MESA is a population-based study of subclinical cardiovascular disease in 6,814 participants of White, Black, Hispanic, and Chinese ethnicities and free of overt cardiovascular disease at entry. MESA participants were initially recruited and examined in 2000–2002 from persons ages 45–84 years in six U.S. communities as previously described [12
]. Protocols were approved by the participating Institutional Review Boards and all participants gave written informed consent.
Coronary calcium was measured by multidetector computed tomographic scanning and electron beam computed tomography as previously described [12
]. Kappa statistics for intra-and inter-reader reproducibility of coronary calcium prevalence were both 0.92. Intra-class correlation coefficients for intra- and inter-reader reproducibility of coronary calcium scores exceeded 0.99. Carotid ultrasonography was performed and read centrally as previously described [12
]. Intra-class correlation coefficients for intra-reader reproducibility of common and internal carotid IMT both exceeded 0.98 and for inter-reader reproducibility were 0.87 and 0.94, respectively.
Prevalence of coronary calcium, geometric mean coronary calcium scores, and mean carotid IMT were compared across ethnicities using chi-square and analysis of variance, respectively. Due to the large proportion of persons with coronary calcium scores = 0 (coronary calcium absent) and the resulting strong departures from normality in distribution of these scores [16
], bivariate associations with scores as continuous variables are presented only for those with coronary calcium present.
The hypothesis of ethnic differences in the relationship of carotid IMT to coronary calcium was tested in a two-part model, by regressing ln-transformed IMT in mm as the dependent variable on ln-transformed total coronary calcium score (ln [score + 1], including those with 0 calcium) in Agatston units (AU), along with an indicator for presence of coronary calcium, as independent variables. Since neither coronary calcium nor carotid wall thickness is accepted as a “gold standard” of cardiovascular disease risk, testing was repeated by regressing ln coronary calcium score, when >0, on carotid IMT. In addition, presence of coronary calcium was regressed separately on carotid IMT. The probability of a positive calcium score was modeled as an exponential function of the covariates using nonlinear least squares to obtain asymptotically unbiased estimates of the relative risk of a positive CAC score. As results of modeling carotid IMT as a function of coronary calcium, and coronary calcium as a function of carotid IMT, were generally similar, only the former are presented because they allow simultaneous modeling of coronary calcium prevalence and scores in these cross-sectional data.
The primary variable of interest was ethnicity used as a class variable, defined by three indicator variables, with Whites as the comparison group. The interaction of ethnicity with both presence of coronary calcium and ln-transformed calcium score was tested to assess ethnic differences in the relationship of coronary calcium to carotid atherosclerosis. If the 6-degrees of freedom interaction term (3 ethnicity terms × 2 calcium terms) was significant at p < 0.10, separate interactions between ethnicity and presence of calcium, and between ethnicity and calcium score, were tested.
Interactions were also tested between age and ethnicity and between age and each of coronary calcium presence, coronary calcium score, and carotid IMT. Significant interactions were found between age and ethnicity and between age and coronary calcium score. Due to the complexity of interaction models, results are presented both unadjusted and after full adjustment accounting for the age interactions. Interactions between sex and each of age and ethnicity, and between sex and coronary calcium, were also tested. When significant interactions with sex were present, models were stratified by sex.
To determine whether any ethnic differences in the relationships of coronary calcium to carotid IMT were explained by differences in prevalence of the risk factor covariates, a final level of adjustment included adding these covariates. Covariates were selected based on prior published risk factor associations with either measure of atherosclerosis and included age, sex, education, body mass index, systolic and diastolic blood pressures, anti-hypertensive medication use, diabetes, LDL- and HDL-cholesterol, cholesterol-lowering medication use, C-reactive protein, IL-6, fibrinogen, current/former smoking, pack-years of smoking, and estrogen use in women. All analyses were done using SPSS, version 13 (SPSS, Inc. Chicago, IL) or STATA, version 8 (STATA Corp., College Station, TX).