There is a strong inverse association between HDL cholesterol

Transcription

1 Elevated High-Density Lipoprotein Cholesterol Levels Are Protective Against Plaque Progression A Follow-Up Study of 1952 Persons With Carotid Atherosclerosis The Tromsø Study Stein Harald Johnsen, MD; Ellisiv B. Mathiesen, MD, PhD; Einar Fosse, MD; Oddmund Joakimsen, MD, PhD; Eva Stensland-Bugge, MD, PhD; Inger Njølstad, MD, PhD; Egil Arnesen, MD Background There is an inverse relationship between HDL cholesterol and coronary heart disease. Experimental studies have indicated that HDL cholesterol may exert an antiatherogenic effect by inducing regression of atherosclerotic plaques and by turning lipid-rich plaques into more fibrotic lesions. In this prospective, population-based ultrasound study, we investigated how HDL cholesterol relates to carotid plaque progression. Methods and Results The study included 1952 men and women aged 25 to 82 years who had at least 1 plaque present in the right carotid artery at baseline examination (1994). All plaque images were computer processed to yield a measure of plaque area in square millimeters and echogenicity, expressed as the gray-scale median. After 7 years of follow-up, a new ultrasound screening was performed, and the changes in plaque area and echogenicity were assessed. In a multivariable adjusted model, HDL cholesterol, age, systolic blood pressure, and current smoking were independent predictors of plaque growth. For a 1-SD (0.41 mmol/l) lower HDL cholesterol level, mean (SE) plaque area increased by 0.93 mm 2 (0.44 mm 2 ; P 0.03). When users of lipid-lowering drugs were excluded from analysis, the HDL estimate was strengthened ( 1.46 mm 2, P 0.002). Although plaque area increased in 70% of cases, and most plaques became more echogenic over the follow-up interval, the plaques that became more echolucent grew more in size than those that became more echogenic (P 0.002). Conclusions This study shows that a high level of HDL cholesterol reduces plaque growth in subjects with preexisting carotid atherosclerosis. Transformation of the plaque mass into higher echogenicity is associated with reduced growth. Our findings may indicate that HDL cholesterol stabilizes plaques and counteracts their growth by reducing their lipid content and inflammation. (Circulation. 2005;112: ) Key Words: cholesterol, HDL plaque, carotid atherosclerosis carotid arteries ultrasonics There is a strong inverse association between HDL cholesterol and risk of coronary heart disease (CHD). 1,2 An inverse relationship between HDL cholesterol and carotid atherosclerosis has also been demonstrated, although the results are not as consistent as for CHD. 3 6 The antiatherogenic effect of HDL may involve reversal of LDL transport from the vessel wall back to the liver, inhibition of lipoprotein oxidation, and direct protection of the vessel wall from damages. 7 Low levels of HDL cholesterol are associated with echolucent plaques in cross-sectional ultrasound studies. 8,9 Such plaques are characterized by elevated contents of lipid and necrotic debris and are predisposed to fissuring, ulceration, hemorrhage, and superimposed thrombosis, features that cause clinical events. 10,11 Little is known about how plaque morphology interferes with plaque growth. In a previous cross-sectional study from our group, presence of echolucent plaques was associated with higher degree of stenosis, which suggests that echolucent plaques are more likely than fibrous and calcified plaques to develop into advanced stenosis. 8 If one presumes that the natural history of atherosclerotic plaques is to become more fibrotic, calcified, and thereby echogenic as time goes by, 12,13 one might postulate that plaques that remain echolucent will tend to grow faster than plaques that become echogenic. Recent data suggest that repeated measurements of plaque area are sensitive markers of atherosclerotic progression. 14,15 In this population-based study, we used computerized image analysis to measure changes in total plaque area and plaque echogenicity after 7 years of follow-up. We measured HDL cholesterol together with other risk factors for atherosclerosis Received December 1, 2004; revision received March 22, 2005; accepted April 11, From the Institute of Community Medicine (S.H.J., E.B.M., E.F., I.N., E.A.), University of Tromsø, and Department of Neurology (E.B.M., O.J., E.S.-B.), University Hospital of Northern Norway, Tromsø, Norway. Correspondence to Stein Harald Johnsen, Institute of Community Medicine, University of Tromsø, N-9037 Tromsø, Norway. Stein.Johnsen@ism.uit.no 2005 American Heart Association, Inc. Circulation is available at DOI: /CIRCULATIONAHA

