Genetic Effects on Carotid Intima-Media ThicknessCLINICAL PERSPECTIVE
Systematic Assessment and Meta-Analyses of Candidate Gene Polymorphisms Studied in More Than 5000 Subjects
Background— Carotid intima-media thickness (CIMT) is highly heritable and associated with stroke and myocardial infarction, making it a promising quantitative intermediate phenotype for genetic studies of vascular disease. There have been many CIMT candidate gene association studies, but no systematic review to identify consistent, reliable findings.
Methods and Results— We comprehensively sought all published studies of association between CIMT and any genetic polymorphism. We obtained additional unpublished data and performed meta-analyses for the 5 most commonly studied genes (studied in at least 2 studies in a total of >5000 subjects). We used a 3-step meta-analysis method: meta-analysis of variance; genetic model selection; and random effects meta-analysis of the mean CIMT difference between genotypes. We performed subgroup analyses to investigate effects of ethnicity, vascular risk status, and study size. We accounted for potential reporting bias by assessing qualitatively the possible effects of including unavailable data. Polymorphisms in 3 of the 5 genes (apolipoprotein E, angiotensin I converting enzyme, and 5,10-methylenetetrahydrofolate reductase) had an apparent association with CIMT, but for all these, we found evidence of small study bias. Apolipoprotein E ε2/ε3/ε4 was the only polymorphism with a persistent, statistically significant but modest association when we restricted analysis to larger studies (>1000 subjects).
Conclusions— Of the most extensively studied polymorphisms, apolipoprotein E ε2/ε3/ε4 is the only one so far with a convincing association with CIMT. Larger studies than have generally been performed so far may be needed to confirm the associations identified in future genome-wide association studies, and to investigate modification of effect according to characteristics such as ethnicity and vascular risk status.
Received November 8, 2008; accepted August 21, 2009.
Studying intermediate, quantitative traits is a potentially useful approach to identify genetic risk factors for ischemic stroke and ischemic heart disease.1 Carotid intima-media thickness (CIMT) is an intermediate phenotype for early atherosclerosis, is a strong predictor of future vascular events, including myocardial infarction and ischemic stroke, and is significantly greater in large artery (atherothrombotic) than small artery ischemic stroke.2–4 Estimates of its heritability range from 30% to 86%.5–7
Clinical Perspective on p 15
There have been many studies on the effects of a range of candidate genes on CIMT, the results of which could provide useful insights into genetic influences on atherothrombotic disease and on large artery ischemic stroke and ischemic heart disease. We aimed to identify all published studies of the influence of any genetic polymorphism on CIMT and, for the most commonly studied polymorphisms, to carry out detailed methodological appraisals and meta-analyses of relevant studies. In doing so, we aimed to establish which candidate gene polymorphisms have been most extensively studied, and of these, which have shown reliable associations with CIMT, and how studies in this area might be improved in the future.
Identification of Studies
We used a comprehensive, 2-stage search strategy. In stage 1, we used an electronic search in Medline (1966 to end of 2007) and Embase (1980 to end of 2007) combining general genetics terms with terms for CIMT and carotid atheroma (see Methods in the online-only Data Supplement). From an initial screening of the titles, abstracts, and occasionally full articles, we identified all potentially relevant studies.
In stage 2, we carried out a series of supplementary searches in Medline and Embase for genetic polymorphisms that had been studied in at least 2 studies (because of the recognized importance of the need for independent replication of genetic associations) in a total of >5000 subjects, combining gene-specific terms with terms for CIMT and carotid atheroma, to ensure that we had identified all potentially relevant articles for the selected polymorphisms (see Methods in the Data Supplement). We used the subject-number cutoff of 5000 to restrict our detailed analysis to a manageable number of polymorphisms for which results were likely to be the most precise and reliable.
We sought studies in all languages, obtaining translations where necessary. We obtained full articles for all potentially relevant studies of the selected polymorphisms, as well as for relevant reviews, and checked the reference lists of these for any further relevant studies.
