The 2017 rules for consumer genetic risk tests covered accuracy and labeling, and set no controls for the data
On November 7, 2017, the Food and Drug Administration codified special controls for consumer genetic risk tests and proposed one review per firm. The controls covered accuracy, labeling and user comprehension. In our reading, the data those tests created was left to other law.
Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.
Two-minute summaryThe dek and the first sentence of each section, quoted from the article
On November 7, 2017, the Food and Drug Administration codified special controls for consumer genetic risk tests and proposed one review per firm. The controls covered accuracy, labeling and user comprehension. In our reading, the data those tests created was left to other law.
- The week On Monday, November 6, 2017, the Food and Drug Administration published a statement from its Commissioner on what it called a streamlined pathway for consumer tests that evaluate genetic health risks.
- The evidence: accuracy and clinical meaning, specified to the sample The special controls start from a premise.
- The words: labeling, web pages and a test of understanding The second half of the special controls governs communication.
- The review: one look per firm, then new tests without another The third part is procedure.
- The data: outside the order, inside other regimes The fourth part is what the order did not cover.
- How the split reached different health care organizations The order applies to test makers, but its effects reach organizations that never make a test.
- Objections, in order The argument that FDA’s order left data protection to other regimes faces several objections.
- What a health system could check now The 2017 order requires none of these steps.
- With hindsight FDA finalized the exemption on June 5, 2018.
The week
On Monday, November 6, 2017, the Food and Drug Administration published a statement from its Commissioner on what it called a streamlined pathway for consumer tests that evaluate genetic health risks. The statement said the agency was taking steps that day to implement a novel approach to these tests. It also said the approach built on lessons from the agency’s authorization of the first genetic health risk and carrier screening tests sold directly to consumers.
The statement named the benefit and the risk. Genetic risk tests could prompt people to engage with healthy lifestyle choices. They could also mislead. The statement gave the case of a person told, incorrectly, that they were not at risk of coronary heart disease, who then skips dietary changes or drugs that would lower their risk. It said these technologies did not fit squarely into the agency’s traditional risk-based approach to device regulation.
Two documents appeared in the Federal Register on Tuesday, November 7. The first was a final order classifying the genetic health risk assessment system into class II, with special controls. The order said 23andMe had requested De Novo classification of its Personal Genome Service test on June 28, 2016. The agency had issued its classification order to the company on April 6, 2017. The November order codified that classification as a new regulation, 21 CFR 866.5950.
The order defined the device type. A genetic health risk assessment system is a qualitative in vitro molecular diagnostic system that detects variants in DNA. It gives users information about their genetic risk of developing a disease, to inform lifestyle choices or conversations with a health care professional. It is for over-the-counter use. The definition added that the device does not determine a person’s overall risk of developing a disease.
The order listed the risks the agency saw and the controls that answered each one. The risks were an incorrect understanding of the device, incorrect test results through false positives or false negatives, and incorrect interpretation of results. The controls ran to several pages of labeling, website, performance and study requirements. The order also set four uses the device type could not claim: prenatal testing, certain cancer predisposition testing, testing that informs medication decisions, and testing for deterministic autosomal dominant variants.
The second document was a notice of the agency’s intent to exempt several class II product types from premarket notification, known as 510(k). The genetic health risk assessment system was on the list, with a limit. The exemption would apply only to a system that had already received a first-time marketing authorization. The notice called this a one-time reviewed genetic health risk assessment system. Comments were due by January 8, 2018.
The notice explained the legal route. The 21st Century Cures Act, signed on December 13, 2016, had amended section 510(m) of the Food, Drug, and Cosmetic Act. As amended, the agency could exempt a class II product type from 510(k) on its own initiative. It had to publish a notice of intent, take comments for 60 days, and then publish a final order within 120 days of the notice.
The notice gave the agency’s reason for keeping the one-time review. It said such a review was needed to mitigate the risk of false negatives and false positives. The review let the agency assess safety and effectiveness and confirm that devices perform to acceptable standards. After it, the Commissioner’s statement said, manufacturers could enter the market with new genetic health risk tests without further review, once the policy was final.
