Clinical applications of polygenic risk score for coronary artery disease through the life course

Abstract

Coronary artery disease (CAD) remains a leading cause of morbidity and mortality worldwide, highlighting the limitations of current primary and secondary prevention frameworks. In this review we detail how the polygenic risk score for CAD can improve our current preventive and treatment frameworks across three clinical applications that span the life course: (i) identification and treatment of people at increased risk early in the life course prior to the onset of clinical risk factors, (ii) improving the precision around risk estimation in middle age, and (ii) guiding treatment decisions and enabling more efficient clinical trials even after the onset of CAD. We end by summarizing the efforts needed as we head towards more widespread use of polygenic risk score for CAD in clinical practice.

1. Introduction

Identifying individuals at high risk for CAD as early as possible remains a principal strategy for contemporary CAD prevention. CAD is prevalent throughout most of the life course manifesting for some clinically even in their late thirties, and even more frequently subclinically early in life [1–4]. While advances in primary prevention through behavioral and pharmacological interventions have curbed initial troubling trends in the 20th century, cardiovascular disease still retains the distinction as being the leading driver of mortality [1,5,6]. We live in an unprecedented era of innovation in biomedicine, and one might wonder: why do millions of people still die of CAD every year?

To completely understand our inability to curb the burden of CAD, it is important to acknowledge that the predisposition to this disease starts from the time of birth. The proportion of phenotypic variation of CAD explained by genetic factors, also known as heritability, is estimated to be 40 to 60% [7]. Analyses of exceptional families led to the discovery and molecular characterization of familial hypercholesterolemia, an incomplete autosomal dominant monogenic condition increasing the risk of CAD through elevation of low-density lipoprotein (LDL) cholesterol [8,9]. Large population datasets with genetic data however have now confirmed that genetic variants leading to familial hypercholesterolemia are only present in about 1 in 250 people and on average triple the risk of CAD [10–12]. Now CAD genome-wide association studies (GWAS), which have been increasing in scale ever since 2007, have uncovered a broad genetic architecture often unexplained by traditional risk factors [13,14]. The latest GWAS for CAD published at the end of 2022 compared about 180,000 people with and a million without CAD and identified more than 250 genetic loci that modify the risk of CAD [15]. Regardless of mechanism, it is now well-established that those loci could be aggregated into a single metric, the polygenic score (PRS) for CAD, that can be used to predict CAD risk for any one individual based on their genomic profile [16,17].

Genomic risk stratification augments our ability to identify individuals at risk that could benefit from preventive efforts earlier in their life course. Primordial prevention targeted at early life (child and adolescent) prevention of risk factors of cardiovascular disease has been proposed as a solution for prevention in general, but identifying who to target is a critical first step [18]. Risk for CAD is due to a combination of genomic factors present from time of birth and nongenomic factors that accumulate through the life course and every individual’s profile of risk for both components is unique [19]. Accurate and dynamic risk estimation for the individual over their life course is important to successful and efficient prevention.

There are three key limitations of the current framework of prevention, and for which the application of PRS in clinical practice could help solve as we will highlight in this review paper. First, we do not initiate preventive strategies early enough for high-risk patients. We will discuss how a PRS could be used to incentivize and prioritize behavioral changes and pharmacological LDL cholesterol lowering at younger age prior to the onset of clinical risk factors. Second, we are not precise enough in our risk estimation. We will discuss how PRS could improve precision in risk estimation when added to clinical risk estimators typically used in middle age. Third, even among patients who develop CAD, recurrent risk remains high in middle age and beyond, and we are not successful at reversing the trend. We will discuss the promise of PRS as a potential biomarker to personalize secondary prevention and enrich clinical trials for novel therapies.

2. Polygenic risk score for coronary artery disease

2.1. Polygenic risk score development

The precursor to the development of a PRS is a GWAS on a large and diverse set of participants that has identified genetic variants associated with the trait of interest. Each variant is associated with an effect size (i.e., relative enrichment of a risk allele among CAD cases versus controls in a GWAS). In simple terms, the PRS is a single number that captures the additive effects of risk-increasing variants for a single individual (Figure 1). Early versions of polygenic score calculation consisted of simply summed effects of the number of variants that achieved genome-wide significance in a GWAS study. Since then, several computational approaches have been developed to compute more powerful polygenic scores from GWAS data [20–23]. These methods overall improve performance by (i) weighting by GWAS effect size, (ii) using information from many more variants (sometimes millions) rather than limiting to those that are genome-wide significant, (iii) borrowing information from related traits such as blood pressure, lipids, peripheral artery disease, stroke, and others for CAD, and (iv) factoring in population structure and other methods to improve cross-ethnic performance.

