Family history and polygenic risk of cardiovascular disease: independent factors associated with secondary cardiovascular events in patients undergoing carotid endarterectomy

Background Family history (FHx) of cardiovascular disease (CVD) is a risk factor for CVD and a proxy for cardiovascular heritability. Polygenic risk scores (PRS) summarizing >1 million variants for coronary artery disease (CAD) are associated with incident and recurrent CAD events. However, little is known about the influence of FHx or PRS on secondary cardiovascular events (sCVE) in patients undergoing carotid endarterectomy (CEA). Methods We included 1,788 CEA patients from the Athero-Express Biobank. A weighted PRS for CAD including 1.7 million variants was calculated (MetaGRS). The composite endpoint of sCVE during three years follow-up included coronary, cerebrovascular and peripheral events and cardiovascular death. We assessed the impact of FHx and MetaGRS on sCVE and carotid plaque composition. In CEA patients, both positive FHx and higher MetaGRS were independently associated with increased risk of sCVE. Moreover, higher MetaGRS was associated with vulnerable plaque characteristics. Future studies should unravel underlying mechanisms and focus on the added value of PRS and FHx in individual risk prediction for sCVE.


Introduction
Family history of cardiovascular disease (FHx) is a major risk factor for primary cardiovascular disease (CVD) and serves as a surrogate for genetic predisposition. (1,2) Risk prediction for secondary cardiovascular events remains challenging as traditional risk factors have limited discriminative performance. (3) The main underlying mechanism of CVD is atherosclerosis, and atherosclerotic plaque composition, exemplified by intraplaque haemorrhage (IPH), was shown to be associated to adverse secondary cardiovascular events (sCVE).(4) Yet, the relevance of FHx for secondary outcome of cardiovascular events is still unclear. (5)(6)(7)(8)(9) Large-scale genome-wide association studies (GWAS) have identified hundreds of common genetic variants (single-nucleotide polymorphisms or SNPs) robustly associated with coronary artery disease (CAD) (10)(11)(12)(13)(14) and ischemic stroke (15)(16)(17) predisposition, albeit with small individual effects. Exact pathobiological mechanisms leading to cardiovascular symptoms are still poorly understood, but CAD-and ischemic stroke genetic variants were previously associated to atherosclerotic plaque composition. (18) Polygenic risk scores (PRS) summarize the small individual genetic effects into a quantitative measure of genetic disease susceptibility. Recent studies showed that PRS were strongly correlated with prevalent and incident CAD independent of traditional risk factors including family history in the UK Biobank population. (19,20) For example, individuals with higher scores of the MetaGRS (a PRS for CAD including 1.7 million SNPs) were at 1.7-4.2 fold higher risk for a first coronary event compared to individuals with lower MetaGRS scores. In 7 addition, a very recent study in a French-Canadian population with established CAD showed that MetaGRS was also associated with increased risk of recurrent CAD events. (21) As atherosclerosis is known for its complex nature with genetic overlap in different diseases such as stroke, CAD, abdominal aortic aneurysm (AAA) and peripheral artery disease (PAD) (22), we aimed to investigate the association between MetaGRS and long-term secondary cardiovascular events in a different cohort of severe atherosclerotic patients with carotid artery stenosis undergoing carotid endarterectomy. Given that FHx is used in clinical practice as a derivative of genetic background, we also examined the association between FHx and secondary cardiovascular events. Moreover, to explore possible underlying pathophysiological mechanisms, we studied the impact of MetaGRS and FHx on carotid histological plaque characteristics.

Athero-Express Biobank
All patients in this study were included in the Athero-Express Biobank, a prospective cohort study that included consecutive patients with severe carotid artery stenosis undergoing CEA in two large tertiary referral hospitals in The Netherlands, the University Medical Centre Utrecht Patients provided written informed consent before study participation.
. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not peer-reviewed) (which was The copyright holder for this preprint . http://dx.doi.org/10.1101/19006718 doi: medRxiv preprint first posted online Sep. 18, 2019 ;

Definitions
A positive FHx was defined as having a first-degree relative (either a parent or sibling) with onset of cardiovascular disease (myocardial infarction (MI), coronary artery stenosis, stroke, AAA, or cardiovascular death including sudden death) before the age of 60 years. The primary outcome of this study was defined as a composite secondary cardiovascular event (sCVE) within three years of follow-up including fatal or non-fatal MI, fatal or non-fatal stroke, ruptured AAA, fatal cardiac failure, coronary or peripheral interventions (either percutaneous or bypass surgery), leg amputation due to cardiovascular causes and cardiovascular death. Secondary outcomes were histological atherosclerotic carotid plaque characteristics.

