Framework for In Silico Clinical Trials to assess the Performance of Medical Devices

The increasing incidence of cardiovascular disease demands that novel medical devices be developed rapidly and within tight budgets, yet be safe and efficacious enough to withstand stringent regulatory scrutiny. Moreover, the expectation of measurable improvement in patient outcomes is driving the need for greater personalization of therapies.

However excessive reliance on physical, animal, and human testing continues to drive development time and cost without necessarily improving device safety and effectiveness. To address these challenges, the Living Heart Project was launched with leaders from academia, industry, clinical practice and regulatory bodies with a mission to revolutionize cardiovascular science through realistic simulation.

As part of this project, In Silico Clinical Trials (ISCT), in which modeling and simulation are used to assess the performance of a medical device in a virtual patient cohort is being explored. ISCT is expected to reduce animal testing and the number of trial participants required while still ensuring safety and efficacy of the device. In this session, we discuss the development and application of a framework for ISCT using an example of a mitral clip device to treat Secondary Mitral Valve Regurgitation (SMVR).

We begin by defining a physics-based human heart mitral valve model representative of an SMVR patient and performing a sensitivity analyses to determine the set of input parameters that most influence clinically relevant metrics. Input parameters are then systematically varied to create a set of patients that are used to train and test a machine learning based surrogate model that predicts the response of SMVR patients. We then use the surrogate model to rapidly generate a large set of potential trial participants for our ISCT. From this set, we then down sample virtual patients to eliminate those displaying non-physiological characteristics or those that would not meet specific inclusion criteria.

Next we use a clustering technique to converge on the final virtual patient cohort that best represents the real world target population. Using this cohort, we then conduct the ISCT to assess device performance.

We conclude by discussing how ISCT of medical devices can help inform clinical trial designs, identify the most relevant patients to study, and support evidence of device safety and effectiveness.