From Data to Decisions: Evaluating Bayesian Dynamic Borrowing Before Your Trial Begins

Clinical trials continue to face significant operational and scientific challenges. Slow enrollment, small patient populations, high placebo response rates, and increasing pressure to accelerate development timelines can make it difficult to generate robust evidence efficiently. These challenges are particularly acute in therapeutic areas such as rare diseases and CNS disorders, where patient recruitment is often limited and trial variability can be substantial. Historical data may offer an opportunity to address some of these issues, but only when used appropriately.

What Is Bayesian Dynamic Borrowing?

Bayesian dynamic borrowing is a statistical approach that allows investigators to incorporate relevant historical information into the analysis of an ongoing clinical trial. Rather than treating historical data as either fully relevant or completely irrelevant, dynamic borrowing evaluates the similarity between historical and current study populations and adjusts the amount of borrowing accordingly. When the data appears comparable, more information is borrowed. When substantial differences exist, borrowing is reduced.

This adaptive approach helps organizations leverage valuable prior knowledge while reducing the risk of drawing conclusions from data that may no longer reflect the current study population.

What Types of Data Can Be Used?

Potential sources of historical information extend well beyond previous sponsor studies. Organizations may be able to leverage data from:

• Previous clinical trials
• ClinicalTrials.gov datasets
• Natural history studies
• Epidemiological studies
• Real-world data sources

Each source presents unique opportunities and challenges. The critical question is whether the historical population is sufficiently comparable to the current study population to justify borrowing.

Where Bayesian Dynamic Borrowing Can Add Value

When historical information is relevant and comparable, Bayesian dynamic borrowing can help sponsors make more efficient use of existing evidence while maintaining scientific rigor. By dynamically adjusting the amount of information borrowed from historical data based on its similarity to the current trial population, Bayesian dynamic borrowing provides a flexible framework for enhancing study performance without blindly relying on past results.

Greater Statistical Power

One of the primary advantages of Bayesian dynamic borrowing is its ability to improve statistical power. By incorporating relevant historical information, researchers can increase the precision of treatment effect estimates and reduce uncertainty in study outcomes. This added precision can make it easier to detect meaningful treatment effects, particularly in studies where patient numbers are limited or endpoint variability is high.

For sponsors, greater power can translate into a higher probability of demonstrating efficacy without increasing trial size. In competitive development environments, this can mean stronger evidence generation while making more efficient use of available resources.

Reduced Sample Size Requirements

Recruiting patients is often one of the most significant challenges in clinical development. This is especially true in rare diseases, pediatric indications, and certain CNS conditions where eligible participants may be difficult to identify and enroll.

When suitable historical data are available, Bayesian dynamic borrowing can potentially reduce the number of patients required in a control arm while still maintaining robust statistical performance. Fewer patients can lead to faster enrollment, reduced operational burden, and lower overall trial costs. For programs operating under tight timelines or budget constraints, these efficiencies can have a meaningful impact on development strategy.

Ethical Advantages

Reducing the need for large concurrent control groups may also offer important ethical benefits. In many therapeutic areas, patients and investigators may be reluctant to participate in studies with a high likelihood of receiving placebo or standard of care when promising investigational options exist.

By leveraging historical information where appropriate, sponsors may be able to reduce the number of participants assigned to less desirable treatment groups while still generating credible evidence. This can be particularly valuable in serious, progressive, or life-threatening diseases where patient access to potentially beneficial therapies is a significant consideration.

Better Decision-Making

Beyond improving individual study metrics, Bayesian dynamic borrowing supports more informed decision-making throughout the development process. Instead of viewing historical evidence as either fully relevant or entirely unusable, sponsors can quantitatively assess how prior information should influence current analyses.

This approach allows teams to better understand the tradeoffs between efficiency and risk, evaluate alternative design strategies, and make evidence-based decisions before a study begins. The result is greater confidence that the chosen design aligns with both scientific objectives and program goals.

The Challenge: Knowing When to Borrow

Despite its advantages, Bayesian dynamic borrowing is not without risks. Excessive reliance on historical data can increase the risk of Type I error if the historical data are not sufficiently representative of the current study population. Additionally, regulators may require strong justification for the assumptions used in borrowing strategies.

The central challenge is determining where the tipping point lies. At what level of similarity does historical data improve study performance, and at what point does it begin to introduce unacceptable risk?

Why Simulation Matters

This is where simulation-based design evaluation becomes valuable. Tools such as KerusCloud^® allow study teams to evaluate borrowing strategies before a trial begins by creating realistic virtual patient populations and assessing thousands of potential design scenarios.  Rather than relying solely on theoretical assumptions, teams can quantify how borrowing affects key operating characteristics such as:

• Power
• Type I error
• Sensitivity to placebo response
• Impact of population drift
• Probability of study success

A Realistic Example

How do different analytical approaches perform when historical information suggests a control outcome differs from current assumptions? In a simulated example, a traditional frequentist approach achieved 72.0% power with a Type I error rate of 2.8%, while a pure Bayesian approach increased power to 85.9% but also increased Type I error to 6.5%. The Bayesian dynamic borrowing approach achieved a middle ground, delivering 81.2% power while maintaining a Type I error rate of 5.0%.

The lesson is not that dynamic borrowing is always superior. Rather, performance depends heavily on the quality, relevance, and comparability of the historical data being incorporated.

Therapeutic Areas Poised to Benefit

Bayesian dynamic borrowing may be especially valuable in settings where patient populations are limited, recruitment is difficult, or existing historical data are abundant.

For sponsors pursuing innovative development programs in these settings, the ability to quantitatively evaluate borrowing strategies before initiating a study could provide a significant advantage.

Turning Historical Data into Smarter Trial Design Decisions

Historical data represent an increasingly valuable asset in modern drug development. However, the question is no longer whether sponsors should consider using these data. The more important question is how to determine when historical information is trustworthy enough to improve study performance without compromising scientific rigor.

Bayesian dynamic borrowing provides a framework for striking that balance, while simulation platforms such as KerusCloud allow teams to evaluate opportunities and risks before a study begins. By combining advanced statistical methodologies with large-scale simulation, sponsors can make more informed trial design decisions and increase confidence in their development strategies.