Traditional clinical trials are time-consuming, expensive and extremely difficult to run. They require massive budgets, take years to complete and are bogged down with difficulties, from finding the correct patient population to ensuring successful outcomes for those patients.
There’s an emerging solution to these issues, and it’s no surprise that it’s powered by artificial intelligence. Computer-based in silico clinical trials, also called virtual clinical trials, offer a launching point for human clinical trials to be conducted with higher rates of precision and success, and they offer a valuable tool in healthcare technology that can bring drugs to market faster and save more lives.
What’s the problem with traditional clinical trials?
Traditional clinical trials suffer from a number of difficulties.
“You have to recognize that more than 80% of treatments that go to clinical trials are unsuccessful,” explained Steve Levine, senior director of the Virtual Human Modelling program at Dassault Systèmes. It’s not hard to see why – nearly every aspect of running a traditional clinical trial hosts an opportunity for failure.
First, it’s incredibly difficult to create patient cohorts that are demographically representative.
A study of clinical trials completed in the US in 2020 indicated that 8% of participants self-identified as Black; in reality, the Black community comprises 14% of the US population. In 2023, researchers found that just 13% of participants in a clinical trial for a cancer drug were from rural areas; rural populations actually make up closer to 20% of the country’s populace. Beyond simply being able to identify representative cohorts, clinical trials often struggle with ensuring end-to-end patient participation. Participant drop-out rates can reach 30%, according to some estimates, and this figure doesn’t include the percentage of patients who
The differences in these numbers may appear small, but consider that environmental, social, racial and ethnic factors significantly impact one’s health. The inability to accurately include patients in clinical trials means that clinicians lack a complete understanding of the efficacy of a drug or medical device on patients. While their understanding is informed by the data collected in a study, that data doesn’t accurately depict reality.
Second, traditional clinical trials can be slow, often taking more than a decade to complete. Patient recruitment can take months or years. Supply chain issues can bog down a lab’s ability to move forward on development and research projects. Low rates of success in the clinical phases send trials back to the proverbial drawing board, stretching their timelines further. For patients, the years spent waiting for drugs, therapies and devices to be brought to market can be excruciating, and in some cases, a matter of life and death.
Last, traditional trials are expensive. Estimates from a 2024 study from the United States Department of Health and Human Services put the cost per trial at $117 million. Other figures suggest the number is much higher, reaching upward of $2 billion. In Europe, those costs are slightly lower, but in general, the total expenditure for trials tends to hover around $100 million. These figures include the costs of patient recruitment, site monitoring, inspections and investigations, salaries for physicians, researchers and nurses, equipment costs, fees for data collection and much, much more. And the costs grow higher each year.
There is a solution to these obstacles.
AI’s solution to the patient procurement problem
Compared to traditional trials, in silico trials are fast, inexpensive and inclusive.
They begin with virtual twins.
First, clinicians identify a small cohort of patients for inclusion in a study, say between 10 and 20 individuals. They map out virtual twins of these individuals, creating 3D replicas of their unique anatomy and physiology. They don’t just look like the patients, they respond like them as well. Then, using machine learning, they’re able to train an artificial intelligence model to extrapolate on these twins to make hundreds or thousands of synthetic virtual twins – twins that are representative of real-life populations but which aren’t modeled after real-life patients, overcoming recruitment barriers. Using these, clinicians can run a computer-simulated in silico clinical trial.
“Despite being an emerging technology, synthetic datasets can enhance clinical research,” found researchers Mauro Guiffre and Dennis Shung in a 2023 study on the power of synthetic data in healthcare.
This synthetic data is both useful and safe, and it lowers the need for procuring — and keeping — hundreds or thousands of actual trial participants. If properly built, these virtual twins can mimic how a real human body would react to different treatments. AI can spot potential safety issues or ineffective treatments before human testing begins and ensure that only the best and safest drugs — those that pass virtual trials — are then passed on to traditional human trials.
Using AI and virtual twins, researchers can design clinical trials that solve the pressing issue of creating viable patient cohorts that not only represent actual populations but can reliably complete the trial from start to finish. Working similarly to a large language model, these virtual twins use scientific reasoning and knowledge to generate realistic behavior, much like how the LLMs that power ChatGPT or Grok generate text based on massive inputs of human language.
How AI can accelerate time in clinical trials
In silico clinical trials offer a variety of advantages over traditional trials. In addition to creating “ideal” patient populations, virtual trials also enable what was previously impossible to achieve. The AI can power them with artificial patients and provide insights not available in traditional settings: the long-term effects of drugs and treatments. With a virtual trial, you can bring the future to the present.
“Consider a neurodegenerative disease — it might take 10 or 20 years for that disease to progress, but you’re not going to run a clinical trial for 20 years,” explained Levine. Using clinical data alone, it’s virtually impossible to predict the impact of a drug or treatment 20 years in the future if you only have two years of data.
