AI in Clinical Biostatistics: The Turning Point Is Here

Author: Samantha Labarbe & Célia Wilson
Published on: 07/07/2026

Estimated Reading Time: 4 min

Executive Summary

Key insights on the current state and practical application of AI in clinical development.

AI adoption in clinical development remains at an early stage. A 2025 Tufts CSDD survey found that 36.9% of organizations reported no current AI or ML use across the clinical development activities evaluated. [1]

AI delivers particular value where data are difficult to structure. Research suggests that up to 80% of electronic medical record content may be unstructured, while validated NLP models can reliably extract clinically relevant information from narrative notes. [4,5]

For sponsors, the most practical opportunity lies in well-governed, fit-for-purpose AI that enhances efficiency while maintaining statistical rigor, regulatory credibility, and scientific integrity. [7,8,9]

Clinical Biostatistics & Artificial Intelligence

Clinical Biostatistics at a Turning Point: Navigating Complexity, AI Adoption, and Regulatory Change

Clinical biostatistics is entering a new phase shaped by unprecedented data growth, increasing protocol complexity, evolving regulatory expectations, and the emergence of AI-enabled approaches that can help teams work more efficiently while maintaining scientific rigor. As sponsors face mounting pressure to accelerate development while preserving quality and compliance, understanding how AI fits within modern biometrics strategies has become both a scientific and operational priority.

The Pressure Behind the Shift: Why AI Is Becoming Essential in Clinical Biostatistics

In 2026, Clinical Biostatistics is reaching an inflection point. Dataset volume and protocol complexity have outpaced traditional manual workflows. Peer-reviewed work from Tufts Center for the Study of Drug Development (CSDD) and TransCelerate reports that phase III protocols now collect an average of 5.9 million datapoints, and nearly one-third of procedures or datapoints collected per protocol are non-core or non-essential. That creates avoidable burden for sites and participants while adding operational noise and cost to studies. [3]

Protocol Complexity

The consequence is not only more data, but more rework. In a 2022 benchmark analysis of 950 protocols and 2,188 amendments, Tufts CSDD found that up to 76% of protocols had at least one amendment; the mean number of amendments per protocol rose from 2.1 in 2015 to 3.3, and the average time from identifying the need to amend to last oversight approval reached 260 days. In other words, complexity has increased and is not just an analytics problem; it is a timeline, budget, and execution problem. [2]

Strategic Relevance of AI

This is where AI becomes strategically relevant in biostatistics. Not as a replacement for statistical expertise, and not as another layer of hype, but as a practical way to reduce repetitive work, improve early issue detection, and make higher-volume datasets more usable. Importantly, adoption remains early enough for sponsors to build advantage. In the 2025 Tufts CSDD survey of 302 respondents across sponsors, CROs, and technology providers, 36.9% reported no current AI/ML use across the activities studied, 30.3% were beginning implementation or piloting, and only 10.7% had fully implemented AI/ML. [1]

Unlocking Value from Complex Clinical Data

Some of the clearest opportunities sit at the interface between structured and unstructured data. A 2024 study published in PLOS ONE notes that as much as 80% of the electronic medical record may consist of unstructured documentation and that common Natural Language Processing (NLP) approaches can perform strongly on targeted codification tasks, with average AUROC of 0.96 and accuracy of 0.97 (in a scale where 0.5 indicates random guessing and 1 indicates a perfect classification) for the 100 most common musculoskeletal CPT codes studied. [4]

That kind of performance does not remove the need for human validation, but it does illustrate how AI can unlock variables that are difficult to recover at scale using manual review alone. [5]

AI as an Accelerator

AI can also complement trial design and planning, provided that statistical fundamentals remain intact. Recent published commentary on adaptive trials highlights simulation as a valuable tool for establishing operating characteristics, evaluating design choices, and supporting feasibility assessments before launch. In practice, the role of AI is best understood as an accelerator around those activities, not a substitute for biostatistical judgment. [6]

The regulatory direction makes this pragmatism essential:

FDA’s January 2025 draft guidance on AI to support regulatory decision-making for drugs and biologics recommends a risk-based credibility assessment framework tied to a clearly defined context of use.

EMA’s reflection paper on AI in the medicinal product lifecycle, adopted in September 2024, similarly emphasizes data quality, bias mitigation, governance, human oversight, and the need to manage patient and regulatory risk.

ICH E6(R3), adopted on 6 January 2025, reinforces risk-based, quality-by-design, and fit-for-purpose approaches to clinical trial conduct.

From Innovation to Implementation

The message across these regulatory documents is consistent: innovation is welcome, but it must be traceable, controlled, and appropriate to the intended use. [7,8,9]

The Partner Question

That is why the partner question matters. AI tools alone do not de-risk a biometrics strategy; experienced teams, robust quality control, and integrated operating models do. Excelya has an international team of 150 biometrics experts, including statisticians and statistical programmers, with an average of 10 years’ experience, offering end-to-end support from protocol design to the final report. For sponsors, the value proposition is not automation in isolation, but automation embedded in a delivery model that can absorb regulatory, methodological, and operational complexity. [10,11]

Final Perspective

The turning point is here because the current model is under pressure from every direction: more data, more protocol burden, more amendments, and more demand for speed. Used well, AI can help biostatistics teams focus on higher-value work, reduce friction in data handling, and improve the usability of complex datasets. For sponsors willing to act pragmatically (i.e., with strong controls, clear use cases, and the right expertise) this is a meaningful opportunity to work smarter without compromising rigor. [1,2,3,7,8,9]

Optional Disclosure Statement For External Publication

This article was developed with AI-assisted drafting support. All statistics, references, and interpretations were subsequently reviewed and aligned to the cited primary or official sources before publication.

