510(k)Premarket Notification 510(k)

A 510(k) is a Premarket Notification — a submission to FDA's Center for Devices and Radiological Health (CDRH) under section 510(k) of the Federal Food, Drug, and Cosmetic Act and the procedural rules of 21 CFR Part 807 Subpart E. It tells FDA that a manufacturer intends to introduce, or reintroduce after a significant change, a device into commercial distribution in the United States. The submission must demonstrate that the device is 'substantially equivalent' (SE) to a legally marketed predicate device — one that itself was either pre-Amendments (in commerce before 28 May 1976), reclassified out of Class III, or previously cleared through 510(k).

If FDA agrees, the agency issues a substantial-equivalence determination — colloquially 'clearance' — and the device may be commercially distributed. Substantial equivalence is not approval (that word is reserved for PMA / De Novo / HDE pathways). 510(k) clearance does not mean FDA endorses safety or effectiveness in absolute terms; it means FDA agrees the device is at least as safe and effective as the predicate. This distinction matters in advertising and in litigation — labelling a 510(k)-cleared device as 'FDA-approved' is a misbranding violation.

510(k) is the dominant route to market by volume. CDRH clears roughly 3,000 510(k)s per year vs roughly 50 PMA approvals. Most Class II devices (and a small number of low-risk Class III devices not yet called for PMA) reach the US market via 510(k). The performance commitment under MDUFA V is a 90 FDA-day target for an SE / NSE decision; the real-world median across MDUFA reporting is closer to 177 calendar days once Additional Information (AI) hold cycles and interactive review are counted.

21 CFR 807.81 sets out when a 510(k) must be submitted. Submission is required before any of the following:

• Introducing into commercial distribution for the first time a device that is not exempt from 510(k) (most Class II devices and some Class I).
• Reintroducing a device that has been significantly changed or modified in design, components, method of manufacture or intended use such that the change could significantly affect safety or effectiveness. The decision tree in FDA's Deciding When to Submit a 510(k) for a Change to an Existing Device (2017) is the controlling guidance.
• Repurposing a device for a new intended use beyond the cleared indications.

Exemptions: many Class I devices and a subset of Class II devices are 510(k)-exempt per 21 CFR 862-892 and the Medical Device User Fee and Modernization Act exemption lists. Exempt devices must still comply with the QMSR (21 CFR 820, effective 2 Feb 2026 / current QSR until then), MDR / vigilance, UDI and establishment registration / device listing. Exemption from 510(k) is not exemption from CGMP or post-market obligations.

Change-decision discipline is the most-misunderstood area. A common 483 / Warning Letter pattern is design changes implemented under change control without a new 510(k) when the change-decision rationale was either absent or not defensible. Document every change-control decision against the 2017 decision tree, even when the conclusion is 'no new 510(k) required' — that rationale is the inspector's audit trail.

FDA recognises three 510(k) submission formats. Picking the right one shortens the review clock and reduces AI cycles.

The Special 510(k) Program (2019 guidance) materially broadened eligibility — including for changes to indications for use that do not affect the intended use, and for some technological changes — provided the change can be demonstrated through design-control verification and validation. Use Special when criteria are met; do not default to Traditional.

The substantive determination is made through FDA's SE decision tree (The 510(k) Program: Evaluating Substantial Equivalence, 2014). The tree asks, in order:

• Is the predicate device legally marketed? (Pre-Amendments, reclassified, or previously cleared. Predicates withdrawn for safety reasons are not eligible.)
• Does the new device have the same intended use as the predicate? Intended use = the general purpose stated in labelling. If different — NSE (Not Substantially Equivalent) — and the device drops into Class III by default unless De Novo is pursued.
• Does the new device have the same technological characteristics as the predicate? If yes, proceed to performance-data review.
• If technological characteristics differ, do the differences raise different questions of safety and effectiveness? If yes — NSE.
• If differences do not raise different questions, do the performance data demonstrate the new device is as safe and effective as the predicate? If yes — SE clearance issued.

Within 15 calendar days of receipt, FDA conducts an RTA review against the published RTA checklist. The checklist verifies the submission contains the required administrative and organisational elements — user fee paid, eCopy / eSTAR submitted, CDRH Premarket Review Submission Cover Sheet (Form FDA 3514), indications for use, 510(k) summary or statement, declarations of conformity, performance-testing summaries, labelling, etc. RTA is administrative completeness, not scientific adequacy.

If the submission fails RTA, FDA issues an RTA hold letter listing missing items. The submitter has 180 days to respond; the 90-day review clock does not start until acceptance. Repeated RTA holds are the single largest avoidable cause of multi-month delays — most failures are missing declarations of conformity, missing electrical-safety / EMC summaries (IEC 60601-1 / -1-2 / -2-x), missing software documentation level rationale, or missing cybersecurity premarket documentation per the 2023 guidance.