“But if you run a clinical trial for two years and say, ‘OK, our models match what we’ve seen in the first two years, providing a solid scientific foundation to trust our accelerated time models,’ we can then predict how the disease will progress over time on a given patient or a population. Once properly validated, these models can accelerate time so that we can do 20 years in 20 minutes,” he explained. From there, the models get refined as additional clinical data is collected, so future treatments benefit from more reliable and diverse virtual twin patients.
That’s the power of AI.
How virtual twins enable cost-effective trials
It’s unsurprising that virtual modeling of human patients and computer-simulated trials offer a reduction in cost to traditional methods. The same phenomenon can be observed across industries: vehicle design and production costs can be lowered with virtual testing, entire factories can be run at reduced cost and sustainable supply chains can be managed with a raw materials savings of $131 billion. Healthcare is no exception.
One study on the cost-effectiveness of virtual trials “observed that in silico clinical trials are almost 90 times cheaper than real clinical trials.” Clinicians have been speculating for years that computer-simulated clinical trials can offer a significant reduction in the expenditure per patient and drug.
“There is a real opportunity because in silico trials can be done, basically, based on the huge amount of data that clinical and research centers around the world have been accumulating, so we can cut a lot the cost of recruiting new patients, especially at the early stages,” suggested Claudio Capelli, a research fellow at University College London, at the 10th International Virtual Human Twin Experience Symposium.
Virtual trials already lower the barrier to entry by presenting a less expensive alternative to their traditional counterparts. However, considering that pre-existing data can be leveraged for future trials makes this method even more appealing. It allows for the proliferation of innovation in the medical field that has been historically bogged down by high costs.
Today’s clinical applications: In silico in use
While a lot of this might sound speculative, in silico clinical trials and the use of virtual twins in the medical field are already happening. The demand for efficiency and precision in clinical trials and drug development has never been greater, and virtual twins enable a greater level of precision than previously thought possible.
Striving for that understanding is what influenced the founding of the Living Heart Project, an initiative spearheaded by Levine that aims to create virtual twins of individual patients’ human hearts. The project was launched in support of the FDA’s 21st Century Cures Act, which aims to accelerate the development of medical devices through innovative methods like in silico trials.
The opportunity for efficiency and precision in clinical trials and drug development is greater now than it’s been before, in large part due to technological advancements like the proliferation of AI and the more widespread adoption of virtual twins in the healthcare industry. The demand has also never been greater, as the timeline and cost of trials have risen steadily in recent years, suggesting bottlenecks and inefficiencies across the board. Since 2019, Dassault Systèmes has collaborated with the FDA to explore how in silico clinical trials can be developed and deployed to accelerate the evaluation of medical devices. After five years of cooperative research, including with clinicians, industry professionals and researchers, they published a playbook with their findings, making public a guide for others in the healthcare industry to move forward in implementing credible, successful virtual trials.
The precision that virtual trials can offer combats the resource-intensive, time-consuming nature of traditional trials that often experience low rates of success.
Heading toward a virtual future
It’s quite possible, given all the advantages that virtual trials offer, that they’ll become the standard for the industry.
The FDA’s endorsement of alternatives to traditional clinical trials, including through in silico methods, symbolizes a stamp of approval that indicates the viability of virtual efforts. Successful in silico trials, like one in 2017 that explored antipsychotic drugs for Alzheimer’s patients using computational models, signal a willingness from within the medical research community to adopt technology for these purposes. In fact, cooperation between technology companies, regulatory bodies and research facilities highlights a general openness to introducing innovative alternatives to approaches we can perhaps call antiquated. In addition to Dassault Systèmes’ collaboration with the FDA, the company has also worked with 13 French research hospitals to launch the Living Brain project in 2019, which aims to create virtual twins of patients’ brains to enable more accurate neurosurgical outcomes.
Levine believes that over the next few decades, human testing with in silico trials will overtake traditional ones, and it’s easy to see why. A virtual trial does what a traditional one can’t: it creates a representative cohort, enables swift experimentation that doesn’t harm patients, speeds up the approach to regulation and provides a system that allows for long-term insights to be gleaned. Even in a traditional-style trial, virtual twins can become the control group, eliminating the costs and ethical challenges of the placebo group. By leveraging AI and virtual twins, the timeline for trials can be significantly shortened, and the budget needed to create and run them can be reduced. By all accounts, in silico trials are a necessary “missing link” for progress in the medical field.
This blog is part of a series on Dassault Systèmes’ 10th International Virtual Twin Human Experience Symposium.
Watch the symposium, learn more about out Virtual Human Twin Initiatives, and keep an eye out for the next blog in this series.
Shoshana is a Senior Communications Manager on the Corporate Publishing team here at Dassault Systèmes. With experience in journalism, marketing, and PR, her work has been published around the world and across the internet.