Plasma Sphingomyelin and Subclinical Atherosclerosis: Findings from the Multi-Ethnic Study of Atherosclerosis

Transcription

1 American Journal of Epidemiology Advance Access published April 12, 2006 American Journal of Epidemiology Copyright ª 2006 by the Johns Hopkins Bloomberg School of Public Health All rights reserved; printed in U.S.A. DOI: /aje/kwj140 Original Contribution Plasma Sphingomyelin and Subclinical Atherosclerosis: Findings from the Multi-Ethnic Study of Atherosclerosis Jennifer Clark Nelson 1, Xian-Cheng Jiang 2, Ira Tabas 3,4, Alan Tall 3, and Steven Shea 3,5 1 Center for Health Studies, Group Health Cooperative, and Department of Biostatistics, University of Washington, Seattle, WA. 2 Department of Anatomy and Cell Biology, State University of New York, Downstate Medical Center, Brooklyn, NY. 3 Department of Medicine, Columbia University, New York, NY. 4 Departments of Anatomy and Cell Biology and of Physiology and Cellular Biophysics, Columbia University, New York, NY. 5 Department of Epidemiology, Columbia University, New York, NY. Received for publication August 29, 2005; accepted for publication December 8, Plasma sphingomyelin has been shown to be an independent risk factor for coronary heart disease, but the relation of plasma sphingomyelin to earlier, subclinical atherosclerotic disease has not been reported. The authors examined the association between plasma sphingomyelin and three measures of subclinical cardiovascular disease (carotid intimal-medial wall thickness, ankle-arm blood pressure index, and Agatston coronary artery calcium score) among 6,814 middle-aged, asymptomatic adults in the Multi-Ethnic Study of Atherosclerosis, which was initiated in The sphingomyelin level was positively correlated with lipids and the Framingham risk score (p < 0.01 for both), and the mean level was higher in women than men (50 (standard deviation (SD), 16) vs. 45 (SD, 15) mg/dl) (p < 0.01) and higher in never versus current smokers (49 (SD, 16) vs. 45 (SD, 17) mg/dl) (p < 0.01). Women with sphingomyelin levels of 60 or more mg/dl had more severe subclinical disease by all three measures than did the referent group with sphingomyelin levels of 39 or less mg/dl, although associations were not significant after multivariate adjustment for standard cardiovascular disease risk factors. Men with sphingomyelin levels of 60 or more mg/dl versus those with sphingomyelin levels of 39 or less mg/dl had higher calcium scores (135 vs. 99 Agatston units) (p ¼ 0.01). These observations are consistent with the hypothesis that plasma sphingomyelin is in the biologic pathway that mediates the risk for subclinical disease attributable to standard cardiovascular disease risk factors. arteriosclerosis; calcium; cardiovascular diseases; carotid artery diseases; cohort studies; coronary disease; sphingomyelins Abbreviations: MESA, Multi-Ethnic Study of Atherosclerosis; RR, relative risk; SD, standard deviation. Evidence from cell culture and animal studies and from one study in a human population supports the hypothesis that plasma sphingomyelin participates in the development of atherosclerosis. It has long been known that sphingomyelin accumulates in atheromas formed in human and animal models (1 8). In a comparison of human atherosclerotic lesions and plasma, low density lipoprotein cholesterol extracted from lesions is enriched in sphingomyelin (9 11), rendering it more susceptible to acid sphingomyelinase-induced aggregation, a proatherogenic process (12). A substantial amount of the sphingomyelin found in arteries and atherosclerotic lesions appears to arise from synthesis in the arterial tissue (13). However, even in atherosclerotic lesions, the apparent rate of sphingomyelin formation is relatively slow compared with the rate of total phospholipid synthesis (14), suggesting that Correspondence to Dr. Jennifer Clark Nelson, Group Health Cooperative, Metropolitan Park East, Suite 1600, 1730 Minor Avenue, Seattle, WA ( nelson.jl@ghc.org). 1

2 2 Nelson et al. other factors might also contribute to intimal sphingomyelin accumulation. Plasma lipoprotein sphingomyelin content may therefore contribute to arterial sphingomyelin accumulation, and several research findings are consistent with this hypothesis. The ratio of sphingomyelin to plasma cholesterol is increased fivefold in very low density lipoprotein cholesterol from hypercholesterolemic rabbits (15). Plasma sphingomyelin levels in mice deficient for apolipoprotein E (apolipoprotein E-knockout mice) are fourfold higher than in wild-type mice (16), which may partly explain the increased atherosclerosis in these mice (17, 18). Recently, it was reported that oral administration of myriocin, a specific inhibitor of de novo sphingomyelin biosynthesis, prevents development of atherosclerotic lesions in apolipoprotein E-knockout mice (19). At almost the same time, Hojjati et al. (20) reported that introperitoneal administration of myriocin significantly decreases plasma sphingomyelin levels and increases phosphotidylcholine levels, thus dramatically increasing the phosphotidylcholine/sphingomyelin