2 Johnsen et al HDL Cholesterol and Plaque Progression 499 and examined the relationship between HDL levels, change in plaque area, and change in echogenicity. Methods Subjects The Tromsø Study is a single-center, population-based prospective study with repeated health surveys of inhabitants in the municipality of Tromsø, Norway. The main focus is on cardiovascular diseases. In the fourth survey conducted in 1994 to 1995 (baseline), all subjects aged 55 to 74 years and a 5% to 10% samples in the other 5-year birth cohorts 24 years of age were invited to an ultrasonographic examination of the right carotid artery. A total of 6889 attended, and ultrasound examination was performed in 6727 persons. 16 At the fifth survey in 2001, all subjects who attended the baseline ultrasound study and were still alive and residing in Tromsø were invited to a follow-up ultrasound screening. Between baseline and follow-up screening, 532 persons (7.9%) had died, 271 (4%) had migrated, and 956 (14%) did not attend in the reexamination. Another 110 subjects attended the follow-up visit but were not rescanned because of logistic problems, which brought the total number of subjects examined at both baseline and follow-up to Of these, 2248 persons (46.3%) had at least 1 plaque present at baseline. In 296 persons, there were missing values on plaque area or echogenicity because of low image quality or missing data. Thus, 1952 subjects (1059 men and 893 women) were included in the present study. The Norwegian Data Inspectorate licensed all data. The Regional Committee for Research Ethics approved the study. Written informed consent was obtained from all participants. Cardiovascular Risk Factors Standardized measurements of height, weight, blood pressure, nonfasting lipids, monocytes, and fibrinogen were done. Blood pressure was recorded with an automatic device (Dinamap Vital Signs Monitor) by specially trained personnel. Nonfasting serum total cholesterol and triglycerides were analyzed by enzymatic colorimetric methods with commercial kits (CHOD-PAP for cholesterol and GPO-PAP for triglycerides; Boehringer-Mannheim). Serum HDL cholesterol was measured after precipitation of lower-density lipoprotein with heparin and manganese chloride. Fibrinogen was measured with the PT-Fibrinogen reagent (Instrumentation Laboratory). Monocytes and white blood cells were counted with automated cell counters by standard techniques. Total cholesterol, HDL cholesterol, triglycerides, and blood pressure were measured twice at an interval of 4 to 12 weeks, and the average of these values was used in analyses. Blood sample analyses were performed at the Department of Clinical Chemistry, University Hospital of Tromsø. A questionnaire on previous myocardial infarction and stroke, prevalent angina pectoris and diabetes mellitus (yes/no), use of antihypertensive and lipid-lowering drugs, and cigarette smoking (never/previously/currently, number of cigarettes per day) was enclosed in the letter of invitation. CHD was defined as prevalent angina pectoris or previous myocardial infarction. Ultrasonography and Measures of Atherosclerosis At baseline and follow-up, the same ultrasound imaging system and transducer (Acuson Xp10 128, ART upgraded, with a 7.5-MHz linear-array transducer, aperture size 38 mm) were used. The B-mode image adjustment parameters were preset to fixed values and were not changed during the course of either survey. With the subject in a supine position, head turned slightly to the left, the right