The statement listed two related actions taken at the same time. The agency issued a final order exempting genetic carrier screening tests from premarket review. It also classified certain vitamin D tests into class II and announced its intent to exempt them. The statement called the genetic test approach similar to the proposed firm-based precertification model the agency had developed for digital health technologies. It said the agency would keep looking for chances to use this firm-based approach, for new tests and for other novel medical products.
None of the three documents set requirements for the privacy or security of the genetic data the tests would produce. The special controls dealt with what a test must measure, how well, and what users must be told. They did not deal with how a firm must store, share or protect the results. That gap is the subject of this article.
FDA’s November 2017 actions on genetic health risk tests read, at first, as a deregulatory story. A class of consumer tests would get one review per firm and then a lighter path for each new test. Read closely, the order is a demanding piece of rule-writing. It specifies sample counts for accuracy studies, confidence intervals, comprehension thresholds and the content of web pages.
Our argument is about what that rule-writing covered and what it left out. The special controls governed the science of the test and the words around it. They said nothing about the genetic data the test created, which stays with the firm, sits in customer accounts, and links people to their relatives. In our reading, FDA did what its device authority equips it to do, and the data question went to other regimes by default.
For health systems, that split matters in two directions. Patients arrive with consumer results whose FDA-required labeling says they are not intended to diagnose disease. And health systems, researchers and plans may be asked to partner with firms whose data obligations come from consumer-protection law, contracts and state statutes, not from the device authorization. We take the November actions in four parts:
- The evidence: what the special controls required a firm to prove about accuracy and clinical meaning.
- The words: what the order required firms to tell users and clinicians, and how it tested that they understood.
- The review: what the one-time review per firm kept, and what dropping later reviews gave up.
- The data: what the order did not address, and which other rules were left to govern it.
The four parts do not overlap. The first two are the content of the special controls. The third is the procedure for applying them. The fourth is the space outside them. After the four parts, we look at how the split reached different kinds of health care organization, at objections to the argument, and at what a health system could check now.
A note on sources. We describe the order, the notice and the statement from FDA’s own documents. We describe later events from FDA’s 2018 final order, securities filings and an FTC release. Where we interpret a rule or say who it covers, that is our reading, and organizations should check with counsel before relying on it.
The evidence: accuracy and clinical meaning, specified to the sample
The special controls start from a premise. A consumer genetic test fails in two ways that matter. It can call a variant wrong, or it can attach the wrong meaning to a variant it called right. The order treats these as separate problems with separate evidence.
Analytical accuracy
For the first problem, the order sets out an accuracy study in detail. Accuracy must be compared against bidirectional Sanger sequencing or another method FDA finds appropriate. Fresh clinical specimens come first, with archived or contrived samples allowed only in stated circumstances. Results from clinical specimens and from contrived samples must be reported separately.
The order sets minimum sample counts tied to how common a variant is. At least 20 unique wild-type samples must be tested. Common heterozygous variants need at least 20 unique samples, with smaller minimums for rarer variants. The point estimate of agreement must reach 99 percent per reported variant and overall. A variant that falls below that line may not be included in test claims and reports.
That last rule is a design constraint, not a disclosure. It means a firm cannot report a variant it measures poorly and then explain the weakness in small print. The variant drops out of the product. For a clinician reading a consumer report, this gives some assurance that a variant listed was called with high agreement against a reference method in the firm’s own study.
Precision, interference and failure
The controls go further than accuracy. Precision and reproducibility data must cover multiple instruments, operators, non-consecutive days and reagent lots. If more than one laboratory runs the test, the study must include them. The firm must study interfering substances and interfering mutations. Where it has not studied a suspected interfering variant, the labeling must say so in prescribed words.
The firm must also state the minimum amount of DNA that lets the test perform correctly in 95 percent of runs. It must report the share of tests that fail quality control and how users are told about a failed test. These are the parts of the order a laboratory director would recognize. They make the device type closer to a validated assay than to a wellness product.
Clinical meaning
For the second problem, the order requires a clinical performance summary for each variant combination. Where odds ratios exist, the firm must calculate likelihood ratios with 95 percent confidence intervals. It must then show post-test risk from pre-test risk. The order sets a minimum for showing that a variant affects risk at all: its likelihood ratio must be statistically higher than 1, meaning the lower bound of the 95 percent confidence interval is above 1.
