Cross‐Ancestry Associations of Spontaneous Coronary Artery Dissection Genetic Risk With Coronary Atherosclerosis and Migraine Headache

Abstract

Background

Research studies of spontaneous coronary artery dissection (SCAD) have been primarily focused on European‐ancestry individuals, with limited recognition and investigation in non–European‐ancestry individuals. While SCAD has not been well ascertained in non–European‐ancestry groups, pleiotropic associated traits identified in those of European ancestry have been assessed in individuals of other ancestries. Whether these traits are associated with the complex genetic architecture of SCAD in those of non‐European ancestry has not been previously investigated.

Methods

We investigated the associations of an established SCAD polygenic score with multiple vascular diseases in ≈900 000 ancestrally diverse participants of large‐scale studies. Individual‐level data from the UK Biobank and the Million Veteran Program and summary statistics of publicly available databases were analyzed.

Results

A set of associations between SCAD polygenic score and related vascular diseases were replicated in non‐European samples. Notable associations with the SCAD polygenic score included (1) coronary artery disease, myocardial infarction, and migraine headache in a Hispanic group (coronary artery disease: odds ratio [OR], 0.93 [95% CI, 0.90–0.95]; P=2.35×10⁻⁷; myocardial infarction: OR, 0.88 [95% CI, 0.80–0.96]; P=5.73×10⁻³; migraine headache: OR, 1.03 [95% CI, 1.01–1.06]; P=1.86×10⁻²) of the Million Veteran Program; (2) headache in an African‐ancestry group (OR, 1.22 [95% CI, 1.06–1.41]; P=6.94×10⁻³) and a South Asian–ancestry group (OR, 1.18 [95% CI, 1.02–1.37]; P=2.43×10⁻²) of the UK Biobank; and (3) coronary artery disease, myocardial infarction, and migraine headache in East Asian–ancestry cohorts (coronary artery disease: OR, 0.95 [95% CI, 0.93–0.98]; P=2.66×10⁻³; myocardial infarction: OR, 0.86 [95% CI, 0.83–0.89]; P=9.51×10⁻¹⁶; migraine headache: OR, 1.27 [95% CI, 1.10–1.47]; P=1.03×10⁻³).

Conclusions

Pleiotropic associations of SCAD polygenic risk with related vascular diseases previously identified in European‐ancestry groups showed notable, largely consistent patterns in non–European‐ancestry groups.

Nonstandard Abbreviations and Acronyms

BBJ

Biobank Japan

MVP

Million Veteran Program

PGS

polygenic score

PGS_SCAD

spontaneous coronary artery dissection polygenic score

SCAD

spontaneous coronary artery dissection

UKB

UK Biobank

VGH

Veterans General Hospital, Taiwan

Research Perspective.

What Is New?

• The pleiotropic associations between spontaneous coronary artery dissection polygenic risk and related cardiovascular diseases were consistent in directions and relative magnitudes with that of previously conducted European ancestry‐based studies in African, East Asian, Hispanic, and South Asian ancestry samples, with notably stronger effects observed in Hispanic individuals.

• A new inverse association of the polygenic basis of spontaneous coronary artery dissection with coronary artery calcification was defined in a multiethnic cohort from a coronary artery calcification genome‐wide association study meta‐analysis, which is consistent with the inverse association of spontaneous coronary artery dissection polygenic risk and atherosclerotic coronary artery disease.

What Question Should Be Addressed Next?

• Future efforts will be needed to pursue the pathogenesis of common signals shared between spontaneous coronary artery dissection and related atherosclerotic vascular diseases in well‐powered and multiancestry cohorts.