carotid artery was scanned with several different angles of insonation, both longitudinally and transversely, from just above the clavicle to as far distal to the bifurcation as possible. A plaque was defined as a local protrusion of the vessel wall into the lumen of at least 50% compared with the adjacent intima-media thickness. In each subject, a maximum of 6 plaques were registered in the near and far walls of the common carotid, bifurcation, and internal carotid, respectively. For each plaque, a still image was recorded with the transducer parallel to the vessel wall and as perpendicular to the point of maximum plaque thickness as possible, with the regional expansion selection set to 38 mm 20 mm. All recordings were done on a Panasonic 7650 video player with Super VHS tape. The sonographers were blinded to the laboratory data and data from the questionnaires. Digital Image Acquisition and Standardization Still image of each plaque was digitized offline with a PC with the Matrox Meteor II frame-grabber card and Matrox Intellicam version 2.07 software at a resolution of pixels. Digital images were saved in RGB format as uncompressed TIFF files. The following steps were performed with the imaging software Adobe Photoshop version 7.0 after all images were digitized: The color information was discarded, with the image to changed grayscale mode. Plaque echogenicity was assessed by the grayscale median (GSM). GSMs of the lumen and adventitia were obtained by the method described previously, 17 with the following modifications: to select areas of lumen and adventitia, we used a preset region of interest ( pixels 1.8 mm 2, and pixels 0.9 mm 2, respectively). Each region of interest was moved to several different locations to find the lowest GSM value in the lumen, and the highest GSM value in the adventitia of the same vessel wall as the plaque and the GSM of the lumen and the adventitia were obtained in the histogram function. Each plaque was then outlined with the Lasso tool, with no feathering or antialiasing, and the cropped image was saved as an uncompressed and nonlayered TIFF image file. The image was then standardized with the levels function, with the recorded GSM values for lumen and adventitia as the low and high input values and the low- and high-output values preset to 1 and 200, respectively. The standardized image was subsequently saved in the same file format as above. The preset output value of 200 for adventitia was predetermined through a pilot study in which a number of images visually assessed to be of high quality were measured by the technique described above, with GSM recordings of the adventitia showing a mean value of 200. In addition to GSM, the histogram function generates pixel values of the cropped plaque. For the resolution used in the present study, a plaque area of 167 pixels corresponded to 1 mm 2. In subjects with more than 1 plaque, the sum of plaque areas was taken as the total plaque area, and the GSM of the total plaque area was estimated as a weighted mean of the GSM value of each single plaque. Plaques included in an acoustic shadow were analyzed if 50% of the plaque area provided real acoustic information, and only this part of the plaque was subjected to analysis of echoes. Figure 1 illustrates examples of plaques with high and low echogenicity (GSM). The plaque area is given as the pixel value. The interobserver mean arithmetic difference (SD) of plaque area was 1.0 (4.4) mm 2, and the limits of agreement were 8.6 mm 2. The intraobserver mean arithmetic difference (SD) for sonographer 1 was 0.2 (3.1) mm 2, and the limits of agreement were 6.1 mm 2. The corresponding values for sonographer 2 were 0.01 (3.8) mm 2 and 7.5 mm 2. The mean GSM of the standardized plaques in the interobserver and intraobserver studies was 51. The interobserver mean arithmetic difference (SD) of GSM was 1.7 (9.8), and the limits of agreement were The intraobserver mean arithmetic difference (SD) for sonographer 1 was 0.1 (6.1), and the limits of agreement were The corresponding values for sonographer 2 were 0.9 (8.7) and In the continuation of this report, the more familiar term echogenicity will be preferred and used in the same meaning as GSM. Correspondingly, a decrease or increase in GSM values will be referred to as a decrease or increase in echogenicity. Statistical Analyses Change in total plaque area ( plaque area) was defined as the dependent variable in the regression models. Risk factors measured at baseline were used as independent variables. We first assessed the univariable effect of risk factors using a general linear model (the GLM procedure in the SAS statistical software). In these analyses,