Since 1 October 2023, the eSTAR template is mandatory for 510(k) submissions. eSTAR enforces structural completeness — required attachments cannot be skipped — and has substantially reduced first-pass RTA failures. The eSTAR build-up is itself the discipline; treat it as the construction of the submission, not a post-hoc packager.

Per the Format for Traditional and Abbreviated 510(k)s guidance, a 510(k) contains 20 structured sections. The eSTAR template enforces this structure.

1. Medical Device User Fee Cover Sheet (Form FDA 3601A).
2. CDRH Premarket Review Submission Cover Sheet (Form FDA 3514).
3. 510(k) Cover Letter.
4. Indications for Use Statement (Form FDA 3881) — verbatim what will be on the label.
5. 510(k) Summary (per 21 CFR 807.92) or 510(k) Statement (per 807.93). Summary is the public-facing summary published on FDA's 510(k) database.
6. Truthful and Accuracy Statement (per 807.87(j)) — signed.
7. Class III Summary and Certification (only if the device is Class III).
8. Financial Certification or Disclosure Statement (Form FDA 3454/3455) — only if clinical data submitted.
9. Declarations of Conformity and Summary Reports — for every FDA-recognised consensus standard relied upon.
10. Device Description — physical and performance characteristics, principles of operation, components, accessories, photos / drawings.
11. Executive Summary — the technical narrative for the reviewer; predicate comparison, performance-data summary, conclusion.
12. Substantial Equivalence Discussion — predicate identification, side-by-side comparison table (intended use, indications, technological characteristics), discussion of any differences.
13. Proposed Labelling — IFU, labels, package inserts. Must match Form FDA 3881 indications.
14. Sterilization and Shelf Life — sterilisation method validation (ISO 11135 / 11137 / 17665), shelf-life real-time + accelerated.
15. Biocompatibility — per ISO 10993 series, biological-evaluation plan + report.
16. Software — per FDA Premarket Submissions for Device Software Functions guidance, with documentation level (Basic or Enhanced) and the full software documentation package.
17. Electromagnetic Compatibility, Wireless, Electrical, Mechanical, and Thermal Safety — IEC 60601 series test reports.
18. Performance Testing — Bench — non-clinical performance demonstrating SE.
19. Performance Testing — Animal — if applicable.
20. Performance Testing — Clinical — if applicable. Most 510(k)s do not include clinical data; ~10% do.

Cybersecurity premarket documentation per the September 2023 final guidance applies to any device with cybersecurity considerations and includes: Security Risk Management documentation, Security Architecture views (global system view, multi-patient harm view, updateability view, security-use-case view), Software Bill of Materials (SBOM) in a machine-readable format (SPDX or CycloneDX), security testing (vulnerability + penetration + static / dynamic analysis), and labelling. Cybersecurity is now the single most common AI-letter topic for connected devices.

The 90-day MDUFA review clock is in FDA days — calendar days excluding time the submission is on hold for AI (Additional Information) response or for other applicant-side actions. The typical lifecycle:

1. Day 1 — Submission received; technical screen begins (eSTAR auto-validates structure).
2. Day 15 — RTA decision. Either accepted (clock starts) or RTA hold issued (180 days to respond).
3. Day 60 — Substantive Interaction expected: either an AI request, an Interactive Review email, or an early SE/NSE decision. FDA's MDUFA V commitment is to have substantive interaction with the submitter by day 60.
4. Day 60-90 — Interactive Review cycles; minor data clarifications by email without formal AI hold.
5. Day 90 — MDUFA goal date: SE / NSE decision, or formal AI request placing the clock on hold.
6. AI hold — Up to 180 days for the submitter to respond. Clock resumes when complete response received. Multiple AI cycles are common but discouraged; each cycle compounds calendar-day delay.

If FDA cannot make an SE determination based on the data, the outcome is NSE (Not Substantially Equivalent) and the device is automatically classified Class III unless the manufacturer pursues De Novo classification. NSE letters are public records.

AI / ML-enabled devices have driven a meaningful share of recent 510(k) activity. FDA's December 2024 final guidance on Predetermined Change Control Plans formalised the route by which an applicant can pre-specify, in the original 510(k) submission, a defined set of model modifications that may be implemented post-market without a new 510(k). A PCCP has three components:

• Description of Modifications — the specific, planned modifications to the device's AI / ML function (e.g. periodic retraining on additional data, expansion to a new patient subgroup, performance recalibration).
• Modification Protocol — the methods that will be used to develop, verify, validate, and implement each modification, including data management, retraining protocols, performance evaluation, update procedures, and transparency provisions.
• Impact Assessment — analysis of the benefits and risks of the planned modifications and the mitigations.

