In silico studies involving virtual participants hold great potential in the neurovascular field—and the broader medical industry—with their future role in the generation of scientific evidence likely to be alongside more traditional clinical trials. And, by circumventing some of the ethical and logistical difficulties relating to current in-human and animal studies, they hold the potential to reduce, refine, and complement, these more conventional methods.
This was a key viewpoint presented by Alejandro Frangi, Diamond Jubilee chair in Computational Medicine, Royal Academy of Engineering chair in Emerging Technologies, and scientific director of the Leeds Centre for HealthTech Innovation, at the University of Leeds (Leeds, UK), speaking at the Barts Research and Advanced Interventional Neuroradiology conference (BRAIN 2021; 13–16 December, London, UK).
Frangi began by highlighting the possibility for a “paradigm shift” in the way medical devices, particularly those used in the treatment of cerebral aneurysms, are evaluated. He drew comparisons between the medical sector and the automotive industry, noting that while the latter has introduced “virtual” approaches, such as aerodynamic modelling and crash simulations, in recent years, the former has undergone comparatively little change over the past few decades.
An in silico trial, Frangi told the BRAIN audience, is a computer-based study performed on populations of virtual patients by modelling key features of their anatomy, physics and physiology. Outlining the premise further, he said it is akin to having a library of anatomical models that is near-limitless in size and within which one can conduct virtual implantations—for example, of flow diverters or coiling devices.
Benefits here include the ability to simulate different physiological conditions (e.g. rest versus stress; normotensive versus hypertensive) and test multiple different devices in a single patient—both of which are not feasible with in-human studies, he added. As such, in silico trials also hold the potential to optimise complex interventions that combine several devices or introduce coadjuvant drugs alongside a device. This is in addition to the significant, intrinsic advantage of using a virtual patient population, which is that it carries no risk to humans nor compromises animal welfare.
Frangi then moved on to discuss the FD-PASS (Flow diverter performance assessment) study, the findings of which were published in Nature Communications in 2021. This saw comparisons drawn between FD-PASS—an in silico trial assessing outcomes with the Pipeline embolisation device (PED; Medtronic)—and three conventional studies (ASPIRe, PREMIER and PUFS) also involving this technology.
In addition to predicting the effectiveness of PED successfully, by producing similar angiographic aneurysm occlusion rates in its virtual, 82-patient cohort to those seen in the conventional studies, FD-PASS was able to go a step further by identifying risk factors of PED failure—including the presence of branch arteries, larger aneurysm sizes, and hypertension—and even explaining the findings of in-human clinical trials thanks to advanced computational modelling techniques.
Expanding on this second point, Frangi said FD-PASS produced data indicating that—following flow diversion—haemorrhagic stroke caused by a delayed rupture is more likely to occur in patients with complex-shaped aneurysms that have a high aspect ratio or are large/giant in size, while an ischaemic stroke resulting from a posterior communicating artery (PCoA) aneurysm is more likely to be observed in hypertensive patients with a distal side branch.
Some of these insights would be “very difficult” to demonstrate or hypothesise on with a conventional trial design, he claimed, adding that identification of potential biomarkers for device failure is therefore another possible avenue opened up by in silico models.
Responding to a query from BRAIN course director Paul Bhogal (Barts Health NHS Trust, London, UK), Frangi said: “I think, potentially, we will continue to use conventional trials as confirmatory studies for hypotheses or findings generated in silico over large-scale, virtual patient datasets, rather than them being the first port of call for human evidence on novel designs that have come straight from the bench and animal experiments.”
Commenting on the shortened timeframes that in silico trials can enable, Frangi added that FD-PASS took place over the course of three months, while the ASPIRe, PREMIER and PUFS studies each took between six and eight years to complete—with the former also requiring significantly less funding. And, he continued, while prior development and validation of the computational models used in FD-PASS did take several years, those models can now be used for many similar neurovascular devices—essentially, anything “stent-like”—in the future.