The order also requires a technical positive predictive value for each relevant subpopulation, by ethnicity, sex, age or geography. This matters because, as the order explains, the technical predictive value depends on how common the genotype is in the subpopulation studied. The required question-and-answer pages must also say which population the tested variants are most applicable to. The order makes population context part of the required record.
What the evidence regime did not reach
The evidence regime is strong where it applies, and it applies to the test as a measurement. It does not reach the systems that hold results after the test is run. The order contains no requirement about access controls on customer accounts, encryption of stored genotypes, logging of who views results, or retention of samples and data. A search of the order’s text finds no use of the words privacy or security.
We do not read that silence as an oversight. The order addresses the risks to health that FDA identified, and it lists three: misunderstanding the device, wrong results and wrong interpretation. A breach of a customer database is not a risk to health in that framework, as FDA defined the device’s risks here. The evidence regime was built to answer whether the test is right, and that question ends when the report is delivered.
Why this matters to a laboratory leader
For a health system laboratory, the order offers a useful benchmark. Its accuracy and reproducibility requirements resemble what a clinical laboratory would expect when validating an assay. A laboratory director asked whether a consumer result is trustworthy can point to the specific evidence the firm had to produce. It gives a better starting point than the marketing material a patient brings.
The benchmark has limits. The evidence belongs to the firm and to FDA, and much of it would sit in the firm’s submission. The order does require a public technical information section on the firm’s website, including performance data. A laboratory can read that section. It cannot audit the underlying study without the firm’s cooperation.
Reading a consumer report as a clinician
The evidence regime gives clinicians a short checklist when a patient brings a report. Is the variant on the firm’s list of reported variants, so it met the accuracy line? What population data supports the risk category for this patient? Does the firm’s technical page list an interfering variant that might apply? These questions take minutes with the firm’s public pages open.
The checklist does not replace confirmation. The order’s own clinician section says decisions should rest on testing the clinician considers appropriate. What the checklist does is help a clinician decide how urgently to confirm, and how to explain the result to the patient while confirmation is pending.
The words: labeling, web pages and a test of understanding
The second half of the special controls governs communication. For a test sold to consumers without a clinician, the words around the result carry much of the safety burden. The order treats them that way and specifies their content at length.
Limiting statements for users
The labeling, any prepurchase page and any test report must include a set of limiting statements for users. The test does not report a user’s entire genetic profile. Other companies may test different variants for the same disease. Environmental and lifestyle factors also affect risk. Some people may feel anxious about results, and a user who feels very anxious should speak to a health care professional before giving a sample. The test is not a substitute for visits to a doctor.
Two statements matter more than the rest for clinicians. One says the test is not intended to diagnose a disease, describe current health, or inform medical decisions such as whether to take a medication or how much. Another requires information on how to reach a genetic counselor, a board-certified clinical molecular geneticist or an equivalent professional.
A section addressed to clinicians
The order also requires a separate section for health care professionals who receive results from their patients. It must say the test is not intended to diagnose, determine treatment, or describe current health. It must say the test is meant to inform lifestyle decisions and conversations with a doctor. And it must say that diagnostic or treatment decisions should rest on testing or other information the clinician considers appropriate.
Read from the health system side, this section is a handoff. The firm states that its result is not the basis for care. The clinician decides what testing would be. The order gives the clinician a clear statement to point to. It does not give the clinician or the health system any duty to act on the result, and it does not say who pays for confirmation.
Web pages, opt-ins and population context
The order requires a public web page with summary and technical information. It must explain risk categories consistently across reports, so that increased risk means the same thing in every report. It must give context for population differences, including ethnicity, age, sex, environment and lifestyle. It must include definitions, a prepurchase page and a frequently asked questions page for each variant and disease pair.
For a life-threatening or irreversibly debilitating condition with few or no options to prevent, treat or cure it, the order requires an opt-in for each such disease. The user must be able to accept or decline that specific result. The page must explain the risk involved, cite professional guidelines on testing, and recommend speaking to a professional before getting the result. This is a consent mechanism, but it concerns receiving information, not sharing it.
A comprehension study with a threshold
The control with the least precedent in device labeling rules is the user comprehension study, in our reading. The firm must give naive participants an education module, then test their understanding before and after. Participants must be demographically diverse, untrained and new to the test. The study must meet a predefined primary endpoint of at least 90 percent overall comprehension for each concept. The order says meeting that rate shows the materials are adequate for over-the-counter use.