A cleared PCCP allows modifications within scope to be implemented without new 510(k) clearance — a substantial post-market velocity advantage for ML devices that retrain on real-world data. PCCPs are increasingly expected for any 510(k) involving AI / ML functions; their absence in such submissions frequently triggers an AI hold.

FDA's Quality Management System Regulation amendment (89 FR 7496) takes effect 2 February 2026 and replaces most of the existing 21 CFR 820 QSR with incorporation-by-reference of ISO 13485:2016, supplemented with FDA-specific add-ons. For 510(k) submitters this means: the design-control evidence (design history file) that feeds the performance-data section now follows ISO 13485 Clause 7.3 terminology; the manufacturer's QMS is auditable under either ISO 13485 (third-party) or FDA inspection (with the same substantive content); and MDSAP audits now carry the QMSR semantic. The 510(k) submission content itself does not change materially — the underlying QMS evidence does.

• RTA failure — missing declarations of conformity, missing cybersecurity premarket documentation, missing software documentation level rationale, missing IFU draft, missing Form FDA 3881.
• Predicate identification weak — predicate cleared on different intended use, predicate withdrawn for safety reasons, split predicate, predicate too old without bridging analysis.
• Indications for Use drift — the 510(k) IFU differs from the predicate in ways that affect intended use without justification.
• Performance testing — bench testing does not match the proposed indications (e.g. tested on saline but indicated for whole blood); sample sizes inadequate; acceptance criteria not justified.
• Sterilisation — validation does not cover the worst-case loading configuration; residuals data (ethylene oxide) absent or out of spec.
• Biocompatibility — gap analysis vs ISO 10993-1 (2018) missing; tests run on prototype not production-equivalent material; rationale for not testing a category absent.
• Software — documentation level rationale missing or wrong; cybersecurity package per 2023 guidance absent or incomplete; SBOM missing or not machine-readable.
• Human factors — usability validation absent for a use environment / user group that would normally require it (especially home-use, OTC, lay-user).
• Clinical data (when included) — endpoint not pre-specified, sample size not justified, missing protocol amendments, statistical analysis plan absent.
• Labelling — IFU contradicts performance testing; warnings / precautions inconsistent with risk-management file; UDI placement non-compliant.
• Change-decision rationale — for a modified-device 510(k), the manufacturer's rationale for the change-trigger is absent or does not align with the 2017 guidance decision tree.
• AI / ML — PCCP absent for a learning model where post-market retraining is anticipated; performance evaluation does not include subgroup analysis; training-data documentation thin.
• Submission organisation — eSTAR sections cross-referenced ambiguously, attachments misplaced, summary document inconsistent with detailed reports.

• RTA pass rate (first submission).
• Days from internal lock to FDA receipt.
• FDA days to substantive interaction (target ≤60).
• AI cycles per submission (target ≤1).
• Calendar days submission-to-clearance (median; target trending down).
• Predicate selection — % of submissions using primary predicate from same manufacturer's portfolio (re-use velocity).
• Special 510(k) usage % of all change-driven submissions (eligibility utilisation).
• Abbreviated 510(k) usage % where consensus-standard coverage exists.
• Cybersecurity documentation completeness pre-submission audit score.
• PCCP inclusion rate among AI / ML-enabled submissions.
• Change-decision tree rationales documented for in-scope modifications.
• Post-clearance MDR / FAERS signals correlated to product family — feedback into next-generation submission.

V5 Ultimate treats the 510(k) as a structured workspace anchored to the discrete-industry design-control framework. Each device program carries a predicate-tracker view — primary predicate, candidate alternatives, reference devices, intended-use comparison, technological-characteristic comparison table, gap analysis driving the performance-data plan. The design history file is the source of truth for the performance-data section; verification and validation outputs auto-roll into the relevant 510(k) subsection with traceability to design inputs, risk controls, and labelling claims.

The eSTAR-aligned submission workspace mirrors the 20-section format with completeness gates against the published RTA checklist. Declarations of Conformity are generated from the consensus-standards register; cybersecurity package contents are pre-validated against the September 2023 guidance (Security Risk Management, architecture views, SBOM in CycloneDX / SPDX, security testing summaries). Software documentation level is selected from a decision-tree workspace tied to the IEC 62304 safety classification. For AI / ML programs, a PCCP authoring workspace captures the description of modifications, modification protocol, and impact assessment in submission-ready form.

Change-control entries are scored against the 2017 'when to submit' decision tree; any 'no new 510(k) required' conclusion captures the rationale in inspector-ready form. The submission timeline records every FDA interaction with the AI cycle counter and the calendar / FDA-day deltas. Post-clearance MDR signals from the complaint module feed back into the same workspace so the next 510(k) is informed by the field experience of the cleared device.