We included in our analyses any study that had assessed the association between variation in one of the selected polymorphisms and a measure of the thickness of the intima-media of the carotid artery. We excluded data on associations with IMT of arteries other than the carotid, with frank atheroma or plaque in the carotid or other arteries, or with change or rate of change in IMT. We avoided double counting by using only the largest available published dataset from any study described in >1 published article.
We extracted information from the articles relevant to each selected study on year of publication; total number of subjects studied; country in which the study was conducted; ethnicity of the subjects; types of subjects studied (eg, healthy volunteers, subjects sampled from the general population, subjects with hypertension, or with diabetes); mean age and gender distribution of the subjects; candidate genes and specific polymorphisms studied; whether genotypes were, after calculation or from published data, in Hardy-Weinberg equilibrium; method of CIMT measurement, where possible selecting measurement of the mean of the far wall of the left and right common carotid artery or as close to that as could be achieved; and, for each genotype, number of subjects, and their mean CIMT (and SD). Wherever possible, we treated studies that had presented data separately for groups of subjects defined by criteria such as ethnicity or presence of specific medical condition as separate substudies.
Two authors (L.P. and one among N.A.M.-G., R.C., or M.C.) independently reviewed study eligibility and extracted the information and data from each study, resolving disagreements and uncertainties by discussion and mutual consensus, involving another author (C.L.M.S. or S.L.) if necessary. If key information or data were not presented in the relevant publications, we sought them directly from the authors of the relevant studies.
We used a 3-step approach to investigate the association with CIMT of each genetic polymorphism studied:
We first determined whether there was evidence of an overall association between genotype and CIMT, by carrying out a meta-analysis of variance (meta-ANOVA) of CIMT, with study and genotype as categorical variables, weighting studies by the inverse of the square of the SEM CIMT.8
When we found a statistically significant (P<0.05) overall association from meta-ANOVA, we went on to determine which genetic model (recessive, codominant, or dominant) should be used to determine the size and nature of the association. We used a regression method to estimate the value λ (and its CI), allowing us to select the most appropriate genetic model for meta-analysis, depending on whether λ was closest to 0 (recessive), 0.5 (codominant), or 1 (dominant; see Methods in the Data Supplement). We did not perform further analyses for any polymorphism not reaching statistical significance at the meta-ANOVA stage.
Using the selected genetic model, we calculated the study-specific and random effects pooled mean differences in CIMT between genotype groups (CIMT mean difference per single allele change for the codominant model).
For the apolipoprotein E (APOE) genetic polymorphism, the 3 commonly encountered alleles (ε2, ε3, and ε4) make up 6 genotypes. For genetic model selection and meta-analysis, we adopted the common convention of grouping these as E2 (ε2ε2 or ε2ε3), E3 (ε3ε3), and E4 (ε4ε4 or ε3ε4), such that a codominant model implied equal differences in CIMT between E2, E3, and E4 genotypes.
We used the I2 statistic to assess heterogeneity between studies, where I2 estimates the percentage of variation between studies that cannot be attributed to chance.9 We performed prespecified subgroup analyses to assess the effects of study size (above or below the mean number of subjects per eligible study or substudy and in a post hoc analysis above or below 1000 subjects), ethnicity (Eastern Asian, Southern Asian, white, or black), and high or low vascular risk status (where studies among subjects with a history of vascular disease or included on the basis of one or more vascular risk factors such as hypertension or diabetes were considered high risk, and studies among healthy volunteers or those from the general population were considered low risk). We used χ2 tests to assess the significance of differences between subgroups in the size of the association between genotype and CIMT. We assessed the potential for publication bias through subgroup analyses based on study size (above) and visual examination of funnel plots.
We carried out all statistical analyses in Stata version 7.0. We were unable to include in formal meta-analyses several otherwise eligible studies for which the necessary data remained unavailable even after attempts to obtain it from the studies’ authors. For each polymorphism selected for analysis, we quantified the proportion of all subjects from eligible studies for whom data were unavailable for meta-analysis. Because qualitative statements about the presence or absence of an association between genotype and CIMT were generally available from the relevant articles for the studies with incomplete data, we attempted informally to assess how our results and overall conclusions might have been affected if we had been able to include these studies in our meta-analyses.
