FDA PCCP Explained: What Medical Device AI Companies Need to Know

Your AI medical device learns from new data. Gets better over time. That's the whole point. But every time it changes, FDA wants to know about it. Traditionally that meant a new 510(k) submission. Every. Single. Time. The Predetermined Change Control Plan fixes this. Mostly. What a PCCP Actually Is A PCCP is a deal you make with FDA upfront. You tell them: here are the specific ways our AI will change after it ships. Here's how we'll validate each change. Here's how we'll know if something goes wrong. If FDA approves the plan, you can make those changes without filing new submissions. The key word is "predetermined." You can't just say "the model might change." You have to specify exactly what kinds of changes you're planning. The final guidance dropped in December 2024. It's dense, but the meat is in three sections. The Three Required Components 1. Description of modifications. What will change? New training data? Updated model weights? Expanded patient populations? Different clinical thresholds? You have to be specific. "We might improve the algorithm" won't fly. "We will retrain on additional radiology images from populations not represented in our initial dataset" will. 2.