Medical Device Classification

The Medical Device Amendments of 1976 established three device classes under section 513 of the FD&C Act, codified at 21 CFR 860. The classes are defined by the regulatory controls FDA considers sufficient to provide reasonable assurance of safety and effectiveness, not by an abstract risk score. The three control tiers stack: Class II carries everything Class I carries plus 'special controls'; Class III carries everything plus PMA premarket approval.

Roughly half of all listed devices are Class I (and most of those are 510(k)-exempt), about 43% are Class II, and around 8% are Class III. The disproportion matters: it means the 510(k) review process — not PMA — is the dominant FDA premarket workflow, and the device industry is largely a Class II industry.

Three steps. First, identify the FDA generic device type. The FDA Product Classification Database lists ~1,700 generic device types, each tied to a three-letter product code and a CFR citation (one of 21 CFR 862 through 892). Searching the database by device function usually returns the right type within a few candidates. Second, read the device's classification regulation to confirm intended use match. Third, check the regulation for class assignment (I, II or III) and for any special controls or 510(k)-exemption notes.

If no generic device type fits — the device is genuinely novel — the device defaults to Class III under section 513(f)(1) and would otherwise need a PMA. The De Novo classification request under section 513(f)(2) (21 CFR 860 Subpart D) lets the sponsor argue that the novel device is in fact low or moderate risk and should be reclassified into Class I or II with appropriate controls. A successful De Novo creates a new generic device type and product code that future devices can use as a 510(k) predicate.

Disagree with FDA's classification? Submit a 513(g) Request for Information. CDRH responds within 60 days with the agency's classification view. This is non-binding but persuasive, and the small user fee is usually money well spent before committing to a regulatory strategy.

In vitro diagnostic devices follow the same three-class scheme but appear in 21 CFR 862 (chemistry / toxicology), 864 (haematology / pathology), and 866 (immunology / microbiology). Class III IVDs cover the highest-impact diagnostics — blood-bank screening, HIV diagnostics, life-threatening infectious-agent assays, and companion diagnostics for high-impact therapies. Most clinical chemistry analyte assays are Class II; controls, calibrators and general laboratory reagents are Class I.

CLIA test categorisation (waived, moderate, high) is layered on top of FDA classification and is assigned at clearance. Classification under the FD&C Act and CLIA category are independent — a Class II IVD can be CLIA-waived, moderate, or high complexity depending on operational characteristics.

EU MDR (2017/745) Annex VIII has 22 rules. Devices fall into Class I (lowest risk), Class IIa, Class IIb, or Class III (highest risk). Rules group by device characteristic: non-invasive (Rules 1–4), invasive (Rules 5–8), active (Rules 9–13), special cases (Rules 14–22) including reusables, surgical instruments, software (Rule 11), nanomaterials (Rule 19), CMR/endocrine substances (Rule 20), implantable substances (Rule 21), and active therapeutic devices with diagnostic functions (Rule 22).

The classification rule of thumb: duration of contact (transient < 60 min, short-term ≤ 30 days, long-term > 30 days), invasiveness (surface, body orifice, surgical, implantable), active vs non-active, and intended use (diagnostic, monitoring, therapeutic, life-supporting). MDCG 2021-24 walks through every rule with worked examples and should be the first reference for any EU classification call.

Up-classification under MDR was significant. Devices that were Class I under MDD frequently land in Class IIa or IIb under MDR. Software under Rule 11 has been the most dramatic: software providing information used for diagnostic or therapeutic decisions is now Class IIa minimum, IIb if the decisions could cause serious deterioration, Class III if death or irreversible deterioration. Standalone software that was self-certified Class I under MDD typically needs a Notified Body under MDR.

Classification is not permanent. FDA can reclassify a device under section 513(e) on its own initiative or in response to a citizen petition under 513(f)(3). Down-classifications of legacy Class III devices have been a recurring CDRH workstream; AED automated external defibrillators were down-classified from Class III to Class II in 2015 after long FDA scrutiny. Up-classification is rarer but happens when post-market data reveals risk a manufacturer cannot mitigate with general or special controls alone — recalled or restricted device types are the typical trigger.

In the EU, classification can shift when MDCG issues new interpretive guidance — Rule 11 software classification is the live example. Manufacturers should monitor MDCG output and update their classification rationale in the technical documentation at every periodic safety update.

• Mapping the device to a product code that almost fits and skipping the 513(g). The product code drives review division assignment and predicate set; a wrong code leads to wrong reviewers and wasted AI cycles.
• Treating an accessory as unclassified. Accessories take the class of their parent device unless FDA has classified them separately under section 513(b).
• Assuming EU and FDA classes align. They don't. A Class II FDA device can be Class IIa, IIb or III under MDR depending on intended use, duration and invasiveness.
• Ignoring Rule 11 for clinical decision-support software. A Rule 11 self-assessment is mandatory and is a routine Notified Body scoping question.
• Forgetting reclassification at periodic review. MDR Article 61 expects classification to be re-evaluated at PSUR cadence.