3 500 Circulation July 26, 2005 TABLE 1. Selected Baseline Characteristics of the Study Population (n 1952) Continuous variables Age, y 63.2 (6.8) BMI, kg/m (3.8) Serum lipids, mmol/l Total cholesterol 7.00 (1.19) HDL cholesterol 1.51 (0.41) Triglycerides 1.68 (0.86) HbA Ic,% 5.5 (0.7) Systolic blood pressure, mm Hg (20.3) Inflammatory markers Fibrinogen, mmol/l 3.49 (0.84) Monocyte count, 10 9 /L 0.60 (0.18) White blood cells, 10 9 /L 6.95 (1.83) Categorical variables, % Male sex 54.3 Current smoking 32.8 CHD 16.7 Stroke 2.8 Diabetes mellitus 2.9 Use of antihypertensive drugs 16.9 Use of lipid-lowering drugs 3.5 BMI indicates body mass index. Values are unadjusted means (SD) or percentages. Figure 1. A, Echogenic plaque with histogram of grayscale frequency distribution of pixels in selected area (plaque). GSM is 93 and pixel value is 2068, corresponding to an area of 12.4 mm 2. B, Echolucent plaque. GSM is 15, and pixel value is 1818, corresponding to an area of 10.9 mm 2. plaque area was calculated in quintiles of risk factor levels. Linear trend across quintiles was tested by linear regression. The independent relationship between risk factors and plaque area was tested by multiple linear regression. Model assumptions were carefully assessed by residual analysis. Interaction with sex was examined with plaque area as the dependent variable and the following independent variables: sex, risk factor, and sex risk factor. Twosided probability values 0.05 were considered statistically significant. The SAS software package was used for all statistical analyses (version 8). Results Table 1 presents selected baseline characteristics of the study population. In 1994, the mean age (range) in men and women was 62.2 (36.0 to 78.0) and 64.4 (37.0 to 81.0) years, respectively. Men had a higher monocyte count and more CHD but lower HDL cholesterol, total cholesterol, body mass index, systolic blood pressure, and fibrinogen than women. The proportion of current smokers was equal among sexes. During the follow-up period, certain risk factors changed significantly. Body mass index increased by 0.7 kg/m 2. There was an increase in systolic blood pressure of 7.2 mm Hg. There was a reduction in total cholesterol and HDL cholesterol by 0.68 mmol/l and 0.06 mmol/l, respectively. From 1994 to 2001, the percentage of current smokers was reduced by 7%. In the same period, the use of antihypertensive and lipid-lowering drugs increased by 17% and 20%, respectively. At baseline, 1059 men and 893 women had plaque present (Table 2). Carotid plaque area increased by age in both sexes, and at any age, men had more atherosclerotic plaque than women. Mean total plaque area (SE) at baseline was 24.1 (0.6) mm 2 in men and 17.5 (0.6) mm 2 in women (P ). In the follow-up period, 1362 persons (70% in both sexes) had an increase in total plaque area. The mean increase was 9.0 (0.6) mm 2 in men and 7.3 (0.6) mm 2 in women (P 0.06). Men had somewhat lower echogenicity at baseline than women (P for differences between sexes). Approximately 60% had an increase in echogenicity, and this increase was similar among sexes. Predictors of Changes in Plaque Area Table 3 presents mean plaque growth in quintiles of risk factor levels. For age, HDL cholesterol (inverse), systolic blood pressure, and smoking, there were significant linear dose-response relationships between these variables and plaque progression. Figure 2 illustrates the relationship between HDL cholesterol levels and plaque growth. The plaque area decreased from 9.0 mm 2 in the lowest HDL quintile to 6.0 mm 2 in the highest HDL quintile. Although there was a significant linear trend for plaque area across strata of HDL cholesterol, plaque area decreased mainly in the upper quintile. The relationship between HDL cholesterol and plaque growth was stronger in men than in women;