Figure 1:

Simplified framework of how a polygenic risk score is developed and reported

In a simplified three-step framework of how polygenic risk score (PRS) for CAD is developed, first risk-increasing (red) and risk-decreasing (green) variants are compared in people with and without CAD to estimate effect size. This is known as a genome-wide association study (GWAS). Second, in any individual the summed effect of all risk-increasing and risk-decreasing variants is computed. At a population level, this results in a single number that quantifies risk and is normally distributed in the population. Third, the number is expressed as a percentile, with higher percentiles associating with a relative risk increase compared to the population average.

This illustration was adopted from www.polygenicscores.org which includes a more detailed explanation.

The typical development of any risk score (genetic or non-genetic) also consists of two steps: training and validation in independent populations. In the training step, multiple PRS across a range of methods and tuning parameters are calculated and the association of each with the trait is compared to select a best-performing score. In a validation step on a separate set of participants, the best-performing score is reported. Multiple performance metrics are used including the correlation coefficient of the score with a trait for continuous traits such as blood pressure or lipid measures, and the odds ratio or hazard ratio of disease per standard deviation of the score for binary traits such as prevalence or incident CAD, respectively [24]. Additionally, the area-under-the-curve or C-statistic is computed and could be compared for a model consisting of PRS alone vs. PRS with other conventional clinical variables. Models are typically adjusted for at least age, sex, and principal components of genetic ancestry in their development, with conventional clinical risk factors also considered for CAD [25].

The ability to calculate a PRS is becoming increasingly egalitarian. Easy-to-use software packages are now available [21,26,27]. Even more importantly, the PGS Catalog is a public repository that now includes most previously published PRS organized by trait, number of variants, with reported performance metrics, and available for download and secondary use in an unrelated target dataset of interest [27]. For example, there are 38 CAD PRS available in the PGS Catalog at the time of writing this paper [27].

2.2. Risk prediction

The primary intention of CAD PRS is CAD risk prediction. PRS is a continuous measure that is normally distributed in a population, and risk prediction is often modeled based on the liability threshold model, where CAD manifests after reaching a certain threshold of risk on the continuous polygenic score distribution [28]. CAD PRS however is a risk measure and not a diagnostic test for CAD and should be evaluated similarly to other conventional risk factors that are measured continuously such as LDL cholesterol and systolic blood pressure.

A common way of defining PRS is by using percentile rank in the target population of interest. Since PRS is a unitless measure derived from comparing people with disease to people without the disease, it has to be converted to a more useful unit to measure and communicate risk. When PRS is computed in any target reference population, it is normally distributed and often normalized and converted to percentile ranks. Unlike gene variant findings which can be interpreted by sequencing single individuals, making sense of a PRS number for an individual requires a reference population, to which this number is projected to define percentile rank and relative risk. Implementation studies of PRS in clinical practice are helping standardize the use of reference populations so that a PRS for prospectively recruited patients could be quickly interpreted. Using this model, implementation researchers and genetic testing companies offering PRS can generate a report to one patient at a time.

Multiple studies by our group and others showed that PRS predicts incident CAD [17,29–32]. In many of those analyses, an individual can fall in each of the 100 PRS percentile ranks (1st to 99th) and as such have a specific risk level associated with this percentile rank – for example, individuals in the top 1% of the population have the highest risk increase. A typical PRS cutoff to define high risk is the top 20% of the population. In a contemporary CAD PRS, the top 20% of the population has three-fold increased risk of CAD compared to the population average [32]. Notably, this is equivalent to the risk of CAD among carriers of monogenic familial hypercholesterolemia related variants, who represent only ~0.4% of the population [10,11]. We and others have also shown that the risk from monogenic and polygenic drivers is additive [29,30,33,34].

PRS needs to be considered in the context of other available clinical risk factors. Uniquely, PRS may be “the first risk factor” as it is established at conception prior to the onset of clinical risk factors [35]. Even in middle age however, CAD PRS provides information that is orthogonal to conventional risk factors. In analyses exploring the interplay of PRS and conventional risk factors, (i) PRS alone provided more information that any single conventional risk factor, (ii) combined models that included both PRS and conventional risk models provided the best performance in risk stratification, and (iii) individuals identified at high risk using PRS are often found to be “flying under the radar” – not often having the the typical clinical risk factors that would identify individuals at risk in clinic [36].

