Building the environment for genetic risk prediction for cardiovascular disease

7 min readMar 5, 2024


George Busby, Allelica CSO & Co-Founder

Bringing a new medical test to market requires generating and compiling a range of evidence. A commonly used framework for evaluating novel genetic tests involves describing an assessment of three main aspects of the test: its analytical validity, clinical validity and clinical utility.

A recent publication led by the Broad Institute (Lennon et al 2024) articulated many of the processes necessary for the development of a clinical test, and we’re proud that much of the work we’ve done at Allelica on PRS test development over the last few years is captured in their paper. Here we’ll discuss the three components of genetic test evaluation in the context of cardiovascular disease PRSs.

Our aim is to discuss whether — to paraphrase Eric Topol — if PRSs are not ready for prime time yet, then when will they be?

PRSs identify people at high genetic risk of disease

First, a reminder that a polygenic risk score (PRS) is a number that represents an individual’s genetic risk of disease from common genetic variants. We have covered how PRSs are developed and tested in the past, and will continue to discuss how Allelica and others are innovating to build better polygenic risk scores in the future. But for now, let’s assume that we have developed a robust PRS for cardiovascular disease: what do we need to do to turn this PRS into a clinical test?

Analytical validity: is the input data for the genetic test reliable?

The raw data that goes into a clinical PRS test is genome-wide genotyping data from a microarray or whole genome sequencing data. This needs to be generated in a systematic way that is robust, reproducible and scalable. Since we’re interested in building a clinical test, the lab processing the raw data will need to develop a pipeline to demonstrate this reproducibility.

To achieve this, the lab can run DNA from samples through their sequencing machines for which we also have so-called “orthogonal” data that acts as our truth. Orthogonal data is data that has been generated from a sample that can be used, in our case, for computing a PRS value, but which has not been generated within the same lab environment or technology as the data that will be generated for our clinical test.

Individuals from the 1000 Genomes Project, for example, have both publicly available whole genome sequence (WGS) data and DNA available, making them good candidates for our validation. Obtaining DNA from these samples and running them through the genotyping platform of choice within a lab means that the lab results can be compared to the WGS truth data. A range of statistics, such as precision and accuracy can be computed to provide an objective test of how reliably a lab can generate PRS values from DNA samples moving forward.

Developing an experimental setup to prove analytical validity for genotypes and PRSs means that the data generated by a lab, or set of labs, can be trusted to produce accurate results for a clinical test.

Clinical validity: does the test predict meaningful clinical risk?

So, we’ve built a robust pipeline for the accurate generation of genetic data and PRSs. Next, we need to be able to show that the PRS actually predicts risk accurately. There is little point, and it can in fact be potentially dangerous, to develop a clinical test for a PRS that doesn’t actually predict risk.

Here we need to carefully choose a range of different clinical datasets upon which we can demonstrate performance of our PRS. These clinical datasets should match as closely as possible the populations on which the test is to be performed. As we showed in our recent paper (Busby et al 2023) a key consideration is the genetic ancestry distribution of the datasets that are used to perform this part of the validation.

Allelica’s PRSs for Coronary Artery Disease show good prediction across a range of genetic ancestry groups and cohorts (source: Busby et al 2023)

Ideally we want to show that the PRSs that will be used for a given patient population have the ability to predict an established level of high risk. We use a two fold increase in risk compared to the average as our default high risk threshold. We chose this threshold due to its equivalence with other commonly used risk factors as well as its ability to capture a significant number of individuals (between 8% and 20% of individuals depending on genetic ancestry) at high risk. Tests of clinical validity will show that this level of predictive accuracy can be achieved in a range of different populations.

Clinical utility: do test results lead to differences in disease management?

Tests of clinical utility involve understanding whether the results of a test can lead to improved outcomes. Clinical utility is actually a relatively elastic concept, with definitions that range from “the ability of a screening or diagnostic test to prevent or ameliorate adverse health outcomes such as mortality, morbidity, or disability through the adoption of efficacious treatments conditioned on test results“ to “any use of test results to inform clinical decision-making… [or] any outcomes considered important to individuals and families.”

One way to assess clinical utility is through a large-scale randomized clinical trial (RCT), where people are randomized into a treatment arm in which individuals are provided with information from their PRS, and a control arm in which they aren’t. However, as we’ll see below, some clinical trials have already found positive results and a number of additional trials are already underway, and not just for cardiovascular disease. Based on the definition above, and the nature of a PRS, it is actually possible to assess utility in a number of ways that don’t require the time and resources required for an RCT.

In the context of PRSs for cardiovascular disease, a range of studies have focused on the perception and information of disease risk gleaned from PRSs, which as we have discussed above, are aspects for clinical utility. For example:

  • In a randomized clinical trial, individuals who were told they had a high PRS for coronary artery disease reduced their LDL cholesterol to lower levels than who were were only provided risk estimated from clinical factors (Kullo et al 2016)
  • Primary care providers understood and endorsed the use of PRS to guide decision making in cardiovascular disease prevention (Vassy et al 2023)
  • Augmenting assessments of monogenic cardiovascular disease risk are perceived by patients to be beneficial and lead to changes in clinical management of risk (Maamari et al 2023).
  • A tool that integrated clinical cardiovascular disease risk with PRS provided usable preventative information for both physicians and patients (Fuat et al 2024).

The test works. Now what?

So there is clear evidence that analytical validity of a genotyping assay can be achieved and that PRSs for cardiovascular disease have both clinical validity and utility — what else needs to happen for this test to be broadly available?

One route forward is to utilize the US Laboratory Developed Test (LDT) framework. And in fact, Allelica has already developed a number of clinical PRSs with CLIA/CAP laboratories. Our support has enabled several labs to perform the required validations to run these tests, and our publications provide additional support for the clinical validity and utility of these tests.

To move to widespread adoption of these beneficial clinical tests, we also need to consider the economics associated with performing these tests and move towards a model where the costs of the test are shouldered by health systems and insurers because of their benefit to the population. From this point of view, we have also provided evidence of economic benefit to deploying PRS tests for cardiovascular disease at the health system level (Mujwara et al 2022). By identifying people at high genetic risk with a PRS and providing risk mitigation approaches such as lifestyle modification and therapeutics, PRSs reduce the overall burden of cardiovascular disease and save health systems money.

While the development of a clinical PRS test requires a number of critical steps, these have already been accomplished for cardiovascular disease. What’s more, PRSs are also able to identify people who are at high risk, but who are invisible to current clinical risk assessments (Aragam et al 2020). This means that PRSs capture an aspect of cardiovascular disease risk which cannot be identified through other means. We therefore have a tremendous opportunity to use PRSs in cardiovascular disease prevention now.

In theUS today, there are approximately 160 million adults aged between 35 and 75 years. About 40 million (25%) of these adults are at intermediate 10 year risk of developing cardiovascular diseases. If PRS tests are used in combination with clinical risk assessments on this population, we estimate that around 200,000 coronary artery disease events, or an average of 20,000 heart attacks per year, could be averted over the next 10 years. Genetic risk can reclassify their clinical risk and be used to accelerate the uptake of interventions such as behaviour and dietary modification, or even lipid-lowering therapy, that are known to reduce risk. Combined with a strong case for economic benefit for these tests, the question that now needs to be answered is: if PRSs are not ready for primetime now, then when will they be?

If you’re interested in providing PRS tests to your patients or developing clinical PRS tests in your lab please get in touch with us today.




Allelica is a Software Genomics Company developing algorithms and digital tools to accelerate the integration of Polygenic Risk Score in the clinical practice