ratio, and that such changes make a major contribution to the prevention of atherosclerosis in apolipoprotein E-knockout mice. These studies are an important step in understanding the associations among sphingolipid metabolism, lipoprotein metabolism, and atherogenesis and in considering how these associations might someday be translated into a novel antiatherogenic class of drugs (21). Finally, epidemiologic data from human subjects enrolled in a recent case-control study by Jiang et al. (22) showed that plasma sphingomyelin levels are positively and independently related to the presence of coronary artery disease. However, studies of the relation of plasma sphingomyelin to earlier, subclinical atherosclerotic disease in human populations have not been reported. We therefore investigated whether plasma sphingomyelin is an early atherogenic risk factor among participants in the Multi-Ethnic Study of Atherosclerosis (MESA). In particular, we examined the association between plasma sphingomyelin level and three measures of subclinical atherosclerosis, namely, carotid intimal-medial wall thickness, ankle-arm blood pressure index, and the Agatston coronary artery calcium score. MATERIALS AND METHODS Study participants MESA is a population-based study of 6,814 men and women aged years, without clinical cardiovascular disease at the time of entry, recruited from six US communities (Baltimore, Maryland; Chicago, Illinois; Forsyth County, North Carolina; Los Angeles County, California; northern Manhattan, New York; and St. Paul, Minnesota). The main objective of MESA is to determine risk factors for subclinical cardiovascular disease and its progression. Sampling and recruitment procedures have been described previously in detail (23). Subjects with symptoms or history of medical or surgical treatment for cardiovascular disease were excluded. Classification of race/ethnicity was based on self-identification by use of questions from the 2000 US Census questionnaire. Subjects who self-reported their race/ethnicity group as White or Caucasian, Black or African American, Chinese, or Spanish/Hispanic/Latino were asked to participate. Race/ethnicity was then categorized as White (non-hispanic), Black (non-hispanic), Chinese, and Hispanic. Subjects were enrolled between August 1, 2000, and July 30, Centrally trained and certified study staff members performed all measurements of participants and obtained informed consent. Institutional review board approval was obtained at all MESA sites. All data collection forms used in MESA can be found at the following website: Laboratory measurements Blood specimens from fasting participants were processed within 30 minutes of phlebotomy and frozen at ÿ70 C. Total cholesterol was analyzed by use of a cholesterol oxidase method (Roche Diagnostics, Indianapolis, Indiana) (laboratory coefficient of variation, 1.6 percent). High density lipoprotein cholesterol was measured in ethylenediaminetetraacetic acid/plasma by use of the cholesterol oxidase method (Roche Diagnostics) after precipitation of non-high density lipoprotein cholesterol with magnesium/dextran (laboratory coefficient of variation, 2.9 percent). The triglyceride level was measured in ethylenediaminetetraacetic acid/plasma by use of triglyceride GB reagent (Roche Diagnostics) on the Roche COBAS FARA centrifugal analyzer (laboratory coefficient of variation, 4.0 percent). Low density lipoprotein cholesterol was calculated in plasma specimens having a triglyceride value of less than 400 mg/dl with the use of the formula of Friedewald et al. (24). Serum glucose was measured by rate reflectance spectrophotometry on the Vitros analyzer (Johnson & Johnson Clinical Diagnostics, Inc., Rochester, New York) (laboratory coefficient of variation, 1.1 percent). C-reactive protein was measured by use of an immunonephelometric assay (N High Sensitivity CRP; Dade Behring, Inc., Deerfield, Illinois). For C-reactive protein, the intraassay coefficient of variation ranged from 2.3 to 4.4 percent, and the interassay coefficient of variation ranged from 2.1 to 5.7 percent. Plasma homocysteine was measured by a fluorescence polarization immunoassay (IMx homocysteine assay; Axis Biochemicals ASA, Oslo, Norway) with use of an Abbott IMx analyzer (Abbott Diagnostics, Abbott Park, Illinois) (laboratory coefficient of variation, percent). Enzymatic measurement of plasma sphingomyelin levels was carried out at Columbia University using a novel fourstep procedure (22). In the first step, bacterial sphingomyelinase hydrolyzed sphingomyelin to phosphorylcholine and N-acylsphingosine. Thereafter, the addition of alkaline phosphatase generated choline from phosphorylcholine. The newly formed choline was used to generate hydrogen peroxide in a reaction catalyzed by choline oxidase. Finally, with peroxidase as a catalyst, hydrogen peroxide was used together with phenol and 4-aminoantipyrine to generate a red quinone pigment, with an optimal absorption at 505 nm. The plasma sphingomyelin levels were measured in a blinded fashion, and the interassay coefficient of variation ranged from 2.5 to 3.1 percent.

