For much of the pharmaceutical industry’s discussion around artificial intelligence (AI), attention has focused on drug discovery. Yet as sponsors race to shorten development timelines, another area of the clinical trial ecosystem is emerging as a candidate for technological disruption: biostatistics.
Despite decades of advances in data capture and trial management, many of the processes underpinning statistical analysis remain heavily reliant on manual programming, bespoke workflows, and labour-intensive quality control.
Discover B2B Marketing That Performs
Combine business intelligence and editorial excellence to reach engaged professionals across 36 leading media platforms.
For an industry under increasing pressure to accelerate development timelines while maintaining regulatory compliance, the question is becoming harder to ignore: why are some of clinical research’s most critical workflows still operating much as they did decades ago?
That is the challenge that Veristat is attempting to address through InStat, an AI-enabled biostatistics platform designed to automate elements of statistical analysis and reporting while preserving the traceability required by regulators.
“We’ve been writing, checking and executing thousands of lines of study-specific code by hand for decades,” Kyle McBride, Vice President of AI and Innovation at Veristat tells Clinical Trials Arena. “At some point you have to ask: there’s got to be a better way to do this.”
A legacy process meets modern technology
The pharmaceutical industry has historically adopted new technologies cautiously, particularly in areas directly tied to regulatory submissions. While electronic data capture and risk-based monitoring have transformed certain aspects of clinical trial execution, statistical programming remains largely dependent on customised code developed on a study-by-study basis.
According to McBride, a recent analysis of a single study required more than 21,000 lines of code to generate submission-ready statistical outputs. Such processes can take months to complete and require extensive review cycles to ensure quality and compliance.
The challenge is not simply one of efficiency. As sponsors seek to reduce development timelines, operational bottlenecks across the trial lifecycle are attracting greater scrutiny.
“AI is creating an environment where the conditions are right to solve some of the problems we’ve been living with for years,” McBride says.
Yet he argues that focusing exclusively on AI risks overlooking a broader opportunity. Rather than viewing generative AI as a standalone solution, Veristat has drawn heavily from practices established in the software engineering sector, including automated workflows, infrastructure-as-code principles and continuous integration pipelines.
The goal, he says, is not merely to perform existing tasks faster, but to redesign how statistical analysis is conducted altogether.
Responsible AI in a regulated environment
For all the enthusiasm surrounding AI, concerns over reliability remain acute in clinical research.
Large language models continue to be susceptible to hallucinations, creating challenges for any environment where accuracy is non-negotiable. In drug development, where regulatory decisions ultimately affect patient safety, the tolerance for error is particularly low.
McBride is explicit about the limits of AI’s role. “AI should not be determining whether a drug is safe or effective,” he says. “It’s not responsible to let AI decide whether the primary endpoint of a pivotal trial has succeeded.”
Instead, he sees AI functioning as an analytical assistant rather than an autonomous decision-maker. Areas such as data exploration, pattern recognition and insight generation may be well suited to AI augmentation, while core statistical calculations and regulatory decision-making remain under human control.
This distinction reflects a broader shift emerging across the industry. While initial conversations about AI often focused on automation and workforce reduction, many organisations are increasingly framing deployment strategies around augmentation and risk mitigation.
The result is a more pragmatic vision of AI adoption: accelerating workflows without compromising scientific oversight.
Traceability becomes a competitive advantage
One of the more intriguing aspects of AI adoption in clinical development is the possibility that automation could improve regulatory transparency rather than weaken it.
Historically, sponsors have relied on statistical programming code as a means of demonstrating how results were generated. Regulatory agencies reviewing submissions often trace outputs back through layers of programming logic to understand how analyses were conducted.
McBride argues that newer platforms may offer a more direct route to transparency.
Instead of reconstructing analysis pathways retrospectively, Veristat’s platform automatically captures metadata, transformations and analytical decisions as results are generated. The outcome is a real-time audit trail that can provide a detailed record of how data moved through the analysis process.
“The industry has traditionally submitted programs because regulators needed visibility,” McBride says. “What regulators really want is traceability.”
As agencies including the FDA continue exploring their own use of AI, questions around explainability and auditability are likely to become increasingly important. Technologies capable of enhancing efficiency and transparency may therefore find a more receptive audience among regulators than many initially assumed.
The talent challenge no one wants to discuss
While much of the industry conversation focuses on productivity gains, AI’s implications for workforce development remain less explored. Biostatistics has traditionally relied on a pyramid structure in which junior programmers and statisticians build expertise through hands-on execution of routine tasks before progressing into more senior roles.
AI threatens to alter that model. A senior statistician supported by advanced AI tools may require fewer junior resources to complete a project. While this increases efficiency, it also raises questions about how future experts will be trained.
“If AI performs much of today’s entry-level work, where do tomorrow’s senior experts come from?” McBride asks.
The issue extends beyond biostatistics. Similar concerns are emerging across clinical operations, data management and medical writing, where AI is increasingly capable of handling tasks once considered foundational training experiences.
For CROs and sponsors alike, balancing productivity gains with long-term talent development may become one of the defining workforce challenges of the next decade.
Beyond biostatistics
The industry’s ambitions extend far beyond statistical analysis. Enrollment delays continue to represent one of the largest sources of clinical trial inefficiency, often adding months to development timelines. Sponsors are increasingly exploring how AI can support feasibility assessments, site selection strategies, patient identification and regulatory planning.
In parallel, organisations are beginning to leverage AI-driven analytics to extract value from vast repositories of historical trial data, using insights to inform future development programmes and regulatory engagement strategies.
The opportunity, according to McBride, lies in rethinking clinical development as a whole rather than optimising isolated functions.
“We’re looking at an opportunity to fundamentally redefine how clinical trials are run,” he says.
Whether the industry ultimately embraces that vision remains uncertain. Pharmaceutical companies have long balanced innovation with caution, particularly when patient safety and regulatory compliance are at stake.
Yet as development costs continue to rise and pressure mounts to accelerate timelines, the appetite for meaningful transformation appears to be growing.
The next phase of AI adoption in clinical research may therefore have less to do with replacing scientists and more to do with redesigning the systems in which they work.
And for an industry built on evidence, that may prove to be the most consequential experiment of all.
