Use of Real-World Data for Enhancing Model Credibility: Applications to Medical Device Development

Abstract Evaluating the credibility of computational models used in medical device development is increasingly important as medical devices become more complex and modeling takes on a more critical role in the device development process. While bench-testing based comparisons are common for assessing model credibility and have many advantages, such as control over test specimens and the ability to quantify outputs, the credibility assessments performed with bench tests often do not evaluate the clinical relevance of key aspects of model form (such as boundary conditions, constitutive models/properties, and geometries) selected when simulating in vivo conditions. Real-world data (outcomes data generated through clinical use of a device) offer an opportunity to assess the applicability and clinical relevance of a computational model. Although real-world data are frequently less controlled and more qualitative than benchtop data, real-world data are often a direct assessment of a particular clinical complication and therefore of high clinical relevance. Further, real-world data have the potential to reveal failure modes not previously identified in preclinical failure modes analysis, thereby motivating testing advancements. To review the use of clinical data in medical device modeling, this paper presents a series of examples related to tibial tray fracture that incorporate varying levels of benchtop data and real world data when evaluating model credibility. The merits and drawbacks of the credibility assessment for each example are discussed in order to provide practical and actionable guidance on the use of real-world data for establishing and demonstrating model credibility.