Author + information
- Received June 29, 2012
- Revision received August 13, 2012
- Accepted September 4, 2012
- Published online December 11, 2012.
- Connie W. Tsao, MD⁎,†,
- Sarah Rosner Preis, ScD⁎,‡,
- Gina M. Peloso, PhD⁎,‡,
- Shih-Jen Hwang, PhD⁎,
- Sekar Kathiresan, MD§,
- Caroline S. Fox, MD, MPH⁎∥,
- L. Adrienne Cupples, PhD⁎,‡,
- Udo Hoffmann, MD, MPH¶ and
- Christopher J. O'Donnell, MD, MPH⁎,§,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Christopher J. O'Donnell, Framingham Heart Study, 73 Mount Wayte Avenue, Suite 2, Framingham, Massachusetts 01702
Objectives This study evaluated the association of timing of lipid levels and lipid genetic risk score (GRS) with subclinical atherosclerosis.
Background Atherosclerosis is a slowly progressive disorder influenced by suboptimal lipid levels. Long-term versus contemporary lipid levels may more strongly impact the development of coronary artery calcium (CAC).
Methods Framingham Heart Study (FHS) Offspring Cohort participants (n = 1,156, 44% male, 63 ± 9 years) underwent serial fasting lipids (low-density lipoprotein cholesterol [LDL-C], high-density lipoprotein, and triglycerides), Exam 1 (1971 to 1975) to Exam 7 (1998 to 2001). FHS Third Generation Cohort participants (n = 1,954, 55% male, 45 ± 6 years) had fasting lipid profiles assessed, 2002 to 2005. Computed tomography (2002 to 2005) measured CAC. Lipid GRSs were computed from significantly associated single-nucleotide polymorphisms. The association between early, long-term average, and contemporary lipids, and lipid GRS with elevated CAC was assessed using logistic regression.
Results In FHS Offspring, Exam 1 and long-term average as compared with Exam 7 lipid measurements, including untreated lipid levels, were strongly associated with elevated CAC. In the FHS Third Generation, contemporary lipids were associated with CAC. The LDL-C GRS was associated with CAC (age-/sex-adjusted odds ratio: 1.14, 95% confidence interval: 1.00 to 1.29, p = 0.04). However, addition of the GRS to the lipid models did not result in a significant increase in the odds ratio or C-statistic for any lipid measure.
Conclusions Early and long-term average lipid levels, as compared with contemporary measures, are more strongly associated with elevated CAC. Lipid GRS was associated with lipid levels but did not predict elevated CAC. Adult early and long-term average lipid levels provide important information when assessing subclinical atherosclerosis and cardiovascular risk.
Coronary artery calcium (CAC), a measure of calcified coronary atherosclerotic plaque, is strongly and consistently associated with cardiovascular morbidity and mortality (1–7). Dyslipidemia is an atherosclerotic risk factor (8), and remote and long-term averaged lipid levels in young adulthood are associated with the presence and extent of CAC by middle age (9,10). The relations of lipid levels with CAC may be modulated by the time course and duration of elevated lipid levels (1,2,4,9–12). Few studies have examined the association of timing and duration of lipid levels with CAC in an older population.
Recent genome-wide association studies (GWAS) have discovered 95 genetic loci associated with low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and triglycerides (TG) in multiple adult populations (13). TG genetic risk score was found to improve risk prediction over long-term TG level for elevated TG in adulthood (14), lending support to the impact of the genetic risk score (GRS) on ultimate lipid levels. Given the association of lipid levels with atherosclerosis, GWAS provides a unique tool to examine the impact of lipid GRS on measures of atherosclerosis such as CAC. We hypothesized that elevated lipid levels over many years reflected in both long-term lipid levels and lipid GRS, would be more strongly associated with CAC than contemporary lipid levels. We thus sought to examine the association between early, long-term average, and contemporary measures of LDL-C, HDL-C, and TG, and their corresponding lipid GRS, with extent of CAC in a broad community population.