Spontaneous coronary artery dissection (SCAD) has been acknowledged as an important cause of myocardial infarction (MI), particularly in women aged <50 years. ¹ Many important factors remain to be established regarding the underlying pathophysiological mechanisms of SCAD, including the genetic risk factors and the potential gene–environment interaction. ² , ³ , ⁴ , ⁵ The complex genetic basis of SCAD has been recently defined by modest‐scale genome‐wide association studies (GWASs), identifying common variants reaching genome‐wide significant association (P<5×10⁻⁸) with odds ratios (ORs) of ≈1.5 to 2.0 at the top associated loci. ⁶ , ⁷ Rare variant analysis in SCAD cases has also been performed by several studies, which were challenged by low sample size and limited power for association. ⁸ , ⁹

SCAD has been reported in all ancestry groups, with the majority of cases being individuals of European ancestry, which may be subject to referral and sampling bias. ¹ , ¹⁰ To date, most studies of SCAD have been conducted in individuals of European ancestry, ⁶ , ⁷ , ¹⁰ , ¹¹ , ¹² with a few studies inclusive of Asian samples. ⁷ , ¹⁰ When SCAD is observed in non–European‐ancestry individuals, the clinical characteristics and presentations are similar to those of European‐ancestry individuals. ¹³ , ¹⁴ Given this observation in this relatively underdiagnosed disease, ¹ in all ancestry groups, we hypothesized that the genetic risk for SCAD and pleiotropically associated traits with the complex genetic architecture of SCAD is similar across multiancestry groups. This may help to support the need to maintain a high index of clinical suspicion for SCAD in any patient presenting with MI, regardless of ancestry.

Clinical and biorepository resources of SCAD in non‐European samples are currently lacking, and SCAD itself has not been ascertained in adequately large‐scale population cohorts, which limits the direct assessment of SCAD genetic risk in non‐European samples. Given that genetic variants associated with SCAD have been associated with multiple related vascular traits and diseases in European‐ancestry individuals, both individually and combined in a polygenic score (PGS) for SCAD (PGS_SCAD), ¹ , ² , ⁶ , ⁷ , ¹¹ we exploited these relationships and investigated the associations of a PGS_SCAD with related vascular diseases in ancestrally diverse participants of large‐scale studies. Our objective was to assess whether vascular diseases previously identified in European samples to be pleiotropic with SCAD polygenic risk, including coronary artery disease (CAD), MI, and migraine headache, show consistent patterns in non‐European samples.

Methods

Data Availability Statement

The individual‐level data of the UK Biobank (UKB) are available at https://www.ukbiobank.ac.uk/ with formal application for access. Individual‐level data for the Million Veteran Program (MVP) are available through a Veterans Affairs Central Institutional Review Board–approved research protocol to qualified investigators. Additional data used in the analyses of this article may be made available from the corresponding author on reasonable request.

SCAD PGS Computation

The SCAD polygenic risk in multiancestry groups were quantified using a PGS approach. The PGS_SCAD was defined as previously published in Saw et al, ⁷ as this PGS is based upon the only multiethnic SCAD GWAS done to date; other GWAS and GWAS/meta‐analysis publications to date exclusively studied individuals of European ancestry. ⁶ , ¹² The weighted (PGS_SCAD) was based on 7 independent SCAD‐associated single‐nucleotide polymorphisms (SNPs) (rs11207415, rs12740679, rs78377252, rs9349379, rs78349783, rs11172113, and rs28451064) (Table S1). The PGS_SCAD for the kth individual was calculated using the formula: $\sum _{i=1}^7{\beta }_i\times {\mathit{SNP}}_{\mathit{ik}}$, where ${\beta }_i$ refers to the effect size of the ith SNP, and SNP _ik refers to the number of SCAD‐associated risk alleles for the ith SNP of the kth individual. We computed PGS_SCAD across 5 ancestry groups (African, American, East Asian, European, and South Asian) of 1000 Genome phase 3 data ¹⁵ to investigate the distribution patterns of PGS_SCAD across different ancestry groups. In addition, for each of the 7 SNPs in terms of index SNP, proxy SNP R ² (±500 kb) and linkage disequilibrium blocks (±60 kb) were computed by LDlink ¹⁶ to examine the linkage disequilibrium patterns using the YRI (African), CHB+JPT+CHS (East Asian), and CEU (European) populations of 1000 Genome phase 3 data as the reference panel, respectively.