Genotyping
Methods for genotyping, quality control and imputation in the Athero-Express biobank have been published elsewhere (25, 26). Briefly, DNA was extracted from EDTA whole blood samples or if not present from atherosclerotic plaque tissue according to validated protocols.
Genotyping was performed with two commercially available chips: the first batch by Affymetrix Genome-Wide Human SNP array 5.0 (previously used in the Athero-Express Genomics Study 1 (AEGS1), covering samples obtained in [2002][2003][2004][2005][2006][2007] and the second batch by Affymetrix Axiom GW CEU 1 array (previously used in Athero-Express Genomics Study 2 (AEGS2), covering samples obtained in 2002-2013). Procedures for data quality control and data cleaning were in accordance with global standards. (27) After genotype calling according to Affymetrix' specification, data was filtered on 1) individual call rate > 97%, 2) genotype call rate > 97%, 3) minor allele frequencies (MAF) > 3%, 4) average heterozygosity rate ± 3.0 standard deviations, . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Polygenic risk score (MetaGRS)
To estimate the polygenic cardiovascular disease susceptibility for included patients in our cohort, we used the previously published polygenic risk score for CAD (MetaGRS).(19) Its construction was described elsewhere. (19) Briefly, the MetaGRS comprises 1.7 million genetic variants associated with CAD and was constructed through meta-analysis of three genomic risk scores: GRS46K (comprising 46,000 cardiometabolic genetic variants), FDR202 (including 202 genetic variants associated with CAD at false discovery rate p<0.05 in the recent GWAS CARDIoGRAMplusC4D), and the 1000Genomes genetic score also created with CARDIoGRAMplusC4D. The MetaGRS was internally and externally validated for the primary risk of prevalent and incident CAD in the UK Biobank. (19) We calculated the MetaGRS for each included patient in this study and standardized it to mean-zero and unit-variance for each genotyping batch separately, i.e. AEGS1 and AEGS2, respectively.
The segment with largest plaque burden was chosen as the culprit lesion and immunohistochemically analysed for macrophages, smooth muscle cells (SMC), lipid core, calcification, collagen, intraplaque haemorrhage (IPH) and microvessel content. Extensive description of the standardized protocol for atherosclerotic plaque processing and analysis of plaque characteristics has been previously reported and is added to the Supplemental. (4,23,29,30) To assess the overall vulnerability of the atherosclerotic plaque, a vulnerability score was created ranging from 0-5 with 1 point for plaque characteristics that are considered hallmarks of a vulnerable plaque (moderate/heavy macrophages, no/minor collagen, no/minor SMC, lipid core>10% and presence of IPH), based on a previous publication.(31)

Statistical analysis
Baseline characteristics were compared between patient groups (FHx and MetaGRS) by chisquare test for categorical variables and Student's t-test for continuous variables (lipid levels were log-transformed). We analysed the association between FHx and MetaGRS and sCVE by Kaplan-Meier curves were constructed to graphically illustrate univariate associations.
Confounders for multivariable analyses were selected based on literature (19,31,32) (for sCVE these were age, sex, diabetes, BMI, smoking and hypercholesterolemia and for plaque . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not peer-reviewed) (which was The copyright holder for this preprint . http://dx.doi.org/10.1101/19006718 doi: medRxiv preprint first posted online Sep. 18, 2019 ; characteristics these were age, sex, surgery year and type of cerebrovascular symptoms).
Additional confounders were added when showing an association of p<0.20 with the determinant (FHx or MetaGRS) and outcome of interest (sCVE or plaque characteristics). For MetaGRS models genotype array and principal components 1-4 were also added. Full model description is displayed in the Supplemental Tables S1 and S2. Because a previous study in our biobank showed that IPH is associated with sCVE(4), IPH was added to multivariable models of FHx, MetaGRS and sCVE to explore whether IPH could be one possible underlying mechanism. Sexstratified analyses were performed to unravel sex-dependent differences in associations. Values with p<0.05 were considered statistically significant. All analyses were performed in IBM SPSS Statistics version 25.0.
. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not peer-reviewed) (which was The copyright holder for this preprint . http://dx.doi.org/10.1101/19006718 doi: medRxiv preprint first posted online Sep. 18, 2019 ;

Results
Patient selection from the Athero-Express Biobank and characteristics of the study population are displayed in Figure 1 and Table 1. Patients had a mean age 69 years and 70% were men. The cohort represented a typically severe atherosclerotic cohort with high prevalence of traditional risk factors and atherosclerotic manifestations in other vascular beds (coronary or peripheral arteries, respectively 30% and 20%). Baseline characteristics were similar between the total cohort with FHx data and the sub-cohort with genotyped data (Table 1).