The Framingham Heart Study (FHS) is a longitudinally followed prospective cohort study of community-dwelling adults evaluated every 4 to 6 years, as described previously (15,16). This study included participants from the Offspring Cohort (original FHS cohort members' children and their spouses) and Third Generation Cohort (children of the Offspring). Included FHS Offspring attended at least 3 of 7 serial examinations, Exam 1 (1971 to 1975) through Exam 7 (1998 to 2001), inclusive of Exams 1 and 7. Third Generation participants were evaluated at Examination 1 (2002 to 2005). A subset of participants from the Offspring and Third Generation Cohorts, weighted toward larger FHS families, received multidetector computed tomography (MDCT) to assess CAC in 2002 to 2005. Participants were excluded for weight ≥160 kg, age <35 years in men or <40 in women, and pregnancy. A total of 3,529 FHS members (1,422 Offspring and 2,093 Third Generation) participated in the MDCT substudy. Individuals were excluded if they had missing data for CAC measurements, lipid measurements, or prevalent coronary heart disease (myocardial infarction, angina, coronary insufficiency, coronary artery bypass graft, or angioplasty). A total of 3,110 individuals (n = 1,156 Offspring 63 ± 9 years of age, and n = 1,954 Third Generation 45 ± 6 years of age) were included in the analysis.
Measurement of coronary artery calcium
Coronary artery calcium was imaged using an 8-slice MDCT scanner (LightSpeed Ultra, General Electric, Milwaukee, Wisconsin) (17). Forty-eight contiguous 2.5-mm-thick slices were acquired. Each participant was scanned twice. Using a dedicated offline workstation (Aquarius, Terarecon, San Mateo, California), an experienced reader assessed the presence and amount of CAC. A calcified lesion was identified as an area of ≥3 connected pixels of attenuation >130 Hounsfield units, and an Agatston score was calculated as described (18). For the present analysis, we defined CAC using ≥75th percentile age-, sex-, and cohort-specific cut points based on a healthy referent sample (19).
Measurement of lipid levels
At each study visit, each participant underwent a routine physical examination, medical history interview, and fasting laboratory tests, including total cholesterol, HDL-C, and TG. LDL-C was calculated according to the Friedewald equation (20). For the present analysis, we considered lipid measurements at 3 time points: 1) early, at Offspring Exam 1; 2) long-term average, the mean of all available lipid levels from Offspring Exams 1 to 7; and 3) contemporary, at both Offspring Exam 7 (1998 to 2001) and Third Generation Exam 1 (2002 to 2005), contemporaneously with MDCT. The use of lipid-lowering medication was assessed at each examination.
Genotyping methods and lipid genetic risk scores
FHS participants in the Offspring and Third Generation Cohorts with available genomic DNA with cell line backups and with prior genotyping were genotyped for lipid single-nucleotide polymorphisms (SNPs) in the SNP Health Association Resource (SHARe) project using an Illumina Golden Gate assay (21). Subjects also had 550K SNPs available from the Affymetrix platform (Affymetrix, Santa Clara, California) and imputation to 2.5 million HapMap SNPs, as described (13,22). SNP genotypes were available in all n = 3,110 participants in this study. SNPs included in the genetic risk scores are shown in Online Table 1.
Lipid GRSs were created from the significant SNP associations reported by Global Lipids Genetic Consortium best SNP (Online Table 2) (13). Scores were calculated for each individual in our sample, separately for HDL-C (47 SNP score), LDL-C (37 SNP score), and TG (32 SNP score) by weighted summation of genotypes (coded additively for the risk allele, increasing for LDL-C and TG and decreasing for HDL-C) multiplied by the reported beta estimates for each trait. Genotypes from “second-tier” genotyping filled in missing genotypes.
All values for TG were natural logarithm transformed. Values for LDL-C, TG, and each of the lipid GRSs were standardized to a mean of zero and 1 SD to facilitate comparisons between the measures. HDL-C was standardized to a mean of zero and 1 SD separately for each sex due to known sex differences in HDL-C. Pearson correlation coefficients were calculated between the GRSs and the lipid levels. Logistic regression models were constructed to calculate odds ratios (ORs) and 95% confidence intervals (CIs) for the association between each lipid measure and elevated CAC. Models were first adjusted for age at CAC measurement and sex, and then adjusted for lipid levels and lipid treatment status at corresponding time points. Because lipid-lowering therapy may confound the association of LDL-C and other lipid levels with CAC, we imputed LDL-C to approximate untreated levels in participants on lipid-lowering therapy, as previously described (23). C-statistics were calculated to assess the impact of both untreated lipid levels at each exam cycle and the addition of lipid GRSs to models on the associations of lipid levels with elevated CAC. All analyses were performed using SAS version 9.1 (SAS Institute, Cary, North Carolina). A p value <0.05 was considered statistically significant.
