Predetermined Change Control Plans for AI/ML Medical Devices: What You Need to Know Before Your Next FDA Submission

Why the PCCP Changes Everything for AI/ML Device Makers

If your medical device incorporates artificial intelligence or machine learning, you already know the core regulatory tension: FDA clearance or approval locks in a specific device configuration, but your algorithm is designed to evolve. Traditional change control pathways were built for static hardware and firmware -- not models that retrain, adapt, and improve over time.

The Predetermined Change Control Plan (PCCP) is FDA's answer to that tension. Introduced formally through the Consolidated Appropriations Act of 2023 and supported by FDA's January 2025 guidance document, 'Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence-Enabled Devices,' the PCCP gives manufacturers a structured mechanism to pre-specify certain future modifications -- and get advance regulatory acceptance for those changes -- without submitting a new 510(k), De Novo, or PMA supplement for each one.

Used correctly, the PCCP is a genuine competitive advantage. Used incorrectly, it creates a compliance gap that can trigger enforcement action or force a costly re-submission. Here is what your team needs to understand before your next submission.

The Regulatory Foundation: What the Law Actually Says

Section 3308 of the Consolidated Appropriations Act of 2023 amended the Federal Food, Drug, and Cosmetic Act to explicitly authorize PCCPs as part of a marketing submission. This was a significant shift -- previously, FDA had only addressed the concept in its 2019 discussion paper and the 2021 AI/ML-Based SaMD Action Plan.

Under the statutory framework, a PCCP must include three core components:

• Description of Modifications: A precise, bounded description of the types of changes the manufacturer anticipates making to the device.
• Modification Protocol: The specific methods, validation procedures, and performance testing the manufacturer will follow before implementing each change.
• Impact Assessment: An analysis of how the proposed modifications and the protocol itself affect the device's safety and effectiveness profile.

FDA's January 2025 guidance elaborates substantially on each element and is now the primary reference document for any team drafting a PCCP. If your regulatory team is not working directly from that guidance, stop and reorient your effort around it.

What Modifications Can -- and Cannot -- Be Covered

One of the most common mistakes startups make is drafting a PCCP that is either too vague to be accepted or too broad in scope. FDA is explicit: a PCCP cannot be used to pre-authorize modifications that would change the intended use of the device, introduce new risks not assessed in the original submission, or alter the device's fundamental technology in ways that would require independent safety and effectiveness review.

Modifications that are generally appropriate for PCCP coverage include:

• Performance improvements to an existing algorithm driven by additional training data from the same patient population and clinical context
• Updates to model architecture that do not change the device's indications for use or output interpretation
• Expansion of demographic or imaging subgroups that fall within the originally cleared intended use
• Retraining schedules with pre-specified performance thresholds and drift-detection criteria

Modifications that require a new or supplemental marketing submission -- regardless of how your PCCP is written -- include changes to indications for use, changes to the device's risk classification, and modifications that introduce cybersecurity vulnerabilities or alter predicate reliance in a material way.

Building a Modification Protocol That Will Survive FDA Review

The modification protocol is where most PCCP submissions succeed or fail. FDA reviewers are looking for specificity, rigor, and transparency. A protocol that reads like a general quality management procedure will not satisfy the requirement.

Your modification protocol should define:

• Pre-specified performance metrics (sensitivity, specificity, AUC, or task-specific measures) with quantitative acceptance criteria
• The datasets to be used for validation, including diversity and representativeness requirements aligned with FDA's 2024 guidance on AI/ML transparency
• Statistical methods for demonstrating that post-change performance is non-inferior to the cleared version
• Real-world performance monitoring methods under 21 CFR Part 820 post-market surveillance obligations
• Clear stop criteria that trigger escalation to a formal FDA submission if a proposed change falls outside the PCCP's defined scope

Your Design History File and Risk Management File under ISO 14971 should be structured to accommodate iterative updates, because every PCCP-authorized change still generates design change documentation under 21 CFR 820.30.

Practical Submission Strategy: Where the PCCP Fits

For 510(k) submitters, the PCCP is included as a standalone section within the submission, and FDA will review it concurrently with the primary clearance review. For De Novo requestors, the same structure applies. PMA holders can include a PCCP in an original PMA or in a PMA supplement.

One strategic point that is often overlooked: FDA can reject or condition a PCCP independently of the primary device review. A weak PCCP can delay your overall clearance timeline even if the core device data is solid. Treat the PCCP as a first-class regulatory deliverable, not an appendix.

Start with a Pre-Submission Meeting

If your AI/ML device is heading toward a marketing submission and you are planning to include a PCCP, a Pre-Sub meeting with FDA is not optional -- it is essential. Use the Pre-Sub to align on the scope of your modification descriptions, the adequacy of your validation framework, and FDA's expectations for your specific device type and risk classification. The feedback you receive will save months of back-and-forth during formal review.

At ADB Consulting and CRO Inc., we have helped AI/ML device companies structure PCCPs that satisfy FDA reviewers while giving development teams the operational flexibility they actually need. Whether you are drafting your first Pre-Sub request or rebuilding a rejected PCCP, the regulatory pathway is navigable -- but only if your documentation strategy is built on a precise understanding of what FDA requires.

Ready to get your PCCP strategy right the first time? Book a free discovery call with Andre Butler and the ADB Consulting team at adbccro.com. We will review your device classification, intended use, and AI/ML architecture to give you a concrete roadmap -- no generic advice, no wasted time.