PGS Association Analysis in MVP

We investigated the association of PGS_SCAD with CAD, MI, and migraine headache using data from 4 populations of MVP as defined by the Harmonizing Genetic Ancestry and Self‐Identified Race/Ethnicity algorithm ¹⁷ : African, Asian, European, and Hispanic. The study received ethics and study protocol approval from the Veterans Affairs Central Institutional Review Board, with informed consent obtained from all participants in the study. The details of the data processing procedures for the phenotype and genotype data of MVP were presented in a recently published multipopulation GWAS of CAD. ¹⁸ Briefly, for each Harmonizing Genetic Ancestry and Self‐Identified Race/Ethnicity group in turn, we computed the PGS_SCAD for each individual using individual‐level genotype data and defined cases and controls based on phecodes using individual‐level phenotype data. We performed association testing by fitting a logistic regression with year of birth, sex, and top 10 principal components as covariates in each ancestry group. Odds ratios (ORs) were reported per 1‐SD unit increase for all results of this study, with SD computed specifically in each analysis cohort. Sex‐stratified analyses were also performed by fitting the logistic regression model in female and male subgroups with year of birth and top 10 principal components as covariates, respectively. In addition to the main analyses based on PGS_SCAD, we conducted 2 sensitivity analyses. First, as the effect sizes of the 7 SNPs were obtained from a GWAS with majority of European‐ancestry samples, we also investigated the association of CAD, MI, and migraine headache with the unweighted PGS_SCAD calculated as the sum of SCAD‐associated risk allele counts. Second, as the chromosome 6p24.1 PHACTR1 locus rs9349379 was previously identified to be associated with atherosclerosis‐associated CAD, MI, and migraine headache, ⁷ , ¹¹ , ¹⁹ , ²⁰ we investigated the association of these 3 diseases with the weighted PGS_SCAD removing rs9349379.

PGS Association Analysis in UKB

We evaluated the SCAD polygenic risk using individual‐level phenotype and genotype data from non‐European participants of the UKB. ²¹ This study has been conducted using the UKB resources under Application Number 30686, with informed consent of all participants obtained through the UKB. The UKB recruited 502 618 participants aged 40 to 69 years from across the United Kingdom between 2006 and 2010. Each participant was assessed for comprehensive phenotypic and health‐related information, including biological measurements, lifestyle indicators, biomarkers in blood and urine, and imaging of the body and brain. ²² The identification of non–European‐ancestry individuals was primarily based on the self‐reported ancestry background (data field 21 000 in the UKB phenotype file) and was confirmed by genotype‐inferred ancestry obtained through the SCOPE (scalable population structure inference) software. ²³ Due to the small sample size of non–European‐ancestry groups in the UKB, we mainly focused on participants who self‐reported as Africans (code 4, 4001, 4002, 4003) and South Asians (code 3001, 3002, 3003). We filtered out individuals (1) who are not included in the UKB version 3 imputed genotype data; (2) who are included in the White British ancestry subset; (3) who are identified as outliers in heterozygosity and missing rates; (4) who have unmatched self‐submitted and genotype‐inferred sex; (5) who have sex chromosome aneuploidy, defined as putatively carrying abnormal sex chromosome configurations; and (6) who are not included in the genotype principal component computation. A total of 6996 African‐ancestry individuals and 6915 South Asian–ancestry individuals were included in our analysis.

To determine whether direct assessment of SCAD polygenic risk was feasible in the UKB, we first queried the case sample sizes of coronary artery aneurysm and dissection in the African‐ and South Asian–ancestry groups of the UKB. The case individuals were identified using phecode 411.41, which consists of 3 International Classification of Diseases, Tenth Revision (ICD‐10) codes I25.3, I25.4, and I34.1 on the basis of the Phecode Map 1.2 ²⁴ (https://phewascatalog.org/phecodes_icd10). In the UKB, only 7 cases were identified in the South Asian–ancestry group and no cases were identified in the African‐ancestry group.

