Study of Atherosclerosis random sample at baseline (2000-2002)
All biomarkers were higher in those with than in those without MetS, including after adjustment for age, sex and race (), except for vWF. The association of shows that sICAM and E-selectin were associated with greater prevalence of CAC, when adjusted for age, sex, race, total cholesterol and current smoking. The correlation of oxLDL with total cholesterol was of a magnitude that neither total cholesterol nor oxLDL were separately associated with CAC prevalence, though they were jointly associated using likelihood ratio test. Thus we tabulated the association in a model not adjusted for total cholesterol. These associations remained significant after adjustment for either the metabolic syndrome or the five individual metabolic abnormalities. None of the biomarkers were associated with the magnitude of calcification among those with detectable calcium.
Median and interquartile ranges of circulating biomarkers, and differences according to metabolic syndrome (MetS), Multi-Ethnic Study of Atherosclerosis, 2000-2002.
Measures of association of biomarker levels [95% confidence intervals] with the presence of detectable coronary artery calcium (CAC) and magnitude of calcification (if detectable), Multi-Ethnic Study of Atherosclerosis, 2000-2002.
E-selectin was associated with thicker IMT in the common carotid artery after age, sex, race, total cholesterol and smoking-adjustment, but this association was marginal on adjustment for MetS () or the five individual metabolic abnormalities. Ox-LDL was associated with thicker internal carotid IMT, when adjusted for age, sex, race and current smoking, and also after further adjustment for MetS or the five individual metabolic abnormalities.
Measures of association of biomarker levels [95% confidence intervals] with common and internal carotid intimal-medial thickness (IMT), Multi-Ethnic Study of Atherosclerosis, 2000-2002
After adjusting for age, sex, total cholesterol, current smoking and race, MetS was not associated with prevalence of CAC, but it was associated with a higher level of CAC if detectable (ratio of geometric means of CAC levels 1.55) (). This association was unchanged when biomarker variables were added as covariates. None of the individual dichotomized components of the metabolic syndrome were significantly associated with CAC prevalence or extent in any model in this subsample of MESA. Those with MetS had significantly greater IMT in both the common and internal carotid arteries. These associations remained after adjustment for the markers of oxidative stress and endothelial dysfunction. The individual components large waist circumference (β = 0.04 mm, p<0.001), low HDL-C (β = 0.02 mm, p=0.025), high blood pressure (β = 0.03 mm, p=0.001) and glucose abnormality (β = 0.02 mm, p = 0.065) were associated with common carotid IMT, significantly or at the borderline and these associations were not much changed by addition of biomarkers (large waist circumference: β = 0.04 mm, p<0.001, low HDL-C: β = 0.02 mm, p=0.030, high blood pressure: β = 0.04 mm, p=0.001, and glucose abnormality: β = 0.02 mm, p = 0.049).
Measures of association [95% confidence intervals] of subclinical atherosclerosis according to the presence of metabolic syndrome (MetS), Multi-Ethnic Study of Atherosclerosis, 2000-2002
For internal IMT there were significant of borderline associations with the individual components of low HDL (β = 0.07 mm, p=0.057) and high blood pressure (β = 0.09 mm, p=0.002), and on addition of biomarkers in the model only the association of high blood pressure (β = 0.10 mm, p=0.002) remained significant.
Sensitivity analyses: All associations for ox-LDL and sICAM-1 remained unchanged in models adjusting for site, education level (categorized by completion of school and various kinds of higher education) and exercise (total intentional weekly exercise in METS). The associations of E-selectin with coronary calcium prevalence became significant at the borderline (p=0.1), and with common carotid intimal medial thickness became non-significant (p=0.15) in models adjusting for education level.