Offspring and Third Generation Cohorts
Baseline characteristics of the Offspring and Third Generation Cohorts are presented in Table 1. Compared with participants with CAC <75th percentile, those with CAC ≥75th percentile had higher LDL-C and TG, lower HDL-C, and were more likely to be on lipid-lowering medication. The overall GRSs were similar between those with and without CAC ≥75th percentile.
Correlation of lipid measures with genetic risk score
Lipid GRSs and corresponding lipid levels were modestly associated (Online Table 2). For nearly all lipid GRSs, the correlation was strongest with the long-term average lipid measure in the total study sample. Imputed LDL-C and TG were similarly modestly associated with their respective GRSs (data not shown). Among lipid GRSs, the correlation was strongest between HDL-C GRS and TG GRS.
Relations of LDL-C measures and LDL-C GRS with CAC
The association of LDL-C with elevated CAC is shown in Table 2. In FHS Offspring, every SD unit increase in LDL-C at Exam 1 was associated with a 42% increased risk of CAC ≥75th percentile in age- and sex-adjusted analyses (95% CI: 1.24 to 1.63, p < 0.0001). The OR remained similar after further adjustment for LDL-C GRS. By contrast, Exam 7 LDL-C was not significantly associated with elevated CAC (OR: 1.06 per SD increase in LDL-C, 95% CI: 0.93 to 1.20; p = 0.41). Exam 1 to 7 averaged LDL-C was less strongly associated with elevated CAC than that in Exam 1 (per SD increase in LDL-C, OR: 1.24, 95% CI: 1.09 to 1.42, p = 0.002) and remained significant after adjustment for lipid treatment and LDL-C GRS. In the Third Generation Cohort (Table 2) and pooled analysis of the Offspring and Third Generation Cohorts (Table 3, top panel), contemporary LDL-C was positively associated with CAC ≥75th percentile in fully adjusted models (per SD increase in LDL-C) (OR: 1.22, 95% CI: 1.08 to 1.38, p = 0.001 in Third Generation; OR: 1.17, 95% CI: 1.07 to 1.27, p = 0.0006 in the pooled analysis).
The LDL-C GRS was modestly associated with increasing CAC. In the Offspring, each SD increase in the LDL-C GRS conferred an OR of 1.14 for CAC ≥75th percentile (95% CI: 1.00 to 1.29, p = 0.04) in the age- and sex-adjusted model. The significance was attenuated after further adjustment for LDL-C at Exam 1, Exam 7, averaged Exams 1 to 7, and lipid treatment (Table 2). The LDL-C GRS did not attain significance in combined analysis of both cohorts (Table 3 top panel).
In sensitivity analyses, we further evaluated the association between untreated LDL-C and CAC ≥75th percentile using an imputed estimate of the predicted untreated LDL-C for the minority of participants on lipid-lowering therapy. Exam 1 and long-term average, but not Exam 7, LDL-C were significantly associated with elevated CAC (Online Table 3). Similarly, the marginal association of LDL-C GRS with CAC was attenuated in further multivariable adjustment. Every 20 mg/dl increase in LDL-C at Exams 1 to 2 were associated with an OR of 1.22 of elevated CAC (95% CI: 1.12 to 1.33, p < 0.0001 in age- and sex-adjusted analysis), with the greatest C-statistics at these exams (Table 4). By contrast, Exam 7 lipid levels were not associated with elevated CAC.
Relationship of HDL-C measures and HDL-C GRS with CAC
Table 5 shows the association of HDL-C with elevated CAC in the FHS. Increased HDL-C at Exam 1 and average HDL-C over Exams 1-7 were associated with a decreased risk of CAC ≥75th percentile (per SD increase in HDL-C, OR: 0.84, 95% CI: 0.74 to 0.95, p = 0.007 for both). The ORs remained significant and similar after further adjustment for the HDL-C GRS. Exam 7 HDL-C was not associated with increased CAC in the Offspring. In the Third Generation Cohort alone and the combined cohort, however, increases of contemporary HDL-C were significantly associated with decreased risk for elevated CAC in all models (per SD increase in HDL-C, OR: 0.82, 95% CI: 0.72 to 0.93, p = 0.002 in Third Generation Cohort; OR: 0.87, 95% CI: 0.80 to 0.95, p = 0.003 in pooled analysis) (Table 5 and middle panel of Table 3, respectively). The HDL-C GRS was not significantly associated with CAC ≥75th percentile in either cohort (Tables 3 and 5).