3 Sphingomyelin and Subclinical Atherosclerosis 3 Measures of subclinical atherosclerosis Carotid ultrasonography was performed by use of highresolution, B-mode ultrasound (LOGIQ 700; General Electric Medical Systems, Waukesha, Wisconsin). Arterial wall thickness was measured by acquiring longitudinal images of both the right and left distal common carotid artery, as well as the site of maximal thickening in the internal carotid artery or in the carotid bulb. The site of maximal thickening was imaged from three angles: anterior, lateral, and posterior. Videotaped images were digitized and analyzed centrally at Tufts-New England Medical Center (Boston, Massachusetts). The internal carotid intimal-medial wall thickness was computed as the mean of the maximum thickness of the near and far walls from the left and right sides, taken across the three views (i.e., a mean of 12 measurements) (25). Ankle and arm blood pressures were assessed with a Doppler apparatus (EN50 LE 100; Nicolet Vascular, Inc., Golden, Colorado) with an earpiece and mercury sphygmomanometer. The ankle-arm blood pressure index was computed as the mean ratio across the left and right sides of the ankle to arm systolic pressure, where ankle pressure is the higher of posterior tibial or dorsalis pedis ankle pressures, and arm pressure is the average of the right and left brachial pressures or the higher brachial pressure if they differed by 10 mmhg or more. Calcified plaque in the coronary arteries was assessed by use of computed tomography. Technologists performed a standardized scanning protocol (26) using either a cardiacgated, electron-beam computed tomography scanner (Imatron C-150; Imatron, Inc., San Francisco, California) or a prospectively electrocardiogram-triggered multidetector system (LightSpeed; General Electric Medical Systems, Waukesha, Wisconsin; or Volume Zoom; Siemens, Erlanger, Germany). Images were read (27) and calibrated (28) centrally at the Harbor-UCLA Research and Education Institute (Torrance, California). The mean (across two successive scans at the same examination per participant) Agatston coronary artery calcium score was computed (29). Calcium prevalence was defined as a positive Agatston coronary artery calcium score (>0 Agatston units). Other measures Questionnaires administered by trained and certified research staff were used to assess age, gender, race/ethnicity, smoking status, educational level, aspirin use, and medication use for diabetes, lipid lowering, and hypertension. Height and weight were measured, and body mass index (weight (kg)/height (m) 2 ) was computed. Blood pressure was measured three times at 1-minute intervals after 5 minutes of rest in the seated position, with an appropriate cuff size, with a Dinamap model PRO 100 automated oscillometric blood pressure device (Critikon Company, LLC, Tampa, Florida) following a standardized protocol (30). The mean of the second and third resting systolic and diastolic blood pressures was computed. Hypertension was defined as systolic blood pressure of 140 or more mmhg, diastolic blood pressure of 90 or more mmhg, or selfreported high blood pressure and on treatment with medication for hypertension (31). Diabetes was defined as being on treatment with insulin or oral medication for diabetes or a fasting glucose measurement of 126 or more mg/dl (32). A resting 12-lead electrocardiogram was obtained with a Marquette MAC-PC instrument (Marquette Electronics, Inc., Milwaukee, Wisconsin). The presence of left ventricular hypertrophy was assessed using Novacode criteria (33). The Framingham risk score was computed to quantify the risk of incident coronary heart disease. A higher risk score corresponds to a higher estimated probability for developing coronary heart disease over a 10-year time period (34). Statistical analysis Scatterplot smoothers were used to graphically explore the association between sphingomyelin and other participant characteristics and risk factors. Exploration did not indicate evidence of nonlinear associations and, thus, bivariate tests for trend in the mean or prevalence of participant characteristics were performed across sphingomyelin groups (39 mg/dl, deciles 1 3; mg/dl, deciles 4 6; mg/dl, deciles 7 and 8; 60 mg/dl, deciles 9 and 10) by linear or logistic regression, respectively. A chi-squared test was used to test for race/ethnicity and smoking status differences between sphingomyelin groups. Correlations with sphingomyelin (defined as a continuous variable) were quantified using Pearson s correlation coefficients. Trend and correlation analyses were performed on the log scale for triglycerides and for C-reactive protein because of skew in the distributions. Mean sphingomyelin levels were compared between covariate groups by analysis of variance methods, and p values were adjusted for multiple comparisons with Scheffe s method. Multivariate linear regression was used to model the mean for each subclinical disease outcome (ankle-arm blood pressure index, carotid intimal-medial wall thickness, and Agatston coronary artery calcium score). The mean Agatston coronary artery calcium score was modeled only among participants with a positive score and was log transformed because of skew in the distribution. We modeled the probability of a positive Agatston coronary artery calcium score as an exponential function of risk factors and performed nonlinear least-squares estimation to obtain asymptotically unbiased estimates of the relative risk of a positive Agatston score. To account for misspecification of the variance, we computed model-robust (Huber-White) standard errors. We did not use logistic regression because the prevalence of calcium is common (~50 percent), and the odds ratios obtained from logistic regression models will overestimate the relative risk except when the relative risk is zero (35). To develop the multivariate models, we first used scatterplot smoothers to explore the association between each outcome and sphingomyelin as a continuous variable. This exploration motivated the selection of a simple set of sphingomyelin categories that accurately represented the continuous association. Thus, sphingomyelin was included in the models as the four-category variable described previously, for both ease of interpretation and consistency with the presentation of the univariate analyses. Inferences using this approach were similar to those using sphingomyelin as