The comprehension concepts named in the order are the test’s limitations, purpose, appropriate action, results, and other factors that affect results. Data handling is not among them. Nothing in the order requires a firm to show that users understood who could see their genotype, whether it would be used for research, or what would happen to it if the company were sold.
Consent to receive, not consent to share
The distinction is worth drawing out for compliance teams. The order builds a careful consent process around receiving bad news. It builds none around the reuse of the underlying data. A user who passes through every required opt-in and comprehension screen has been told a great deal about genetics and nothing, by force of this order, about data. Whatever data disclosures a firm made came from its own privacy policy and other law.
The review: one look per firm, then new tests without another
The third part is procedure. FDA’s notice proposed that, once a firm’s genetic health risk system had been authorized once, later tests in the same system could reach the market without a 510(k). The statement described this as similar to the proposed firm-based precertification model FDA had developed for digital health technologies.
What the one-time review kept
The one-time review kept FDA’s look at the firm’s methods and its study design. The notice tied the review to the risk of false negatives and false positives. It said a one-time review was necessary to assess safety and effectiveness and to confirm acceptable performance. The special controls stayed binding on every later test, whether or not FDA saw it before launch.
That structure shifts the evidence burden in time. Before launch, the firm must still generate the accuracy, clinical and comprehension evidence the controls require, and post it. FDA no longer checks it in advance for each new test. Enforcement, if any, would come after a problem surfaced.
What it gave up
Dropping per-test review removed a point at which FDA could see a firm’s growing catalog of reports. Under the proposal, a firm could add disease reports over time, each meeting the controls on paper. FDA would not, by default, review the clinical evidence for each new variant and disease pair before customers saw it.
This trade was explicit and argued in the record, as we describe in the hindsight section. Our interest here is narrower. A firm-based model rests on confidence in the firm’s processes. For accuracy and labeling, the special controls spelled those processes out. For data handling, the firm-based model had no process to rest on, because the controls never named one.
Firm-based regulation and data
A firm-based model can be a strong fit for data risk, in principle. Data risk lives in a firm’s systems, staff and contracts, not in a single product version. A regulator that certifies a firm could ask about its security program. The 2017 approach certified the firm’s science and communication. It did not extend the idea to the firm’s data practices.
That choice reflects FDA’s mandate as the order framed it. The agency’s device authority concerns safety and effectiveness. A consumer genetics database is not a device controlling therapy. Its compromise harms privacy and family relationships rather than bodily safety, and the order did not treat it as a device risk.
Change inside a firm after the first review
Under the exemption, the firm’s own processes decide when a new report is ready. The special controls say what evidence each report needs and require changes that could significantly affect safety or effectiveness to be supported by new data, posted on the firm’s website. That posting rule makes the firm’s website the record of change. A clinician or partner who wants to know what changed must read it.
The posting rule also creates a practical evidence trail for buyers. A research partner can ask the firm for the dated history of its technical information pages. It can compare that history with the reports customers were shown. Neither step is required of the partner. Both are cheap ways to test whether a firm’s change process matches what the controls demand.
How the procedure looked to buyers
For a health system or research institution considering a partnership, the procedure has a practical effect. FDA authorization of a firm’s system is a statement about the first review and the controls. It is not a statement that each later report was reviewed. A partner relying on the authorization for any purpose should ask which reports were in the reviewed submission and which came later.
The data: outside the order, inside other regimes
The fourth part is what the order did not cover. A consumer genetic test produces a genotype file, and a firm may build ancestry and family-matching features on it. The data persists after the report, lives in an online account, and reveals information about blood relatives who never bought the test. The special controls addressed none of that.
Which rules reached the data
In our reading, HIPAA’s privacy and security rules apply to covered entities and their business associates. A firm selling a test directly to consumers, with no health plan or provider in the transaction, is generally neither. That leaves federal consumer-protection law, enforced by the FTC against unfair or deceptive practices, plus state privacy, genetic privacy and breach notification laws. Organizations should ask counsel how these apply to any specific firm or deal.
The consumer-protection route works mainly through promises. A firm that says it encrypts data, deletes samples or limits sharing can be held to what it said. A firm that says little can be harder to reach under a deception theory, though the FTC also uses unfairness. The protection a customer gets depends on the firm’s own privacy policy and on which state the customer lives in.