2.3. Clinical utility and reporting

The promise of clinical utility of PRS followed from its ability to identify a large proportion of people at increased risk based on genetic information alone, and that are not readily identified using conventional risk factors. However, identifying people at risk alone is not enough, as it should be coupled with the ability to mitigate that risk. Retrospective studies have shown that individuals with high polygenic risk have an attenuated risk of CAD if they adhered to a heart-healthy lifestyle, highlighting that genetic risk is not immutable and may be offset by adopting healthy lifestyle behaviors [37]. Post hoc analyses of clinical trials have also shown that individuals with high polygenic risk derive greater absolute and relative cardiovascular risk reduction from LDL cholesterol-lowering therapy compared to the general trial population [38–40]. Such relative effect differences have not been observed for clinical risk factors. These observations highlight an opportunity to use PRS to enrich clinical trials for new therapies for CAD, at least for LDL cholesterol-lowering therapies [41].

Prospective implementation data of CAD PRS is also starting to emerge, with early results showing positive impact on healthy behavior and LDL cholesterol lowering [42–44]. One of the first challenges of implementation is figuring out the best mechanism of how CAD PRS could be integrated in clinical practice. This is then followed by prospective evidence generation for efficacy. Multiple implementation models are already published with early positive results, including a combined clinico-genomic risk score that results in lower LDL cholesterol compared to clinical risk score alone in the MIGENES trial, a preventive genomics clinic with disclosure of combined monogenic and polygenic results paired with clinic visits that resulted in change in clinical management in 40% of participants, and even a smartphone-based application disclosure of CAD PRS result which increased statin uptake [42–44]. All three studies had no clear signal for harm and overall provided encouraging early outcomes data to justify larger prospective trials. Some of those are already ongoing such as the eMERGE network’s effort to return genome-informed risk assessment for multiple conditions including CAD to 25,000 participants in the US, the Genomic Medicine at VA Study (GenoVA) which will measure the clinical effectiveness of PRS in reducing time to diagnosis in ~1000 participants, the Our Future Health which will be returning PRS for up to 5 million volunteers in the UK, and others [45–48]. Notably, CAD PRS is increasingly available in clinical settings already. While some direct-to-consumer genetic testing companies report CAD PRS, such reports still require clinical confirmation at this time [49].

3. Clinical application of PRS for coronary artery disease through the life course

Based on the developments to-date in CAD PRS, its clinical application is emerging in three key areas that will increasingly be part of our routine practice, and that span the life course of an individual (Fig. 2). First, as “the first risk factor” PRS can identify people at high risk of CAD early in their life course prior to the onset of risk factors and serve as a single reason for initiation of preventive therapies such as lifestyle changes and statin to lower LDL cholesterol. Second, in middle-age and in the presence of other conventional risk factors, PRS can improve precision in risk estimation to guide initiation and dosing of preventive therapies. Third, even after the onset of CAD, PRS might have a role in prevention of recurrent events and enrichment of clinical trials to study new interventions for CAD. In the below sections, we will summarize the current state of the evidence that supports each of the three use cases and highlight what remains to be done in the years to come.

Fig. 2:

Clinical application of polygenic risk score for CAD through the life course

Polygenic risk score (PRS) for CAD could identify people at high risk in young age such as those in the 18 to 40 years group prior to onset of clinical risk factors and serve as the single risk factor to decide on initiation of targeted interventions including LDL cholesterol lowering. In early middle age (40-55 years) after the onset of clinical risk factors such as diabetes, hypercholesterolemia and hypertension, CAD PRS has been shown to improve precision in risk estimation when added to clinical risk calculators. Finally, even after the onset of CAD such as in people older than 55 years, CAD PRS predicts recurrent events and could be used to personalize treatment strategies and enrich clinical trials.

3.1. Earlier identification and treatment of people at risk

The risk for CAD starts accumulating early in life due to the cumulative exposure to risk factors of atherosclerosis. Plaque deposition starts building up in childhood and accumulates with age due to the exposure to risk factors [18,19]. Both the duration and dose of exposure (i.e. the “area under the curve”) to a modifiable risk factor, such as smoking or LDL cholesterol, matter [50]. Conversely, minimizing the exposure dose and duration (i.e., “primordial prevention”) is an effective strategy to prevent atherosclerosis from developing or progressing into clinical CAD [18,51,52]. In clinical practice however, most targeted prevention consideration for CAD starts in middle age when risk factors are detected prompting risk estimation, which is overweighted by chronological age [53]. This is often too late. Furthermore, even though risk factor burden is increasing in younger adults, the age-dependent intermediate-term risk estimation framework still misses earlier opportunities for prevention [54].