4 Johnsen et al HDL Cholesterol and Plaque Progression 501 TABLE 2. Changes in Plaque Characteristics Between Baseline and Follow-Up Age n Plaque Area at Baseline, mm 2 Plaque Area, mm 2 Baseline Echogenicity Echogenicity Men Total, age-adjusted (SE) (0.6) 9.0 (0.6) 44 (0.6) 6 (0.9) Women Total, age adjusted (SE) (0.6) 7.3 (0.6) 47 (0.7) 7 (0.9) P for sex difference however, because the trends in each sex did not differ significantly (P 0.1 for sex difference after multiple adjustments), we only present the results of the pooled analyses. Adjusted for the other risk factors, a 1-SD (0.41 mmol/l) greater HDL cholesterol level was associated with a mm 2 lower change in plaque area (Table 4). There was no interaction between HDL cholesterol and sex. In nearly 20% of the subjects included in the present study, treatment with lipid-lowering drugs and antihypertensive drugs had been initiated between baseline and follow-up examination. When the analysis was repeated with lipid-lowering drugs and TABLE 3. Plaque Growth (mm 2 ) by s of Risk Factor Level at Baseline First Second antihypertensive drugs (ever-users) included in the model as covariates, no change in HDL estimate was observed. The same was true when subjects who had ever been treated with antihypertensive drugs were excluded from the data set. When we excluded ever-users of lipid-lowering drugs (n 442), the HDL estimate was strengthened ( 1.46 mm 2, P ) A strengthening of the HDL estimate was also seen when the analysis was run with adjustment for plaque area at baseline ( 1.15, P 0.005). Controlling for change in echogenicity did not influence the relationship between HDL cholesterol and plaque growth. Continuous Variables (Range) Third Fourth Categorical Variables Fifth P trend Yes No P Age, y* (36 81) BMI, kg/m 2 ( ) Serum lipids, mmol/l Total cholesterol ( ) HDL cholesterol ( ) Triglycerides ( ) HbA Ic, % ( ) Systolic blood pressure, mm Hg ( ) Inflammatory markers Fibrinogen, mmol/l ( ) Monocyte count, 10 9 /L ( ) White blood cell count, 10 9 /L ( ) Smoking, no. of cigarettes/day (0 60) Echogenicity, GSM (3 174) Smoking Use of antihypertensive drugs Use of lipid-lowering drugs Values are age- and sex-adjusted means. *Adjusted for sex only.

5 502 Circulation July 26, 2005 Figure 2. Plaque growth in quintiles of HDL cholesterol. Values are age-and sex-adjusted. HDL cholesterol levels in quintiles were 1.16, 1.16 to 1.36, 1.37 to 1.54, 1.55 to 1.83, and 1.83 mmol/l. Figure 3. Plaque growth by changes in echogenicity. Change in echogenicity is measured as GSM. Values are age-and sexadjusted. GSM values in quartiles were less than 9, 9 to6, 7 to 21, and 21. Negative values indicate plaques have become more echolucent; positive values indicate plaques have become more echogenic. Changes in Echogenicity in Relation to Changes in Plaque Area There were no differences in plaque growth across strata of baseline echogenicity (Table 3). The change in plaque area was inversely correlated with change in echogenicity (r 0.1, P ). Figure 3 shows plaque growth by quartiles of echogenicity in the follow-up period (P for trend 0.002). The trend over quartiles of echogenicity was similar in both sexes (P 0.2 for sex difference, after multiple adjustments). In the first quartile (all of which decreased in echogenicity), plaque growth was 9.5 mm 2 ; in the second quartile, plaque growth was 9.4 mm 2 (54% decreased in echogenicity); and in the third and fourth quartiles (all of which increased in echogenicity), plaque growth was 7.5 and 6.4 mm 2, respectively. This means that plaques that became more echogenic had a lower growth rate than those that became more echolucent. Correspondingly, the proportions with regressed lesions ( plaque area 0) within each TABLE 4. HDL