4 4 Nelson et al. a continuous or more finely categorized variable, and thus our conclusions were robust to the choice of the functional form for sphingomyelin, including the choice of category cutpoints and the number of categories. Models were adjusted for Framingham risk factors. Conclusions were robust to the exclusion of lipid levels and the inclusion of C-reactive protein, homocysteine, and the use of medications for lipid lowering in the adjustment. Model results were stratified by gender. The unadjusted and covariate-adjusted mean carotid intimal-medial wall thickness, ankle-arm blood pressure index, and Agatston coronary artery calcium score are presented for each sphingomyelin group, and the relative risk for a positive Agatston score is presented for each sphingomyelin group relative to the referent group (sphingomyelin, 39 mg/dl). We tested for and found no significant (p < 0.05) interactions between sphingomyelin and other covariates. However, we further investigated interactions with covariates that were modestly suggestive (p < 0.20) by performing additional subgroup analyses. In particular, we examined models stratified by race/ethnicity and by smoking status. Results were similar to the main reported findings and thus are not presented. All statistical tests were two sided. Analyses were performed using SPSS (SPSS, Inc., Chicago, Illinois) and STATA (StataCorp LP, College Station, Texas) software. RESULTS Of the 6,814 MESA baseline examination participants, 6,703 had available plasma sphingomyelin data. Eighty-five participants with triglyceride values greater than 400 mg/dl were excluded. Among the 6,618 participants included in the analyses, 53 percent were women, 62 percent were of minority ethnic background, 15 percent were diabetic, and 44 percent were hypertensive; the mean age was 63 (standard deviation (SD), 10) years. Plasma sphingomyelin levels ranged from 17 to 125 mg/dl, and the distribution had a mild positive skew (skewness coefficient ¼ 0.79 (SD, 0.03)), with mean and median sphingomyelin values of 47 (SD, 14) mg/dl and 45 mg/dl, respectively. Approximately half of the participants had a positive Agatston coronary artery calcium score, and the mean score among those with calcium (n ¼ 3,307) was 292 (SD, 556) score units. The mean carotid intimal-medial wall thickness was 1.07 (SD, 0.60) mm, and mean ankle-arm blood pressure index was 1.11 (SD, 0.12). Subjects in higher deciles of the sphingomyelin level were at higher risk for cardiovascular disease based on the Framingham risk score (p trend < 0.001) (table 1). Increasing trends across increasing sphingomyelin groups were seen in mean age, body mass index, prevalence of hypertension, and systolic blood pressure level. Higher mean lipid levels (total, low density, and high density lipoprotein cholesterol and triglycerides) were also associated with a higher mean sphingomyelin level. The correlations between sphingomyelin and other lipid levels were modest but significant, with correlation coefficients highest for total cholesterol (r ¼ 0.23; p < 0.01) and triglycerides (r ¼ 0.20; p < 0.01) and less strong for low density lipoprotein (r ¼ 0.13; p < 0.01) and high density lipoprotein (r ¼ 0.08; p ¼ 0.02) cholesterol. An increasing trend across sphingomyelin groups (p < 0.001) and positive correlation with sphingomyelin (r ¼ 0.06; p < 0.01) were also observed for C-reactive protein. The modest difference in Framingham risk score across sphingomyelin groups was in part explained by a lower mean sphingomyelin level among men compared with women (45 (SD, 15) mg/dl vs. 50 (SD, 16) mg/dl; p < 0.01) (figure 1) and among current smokers. Current smokers had a lower mean level of sphingomyelin compared with former smokers who, in turn, had a lower mean sphingomyelin level than did never smokers (45 (SD, 17), 48 (SD, 16), and 49 (SD, 16) mg/dl, respectively; p < 0.01 for all pairwise smoking-group contrasts). Caucasians had a lower mean sphingomyelin level than did other ethnic groups (46 (SD, 16) mg/dl vs. 48 (SD, 14) mg/dl for Chinese Americans, 49 (SD, 16) mg/dl for African Americans, and 50 (SD, 17) mg/dl for Hispanics; p < 0.03 for all pairwise comparisons with Caucasians) (figure 1). Gender-stratified results of unadjusted and covariateadjusted models for each subclinical disease measure and plasma sphingomyelin are presented in tables 2 and 3. Multivariate analyses were restricted to participants with nonmissing cardiovascular disease risk factor data and nonmissing outcome data. To maximize the amount of information used in each analysis, we used all available data for each outcome (n ¼ 6,398 for carotid intimal-medial wall thickness; n ¼ 6,495 for ankle-arm blood pressure index; n ¼ 6,571 for Agatston coronary artery calcium prevalence; and n ¼ 3,285 for Agatston coronary artery calcium score among those with a positive score) as opposed to restricting all analyses to the subset of participants who had nonmissing data for all outcomes. p values indicate the significance of differences in subclinical disease from the referent sphingomyelin group (sphingomyelin, 39 mg/dl). As shown in table 2 and table 3, women in the highest sphingomyelin group had significantly more severe subclinical disease than did those in the referent group. Specifically, compared with women in the sphingomyelin referent group, women with sphingomyelin levels of 60 or more mg/dl had a thicker mean carotid intimal-medial wall thickness (1.08 vs mm; p ¼ 0.001), a lower mean ankle-arm blood pressure index (1.08 vs. 1.09; p < 0.001), a higher prevalence of Agatston coronary artery calcium (relative risk (RR) ¼ 1.28, 95 percent confidence interval: 1.15, 1.44), and a marginally higher mean Agatston coronary artery calcium score among those with some calcium (68 vs. 55; p ¼ 0.09). In men, those with sphingomyelin levels of 60 or more mg/dl versus 39 or less mg/dl had a higher prevalence of Agatston coronary artery calcium (RR ¼ 1.09, 95 percent confidence interval: 1.01, 1.19) and a higher mean Agatston coronary artery calcium score among those with some calcium (135 vs. 99 Agatston units; p ¼ 0.01). No associations were detected between the plasma sphingomyelin level and carotid intimalmedial wall thickness or the ankle-arm blood pressure index among men. After multivariate adjustment for standard cardiovascular disease risk factors, only the differences in mean Agatston coronary artery calcium score between sphingomyelin groups among men remained significant.

