510(k) Audit Readiness: What FDA Reviewers Flag First in Class II Device Submissions

Most 510(k) submissions that stall at FDA don’t fail on the science. They stall on documentation gaps — structural deficiencies that trigger an Additional Information (AI) request and reset the review timeline by weeks or months. For Class II device manufacturers, the difference between a 90-day clearance and a 6-month back-and-forth often comes down to three decisions made before the submission is even assembled.

After reviewing dozens of pre-submission packages for clients across the device spectrum — from connected diagnostics to Class II surgical instruments — the pattern is consistent enough to be almost predictable. FDA’s reviewers at CDRH aren’t adversaries; they’re working from a structured checklist governed by 21 CFR Part 807, Subpart E, and a set of internal guidance documents that are publicly available but rarely read in full. Understanding exactly what that checklist prioritizes is, frankly, the most valuable thing regulatory compliance consulting services can offer a device manufacturer.

Predicate Selection: Where Most Submissions Lose the Argument

The concept of “substantial equivalence” sounds deceptively simple. Find a legally marketed predicate device, demonstrate that your device has the same intended use and either the same technological characteristics or safe and effective differences — done. In practice, predicate selection is where I see more preventable damage than anywhere else in the 510(k) process.

The most common trap: selecting a predicate that’s old enough to be defensible but different enough to require performance bridging data that was never collected. FDA’s reviewers compare your device’s intended use language word-for-word against the predicate’s cleared language. A single phrase mismatch — “point-of-care setting” vs. “clinical laboratory setting,” for example — can shift the equivalence burden significantly and invite an AI request for additional clinical validation.

Split predicate strategies (using different predicates to satisfy intended use and technological characteristics separately) are explicitly allowed under FDA guidance. But they require precise documentation of which predicate addresses which equivalence arm. When that mapping is implicit rather than explicit, reviewers ask for clarification. That question costs you 30 to 60 days.

Before any 510(k) goes out the door, the predicate rationale should be a standalone, reviewable document — not something reconstructed from the comparison table. That’s a discipline we apply in every pre-submission gap review at Aurora TIC, and it’s one of the first things our AI-augmented review process flags for completeness.

What the February 2026 QMSR Transition Changed for Your Quality Evidence

If you’ve been using quality system documentation drafted against the old QSR framework — 21 CFR Part 820 as it existed before February 2, 2026 — pay close attention here. FDA’s Quality Management System Regulation (QMSR) became effective on that date, replacing the legacy QSR and aligning 21 CFR Part 820 with ISO 13485:2016.

This matters for 510(k) submissions because quality system references in a device submission package aren’t just compliance boilerplate. When you describe your manufacturing processes, design controls, and post-market surveillance procedures in a 510(k), FDA reviewers cross-reference those claims against the regulatory standard that now governs them. Submissions that reference QSR-era documentation frameworks without acknowledging the QMSR transition create confusion for reviewers trained on an ISO 13485-aligned lens.

Specifically, the QMSR introduces explicit requirements for risk management integration — consistent with ISO 14971:2019 — as well as supplier controls and complaint handling structures that map directly to ISO 13485 section architecture. If your Design History File (DHF) and Device Master Record (DMR) language was built against 21 CFR 820.30 design controls terminology, verify that it speaks to the QMSR’s updated expectations before it appears in a regulatory submission.

The practical implication: a 510(k) submission package assembled in early 2025 under the old QSR framework may still be technically compliant for devices cleared under that timeline, but any submission filed today should explicitly reflect QMSR alignment. Documentation that reads as legacy creates a credibility gap — even when the underlying data is sound.

The Biocompatibility Blind Spot That Triggers AI Requests

Biocompatibility testing under ISO 10993-1:2018 remains the single most common source of Additional Information requests in device submissions with patient-contact components. The 2018 revision shifted the framework from a test-selection grid to a risk-based approach — meaning the burden now falls on the manufacturer to justify which tests were not conducted, not just to document the results of those that were.