We then performed association analyses to investigate the pleiotropic associations of PGS_SCAD with migraine headache in the African‐ and South Asian–ancestry groups of the UKB. There was insufficient phenotyping or too few CAD and MI cases for us to further investigate associations of PGS_SCAD with other SCAD‐pleiotropic cardiovascular disease in African‐ and South Asian–ancestry groups of the UKB. Specifically, we first computed PGS_SCAD for each individual using the genotypes of the 7 SNPs in the UKB version 3 imputed genotype data and the corresponding SNP effect sizes as described previously; the PGS_SCAD was standardized to have a mean of 0 and a SD of 1. We defined migraine cases as we have done previously ⁷ on the basis of the self‐reported code for data field 20 002. Specifically, we defined migraine cases as individuals who self‐reported as having migraine (code 1265) at least 1 time for their 3 assessment center visits. To further increase the power of analysis and because migraine is classified on the basis of subjective criteria, we also included individuals who self‐reported as having nonmigraine headache (code 1436) at least 1 time for their 3 visits and combined them with the previously defined migraine cases as patients with headache. In each ancestry group, we fitted a logistic regression model for association testing with age, sex, and top 10 principal components adjusted as covariates. Sex‐stratified analyses were performed to evaluate potential sex‐specific associations, and interaction tests of PGS_SCAD and sex were also performed to investigate the association differences across sex strata. Sensitivity analyses based on individual SNP, unweighted PGS_SCAD, and the weighted PGS_SCAD removing rs9349379 were also performed. In addition to the main analyses based on self‐reported headache, we also investigated the association of PGS_SCAD with phecode‐based migraine headache. The phecode‐based migraine headache cases were defined as individuals with phecodes of migraine headache (340 or 340.1). ¹⁹

PGS Association Analysis Using GWAS Summary Statistics

In addition to the PGS association analysis using individual‐level data of MVP and UKB, we leveraged publicly available GWAS summary statistics of external large‐scale non‐European studies to investigate the association of PGS_SCAD with CAD, MI, and migraine headache in East Asian–ancestry individuals. Specifically, we obtained the summary statistics of the 7 SCAD‐associated SNPs described previously from CAD and MI GWAS conducted by the Biobank Japan (BBJ) ²⁵ , ²⁶ and migraine headache GWAS conducted by the Veterans General Hospital, Taiwan (VGH). ²⁷ For SNPs that were missing in the external summary statistics, we replaced them with the best proxy SNP available in the data set, which was defined as the SNP achieving the maximum r ² among all SNPs having r ²>0.5 with the index SNP using 1000 Genomes Project phase 3 data ¹⁵ as the reference panel. The search of proxy SNPs was conducted using PLINK software version 1.90b6.9). ²⁸ For PGS association testing, we leveraged a summary statistics–based method implemented in the R package gtx (https://www.rdocumentation.org/packages/gtx/versions/0.0.8), which approximates the regression of vascular disease status onto PGS_SCAD in the corresponding data set, and we aligned the β coefficients and effect alleles between summary statistics of external studies and SCAD GWAS for model fitting. In addition, we also investigated the association between PGS_SCAD and coronary artery calcification (CAC) in a multiethnic cohort following the same procedure described above, given that CAC is a proxy intermediate phenotype for CAD. The summary statistics of the 7 SNPs were obtained from a multiancestry GWAS meta‐analysis of CAC, ²⁹ which comprised 26 909 individuals of European ancestry and 8867 individuals of African ancestry.

Results

We first queried for SCAD cases available for analysis and found the number of individuals of non‐European ancestry with SCAD was insufficient for any analysis (<10 cases), whereas vascular traits previously demonstrated to be pleiotropic with the polygenic basis of SCAD in those of European ancestry were identified in sufficient sample sizes for analysis. For the current multiethnic analysis, we leveraged an existing PGS_SCAD based upon a GWAS of SCAD that did not restrict its analysis by ancestry or ethnicity ⁷ and was previously demonstrated in European‐ancestry individuals to be associated with CAD, MI, and migraine headache, as well as familial risk of SCAD. ³ While in the UKB atherosclerotic forms of CAD were insufficient for analysis (<100 cases), in the MVP, larger sample sizes of cases were identified (N_African=18 440, N_{East Asian}=858, N_European=101 794, N_Hispanic=7299). CAC GWAS summary statistics were available from a study of 26 909 European‐ancestry individuals and 8867 African‐ancestry individuals. ²⁹ The number of cases with headache (N_African=179, N_{South Asian}=190) diagnoses were adequate for a modest‐scale analysis in the UKB, and a higher‐powered analysis was feasible in the MVP resource (N_African=12 879, N_{East Asian}=900, N_European=38 219, N_Hispanic=5946). Summary statistic data from East Asian analyses of CAD (N_CAD=29 319, N_control=183 134) and migraine headache (N_headache=1005, N_control=1053) diagnoses allowed for analyses of pleiotropy as well.