5 Sphingomyelin and Subclinical Atherosclerosis 5 TABLE 1. Demographic characteristics and standard cardiovascular disease risk factors for study participants by plasma sphingomyelin groups in the Multi-Ethnic Study of Atherosclerosis, initiated in mg/dl (n ¼ 2,178) mg/dl (n ¼ 1,857) Sphingomyelin mg/dl (n ¼ 1,302) 60 mg/dl (n ¼ 1,281) Gender (%) Women <0.001 Men Age, years (mean) 61.7 (10.2)y 61.5 (10.1) 63.6 (10.3) 65.3 (10.1) <0.001 Race (%) Caucasian <0.001z Chinese American African American Hispanic Smoking status (%) Never <0.001z Former Current Pack-years, if smoker (mean) 32 (41.9) 24 (20.2) 26 (23.1) 26 (26.6) 0.04 Framingham risk score (mean) 5.9 (3.4) 6.0 (3.7) 6.4 (3.6) 7.0 (3.6) <0.001 Body mass index, kg/m 2 (mean) 28.1 (5.4) 28.3 (5.4) 28.5 (5.5) 28.6 (5.6) Diabetic (%) Left ventricular hypertrophy by electrocardiogram (%) Hypertensive (%) <0.001 Systolic blood pressure, mmhg (mean) 125 (20.8) 125 (20.7) 128 (21.8) 129 (23.0) <0.001 Diastolic blood pressure, mmhg (mean) 73 (10.1) 72 (10.3) 72 (10.4) 71 (10.2) <0.001 Total cholesterol, mg/dl (mean) 186 (33.5) 193 (33.7) 194 (34.6) 206 (34.1) <0.001 Low density lipoprotein cholesterol, mg/dl (mean) 114 (30.5) 117 (31.2) 116 (31.5) 124 (32.1) <0.001 High density lipoprotein cholesterol, mg/dl (mean) 50 (13.8) 51 (14.3) 51 (15.2) 54 (16.3) <0.001 Triglyceride, mg/dl (median (interquartile range)) 100 (73 137) 115 (81 164) 114 (78 169) 123 (82 189) <0.001 Taking an antihypertensive (%) <0.001 Taking lipid-lowering medication (%) C-reactive protein, mg/liter (median (interquartile range)) 1.7 ( ) 1.9 ( ) 1.9 ( ) 2.4 ( ) <0.001 Homocysteine, lmol/liter (mean) 9.4 (3.5) 9.2 (4.0) 9.2 (3.0) 9.7 (4.5) 0.09 p trend * * p values denote the significance level for trend across sphingomyelin groups. Because of skew, triglyceride and C-reactive protein levels were log transformed, and tests were performed on the log scale. y Numbers in parentheses, standard deviation except where noted otherwise. z A chi-squared test of independence was performed. n ¼ 826. DISCUSSION These findings from the MESA cohort are the first to examine the associations between plasma sphingomyelin and subclinical atherosclerosis in a population-based sample of asymptomatic men and women. Among women, we observed that those in the highest sphingomyelin group had more severe subclinical disease by all three subclinical outcome measures compared with those in the referent group. In men, the highest sphingomyelin group had a higher prevalence and extent of Agatston coronary artery calcium compared with the referent group, but no associations were detected between plasma sphingomyelin and carotid intimalmedial wall thickness or the ankle-arm blood pressure

6 6 Nelson et al. FIGURE 1. Mean plasma sphingomyelin level (mg/dl) and 95% confidence intervals by gender, ethnic group, age group (45 54, 55 64, 65 74, years), and smoking status group in the Multi-Ethnic Study of Atherosclerosis, initiated in index. Further, only the difference in mean Agatston coronary artery calcium score between these sphingomyelin groups among men remained statistically significant after regression adjustment for standard cardiovascular risk factors (138 vs. 107 Agatston units; p ¼ 0.03). These observations are consistent with a biologic model in which plasma sphingomyelin partially mediates the associations between traditional risk factors and subclinical atherosclerosis by participating in the intermediate pathways. The only other study in a human population of sphingomyelin as a risk factor for cardiovascular disease that has been reported was conducted by Jiang et al. (22) as a casecontrol study of 279 angiographically documented cases of coronary artery disease and 277 similarly documented controls. In that study, conducted primarily in men, the plasma sphingomyelin level was assayed with the same methods as those used here and was associated with the presence of coronary artery disease after adjustment for age, diabetes, smoking, hypertension, low density and high density lipoprotein cholesterol, triglycerides, remnant cholesterol, fibrinogen, and C-reactive protein (for the third sphingomyelin quartile: adjusted RR ¼ 2.83 (p < 0.001); for the fourth sphingomyelin quartile: RR ¼ 2.59 (p ¼ 0.001), compared with the referent (lowest) sphingomyelin quartile). The carotid and peripheral vascular arterial beds were not examined in the study of Jiang et al. Our findings from the MESA cohort may be interpreted in relation to this earlier study in at least three ways. First, the plasma sphingomyelin level, like other risk factors for atherosclerosis, may affect different vascular beds to different degrees. Among men in our study, there was evidence of an association with the Agatston coronary artery calcium score but not with indicators of subclinical atherosclerosis in the carotid or peripheral arterial beds. Second, the Agatston