Cholesterol as a Predictor of Plaque Area Progression quartile of echogenicity were 0.24, 0.30, 0.33, and 0.34 (P for trend ). Discussion The present study showed that HDL cholesterol, age, systolic blood pressure, and smoking were independent predictors of plaque growth. Although plaque area increased in 70% of cases, and most plaques became more echogenic over the follow-up interval, plaques that became more echogenic had a lower growth rate than those that became more echolucent. Our group has previously demonstrated that echolucent plaques are associated with lower levels of HDL cholesterol and a higher degree of stenosis. 8 We found no association between baseline echogenicity and later plaque growth. The prospective findings in the present study rather appear to show that it is the change in echogenicity that is associated with plaque growth, which suggests that plaques that remain echolucent may have a higher growth potential than plaques that remain echogenic. Model I Model II Model III Model IV Risk Factor * (SE) P * (SE) P * (SE) P * (SE) P HDL cholesterol 0.95 (0.43) (0.44) (0.44) (0.48) Age 1.47 (0.43) (0.45) (0.48) Male sex 1.07 (0.89) (0.91) (0.99) 0.2 Total cholesterol 0.28 (0.44) (0.51) 0.2 Systolic blood pressure 1.21 (0.44) (0.48) 0.03 Smoking, yes vs no 3.18 (0.92) (1.00) *Values are regression coefficients (SE) expressed in mm 2 for a 1-SD change in continuous variables and for presence vs absence of categorical variables. Model I was unadjusted. Model II was adjusted for age and sex. Model III was adjusted for age, sex, total cholesterol, systolic blood pressure, and smoking. Model IV excluded persons who had ever used lipid-lowering drugs (n 442).

6 Johnsen et al HDL Cholesterol and Plaque Progression 503 To the best of our knowledge, this is the first study to show an association between HDL cholesterol and plaque growth in humans. In a previous study on cholesterol-fed rabbits, intravenous administration of HDL and VHDL induced regression of established atherosclerotic lesions. 18 In another study, increased levels of apolipoprotein A-I, the major protein component of HDL, induced regression of atherosclerotic lesions in mice. 19 Regressed lesions were less rich in foam cells and macrophages and more fibrotic. Our findings are in accordance with these observations. We found that plaque mass that increased most in echogenicity (Figure 3, 4th quartile) had the lowest growth rate (6.4 mm 2 ) and the highest proportion with regressed lesions (34%). This is plausible if one mechanism of plaque growth and plaque regression is accumulation and removal of lipids. Most of the plaques were heterogeneous, containing a mixture of echogenic and echolucent components. Lipids appear echolucent in ultrasound, and accumulation of lipids within a plaque may therefore increase the echolucent proportion of the plaque as it grows. Conversely, removal of lipids may increase the echogenic proportion, so the regressed plaque will appear more echogenic (fibrous). By transforming the plaques into more echogenic (fibrotic) lesions, HDL particles may stabilize them and counteract their growth. The mechanism may be by reversing cholesterol transport or active LDL removal from the vessel wall and from the macrophage foam cells. 18 HDL particles contain apolipoprotein A-I, apolipoprotein E, paraoxonase, and the lipid-processing enzymes LCAT (lecithin-cholesterol acyltransferase) and CETP (cholesteryl ester transfer protein). In addition to roles in cholesterol efflux, there are data implicating these proteins in metabolic, oxidative, or inflammatory processes that affect atherosclerosis. 