WHO Global Model Regulatory Framework for Medical Devices including In Vitro Diagnostic Medical Devices 8.4-SaMD: For SaMD and software in devices, including machine learning devices, demonstrate scientific validity, analytical validity and clinical performance, run life-cycle processes under a QMS, manage updates, cybersecurity and data quality, and monitor performance after deployment

What does 8.4-SaMD require?

Software meeting the device definition must be safe, secure and effective. Manufacturers should show scientific validity (the output is clinically well founded), analytical validity (it accurately and reliably produces the intended output from its input) and clinical performance (results correlate with the clinical condition in the target population and users), and should run ongoing life-cycle evaluation of real-market performance under QMS principles. They should plan clear, timely communication with users about updates (often installed by users), secure the software against cybersecurity threats, apply recognised standards because software cannot be verified by inspection alone, manage data quality with training and test datasets kept independent and representative of intended populations, design for appropriate human oversight, and monitor machine learning devices after deployment because real-world data shifts can affect performance.