Identification and Selection of Genetic Polymorphisms and Studies for Meta-Analyses
Our stage 1 search strategy yielded 2319 articles, 384 of which seemed to be potentially relevant from titles and abstracts. Polymorphisms in 5 genes (APOE, apolipoprotein E; ACE, angiotensin I converting enzyme; MTHFR, 5,10-methylenetetrahydrofolate reductase; NOS3, nitric oxide synthase 3 [endothelial cell]; ADD1, adducin 1) had been studied in at least 2 studies in a total of >5000 subjects. We selected these genes for stage 2 gene-specific searching and analyses.
For each of the 5 selected genes, Table 1 gives summary information on the polymorphisms studied, the function of their protein products, and the numbers of relevant studies (and subjects), together with information on numbers of studies (and subjects) that could not be included in meta-analyses because the necessary data were unavailable.
We identified 95 independent studies (77 118 subjects) that analyzed the association between CIMT and a polymorphism or polymorphisms in 1 of the 5 genes of interest (Supplemental references W1–W95). Twenty-nine studies (including 12 982 subjects) did not have all the necessary data available for analysis in the relevant published articles (supplemental references W4, W5, W15, W16, W25, W27, W29, W36, W47, W50, W53, W54, W59, W61, W63, W64, W67, W69, W70, W72, W76, W78, W80, W82, W84, W86, W92, and W94), and authors from 9 of these were able to provide us with additional unpublished data (Supplemental references W15, W16, W27, W29, W59, W76, W78, W80, and W84), enabling us to retrieve 24% of the unavailable data (5457/12 982 subjects).
Characteristics of Included Studies
The summary characteristics of all relevant studies for the 5 selected genes are shown in supplemental Table I. Sample sizes ranged from 47 to 9304. White subjects from Europe, Australia, and the United States made up the majority of subjects; Eastern Asian subjects were mostly from China, Japan, and Taiwan; and 1 study included Southern Asian subjects. One study was carried out in black Americans.
Subjects were mostly middle aged to elderly. Most were from population samples or healthy volunteers, but some were selected groups of patients at high vascular risk. Genotypes were mostly in Hardy-Weinberg equilibrium, and where they were not, the subjects were generally selected patient groups for which Hardy-Weinberg proportions would not necessarily be expected. CIMT measurement methods varied between studies, but the majority measured the far wall of the common carotid artery.
Tests for Overall Association and Selection of Genetic Models
Table 2 shows the results of the 3-step meta-analysis for the 5 selected genes. Meta-ANOVA found an overall association between genotype and CIMT at P<0.05 for only 3 polymorphisms: APOE (ε2/ε3/ε4), ACE (I/D), and MTHFR (677 C/T). Linear regression showed that APOE (ε2/ε3/ε4) and ACE (I/D) should be analyzed according to a codominant genetic model (λ=0.4 and 0.5, respectively). For MTHFR (677 C/T), λ was estimated to be 0.2 (95% CI, 0.1 to 0.4). Although the estimated 95% CI did not include any of the expected values (0, 0.5, or 1), we analyzed the data using a recessive model, because 0.2 is closest to 0.
Meta-Analyses of Polymorphisms in APOE, ACE, and MTHFR
We found 30 relevant studies (36 substudies, 32 995 subjects) for APOE (ε2/ε3/ε4). Full data for meta-analysis were unavailable in relevant publications or from the authors for 3 studies (742 subjects, 2% of subjects; supplemental references W4, W5, and W25). Meta-ANOVA yielded a significant overall association between APOE and CIMT (P<0.001), with a pooled random effects estimate of the difference in mean CIMT per step from E2 to E3 or from E3 to E4 genotypes of 25 μm (95% CI, 17 to 33; Table 2 and supplemental Figure I).