Relationship of TG measures and TG GRS with CAC
The OR of CAC ≥75th percentile per SD increment of TG in the FHS is shown in Table 6. Every SD increase in TG at Exam 1 and averaged over Exams 1 to 7 was associated with a 36% increased risk of significant CAC in fully adjusted models (95% CI: 1.19 to 1.56, p < 0.0001 in Exam 1; 95% CI: 1.18 to 1.57, p < 0.0001 in averaged Exams 1 to 7). TG measured at Exam 7 was also significantly associated with high CAC, though the magnitude of association was less than seen for Exam 1 and averaged Exams 1 to 7 (per SD increment of TG, OR: 1.18, 95% CI: 1.03 to 1.34, p = 0.01 in the fully adjusted model). The ORs remained similar after further adjustment for the TG GRS. Similar to results for LDL-C and HDL-C, contemporary TG levels were significantly associated with CAC in the Third Generation (OR: 1.32 per SD increment of TG, 95% CI: 1.17 to 1.48, p < 0.0001) and pooled cohorts (OR: 1.30 per SD increment of TG, 95% CI: 1.19 to 1.41, p < 0.0001) (Table 6 and bottom panel of Table 3, respectively). The TG GRS was not significantly associated with CAC ≥75th percentile in either cohort (Tables 3 and 6).
Longitudinal lipid levels and impact of lipid-lowering therapy
The relationships of CAC and lipid GRS to longitudinal measures of LDL-C, HDL-C, and TG in the Offspring are shown in Figures 1A,1B, and 1C, respectively. Lipid levels progressed toward unfavorable levels through the early exam cycles, with trends toward improvement in Exams 5 to 7. Corresponding use of lipid-lowering therapy in this cohort rose from minimal (0.3% to 0.5%) at Exams 1 to 3, to 2.1% at Exam 4, approximately doubling each subsequent exam to 15.2% at Exam 7. In Exams 6 to 7, use of statin therapy (6.6% and 13.8%, respectively) comprised the majority of lipid-lowering treatment. However, similar time trends of lipids were noted after conducting secondary analyses excluding Offspring on lipid-lowering therapy (Online Fig. 1).
Incremental impact of lipid GRS
For both strata of CAC, those with lipid GRS ≥75th percentile had greater mean LDL-C. The mean LDL-C difference between the lowest group (CAC and GRS both <75th percentile) and the highest group (CAC and GRS both ≥75th percentile) was greatest in earlier exam cycles, with a decrease in mean LDL-C at Exam 7 in the group with both CAC and LDL-C GRS ≥75th percentile. By Exam 7, there were no differences in LDL-C between those with CAC ≥75th percentile or <75th percentile, regardless of GRS stratum. However, even at Exam 7, the presence of high GRS defined separate strata of LDL-C levels for individuals above and below CAC ≥75th percentile (Fig. 1A). Similarly, high (≥75th percentile) CAC and high GRS was associated with less favorable HDL-C and TG levels, with the lowest HDL-C and highest TG in those with greatest lipid GRS (Figs. 1B and 1C, respectively). In a secondary analysis of participants who attended Exam 1 and at least 2 of the final 3 exams, the pattern of serum lipids according to CAC ≥75th percentile or <75th percentile appeared similar (data not shown).
The C-statistics for LDL-C, HDL-C, and TG adjusted models to predict CAC ≥75th percentile were modest (Online Table 4). Overall, addition of the GRS to models resulted in very small increases in the C-statistic that were not statistically significant for any of the lipid measures.
Our longitudinal study of middle-aged to elderly adults, with initial lipid measurements beginning up to 30 years earlier in the Offspring Cohort, reveals the effect of long-term measures of lipids and the possible effect of age at lipid measurement on the association of lipids with elevated CAC. In the older age population of FHS Offspring, early and long-term average levels of LDL-C, HDL-C, and TG were strongly associated with elevated CAC score. With the exception of TG, however, contemporary lipid levels were not associated with elevated CAC. In contrast, in the younger Third Generation Cohort and pooled cohort, contemporary measures of all lipids were associated with CAC ≥75th percentile. The GRSs for each of the lipid types were modestly associated with early and long-term levels of their respective lipid types. However, while the presence of high GRS and high CAC divided individuals into clear strata of long-term lipid levels, the GRS were not associated with elevated CAC in multivariable-adjusted models.
