How to Write an FDA Predetermined Change Control Plan (PCCP): Step-by-Step Template
An FDA PCCP is written by defining the locked initial version of an AI/ML-enabled software function, then documenting a bounded set of anticipated modifications, the protocol used to verify and validate each one, and the risk controls tied to each change. A usable draft organizes this into three linked sections: description of modifications, modification protocol, and impact assessment, with clear traceability between them.
What a PCCP Is and Why It Matters for AI/ML SaMD
A Predetermined Change Control Plan lets a manufacturer define, in advance, how an AI/ML-enabled software function is allowed to change after FDA authorizes the initial, locked version. Rather than resubmitting for every model update, the manufacturer commits to a bounded set of anticipated modifications, each backed by its own verification and validation protocol and mapped to specific risk controls.
The Three Core Components of a PCCP
A usable draft organizes this commitment into three linked sections, with clear traceability from the modification that is described, to the protocol used to test it, to the risk it addresses.
Description of Modifications
The specific, bounded changes anticipated after initial authorization.
Modification Protocol
How each anticipated change will be verified, validated, and approved before deployment.
Impact Assessment
How each modification affects safety, effectiveness, and risk profile relative to the authorized device.
Step-by-Step: Drafting Your PCCP
Structuring the Modification Protocol for Verification and Validation
A common weakness in draft PCCPs is a single protocol written to cover all future changes at once. A more defensible structure separates concerns explicitly, giving each anticipated modification its own protocol, acceptance criteria, and sign-off path.
PCCP Readiness Checklist
Use the following checklist to confirm each modification protocol is specified narrowly enough to stand on its own before submission.
- Every anticipated modification is described narrowly enough to be independently tested.
- Each modification has its own verification and validation protocol with pre-specified acceptance criteria.
- Risk controls are mapped to specific modifications, not stated as general mitigations.
- A clear boundary condition defines when a change falls outside the plan's scope.
- Internal sign-off roles are documented for each modification protocol before submission.
- A version control process exists for amending the plan itself over the product lifecycle.
Documentation Practices That Support Ongoing Compliance
Writing the PCCP is only the first step. The practices below support demonstrating, after authorization, that executed changes stayed within the plan's authorized scope.
- Maintain an execution log: Keep a running record of every modification executed against the authorized plan, including the verification results and sign-off.
- Pre-define data sources: Fix the data sources and sample sizes for each anticipated change in advance to avoid ambiguity when the modification is executed.
- Review the plan periodically: Reassess whether the PCCP still reflects current product architecture and risk posture, and version any amendments explicitly.
- Close the loop on monitoring: Document how post-deployment monitoring results were evaluated against the plan's acceptance criteria and what decision followed.
- Centralize approval records: Retain evidence of which role approved each modification's execution, distinct from who authored the protocol.
A PCCP defines what changes are authorized on paper. Demonstrating that executed changes stayed within that scope, with monitoring, audit logging, and controlled approval workflows, is a separate operational discipline that continues after authorization.
Operationalizing Change Control Beyond the Submission
A PCCP defines what changes are authorized on paper. Demonstrating that executed changes stayed within that scope, with monitoring, audit logging, and controlled approval workflows, is a separate operational challenge that continues after authorization.
Explore Runtime Governance