Consumer DNA tests and polygenic scores: what your genome can and cannot tell you

A tube of saliva, a few weeks of waiting, and a report on your ancestry, your caffeine metabolism and your "genetic risk" for dozens of diseases. Direct-to-consumer DNA tests have been bought by many millions of people, and polygenic risk scores — single numbers that summarise thousands of small genetic effects — are increasingly marketed as the future of prevention. Genetics can matter enormously for a small number of people. For most, a consumer test adds little to what a good family history and ordinary risk factors already reveal — and it carries risks of its own.

Two very different kinds of genetic risk

It helps to separate two things that consumer reports often blend together.

These three conditions are classed by the US Centers for Disease Control and Prevention as "Tier 1": conditions where finding the variant leads to proven, effective action. In the Healthy Nevada Project, 26,906 people had their genomes sequenced; 1.33 % carried a Tier 1 variant, and about 90 % of those carriers had not been identified by the healthcare system. Of 273 carriers with medical records, only 22 had any documented prior suspicion. Monogenic risk is uncommon — roughly 1 in 75 people in that study — but important, and often missed.

What consumer tests actually check

Most consumer tests use genotyping chips that read a preselected set of known positions in the genome, not the full sequence of a gene. The consequences are clearest for BRCA. When the US FDA authorised 23andMe's BRCA report in March 2018, the test covered three variants — the ones most common in people of Ashkenazi Jewish descent, carried by about 2 % of Ashkenazi Jewish women but by 0 to 0.1 % of the general population. More than 1,000 other known harmful BRCA variants were not tested. As the FDA's own reviewer put it, most BRCA mutations that increase risk are not detected by this test.

Raw data and false positives

Many customers download their raw data and upload it to third-party interpretation websites. This creates the opposite problem: false alarms. A clinical laboratory reviewed 49 patient samples sent for confirmation after consumer raw data had flagged a variant. Forty percent of the variants reported in the raw data were false positives — they were simply not there when the gene was properly tested. Some variants labelled "increased risk" by third-party services were, in fact, common harmless variants classified as benign by clinical labs. Chips are good at common variants and unreliable for rare ones — precisely the ones that matter medically.

Any potentially serious finding from a consumer test therefore needs confirmation in an accredited clinical laboratory, ideally arranged through genetic counselling, before anyone acts on it.

Polygenic scores: modest predictive value

Polygenic risk scores are a genuine scientific advance and useful in research. As screening tools for individuals they are weak. A 2023 analysis in BMJ Medicine examined 926 published scores for 310 diseases. If the top-scoring 5 % of people were flagged as "high risk", the scores identified a median of only 11 % of those who would develop the disease. For coronary heart disease, the figure was 12 %; for breast cancer, 10 %. For coronary disease, being flagged moved the 10-year odds from about 1 in 19 in the population to 1 in 8 — a shift, but one that leaves most flagged people unaffected and misses most future cases. The authors concluded that the scores performed poorly for screening, individual prediction and risk stratification.

In practice, blood pressure, cholesterol (including ApoB and lipoprotein(a)), smoking, diabetes, age and family history already capture much of what a polygenic score would add — and they point directly to treatments.

Privacy: data you cannot change

A password can be reset; a genome cannot. And your DNA also reveals information about your parents, siblings and children, who never consented. In October 2023, 23andMe disclosed a data breach affecting 6.9 million customers worldwide; parts of the data were later offered for sale online. The company filed for bankruptcy in March 2025, and its customer data was sold to TTAM Research Institute, a nonprofit set up by its co-founder, for 305 million US dollars. A multistate settlement over the breach, announced in July 2026, was capped at 18 million dollars because of the limited bankruptcy estate. Customers had little say over where their genomes ended up.

The German framework

In Germany, the Genetic Diagnostics Act (Gendiagnostikgesetz, GenDG) regulates genetic testing for medical purposes. Diagnostic genetic tests may be carried out only by physicians; predictive tests — those that estimate future disease risk in healthy people — only by specialists in human genetics or physicians with a specific qualification. Before and after a predictive test, genetic counselling by a qualified physician is mandatory; it can be waived only in writing, and a reasonable period for reflection must be allowed before testing. The counselling must be non-directive and should recommend counselling for relatives when a preventable or treatable condition is found. Consumer tests ordered online from abroad sit outside this physician-led framework — which is exactly why their results deserve caution.

What this means for you

If you enjoy ancestry results, that is a personal choice — but read the privacy terms, consider whether you want your data kept, and do not treat health reports as medical findings. Do not let a reassuring result override a worrying family history, and do not let an alarming result trigger decisions without clinical confirmation. Polygenic scores should not replace standard risk assessment.

This article is general information and not individual medical advice. Decisions about tests and treatment belong in a conversation with your own physician.