Rethinking oversight as clinical trials become more data-driven and algorithm-informed
As artificial intelligence (AI) becomes more embedded in clinical trial design and execution, a natural question is emerging: Do data safety monitoring boards (DSMBs) need AI expertise?
It’s a fair question. AI is increasingly used to support patient selection, optimize trial design, and generate predictive insights from large, complex datasets. In adaptive trials, AI-enabled tools may influence randomization strategies, simulate potential outcomes, and help identify emerging safety or efficacy signals earlier than traditional approaches.
As more of the trial ecosystem becomes data-driven—and, in some cases, algorithm-informed—it’s reasonable to ask whether DSMBs should evolve in parallel by adding AI specialists to their membership.
The answer is more nuanced than it first appears.
What DSMBs are designed to do
To understand whether DSMBs need AI expertise, it helps to revisit their core purpose.
DSMBs are independent groups of experts responsible for reviewing accumulating clinical trial data, safeguarding participant safety, and preserving study integrity. As data emerge, DSMBs evaluate the evolving risk-benefit profile and make recommendations about whether a trial should continue as planned, be modified, or stop early.
This responsibility has always required a careful balance of clinical insight, statistical understanding, and independence from those conducting the trial. DSMBs are not designed to build models or develop analytical tools. Their role is to evaluate the evidence in front of them and determine whether it supports continued study conduct.
That distinction becomes especially important as new analytical approaches, including AI, enter the picture.
Where AI fits into the modern trial
AI is not replacing traditional statistical methods, but it is expanding what is possible.
Today, AI may be used to identify patterns in large datasets, predict patient outcomes, or inform aspects of trial design. In some adaptive trials, AI can support simulations or contribute to decision-making frameworks that adjust trial parameters based on accumulating data.
As a result, DSMBs may find themselves reviewing outputs that extend beyond traditional tables, listings, and graphs. In some cases, they may be asked to consider model-informed insights—analyses that depend not only on observed data, but also on underlying assumptions, training datasets, and algorithm design.
This does not fundamentally change the DSMB’s role, but it does change the context in which that role is carried out.
The real question: Expertise or understanding?
Rather than asking whether DSMBs need AI experts, it may be more useful to ask whether DSMB members need to understand how AI is being used within a given trial.
Expertise or understanding? The answer is both.
DSMB members must be able to interpret the information presented to them, understand its limitations, and assess whether the conclusions drawn from it are justified. In that sense, their responsibility remains consistent with what it has always been. The difference is that some of the analyses they review may now be more complex or less transparent than traditional statistical outputs.
This does not necessarily mean that every DSMB requires a dedicated AI specialist. It does mean the outputs provided to the DSMB must be clear, contextualized, and supported by appropriate explanation. Without that clarity, even sophisticated analyses risk being misunderstood or misapplied.
The critical role of the DSMB statistician
If any role becomes more central in AI-enabled trials, it is the DSMB statistician.
The unblinded DSMB statistician is responsible for preparing and validating the data the board reviews, ensuring analyses are conducted appropriately and aligned with the study’s predefined statistical framework.
As AI-driven methods are introduced, this responsibility may expand to include translating model outputs into interpretable summaries and helping the DSMB understand the assumptions and limitations underlying those outputs. In practice, the statistician often serves as the bridge between complex analytical methods and the board’s decision-making process.
That bridging function becomes even more important as the analytical tools themselves become more sophisticated.
For organizations that administer DSMBs, this reinforces the importance of matching each board with statisticians who have the right expertise for the study design, data profile, and oversight needs.
New risks require careful oversight
The introduction of AI into clinical trials does not eliminate risk. In many cases, it shifts the nature of the risk.
Models are only as reliable as the data used to train them. If those data are incomplete or unrepresentative, the resulting outputs may introduce bias. Similarly, some AI models lack transparency, making it difficult to understand how conclusions are generated. There is also the potential for overreliance on predictive outputs, particularly when those outputs are presented with a level of precision that may not reflect the underlying uncertainty.
These are not abstract concerns. They have direct implications for how safety signals are interpreted and how decisions about trial continuation or modification are made.
DSMBs are well positioned to serve as an independent checkpoint in this environment. Their role is not to validate algorithms in a technical sense, but to ensure the evidence used to guide decisions is credible, appropriately interpreted, and consistent with the study’s objectives.
When might AI expertise be needed?
There may be situations where additional expertise is helpful.
In trials that rely heavily on machine learning models to inform endpoints or adaptive decisions, it may be appropriate to involve individuals who can explain how those models function and how they were validated. This does not necessarily require a permanent AI specialist on every DSMB, but it may call for access to that expertise when needed.
The goal is not to transform the DSMB into a technical development team. It is to ensure the board has enough context to evaluate the information it is reviewing and to ask the right questions when something is unclear.
Preserving independence in a data-rich environment
As trials become more technologically integrated, maintaining independence remains a central concern.
AI tools are often embedded within sponsor or CRO workflows, and in some cases, they may be part of proprietary platforms used to generate or analyze trial data. This can create additional layers between the raw data and the outputs presented to the DSMB.
In this environment, it becomes even more important to ensure the information provided to the DSMB is independently validated and aligned with the study’s statistical analysis plan. The longstanding expectation that DSMBs operate independently from those conducting the trial continues to apply and may carry even greater weight as analytical complexity increases.
A practical path forward
So, do DSMBs need AI expertise?
Not in the sense that every board must include a data scientist or machine learning engineer. But DSMBs do need support through clear, transparent communication about how AI is being used within a trial and what its outputs represent.
For sponsors, this means ensuring AI-driven components are well documented, appropriately validated, and integrated into the broader statistical and oversight framework. It also reinforces the importance of experienced statisticians and independent oversight structures that can translate complex analyses into meaningful insights.
Ultimately, the effectiveness of a DSMB depends less on whether it includes AI expertise and more on whether it has the tools and context needed to evaluate increasingly complex data responsibly.
Clarity matters more than specialization
As clinical trials continue to evolve, so will the tools used to design and analyze them. AI will likely play a growing role, and adaptive methodologies will continue to shape how studies are conducted.
But the core responsibility of the DSMB remains unchanged. Its role is to provide independent, expert oversight that helps protect participants and preserve trial integrity.
Meeting that responsibility does not require DSMBs to become AI experts. It requires clear, reliable information and access to the right expertise when needed. In a landscape defined by growing complexity, clarity is what enables effective oversight.
Want to learn more? Check out this overview of Advarra’s review services or contact us to ask an expert a specific question.
