R. V. Marrey, R. Burgermeister, M. Krever, Cordis Corporation, a Johnson & Johnson company, Warren, NJ; A. Sleeper, Successful Statistics, LLC, Fort Collins, CO
Background: Medical device design and testing methods are often based on usage of worst-case inputs, such that device safety and performance attributes can be conservatively characterized. Such a deterministic analysis is called worst-case analysis, due to the concurrent “worst-case” assumption for all the inputs to the design or testing process (Fig. 1). These methods can be excessively conservative, and in addition, do not quantify the degree of conservatism. Probabilistic analysis methods such as Monte Carlo simulations provide a way of understanding risk by generating mathematical distributions for device safety. These distributions allow predictions of device safety at desired levels of product reliability. Probabilistic method using input distributions to predict output distributions |
Figure 1. Comparison of deterministic and probabilistic methods
Methods: The probabilistic analysis process was used for numerical predictions of fatigue safety of a coronary stent under pulsatile loading conditions. For the stent design to be acceptable from a structural fatigue perspective, the predicted “Fatigue Safety Factor” was required to be greater than or equal to one (1.00). The Fatigue Safety Factor was a function of cyclic stresses in the stent structure during pulsatile loading, as well as stent material properties. A FEA model was used to predict the response of stent cyclic stresses as a function of stent strut width and wall thickness, stent material data, and stent loading variables. The probabilistic results were compared to a deterministic worst-case analysis.
Results and Conclusions: Implementation of a probabilistic approach provided insight into stent fatigue safety by assessments of its distribution mean and variability. Knowledge of the safety distribution helped manage risk through the understanding of product safety at various reliability levels.
Summary: Medical device design and testing methods are often based on usage of worst-case inputs, such that device safety and performance attributes can be conservatively characterized. Such a deterministic analysis is called worst-case analysis, due to the concurrent “worst-case” assumption for all the inputs to the design or testing process (Fig. 1). These methods can be excessively conservative, and in addition, do not quantify the degree of conservatism. Probabilistic analysis methods such as Monte Carlo simulations provide a way of understanding risk by generating mathematical distributions for device safety. These distributions allow predictions of device safety at desired levels of product reliability.