Remote, long-term, and contemporary lipid measures and cardiovascular risk
LDL-C particles initiate inflammation and atherosclerosis in the vessel wall (24). This process starts early in life, and childhood LDL-C and total cholesterol levels correlate with adult cardiovascular risk burden (25–27). Thus, a long-term average of LDL-C may be a better measure of cardiovascular risk than a single contemporary assessment. Indeed, we found that both remote and long-term averaged, but not contemporary, LDL-C were strongly predictive of elevated CAC by middle to elderly age. Although contemporary statin use may be partially implicated in findings, our multivariable models adjusted for lipid-lowering therapy. In addition, in analysis of untreated lipid levels, we noted similar time trend of lipid levels, similar associations of LDL-C with CAC, and a consistent strong association of the earliest LDL-C with elevated CAC. Moreover, the results of other population studies lend support to our findings. In the CARDIA (Coronary Artery Risk Development in Young Adults) study, contemporary lipid levels were less strongly associated with CAC compared with remote levels, despite a low (2%) prevalence of lipid-lowering therapy (9). In the MESA (Multi-Ethnic Study of Atherosclerosis) study, untreated contemporary lipid levels were not significantly associated with prevalent CAC (11). Together with CARDIA (9,10) and MESA (4), our results collectively provide support for the longitudinal impact of dyslipidemia on atherosclerosis. Thus, the estimation of cardiovascular risk in older adult populations should include remote lipid measures, as elevated CAC is not well predicted by contemporary lipid measures.
The association between contemporary lipids and elevated CAC differed between the 2 FHS cohorts. Although contemporary LDL-C and HDL-C were not associated with elevated CAC in the Offspring (mean age 63 years), all contemporary lipid measures were associated with elevated CAC in the Third Generation Cohort (mean age 45 years). Similar associations in the pooled analysis likely reflect weighting by the significantly greater population size of the Third Generation Cohort. Our findings are consistent with a differential association of contemporaneous lipid measures with CAC in younger, rather than older, adults. In the CARDIA study (mean age 42 years), contemporaneous LDL-C was associated with prevalent CAC (9). In addition, the MESA study investigators reported progressive attenuation of the association between lipid measures and CAC with increasing decade from middle to elderly age (11). Interestingly, in the CARDIA study, although contemporaneous LDL-C was associated with prevalent CAC, remote and long-term averaged LDL-C were more strongly associated with CAC (9). Although the explanation for the age effect is not clear, the lack of association of contemporaneous lipids with elevated CAC could be due to the interplay with and contributions of nonpharmacological lipid-lowering lifestyle interventions or other cardiovascular risk factors for CAC at older ages, which were not measured or adjusted for in our analyses.
SNP scores, lipids, and cardiovascular risk
One of the major theoretical advantages of genetic risk prediction is that genes are present at birth and remain largely unchanged over lifetime. Hence, genetic polymorphisms affecting serum lipids can be expected to correlate with lifelong average (or integrated) lipid levels. Indeed, we found that our SNP scores for LDL-C, HDL-C, and TG correlated more consistently with 30-year average lipids than with single measurements. However, this was particularly apparent for LDL-C where lipid-lowering therapy may confound the association between unadjusted contemporary serum lipid levels and genetic background.
Reflecting a lifelong burden of a particular phenotype, genetic scores may be superior to single risk factor measurements for assessing subclinical disease. However, our data do not support the current use of genetic assessment of lipid profiles. In age- and sex-adjusted models, the LDL-C GRS showed only modest associations with CAC, which were attenuated in fully adjusted models. Our findings are consistent with previous studies that found no significant improvement in CAD prediction models after incorporating SNP score information (21,28,29). One possibility is that SNPs captured in the lipid genetic risk score, although associated with lipid levels, may not necessarily reflect the best combination of SNPs predisposing to development of CAC. Another explanation is that SNP scores generally explain only a small proportion of the interindividual variance of an associated trait. In our data, the SNP scores accounted for 6.2%, 8.6%, and 3.8% of the variance in early, long-term average, and contemporary LDL-C, respectively. Similar proportions were found for HDL-C (4.7%, 5.8%, and 5.7%) and triglycerides (2.8%, 6.0%, and 5.8%). These values are low, given that these lipid traits have estimated heritabilities of 40% to 70% (30,31). This obvious mismatch (the “missing heritability”) has been observed in almost any disease or trait where GWAS have been performed, and may be attributed to several factors, including a large number of yet undiscovered SNPs with low minor allele frequencies and/or weak effect sizes, suboptimal fit of SNP effect estimates, gene-gene and gene-environment interactions, structural genetic variation not captured by SNPs, epigenetic modification, and familial/social clustering of cardiovascular risk behaviors (29,32).
