Assessing Computational Model Credibility Using a Risk-Based Framework: Application to Hemolysis in Centrifugal Blood...

Area of Science:

• Biomedical Engineering
• Computational Science
• Regulatory Science

Background:

• Medical device manufacturers traditionally rely on internal best practices for computational modeling.
• A lack of consensus on model validation standards impedes broader acceptance, especially in regulatory contexts.
• The US Food and Drug Administration and the American Society of Mechanical Engineers (ASME) partnered to develop a structured approach for computational model credibility.

Purpose of the Study:

• To introduce the ASME V&V 40 standard for establishing computational model credibility in medical device design.
• To present a risk-informed credibility assessment framework where validation requirements are proportional to the model's associated risk.
• To illustrate the application of this framework using a centrifugal blood pump model.

Main Methods:

• Development of a risk-informed credibility assessment framework by the ASME V&V 40 Subcommittee.
• Application of the framework to a generic centrifugal blood pump model.
• Evaluation of model credibility requirements based on different contexts of use and associated risks.

Main Results:

• The framework demonstrates that computational model credibility requirements should be commensurate with the risk of model use.
• Experimental evidence, such as in vitro testing, can effectively support computational modeling for medical device evaluation.
• Different contexts of use for the same model necessitate varying Verification and Validation (V&V) activities and outcomes.

Conclusions:

• The ASME V&V 40 standard provides a structured, risk-informed approach to computational model validation for medical devices.
• This framework facilitates regulatory acceptance by standardizing the evidentiary requirements for model credibility.
• The study highlights the importance of tailoring V&V efforts to the specific risks associated with the intended use of computational models in medical device development.