The trajectories of risk based on CAD PRS start diverting early in the life course. The ability of the PRS to identify people at risk is even more relevant and pronounced at younger ages [30,55–57]. For example, the hazard ratio of CAD per standard deviation of PRS is higher at younger ages [51] highlighting its unique strengths in risk prediction earlier in life. Specifically for LDL cholesterol lowering, it is now well-established that “the earlier, the lower, the better” [50–52,59]. As such, for individuals at increased risk based on their PRS, it is reasonable to consider initiation of statin therapy for LDL cholesterol lowering based on this information.

Comparing risks to other risk factors used to prompt statin initiation may better inform its utility. The top 20% of the population of PRS distribution have an equivalent risk of CAD to carriers of familial hypercholesterolemia (FH) variants, which is about a three-fold increase [29,32]. The top 21% of the PRS have an equivalent risk to diabetes, and the top 29% have an equivalent risk to severe hypercholesterolemia (LDL cholesterol ≥ 190 mg/dL), both of which are risk factors that alone justify initiation of statin therapy under the current American College of Cardiology (ACC) and American Heart Association (AHA) guidelines [32,53]. Overall, there is sufficient evidence to justify initiation of statin therapy based on extreme polygenic risk alone. As PRS are improving in their power to stratify disease risk through more training data and advanced methods, the population that could benefit from may further increase. For example, in a 2018 score, the top 8% had an equivalent risk to FH, but in the most recent score (GPSmult), this proportion increased to the top 20% [17,32].

3.2. Improved discrimination of risk in middle-age

Modern preventive cardiology is centered on 10-year risk prediction optimized in middle-aged individuals. As most cardiovascular cohorts are also middle-aged participants, most PRS evaluations have been in this context. When compared to risk models including individual conventional risk factors such as LDL cholesterol, blood pressure, and smoking, a model including CAD PRS has the highest C-statistic for a single risk factor [36]. The ACC/AHA Pooled Cohort Equations (PCE) is used in the US to estimate 10-year atherosclerotic cardiovascular (ASCVD) risk and consequently inform statin therapy initiation. Models that include the PCE and CAD PRS have an increased C-statistic by about 1 to 3% depending on the cohort age and characteristics compared to a model including PCE alone [32,36,57]. Comparable results are observed for the QRISK3 risk prediction algorithm used in the UK [60,61]. Within the strata of the PCE used in clinical guidelines (low, borderline, intermediate, and high), CAD PRS can further stratify risk. At the intermediate risk cut-off (10-year ASCVD risk of 7.5%) above which the ACC/AHA guidelines recommend considering statin therapy, the net reclassification index achieved by adding PRS is about 7% [32]. Since the PCE is highly driven by age while PRS is not, the value of adding PRS to risk prediction using the PCE is higher at younger age (eg. 40-55) compared to over 55 years [57,58]. Notably because PRS is a continuous measure of risk, individuals with low CAD PRS benefit from increased protection throughout their life course compared to the rest of the population and a recalibration of the PCE absolute risk measure using CAD PRS information might help de-risk people with low CAD PRS below thresholds for treatment with statins [36,57].

Multiple ‘risk enhancers’ meeting similar levels of evidence are in the current ACC/AHA guidelines [53]. Since the risk from high CAD PRS is comparable to the effect size of risk enhancers, CAD PRS may also be considered a risk enhancer [62]. An alternative model is the development, validation, and dissemination of integrated risk calculators that factor both conventional and genomic risk factors. While this has been performed in retrospective datasets, moving into implementation is more challenging as it requires standardization of testing and reporting of PRS, validation and calibration of a novel risk calculator across different populations, and development of a scalable implementation framework [24,63]. We will discuss the challenges on implementation that remain to be overcome in a subsequent section of this review.

