The Role of Virtual Twins in Clinical Trials
By Allison Proffitt
July 13, 2020 | Last summer Dassault Systèmes announced a five-year extension of its ongoing in silico clinical trial experiment with the US Food & Drug Administration to explore how virtual or digital twins might be used in drug and device development.
Steve Levine, the senior director of virtual human modeling at Dassault Systèmes, has been immersed in this world for years. He also believes that right now—in the midst of a pandemic that is impacting everything from engineering to healthcare—there are great opportunities for virtual models.
Dassault Systèmes’ collaboration with FDA was originally announced in 2014, a few months after Dassault Systèmes announced its Living Heart Project, a realistic 3D simulation model of a human heart that Levine championed and leads. 2019’s renewal of the collaboration supports the 21st Century Cures Act, using virtual patients based on computational modeling and simulation to improve efficiency of clinical trials for new device designs.
The team is using the Living Heart Project to create a cohort of virtual patients to test a mock artificial heart valve and submit the data to FDA. Two teams of regulators—one blinded to experiment, one not—will review the dataset. The goal is to see if FDA would be confident taking the virtual patients as a replacement or augmentation to real patients, reducing animal testing or the number of patients required while still ensuring safety and efficacy of the device. The new digital process is intended to be more efficient and less expensive than current norms—where delays and costs can impede patient access to novel treatments—without losing rigor or confidence in a device’s safety and efficacy.
Since the renewal of the collaboration, Dassault Systèmes has met with FDA twice, Levine said, and has assembled several advisory boards to guide them through the process. The clinical advisory board is “helping us to design the models to be representative of the patients and helping us understand what virtual biomarkers will correlate to physical cases,” Levine said. A scientific advisory board is helping build the first-of-its-kind virtual population. Finally, an industry advisory board “will help us with the translation of the process we’re developing with the FDA into something they can use to make real submissions more efficient.”
Dassault Systèmes plans to release all of the data from this experiment to the industry to inform next steps, and Levine hopes this experiment will serve as a playbook for future in silico trials. It’s a promising way to conduct research, he believes.
The Risk of Virtual Patients
Of course, using virtual patient data does pose risks. In a LinkedIn post last year, Craig Lipset, Managing Partner, Clinical Innovation Partners, asked if the real life people providing data for a digital twin will be asked for consent before the digital twin enrolls in a clinical trial.
“It is seeming more and more feasible to envision futures where digital twins are the participants in clinical trials. This future is already on the horizon for the control arm of studies — whether through synthetic controls or placebo data sharing initiatives. Some disrupters may even envision how digital twins fueled with sufficient digital data may even be used to model the active arm of future studies,” Lipset wrote. “But can my digital twin go rogue? Can he join a clinical trial in the future without my ever knowing?”
I asked Levine for his thoughts.
“It’s a fantastic ethical question. This is an area I, personally, struggle with a lot,” Levine said. “One of the things [the digital twin] is best at is helping the body tell its story… All of our bodies are telling us how diseases work, and they have mechanisms to stop those diseases if only we could understand them. So I understand the value of getting access to the data. In the same respect, I also know the risks of losing privacy.”
He said that in the current FDA experiment, the patient model is only of the heart, with little to no identifying characteristics, but it is still based on real patient data. He echoed Lipset’s questions about how we, as a society, manage virtual data. What if a device company wanted to save that model and use it to test the next iteration of its heart valve? And the next? And the next? What consent would be required?
Virtual vs Hypothetical
There is an alternative, Levine pointed out. Instead of being based on real people, virtual patients could be complete hypotheticals, computer-generated to represent the range of human variables.
In the current study, FDA wants real patient data to validate the model. Their belief is, Levine explained, that to show that a device or compound works in a real person, you have to base the data on a real person. However, all clinical trial cohorts are really just a model of a population, Levine argued. “If our goal is to get to the population, do we really need to go through the individual to get there?”
Levine doesn’t think so. “We should not limit ourselves by how the real world limits us. We can’t create a person that represents more than that person, but we can create a model that represents more than one person. Why not take advantage of that? But we have to know enough to be able to do that and be confident in it.”
Virtual or in silico cohorts may also be a tool for introducing much-needed diversity into clinical research. To start, we still need to identify diversity from real patients. “We start from the same, but we need to make sure we make those adaptations to take on the full diversity,” he explained. “But the advantage of doing it with virtual patients is you don’t have to do it every time. Once you understand the diversity, you can build that into the [future, virtual] patient population,” Levine said. “We’re far more alike than we are different.”