A New Regulatory Road in Clinical Trials: Digital Twins

Gen Li, PhD, MBA, CEO and Founder, Phesi

In recent decades, the complexity of clinical development has grown considerably. One study found that in the last decade, the average complexity score across all trials has risen by more than 10%—with the biggest complexity increase seen in oncology trials, followed by Crohn’s disease, multiple sclerosis, and stroke.¹

But despite the backdrop of complexity, the pressure to drive down costs of clinical development while increasing successful outcomes has remained consistent for many years, and trial attrition is still a real issue. A mid-year analysis of 66,935 trials published by Phesi in July 2024 finds that the rise in Phase II study attrition first identified in 2022 is continuing. In the first six months of 2024, a third (32%) of trials were terminated during Phase II—a 56% increase on pre-pandemic levels.²

In a bid to mitigate these challenges, sponsors are increasingly interested in how technology can help. That includes how data can be applied to clinical trials by exploring sophisticated predictive analytics and scenario modeling in trial planning, and using data to simulate clinical trials. One area of particular interest is the creation and use of “digital twins” in the control arm of a clinical trial. Digital twins offer huge potential to not only reduce patient burden but accelerate cycle times and lower costs.

However, the era of digital twins opens a new frontier for regulatory agencies. Bodies such as the FDA and the European Medicines Agency (EMA) face the same pressure as biopharmaceutical companies to increase success and decrease costs. Across the industry, everyone is united in wanting more productive and safe trials. But with innovation will inevitably come a need for new regulations and adaptations to existing rules. For sponsors, the key to navigating this evolving regulatory landscape and working with global regulators to implement digital twins effectively is smarter use of data. This will include exploring real-world data and evidence, and generating a method and approach that can be discussed early with regulatory agencies to accelerate adoption.

Improving diversity

Technology breakthroughs in big data and artificial intelligence (AI) have enabled clinical development organizations to gather, tabulate, and interpret large volumes of contextualized real-world patient data. Digital twin analytics is a natural byproduct of accumulating this patient data from placebo arms, active comparator arms, or even failed intervention arms, and other settings. It’s important to note that digital twins do not exist to replace any clinical development processes, per se. At Phesi, we consider digital twins part of an ongoing effort to enhance the current clinical trial toolkit and enrich the development process.

Digital twin technology allows sponsors to “meet the patients” before starting a clinical trial; eliminate costly protocol amendments through better alignment between trial design and the target patient population; and improve commercial viability. The construction of a digital twin begins with a digital patient profile (DPP).³ A DPP is a statistical view of patient attributes—such as demographics, including ethnicity, race, and sex; comorbidities; stage and severity of disease; outcome measures; and concomitant medications. This analysis enables the generation of a patient access score, which allows for precise selection of investigator sites with a clear record of success in accessing the target patient population.

It’s this precise selection of investigator sites that will improve the enrollment of the target patient population, including the targeted diversity of patients. This is critical for sponsors. The FDA recently published guidance to assist companies in creating a diversity action plan for submission with regulatory packages.⁴ The action plan is designed to ensure investigators enroll populations that are representative of the patient who will receive the treatment if the drug or device is approved. Given that a Phesi analysis of US-recruiting cancer trials over the past 15 years found that 42% do not include African-American patients and 48% have no Hispanic American patients,⁵ clinical development organizations must explore how applying digital twins will enable them to satisfy regulations, such as the FDA’s, that are issued to improve diversity. At Phesi, we work with sponsors to define a level playing field for diversity, and to quantify actionable and achievable diversity targets that are agreed by both the FDA and the sponsor.

Use in the control arm

Another important role digital twins are set to play is their potential use in control arms and in simulating the planning, implementation, and outcomes of trials. The outcomes of a clinical trial have three essential components: patient characteristics at baseline, summary of efficacy outcomes, and summary of safety outcomes. The outcomes of a digital twin have the same three components. Unlike a clinical trial, the data to construct a digital twin include data gathered on patients in a healthcare setting, as well as clinical trials data from a database, where patient data was previously collected from patients with clearly defined context. That context includes who collected the data, when the data was collected, where the data was collected, and according to what design the data was collected.

In addition, the scale of data over time afforded by a digital twin approach provides for much deeper analyses. The digital twins that Phesi has constructed across multiple disease areas include many more patients than a single clinical trial arm can possibly include, with much larger geographic coverage, and by many more healthcare providers in more diversified settings. Digital twin-led analysis also covers a longer time frame than a clinical trial is able to, which gives us the opportunity to observe trends in shifting treatment paradigms in a way that is not usually possible in a regulated clinical trial. Just like in “wet” clinical trials, there can be data quality issues in digital twins, yet it is not complicated to identify one or more statistical outliers. What’s more, with a successfully constructed digital twin, these statistical outliers do not impact the statistical validity of the conclusions.

Therefore, a digital twin can be constructed to replace a control arm completely or partially in a clinical trial. Additionally, the digital twin can be used as a historical control to enhance the submission package for regulatory approval. There are circumstances where a placebo control or active comparator arm is not possible, because of ethical reasons or the size of patient population (e.g., in a rare disease setting). In such cases, a digital twin can be considered. Increasingly, a single-arm trial in Phase II with a limited number of patients is allowed by regulators to proceed to Phase III. However, sponsors can increase the chance of Phase III trials being successful if they harness digital twin technology to understand and interpret the outcomes of early phase trials—avoiding costly failures further down the line. A digital twin provides a framework for more accurate interpretation of clinical trial results when there are a small number of patients, helping inform decision-making in the next steps of development.