Particularly relevant to improving risk discrimination in middle age is the use of coronary artery calcium (CAC) score. Current guidelines support measuring CAC in selected middle-aged adults of intermediate clinical risk to aid in risk decision and initiation of statin [53]. A CAC of zero is most helpful as it indicates lower risk and supports a decision to defer initiation of statin [53]. There are three important concepts to consider in understanding the interplay between CAC and CAD PRS. First, CAD PRS is a static measure based on genetic information available from time of birth and as such provides information about lifetime risk, while CAC is a marker of subclinical atherosclerosis often present in middle age and provides strong predictive information about short-term risk. Second, CAD PRS predicts CAC, consistent with its ability to predict CAD phenotypes [64]. One framework could consider CAC as a clinical test that might be helpful in individuals with high CAD PRS [44]. This approach of using CAD PRS to triage who should get a CAC score is being tested in the ESCALATE clinical trial [65]. Third, among late middle-aged individuals with prevalent clinical risk factors, CAC provides superior information to CAD PRS but there are mixed reports on how CAD PRS might further augment risk stratification such as with identification of non-calcified plaque [64,66,67].

3.3. Prevention of recurrent events

Despite maximal therapies after the diagnosis of CAD, recurrent event rate remains high, necessitating new opportunities to refine and mitigate residual risk [68,69]. Even in contemporary clinical trials of secondary prevention, ~10% still develop a myocardial infarction at one year [70–72]. There are three arguments why this is the case, and it is likely a combination of all three. First, once CAD has manifested itself clinically, you are already too late and the “horse is out of the barn”. This argument is partially true, and it remains critical to prevent CAD earlier in the life course as we have discussed in this review. Age at the first CAD event is one of the strongest predictors of recurrent events [73]. Second, after the onset of CAD, we are not aggressive enough in reversing the trajectory and controlling risk factors. This argument is also true as many people with CAD do not achieve guideline-defined targets for treatment due to non-compliance, general failure of our chronic disease care model, and suboptimal personalization of risk prediction and risk of recurrent events [74–76]. Third, we are not targeting the full spectrum of biological risk of CAD and there is an opportunity to target novel pathways. As CAD PRS predicts clinical response to LDL cholesterol lowering and has been shown to predict recurrent events, it may have potential role in treatment intensification as well as efficient clinical trial design [39–41,77]. At minimum, CAD PRS may help identify individuals more likely to have recurrent events enabling smaller clinical trial sizes or shorter durations.

In a recent study, CAD PRS is the top predictor for predicting recurrent CAD events across a range of studied factors [78]. Among approximately 7000 middle-aged participants with a prior CAD event at mean age of 57 years and more than a decade of follow-up, the CAD PRS was independently associated with recurrent CAD events with a hazard ratio per standard deviation (HR/SD) of 1.12 (95% confidence interval [CI] 1.05-1.19). Among placebo-treated individuals with CAD in a PCSK9 inhibitor trial, a high CAD PRS was associated with 1.65-fold increased risk of recurrent coronary events compared to a low PRS [39]. As CAD PRS is associated with the burden of subclinical coronary atherosclerosis, it may also capture more advanced atherosclerosis among those with clinical CAD [37,38,79]. Additionally, since CAD PRS is derived from GWAS, it captures biological information related to CAD not explained by conventional risk factors. Indeed approximately half of associated risk alleles do not exhibit notable pleiotropy with clinical risk factors [36,80]. Collectively, these observations provide the opportunity for using CAD PRS as a biomarker to capture phenotypic and biologic heterogeneity of CAD at its onset and guide secondary prevention decisions.

Finally, CAD PRS can be used as a tool to enrich clinical trials of CAD to help explore novel mechanisms of treatment. In post hoc analyses of clinical trials, we have previously shown that CAD PRS provides powerful predictive and prognostic enrichment – there is a markedly increased absolute and relative effect size among individuals with high PRS compared to the rest of the trial population [41]. This approach for enrichment may improve power and reduce sample size and follow-up time of clinical trials, making the clinical investigation of novel therapeutics for CAD more efficient. In our upcoming PROACT studies (polygenic-based identification of subclinical coronary atherosclerosis), we will be using this approach of PRS-based enrichment and coronary plaque phenotyping using CCTA to prospectively test lifestyle and pharmacological interventions [81,82].

