We found that a genetic risk score formed of variants previously associated with the risk of CAD on a genome-wide -significant level (SNPs) does not predict the development of subclinical atherosclerosis measured as intimal thickening, the arterial elasticity of the carotid artery, or the incidence of high CIMT or plaques during a six-year follow-up among healthy adults. The results of the present study also showed that individual polymorphisms do not have a substantial impact on the development of subclinical atherosclerosis. The few observed significant associations were moderate at best and were not replicable. A meta-analysis of three independent study populations confirmed the negative results.
CIMT is a good non-invasive method for evaluating the development of atherosclerosis and it is associated with the risk of CAD 
. In apparently healthy young individuals, it provides an excellent means to follow the development of the early phases of atherosclerosis. A recent meta-analysis revealed that an approximately 0.1 mm change in CIMT corresponds to an increase of 15% in the risk of myocardial infarction 
. In the present study, the main cohort (YFS) alone had an 80% power to detect at least 0.0155 mm changes for all SNPs in assuming additive genotypic effects, which should therefore be enough to reveal an at least 2.3% increase in the risk of MI. In the present study, the genetic risk score did not associate with increased CIMT. Previously, each of the studied 24 SNPs have been associated with a 6%–29% increase in the risk of CAD per one risk allele, depending on the variant 
. In the YFS population, the maximal significant difference in individual genotypic means was 0.013 mm, and it was not replicated in other cohorts.
Although similar GRSs have shown potential for being an effective tool in risk stratification for CAD 
, the CAD GRS used in the present study does not seem to have any association with CIMT or CAE measured in a cross-sectional study setting or with the development of these end-points in a six-year follow-up. In the case of the GRS in the YFS, we achieved an 80% power to detect an at least 0.017 mm difference in CIMT and a 0.120%/10 mmHg difference in CAE between the individual population groups when the YFS population was divided into four categories/quartiles by CAD risk score dosage. The 0.017 mm difference in CIMT would correspond to a maximum of 2.6% increase in the risk for MI (as discussed above). As for CAE, the power of the current study was sufficient to detect a difference of less than half (0.120%/10 mmHg) of the observed difference between men and women in CAE (0.280%/10 mmHg) when comparing any two quartiles. Considering our findings (or the lack of significant associations), it is plausible to conclude that the studied polymorphisms would have some other more powerful ways of mediating the risk for CAD than through early intimal thickening and decreased arterial elasticity, at least when measured from the carotid artery among healthy adults. However, the results of GRSs should be interpreted with caution especially when only 24 variants are used to form a risk score. Thus far, genetic risk scoring for Crohn's disease, Bipolar disorder and even CAD, has been shown effective in risk prediction, provided that a sufficiently large amount (hundreds or even thousands) of variants is used (even if their individual effects would be merely nominal) 
. We have previously shown in the YFS population that it is possible to predict, with good accuracy, the risk for extreme CIMT by using multiple SNPs of good predictive value even though they would not associate statistically significantly with CIMT. In fact, when combined with traditional risk factors, the SNPs with high predictive value in a multivariate model (although not statistically significantly associating with CIMT) predicted the risk of extreme CIMT better (AUC of 0.844) than traditional risk factors alone (AUC of 0.741). This method of comparing low and high extremes of quantitative traits has been shown to provide good statistical power for the studied variants 
. In the present study, the CAD SNPs did not seem to increase the predictive value of the analysis when added to the model with age, sex and BMI Supporting our conclusion, these CAD variants have not been associated with CIMT or higher risk of carotid plaque in a recent GWAS published during the preparation of the current manuscript 
. However, the application of the results of genetic association studies and especially the use of genetic risk scores has proved difficult in regard to most diseases with a complex etiology 
. Confounding factors such as different phenotypic diversity within one disease group, as is the case with coronary heart disease, gene–gene interactions and gene–environment interactions as well as the modulating effect of time on genetic risk factors are challenging.
Thus far only few studies have investigated the role of individual SNPs associated with CAD in previous GWASs. The study by Cunnington et al. failed to find any association between CIMT and three individual CAD variants (rs1333049, rs6922269 and rs2943634) 
. These results are in line with the meta-analysis results of the present study. We have also published a negative result between the SNP rs1333049 and CIMT based on the earlier measurements of the present YFS population from 2001 (at age 23–39 years) 
. However, it is possible that some risk variants with modest effect can only be discovered in older populations with a heavier risk factor burden and more advanced CIMT changes. This could explain the positive association (if not due to chance) between CIMT and rs4977574 in 2007. The rs4977574 is in high linkage disequilibrium with rs1333049 which in 2001 was not observed to associate with CIMT. Supporting this, Ye et al. have found rs1333049 to associate with CIMT among substantially older subjects with a mean age of over 50- years. They also found a positive connection between this variant and the progression of CIMT during a 10- year follow-up 
At the time of the review process for the present publication, another study was published with results supporting our findings. The study by Conde et al. presented a meta-analysis of the association between CIMT and eleven independent variants previously associated with CAD in GWASs 
. However, as the meta-analysis also included the YFS population, the results of the study by Conde et al. cannot be considered completely independent of our current results. Although eight of the variants overlap with the SNPs in the present publication, the SNP selection of the present study is more extensive (based on the results of the recent and thus far largest GWAS by the CARDIoGRAM consortium). Furthermore, the current study uniquely presents data from several measurements over time, thus also enabling the investigation of the possible association between individual variants associated with CAD on a genome-wide significant level and the development of CIMT and CAE. No other study has been published concerning the possible connection between the development of subclinical atherosclerosis and an extensive genetic risk score for CAD formed by using a similar selection of SNPs.
