AI models changing clinical trial designs

Clinical trials are expensive, slow, and often restricted by outdated design constraints. In particular, placebo arms create ethical and logistical hurdles, particularly in areas such as rare diseases and oncology.

Digital twins provide a path forward.

They provide patient-specific outcome predictions generated using machine learning models trained with real historical clinical data. These digital twins are created for each trial participant using baseline data, whether assigned to a placebo or treatment arm, to simulate how the individual responded under controlled conditions. Rather than replacing control patients, these predictions of control outcomes are integrated into the analysis, reducing sample size requirements and increasing statistical power.

Ulrearn is one of the few companies that have applied this approach at scale. Steve Herne's CEO has spent more than 25 years in clinical research, taking senior roles at WCG, Bioclinica and Covance. At Ullearn, he focuses on actively using digital twin technology in both early and late exams.

“Today, we are focusing on scaling AI-driven clinical solutions that help sponsors optimize their clinical programs,” Herne says. “Accelerating the timeline across the pipeline and deciding on development.”

What is a digital twin?

In Ailan, digital twins are not virtual avatars or speculative models. They are patient-specific outcome predictions generated using disease-specific machine learning models trained on a large, longitudinal clinical data set. Each digital twin represents a data-driven prediction of how individual participants progressed under a placebo, based on their baseline characteristics.

“We use the term “digital twin” to mean comprehensive predictions of longitudinal clinical outcomes for individual trial participants,” explains Herne. “These are generated using disease-specific machine learning models called digital twin generators or DTGs.”

By incorporating digital twins into statistical analysis of the trial, sponsors can reduce the number of participants required for the control arm without compromising statistical power. This allows for faster and more efficient testing with clear implications for patient burden, recruitment timelines and ethical design.

“By incorporating digital twins into analysis, sponsors can maximize the availability of research,” Herne says. “The power can be used in a variety of ways: to increase confidence in the results of the trial, or to speed up the research by reducing the number of participants needed (particularly the control arms).”

This approach does not replace traditional controls, but rather provides a validated method to complement them and extract larger values ​​from baseline data.

Responding to regulations

The pharmaceutical industry is often open to innovation in principle, but regulatory compliance demands can delay the adoption of new technologies, particularly technologies that challenge traditional testing designs.

Allearn's early collaboration with Regulators helped us navigate this space effectively. The company works closely with agencies such as the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA) to work with evolving guidance on its methodology. In 2022, the EMA officially accredited Illern's Procova method (a specific covariate adjustment approach) for use in phase I and III trials with ongoing outcomes. The FDA also confirmed that Ulrearn's covariate adjustment strategy is consistent with current guidance.

“We take transparency seriously, document every aspect of the model, from how they are trained to how they are validated in a particular context of use,” Herne says.

The emphasis on transparency and scientific verification supports a wider adoption. Illern's digital twins are currently being used in both early and late clinical trials, and adoption continues to grow.

Solve old problems in new ways

Herne is well versed in sustained operational inefficiencies throughout clinical development. One of the most sustained challenges remains a control group that can delay recruitment, extend timelines and present ethical concerns.

“In a randomized controlled trial, digital twins reduce the number of control participants needed while maintaining power,” Herne says. “Alternatively, digital twins can increase trial power without the need to increase sample size.”

In oncology and rare disease studies where traditional control arms are often impractical, digital twins act as simulated control groups in the context of exploration or planning. This approach allows for scientifically effective comparisons without placing additional patients in placebo or standard withholding care, but regulatory decisions are not yet accepted.

“By generating scientifically valid comparisons without the need for external or placebo controls, digital twins expand what is possible in trial designs.”

This is not just an improvement in operation. Future research could reconstruct how control data is generated and used.

AI under the hood

Illern's DTG is built on a proprietary neural network architecture developed specifically for clinical prediction. Unlike more generalized models, these are designed to reflect the complexity and variation of actual patient data.

“These models feature a unique neural network architecture built for clinical prediction,” Herne says. “Their accuracy and granularity reflect a deep investment in designing models optimized for complex, real-world clinical data.”

The model is designed for compatibility with existing trial infrastructures and does not require any changes to endpoints, processing sets, or randomization schemes.

“Integration is easy,” Herne points out.

For sponsors, this creates a clear value proposition. It's meaningful innovation without disrupting current workflows and regulatory expectations.

AI with human purpose

Herne talks about AI in clinical research with a practical focus. Rather than making a radical argument about healthcare transformation, he emphasizes the need for practical tools to improve clinical decision-making and integrate it into current systems.

“Our goal is to make adoption feel familiar. We use proven ways of doing things in smarter, personalized ways that will give you additional confidence in your clinical decision-making.”

This vision is consistent with the core concept of the Digital Twin. Rather than replacing traditional approaches, it's about reflecting them more accurately and improving them.

In landscapes where AI is often exaggerated, Illern's work represents a case study of how measured application of machine learning can improve the efficiency, ethics and outcomes of clinical trials.

Key takeout

• what: Elearn builds digital twins – AI-generated predictions about how patients progress in clinical trials – is used to optimize trial design and execution
• Why is it important?: No more statistical power, fewer patients with placebo arms, faster timelines, and more confident data
• Responding to regulations: EMA qualification and FDA alignment for 2022 has been confirmed
• Use Cases: Randomized trials with control group and single arm studies in rare diseases and oncology
• technology: Disease-specific neural networks trained with extensive clinical data
• Future direction: Integrating existing test designs with workflows enables smarter, patient-centered research.

Many in the industry are cautious about adopting new technologies. Ullearn takes a practical, evidence-based approach to AI to use digital twins to improve testing efficiency, support ethical research design, and produce reliable clinical outcomes.