We found 39 relevant studies (43 substudies, 20 105 subjects) for ACE (I/D). Full data for meta-analysis were unavailable for 9 studies (3067 subjects, 15% of subjects; supplemental references W25, W36, W47, W50, W53, W54, W61, W63, and W64). Meta-ANOVA yielded a significant overall association between ACE and CIMT (P=0.005), the pooled estimate of the per allele (D) difference in mean CIMT being 14 μm (95% CI, 5 to 22; Table 2 and supplemental Figure II).
We found 20 relevant studies (22 substudies, 10 487 subjects) for MTHFR (677 C/T). Full data for meta-analysis were unavailable for 7 studies (2542 subjects, 24% of subjects; supplemental references W50, W63, W67, W69, W70, W72, and W82). Meta-ANOVA yielded a significant overall association between MTHFR and CIMT (P=0.02). The pooled estimate of the difference between TT and CT/CC genotypes was 29 μm (95% CI, 0 to 58; Table 2 and supplemental Figure III).
For each of these 3 genetic polymorphisms, there was substantial heterogeneity between studies (I2 values ≈80%; supplemental Figure I through III). The results of prespecified subgroup analyses to investigate possible causes of this heterogeneity were strikingly similar for all 3 polymorphisms (Figure). We found substantially larger pooled mean CIMT differences among Eastern Asian compared with white population (the subgroup difference was highly significant for APOE), in high-risk populations compared with low-risk populations (subgroup differences clearly significant for all 3 polymorphisms), and in smaller compared with larger studies (subgroup differences clearly significant for all 3 polymorphisms), suggesting the existence of small study bias. For each polymorphism (especially APOE and ACE), there was less heterogeneity between the results of the larger studies. Study size could explain the apparent difference in size of association between different ethnicities and risk groups, because studies in high-risk populations and among Eastern Asian subjects were on average much smaller than among low-risk populations and white subjects (Figure). In keeping with the results of subgroup analyses based on study size, funnel plots of the study-specific mean difference versus its standard error were asymmetrical for all 3 polymorphisms, suggestive of small study bias, most likely publication bias.
Focusing attention on just the larger (and presumably more reliable) studies, an association between CIMT and APOE remained, albeit much smaller than the overall pooled estimate, with an estimated mean CIMT difference of 8 μm (95% CI, 6 to 11) per step from E2 to E3 to E4 genotype groups, but associations between CIMT and both ACE (I/D) and MTHFR (677 C/T) were smaller and no longer statistically significant (for ACE, mean CIMT difference per additional D allele: 4 μm, 95% CI, 0 to 8; for MTHFR, mean CIMT difference between TT and CT/CC: −9 μm, 95% CI, −32 to 13). In a further post hoc subgroup analysis, we restricted analyses to studies in >1000 subjects and found a similar, significant result for APOE and similar, nonsignificant results for ACE and MTHFR (data not shown).
Of the 3 studies with data unavailable for the APOE meta-analysis, one found an association between E4 genotypes and higher CIMT (W4), another found a similar association in the nondiabetic subgroup only (W5), and the third found no association (W25). All 3 were small (66, 206, and 470 subjects) and so would not have contributed to the analysis including only larger studies (supplemental Figure I).
Most of the 9 studies with data unavailable for the ACE meta-analysis reported no association. Three would have contributed to our “larger studies” analysis, using the mean study size cutoff criterion (W25, W61, and W63). Of these, 2 found that the D allele was associated with increased CIMT, so their inclusion could potentially have strengthened the association between ACE and CIMT. However, none of the studies with data unavailable for meta-analysis would have been large enough to merit inclusion in our analysis of studies including >1000 subjects (supplemental Figure II).
Of the 7 studies with data unavailable for the MTHFR meta-analysis, 2 were larger than the mean study size cutoff for the study size subgroup analysis (W63 and W69). The smaller of these found an association between MTHFR (677 C/T) and CIMT whereas the other did not, and so their inclusion would seem unlikely to materially alter the results of any of the MTHFR analyses (supplemental Figure III).