4. Ushering a new era of polygenic risk score for CAD in clinical practice

Despite the strong evidence for clinical utility of CAD PRS through the life course, its widespread adoption in routine clinical practice requires continued efforts on five key dimensions (Table 1). First, because genomic data used to develop polygenic scores are enriched for individuals of European ancestry, the performance of CAD PRS is variable and often reduced across ancestries outside Europe – most notably, it under-performs in individuals of African ancestries [83–85]. This cross-ancestry difference in performance has been a top priority in genomics research for the past few years with both efforts to increase large dataset representation and improved computational methods to improve portability of PRS across ancestries [22,86]. Both better scores that perform comparably across multiple ancestries as well as ancestry-optimized scores have been approached [87]. For example, the performance of the most recent GPSmult score in predicting incident CAD events was comparable among European, East Asian, and South Asian groups, with HR/SD of 1.75 (95% CI 1.71-1.78), 1.72 (95% CI 1.13-2.60), and 1.65 (95% CI 1.49-1.77), respectively. However, score performance was weaker among individuals of African ancestry with HR/SD of 1.25 (95% CI 1.07-1.46) [32]. Several biomarkers in cardiovascular medicine, including lipoprotein(a), show differential distributions and risk profiles across populations. Expected differences of PRS across populations need to be defined. The strength of association of the PRS with CAD is not the only determinant of public health impact, but also the population prevalence of CAD which can impact the net benefit. For example, in one analysis in the US, the CAD PRS improved the estimation of absolute risk of myocardial infarction in Black individuals more than in White individuals despite a weaker association of the score with CAD [84].

Table 1:

Challenges and ongoing efforts for widespread implementation of polygenic risk score for CAD in clinical practice

Challenge	Ongoing effort(s)
Reduced cross-ancestry performance	Increasing diversity of large genomic studies Improved computational methods
Standardization of reporting	Hospital system implementation efforts Direct-to-consumer genetic testing companies
Scalable implementation frameworks	Our Future Health eMERGE
Guideline adoption	AHA Statement Increasing advocacy and awareness
Insufficient prospective studies	GenoVA PROACT

eMERGE: Electronic Medical Records and Genetics Network, AHA: American Heart Association, GenoVA: The Genomic Medicine at VA Study, PROACT: Polygenic Risk-Based Detection of Subclinical Coronary Atherosclerosis Studies.

Second, standardization in reporting of PRS is needed with calibration across populations [24,86]. Currently, multiple PRS exist due to different population/ancestries and methods, and all are reported as percentiles of a population distribution according to the cohort studied. Each percentile is associated with a relative risk compared to a population average. Emerging methods capturing radial distance of genetic similarity may advance current practices of binning by genetic ancestry [88]. Universal harmonization efforts are needed to improve the reporting of PRS beyond single studies or health systems. Additionally, the fact that interpreting a PRS requires a population reference population might complicate its implementation and adoption by clinically-oriented researchers. Third, scalable implementation efforts are necessary to establish that the early observations of no harm and positive changes seen in smaller studies are sustained in larger and more diverse populations. Many of those efforts are already under way. For example, the GenoVA study in the US Veterans Affairs health system has already established analytic validity and reporting framework and is prospectively returning results to participants [63]. Similarly, the eMERGE network and the Our Future Health in the US and UK respectively are large scale implementation efforts [45,46]. Fourth, despite similar evidence in some scenarios as risk factors (i.e., like LDL cholesterol early in life, and like risk enhancers in middle age), society guidelines have yet to incorporate CAD PRS into practice guidelines. A recent statement by the ACC/AHA on PRS for cardiovascular disease helped chart this path [89]. Fifth, we may see more clinical trials leveraging CAD PRS as an entry point in the years to come in order to further establish clinical utility of novel pharmacological interventions for CAD beyond the current standard of care [41,80].

5. Conclusions and perspectives

Polygenic risk score for CAD can identify people at risk of future CAD events early in the life course prior to the onset of conventional risk factors, improve precision in risk estimation in middle age even in the presence of risk factors, and serve as an enriching prognostic and predictive biomarker even after the onset of clinical CAD. Its clinical utility across the life course has been established by massive global cohorts and biobanks, post hoc analyses of clinical trials where genetic data was collected, and small-scale implementation efforts. Larger scale prospective implementation studies, standardization in reporting, adoption in clinical guidelines, and increasing use in clinical trials may further facilitate its use in clinical practice.

Highlights.

• CAD PRS can identify people at high risk of CAD early in the life course prior to the onset of risk factors who are likely to derive great benefit from targeted behavioral interventions and LDL cholesterol lowering with statin therapy.

• In middle age and in the presence of conventional risk factors, CAD PRS can augment performance of conventional risk calculators and improve precision in risk prediction.

• PRS can predict recurrent events even after the onset of clinical CAD and holds a promise to serve as a biomarker in guiding treatment modalities and enriching clinical trials of new therapies.

• Early implementation studies of CAD PRS validate the proposed clinical utility and suggest no harm, but larger scale implementation efforts are underway to establish those observations in diverse populations.