The FDA medical device approval process is classify the device by risk, pick 510(k), De Novo, or PMA, then prove safety and effectiveness. The path is not a preference. Risk and novelty pick it.
If you meant QMS / ISO 13485 → medical device quality management system. If you meant post-market surveillance → medical device post-market surveillance. If you meant device software testing / validation → medical device software testing / medical device software validation. If you meant device V&V → medical device verification and validation. If you meant development / GTM / risk → medical device development process. If you meant HIPAA transfer → HIPAA-compliant data transfer. If you meant generic AI regulation (EU AI Act, not a device) — that is a different job; leave it.
If the device is imaging software, classification and the 510(k)/De Novo choice still sit with the manufacturer. PYCAD is the imaging stack, not the regulatory agent.
Four things the file has to show
Approval is one gate. The file still has to stand on four pillars: a quality system, a class, clinical evidence, and labeling plus post-market surveillance. The how-to for QMS is ISO 13485 / QMS. The how-to for PMS is post-market surveillance. This page is the class → pathway tree.
Three classes
The FDA has catalogued more than 1,700 device types. Start in the product classification database: regulation number, three-letter product code, class. Class is the single decision that sets the rest of the file.
Class I — low risk. Elastic bandage, tongue depressor, hospital bed. General controls (registration, listing, labeling, quality). Most are 510(k)-exempt.
Class II — moderate risk. Infusion pump, powered wheelchair, most imaging software that is not life-sustaining. General controls plus special controls. Usual path: 510(k).
Class III — high risk. Pacemaker, heart valve, anything that sustains life or is a significant implant. Premarket Approval. You prove safety and effectiveness from scratch, not by comparison.
A new electronic thermometer is typically product code FLL, Class II — a 510(k) unless it is genuinely novel. Get the class wrong and every later document is built on the wrong foundation.
Pathways at a glance
| Pathway | Typical class | What you prove | FDA’s review job |
|---|---|---|---|
| 510(k) | Class II | Substantial equivalence to a predicate | As safe and effective as the predicate |
| PMA | Class III | Valid scientific evidence of safety and effectiveness | Benefits versus risks, from scratch |
| De Novo | Class I or II, novel | Low-to-moderate risk; general/special controls are enough | Create a new classification |
| Exempt | Most Class I, some Class II | General controls; no 510(k) | Registration, listing, labeling |
CDRH’s 2022 annual report authorized about 5,700 marketing submissions — 3,229 of them 510(k)s, 23 De Novos, 22 original PMAs plus 2,126 PMA supplements. A JAMA review of 1987–2020 put novel PMAs at a median of 32 a year (range 8–56). Most devices never see a PMA.
510(k): predicate, substantial equivalence, pre-sub
A 510(k) is a clearance, not an approval. You notify FDA that the new device is substantially equivalent to a legally marketed predicate that did not need a PMA. Same intended use. Technological characteristics similar enough that they do not raise new safety or effectiveness questions. If the tech differs, performance data has to close the gap.
Predicate choice is the whole strategy. Same intended use is a hard gate — a shift in patient group or clinical use kills SE. Then materials, design, energy source, software. Then the bench (and, when needed, animal or limited clinical) data that shows the differences do not introduce new risk.
A handheld ultrasound with a new transducer material and the same diagnostic-imaging indication is still a 510(k), if resolution, signal-to-noise, thermal, and electrical data sit next to the predicate without a new hazard. Software in the device follows FDA software documentation; the V&V how-to is not this page — software testing, software validation, device-level V&V.
FDA’s review clock for a 510(k) is 90 days. The clock stops on an Additional Information request. A messy file invites one. Common holes: ISO 10993 biocompatibility not justified by contact type and duration; sterilization and shelf-life not validated; software architecture / risk / V&V summary missing.
Unsure of the predicate or the test plan? File a Q-Submission (Pre-Sub) before you spend the study. Specific questions, not “what do we need to do.” FDA’s own Q-Sub guidance: early interaction on planned studies can shorten later review.
PMA: the clinical bar
Class III does not get a predicate comparison. A PMA is a scientific report that gives reasonable assurance of safety and effectiveness. Almost always that means human data. In the U.S. you usually need an Investigational Device Exemption before those trials start.
The protocol has to name endpoints, the population, the sites, and the statistics before the first patient. Manufacturing process validation and a quality-system inspection (including BIMO of the trial) ride with the file. MDUFA performance goals for original PMAs sit around 285 days of review — after the years of clinical work. Document from the first design decision; a gap in the history is a delay.
De Novo: novel, still moderate risk
No predicate, low-to-moderate risk, not automatically Class III — that is De Novo. You show the risk and propose the controls. If FDA grants it, they create a new regulation and product code. Your device becomes the predicate everyone else will 510(k) against. A first-of-kind diagnostic on a new biomarker, still Class II risk, is the textbook case.
Breakthrough and EUA (not extra pathways)
Breakthrough Devices Program is for a device that may more effectively treat or diagnose a life-threatening or irreversibly debilitating condition. You get more FDA interaction and a prioritized review. Designation is not authorization. It does not lower the evidence bar and it is not a fourth path next to 510(k) / PMA / De Novo — those still apply.
Emergency Use Authorization is temporary. It exists only during a declared public health emergency, under FD&C Act §564, when no adequate approved alternative is available. COVID diagnostics and ventilators used it. An EUA is not a clearance or an approval, and it expires with the emergency unless you convert through a real pathway.
If you also need Europe
This is still a U.S. process page. Europe is a different system, not a second article. FDA reviews in-house. The EU uses independent Notified Bodies; their certificate is the CE mark. The technical file is a living document — clinical evaluation, UDI, and post-market data have to keep updating it. Do not assume a U.S. class maps one-to-one onto MDR.
| Risk | US FDA | US example | EU MDR | EU example |
|---|---|---|---|---|
| Lowest | Class I | Elastic bandage | Class I | Stethoscope |
| Low–medium | Class II | Infusion pump | Class IIa | Hearing aid |
| Medium–high | Class II | Ventilator | Class IIb | Blood bag |
| Highest | Class III | Pacemaker | Class III | Prosthetic heart valve |
FAQ
What is a Q-Submission, and do I need one?
A Q-Sub is written FDA feedback before the marketing file. A Pre-Sub is the usual flavor. Not mandatory. Worth it when the predicate is fuzzy, the tech is new, or you have not filed before. Go in with a device description, a proposed class / product code / predicate, and specific questions.
Can I change the device after clearance?
Yes, after you evaluate the change. A shift that can significantly affect safety or effectiveness needs a new 510(k) (or a PMA supplement on the PMA side). New intended use, a different energy source, a new patient-contacting polymer — those come back. A color change or a small bug-fix patch usually does not. Every change still goes in the quality system. When it is close, ask FDA.
How long, and what does it cost?
510(k): FDA’s review goal is 90 days; the clock stops when they ask for more. Plan on about five months of review, plus the months to build the file. PMA: years of clinical work, then on the order of 285 days of review. De Novo review goal is 150 days. User fees are the small line. Generating the evidence — bench, biocompatibility, software, clinical — is the cost. Do not budget the fee and call it a plan.
Can I sell while FDA is reviewing?
No. A device that needs a 510(k) or PMA cannot be marketed in the U.S. until the clearance or approval letter is in hand.
If the device is imaging software, classification and the 510(k)/De Novo choice still sit with the manufacturer. PYCAD is the imaging stack, not the regulatory agent. Case studies.