7 Sphingomyelin and Subclinical Atherosclerosis 7 TABLE 2. Association of plasma sphingomyelin with carotid intimal-medial wall thickness and with the ankle-arm blood pressure index, by gender, in the Multi-Ethnic Study of Atherosclerosis, initiated in 2000 Sphingomyelin group, by gender and adjustment Carotid intimal-medial wall thickness coronary artery calcium score in men, which was computed only in the subset in whom coronary artery calcium was present (i.e., an Agatston score of >0), may be an indicator of more severe or advanced subclinical disease, compared with carotid intimal-medial wall thickness or the ankle-arm blood pressure index, and also compared with the Agatston score in women. Thus, it is possible that sphingomyelin is associated with more advanced but not less advanced subclinical disease, consistent with findings in the study by Jiang et al. (22), where coronary atherosclerosis was measured by angiography. Third, it is possible that the association that we observed between the sphingomyelin level and the mean Agatston coronary artery calcium score in men is a chance finding. The adjustment for multiple comparisons was applied to the multiple levels of sphingomyelin being compared with the referent category but not to the multiple subclinical endpoints considered or the two gender groups. Thus, it is possible that the finding of a significant association between sphingomyelin and the Agatston coronary artery calcium score in men may represent type I error. No. Mean (mm) 95% confidence interval (mm) p value No. Mean Ankle-arm blood pressure index 95% confidence interval p value* Women 3,369 3,410 Unadjustedy 39 mg/dl , 1.03 Referent , 1.10 Referent mg/dl , , mg/dl , , mg/dl , , 1.08 <0.001 Adjustedz 39 mg/dl , 1.11 Referent , 1.09 Referent mg/dl , , mg/dl , , mg/dl , , Men 3,029 3,085 Unadjustedy 39 mg/dl 1, , 1.17 Referent 1, , 1.15 Referent mg/dl , , mg/dl , , mg/dl , , Adjustedz 39 mg/dl 1, , 1.22 Referent 1, , 1.13 Referent mg/dl , , mg/dl , , mg/dl , , * Indicates significance of differences compared with the referent group (sphingomyelin, 39 mg/dl). y Unadjusted means are presented. z Covariate-adjusted means are evaluated at the mean values of the following model covariates: age, race, smoking status, diabetes status, systolic and diastolic blood pressure, and low and high density lipoprotein cholesterol. Our overall sample size was large (n ¼ 6,618) and adequate, even after partition by gender and into sphingomyelin groups, to detect clinically important associations. For example, in the analysis for the presence of coronary artery calcium, there were 912 women and 1,248 men in the referent group versus 855 women and 417 men in the highest sphingomyelin group. Thus, it is extremely unlikely that the negative findings in our study are due to low statistical power. On the other hand, the methods used in MESA to assess atherosclerotic burden are indirect and may have limited correlation with coronary atherosclerosis. Further investigation of the association between plasma sphingomyelin and atherosclerosis with measures of atheroma volume by intravascular ultrasound or other newer modalities may be warranted. As in the earlier study by Jiang et al. (22), we found that plasma sphingomyelin was associated with lipid risk factors. This reflects the fact that sphingomyelin is a constituent of plasma lipoproteins, along with cholesterol and triglyceride. Sphingomyelin is particularly enriched in remnant lipoproteins (36), but these would not be present at high

8 8 Nelson et al. TABLE 3. Association of plasma sphingomyelin with the Agatston coronary artery calcium score and with the presence of coronary calcium, by gender, in the Multi-Ethnic Study of Atherosclerosis, initiated in 2000 Sphingomyelin group, by gender and adjustment Agatston coronary artery calcium score No. Mean 95% confidence interval p value No. Presence of coronary calcium* Relative risk 95% confidence interval p valuey Women 1,385 3,470 Unadjustedz 39 mg/dl , 66 Referent 912 Referent Referent Referent mg/dl , , mg/dl , , mg/dl , , 1.44 <0.001 Adjusted 39 mg/dl , 87 Referent 912 Referent Referent Referent mg/dl , , mg/dl , , mg/dl , , Men 1,900 3,101 Unadjustedz 39 mg/dl , 112 Referent 1,248 Referent Referent Referent mg/dl , , mg/dl , , mg/dl , , Adjusted 