Patients with positive FHx have a higher risk of sCVE
Patients with a positive FHx (744/1,788, 41.6%) were younger and had on average more cardiovascular risk factors (Table 1). During a median follow-up of 2.9 years, 418 patients  Table S3). This association remained significant after correction for confounders with adjusted HR 1.287, 95%CI 1.033-1.604, p=0.024 ( Figure 3; Supplemental Table S3) and was independent of genetic predisposition as measured by MetaGRS (adjusted HR 1.397, 95%CI 1.074-1.819, p=0.013, Figure 3; Supplemental Table S3). Sex-stratified analyses confirmed results in men (with adjusted HR after correction for confounders of 1.380, 95%CI 1.068-1.783, p=0.014; adjusted HR after . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not peer-reviewed) (which was The copyright holder for this preprint . http://dx.doi.org/10.1101/19006718 doi: medRxiv preprint first posted online Sep. 18, 2019 ; correction for confounders including MetaGRS was 1.513, 95%CI 1.115-2.052, p=0.008).
However, in women the univariate association between FHx and sCVE was not significant (unadjusted HR 1.187, 95%CI 0.822-1.171, p=0.360) but multivariable analyses could not be performed because of limited power (Supplemental Table S3).

Patients with higher MetaGRS have a higher risk of sCVE
The MetaGRS, standardized to mean-zero and unit-variance, approximated a normal distribution in the study population (Supplemental Figure S1). Patients in the top 20% of MetaGRS were relatively more often females, younger of age and had less often diabetes compared to the remaining 80% of the cohort (Table 1). Differences in baseline characteristics between high (highest quintile of MetaGRS) and low genetic risk patients (lowest quintile of MetGRS) can be found in the supplemental (Table S4) Figure 2C) and for multivariable analysis adjusted . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not peer-reviewed) analysis was not possible due to lack of power (Supplemental Table S3).

MetaGRS is associated with vulnerable carotid plaque characteristics
To unravel possible underlying pathophysiological mechanisms of the associations between MetaGRS, FHx and CVD, we explored the impact of FHx and the MetaGRS on atherosclerotic plaque characteristics. We found no associations between histological plaque characteristics and FHx in the total cohort or in women although not all multivariable analyses could be performed (Supplemental Table S7  Because IPH has been associated with increased risk of sCVE(4), we added IPH to the multivariable models of FHx, MetaGRS and sCVE to unravel whether the association between FHx, MetaGRS and sCVE could be explained by IPH. We found that all associations of FHx and MetaGRS were independent of IPH given that adding IPH to multivariable models of sCVE did not alter the effect sizes (Supplemental Table S3).
. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not peer-reviewed) (which was The copyright holder for this preprint . http://dx.doi.org/10.1101/19006718 doi: medRxiv preprint first posted online Sep. 18, 2019 ;