The distributions of PGS_SCAD across 5 ancestry groups of 1000 Genome phase 3 data are displayed in Figure S1, showing largely consistent distribution patterns. As shown in the scatterplots of proxy SNP R ² and heat maps of the linkage disequilibrium matrix in Figures S2 through S8, no major differences in linkage disequilibrium structures were identified across African‐, East Asian–, and European‐ancestry groups of 1000 Genome phase 3 data, except for rs78377252 (chr2) and rs28451064 (chr 21) that are monoallelic in the East Asian– and African‐ancestry groups.

Association analyses of the PGS_SCAD with CAD, MI, and migraine headache were performed in 4 different ancestry groups of MVP: African (≈92 000), East Asian (≈7000), European (≈350 000), and Hispanic (≈41 000) (Figure). For CAD and MI, we observed at least nominally significant (P<0.05) inverse associations with PGS_SCAD in European‐ancestry individuals (CAD: N_CAD=101 794, N_control=240 886; OR, 0.94 [95% CI, 0.93–0.95]; P=4.25×10⁻⁵³; MI: N_MI=5250, N_control=240 886; OR, 0.95 [95% CI, 0.93–0.98]; P=4.85×10⁻⁴) and Hispanic‐ancestry individuals (CAD: N_CAD=7299, N_control=33 546; OR, 0.93 [95% CI, 0.90–0.95]; P=2.35×10⁻⁷; MI: N_MI=481, N_control=33 546; OR, 0.88 [95% CI, 0.80–0.96]; P=5.73×10⁻³) with OR reported per 1‐SD unit increase for all results of this study (Figure [A]). These results are consistent with our prior report of this association in the European‐ancestry groups of the UKB and MVP with a smaller sample size. ⁷ The associations of PGS_SCAD with CAD and MI in the African‐ancestry group (N_CAD=18 440, N_MI=1467, N_control=72 647) and East Asian–ancestry group (N_CAD=858, N_MI=60, N_control=6150) did not reach nominal significance. For migraine headache, nominally significant associations with PGS_SCAD were observed in European‐ancestry individuals (N_migraine=38 219, N_control=312 845; OR, 1.04 [95% CI, 1.03–1.05]; P=6.37×10⁻¹²) and Hispanic individuals (N_migraine=5946, N_control=35 426; OR, 1.03 [95% CI, 1.01–1.06]; P=1.86×10⁻²), whereas significant associations were not observed in the African‐ancestry group (N_migraine=12 879, N_control=79 769) or East Asian–ancestry group (N_migraine=900, N_control=6216) (Figure [B]). In sex‐stratified analyses, the association results of the male subgroups were largely consistent with that of the combined‐sex group, and we did not observe significant associations in most of the female subgroups, which had approximately 10% of the sample size of men, with the only exception of a nominally significant (P<0.05) positive association of PGS_SCAD with migraine headache observed in the female subgroup of the European‐ancestry group (Figure [C] and Table S2). Sensitivity analyses were performed with the unweighted PGS_SCAD and the weighted PGS_SCAD removing rs9349379; these results were largely consistent with the results of the analysis of the PGS_SCAD (Tables S3 and S4).

Figure 1. Analyses of PGS_SCAD in ancestrally diverse samples.

The forest plots of the ORs and 95% CIs show associations of PGS_SCAD with (A) CAD/MI in the MVP and BBJ; (B) headache in the UKB and migraine headache in the MVP and VGH; and (C) CAD, MI, and migraine headache in the sex‐stratified subgroups of the MVP and headache in the sex‐stratified subgroups of the UKB. Cohorts marked with an asterisk indicate that results were obtained from analyses previously reported in Saw et al. ⁷ OR is reported per 1‐SD unit increase. The MVP and UKB analyses were performed by fitting logistic regression adjusting for age (year of birth for the MVP analyses), sex (in the combined‐sex analyses only), and top 10 principal components using individual‐level genotype/phenotype data. The BBJ and VGH analyses were performed using genome‐wide association study summary statistics from each specific ancestry cohort. BBJ, BioBank Japan; CAD, coronary artery disease; MI, myocardial infarction; MVP, Million Veteran Program; OR, odds ratio; PGS_SCAD, SCAD polygenic score; UKB, UK Biobank; and VGH, Veterans General Hospital, Taiwan.

