Telling people their polygenic risk score changes almost nothing

Being told your genetic risk of disease, as a polygenic risk score, does not appear to change what you eat, whether you attend screening, whether you start statins, how anxious you feel, or your cholesterol. A June 2026 systematic review pooled all 27 randomised trials that have tested this and found no significant effect on any of the 22 outcomes measured in two or more studies [s1].

The distinction the finding rests on is important. A polygenic risk score can be a real predictor of who develops a disease; that is a separate claim from whether telling people the score improves their health. The trials say it largely does not — at least in the short term, and at least so far.

What a polygenic risk score is meant to do

A polygenic risk score adds up the small effects of many common genetic variants to estimate a person's inherited risk of a condition — a common heart disease, a cancer, type 2 diabetes. The commercial and clinical pitch is that handing someone this number will motivate healthier behaviour or better-targeted prevention: the "high-risk" patient exercises, takes a statin, or turns up for screening.

That motivational claim is what the review set out to test, searching the Cochrane CENTRAL register and PubMed from inception to 1 March 2025 for randomised trials that compared disclosing a polygenic risk score against giving no genetic information [s1].

The pooled result

Of 7,830 articles screened, 27 randomised trials qualified [s1]. Most scored risk for cancer (9 trials), cardiovascular disease (8) or diabetes (6); 21 were run mainly in healthy populations [s1]. Across every outcome that appeared in at least two trials, the meta-analysis found no significant effect [s1].

The specific estimates cluster tightly around no effect. For diet, the standardised mean difference in daily energy intake was −0.11 (95% CI −0.22 to 0.01) and for alcohol consumption −0.11 (−0.28 to 0.06) [s1]. Physical activity barely moved: −0.01 (−0.13 to 0.11) [s1]. Screening attendance carried a relative risk of 1.12 (0.77 to 1.61), statin use 1.50 (0.98 to 2.29) and disease incidence 0.95 (0.32 to 2.79) — every interval crossing 1, meaning no reliable effect [s1]. For LDL cholesterol, the pooled mean difference was −3.64 mg/dL (−7.88 to 0.60) [s1].

Two anxieties often raised about genetic risk information — that it either frightens people or falsely reassures them — were not borne out at the group level either. The standardised mean difference for anxiety was −0.02 (−0.13 to 0.08), with similarly null results for worry, perceived risk and depression [s1]. Disclosure neither harmed nor helped psychologically, on average.

The quieter, sharper finding

The review's most useful contribution is about how this literature is written. Of the 27 trials, 15 concluded their abstracts with favourable claims about the polygenic risk score — but only 5 of those 15 had any significant result to justify the claim [s1]. Nine of the 27 trials were at high risk of bias [s1].

There is also a fragmentation problem that makes the field look busier than it is. The trials reported 80 outcomes across multiple studies, yet each of those 80 was reported in only a single trial [s1]. Of the 19 that were primary outcomes, just two were statistically significant [s1]. When almost every outcome is measured once, positive results cannot be replicated and cannot be pooled — and isolated positives are exactly what marketing quotes.

Where a single trial did move a number

The pooled null does not mean disclosure never changes anything. The MI-GENES trial, one of the more rigorous single studies, randomised 203 adults aged 45 to 65 at intermediate heart-disease risk to receive their risk estimate either with or without a genetic risk score, followed by counselling and a shared decision about statins [s2]. At six months, the group that got the genetic score had a lower LDL cholesterol than the group that did not — 96.5 versus 105.9 mg/dL (P = 0.04) — with the largest drop among those told they were high-risk (92.3 mg/dL; P = 0.02) [s2].

MI-GENES shows the effect is not impossible; it is small, inconsistent across trials, and — in that study — tied to counselling and a statin decision rather than to the number alone. Pool that trial with the rest and the average benefit disappears [s1].

What it means

Polygenic scores are advancing fast as predictors, and their clinical role is real where they refine who to screen or treat — as in risk-based breast screening or melanoma risk stratification. That is a different use from the one tested here: handing a consumer or patient a risk number and expecting behaviour to follow.

The site has covered how these scores are being built into national programmes, from Taiwan's precision-medicine effort to premarital genomic screening in Dubai. This meta-analysis is the counterweight: prediction is not persuasion, and a score by itself is not an intervention. It also fits a broader pattern in which consumer risk and "biological age" tests return numbers that rarely change outcomes.

The authors' own conclusion is that, despite frequent promising claims, disclosing polygenic risk scores did not produce meaningful changes in behavioural, psychological or clinical measures — while noting the trials' heterogeneity and limitations [s1]. Longer trials, embedding scores in genuine clinical pathways rather than testing the number in isolation, are what would settle whether the prediction can be turned into benefit.

Nothing here is medical advice. Whether a polygenic risk score is useful for any individual is a clinical question, and this article does not recommend for or against testing.

Sources

1. [s1] Effects of polygenic risk score communication on short term health outcomes: systematic review and meta-analysis. BMJ Medicine, published online June 12, 2026. https://doi.org/10.1136/bmjmed-2025-002347
2. [s2] Incorporating a Genetic Risk Score Into Coronary Heart Disease Risk Estimates: Effect on Low-Density Lipoprotein Cholesterol Levels (the MI-GENES Clinical Trial). Circulation, published online February 26, 2016. https://doi.org/10.1161/CIRCULATIONAHA.115.020109