Projected to be worth over $38 billion in the global healthcare market by 2032, AI simulations have the potential to streamline clinical trials and help address inequities in underserved patient populations.
A glowing, 3D organ with pulsating veins rotates on a flashing dashboard. This generative-trained AI simulation, known as a digital twin, is a virtual replica of a person. It possesses the unusual ability to mimic biological complexities of humans with meticulous precision—a powerful tool when treating patients, for example, by running simulations to forecast outcomes of different treatment regimens. And in the research realm, experts say digital twins could help streamline clinical trials and ease racial and ethnic data disparities.
Worth $1.17 billion in 2022, the global healthcare digital twin market is predicted to exceed $38 billion by 2032, according to a 2024 report by Towards Healthcare. As digital twins model patient outcomes, they generate synthetic or artificial data that is often used to supplement real-world data, said Rashidi Hooman, the new associate dean of AI in Medicine at the University of Pittsburgh School of Medicine. And this, he continued, is beginning to alter the life sciences landscape beyond clinical practice. Digital twins can also be used directly in clinical trials, predicting outcomes for participants had they been in different treatment groups.
“We’re going to have an ‘AI tsunami,’” Hooman told BioSpace in an email. “We haven’t even seen a glimpse of it yet. It will hit in three to five years.”
“[Digital twins] are now starting to address our various needs (e.g. drug development or biomarker discovery) and helping to expedite our research and development domains,” he continued.
Pharma Adopts Digital Twins
One early adopter of digital twins in the pharmaceutical space is Bayer. Sai Jasti, head of data science and artificial intelligence at Bayer Research and Development, told BioSpace in an email that he sees digital twins as a “transformational tool” in the pharma giant’s arsenal of resources that it hopes will drive innovation in drug development.
“Leveraging [digital twins] allows us to improve precision in clinical decision-making and streamline trial processes,” he said. “This data enhances our understanding of patient responses and treatment efficacy, providing a more comprehensive view that is often lacking in traditional methodologies. By employing these models, we aim to improve patient stratification and enhance the overall efficiency of our trials.”
In 2023, Bayer and AstraZeneca entered into a collaboration with Toronto-based AI outfit Altis Labs to help accelerate and improve cancer trials using AI-generated digital twins. As part of the deal, the companies gained early access to test and implement Altis’ digital twin technology.
Started by Felix Baldauf-Lenschen in 2019, Altis intends to advance precision medicine by partnering with healthcare systems to train and validate its AI models on historical de-identified patient data. The company trains its models on historical real-world data to predict standard-of-care patient outcomes “so that in a clinical trial, we can predict what a patient’s outcome would be if they got the standard-of-care treatment,” he told BioSpace.
Altis’ approach was on display at the 2024 European Society for Medical Oncology (ESMO) conference, where a poster presentation highlighted its prognostic AI models in non-small cell lung cancer. According to the poster, Altis’ AI model predicted overall survival and showed potential to enhance treatment effect quantification beyond tumor size measurements that have historically been used as the basis for evaluating treatment response and markers used in regulatory approvals.
Alleviating Data Disparities
Digital twins also have the potential to help address data disparities among underrepresented patient groups, Baldauf-Lenschen said.
The FDA recently issued draft guidance intended to standardize the collection and reporting of race and ethnicity data in clinical trials. AI and digital twins could help with these efforts, Baldauf-Lenschen added.
Subpopulations such as pregnant women, people of color and children are “extremely underrepresented . . . in terms of evidence of treatments being evaluated in clinical trials, and that’s where there is a really interesting opportunity to start to use AI to enhance evidence generation,” he said.
Differences in patient demographics between clinical trial populations and real-world populations can result in different efficacy profiles, Baldauf-Lenschen explained. For example, there may be a more muted treatment effect in older, sicker patients compared to what was observed in a clinical trial population, he said.
One area in which digital-twin models could improve AI performance is in people of color who are battling skin cancer, according to the Melanoma Research Alliance. While people of color are diagnosed with melanoma less often, they are up to four times more likely to be diagnosed with advanced melanoma and 1.5 times more likely to die from the disease.
A study published in JMIR Dermatology in 2022 noted a “white lens phenomenon,” leading to the underrepresentation of dark skin pathology images in dermatology resources, which the authors noted has disadvantaged people of color by having AI diagnostic systems trained with light skin color images. They added that AI has the potential to fix this problem.
“Deep learning approaches can generate realistic skin lesion images that improve the skin color diversity of dermatology atlases. The diversified image bank, utilized herein to train a [convolutional neural network], demonstrates the potential of developing generalizable artificial intelligence skin cancer diagnosis applications,” the authors wrote.
Challenges in the Twinning Space
The experts who spoke with BioSpace all acknowledged, however, that there are myriad challenges that still need to be addressed.
Hooman said that while he sees vast innovation in the digital twins space, the technology is not infallible.
“None of these frameworks are perfect. They may seem magical but it’s not magic. They’re following basic statistical parameters and predictive analysis,” he said. “There is an error component built in regardless of which AI you’re using. That being said, knowing what they can do, we know a lot of their limitations as well.”
One key challenge, Jasti said, is “navigating regulatory requirements, which necessitates early engagement with authorities to ensure compliance.” A second is the quality of historically collected data, which “is vital,” especially in terms of image annotation. Scientific feedback and collaboration help Bayer to navigate these issues, he said.
Another challenge is the sheer quantity of data collected. Digital twins require an “excessive amount of patient data” to create representations that are accurate, according to a report by Cromos Pharma. “The complexity and volume of the data needed can be a significant barrier.”
Baldauf-Lenschen agreed, noting that it becomes more difficult to train AI models when dealing with smaller subsets of populations, such as underrepresented patient groups.
Hooman said that in healthcare, when AI makes a mistake, there can be deadly consequences. “In our world, things are a lot more sensitive. If AI makes a recommendation on Amazon . . . versus if AI makes the wrong recommendation for your chemo treatment, that could hurt or kill somebody.”
Nonetheless, he indicated he is hopeful. “I believe the amount of positives will be so much greater than the negatives,” he said.