Other Selected Genes
There was no association between CIMT and polymorphisms in nitric oxide synthase 3 or adducin 1 (Table 2). Although the proportion of data unavailable for meta-analysis was >40% for nitric oxide synthase 3 (Table 1), almost all studies with unavailable data reported either no association or association only in a particular subgroup (eg, in males or black diabetics only). Thus, it seems unlikely that the inclusion of these data would have materially affected the results for these polymorphisms.
Other Potential Genes of Interest
Our search strategy identified polymorphisms in >140 genes that had been studied for an association with CIMT, but less than half of these had been assessed in >1 study. Of those studied in <5000 subjects, 19 genetic polymorphisms had been studied in a total of ≈3000 to 5000 subjects, a further 53 in a total of ≈1000 to 3000 subjects, and a further 68 in an estimated total of <1000 subjects. Excluding those included in our meta-analyses, for 46 genetic polymorphisms, the largest study done had >1000 subjects (and so could, according to the results of our analyses based on study size, be considered reliable), but in no more than a handful of cases was the genetic polymorphism assessed in at least 2 studies including >1000 subjects each. Many of the genetic polymorphisms studied but not included in our meta-analyses showed preliminary evidence for an association with CIMT, but most if not all of these would need replication in much larger samples before the results could be considered reliable. For many others, the study or studies performed were too small reliably to detect effects of moderate size.
Although there have been narrative reviews of the genetics of CIMT,10 to our knowledge, there has not previously been an attempt to review the evidence for genetic associations with this phenotype systematically. Narrative reviews draw attention to exciting new findings and may stimulate new research but may selectively emphasize results of particular studies, and so can sometimes be misleading.11
Our systematic review identified >140 genes studied as candidates for association with CIMT. We reviewed in detail 95 independent studies of the association between CIMT and polymorphisms in the 5 most commonly studied genes and found clear evidence of an association for only one of these. Polymorphisms in the APOE, ACE, and MTHFR genes all showed a significant association with CIMT in meta-ANOVA analysis. But, of these, APOE (ε2/ε3/ε4) was the only polymorphism whose association withstood restricting analysis to larger studies only, suggesting that although there almost certainly is an association, its size is overestimated in the literature because of small study bias. Our subgroup analyses suggest that the apparent associations with CIMT of ACE and MTHFR, and apparent modification of genetic effects according to ethnicity and vascular risk status, may well be due to small study bias. The 95% CI from analyses restricted to the larger and more reliable studies of the relevant polymorphisms in these genes would suggest that, for both ACE I/D and MTHFR 677 C/T, an association with CIMT remains possible, but is unlikely to be larger in magnitude than 8 μm per additional D allele for ACE I/D and could range between approximately −35 and 13 μm for MTHFR 677 TT versus CT/CC.
Our study has some particular strengths. We used a series of explicit, predefined methods to select genes for further investigation, to assess whether there was an overall association between genotype and CIMT, and to choose the most appropriate genetic model for meta-analysis. We also sought additional unpublished data where it was unavailable in publications and considered the implications of any data that remained unavailable for analysis. Publications identified by our search in which the association between a genetic polymorphism and CIMT was not the main feature of the article tended not to identify an association and not to report results in full, although sometimes reported positive results for particular subgroups. Thus, we identified and accounted for reporting bias, where positive results tend to be highlighted and reported in more detail than negative ones. Previous meta-analyses in stroke genetics have used adequate reporting of data within publications as a criterion for inclusion,12 which might make them prone to reporting bias.