We then investigated the association of PGS_SCAD with migraine headache in the African‐ and South Asian–ancestry groups of the UKB. There were 114 individuals with migraine headache and 69 individuals with general headache diagnosis codes out of 6996 African‐ancestry individuals, and 141 individuals with migraine headache and 51 individuals with general headache diagnosis out of 6915 South Asian–ancestry individuals. Because some individuals reported both migraine headache and general headache without specification, and to optimize statistical power in this analysis, the 2 headache categories were combined for analysis, yielding a total of 179 African‐ancestry headache cases and 190 South Asian–ancestry headache cases. Nominally significant (P<0.05) associations between PGS_SCAD and headache were observed in the analyses of African‐ancestry individuals (N_headache=179, N_control=6817; OR, 1.22 [95% CI, 1.06–1.41]; P=6.94×10⁻³) and South Asian–ancestry individuals (N_headache=190, N_control=6725; OR, =1.18 [95% CI, 1.02–1.37]; P=2.43×10⁻²) (Figure [B]). These associations were consistent in direction and magnitude with associations previously defined in European‐ancestry individuals of the UKB. ⁷ The PGS_SCAD results appeared to be driven primarily by an association of rs12740679 on chromosome 1q21.2 (OR, 1.41 [95% CI, 1.12–1.79]; P=4.21×10⁻³) in the African‐ancestry group and by an association of rs11172113 on chromosome 12q13.3 (OR, 1.38 [95% CI, 1.11–1.73]; P=4.14×10⁻³) in the South Asian–ancestry group, respectively (Table S5). Sex‐stratified analyses identified nominally significant associations in the male subgroup of African‐ancestry individuals and the female subgroup of South Asian–ancestry individuals (Figure [C] and Table S2), and no statistically significant interactions between PGS_SCAD and sex were identified in the African (β=0.19 [95% CI, −0.11 to 0.50]; P=2.19×10⁻¹) or South Asian (β=−0.16 [95% CI, −0.48 to 0.17]; P=3.50×10⁻¹) ancestry groups of the UKB. The association of PGS_SCAD with phecode‐based migraine headache in the African‐ancestry group (N_migraine=49, N_control=6947; OR, 1.03 [95% CI, 0.77–1.36]; P=8.64×10⁻¹) and South Asian–ancestry group (N_migraine=64, N_control=6851; OR, 0.96 [95% CI, 0.75–1.23]; P=7.40×10⁻¹) did not reach nominal significance, likely due to the low case sample sizes defined by phecodes. Sensitivity analyses using the unweighted PGS_SCAD and the weighted PGS_SCAD removing rs9349379 remained nominally significant and were consistent with the results based on PGS_SCAD (Tables S3 and S4).

To further maximize our evaluation of pleiotropic associations in multiancestry samples, we leveraged a summary statistics–based method to investigate the associations of PGS_SCAD with CAD, MI, and migraine headache in East Asian individuals (Figure [A] and [B]). From the CAD GWAS conducted by the BBJ, ²⁵ 4 SNPs comprising the PGS_SCAD were available, and no proxy SNPs were identified for the remaining 3 SNPs (rs78377252, rs9349379, and rs28451064). Despite this, we observed a nominally significant inverse association of PGS_SCAD with CAD (N_CAD=29 319, N_control=183 134; OR, 0.95 [95% CI, 0.93–0.98]; P=2.66×10⁻³), consistent with our previous findings of the protective effect of increased SCAD polygenic risk against CAD. ⁷ rs9349379, which has previously been reported as CAD associated, was not included in this analysis. From the MI GWAS conducted by the BBJ, ²⁶ 6 SNPs of the PGS_SCAD were available; no proxy SNP was identified for rs78377252. We observed a strong inverse association of PGS_SCAD with MI (N_MI=14 992, N_control=146 214; OR, 0.86 [95% CI, 0.83–0.89]; P=9.51×10⁻¹⁶), which is again consistent with the inverse association previously identified in the European‐ancestry individuals of MVP. ⁷ Finally, we investigated the association of PGS_SCAD with migraine headache on the basis of the summary statistics from VGH, ²⁷ with 5 SNPs available and no proxy SNP identified for the remaining 2 SNPs (rs78377252 and rs28451064). We observed a nominally significant positive association of PGS_SCAD with migraine headache (N_migraine=1005, N_control=1053; OR, 1.27 [95% CI, 1.10–1.47]; P=1.03×10⁻³), which is consistent with the associations identified in the previous European‐ancestry–based study ⁷ and in the African, European, Hispanic, and South Asian individuals of the current study.