In 2007 the present study population consisted of six age groups between the ages of 30 and 45 years. As it is possible that with increasing age the effect of hereditary risk factors could change, we also screened for possible age-by-CAD risk score interactions affecting CIMT and CAE. Although, the interaction analyses showed no apparent age-dependent associations in the YFS or BHS, the possibility that age would modulate the effect of the individual genetic factors in older age groups or among populations with a heavier atherosclerotic burden cannot be excluded. A good example of this is the SNP rs501120 (locus 10q11.2) which is located upstream of the CXCL12 gene and is in complete linkage disequilibrium with rs1746048 identified as a risk variant for CAD by the CARDIoGRAM consortium 
. In the current study population of young adults, rs1746048 did not associate significantly with CIMT or CAE. However,rs501120 and rs1746048 did seem to associate with CIMT among the African American population of the Bogalusa Heart Study as well as among the study population of the Health 2000 Survey where only rs501120 was genotyped. Both of these study populations have higher mean CIMT values when compared to the study population of the YFS orthe subjects of European ancestry in the BHS. The possible confounding effect of genetic heterogeneity can also explain some of the replication failures between these samples. Furthermore, we have also shown in a larger meta-analysis integrating three independent cohorts with a mean age over 50- years (Bruneck, Health2000 and HTO), that rs501120 associates with CIMT. These separate cohorts are based on older populations (Bruneck with an age range of 45–94 years, Health2000 with an age range 46–76 years, and HTO with and age range 19–88 years) 
The foremost problem of the present study is that we lacked the replication data for some of the SNPs in the replication cohorts, especially in the Health 2000 Survey. Therefore, we were unable to evaluate the effect of the risk score on carotid artery elasticity. However, we were able to replicate the findings regarding CIMT for most of the variants, and the consistency of the results obtained from each study population makes the interpretation of the main results relatively safe. The main strength of the study is the size of the combined study population (4,581 subjects with measurements of CIMT and 3,401 with measurements of CAE) and the extensiveness of the risk factor data in each cohort. Furthermore, the YFS population comprises randomly selected healthy individuals from the general population. Therefore, the study is free of selection bias. The YFS also provides a perfect opportunity to follow the development of subclinical atherosclerosis among healthy young adults because both CAE and CIMT have been measured from the same population twice six years apart (2001 and 2007). Although the Health 2000 survey population also represents a randomly selected general population, it is not entirely free of selection bias as the mean age is older (~58 years) and cardiovascular mortality is likely to affect the study population. Fortunately, we were able to replicate our findings in the BHS with measurements of healthy adults of the same age as in the YFS. Due to the fact that the study populations of the YFS and BHS are relatively young (mean age ~38 years in both), the result cannot be applied to older populations. Although the interaction analyses did not suggest that age would clearly modulate the effect of the genetic risk score on sub-clinical carotid atherosclerosis, it is possible that, in older populations, the association would be significant, as could also be the case for individual SNPs. We withheld from further age or sex stratifications in the current study population due to the negative results of the interaction analyses. Furthermore, such stratifications would have greatly decreased the power of the study. We also did not analyse the possible interactions between individual SNPs and risk factors in order to limit the number of analyses.
While the 24-SNP risk score did not identify subjects with increased subclinical risk for atherosclerosis, it is still quite possible that factors associated with pre-symptomatic stages of the disease could be used for identifying patients more predisposed to severe endpoints. Thus, in the future more precise methods are required for studying the mechanism (gene-by-gene interactions and gene-by-environment interactions) by which these risk factors influence the development of the disease. Furthermore, although CIMT and CAE are excellent surrogate phenotypes for studying the development of subclinical atherosclerosis (the early stages of intimal thickening), their results should be interpreted with caution as not all etiological factors are identical in the pathogenesis of coronary atherosclerosis. Many of the genetic risk factors might have different effects in different vascular beds and the phenotypes they promote are diverse as well (from stable manifestations of coronary artery disease to the most severe forms of acute coronary syndromes).
In conclusion, most of the studied SNPs should not influence CAD risk through a mechanism that would also affect the development of subclinical atherosclerosis measured by CAE or CIMT in a healthy general population of young adults.