The recent introduction of a new FDA regulation for software devices incorporating a pre-approved change control plan (PCCP) marks an interesting shift in the field. A new product code (QVD), under 892 2055, has been granted for a cardiac ultrasound device employing machine learning technology, signaling that the FDA is now accepting PCCPs in 510Ks. This signifies a significant change, enabling companies to navigate this route as opposed to the more complex de novo route for comparable devices. However, this comes with certain stipulations and special controls tied to identified risks associated with device output accuracy and potential misuse. Requirements include providing detailed descriptions of image post-processing algorithms, training data, performance testing protocols, and results, alongside maintaining a strict demarcation between training data and testing/validation data. This new regulation is anticipated to augment the number of devices implementing inhouse modifications under the PCCP, thereby eliminating the necessity to submit a 510K for these amendments.
