Assessing the use of finite element analysis for mechanical performance evaluation of intervertebral body fusion devices

Abstract Background Intervertebral body fusion devices (IBFDs) are a widely used type of spinal implant placed between two vertebral bodies to stabilize the spine for fusion in the treatment of spinal pathologies. Assessing mechanical performance of these devices is critical during the design, verification, and regulatory evaluation phases of development. While traditionally evaluated with physical bench testing, empirical assessments are at times supplemented with computational models and simulations such as finite element analysis (FEA). However, unlike many mechanical bench tests, FEA lacks standardized practices and consistency of implementation. Objectives The objectives of this study were twofold. First, to identify IBFD 510(k) submissions containing FEA and conduct a comprehensive review of the elements provided in the FEA reports. Second, to engage with spinal device manufacturers through an anonymous survey and assess their practices for implementing FEA. Methods First, a retrospective analysis of 510(k) submissions for IBFDs cleared by the FDA between 2013 and 2017 was performed. The contents of FEA test reports were quantified according to FDA guidance. Second, a survey inquiring about the use of FEA was distributed to industry and academic stakeholders. The survey asked up to 20 questions relating to modeler experience and modeling practices. Results Significant gaps were present in model test reports that deemed the data unreliable and, therefore, unusable for regulatory decision‐making in a high percentage of submissions. Nonetheless, the industry survey revealed most stakeholders employ FEA during device evaluation and are interested in more prescriptive guidelines for executing IBFD models. Conclusions This study showed that while inconsistencies and gaps in FEA execution do exist within the spinal device community, the stakeholders are eager to work together in developing standardized approaches for executing computational models to support mechanical performance assessment of spinal devices in regulatory submissions.

exist to help align modeling practices across stakeholder groups.
As with bench testing, having best practices for executing computational analyses helps to ensure credibility and comparability of results. Modeler expertise and availability of resources can lead to significant differences in model form, execution, and reporting of results.
Documents have been developed to help guide FEA and subsequent reporting, such as the FDA guidance on Reporting of Computational Modeling Studies in Medical Device Submissions, ASTM International's standards, and the American Society of Mechanical Engineering (ASME) Verification and Validation (V&V) documents. [4][5][6][7][8][9] However, no such standard exists specifically for IBFDs, and implementation of FEA remains at the discretion of the modeler. Thus, it is important to define a best practices approach for FEA of IBFDs to help standardize numerical techniques and advance the use of FEA in the spinal device industry.
In order to define best practices for using FEA to evaluate mechanical performance of IBFDs, it is important to first understand the current state of simulations as well as the needs of stakeholders.
Therefore, the objectives of this current study were 2-fold. First, to identify IBFD 510(k) submissions containing FEA and conduct a comprehensive review of the reporting elements provided in the FEA reports. Second, to engage with spinal device manufacturers through an anonymous survey and assess their practices for implementing FEA in research and development activities. Together, these tasks would identify gaps in how FEA is currently being used and guide future efforts in defining best practices.

| 510(k) review
Previously, a retrospective analysis of 510(k) submissions was conducted and mechanical performance of FDA-cleared IBFDs were summarized. 10
General introduction and background language in the test reports indicated that all FEA was used to determine a worst-case device size or shape to then be selected for bench testing according to ASTM F2077 (Table 1). Worst-case selection was made by simulating the relevant loading modes outlined in ASTM F2077 on the range of IBFD designs in the 510(k) submission. Cage geometry was simplified more often than using the exact cage geometry. Load fixture geometry was included in 33 of the reports (51%), and most of these fixtures included pockets which mated with the cage geometry (35%). For constitutive laws, linear elasticity was used most often (42%). The most common loading and boundary conditions simulated compression testing (92%), followed by compression-shear (49%), and torsion (34%). Among the 33 reports which included fixture geometry, 70% described contact conditions used between cages and fixtures. These included bonded (27%), no separation (24%), friction (12%), and fric-

| Stakeholder survey
A total of 31 stakeholders from within the United States participated in the survey ( Results of the FDA 510(k) submission review showed large disparities in the information provided in FEA reports. Some sections like system geometry, boundary and initial conditions, and results were nearly always reported. Conversely, other sections such as code verification, validation, and a mesh convergence study were provided no more than one third of the time. In order for regulators to establish credibility in computational models, all sections should be adequately represented. While most of the recommended information may be present in a FEA report, the information is only useful if the report contains all sections. Therefore, it is critical for regulators to clearly articulate expectations, and for stakeholders to adhere to them. Failure to do so may produce FEA results that are unreliable or difficult to interpret, which may in turn undermine the overall utility of the model for the purposes of regulatory review and decision making. Modelers with a formal education in FEA are generally exposed to these principles in textbooks. 18,19 Additionally, these sections are described to varying degrees in ASME V&V 10, 7 NASA-STD-7009, 20 SAND2007-5948, 21  warranted to ensure these approaches are clearly communicated and well-understood by all stakeholders. These steps will improve our confidence in modeling and simulation and enhance its role in regulatory review, thereby reducing costs and burden for the community.