Understanding the process before applying the technology
Clinical endpoint adjudication is a well-established component of many clinical trials. Its purpose is to reduce investigator and systematic bias through the independent review of complex clinical endpoints by qualified experts. Because adjudicators typically work remotely, often across different time zones, and without direct access to patients or source records, the process relies on a robust technical infrastructure.
This article builds on ongoing work to examine where artificial intelligence could realistically support clinical trial operations, using endpoint adjudication as a structured and well-defined example. Meaningful and compliant use of AI depends first on a clear understanding of the adjudication process itself.
This article, the second in a series on AI agents in endpoint adjudication, examines that process and the roles involved as a foundation for evaluating where AI may, and may not, add value.
Operational Requirements of Endpoint Adjudication
Endpoint adjudication platforms must meet several core operational requirements. They must provide secure access to blinded clinical data, support structured assessments and discrepancy resolution, and reliably transmit final outcomes to the clinical database.
Meeting these requirements presents significant technical and operational challenges. Systems must be robust enough to ensure data integrity and compliance, yet simple and intuitive enough to be used efficiently by clinicians and operational staff with limited time.
Beyond the technology itself, endpoint adjudication is a role-based process. Multiple participants interact with the system, each with defined responsibilities, permissions, and timelines. Today, these roles are fulfilled entirely by human operators. Many of the associated tasks are repetitive, rules-based, and resource-intensive, making them candidates for careful AI-supported optimization.
Adjudication Setup: Defining the Process
Endpoint adjudication is planned during trial setup. The events requiring adjudication, the composition of the adjudication committee, the assessment process, and the associated roles are all defined in the adjudication charter. This document is agreed upon by the sponsor and the adjudication committee and serves as the formal reference for all adjudication activities.
At this stage, generative AI could potentially assist with drafting or structuring the charter based on the study protocol and relevant reference material. Any such use would remain supportive, with full human oversight and final approval retained by the sponsor.
Roles in Endpoint Adjudication and Potential AI Support
Once the study is underway, adjudication is triggered by the occurrence of a qualifying event. This may include confirmation of subject eligibility, the occurrence of a predefined adverse event, or outcome measurements related to treatment effect. Each event initiate a structured workflow involving several distinct roles.
- Investigator and Site Personnel
Investigators or site staff may receive requests for additional information from the subject’s medical record. They are responsible for providing the requested data within defined timelines while ensuring that no personal identifiers or treatment allocation details are disclosed.
Software platforms can streamline this exchange, but broader use of AI at the site level raises governance and responsibility considerations that fall outside sponsor control and therefore require caution.
- Adjudicators
Adjudicators are notified when an event requires review. They receive the relevant information, may request additional data within the framework of the adjudication charter, and issue an assessment based on predefined criteria. In cases of disagreement, adjudicators may participate in resolution discussions with the other committee members and the chair.
While the adjudication of complex medical cases requires clinical expertise, adjudicators could benefit from AI-based decision support. Such support may include verifying data completeness, identifying missing or inconsistent information, suggesting the request of additional data, organising materials for review, or presenting relevant prior cases.
Any opinions generated by AI would remain advisory and would not replace expert judgment.
- Adjudication Coordinator
The adjudication coordinator, who may be part of the sponsor organisation or an independent consultant, plays a central operational role. Responsibilities typically include identifying cases requiring adjudication, requesting and verifying data from investigational sites, assigning cases to adjudicators while balancing workload, managing disagreements, and recording final decisions.
Many of these activities are structured and rules-based. With appropriate controls and verification steps, a well-trained AI agent could support some of these tasks efficiently. Human oversight would remain essential, particularly during early use.
Conclusion: Aligning AI Support with Defined Processes
AI agents can support endpoint adjudication only when they are aligned with clearly defined processes, roles, and responsibilities. This article has outlined those elements as a necessary foundation for any assessment of AI applicability.
Opportunities for AI support exist across the adjudication lifecycle. At study setup, generative AI can assist with drafting adjudication charters based on protocols and relevant documentation. Once the study is underway, AI agents can streamline repetitive, rules-based tasks and reduce operational burden. These opportunities are supportive rather than transformative, and they do not replace clinical expertise or sponsor accountability.
What Comes Next
The next article in this series will examine different types of AI agents and assess which are best suited to specific adjudication tasks, helping sponsors distinguish realistic use cases from theoretical promise.
