In Silico Trials
In silico modelling and digital twins enable trailblazing virtual clinical trial
Clinical trials are not only expensive and time-consuming, they also compromise animal welfare and potentially put human volunteers at risk. A team of researchers is working on computer modelling that provides a viable alternative – a virtual (aka in silico) clinical trial. “Medical devices follow a lengthy evaluation through bench to animal experiments, and then through human trials, with a tiny amount of scientific evidence currently derived from computer modelling and simulation. The cost is ever-increasing, delaying lifesaving benefits to patients, and they are not without fundamental limitations. Yet, they are essentially a trial-and-error approach to generating scientific evidence on medical device safety and efficacy,” says Prof. Alex Frangi, Director of Research & Innovation at LIDA. A cardiovascular device trial, for example, can cost £80m and spend up to ten years in Research and Development. Prof. Frangi explains researchers can “never fully know how results from such trials will translate when the device hits the market and is applied to the broader population.”
The technology underpinning medical implants has also grown more complex with time, meaning it can also go wrong in more ways. As devices are generally more effective, current guidelines offer therapy, and these devices can now be put into younger people. In the past, they were only deployed in life-or-death situations, where gaining just a few more years of life was an achievement. Someone who now receives an implant at 45 years old could live with it for twenty or more years. Even if running a clinical trial for that length of time was possible or affordable, we would only know the benefits of the technology too late when technology would have considerably moved on.
One approach to addressing these challenges is to learn from other sectors like aerospace or automotive where innovations make substantive use of computational modelling and simulation well before resourcing to experiments or prototypes. In silico trials use digital medical imaging and complex computational modelling of the mechanisms underpinning the physics, physiology and anatomy of patients, devices, and interventions to create many patient-specific simulations. These trials rely on virtual populations and interventions simulated using patient-specific models. As a result, Prof. Frangi says, “We can test devices faster, more thoroughly, and generate better quality scientific evidence at lower cost.”
This is the most comprehensive such virtual trial yet attempted, with the aim to produce unprecedented scientific evidence of device performance and safety. In silico trials will not fully replace but instead refine, reduce or complement other scientific evidence approaches. This could reduce the time spent in clinical trials from years to a few days, an approach already used in other industries such as aerospace and automotive – where they rarely now use physical wind tunnels or build physical prototypes. These sectors not only have shown step changes into the quality of their final products but also in the precision and costs of their design and manufacturing processes, helping to achieve better quality products, faster product lifecycles, and contributing to net zero with less waste.
Researchers at the University of Leeds’ CISTIB (Centre for Computational Imaging and simulation technologies in Biomedicine) were sponsored by the Royal Academy of Engineering to collaborate with an international interdisciplinary team of data scientists, clinicians, and engineers from Oxford University, KU Leuven and Radbound University completing one of the most complex in silico trials to date. They used complex data processing workflows involving artificial intelligence and biomechanical modelling to explore virtual treatments of brain aneurysms. For the first time, an in silico trial of a medical device was compared to determine whether this approach could replicate the outcomes of conventional clinical trials. The team also examined whether this approach could elucidate novel insights into the performance of medical devices using virtual experimentation that would be otherwise unknown from standard clinical testing.
The team of researchers looked at stent-like devices called intracranial flow diverters (FD), used to reduce blood flow into the aneurysm (a bulge in the arterial wall, which, if burst, causes haemorrhagic stroke). By inserting a FD, the aneurysm is cut off from the circulation, eventually eradicated reducing the rupture risk. The flow diverter performance assessment (FD-PASS) in silico trial used AI-enabled anatomical vascular reconstruction from medical imagery, biomechanics-based blood flow and clot modelling, alongside with a virtual stenting procedure to look at treatment of intracranial aneurysms with a flow-diverting stent in 164 virtual patients with 82 distinct anatomies each in normotensive and hypertensive regimes. Computational fluid dynamics allowed to quantify post-treatment flow reduction.
The results of this in silico trial predicted flow-diversion success rates replicating the values previously reported in three conventional clinical trials (ASPIRE, PREMIER and PUFS) that each took eight years from design to publication at a cost of £20-40m. “The complex modelling FD-PASS project showed we could predict radiological occlusion rates of the cerebral aneurysms to within five per cent of the results of clinical trials yet at a fraction of their costs,” says Prof. Frangi. “The in silico methodology meant a broader and more thorough investigation could be conducted into associated factors linked to device performance and root causes of their failure, which would be challenging if not impossible through a conventional clinical trial.”
By performing virtual experiments and sub-group analyses, the team could also new insights into the heterogeneity of device performance and treatment failure, e.g., in the presence of hypertension and vascular side-branches. This research helped to identify and model new biomarkers, which, as Prof. Frangi explains, means, “we can now identify patients who may be good responders to the therapy.”
The team is now looking to model more subtle or long-term effects of the therapy, novel device designs, and larger virtual populations. Prof Frangi highlighted that “collaboration with the involvement of industry, regulators, health economists, and health decision makers, and policy makers is critical to generate evidence and momentum that will accelerate the adoption of these technologies in industry and regulatory processes”. The US Federal Drug Administration (FDA) already accepts in silico evidence as complementary evidence in regulatory dossiers. The European Medicines Agency is showing signs of heading towards the same, and there are similar initiatives in China. The independent report by the Taskforce on Innovation, Growth and Regulatory Reform (TIGR) states the need for modernising the UK regulatory system and its transformation into a driver of economic growth and industrial innovation in the post-pandemic era. The UK MedTech sector, primarily comprised of SMEs, and the Medicines and Healthcare products Regulatory Agency (MHRA) could benefit enormously from the implementation of the recommendations in this report and consideration of scientific evidence from in silico trials. Thanks to a digitally mature and well-connected healthcare research data ecosystem, the UK offers a unique opportunity to attract international investment. Hence, the potential is there for pioneering the creation of virtual populations at scale from real-world NHS data. This could be achieved through a digital data donation programme akin to organ donation that preserving the benefits of data privacy and security, makes the benefits derived from the acceleration promised by in silico trials available to patients in the UK and beyond.
Professor Alex Frangi