FDA’s guidance on biocompatibility, updated most recently in 2023, is explicit: a biocompatibility evaluation must include a biological evaluation plan, a literature review, an inventory of all chemical substances in patient-contact materials, and either test data or a scientifically justified rationale for each ISO 10993-1 endpoint. Submissions that include a test report without a biological evaluation plan receive an AI request almost automatically. This is not a gray area.

For devices with indirect contact or very short contact duration (under 24 hours, categorized as “limited contact” under ISO 10993-1), there is often a legitimate case for waiving cytotoxicity, sensitization, or other endpoints — based on predicate equivalence and material characterization data. But that waiver argument has to be made explicitly, with citations to the predicate’s cleared biocompatibility position. Leaving the reviewer to infer it is not a strategy.

How AI-Augmented Pre-Submission Review Changes the Odds

Regulatory compliance consulting for medical devices has traditionally operated as a one-expert-reads-the-package model: a consultant reviews the submission, flags issues, and the manufacturer revises. That model is slow, expensive, and highly dependent on the individual consultant’s most recent exposure to FDA reviewer behavior.

What changes with AI-augmented review — and what we’ve built into our process at Aurora TIC — is the ability to cross-reference a submission package against a continuously updated library of FDA guidance, recognized standards, and observed AI request patterns in a fraction of the time. Our tools can scan a 510(k) draft for predicate language consistency, QMSR alignment gaps, and missing sections under 21 CFR Part 807.87 in minutes, not days.

This isn’t AI replacing regulatory judgment. It’s AI handling the structural audit so that expert judgment can focus on the substantive questions that actually require experience: Is the predicate selection defensible? Does the performance data tell a coherent safety and effectiveness story? Is the intended use language tight enough to survive a literal word-for-word comparison?

For clients preparing submissions, this means a pre-submission gap review that might have taken two weeks of consultant time can now surface high-priority issues in the first working session. The manufacturer walks away with a prioritized deficiency list, not a general impression. FDA’s MDUFA V performance goals call for 90% of standard 510(k)s to be reviewed within 90 calendar days. That window is achievable — but only if the submission doesn’t trigger an AI request. Every AI request is effectively a clock reset.

What to Confirm Before Your Next 510(k) Goes Out

If you’re preparing a Class II device submission in the next quarter, here’s what the pre-submission checklist should confirm before the package is assembled:

Predicate documentation: The predicate rationale is a standalone section, not embedded in the comparison table. Intended use language is cross-referenced word-for-word against the cleared predicate. If a split predicate is used, each equivalence arm is mapped explicitly.

QMSR alignment: All quality system references reflect the February 2026 QMSR framework and ISO 13485:2016 terminology. DHF and DMR documentation has been reviewed for QSR-legacy language that may read as outdated.

Biocompatibility position: A biological evaluation plan exists and addresses all ISO 10993-1:2018 endpoints. Any waived tests have a written, scientifically justified rationale with explicit predicate support.

eSTAR completeness: Since FDA mandated the eSTAR (electronic Submission Template and Resource) format for most 510(k)s, the structured template enforces section completeness — but it doesn’t evaluate the quality of what’s in each section. A structurally complete eSTAR package can still contain deficient content that triggers an AI request.

Q-Submission strategy: For novel technologies, combination products, or borderline classification situations, a pre-submission meeting (Q-Sub) with CDRH is almost always worth the approximately 60-day turnaround time. It’s the one opportunity to surface FDA’s informal position before the formal review clock starts.

The submissions that clear on the first review aren’t the ones with the most data. They’re the ones where every question a reviewer might ask has already been answered in writing.

Written by Sam Sammane, Founder & CEO, Aurora TIC | Founder, Qalitex Group. Learn more about our team

Reserve early access to our AI audit tools Contact us

Related from our network

• ISO 17025-Accredited Performance and Biocompatibility Testing for Medical Device Components — When your 510(k) requires third-party test data, ISO 17025-accredited results from Qalitex Laboratories carry the evidentiary weight FDA reviewers expect.
• GMP Compliance Testing for Canadian Medical Device Manufacturers — Health Canada’s medical device quality system requirements share significant structural overlap with FDA’s QMSR — Androxa’s testing services support manufacturers operating on both sides of the border.