Abstract
Pre-clinical validation of implantable devices, including prostheses, generally aims at demonstrating that a new device offers some advantage compared to existing ones, while not introducing additional hazards. This process involves the assessment of a number of possible failure scenarios and claimed benefits, in order to obtain certification of the device (e.g. FDA or CE-mark), and to support its marketing strategy.
While until the 90ies in vitro tests were regarded as the golden standard, nowadays the trend is to rely more and more on numerical models (chiefly Finite Element models, FE). The truth is that neither numerical models nor in vitro tests are self-sufficient. FE models require the support of in vitro tests for a number of reasons. First of all, to construct reliable FE models a number of input parameters are required (e.g. material properties, friction coefficients) that can only be measured experimentally. Furthermore, FE models, like any model, can only address the scenarios they are intended for, and cannot predict something that is totally unexpected: for this reason, some preliminary indication is mandatory from in vitro tests. Finally, FE models cannot be assumed true until this is proven by validation against in vitro measurements. At the same time, in vitroexperiments have several limitations that make them unsuitable in a number of cases, for which FE models are better suited. First of all, experiments need optimization, which can be performed efficiently using FE models. Secondly, experiments typically inspect the outer surface of the in vitro specimen. Finally, in vitro experiments are ineffective in exploring multiple similar conditions (sensitivity analysis).
A possible paradigm for pre-clinical validation can be summarized as follows (Fig. 1):
- 1)
Preliminary in vitro experiments should be performed on implants with a prototype of the prosthesis to understand which failure scenarios should be expected.
- 2)
Potential hazards must be identified. For each hazard, the probability of occurrence and the risk must be identified using either a top-down Fault Tree Analysis (FTA), or a bottom-up Failure Mode and Effect Analysis (FMEA).
- 3)
To assess the risk of occurrence of each mode of failure, the most appropriate approach must be chosen (either experimental, or numerical). For instance, in vitro experiments are necessary to:
Preliminarily assess the intended implant performance, and explore possible failure modes.
Measure the actual material properties and interface conditions.
Perform tests on specimens that include a real bone, the typical uncertainty related to implantation (interface condition, press-fit), etc.
Conversely, numerical models are advantageous to:
Estimate biomechanical quantities (e.g. state of stress/strain) in regions that are not accessible experimentally.
Explore the effect of design factors (material, surface finish, geometric features, etc), surgical factors (e.g. implant malpositioning) on the outcome.
Predict the post-operative evolution of the implant over time, including progressive failure, tissue adaptation, etc.
Therefore, in vitro experiments and numerical models should be designed concurrently, to enable maximal synergy. The aim of this paper is to illustrate a framework where numerical models and in vitro tests synergistically complement each other (Fig. 2).
