DTC Genetic Testing

Consumer genetic tests genotype hundreds of thousands of preselected variant positions — a sampling, not a reading, of the genome, which is why they excel at ancestry and relative-matching (built on exactly such markers) while health inference inherits every limitation of the sampling. The flagship example is BRCA: the FDA-authorised consumer report covers three founder variants common in Ashkenazi Jewish populations out of thousands of known pathogenic BRCA variants — so a negative tells most customers almost nothing about their BRCA status, a false reassurance engineered into the product's design. Pharmacogenomic and carrier reports share the pattern: real information about the variants tested, silence about the rest, and a customer who cannot be expected to know the difference.

The industry's defining event was corporate, not scientific: 23andMe — 15 million customers, a 2023 credential-stuffing breach exposing data on nearly 7 million of them, a collapsed share price — filed for bankruptcy in March 2025, and its genetic database became an asset for sale. A bankruptcy court ruled the data could transfer without fresh customer consent, over the objections of more than 30 state attorneys general, because the privacy policy had always disclosed that possibility; after a contested auction, founder Anne Wojcicki's nonprofit TTAM Research Institute bought the assets for 305 million dollars in July 2025. The episode taught the durable lesson: consumer genomic data is corporate property governed by a click-through policy, and its protections are one merger, breach or bankruptcy away from renegotiation.

Consumer genetics lives outside medical privacy law: HIPAA covers patients, not customers, so the file sits under consumer terms-of-service instead. The US Genetic Information Nondiscrimination Act (2008) bars health insurers and employers from using genetic data — and stops there: life, disability and long-term-care insurers may lawfully ask about and price on genetic test results in most US states, a gap most customers discover only when applying. Add law enforcement: investigative genetic genealogy — triangulating suspects through relatives' uploads to open databases, famous since the 2018 Golden State Killer identification — means one cousin's upload effectively enrols a whole family, consent unasked. None of this is hypothetical; all of it is disclosed, in the sense that nobody reads.

The scientific frontier of consumer genomics is the polygenic risk score — summing thousands of small-effect variants into a single number for heart disease, diabetes or cancers. At population scale these scores genuinely stratify risk; for an individual they are weak predictors, shifting probabilities modestly for most people, and they carry a structural equity flaw: built overwhelmingly from European-ancestry datasets, their accuracy degrades substantially in other populations, systematically shortchanging the customers already underserved. The practical tell: virtually no polygenic score yet changes clinical management in a guideline, because knowing your score rarely alters advice beyond what blood pressure, lipids and family history already dictate — the cheap variables the glossy report quietly leans on.

The riskiest consumer behaviour is exporting raw genotype data into third-party interpretation websites: the raw file itself is research-grade, and a clinical-laboratory study found around 40% of variants flagged in such raw data were false positives on confirmatory sequencing — while interpretation sites layer disease claims of wildly varying quality on top. People have pursued surgery-adjacent decisions on unconfirmed variants. The rule that prevents all of it: no medical decision on a consumer or third-party genetic result until confirmed in a clinical laboratory and interpreted with genetic counselling — the pathway that also, for genuine familial cancer or cardiac syndromes, is available through healthcare at higher quality than any kit.

Bought for ancestry, curiosity or family-finding, consumer tests deliver — with the standing caveats that relative-matching can surface unexpected paternity and donor discoveries, and that uploads expose relatives who never consented. Bought for health answers, they are a screening product with engineered gaps: negatives on partial panels reassure falsely, positives need clinical confirmation, and polygenic scores rarely change what your blood pressure already tells you. Anyone with a real red flag — strong family history of early cancer, sudden cardiac death, known familial syndromes — should route directly to clinical genetics, not retail. And before spitting: read what happens to the sample and data on sale of the company, because 2025 answered that question in court.