How AI and Advanced Statistics Are Helping Research Rise

Clinical trials are undergoing a period of change. Innovative uses of artificial intelligence (AI) are unlocking better insights. Regulators are backing risk-based approaches and a focus on the issues that matter. In this period of evolution, sponsors and CROs need to choose the right solutions—and the right partners—to help keep research rising.

ICH E6(R3) marked a major turning point in how clinical trials are planned and carried out. Regulators want sponsors to identify critical-to-quality (CtQ) factors, monitor those factors deliberately, and demonstrate critical thinking. Risk-based quality management (RBQM) today should be a live operational process—continuously reviewed and adapted to ensure risk and proportionally remain in view as data is collected and processed.

This continuous process can be enhanced with the use of intelligent AI. Tools like machine learning and deep learning (DL) can also revolutionize previously labor-intensive data management work. The key is choosing the right tools for the job.

Challenges preventing research rising

There are several challenges in traditional clinical trial execution which are preventing research from rising. Legacy operating models still linger, including emphasis on traditional on-site monitoring approaches on standard weekly frequencies, over-engineering of auto-queries and edit checks that either don’t yield intended outcomes, and overlapping/redundant functional data review processes where multiple reviewers are seeking to validate the same data point without congruency.

Manual listing review is incredibly resource-intensive with data managers having to spend a huge amount of time on low-value investigations. There can also be extensive back-and-forth with sites, leading to increased cycle time, higher levels of rework and data reconciliation bottlenecks which threaten database lock.

Similarly, medical and safety review is often a static, manual exercise. While the right tools can consolidate and improve review, too often ad-hoc or inappropriate tools delay action, lack traceability and scalability, and provide generic analytics which fail to reflect how data monitors work.

Medical coding is another costly and time-consuming process. Yet current auto-coding tools offer just 50% accuracy on free-text terms.

With our advances in technology, automation, and AI focus, it’s time to break down old techniques and process, and inject a new way of working that drives harmonious and contextually informed actions to drive a pro-active and impactful approach to meeting ALCOA+ principles.

Solution 1: Operational RBQM

Effective RBQM is not about one specific tool, one specific process, or a teams interpretation of updated regulations and best practices. Effective RBQM hinges upon harmony between these factors, such as connecting study-level risk detection and central analytics with site-level prioritization and action to detect issues earlier, focus monitoring where it matters most, and demonstrate ICH E6(R3)/E8(R1)-aligned oversight with confidence.

There are three key elements to this operational approach:

1. Cross-functional stakeholder alignment on future state operating model to be deployed
2. Proper assessment of to-be/future state assessment of business processes and technology stack to support driving updated objectives and key results
3. Change management and strategic plans to enact required organizational changes to drive an RBQM operating model that adapts/scales to the companies pipeline and organizational goals

If we truly want to disrupt and evolve legacy R&D practices to drive higher quality and efficiency, we have to focus on proper replacement of non-value-added activity (i.e. SDV / routine visits) with adoption of new techniques (i.e. critical data review/SDR/adaptive monitoring plans). This includes a hard look at leveraging AI and advanced statistical methods and leading with this as a driving force to then activate central monitoring, on-site monitoring, and data management activities.

This also includes a focus on proper re-training and upskilling of the R&D resources, especially providing scenario and/or simulated related training, so we can confirm legacy models are fading away and allowing for higher quality/more effective process to prove outcomes can be achieved.

By adopting this evolved approach, sponsors and CROs can flag emerging issues before they become protocol deviations or inspection findings, empower more targeted SDR and capture all evidence in a single, audit-ready environment, providing a clear inspection narrative.

Solution 2: Effective data management tools

AI and integrated platforms can transform how we review and manage data in three key areas—manual listing review, medical coding, and medical and safety review.

In listing review, intelligent query detection fundamentally shifts management from time-consuming manual investigation to confident confirmation. Validated agentic AI with proven precision and recall can identify true discrepancies with over 80% precision, eliminating time spent on non-issues. A depth-first approach also ensures effort is concentrated where delays are most likely to occur. These evidence-backed findings decrease subjective interpretation and variability, reducing query cycle time, improving query acceptance rates and reducing critical-path risk to database lock. To ensure audit-ready execution, all workflows should be 21 CFR Part 11–compliant, with full traceability and documented reasoning for all AI-assisted outputs.

Intelligent medical coding, which harnesses the power of DL, can generate automated medical coding suggestions for adverse events and concomitant medications at up to 99% accuracy. A DL medical coding tool has also been shown to automatically handle regular updates of the WHODrug and MedDRA dictionaries with up to 80% accuracy for completely new dictionary terms. The result? Reduced manual effort and accelerated and standardized medical coding without compromising quality or compliance. Improved data quality control can also increase data integrity for activities like medical and safety Review.

For medical and safety review, connecting consolidated clinical data and medical review signals with prioritized action and traceable follow-up, allows medical monitors and safety reviewers to detect issues earlier and demonstrate audit-ready compliance with confidence. By providing near real-time access to refreshed clinical data, with review rules and patient flags to prioritize attention on emerging issues, validated platforms can reduce patient exposure and improve audit readiness with a documented review process and clear evidence of ongoing oversight in line with ICH-GCP expectations.

Conclusion

Clinical trials allow us to unlock new discoveries and improve the health of people worldwide. But, to continue to innovate, sponsors and CROs need to work with partners who can harness the latest advancements in AI and advanced statistics, help them enable these technologies into their standard operating models to drive results, and empower upskilled resources to demonstrate critical thinking at every stage of trial design and execution.

By doing so, they can run smarter clinical trials with more reliable results for the benefit of patients worldwide and ensure research continues to rise.

About the author

Ken McFarlane, vice president, strategic consulting, CluePoints, has spent that last 24 years in the clinical trial R&D space. 12 of those 24 years was spent working in various clinical operations roles within sponsors & CROs, with a large focus in monitoring, project management and clinical trial oversight. The other 12 years Ken has spent working for clinical trial technology vendors, where he's been working to innovate and streamline clinical trial processes to benefit sponsors, CROs, sites, and the patients they serve.