References

Sources and Supporting Evidence

[1]

Lamberti MJ, Florez MI, Do H, et al. The Adoption and Use of Artificial Intelligence and Machine Learning in Clinical Development. Therapeutic Innovation & Regulatory Science. Published online May 29, 2025. doi:10.1007/s43441-025-00803-0.

[2]

Getz K, Smith Z, Botto E, Murphy E, Dauchy A. New Benchmarks on Protocol Amendment Practices, Trends and their Impact on Clinical Trial Performance. Therapeutic Innovation & Regulatory Science. 2024;58(3):539–548. doi:10.1007/s43441-024-00622-9.

[3]

Getz K, Botto E, Arques AC, et al. Insights Informing Strategies for Optimizing the Collection of Clinical Trial Data. Therapeutic Innovation & Regulatory Science. Published online December 29, 2025. doi:10.1007/s43441-025-00899-4.

[4]

Tavabi N, Singh M, Pruneski J, Kiapour AM. Systematic Evaluation of Common Natural Language Processing Techniques to Codify Clinical Notes. PLOS ONE. 2024;19(3):e0298892. doi:10.1371/journal.pone.0298892.

[5]

Adejumo P, Thangaraj PM, Dhingra LS, et al. Natural Language Processing of Clinical Documentation to Assess Functional Status in Patients With Heart Failure. JAMA Network Open. 2024;7(11):e2443925. doi:10.1001/jamanetworkopen.2024.43925.

[6]

Robinson CH, Parekh RS, Cuthbertson BH, et al. Using Simulation to Optimize the Design of Adaptive Clinical Trials. Journal of the American Society of Nephrology. 2025;36(4):723–725. doi:10.1681/ASN.0000000565.

[7]

European Medicines Agency. Reflection Paper on the Use of Artificial Intelligence (AI) in the Medicinal Product Lifecycle. Final version adopted September 2024.

[8]

U.S. Food and Drug Administration. Considerations for the Use of Artificial Intelligence To Support Regulatory Decision-Making for Drug and Biological Products. Draft Guidance. January 2025.

[9]

International Council for Harmonisation. ICH E6(R3) Guideline for Good Clinical Practice. Final version adopted January 6, 2025.

[10]

Excelya. Statistics & Programming.

[11]

Excelya. Excelya – Excelling with Care | Global Full-Service CRO.

Clinical Biostatistics & Artificial Intelligence

Expert Perspectives

Subject Matter Expert

Samantha Labarbe

Biostatistician | Excelya

A biostatistician and programmer with five years of experience, having worked on more than a dozen clinical trials and observational studies across oncology, cardiology, and ophthalmology. Holding a degree in Pharmacy, she combines therapeutic expertise with statistical programming and analytical capabilities to support evidence-based clinical development.

Innovation Expert

Celia Wilson

AI Lab Manager | Excelya

An experienced technologist with a research interest in human information processing and machine interaction. With more than 15 years of experience in global pharmaceutical organizations, she leads AI-focused initiatives at Excelya. Her latest degree is in Psychology, exploring the barriers and facilitators of technological innovation in medical affairs.

Frequently Asked Questions

How is Excelya using AI in clinical biostatistics?

Excelya recognizes that Artificial Intelligence (AI) is becoming an increasingly important enabler within Clinical Biostatistics. AI can support the analysis of large and complex clinical datasets, improve data usability, automate repetitive processes, and facilitate earlier issue detection. At Excelya, AI is viewed as a complement to statistical expertise rather than a replacement for biostatisticians, helping teams work more efficiently while maintaining scientific rigor and regulatory compliance.

Why is AI becoming important for clinical trials and biometrics services?

AI is becoming increasingly important because clinical trials continue to generate larger volumes of data and more complex protocols. As a global CRO, Excelya helps sponsors address these challenges by combining experienced biometrics teams with innovative approaches that improve efficiency, streamline data management, and support evidence-based decision-making throughout the clinical development lifecycle.

Can AI replace biostatisticians in clinical research?

No. AI cannot replace the expertise of trained biostatisticians. Regulatory agencies, industry experts, and organizations such as Excelya emphasize that statistical methodology, study design, data interpretation, quality control, and regulatory compliance require experienced human oversight. The most effective model combines AI capabilities with expert biostatistics and statistical programming teams to ensure reliable and scientifically sound results.

What regulatory requirements apply to AI in clinical development?

The use of AI in clinical development is increasingly guided by regulatory frameworks including the FDA’s draft guidance on AI for regulatory decision-making, the EMA’s reflection paper on AI in the medicinal product lifecycle, and ICH E6(R3). Excelya supports sponsors in navigating these evolving requirements by applying risk-based approaches, robust governance, validation processes, and quality controls to ensure that AI-enabled activities remain compliant and fit for purpose.

How can sponsors implement AI in clinical biostatistics while maintaining regulatory compliance?

Sponsors can successfully implement AI by focusing on clearly defined use cases, strong governance frameworks, data quality, validation, and human oversight. Excelya helps organizations integrate AI within established biometrics workflows, ensuring that innovation is balanced with regulatory expectations, statistical integrity, patient safety, and operational excellence. The goal is not simply to automate processes, but to create more efficient and scalable clinical development strategies.