39 mg/dl , 124 Referent 1,248 Referent Referent Referent mg/dl , , mg/dl , , mg/dl , , * The presence of coronary calcium is indicated by a coronary artery calcium score greater than zero. y Indicates significance of differences compared with the referent group (sphingomyelin, 39 mg/dl). z Unadjusted means or relative risks are presented. Covariate-adjusted means and relative risks are evaluated at the mean values of the following model covariates: age, race, smoking status, diabetes status, systolic and diastolic blood pressure, and low and high density lipoprotein cholesterol. concentrations in fasting plasma. Jiang et al. did not report data describing the associations between sphingomyelin and other traditional cardiovascular risk factors, but in the present study, we found that the plasma sphingomyelin level was also positively correlated with a subset of these standard risk factors, including age, body mass index, and systolic blood pressure level. We also observed, somewhat counterintuitively, lower mean plasma sphingomyelin levels in men compared with women, in current smokers compared with nonsmokers, and in Caucasians compared with other ethnic groups. These associations with sphingomyelin level have not previously been reported or replicated to the best of our knowledge, nor have mechanistic studies been done to address specific hypotheses that might explain them. The main strengths of our study include the diverse, population-based sample of participants, the extensive noninvasive characterization of subclinical cardiovascular disease, and the large sample size that provided statistical power to examine associations within subgroups. The main limitation is the lack of data describing incident cardiovascular and coronary heart disease events in the MESA cohort. It will be possible to perform prospective confirmatory analyses using the MESA sample in the future as participants are followed over time for cardiovascular endpoints including coronary heart disease. It will also be possible to examine whether or not plasma sphingomyelin is associated with progression of subclinical disease, since some subclinical disease measures are being repeated at follow-up examinations. In summary, in this large representative sample with quantitative measures of subclinical atherosclerosis in the coronary, carotid, and peripheral arterial vessels, we found that the plasma sphingomyelin level is associated with subclinical disease in all three of these arterial beds in women. Although these associations were not significant after multivariate adjustment for standard cardiovascular disease risk factors, the unadjusted data provide evidence in support of the hypothesis that plasma sphingomyelin is in the biologic pathway that mediates the risk for subclinical disease attributable to some traditional cardiovascular disease risk factors.

9 Sphingomyelin and Subclinical Atherosclerosis 9 In addition, we found an association between plasma sphingomyelin and the Agatston coronary artery calcium score in men among those with Agatston scores greater than zero, consistent with the recent case-control study by Jiang et al. (22), which included mostly male subjects and had angiographic coronary atherosclerosis as the endpoint. ACKNOWLEDGMENTS MESA is supported by contracts N01-HC through N01-HC and N01-HC from the National Heart, Lung, and Blood Institute. Additional support for this research was provided by grant P50-HL56984 (A. Tall and I. Tabas). The authors thank the other MESA investigators and staff for their valuable contributions. A full list of participating MESA investigators and institutions can be found at Conflict of interest: none declared. REFERENCES 1. Steele JM, Kayden HJ. The nature of phospholipids in human serum and atheromatous vessels. Trans Assoc Am Physicians 1955;68: McCandless EL, Zilversmit DB. The effect of cholesterol on the turnover of lecithin, cephalin, and sphingomyelin in the rabbit. Arch Biochem Biophys 1956;62: Smith EB. Intimal and medial lipids in human aorta. Lancet 1960;1: Newman HA, McCandless EL, Zilversmit DB. The synthesis of C 14 -lipids in rabbit atheromatous lesions. J Biol Chem 1961;236: Phillips GB, Dodge JT. Composition of phospholipids and of phospholipid fatty acids of human plasma. J Lipid Res 1967; 8: Portman OW, Illingworth RD. Arterial metabolism in primates. Primates Med 1976;9: Hakomori SI. Glycosphingolipids in cellular interaction, differentiation, and oncogenesis. Annu Rev Biochem 1981;50: Kummerow FA, Cook LS, Wasowicz E, et al. Changes in the phospholipid composition of the arterial cell can