7 Approximately 20% of the study population was treated with lipid-lowering drugs (statins) after the baseline examination in 1994, of which 63% had CHD. Exclusion of ever-users of statins from the analysis strengthened the HDL estimate significantly. Statins mainly decrease LDL cholesterol but also have some raising effect on the HDL cholesterol level. In persons treated with statins, HDL cholesterol was 1.44 mmol/l compared with 1.53 mmol/l in never-users (P ). Plaque growth was also lower in statin users compared with never-users (7.1 versus 8.5 mm 2, P 0.1). Plaque growth was mainly reduced in the topmost quintile of HDL cholesterol, which may indicate a threshold effect more than a truly linear relationship. There also appeared to be a threshold effect of smoking in which the main difference was between the first and second quintiles (nonsmokers versus smokers), whereas for systolic blood pressure, there appeared to be a true linear relationship. We found that carotid plaque area increased by age in both sexes, and at any age, men had more plaque than women. A similar finding has been reported previously. 12,20 Furthermore, over the follow-up period, men also had a higher plaque growth rate than women. We have measured plaque progression as a continuous variable. Measuring the dependent variable on a continuous scale is beneficial, because the power to quantify the effect of risk factors and interaction among risk factors is increased compared with categorical classification. 21 Furthermore, computerized GSM analysis may have better reproducibility and be more accurate than visual characterization and appears to be a more objective and quantitative method for assessing plaque echogenicity Low GSM values in carotid plaques are predictive of future cerebral infarction. 25 Single measurements of risk factors will tend to underestimate the strength of the relationship between risk factors and disease because of measurement errors and fluctuations of risk factor level within individuals. At baseline, total cholesterol, HDL cholesterol, and blood pressure were measured twice to obtain more precise estimates. We have presented the results without adjustment for baseline levels of atherosclerosis. Adjustment for baseline values of the dependent variable without correction for measurement error may introduce bias that leads to an overestimation of the risk estimates. 26 There are some limitations to the present study. Because of loss of follow-up, severely ill or disabled persons may have been underrepresented. This possible selection bias would probably weaken the true relationship between risk factors and plaque progression. Another potential limitation to this study is that only 1 carotid artery was studied. Inclusion of the left carotid and femoral arteries might have given a better description of the individual plaque burden. Measurement errors of plaque area are present in this study. Even if the plaque analyses were computer assisted, the outlining of the plaques and the selection of regions of interest in the lumen and adventitia were operator dependent and therefore introduced measurement variability into the study. GSM gives a general description of the gray-level distribution of the plaque but does not take into account texture features like heterogeneity and dark areas close to lumen. We conclude that a high level of HDL cholesterol is inversely related to plaque growth. Transformation of the plaque mass into higher echogenicity is associated with reduced plaque growth, and the HDL effect may be due to change in plaque morphology. The findings in the present study support the hypothesis that HDL cholesterol may stabilize vulnerable plaques and counteract their growth by reducing their lipid content and inflammation. Acknowledgments This study was supported by grants from the Norwegian Research Council and was conducted in collaboration with the Norwegian Health Screening Services, Oslo, Norway. References 1. Miller NE, Thelle DS, Førde OH, Mjøs OD. The Tromsø Heart Study: high-density lipoprotein and coronary heart-disease: a prospective casecontrol study. Lancet. 1977;1: Gordon T, Castelli WP, Hjortland MC, Kannel WB, Dawber TR. High density lipoprotein as a protective factor against coronary heart disease: the Framingham Study. Am J Med. 1977;62: Tell GS, Crouse JR, Furberg CD. Relation between blood lipids, lipoproteins, and cerebrovascular atherosclerosis: a review. Stroke. 1998; 19: Heiss G, Sharrett AR, Barnes R, Chambless LE, Szklo M, Alzola C. Carotid atherosclerosis measured by B-mode ultrasound in populations: associations with cardiovascular risk factors in the ARIC study. Am J Epidemiol. 1991;134:

