Genetic Risk Scores in Premature Coronary Artery Disease Still Only One Piece of the Prevention Puzzle

Just over a decade has passed since the publication of the seminal genome-wide association studies identifying the first common susceptibility locus for coronary artery disease (CAD) at the 9p21 locus.^1,2 Considerable progress has ensued through ever-expanding meta-analyses of case–control and cohort studies organized by multiple consortia.³ The most recent large-scale meta-analysis published in 2015 by CARDIoGRAMplusC4D involved genome-wide genetic data imputed to the 1000 genomes panel for the first time.⁴ This analysis extended the number of loci reaching genome-wide significance to 58 and identified 202 variants in 129 loci with a P value lower than a false discovery rate threshold of 5%. Overall, these loci explain ≈25% of the heritability of CAD when assuming a multifactorial liability-threshold model, a prevalence of disease of 5%, and a total heritability of CAD of 40%.⁴

The findings of genome-wide association studies to date provide important insights on the genetic architecture of CAD, the causal mechanisms at play, and the predictive power of genetics. First, they confirm the long-suspected polygenic nature of CAD. The increases in risk among variants on the false discovery rate list were found to be uniformly under 15% per allele for common, and no >53% per allele for rare variants.⁴ This observation combined with the modest overall variance explained to date suggests that there are many more, as yet unidentified, variants that contribute to the risk of CAD. Second, they corroborate the role of known causal biological pathways of atherosclerosis; whereas at the same time illuminate the importance of new ones.⁴ The latter is underscored by the observation that only a minority of the CAD loci uncovered to date are also strong signals in genetic studies of traditional risk factors with nearly all of this pleiotropy being observed for cholesterol and blood pressure.⁵ Last, they provide clinicians with additional means to stratify individuals at risk of CAD.⁶ In the case of CAD, multiple studies have already demonstrated the ability for genetic risk scores (GRS) to predict CAD related outcomes independent of traditional risk factors or clinical risk scores.^6,7

In this issue of Circulation: Genomics and Precision Medicine, Thériault et al⁸ provide us with new data concerning the efficacy of a comprehensive GRS for prediction of very early-onset CAD (EOCAD) as defined by the presence of clinically significant disease before the age of 40 years for males and 45 years for females with no obvious secondary cause. In a set of EOCAD cases among the initial release of UK Biobank genetic data and a local cohort receiving care in Hamilton, Canada, they demonstrate that a GRS of CAD involving 182 out of 202 low false discovery rate variants (GRS182) identified by CARDIoGRAMplusC4D is strongly associated with the development of EOCAD independent of traditional risk factors. The magnitude of increased risk was estimated at 84% per SD increase in GRS in 96 participants of the UK Biobank with EOCAD with the score translating to an increased risk of CAD in 22 out of the 30 subjects in the local cohort with at least 7 of these 22 subjects having more than double the baseline risk.

We congratulate the investigators for their efforts given the challenges of collecting a large number of individuals with clinically significant CAD at an age which is at least 15 years younger for males and 20 years younger for females than traditional cutoffs. The incorporation of such a large number of genome-wide association studies variants into a cogent prediction model that offers clear incremental information about CAD risk is also a notable accomplishment. There are a few important points as well as some cautionary notes to be made about the results of this study. First, the overlap between the distribution of the GRS182 among subjects with EOCAD and among subjects with CAD later in life remains substantial. This suggests that this new GRS is not so uniquely informative about EOCAD as much as it is informative across a wider age range that has been demonstrated in previous GRS studies. Second, as the authors point out, the importance of traditional risk factors cannot be underestimated even for EOCAD. For example, the data presented show that diabetes mellitus more than triples the risk of EOCAD in the UK Biobank, whereas hypertension and smoking each approximately double the risk. Furthermore, a majority of the cases are abdominally obese based on a waist-to-hip ratio of >0.9, whereas a majority of the controls are not. Although other risk factors such as physical activity were not examined, the high triglyceride to HDL (high-density lipoprotein) ratio within the local cohort of EOCAD hints toward the presence of significant insulin resistance among EOCAD most likely secondary to abdominal obesity and physical inactivity. These findings are consistent with prior studies demonstrating that traditional risk factors contribute to the same and likely higher degree toward the genesis of EOCAD compared with later-onset CAD.⁹ Last, the environment is likely to still play a major role for EOCAD even in the presence of high polygenic risk given a majority of traditional risk factors themselves are only modestly heritable and one of the single most important risk factors for CAD, smoking, is not at all captured by current polygenic risk scores.⁵ Thus, high polygenic risk, as it is currently defined, is unlikely to be the sole driving factor for any given case of EOCADs and global risk assessment should continue to factor prominently for all individuals.

Family history would intuitively serve as a substitute for genetic risk and perhaps more so for EOCAD. However, the authors observed little change in the magnitude of the genetic risk of CAD when family history was included in their multivariate model as has been observed in other studies.¹⁰ In fact, the point estimate was attenuated by only ≈7% (from 1.84 to 1.78) after adjustment for multiple risk factors in addition to family history concurrently. No clear explanations for these observations exist, but several possibilities exist. First, family history may reflect exposure to common familial environmental risk factors to a larger degree than previously suspected. Alternatively, current GRSs may not capture enough genetic variance to noticeably erode the component of family history which is derived from the inheritance of genetic susceptibility to CAD. Last, genetic contributions to family history may largely reflect rare variants with large effects that are private and not captured by current GRSs. These findings imply that hundreds of SNPs contribute to the heritability of CAD, and the erosion of predictive power of family history may not become substantial until a much larger fraction of common and rare variants is identified and incorporated into the GRS. In this context, the observation of a statistically significant higher prevalence of family history among noncases in the UK Biobank with a high GRS182 compared with noncases with a low GRS182 reassures us that at least some genetic signal is being captured through family history reports.

Without diminishing the value of the current work, this study also highlights how much more needs to be done to translate knowledge of genetic risk into our ultimate goal of eradication of cardiovascular disease. This GRS is not the first, nor will it be the last to demonstrate incremental risk prediction over earlier risk models. The challenge now is less about finding additional improvements in prediction and more about whether clinicians and patients armed with this additional information are able to make decisions that produce fewer events. The broad effectiveness and safety of current primary prevention strategies for CAD (eg, lifestyle, statins, and aspirin) applied on the basis of global risk suggest that a well-powered trial testing the effectiveness of a GRS will likely be successful. However, it is also possible that some aspects of additional risk identified by genetic variants—many of which may be operative through novel mechanisms—may not be as responsive to current prevention therapies as risk defined by conventional means. Thus, the outcome is not certain and needs to be tested. In addition, critical aspects of implementation including clinical laboratory standards, clinic workflows, results reporting, provider, and patient education etc need to be developed and optimized before this new class of information can be generally used to enhance our application of contemporary prevention interventions.

Ultimately, the truly transformative opportunity rests not in improving the current prevention paradigm, but in using new knowledge gained about novel molecular mechanisms to devise an entirely new one. Ongoing mechanistic studies in laboratories across the world are underway to translate this knowledge to novel therapies. Given the complexities and pitfalls of drug development, this translation is expected to be realized in piecemeal fashion over the long-term but remains the best hope for the results from the landmark genomewide association studies of the last decade to transform the landscape of cardiovascular disease in the decades to come.