Finally, as a proxy intermediate phenotype for CAD ³⁰ and to further validate the inverse association between atherosclerotic CAD/MI and SCAD polygenic risk, we accessed available resources to evaluate CAC, a measure of subclinical atherosclerosis. The summary statistics of a multiancestry GWAS meta‐analysis of CAC ²⁹ were available for the PGS association analysis, with 6 SNPs available (no proxy SNP for rs78377252 was identified) in the multiancestry meta‐analysis summary statistics, and all 7 SNPs available in the European‐ancestry summary statistics. We observed a strong inverse association of PGS_SCAD with CAC in both the meta‐analyzed multiancestry data (N=35 776; β=−0.12 [95% CI, −0.16 to −0.09]; P=8.11×10⁻¹²) and the European‐only data (N=26 909; β=−0.13 [95% CI, −0.17 to −0.10]; P=6.78×10⁻¹²), which is directionally consistent with the inverse associations of PGS_SCAD with CAD and MI identified in the European and Hispanic individuals of MVP and in the East Asian individuals of the BBJ. Sensitivity analyses removing rs9349379 also identified nominally significant associations in both data sets (multiancestry meta‐analysis: β=−0.06 [95% CI, −0.09 to −0.02]; P=3.47×10⁻³; European‐only analysis: β=−0.06 [95% CI, −0.10 to −0.02]; P=3.35×10⁻³).

Discussion

Taking the results of these analyses in total, there were several results corroborating in non‐European groups a consistent pattern of pleiotropic associations of PGS_SCAD that have to date been defined only in European‐ancestry studies. Although the consistency of pleiotropic associations does not provide direct evidence for similar, specific SCAD genetic risk in African, East Asian, Hispanic, and South Asian populations, the results support a potentially shared complex genetic basis of these diseases across ancestries. Notable findings include the results of analyses of Hispanic individuals who demonstrated notably stronger pleiotropic associations, and this deserves particular attention in future studies. The extension of the pleiotropy of SCAD risk to CAC, an intermediate phenotype of coronary atherosclerosis, was defined in a multiethnic cohort from a CAC GWAS meta‐analysis and may be informative about the biology of SCAD. Specifically, this result underscores that SCAD is not an atherosclerotic process, consistent with histologic analysis of SCAD autopsy specimens, which typically show neither atheromatous plaques nor calcifications.

The current understanding of the polygenic and complex genetic architecture of most diseases and traits is primarily based upon studies performed in European‐ancestry individuals, which has been a persistent trend over time. ¹ , ² The low population prevalence of SCAD challenges research efforts for SCAD, as SCAD has been vastly underrecognized and underdiagnosed in the clinic, until more recent years, during which growing awareness and advocacy has led to a surge in diagnoses, facilitating current research of SCAD. ¹ It was notable that in our search for data sets in which to evaluate SCAD polygenic risk in non–European‐ancestry groups, the available resources are extremely sparse and insufficient for analysis; beyond SCAD, even the numbers of individuals with related vascular traits were limited. The resources needed to address our study question, which was already shifted away from primary analysis of SCAD due to lack of ascertainment of the disease in multiancestry groups, were limited. The query of non–European‐ancestry samples in large‐scale data resources indicates the relative low sample size and related underpowering of analyses of non‐European individuals in the current large‐scale genetic databases; their unbalanced proportions in the current study may not reflect the general population structure. Therefore, future efforts with recruiting strategies to allow for increased sampling will be important to enhance our understanding of population‐specific SCAD risk. Our findings particularly highlight potentially high risk in Hispanic individuals, which to date have not been specifically studied. Future corrective action to expand the population resources is urgently needed, ³¹ as has been well documented elsewhere, and ongoing studies of SCAD would benefit from specific efforts to augment ancestrally diverse samples of individuals so that SCAD risk may be directly ascertained in multiple ancestry groups. Importantly, ancestry‐specific pleiotropic associations may also be driven by nongenetic factors that are challenging to be accounted for, in particular social and structural factors that may vary across ancestry groups. Thus, leveraging nongenetic factors may further benefit the assessment of SCAD genetic risk in future research efforts.

