On 20 March 2019, standing at the Royal Society, the man in charge of the National Health Service told the country a genetic test may have saved his life. Matt Hancock, then health secretary, had sent off his DNA and been handed back a number: his risk of prostate cancer by age 75 was almost 15 percent. In a promotional video he described the worry, the doctor’s appointment, the sense of a bullet dodged. “The truth is,” he said, “genomics might have saved my life.”
It had not, and the number did not mean what he thought it meant. Fifteen percent put Hancock near the middle of the pack, his risk perhaps 50 percent above an average that is, for every man, already uncomfortably high. At 40, under NHS guidelines, he had no medical reason to be tested at all. David Curtis of University College London said so at the time: by misreading his score, the health secretary “first suffered unnecessary anxiety and then took up valuable medical time.” The assembled experts were, for scientists, close to scathing.
The company that sold Hancock his reading was Genomics plc, a spinout of Oxford University. Seven years later the same company’s scores are being wired into the NHS itself, scores that, at their own published performance, catch about 11 percent of the people who go on to develop the disease they screen for. The other nine in ten are never flagged.
| Measure | Share |
|---|---|
| of future cases | 11 percent |
The government would rather you read that as prevention. On 3 July 2025 the Department of Health and Social Care published “Fit for the Future,” its ten-year plan for the NHS, and named genomics one of a handful of “big bets” on the future of medicine. The plan commits to a new Genomic Population Health Service, open to all by the end of the decade, pairing whole-genome sequencing of newborns with population-wide polygenic risk scoring. Wes Streeting, the health secretary, cast it as a shift from treating the sick to spotting risk before symptoms show. The pitch is prevention. The instrument is a score that, on the evidence, mostly cannot find the people it is meant to prevent disease in.
This week Nature Medicine ran a commentary under a title that says the quiet part aloud: the debate over putting these scores into the NHS, it reads, “is not primarily about the evidence.” Worth sitting with, because the evidence, when you go and read it, is not ambiguous.
A polygenic risk score adds up the tiny contributions of a million points across your genome, each nudging your estimated risk by a sliver, and returns a percentile: top 5 percent, or the middle, or the bottom. A percentile is not a diagnosis. At the population level the arithmetic holds. For the individual sitting in front of a GP, it mostly does not. In a 2023 secondary analysis in BMJ Medicine, Aroon Hingorani and colleagues at UCL worked through 926 published scores covering 310 diseases and found that a typical score, set to flag the highest-risk group while wrongly flagging 5 percent of healthy people, detected only about 11 percent of those who would go on to develop the disease. For breast cancer it was 10 percent; for coronary artery disease, 12. The other nine in ten were not flagged as high risk at all.
The follow-up, published in the Journal of Medical Screening in 2025, did not soften. Across 28 common diseases the median detection at that same 5 percent false-positive threshold was again 11 percent. One much-praised coronary score improved detection from 11 percent to 14, a gain the authors called “trivial.” Their conclusion is not hedged in the way scientists usually reserve for a rival’s life’s work: “An assessment of the evidence leads to the inescapable conclusion that polygenic risk scores are not of value in the prevention of disease.” Rolling them out, they warned, would divert money and staff “from population-wide approaches that address the major disease burden in the average-risk majority to the follow-up of the many false positive results.”
| Group | Value (percent) |
|---|---|
| Before coronary score | 11 |
| With coronary score | 14 |
Set against that is a very different number, the kind that appears in press releases. Genomics plc projects that combining its scores with standard heart-risk testing could prevent more than 45,000 cardiac events over a decade and save the NHS over £146 million. It has been building toward this for years: a 2021 pilot of roughly a thousand patients in the North East, and a partnership to generate scores for Our Future Health, the biobank enrolling up to five million British volunteers. When the Hingorani work landed, it was the company’s co-founder, Oxford’s Gilean McVean, who stepped forward to argue the scores were never meant to diagnose, only to “support the personalisation and effectiveness of screening programs.” The men defending the tool and the men selling it are, in several cases, the same men.
There is a further problem the prevention pitch skates past. Polygenic scores are built overwhelmingly from studies of people of European ancestry, and their accuracy decays the further a patient’s genome sits from that training data. This is not an obscure caveat; it sits in the field’s own review literature. A national screening service built on that foundation would work least well for the very citizens the NHS already fails most. Building the service first and promising to address the resulting inequalities later is a familiar order of operations.
The independent evidence was public in 2023. The plan came in 2025 and did not slow for it. The Nature Medicine commentary arrived in 2026, still circling a debate its own title concedes is not primarily about the evidence.
Genomics plc’s founders will tell you, correctly, that their scores were never built to diagnose anyone, only to sharpen screening at the edges. It is a reasonable thing to say. It is also close to what the company was saying in 2019, when a health secretary mistook one of its numbers for his fate, booked an appointment he did not need, and told the country his life had been saved. One man’s misread cost him some anxiety and a doctor’s afternoon. The service now on the table would hand a version of that number to millions, and call it prevention.
Sources
- Journal of Medical Screening (2025): assessment of the value of polygenic risk scores in the prevention of disease
- BMJ Medicine (2023): performance of polygenic risk scores in screening, prediction, and risk stratification
- Nature Medicine (25 September 2026): Polygenic scores in the NHS: the debate is not primarily about the evidence
- Department of Health and Social Care: Fit for the Future, the 10 Year Health Plan for England (3 July 2025)
- Genomics plc: welcoming the Genomic Population Health Service in the NHS 10-Year Plan (£146m and 45,000-event projection)
- MIT Technology Review (2019): Britain’s health minister accidentally revealed everything wrong with polygenic risk scores
- Progress Educational Trust: health secretary “massively misinterprets” genetic test results
- Science Media Centre: expert reaction to Matt Hancock’s speech on genetic testing
- Human Genomics (2024): polygenic risk score portability across genetically diverse populations
- Science Media Centre: expert reaction to the Hingorani screening-performance study
- Progress Educational Trust: NHS launches polygenic scores pilot to calculate heart disease risk (2021)
- Our Future Health: Genomics plc to generate polygenic risk scores