In silico trials: Verification, validation and uncertainty quantification of predictive models used in the regulatory evaluation of biomedical products

Historically, the evidences of safety and efficacy that companies provide to regulatory agencies as support to the request for marketing authorization of a new medical product have been produced experimentally, either in vitro or in vivo. More recently, regulatory agencies started receiving and accepting evidences obtained in silico, i.e. through modelling and simulation. However, before any method (experimental or computational) can be acceptable for regulatory submission, the method itself must be considered "qualified" by the regulatory agency. This involves the assessment of the overall "credibility" that such a method has in providing specific evidence for a given regulatory procedure. In this paper, we describe a methodological framework for the credibility assessment of computational models built using mechanistic knowledge of physical and chemical phenomena, in addition to available biological and physiological knowledge; these are sometimes referred to as "biophysical" models. Using guiding examples, we explore the definition of the context of use, the risk analysis for the definition of the acceptability thresholds, and the various steps of a comprehensive verification, validation and uncertainty quantification process, to conclude with considerations on the credibility of a prediction for a specific context of use. While this paper does not provide a guideline for the formal qualification process, which only the regulatory agencies can provide, we expect it to help researchers to better appreciate the extent of scrutiny required, which should be considered early on in the development/use of any (new) in silico evidence.