Data about people who never consented
Genetic data has a feature that ordinary personal data lacks. A genotype carries information about parents, siblings and cousins. Family-matching features make that link explicit by showing users which other customers share DNA with them. A user can opt into such a feature, but the relatives revealed by matches did not consent on behalf of each other in any meaningful sense.
The device order did not address this either. Its required question-and-answer pages must explain how test results may affect a user’s family, including children. That requirement is about health implications, not about data exposure. The family dimension of the data was left entirely to the firm’s design choices and to the other regimes.
Notice when the data is exposed
When consumer genetic data is exposed, the duty to tell customers comes from wherever the firm’s other obligations come from. In our reading, that means state breach notification laws, whose definitions of covered data vary, and any promises the firm made. HIPAA’s breach notification rule would apply only where the firm is a covered entity or business associate. A customer’s right to know depends on facts the device order never touches.
For a health system partner, this is a contract point. If a partner firm’s data about the health system’s patients or research participants is exposed, the system wants to know quickly and in detail. That duty has to be written into the agreement. Without it, the system may learn of an incident from the news or from its own patients.
Data as an asset
A consumer genetics firm’s database is also a business asset. It can support research partnerships, drug discovery and future products. It can be pledged, licensed or sold. The device order does not govern any of this, and nothing in it limits what happens to the data if the firm changes hands. That question falls to the firm’s terms of service, consumer-protection law, state law, and in an insolvency, the bankruptcy court.
Why the split was predictable
We think the split was predictable from the statute. FDA regulates devices for safety and effectiveness. The order applied that frame carefully and well. Data protection for consumer health information outside HIPAA had no single federal owner in 2017, as our reading of the regimes above suggests. The November actions simply made the gap more consequential, by making it easier for a firm to grow its catalog of tests and its customer base.
A health system leader reading the order in 2017 could have drawn two conclusions. The first is that authorized consumer genetic results meet a defined standard of accuracy, worth knowing when patients bring them in. The second is that the data behind those results carries no device-level protection at all. Both conclusions held up.
Diligence questions the order leaves to partners
Because the order is silent on data, a partner has to ask the data questions itself. A short list follows from the gaps we have described. Each item maps to a risk the special controls do not address:
- Account security: whether the firm requires multifactor authentication and detects logins that use credentials exposed elsewhere.
- Relatives features: what one account can see about other customers, and whether a single compromised account can reach data about other customers in bulk.
- Storage: whether genotype data is encrypted at rest, and who inside the firm can read it.
- Samples: how long physical samples are kept, and by which laboratories.
- Sharing: which third parties receive data, and on what consent.
- Change of control: what the firm’s terms say about data in a merger, sale or bankruptcy.
These checks are ordinary for a security or privacy review. What makes them easy to skip here is the FDA authorization, which can make a partner feel the firm has already been vetted. The order vetted the firm’s science. The items above vet everything else.
Who inside a health system owns the question
Inside a health system, the two halves of the question land with different owners. Laboratory medicine and genetics own the clinical meaning of a consumer result. Privacy, security and research compliance own the data relationship with a consumer firm. A system that routes consumer genetics partnerships only through the clinical side may approve one half and never examine the other.
One remedy is to require both reviews for any agreement that involves a consumer genetics firm, whatever the stated purpose. A recruitment partnership, a return-of-results program and a data license all involve the firm’s data practices. Each deserves a security and privacy review alongside the scientific one. The order gives no reason to merge the two, and good reason to keep them separate.
How the split reached different health care organizations
The order applies to test makers, but its effects reach organizations that never make a test. We take five kinds of organization in turn.
Health systems and academic medical centers
Health systems meet consumer results at the front door. A patient brings a report showing an increased risk variant and asks what to do. The order’s clinician section says diagnostic and treatment decisions should rest on testing or information the clinician considers appropriate. In practice, that points toward confirmatory clinical testing and genetic counseling, each with cost and capacity implications.
A health system that has not set a pathway leaves each clinician to decide. Some will order confirmation. Some will document the result as patient-reported history. Some may act on it directly, which the labeling warns against. A written pathway would settle where the result goes in the record, whether it is flagged as unconfirmed, and when to refer to genetics. The order itself requires none of this, so it is a practice choice.
