A novel polygenic risk score for intracranial aneurysms

Since the completion of the first genome-wide association studies in the early 2000s, polygenic risk prediction has raised considerable expectations for research and clinical use.¹ Despite methodological concerns about construction and validity of polygenic risk scores (PRS)², multiple studies have shown that PRS strongly associate with disease status³ and the number of studies related to PRS is increasing exponentially (based on a PubMed search for the terms: “polygenic risk score*” OR “genetic risk score*” OR “genetic score*” OR “polygenic score*”, as of 23/11/2022). Most commonly, polygenic risk scores, also known as genetic risk scores (GRS), represent a defined set of risk variants based on findings from genome-wide association studies. For each individual, the number of risk alleles is summed and weighted by its effect size.³ Therefore, PRS can discriminate individuals based on their genetic susceptibility to a particular trait or disease.

Much of the evidence supporting the clinical utility of PRS comes from well-powered studies of cardiovascular diseases, type 2 diabetes, cancer, and Alzheimer’s disease.³ Conversely, previous PRS studies on intracranial aneurysms (IAs) are small and scarce.⁴ Whereas the first PRS study on IA could not find an association between a PRS and aneurysm size⁵, follow-up work showed that the PRS is higher in individuals with IAs located at the middle cerebral artery as compared to all other locations.⁶ Another study found that an improved PRS associated with aneurysm diameter and volume, but not with aneurysm presence.⁷ Notably, the cited studies were based on no more than 10 genetic variants which amounted to a SNP-based heritability of approximately 4%.^5,6 In addition, associations with IA outcomes were evaluated in small cohorts of fewer than 2000 cases.^5–7 The statistical power of these studies to detect meaningful associations was therefore limited.

In this issue of STROKE, Bakker et al.⁸ address this gap by creating a novel genetic risk score of IA. The authors leverage a recent genome-wide association study of IA and aneurysmal subarachnoid hemorrhage (ASAH)⁹ along with association data of 17 IA-related traits, to create a ‘meta Genetic Risk Score’ (metaGRS). Among the 17 selected traits are established IA risk factors, such as blood pressure and smoking, and diseases genetically correlated with IA, including ischemic stroke.⁸ To construct the score, the authors utilized an ‘elastic net regression’ methodology in a training sample from the UK Biobank (1,161 IA cases, 407,392 controls). In this way a total of 7,078,955 SNPs were included in the score. First, the authors evaluated associations of the metaGRS with ASAH incidence and IA presence in the HUNT study (828 IA cases, 68,568 controls). While the score improved ASAH incidence prediction above a model including clinical risk factors sex, blood pressure, and smoking (C-index 0.63 to 0.65), the prediction of IA presence did not increase after including the score. The authors also showed that the prediction of ASAH incidence was stronger in women than in men by using scores trained and validated in women and men separately. Next, a higher score independently associated with 3 out of 9 tested IA patient characteristics in the ISGC-IA cohort (5,560 IA cases), namely hypertension, smoking status, and lower age at ASAH. In addition, a lower metaGRS was observed in patients with a single IA compared to multiple IAs, and in patients with an IA at the internal carotid artery. The latter two associations, however, were not independent of smoking and hypertension in a multivariate model.

The predictive value of the metaGRS remains limited, given that it did not improve prediction of IA presence above a model including clinical risk factors. One reason for this could be the presence of undetected and unruptured IAs in the control group, as suggested by the authors⁸, which would attenuate the statistical power to predict IAs. Only an improved characterization of the controls via brain vessel imaging (e.g., MRA, CTA) can resolve this issue. Furthermore, a previously shown association with IAs located at the middle cerebral artery⁶ could not be replicated. Conversely, a decreased genetic load in patients with an IA at the internal carotid artery was identified, a location that was not included in the previous study. Lastly, the PRS is based on individuals of European ancestry, resulting in reduced predictive accuracy in other populations.³

Despite these limitations, the novel PRS by Bakker et al.⁸ is a major improvement over previous scores. The statistical power is considerably larger, as the utilized genome-wide association study more than doubled the number of cases compared to previous studies (7,495 IA/ASAH cases, 71,934 controls).⁹ SNP-based heritability increased to 21.6%, thereby explaining more than half of the twin-based heritability of IA (h²=41%).⁹ More than 7000 IA cases were used for creating and evaluating the metaGRS.⁹ Further, the authors used 3 different cohorts to train and validate their score (UK Biobank, HUNT, ISGC-IA).

In closing, the investigators constructed a metaGRS with predictive ability for ASAH, although with limited added value over standard clinical risk factors. This finding emphasizes the need for patient-specific vascular risk factor control in the setting of IA, specifically optimizing blood pressure and smoking cessation. However, their findings that prediction by the metaGRS for ASAH performed better in women than in men, and in subjects at a younger age, independent of both hypertension and smoking, further highlights that genetic drivers also play a key role. While these results do not indicate that regular use of the metaGRS is clinically warranted, they certainly highlight that both genetic and environmental factors jointly contribute to disease risk and that additional research in needed to further clarify these relationships.

Non-standard Abbreviations and Acronyms

ASAH

aneurysmal subarachnoid hemorrhage

GRS

genetic risk score

IA

intracranial aneurysm

PRS

polygenic risk score