Limitations of the current study include the small sample size of non‐European individuals in the UKB and other existing large‐scale studies, limited SCAD or SCAD‐related vascular disease studies targeting non‐European populations, and the lack of comprehensive diagnostic information for accurate phenotype ascertainment in the UKB. Increasing the sample size of non‐European individuals would be expected to improve the power of identifying significant pleiotropic association, and adequate availability of multiethnic vascular disease studies can possibly provide more evidence for the consistent patterns of pleiotropic associations across different ancestries. Insufficient resources to query non‐European individuals highlights the problem of underrepresentation in the current SCAD studies and reference data sets. Current SCAD registry data are likely to highlight the significant disparities, and intentional efforts to enroll non–European‐ancestry individuals with SCAD will be needed to not only define the significance of already defined associations from studies of largely or all European ancestry individuals, but also identify and evaluate potential contributions of novel, ancestry‐specific associations. Finally, we note that the Harmonizing Genetic Ancestry and Self‐Identified Race/Ethnicity–defined Hispanic cohort in MVP is heterogeneous with respect to the proportion of White, Indigenous, and African‐ancestry admixture, ³² while the Asian subgroup also included a small fraction of South Asian individuals. ³³ Future efforts may benefit from more precise grouping of individuals based strictly on genetically inferred ancestry for more accurate and effective quantification of genetic risks.

Despite the limitations to this work, the results of the current study suggest SCAD genetic risk in non‐European groups by leveraging large sample sizes of multiethnic cohort resources to define associations of SCAD polygenic risk with SCAD‐associated vascular diseases in diverse samples. The parallels of association patterns and the observed stronger pleiotropic effects in Hispanic individuals, which to our knowledge has not been evaluated in other studies, highlight a specific gap for which further targeted investigation efforts are needed. Overall, further downstream analyses of common signals shared between SCAD and related atherosclerotic vascular diseases in well‐powered multiethnic cohorts would be useful to clarify the complex genetic architecture of SCAD in multiancestry populations.

Conclusions

Our evaluations of vascular diseases previously identified in European samples to be pleiotropic associated with SCAD polygenic risk defined notable, largely consistent patterns in non‐European samples. The consistent association patterns support a potentially shared complex genetic basis of vascular diseases across ancestries. The previously defined inverse association of the polygenic basis of SCAD with atherosclerotic forms of CAD and MI were further corroborated by our finding of an inverse association with coronary artery calcium, a vascular trait that has not been previously analyzed for association with SCAD. As our understanding of the genetic basis of SCAD is improving from current research investigations, further effort is needed to expand the sufficiency of resources for research of genetic factors contributing to the SCAD pathogenesis in ancestrally diverse samples.

Sources of Funding

This work was supported by the National Institutes of Health (R35HL161016) and Veterans Administration awards I01‐BX003362 and MVP000. S.K.G. was supported by National Institutes of Health (R01HL139672, R01HL122684, R01HL086694, R35HL161016), Department of Defense, and the University of Michigan A. Alfred Taubman Institute. X.Z. was supported by R01HG009124.

Disclosures

The authors declare no competing financial interests. The University of Michigan and University of British Columbia have filed for patents on genetic risk stratification tools for SCAD and other vascular diseases.

Supporting information

Data S1

Tables S1–S5

Figures S1–S8