Ambulatory networks and primary care groups
Primary care practices meet the same patients, sometimes with no genetics service of their own to call. The order requires firms to tell users how to reach a counselor. It does not create counselors. An ambulatory network may need referral agreements or telehealth arrangements to answer the questions the labeling encourages patients to bring.
Health plans
Plans see consumer results when members seek coverage for confirmatory testing or follow-up care. A plan’s medical policy decides whether a consumer result is enough to justify confirmation. That is a coverage decision for the plan, made under its own rules. Plans should note that the order frames these tests as informing lifestyle choices and conversations with a professional, which bears on how a plan might weigh them, though FDA’s framing does not decide coverage.
Researchers and research institutions
Research institutions meet consumer genetics firms as data partners. A partnership may involve receiving genotype data, recruiting through a firm’s customer base or sharing results back. Here the data question dominates. The research institution has its own obligations, which may include HIPAA, the Common Rule or state law, depending on the study. The firm’s obligations come from its own policies and from consumer-protection and state law.
A research data-use agreement is where those two sets of obligations meet. It can set security requirements, limit reuse, require breach notice and govern what happens on a change of control. None of those terms comes from FDA’s order. An institution that treats FDA authorization as due diligence on the partner’s data practices would be relying on something the order does not address.
Life sciences firms
Life sciences firms meet consumer genetics data as a research and development input. For them, the order is reassurance about the genotype calls for the variants it covers. It is no assurance about consent for research use or about the security of the source data. Those come from the firm’s consent documents and contracts, which a buyer of data access would need to read.
Objections, in order
The argument that FDA’s order left data protection to other regimes faces several objections. We take the strongest six in order.
1. FDA was never the right regulator for data, so the gap is no criticism
We agree that FDA’s mandate did not obviously extend to consumer data protection, and our argument does not criticize FDA. The point is descriptive. Organizations that relied on FDA authorization as a signal of overall trustworthiness were relying on something it did not cover. The gap matters to buyers and partners whatever its cause.
2. The FTC filled the gap
The FTC does act against genetic testing firms, as the hindsight section shows. But its tools work mainly after the fact and through the firm’s own promises and unfairness theory. A firm-level FTC order follows a failure. That is a different protection from a premarket requirement, and a partner still needs its own diligence.
3. Firms had strong commercial reasons to protect the data
Commercial incentive is real. A firm whose value rests on customer trust has reason to guard its data. But incentive is not a control. The later incident we describe began with reused passwords and spread through a relatives feature, according to the firm’s own filing. Incentive did not prevent that design exposure.
4. The accuracy rules were the main safety risk and they were handled
For bodily harm, accuracy was the main risk, and the order handled it with care. We do not dispute this. Our argument is that a second category of harm, to privacy and to relatives, sat outside the order’s definition of risk. Both categories can matter to a patient. Only one was regulated by this order.
5. Health systems are not responsible for consumer firms’ data
That is mostly true, and it is why we focus on the points where health systems do touch these firms. Health systems ingest results into records, refer patients, partner on research and sometimes recruit through consumer platforms. At each point, the health system makes choices. Knowing where FDA’s assurance stops helps make those choices well.
6. Consumer genetics is a niche that health systems can ignore
Consumer results reach clinics whether or not a health system plans for them, because patients carry them in. Data partnerships reach research offices through recruitment and data licensing. The order’s definitions apply each time. Ignoring the category leaves those encounters to individual judgment, which is where inconsistency in records and referrals begins.
What a health system could check now
The order’s structure suggests a short set of checks for a health system that meets consumer genetic results or firms:
- Decide where a patient-reported consumer genetic result goes in the record, and flag it as unconfirmed.
- Give clinicians a written pathway for confirmatory testing and genetic counseling referral.
- For any data partnership with a consumer genetics firm, review its security program, account protections and change-of-control terms, separately from its FDA status.
- Require the firm, by contract, to notify the system of incidents affecting data tied to its patients or research participants.
- Require multifactor authentication on the system’s own portals and research sites, and watch for logins with reused credentials.
