Summary
A Nature Medicine correspondence argues that debate over polygenic scores in England’s NHS depends partly on what prevention is meant to achieve. The authors say risk thresholds, outcome measures and comparators shape how the evidence is judged.
Whether polygenic scores should be considered for NHS prevention policy is not only a question of how well they predict risk, argue Hayley Wilson, Chantal Babb de Villiers and Peter Mills in a Nature Medicine correspondence published on 25 September 2026. They say debate also depends on what health services expect the scores to do and which outcomes should count as success.
A polygenic score combines information from many genetic variants into an estimate of a person’s genetic predisposition to a trait or disease. It can help distinguish levels of risk, but a risk estimate is not a diagnosis.
The authors place the discussion within England’s longer-term health plans. The 10 Year Health Plan for England commits to establishing a Genomics Population Health Service by 2035 and names polygenic scores as a component technology. The National Cancer Plan commits to personalised risk assessment, in which the authors say genomic population stratification is likely to play a part. They also point to a February 2026 PHG Foundation roundtable convened to consider what responsible use of scores in the NHS would require.
Why prevention goals change the debate
A central distinction is between identifying people at the highest individual risk and improving health across a whole population. The authors describe a prevention paradox: because the moderate-risk group is much larger, many cases of common disease can arise among its members, even when each person’s risk is lower than that of the small high-risk group.
A more accurate score may therefore make it easier to distinguish people at the high-risk end without addressing most cases that arise across the wider population. The authors argue that this tension matters when deciding what role a score should have in a prevention strategy.
It also affects how a score should be assessed. Evidence may be judged against different endpoints, compared with different alternatives, or held to different performance thresholds. A measure that separates risk groups is not automatically judged by the same criteria as a programme intended to improve health across the population. In the authors’ view, some disagreements about the evidence reflect these differences in priorities, not just disagreement about predictive performance.
That framing matters as England’s plans bring genomics into future population-health and cancer-risk policy. Before weighing a score’s performance, decision-makers need to specify the prevention goal and the outcome against which its use would be assessed. The correspondence offers a policy and evidence framework for that discussion; it does not report a clinical evaluation of a particular score or prevention programme.