Discussion
We validated the polygenic risk score for CAD (MetaGRS) for sCVE in a severe atherosclerotic cohort of carotid artery stenosis patients undergoing carotid endarterectomy. We report four key findings. First, FHx was independently associated with an increased risk of sCVE. Second, higher MetaGRS score was independently associated with an increased risk of sCVE. Third, the association of MetaGRS with sCVE was independent of FHx and the association of FHx with sCVE was independent of MetaGRS. Fourth, high MetaGRS was associated with more vulnerable carotid atherosclerotic plaque characteristics suggesting possible underlying pathobiological mechanisms through which genetic variants could affect CVD.
Although positive FHx is a well-known risk factor for primary CVE(1, 2), previous studies assessing FHx and secondary outcome are inconsistent. (5,(7)(8)(9)33) In patients with first-MI (7,8,33), studies have reported a protective effect of FHx on all-cause mortality, whereas others showed an increased risk of CVE. (5,9) One can assume that patients with positive FHx are identified earlier as at-risk individuals through screening programs resulting in more intensive surveillance and preventive strategies leading to the benefit in overall survival. (7) Indeed, in our cohort patients with positive FHx were also younger at timing of CEA. Of note, it is known that the sensitivity of self-reported FHx can be low (50%-70%) and might therefore be an unreliable estimate. (34) Our results are an independent validation of and consistent with a recent French Canadian study that revealed the association of MetaGRS in patients with recurrent CAD. (21) We now confirm this association in a different patient population with high prevalence of other CVD . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The independent association of FHx and PRS has been reported frequently for primary CAD events (19,20,41), however only one study showed this for sCVE. (40) We confirm the independent association of MetaGRS with sCVE from FHx. Indeed, we also found that positive FHx was associated with increased risk of sCVE independent of genetic risk (MetaGRS). Several reasons could be hypothesized for the non-overlapping associations of FHx and MetaGRS with CVE. MetaGRS includes common genetic variants associated to an increased risk for CAD in the general population, whilst CVD in families may arise in part from more rare genetic mutational events, and thus a positive FHx captures individual yet family specific rare variation.
Another explanation could be that FHx reflects not only genetic factors but also non-genetic factors. Although we corrected for traditional risk factors, other environmental factors that were . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not peer-reviewed) (which was The copyright holder for this preprint . http://dx.doi.org/10.1101/19006718 doi: medRxiv preprint first posted online Sep. 18, 2019 ; not taken into account such as social economic status or nutrition patterns could still be attributable to the risk of secondary events. Exactly how genetic variants affect CVD is currently unknown. CAD-variants have been linked to pathways involved in atherosclerosis (e.g. lipid metabolism, inflammation, blood pressure and vascular remodelling), but most CAD-variants are situated outside protein-coding regions with unknown functions, making them hard to map to pathophysiological mechanisms. (13,42) We found that MetaGRS was associated with a more rupture-prone atherosclerotic plaque displayed by a higher plaque vulnerability score caused by more fat, IPH (in men) and macrophages (predominantly in women). Moreover, carotid plaques from men with positive FHx were associated with less SMC and less collagen, whereas the association in women remains unclear. Similarly, a previous AE study showed that a CAD-PRS was correlated with more fat, whereas a large-artery stroke-PRS was correlated with more IPH and SMC. Admittedly, our study has several limitations. First, although our results may suggest interesting sex-differences as we only observed the independent association of MetaGRS and FHx for sCVE in men, the association in women remains unclear because we were . CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not peer-reviewed) (which was The copyright holder for this preprint . http://dx.doi.org/10.1101/19006718 doi: medRxiv preprint first posted online Sep. 18, 2019 ; underpowered for multivariable analysis. Second, we did not have data regarding medication use during follow-up nor therapy compliance. These factors could have interfered with observed sCVE rates. Second, owing to limited power we were unable to assess associations with separate CVE or determine the predictive value of MetaGRS above clinical risk factors. Last, most included patients are of European ancestries and generalizability to other ethnicities needs further attention. Yet, the Athero-Express Biobank is unique in its scope and its major strengths are 1) the unique population that is relatively unexplored in the field of FHx and PRS, 2) the prospective study design with detailed data on patient characteristics (including FHx), and 3) the extensive data on plaque composition and morphology that enable us to identify putative pathological mechanisms.
In the future, PRS may be a useful tool for personalized risk prediction for primary or secondary CVE. Adding MetaGRS to a model with traditional risk factors improved prediction of first-CAD events, although model improvement was modest. (19) The clinical utility of PRS for sCVE is still unknown because the incremental value of PRS above clinical factors still needs to be established. One study suggested an additive predictive value of PRS above clinical factors but did not include FHx (38), whereas other studies failed to demonstrate this. (35,36,39) The power of these studies may have been limited due to limited number of CVE. Pooling data of several cohorts including detailed data on preventive strategies and medications during follow-up together with use of uniform outcome definitions for sCVE and uniform PRS composition could help elucidate the clinical value of PRS, for example within international collaborations such as the GENIUS-CHD Consortium.(49) Furthermore, possible sex-differences in the role of risk prediction with PRS need to be further elucidated.
. CC-BY-NC 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. not peer-reviewed) (which was The copyright holder for this preprint . http://dx.doi.org/10.1101/19006718 doi: medRxiv preprint first posted online Sep. 18, 2019 ; In conclusion, both higher MetaGRS and positive FHx were independently associated with increased risk of sCVE in carotid artery stenosis patients undergoing CEA. Higher MetaGRS was also associated with more vulnerable carotid plaque characteristics indicating possible underlying mechanism how genetic variants influence CVD. PRS could identify highrisk individuals and may help selecting future study populations when investigating new therapeutic CVD prevention strategies.       Values are displayed as n/total (%), unless otherwise specified. Values indicated as bold* are statistically significant.

Conflict of interest statement:
MetaGRS was standardized to mean-zero and unit-variance.