First, because our study is cross-sectional and we did not acquire CAC data at baseline Exam 1, our results preclude causal inference of the lipid measures directly on the development of CAC. In addition, our work was based on the currently available GWAS data. It is likely that refined genotyping techniques and improved biostatistical tools will augment our collective knowledge of the genetic regulation of lipid traits, increasing the diagnostic power of future GRSs. Finally, the study population consists largely of Caucasians, which may limit generalizability to other races and/or ethnicities.
In this longitudinal study of the associations between timing of exposure to lipids, lipid GRSs, and subclinical coronary atherosclerosis, remote (∼30 years), followed by long-term averaged, lipids were most strongly associated with CAC in our middle-aged to elderly adult population, though contemporary lipid measures were associated with elevated CAC in our younger adults. Although lipid GRSs were modestly correlated with their corresponding lipid measures, overall lipid GRSs were not significantly associated with elevated CAC. Our findings support assessments of early adulthood and long-term lipid profile measurements to assist in determination of cardiovascular risk. With further detailed DNA sequencing studies, the resulting characterization of the complete spectrum of common and uncommon alleles implicated in lipid levels may improve the GRS and its usefulness in prediction of cardiovascular risk.
For supplemental tables and figures, please see the online version of this article.
The Framingham Heart Study (FHS) is supported by the National Heart, Lung, and Blood Institute (contract N01-HC-25195). This work was partially supported by the National Heart, Lung, and Blood Institute's contract with Affymetrix, Inc, for genotyping services (contract N02-HL-6-4278). Analyses are partially based on resources from Framingham Heart Study investigators in the SNP Health Association Resource project. Dr. Tsao is partially supported by an award from the ACCF. Dr. Kathiresan is a member of the Scientific Advisory Board for Merck; and has received research grants from Merck and Pfizer. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Tsao and Preis contributed equally to this work.
- Abbreviations and Acronyms
- coronary artery calcium
- confidence interval
- Framingham Heart Study
- genetic risk score
- genome-wide association study
- high-density lipoprotein cholesterol
- low-density lipoprotein cholesterol
- multidetector computed tomography
- odds ratio
- single-nucleotide polymorphism
- Received June 29, 2012.
- Revision received August 13, 2012.
- Accepted September 4, 2012.
- American College of Cardiology Foundation
- Kronmal R.A.,
- McClelland R.L.,
- Detrano R.,
- et al.
- Greenland P.,
- Alpert J.S.,
- Beller G.A.,
- et al.
- ↵(2004) Third Report of the Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (ATP III Final Report) (National Heart, Lung, and Blood Institute, Bethesda, MD) www.nhlbi.nih.gov/guidelines/cholesterol. Accessed February 29, 2012.
- Loria C.M.,
- Liu K.,
- Lewis C.E.,
- et al.
- Paramsothy P.,
- Knopp R.H.,
- Bertoni A.G.,
- et al.
- Tikkanen E.,
- Tuovinen T.,
- Widen E.,
- et al.
- Kannel W.B.,
- Feinleib M.,
- McNamara P.M.,
- Garrison R.J.,
- Castelli W.P.
- Splansky G.L.,
- Corey D.,
- Yang Q.,
- et al.
- Parikh N.I.,
- Hwang S.J.,
- Larson M.G.,
- et al.
- Agatston A.S.,
- Janowitz W.R.,
- Hildner F.J.,
- Zusmer N.R.,
- Viamonte M. Jr..,
- Detrano R.
- Friedewald W.T.,
- Levy R.I.,
- Fredrickson D.S.
- Thanassoulis G.,
- Peloso G.M.,
- Pencina M.J.,
- et al.
- Psaty B.M.,
- O'Donnell C.J.,
- Gudnason V.,
- et al.
- Mahoney L.T.,
- Burns T.L.,
- Stanford W.,
- et al.
- Napoli C.,
- D'Armiento F.P.,
- Mancini F.P.,
- et al.
- Davis P.H.,
- Dawson J.D.,
- Riley W.A.,
- Lauer R.M.
- Thanassoulis G.,
- Vasan R.S.
- Kaess B.,
- Fischer M.,
- Baessler A.,
- et al.