Two linkage studies have identified quantitative trait loci for CIMT. One reported a maximum LOD score of 4.1 at 161 cM on chromosome 12 and subsequently found evidence of association with an atherosclerosis candidate gene (SCARB1, a high density lipoprotein receptor, cell-surface glycoprotein) from the region of linkage.13 The other identified 2q33-35 as a region with significant linkage (LOD=3.08), including the NOSTRIN, IGFBP2, and IGFBP5 genes, none of which have yet been independently tested for an association with CIMT.14
In the current era of genome-wide association studies, further promising candidates for CIMT are likely to emerge in the near future. The findings of our systematic review suggest that, if CIMT is to fulfill its promise as an intermediate phenotype that will improve our understanding of the genetics of vascular diseases, studies to confirm associations with promising candidates identified in genome-wide studies may have to be very large. For example, if other polymorphisms influencing CIMT have effect sizes similar to that of APOE, then studies would need to be powered to detect between-genotype differences in CIMT of as small as ≈10 μm (the approximate effect size found by pooling data from the larger, more reliable studies). Taking estimates of the population mean and standard deviation of CIMT from a recent relevant review (mean, 700 μm; SD, 160 μm),2 a sample size of >6000 subjects would be needed for 80% power at P<0.05 (and of almost 12 000 subjects for 90% power at P<0.01) to detect a per-genotype mean CIMT difference of 10 μm in a codominant model, assuming a minor allele frequency of 0.2. Estimated sample sizes are larger with more stringent power and significance criteria or for polymorphisms with a lower minor allele frequency but are smaller with a higher minor allele frequency, or if we only want reliably to detect larger CIMT differences of up to, say, 100 μm. For example, reliable detection (90% power at P<0.01) of a per-genotype mean CIMT difference of 100 μm assuming a minor allele frequency of 0.05 would need ≈400 subjects (supplemental Table II). To set this in context, an increase in CIMT of 100 μm is associated with an increased future risk of myocardial infarction and stroke of ≈15% and 18%, respectively.2 At any minor allele frequency, more than ≈2000 subjects would be required reliably to detect a CIMT difference of 25 μm (supplemental Table II), and because it is recommended that findings are replicated in studies that are larger than the discovery dataset, a total of >5000 subjects would be required to detect and confirm such effects confidently. However, our study’s stringent cutoff of 5000 subjects for inclusion means that we may have missed genetic polymorphisms with effect sizes much larger than this.
So far, among the most extensively studied genetic polymorphisms, APOE ε2/ε3/ε4 is the only one to have shown a convincing association with CIMT, with E4 genotypes associated with an increase and E2 genotypes with a decrease in CIMT. This is consistent with meta-analyses showing an association between the APOE polymorphism and ischemic heart disease,15 our previous meta-analysis of the association between APOE and stroke subtypes, which suggested that APOE E4 genotypes may be specifically associated with the large artery subtype of stroke but not with other ischemic subtypes,16 with our previous meta-analysis of APOE and CIMT that used less sophisticated statistical methods,17 and with our meta-analysis of APOE and white matter hyperintensities on brain imaging (a quantitative phenotype linked to small vessel disease lacunar stroke), which did not identify an association.18
The genes we reviewed here were those most commonly studied, mainly because they are key genes in known candidate pathways for vascular disease. This method of selecting genes is limited by current knowledge. Recent successes with the genome-wide association approach show the potential for hypothesis-free methods to identify candidate genes in novel pathways,19 which, if confirmed in studies of adequate size, may lead to new insights into the causes of and treatments for complex disease, including CIMT and so atherothrombotic vascular diseases.
The following investigators provided additional information from their studies: Claudia Altamura, Moniek P.M. de Maat, Mireia Junyent, Heikki Kauma, Linda E. Kelemen (on behalf of the SHARE investigators), Giuseppe Lembo, Stephen McDonald, Natasa Marcun Varda, and Noriaki Yorioka (supplemental references W29, W76, W27, W15, W78, W84, W82, W59, and W16, respectively). We thank Brenda Thomas for her help with the electronic searches and Jim Wilson for helpful comments on an earlier version of this article.
Sources of Funding
Dr Sudlow was funded by the Wellcome Trust (Clinician Scientist Award WT063668), Lavinia Paternoster was funded by the UK-MRC (PhD studentship), and Nahara A. Martinez-Gonzalez was funded by ConacYt-Mexico.