The 2017 order requires none of these steps. Each follows from reading it carefully and noticing where its protections stop. In our view, the order’s main gift to health systems was a precise definition of what a consumer genetic result is and is not. A health system that adopts that definition in its own processes, and adds the data diligence the order omits, is in a stronger position than one that treats FDA authorization as the whole answer.
There is one more practical step. When a firm’s FDA status is cited in a contract, vendor review or board paper, the document can say what that status covers. A sentence noting that the authorization concerns accuracy and labeling, not data protection, prevents a reader from assuming more. It costs little and closes off one easy misreading.
With hindsight
FDA finalized the exemption on June 5, 2018. The final order said FDA had received a submission from one commenter, a professional organization, opposing the exemption for genetic health risk tests. The commenter argued that a one-time review could not assess the validity of later tests, and that consumers would assume FDA had reviewed tests it had not. FDA responded that the comprehension study, labeling and website requirements addressed consumer understanding. On accuracy, it said that once FDA had reviewed one test meeting the 99 percent requirement, the provider had shown it could meet that requirement for similar tests. FDA finalized the exemption with the one-time review limit.
The accuracy and labeling regime held up as written. The special controls remained the standard for the device type, and the one-time review per firm became final. The order’s structure, a precise definition of what a consumer genetic result is, gave health systems a stable reference point for how to treat such results.
The data question developed outside FDA, as the order’s silence suggested it would. On September 7, 2023, the FTC finalized an order with 1Health.io, a genetic testing company formerly known as Vitagene. According to the FTC’s complaint, the company stored unencrypted health, genetic and other personal information in publicly accessible data buckets, despite promising strong security. The FTC also charged that the company failed to keep promises to destroy DNA samples and to delete data on request. These were allegations, resolved by a settlement. The order required a payment of $75,000 and a comprehensive information security program.
The FTC order also prohibited the company from sharing health data with third parties without affirmative express consent. It required the company to notify the FTC of unauthorized disclosures of personal health data and to have laboratories destroy samples retained more than 180 days. These are the kinds of controls the 2017 device order never contained. They arrived through consumer-protection enforcement, firm by firm, after a failure.
The firm whose De Novo request created the device type disclosed a security incident in October 2023. Its amended filing of December 1, 2023, said a threat actor had accessed about 0.1 percent of user accounts, where users had reused passwords exposed elsewhere. Using those accounts, the actor accessed a significant number of files with profile information about other users who had opted into the DNA Relatives feature. The company said it had required password resets and, from November 6, 2023, two-step verification for all users.
That incident illustrates the family dimension we described. The filing separates the 0.1 percent of accounts accessed directly from the significant number of files about other users. Those other users had chosen to share profile information with genetic matches. The filing describes their files as reached through the credential-stuffed accounts, not through their own logins. No device control addressed that design, because the order’s risks were defined around the test result, not the platform.
The company’s later filings show the data becoming a bankruptcy asset. Its March 24, 2025, filing said it and certain subsidiaries had filed chapter 11 petitions on March 23, 2025. The same filing said it had agreed to pay an aggregate of $37.5 million to settle claims from U.S. customers relating to the 2023 incident, subject to conditions, without admitting liability. Its July 14, 2025, filing said the court had approved a sale of substantially all its assets to TTAM Research Institute, a nonprofit affiliated with the company’s co-founder.
The July filing said the sale closed on July 14, 2025, for $305.0 million in cash plus assumed liabilities. It described the assets as substantially all of the debtors’ assets, other than assets the agreement excluded, and said TTAM had separately agreed to sponsor a plan to acquire the company’s telehealth business. The filing does not single out customer data. Whatever data was among those assets moved under a sale a bankruptcy court approved, and the device order of 2017 had nothing to say about it.
What held up, then, is the split itself. FDA’s rules governed the test well. In our reading, the data was governed by a patchwork of FTC enforcement, state law, securities disclosure and, at the end, a bankruptcy court. For a health system, the durable lesson of November 2017 is to treat a consumer genetics firm’s FDA status and its data practices as two separate questions, each needing its own evidence.
What did not hold up is any assumption that the firm-based approach would extend naturally to data. FDA’s idea of certifying the firm rather than each product applied to the science. No parallel certification of data practices followed from the 2017 actions. Organizations that partner with these firms in 2026 still need their own diligence on security, consent and change of control.
Spotted an error? Tell us and we will check it and correct the article where needed.