7 504 Circulation July 26, Bots ML, Breslau PJ, Briet E, de Bruyn AM, van Vliet HH, van den Ouweland FA, de Jong PT, Hofman A, Grobbee DE. Cardiovascular determinants of carotid artery disease: the Rotterdam Elderly Study. Hypertension. 1992;19: Prati P, Vanuzzo D, Casaroli M, Di Chiara A, De Biasi F, Feruglio G, Touboul PJ. Prevalence and determinants of carotid atherosclerosis in a general population. Stroke. 1992;23: van Vlijmen BJ, Herz J. Gene targets and approaches for raising HDL. Circulation. 1999;99: Mathiesen EB, Bønaa KH, Joakimsen O. Low levels of high-density lipoprotein cholesterol are associated with echolucent carotid artery plaques: the Tromsø Study. Stroke. 2001;32: Kofoed SC, Grønholdt ML, Bismuth J, Wilhjelm JE, Sillesen H, Nordestgaard BG. Echolucent, rupture-prone carotid plaques associated with elevated triglyceride-rich lipoproteins, particularly in women. J Vasc Surg. 2002;36: Falk E, Shah PK, Fuster V. Coronary plaque disruption. Circulation. 1995;92: Mathiesen EB, Bønaa KH, Joakimsen O. Echolucent plaques are associated with high risk of ischemic cerebrovascular events in carotid stenosis: the Tromsø Study. Circulation. 2001;103: Joakimsen O, Bønaa KH, Stensland-Bugge E, Jacobsen BK. Age and sex differences in the distribution and ultrasound morphology of carotid atherosclerosis: the Tromsø Study. Arterioscler Thromb Vasc Biol. 1999; 19: Stary HC, Chandler AB, Dinsmore RE, Fuster V, Glagov S, Insull W Jr, Rosenfeld ME, Schwartz CJ, Wagner WD, Wissler RW. A definition of advanced types of atherosclerotic lesions and a histological classification of atherosclerosis: a report from the Committee on Vascular Lesions of the Council on Arteriosclerosis, American Heart Association. Arterioscler Thromb Vasc Biol. 1995;15: Spence JD Ultrasound measurement of carotid plaque as a surrogate outcome for coronary artery disease. Am J Cardiol. 2002;89:10B 15B. 15. Barnett PA, Spence JD, Manuck SB, Jennings JR. Psychological stress and the progression of carotid artery disease. J Hypertens. 1997;15: Joakimsen O, Bønaa KH, Mathiesen EB, Stensland-Bugge E, Arnesen E. Prediction of mortality by ultrasound screening of a general population for carotid stenosis: the Tromsø Study. Stroke. 2000;31: Elatrozy T, Nicolaides A, Tegos T, Zarka AZ, Griffin M, Sabetai Ml. The effect of B-mode ultrasonic image standardisation on the echodensity of symptomatic and asymptomatic carotid bifurcation plaques. Int Angiol. 1998;17: Badimon JJ, Badimon L, Fuster V. Regression of atherosclerotic lesions by high density lipoprotein plasma fraction in the cholesterol-fed rabbit. J Clin Invest. 1990;85: Tangirala RK, Tsukamoto K, Chun SH, Usher D, Pure E, Rader DJ. Regression of atherosclerosis induced by liver-directed gene transfer of apolipoprotein A-I in mice. Circulation. 1999;100: Iemolo F, Martiniuk A, Steinman DA, Spence JD. Sex differences in carotid plaque and stenosis. Stroke. 2004;35: Spence JD, Eliasziw M, DiCicco M, Hackam DG, Galil R, Lohmann T. Carotid plaque area: a tool for targeting and evaluating vascular preventive therapy. Stroke. 2002;33: El Barghouty NM, Levine T, Ladva S, Flanagan A, Nicolaides A. Histological verification of computerised carotid plaque characterisation. Eur J Vasc Endovasc Surg. 1996;11: Iwamoto T, Shinozaki K, Kiuchi A, Umahara T, Takasaki M. Evaluation of B-mode ultrasonographic images of carotid lesions by computer analysis as compared with visual assessment. J Stroke Cerebrovasc Dis. 2003;12: Arnold JA, Modaresi KB, Thomas N, Taylor PR, Padayachee TS. Carotid plaque characterization by duplex scanning: observer error may undermine current clinical trials. Stroke. 1999;30: El Barghouty N, Nicolaides A, Bahal V, Geroulakos G, Androulakis A. The identification of the high risk carotid plaque. Eur J Vasc Endovasc Surg. 1996;11: Yanez ND III, Kronmal RA, Shemanski LR. The effects of measurement error in response variables and tests of association of explanatory variables in change models. Stat Med. 1998;17: