Polygenic risk scores (PRS) aggregate the small effects of thousands or millions of common genetic variants into a single risk estimate for a disease. For years the evidence base was almost entirely retrospective: scores validated for association with disease in large biobanks, without a single randomized trial showing that giving someone their score changes what happens to them. That is finally changing.
What did the first outcome-focused PRS trial actually test?
The MI-GENES trial (Myocardial Infarction Genes), registered on ClinicalTrials.gov, randomized participants at intermediate cardiovascular risk to receive either a Framingham risk score alone or a Framingham score plus a disclosed polygenic risk score for coronary heart disease. The published analysis in Circulation: Genomic and Precision Medicine examined whether disclosure of the polygenic component changed downstream cardiovascular events and risk-factor management. Source: Effect of Disclosing a PRS for CHD on Adverse Cardiovascular Events, AHA Journals.
The core finding across this line of research is that disclosure of elevated polygenic risk is associated with modest increases in statin initiation and, in some analyses, improved LDL cholesterol control, compared to standard risk communication alone. That is a real, actionable behavior change signal, not a large one, but a directionally consistent one across the trials in this space.
What is the newest, larger trial testing?
The PROACT trial program, published in JACC in late 2025, went a step further: it used polygenic risk to identify people at elevated genetic risk who were not being flagged by conventional risk calculators, then tested whether detecting and treating subclinical atherosclerosis in that group changed outcomes. This targets exactly the population conventional risk tools are worst at: people who look low-risk by traditional factors but carry high genetic risk. Source: PROACT Clinical Trials, JACC 2025.
Separately, an ongoing trial sponsored by MyOme (NCT06542432) is specifically measuring the clinical utility of adding an integrated polygenic risk score to physician decision-making for coronary artery disease, enrolling roughly 1,000 participants with completion expected in 2029. Source: Utility and Effectiveness of PRS for CAD, ClinicalTrials.gov.
What about cancer screening applications?
The WISDOM study, a national randomized personalized breast cancer screening trial, published a 2025 paper in Genome Medicine describing how it integrated breast cancer polygenic risk scores at scale to personalize screening frequency and starting age, comparing a risk-based screening strategy against annual screening for everyone. Source: Integrating breast cancer PRS at scale in WISDOM, Genome Medicine 2025.
This is a structurally different and arguably more mature use case than disclosure-only cardiovascular trials, because it changes a screening protocol (who gets mammograms, how often, starting when) rather than relying on a patient or clinician to voluntarily act on a number.
| Trial or program | Disease area | What was tested | Status |
|---|---|---|---|
| MI-GENES | Coronary heart disease | PRS disclosure vs. standard risk score | Completed, published |
| PROACT | Coronary/subclinical atherosclerosis | PRS-guided detection and treatment | Results published 2025 |
| MyOme NCT06542432 | Coronary artery disease | PRS impact on physician decisions | Enrolling, completes 2029 |
| WISDOM | Breast cancer | PRS-personalized screening frequency | Ongoing, at-scale integration published 2025 |
| Genomic Medicine at VA | Six diseases including CAD | PRS clinical effectiveness, high genetic risk | Completion 2025 |
What does the older PRS validation literature get wrong when it is oversold?
Most consumer-facing polygenic risk products still lean on retrospective association data: a score's odds ratio in a large cohort study. That tells you the score correlates with disease. It does not tell you that giving someone the score, and then acting on it, produces a better outcome than not knowing it, which is a fundamentally different and much harder question. The trials above are the first generation actually built to answer it, and they remain few, modest in size, and disease-specific.
There is also a well-documented portability problem: PRS derived predominantly from European-ancestry biobanks perform worse, sometimes substantially worse, when applied to individuals of non-European ancestry. None of the major outcome trials above have yet published adequately powered subgroup results across diverse ancestries.
What this does not prove
This evidence does not establish that polygenic risk scores improve outcomes across the dozens of diseases marketed by direct-to-consumer genomics companies, most of which have no outcome trial at all, only association data. It does not resolve the ancestry portability gap. It does not prove that behavior change from disclosure (more statin starts, earlier screening) translates into fewer clinical events at a population level, since most trials so far are underpowered for hard outcomes like myocardial infarction or mortality.
The takeaway
Polygenic risk scores have crossed from purely retrospective association evidence into a small but real body of randomized outcome trials, concentrated in cardiovascular disease and breast cancer screening. The honest read is that disclosure modestly changes clinician and patient behavior, and screening-integration trials like WISDOM are the most promising near-term application. Anyone selling or buying a PRS product for a disease outside these few trial-backed areas is selling association, not proven clinical utility.