result in severe atherosclerotic lesions. J Nutr Biochem 2001;12: Hoff HF, Morton RE. Lipoproteins containing apo B extracted from human aortas. Structure and function. Ann N Y Acad Sci 1985;454: Guyton JR, Klemp KF. Development of the lipid-rich core in human atherosclerosis. J Lipid Res 1996;16: Schissel SL, Jiang XC, Tweedie-Hardman J, et al. Secretory sphingomyelinase, a product of the acid sphingomyelinase gene, can hydrolyze atherogenic lipoproteins at neutral ph. J Biol Chem 1998;273: Tabas I. Secretory sphingomyelinase. Chem Phys Lipids 1999;102: Zilversmit DB, McCandless EL, Jordan PH, et al. The synthesis of phospholipids in human atheromatous lesions. Circulation 1961;23: Eisenberg S, Stein Y, Stein O. Phospholipases in arterial tissue. IV. The role of phosphatide acyl hydrolase, lysophosphatide acyl hydrolase, and sphingomyelin choline phosphohydrolase in the regulation of phospholipid composition in the normal human aorta with age. J Clin Invest 1969;48: Rodriguez JL, Ghiselli GC, Torreggiani D, et al. Very low density lipoproteins in normal and cholesterol-fed rabbits: lipid and protein composition and metabolism. Atherosclerosis 1976;23: Jeong TS, Schissel SL, Tabas I, et al. Increased sphingomyelin content of plasma lipoproteins in apolipoprotein E knockout mice reflects combined production and catabolic defects and enhances reactivity with mammalian sphingomyelinase. J Clin Invest 1998;101: Zhang SH, Reddick RL, Piedrahita JA, et al. Spontaneous hypercholesterolemia and arterial lesions in mice lacking apolipoprotein E. Science 1992;258: Plump AS, Breslow JL. Apolipoprotein E and the apolipoprotein E-deficient mouse. Annu Rev Nutr 1995;15: Park TS, Panek RL, Mueller SB, et al. Inhibition of sphingomyelin synthesis reduces atherogenesis in apolipoprotein E-knockout mice. Circulation 2004;110: Hojjati MR, Li Z, Zhou H, et al. Effect of myriocin on plasma sphingolipid metabolism and atherosclerosis in apoe-deficient mice. J Biol Chem 2005;280: Tabas I. Sphingolipids and atherosclerosis: a mechanistic connection? A therapeutic opportunity? Circulation 2004; 110: Jiang XC, Paultre F, Pearson TA, et al. Plasma sphingomyelin level as a risk factor for coronary artery disease. Arterioscler Thromb Vasc Biol 2000;20: Bild DE, Bluemke DA, Burke GL, et al. Multi-Ethnic Study of Atherosclerosis: objectives and design. Am J Epidemiol 2002;156: Friedewald WF, Levy RI, Fredrickson DS. Estimation of the concentration of low density lipoprotein cholesterol in plasma without use of the preparative ultracentrifuge. Clin Chem 1972;18: O Leary DH, Polak JF, Kronmal RA, et al. Use of sonography to evaluate carotid atherosclerosis in the elderly: the Cardiovascular Health Study. Stroke 1991;22: Carr JJ, Nelson JC, Wong ND, et al. Measuring calcified coronary plaque with cardiac CT in population based studies: the standardized protocol of the Multi-Ethnic Study of Atherosclerosis (MESA) and Coronary Artery Risk Development in Young Adults (CARDIA). Radiology 2005;234: Yaghoubi S, Tang W, Wang S, et al. Offline assessment of atherosclerotic coronary calcium from electron beam tomograms. Am J Card Imaging 1995;9: Nelson JC, Kronmal RA, Carr JJ, et al. Computed tomography coronary calcium measures from images adjusted for attenuation differences. Radiology 2005;235: Agatston AS, Janowitz WR, Hildner FJ, et al. Quantification of coronary artery calcium using ultrafast computed tomography. J Am Coll Cardiol 1990;15: Multi-Ethnic Study of Atherosclerosis (MESA) field center manual of operations (exam 1). Seattle, WA: Collaborative Health Studies Coordinating Center, Department of Biostatistics, University of Washington, The sixth report of the Joint National Committee on prevention, detection, evaluation and treatment of high blood pressure. Arch Intern Med 1997;157: Genuth S, Alberti KG, Bennett P, et al. Follow-up report on the diagnosis of diabetes mellitus. Diabetes Care 2003;26:

10 10 Nelson et al. 33. Rautaharju PM, Park LP, Chaitman BR, et al. The Novacode criteria for classification of ECG abnormalities and their clinically significant progression and regression. J Electrocardiol 1998;31: Wilson PW, D Agostino RB, Levy D, et al. Prediction of coronary heart disease using risk factor categories. Circulation 1998;97: Greenland S. Model-based estimation of relative risks and other epidemiologic measures in studies of common outcomes and in case-control studies. Am J Epidemiol 2004;160: Schlitt A, Hojjati MR, von Gizycki H, et al. Serum sphingomyelin levels are related to the clearance of postprandial remnant-like particles. J Lipid Res 2005;46: