Clinical Evaluation of Software as a Medical Device (SaMD)
This guidance focuses on the clinical evaluation activities needed for Software as a Medical Device (SaMD), providing a harmonized approach for demonstrating the clinical association, analytical validation, and clinical validation of SaMD. It aims to establish a common understanding of clinical evaluation principles for demonstrating safety, effectiveness and performance of SaMD.
Recommended Actions
- Establish clear SaMD definition statement and risk categorization
- Develop clinical evaluation plan covering:
- Valid clinical association
- Analytical validation
- Clinical validation
- Generate appropriate clinical evidence based on risk category
- Implement independent review process for higher risk SaMD
- Establish continuous monitoring system for real world performance
- Develop post-market surveillance plan
- Maintain documentation of clinical evaluation process
- Implement change management process for SaMD modifications
- Ensure appropriate labeling and user instructions
- Establish process for ongoing assessment of clinical evidence
Key Considerations
Clinical testing
- Clinical evaluation should be an iterative and continuous process as part of the quality management system
- Clinical validation measures the ability of SaMD to yield clinically meaningful outputs
- Level of clinical evaluation should be commensurate with the risk posed by the SaMD
- Clinical validation can be demonstrated through existing data, extrapolated data, or new clinical data
Non-clinical testing
- Analytical validation measures the ability to accurately and reliably generate intended technical output from input data
- Verification and validation activities should be conducted as part of quality management system
- Evidence can be generated through use of curated databases or previously collected patient data
Human Factors
- Clinical usability and user interface considerations must be evaluated
- Safety considerations for device users, use environments and user interfaces must be assessed
Software
- Software verification and validation must be conducted
- Continuous monitoring of software performance should be implemented
- Software changes must be managed through appropriate processes
Labelling
- Clear definition statement about intended use must be provided
- Limitations of the SaMD relevant to clinical performance must be identified
- Instructions for use must be appropriate for intended users
Safety
- Safety data collection and monitoring required
- Real world performance data should be collected to monitor safety
- Risk assessment should be considered when conducting clinical evaluation
Other considerations
- Independent review of clinical evidence recommended for higher risk SaMD
- Real world performance data should be leveraged for continuous learning
- Post-market surveillance should be conducted
- Changes to SaMD definition statement may require regulatory review
Relevant Guidances
- Content of Premarket Submissions for Device Software Functions
- Off-The-Shelf Software in Medical Devices: Documentation Requirements for Premarket Submissions
- Software Validation for Medical Device Production, Quality Systems, and Device Components
- Cybersecurity in Medical Devices: Design, Implementation, and Premarket Submissions
- Clinical Decision Support Software Functions: Criteria for Non-Device Classification and Implementation Requirements
- Policy for Device Software Functions and Mobile Medical Applications
- Changes to Medical Device Definition for Software Functions Under the 21st Century Cures Act
Related references and norms
- ISO 14155:2011: Clinical investigation of medical devices for human subjects - Good clinical practice
- ISO 14971:2007: Application of risk management to medical devices
- ISO 62304/A1:2015: Medical device software – Software life-cycle processes
- IEC 62366-1:2015: Medical Devices – Part 1: Application of usability engineering to medical devices
- ISO 82304-1:2016: Health software – Part 1: General requirements for product safety
Original guidance
This post is licensed under CC BY 4.0 by the author.