Table 2.

Application of the ASME V&V 40 risk-informed credibility framework to two different scenarios using computational models in the development of physiological closed-loop controlled medical devices.

		Scenario 1	Scenario 2
Question of interest		Design a controller to keep the physiological variable within X% of a set-point	Is a controller stable under variable patient conditions
Context of use		The controller will be synthesized by optimizing parameters to the computational patient model	The computational patient model will be used to perform a risk-based evaluation of the controller performance before being used in clinical studies.
Model risk	Influence on decision	High – no other evidence will be used to support the decision	High – no other evidence will be used to support the decision
	Consequence of decision	Low – following design of the controller, a series of studies including additional computational testing and animal studies will be performed to evaluate the controller performance before being used on patients	High – use of the controller on patients could lead to injury
	Overall risk	Medium	High
Example credibility factors		Qualitative: Parsimonious, low order, transparent, physiological relevant Quantitative: Linearizable, identifiable, predictive accuracy, reproduces uncertainty bounds	Qualitative: Transparent, physiologically relevant Quantitative: Identifiable, predictive accuracy, reproduces uncertainty bounds, generalizable
Credibility activities		Develop and perform plan to gather credibility evidence: experimental design, comparator (e.g., animal model), data analysis plan

The two scenarios outline example model risk and credibility factors. The noted model risk components including influence on decision and consequence of decision are different here to highlight how use of a model may influence the necessary evidence to support that use.