Fatigue Verification Test Planning for Prosthetic Heart Valves—Sample Size and Statistical Analysis Considerations
Abstract
Fatigue verification tests confirm whether the structural component(s) of a heart valve meet predetermined reliability requirements with a specified level of confidence when subjected to test conditions equivalent to or surpassing expected in vivo conditions. This verification testing can be conducted using either an attribute-based approach or a variables-based approach, also known as the fatigue-to-fracture method. In both approaches, a comprehensive test program achieves the goal of providing insight into the circumstances and mechanisms under which component fracture may occur. For this purpose, a robust test program should generate data including both units that have survived and those that have fractured. Test planning is essential to establish sample sizes and strain test levels to verify that the component meets the desired reliability requirement. The two risks that the test plan must address are 1) the risk that an unreliable component will pass the test, i.e., type 1 error, and 2) the risk of a test failing for a sufficiently reliable component, i.e., type 2 error. Methods for test planning are provided to design a test plan that effectively fulfills these goals for both attribute-based and fatigue-to-fracture approaches. The attribute approach provides the benefits of straightforward planning and a simple output, with sample size calculated using a well-established formula. Historically, the test has been performed as a test-to-success. However, additional testing intended to generate fractures is recommended to strengthen the traditional attribute testing approach and identify the failure modes during testing. Although the fatigue-to-fracture (i.e., variables) approach requires more complex planning and analysis, it facilitates a deeper understanding of reliability. Sample sizes and strain test levels are established through a Monte Carlo simulated data strategy, clearly depicting potential test outcomes.