Regulatory pathways
Device classes, 510(k) versus De Novo versus PMA, predicates, breakthrough designation, and the guidance written specifically for implanted BCIs. Twenty hours of reading that make you sound like a third-year employee.
You are skimming: the title, the first figure, and the short version. Switch to Read in the header for the full page, or Deep to open every deep dive.
A device that touches a patient in the United States needs the FDA’s permission to be sold, and the kind of permission depends on how risky the device is and whether something like it already exists. Three classesFDA device classesClass I is low risk (a tongue depressor), Class II moderate (most EEG amplifiers), Class III highest (implanted BCIs), each with heavier review. Glossary entry, three main pathways, one special program, and a set of public databases that let you read exactly how every device on the market got there. Learn the vocabulary here and read three real submissions; after that, regulatory conversations stop being intimidating.
Three classes
Class I: low risk. Tongue depressors, manual wheelchairs, most electrodes. General controls (registration, labelling, good manufacturing) and usually no premarket submission.
Class II: moderate risk. Most EEG amplifiers, powered wheelchairs, infusion pumps, sleep monitors, many BCI software products. Requires a premarket submission, almost always a 510(k), plus “special controls” specific to the device type.
Class III: highest risk, life-sustaining or implanted with significant risk. Pacemakers, deep brain stimulators, cochlear implants, implanted BCIs. Requires premarket approval with clinical evidence.
The class is set by the device’s intended use and indications. The same EEG amplifier is Class II as a diagnostic aid and unregulated as a “wellness” product that makes no medical claim, which is the line consumer neurotech companies walk. Change the claim and you change the class.
Three pathways
510(k)510(k)The FDA pathway that clears a device by showing it is substantially equivalent to one already on the market, its predicate. Glossary entry. Show that your device is “substantially equivalent” to a legally marketed predicatePredicate deviceThe already-legal device a 510(k) submission compares itself to. Glossary entry: same intended use, and either the same technology or technology that raises no new questions of safety and effectiveness. Bench testing and sometimes small clinical data. Most Class II devices, and around three thousand clearances a year. Ninety days nominal, six to twelve months typical. “Cleared,” not “approved.”
De NovoDe NovoThe FDA pathway for a low-to-moderate-risk device that has no predicate. Glossary entry. For a low-to-moderate-risk device with no predicate. The FDA creates a new classification and special controls, and your device becomes the predicate for everyone after you. More work than a 510(k), far less than a PMA. Several novel neurotech devices took this path.
Premarket approvalPremarket approval (PMA)The strictest FDA pathway, requiring clinical evidence of safety and effectiveness; the route for Class III implants. Glossary entry. For Class III. Requires clinical trials demonstrating safety and effectiveness, a manufacturing inspection, and often an advisory panel. Years and tens of millions of dollars. “Approved.” Every implanted stimulator went this way.
A company builds a new EEG headband that estimates sleep stages and markets it as helping you 'understand your sleep.' A second company sells the identical headband to 'detect sleep apnea.' What is the regulatory difference?
The claim decides. “Understand your sleep” is a general wellness claim the FDA has said it will not regulate as a device. “Detect sleep apnea” is a diagnostic claim for a disease, making the same hardware a Class II device that needs a 510(k), clinical validation of its detection performance, and special controls. Regulatory strategy is largely the art of choosing claims, and most consumer neurotech lives on the wellness side of this line on purpose.
Getting to a trial
A device that is not yet cleared can be studied in people under an investigational device exemptionInvestigational device exemption (IDE)FDA permission to use an unapproved device in a clinical study. Glossary entry, with IRB approval and, for significant-risk devices, FDA approval of the study. Every implanted BCI participant is in an IDE study. Early feasibility studies with a handful of participants are the first step, and they are what the BCI companies are running now.
Breakthrough designation
The Breakthrough Devices ProgramBreakthrough Device designationAn FDA program that speeds review for devices treating serious conditions with no good alternative; most invasive BCI companies have it. Glossary entry gives priority review and more frequent FDA interaction to devices for life-threatening or irreversibly debilitating conditions with no adequate alternative. Most implanted BCI companies have it. It speeds communication; it does not lower the evidence bar.
The BCI guidance
In 2021 the FDA published guidance specifically for implanted BCI devices for patients with paralysis or amputation: what non-clinical testing to do (biocompatibility, electrical safety, software, cybersecurity, and specific tests for the implant and its wireless link), how to design early feasibility studies, and what outcome measures make sense. It is forty pages, it is written for exactly the devices this site is about, and reading it once puts you ahead of most people in the field.
Read three submissions
The FDA’s 510(k) database has a summary for every clearance: device, predicate, tests performed, sometimes clinical data. Search “EEG” and read three. Then find one De Novo decision summary for a neurotech device and one PMA summary for a neurostimulator. The pattern of what evidence was required, and how the intended use was worded, is the real education. Twenty hours total across this page and those documents.
Deep dive Software and adaptive algorithms 3 min
Software that is itself a medical device (a seizure detector, a sleep stager) follows the same classes and pathways, with its own guidance on documentation. A machine-learning model that is locked at clearance is treated like any software; one that keeps learning after deployment has no clean path yet, and the FDA’s “predetermined change control plan” is the current attempt to allow pre-approved updates. If your decoder adapts to the user, this is your issue.
Deep dive Outside the United States 2 min
The European Union’s Medical Device Regulation uses a similar class structure with conformity assessment by notified bodies and a CE mark; since 2021 it is substantially stricter than before and slower than the FDA for many devices. Japan, China, and others have their own systems. A device company’s regulatory team plans the sequence of markets; the United States is usually first for neurotech because the pathways are the most predictable.
Explain what this page was about to your roommate in three sentences. No jargon they would not know.