Using polygenic risk scores to enhance cardiovascular disease risk prediction
George Busby, Allelica CSO & Co-Founder
Cardiovascular diseases — a general term for a suite of diseases that affect the heart or blood vessels — account for one in three deaths globally and are a major drain on public health spending. Our risk of cardiovascular disease increases as we age, but there are also clear and well established links between lifestyle and environmental factors that can further accelerate disease progression. We know that a healthy diet, low in saturated fat, and regular exercise are important steps in controlling disease risk, and public health campaigns across the world centre on these key messages in an effort to reduce the burden of preventable disease on society.
Therapeutic interventions are also available. Large, randomized clinical trials and meta-analyses over the last few decades have repeatedly shown the benefits of using lipid lowering drugs such as statins to combat cardiovascular disease risk. The use of statins, which lower LDL cholesterol levels in an attempt to stave off the atherosclerotic effect of high cholesterol levels, has become widespread. In the US, an estimated 40 million adults are on statins, making them the most commonly prescribed drug in the country.
Risk prediction models are used to assess whether someone is at high risk of cardiovascular disease, and therefore whether behavioral or therapeutic interventions should be administered. These models take into account a range of known clinical risk factors, such as LDL cholesterol, blood pressure, BMI and age to assess 10 year risk of a cardiovascular event. Examples include the Pooled Cohort Equations in the US, and QRISK3 or SCORE2 used in European populations. These algorithms predict the 10 year risk of a cardiac event, placing individuals into one of three or four risk categories. A less than 5% 10 year risk is generally considered low, 5–20% borderline or intermediate, and greater than 20% is high.
Mitigating high cardiovascular disease risk
National guidelines have clear recommendations for next steps when 10 year risk is low (do nothing and reassess in 5 years) or high (initiate high intensity statins) but in the borderline or intermediate risk category, which captures around a third of the population, the recommendations are less clear. Initiating a behavioral change program, for example by performing more exercise can help to reduce BMI, and dietary changes can lower cholesterol. Moderate intensity statins are also an option. The most recent lipid management guidelines however advocate for clinicians and patients to discuss whether there are factors — known as risk enhancers — that could contribute to additional risk, unmeasured by the risk algorithms, but which are nevertheless important for overall cardiovascular health.
At the moment, a variety of risk enhancers, including diabetes, various biomarkers such as elevated lipoprotein A (LpA), South Asian ancestry, and family history of early cardiovascular disease, are currently available to up-classify risk in individuals with borderline / intermediate 10 year risk. But there are currently no genetic risk factors, despite a growing understanding of the clear role of genetic variation in the heritability of cardiovascular disease.
Polygenic risk scores (PRSs) capture the genetic risk from common, low effect variants spread throughout the genome and are gaining increasing prominence as clinical tools. PRS can act as a risk enhancer by interacting with LDL cholesterol to determine our heart attack risk. Within cardiovascular disease, PRSs have repeatedly been shown to stratify risk of coronary artery disease (CAD), a major component of cardiovascular disease, indicating their potential to identify people at high risk of disease. However, challenges to the implementation of PRSs still remain. In particular, the performance PRSs can often vary when the population on which they are intended to be used are different from the ones on which they were developed.
Allelica’s latest work on PRSs — which is now available as a preprint — aimed to develop CAD PRSs that can be used for all individuals regardless of their ancestry background and provide additional evidence into how genetic risk can be used to augment 10 year risk assessments. Specifically, we investigated whether using CAD PRSs to identify individuals with a polygenic risk enhancing factor could better split individuals at borderline / intermediate 10 year risk into those who were at higher overall risk than those without the risk factor.
Developing polygenic risk scores for everyone
As we mentioned above, one challenge with developing and implementing PRSs is that they typically do not transfer across different populations. To overcome this, we decided to build multiple genetic ancestry specific PRSs to develop an approach to categorizing genetic risk that could be applicable to anyone, regardless of their genetic ancestry. We utilized multiple datasets and data sources to build these PRSs and independently validated their performance on additional datasets. Using a novel statistical approach to combine these ancestry specific scores in individuals of mixed genetic ancestry, we further demonstrated their applicability to diverse populations.
Defining high genetic risk
We next developed an innovative approach to defining high risk with these PRSs. We identified the threshold of the PRS distribution where the risk conferred by being above the threshold was twice that below the threshold. This ‘two fold’ risk threshold is well established for other risk factors, for example family history, and allowed us to variously identify between 12% and 24% of individuals, depending on their genetic ancestry, who were at increased genetic risk of CAD. When we inferred who had this PRS risk factor in a set of around 10,000 diverse genetic ancestry individuals who were not used in the PRS development but were known to have had disease or not, we found that those with the risk factor has approximately twice the number of cardiac events as those without the risk factor.
Our analyses show that it is possible to more accurately define cardiovascular risk by taking genetic risk into account, an approach that has the potential to save lives in a sustainable manner by reducing healthcare costs. Given the large burden of cardiovascular disease on public health systems and the potential route to mitigate this risk with behavioral and therapeutic interventions, as well as the clear role for risk enhancers to be used to aid risk discussions, this work increases our understanding of the potential for genetics to be leveraged to reduce the burden of disease.