Dichgans M, Markus HS. Genetic association studies in stroke. Stroke. 2005; 36: 2027–2031.
Lorenz MW, Markus HS, Bots ML, Rosvall M, Sitzer M. Prediction of clinical cardiovascular events with carotid intima-media thickness: a systematic review and meta-analysis. Circulation. 2007; 115: 459–467.
Dijk JM, van der Graaf Y, Bots ML, Grobbee DE, Algra A. Carotid intima-media thickness and the risk of new vascular events in patients with manifest atherosclerotic disease: the SMART study. Eur Heart J. 2006; 27: 1971–1978.
Pruissen DMO, Gerritsen SAM, Prinsen TJ, Dijk JM, Kappelle LJ, Algra A. Carotid intima-media thickness is different in large- and small- vessel ischemic stroke: the SMART study. Stroke. 2007; 38: 1371–1373.
Duggirala R, Gonzalez VC, O'Leary DH, Stern MP, Blangero J. Genetic basis of variation in carotid artery wall thickness. Stroke. 1996; 27: 833–837.
Fox CS, Polak JF, Chazaro I, Cupples A, Wolf PA, D'Agostino RA, O'Donnell CJ. Genetic and environmental contributions to atherosclerosis phenotypes in men and women: heritability of carotid intima-media thickness in the Framingham Heart Study. Stroke. 2003; 34: 397–401.
Juo SH, Lin HF, Rundek T, Sabala EA, Boden-Albala B, Park N, Lan M-Y, Sacco RL. Genetic and environmental contributions to carotid intima-media thickness and obesity phenotypes in the Northern Manhattan Family Study. Stroke. 2004; 35: 2243–2247.
Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003; 327: 557–560.
Wang D, Yang H, Quinones MJ, Bulnes-Enriquez I, Jimenez X, De La Rosa R, Modilevsky T, Yu K, Li Y, Taylor KD, Hsueh WA, Hodis HN, Rotter JI. A genome-wide scan for carotid artery intima-media thickness: the Mexican-American coronary artery disease family study. Stroke. 2005; 36: 540–545.
Sudlow C, Martinez-Gonzalez NA, Kim J, Clark C. Does apolipoprotein E genotype influence the risk of ischemic stroke, intracerebral hemorrhage, or subarachnoid hemorrhage? Systematic review and meta-analyses of 31 studies among 5961 cases and 17 965 controls. Stroke. 2006; 37: 364–370.
Paternoster L, Martinez-Gonzalez N, Lewis SC, Sudlow CL. Association between apolipoprotein E genotype and carotid intima-media thickness may suggest a specific effect on large artery atherothrombotic stroke. Stroke. 2008; 39: 48–54.
Paternoster L, Chen W, Sudlow CLM. Genetic determinants of white matter hyperintensities on brain scans: a systematic assessment of 19 candidate gene polymorphisms in 46 studies in 19,000 subjects. Stroke. 2009; 40: 2020–2026.
Carotid intima-media thickness (CIMT) is a measure of subclinical atherosclerosis, is associated with future risk of myocardial infarction and stroke, and is highly heritable. Studies of genetic factors influencing CIMT may identify genes that alter risk of ischemic heart disease and stroke, in particular large artery ischemic stroke, thereby helping us to better predict the risk of developing these clinical conditions and better understand their aetiology. We systematically identified and performed meta-analyses of all studies of the association of CIMT with variation in any gene that had been studied in >2 studies with a total of >5000 individuals. We found only 1 gene (apolipoprotein E, which affects cholesterol levels) that had a robust association with CIMT, but the magnitude of the association was quite small, limiting its utility in clinical risk prediction. Genome-wide studies and large confirmatory candidate gene studies are needed for reliable identification of genes associated with CIMT.
The online-only Data Supplement is available at http://circgenetics.ahajournals.org/cgi/content/full/CIRCGENETICS.108.834366.