A new study published in the journal Bone Marrow Transplantation demonstrates the feasibility of AI-powered digital twin solutions to potentially replace the standard-of-care control arm.⁶ A digital twin was created by Phesi for chronic graft versus host disease (cGvHD) in the primary treatment setting. The twin replicated patients receiving prednisone, the current first-line treatment used as the standard of care for cGvHD, which is commonly used in control arms of prospective clinical trials.

^{Regulators’ current positions}

In the US, the FDA has indicated willingness to investigate the use of digital twins. A recent collaboration between the US National Science Foundation, the NIH, and the FDA has sought to explore how digital twins can serve as a “catalyzer” of biomedical innovation.⁷ The collaboration will explore the high potential of digital twins to “revolutionize preclinical and clinical research.” Earlier this year, the FDA also published two position papers exploring AI and machine learning throughout drug development, which includes how digital twins can be applied to model medical interventions.⁸

In Europe, the EMA has also signaled its intent to explore digital twins in clinical development. At the end of 2023 it published a five-year “AI Action Plan,” which includes a commitment to conduct several technical deep dives into specific tools and techniques, including digital twin technology.⁹ Additionally, the EMA places particular emphasis on the value of real-world evidence in clinical decision-making. This emphasis aligns well with the use of real-world data in the construction of digital twins.

There are not necessarily obstacles between digital twins and regulatory approvals, if a sponsor wants to use a digital twin to replace a placebo control arm. The FDA, for example, has made it clear that if sponsors intend to deploy a digital twin/external control arm, they must inform the agency ahead of starting a human trial when an investigational new drug application is filed. Regulators also need to be informed consistently and completely on progress and issues. Sufficient groundwork must be conducted to present compelling evidence to regulatory authorities when a digital twin can be used to replace a placebo and active comparator arm. It requires early collaboration in the development lifecycle among several stakeholders to materialize a digital twin/external control arm.

Finally, it is important to note that a digital twin can yield many internal sponsor decision-making benefits that do not, in fact, require regulatory approval. This includes defining an optimized development pathway; designing a better program and protocols; eliminating the need to make amendments; selecting the best investigator sites with improved precision; recruiting patients faster; collecting better-quality patient data; and interpreting results from a clinical trial in early development more accurately.

We can be certain that regulatory frameworks concerning digital twins will continue to evolve. The shifting sands around regulation may dissuade conservative organizations from committing to the technology—but the potential of digital twin technology is so significant, with the possibility of delivering immediate benefits, that there is no need for sponsors to wait. The future is available today.

Gen Li, PhD, MBA, is CEO and Founder, Phesi

References

1. Markey, N.; Howitt, B.; El-Mansouri, I.; et al. Clinical Trials are Becoming More Complex: A Machine Learning Analysis of Data From Over 16,000 Trials. Sci Rep. 2024. 14, 3,514. https://doi.org/10.1038/s41598-024-53211-z

2. Phesi Mid-Year Global Data Analysis. Phesi. 2024. https://info.phesi.com/2024-midyear-most_studied

3. Digital Patient Profile Catalog Set to Fast-Track Oncology Clinical Development. Phesi blog. May 16, 2023. https://www.phesi.com/digital-patient-profile-catalog-set-to-fast-track-oncology-clinical-development/

4. FDA Guidance Provides New Details on Diversity Action Plans Required for Certain Clinical Studies. FDA press release. June 26, 2024. https://www.fda.gov/news-events/press-announcements/fda-guidance-provides-new-details-diversity-action-plans-required-certain-clinical-studies

5. 48% of US Cancer Clinical Trials Have No Hispanic or Latin American Representation, and 42% do Not Include a Single Black Patient, Phesi Analysis Reveals. Phesi. June 23, 2022. https://www.phesi.com/news/cancer-trials-diversity/

6. Li, G.; Chen, Y.B.; Peachey, J. Construction of a Digital Twin of Chronic Graft vs. Host Disease Patients with Standard of Care. Bone Marrow Transplant. 2024. 59 (9), 1,280-1,285. https://pubmed.ncbi.nlm.nih.gov/38898224/

7. NSF 24-561: Foundations for Digital Twins as Catalyzers of Biomedical Technological Innovation. U.S. National Science Foundation. March 22, 2024. https://new.nsf.gov/funding/opportunities/fdt-biotech-foundations-digital-twins-catalyzers-biomedical/nsf24-561/solicitation?WT_mc_id=USNSF_28&WT_mc_ev=click

8. FDA Releases Two Discussion Papers to Spur Conversation About Artificial Intelligence and Machine Learning in Drug Development and Manufacturing. FDA press release. May 10, 2023. https://www.fda.gov/news-events/fda-voices/fda-releases-two-discussion-papers-spur-conversation-about-artificial-intelligence-and-machine

9. Artificial intelligence Workplan to Guide Use of AI in Medicines Regulation. European Medicines Agency. December 18, 2023. https://www.ema.europa.eu/en/news/artificial-intelligence-workplan-guide-use-ai-medicines-regulation