On Wednesday 4 July, attendees joined an online event organised by the Leeds-Bradford local group. The speaker was Professor Alex Frangi, who spoke about in silico trials – where simulation meets trial design to test what couldn't otherwise be tested.
Prof Frangi’s talk focused on medical devices but in silico trials are applicable to medications. The justification is that 90% of the medical-device market in the UK is made of small- to medium-sized enterprises, SMEs. Traditional trials are costly and take a lot of time, which might make important research out of reach for SMEs.
The medical sector might be able to learn from other industries. The automotive industry have used simulation for many years to quickly learn what might and might not work to improve products safely. Work in the medical domain has already been done, for example, to model the design and development of stents for aneurysms. With the right inputs to represent physiological variability, it is possible to simulate clinical trials.
But what are in silico trials? Simply, they are the use of computers to simulate trials on virtual, chimera participants. A chimera is a mythical animal made of a mixture of features. In silico trials can make virtual chimera patients by combining marginal probabilities of features from various sources of information, under assumptions and methods to handle independence. It differs from Digital Twin studies in the following way: Digital Twin trials make a digital twin of individuals and subsequently sample from the digitised population; In silico trials make a digital twin of each individual and subsequently combine all features to inform a chimera population generator. The result is that the participants in an in silico trial are representative of the population while not being a member of the real-world input participants.
Professor Frangi spoke about some of his recent work with colleagues, which was published in Nature. The project took three months to collect data but eight years to develop the models for in silico trials. That sounds like a long time but the models can run as many trials as needed. This can be much quicker and cheaper than conducting many real-world trials. Key results included predicting efficacy of the medical device intervention, explanations for device failures scenarios, and identification of potential biomarkers for device failure. These insights were offered partly because the in silico approach permits retrospective, deeper dives into the virtual patients – something not possible with traditional trials once participants have participated.
The audience asked questions about buy-in from regluators, the best applications for in silico trials, and about how in silico trials will fit with existing methods to improve our knowledge. Prof Frangi spoke passionately about the potential about the involvement of statisticians in developing in silico trial methods. Particularly, in the model development and joint modelling steps required to produce the generator of virtual, chimera patients. The Leeds-Bradford group will no doubt be inviting further discussion about in silico trials with more speakers to see how the Royal Statistical Society can contribute to this approach.
A recording of the talk will be available on the Leeds-Bradford local group playlist within the RSS’s YouTube channel, shortly.
Ciarán McInerney, PhD, is secretary of the Leeds-Bradford local group. He is a research fellow in the School of Computing at University of Leeds and the NIHR Yorkshire & Humber Patient Safety Translational Research Centre, where he studies the design and evaluation of digital innovation for patient safety.