Core clinical engineering and protocols for FDA GMLP & Change Control Plans.
Historically, FDA medical device regulations assumed static software: any change to an algorithm required a new 510(k) premarket notification. However, modern AI algorithms improve continuously as they encounter diverse patient populations. The FDA's Predetermined Change Control Plan (PCCP) framework establishes a legal mechanism for adaptive SaMD.
AgenticSaMD implements automated pipeline verification for PCCPs: when updated hospital data triggers automated retraining, the candidate model is subjected to automated regression test suites evaluating sensitivity, specificity, and demographic fairness. Only models that strictly meet pre-approved performance criteria are promoted to clinical production.