Clinical and operational data are usually managed as separate streams, each feeding different teams and different reports. But the real advantage comes from converging them at the moment decisions are actually made: trial design, interim go/no-go, and program continuation. This session explores why fragmented data doesn't just slow trials down—it degrades the quality of the decisions themselves. Drawing on real-world examples from feasibility planning, enrollment-risk prediction, and real-time trial oversight, we'll offer a practical framework for identifying exactly where your clinical and operational data need to converge.

Regeneron is successfully accelerating the journey from "Data to Insight" by building a modern, Unified Analytical Platform. This session explores how a multilingual Statistical Computing Environment (SCE) bridges the gap between clinical operations and statistical programming. We will detail our strategic planning, risk-based validation for open-source frameworks, and the integration of AI layers to scale clinical data management and interactive reporting.

Progress update on FDA's real-time clinical trials pilot, sharing how the program is unfolding in practice. Discussion points include what real-time data access looks like in action, infrastructure and data standards enabling continuous monitoring, early wins and unexpected challenges from the pilot, and lessons learned around data harmonization and site-level data flow. The session also touches on balancing speed with data integrity, and what the pilot suggests for scaling real-time oversight across future trials.

Clinical data management is evolving from a reactive, task-driven function toward a more proactive and intelligent approach. This presentation will explore how AI can act as an “always-on teammate,” continuously monitoring clinical data, identifying potential issues, surfacing actionable insights, and supporting data management teams throughout the trial lifecycle.

• Leveraging AI to automate data management activities for clinical study teams, database programmers, and data managers 
• Redesigning clinical data flow for agentic AI monitoring, orchestration, and workflow automation 
• Automating anomaly detection, data reconciliation, and query management to accelerate clinical database locks​

Every protocol amendment triggers hours of manual rework across specifications, data definitions, and analysis plans; rework that quietly drives trial delays, added cost, and last-minute quality risk. CDISC's 360i initiative is building a connected data framework that links protocol design through analysis and submission, so a single change updates everywhere it matters instead of being rediscovered downstream. This session gives clinical operations leaders a plain-language look at what 360i is building, why it should matter to their timelines and budgets, and what to watch for as it heads toward broader adoption.

Digital protocols have the potential to reduce manual study setup and improve interoperability across clinical trial systems, but what is possible today? The SCOPE 365 “Automating the Configuration of Data Acquisition” challenge helped answer that question by bringing together innovators from across the clinical research ecosystem to demonstrate how standards-based digital protocols can automate real-world study configuration activities. Hear directly from winning challenge participants as they share their use cases, how the Unified Study Definitions Model (USDM) supported their solutions, and lessons learned through implementation. Attendees will see how structured protocol data can move the industry beyond manual configuration, improve consistency across systems, and accelerate study start-up while advancing a more automated and interoperable approach to clinical trial execution.

Eligibility criteria and Schedules of Activities sit in every major registry as free text, so no system can query them. This session introduces the Clinical Trial Distribution Network (CTDN), a neutral, open-access augmented registry operated by The Synergist that turns registry data into computable HL7 FHIR and CDISC USDM assets at the source. We'll structure a live protocol on stage and show what ships by March.

Clinical studies are foundational to drug development programs, and Tables, Listings, and Figures (TLFs) constitute the primary statistical reporting outputs of these studies. The first step in producing TLFs is compiling a List of Tables, Listings, and Figures (LoT)—a structured inventory specifying every output to be generated—from the study's Statistical Analysis Plan (SAP). As study designs grow in complexity, timely and consistent drafting of LoTs becomes increasingly challenging and labor-intensive, requiring statistical programming leads to manually interpret lengthy and complex SAPs. This manual effort is time-consuming, typically 2-5+ days, and prone to inconsistencies. To address this, we developed an automated agentic system that generates a draft LoT directly from an SAP document. The system uses semantic similarity search over a curated corpus of historical studies to retrieve relevant few-shot exemplars, which—together with domain-specific mapping rules and therapeutic area templates—guide agentic generation of LoT entries by output type. To improve robustness, multiple independent generation iterations are aggregated through a self-consistency step, followed by a second correction pass informed by accumulated meta-review rules. In evaluations across multiple studies, the system achieved an average concordance of 70%—in both precision and recall—at the individual TLF level when compared against finalized LoTs, comparable to the initial draft a statistical programming lead produces before broader study team review and refinement. The system has been deployed and used to accelerate 30+ studies to date, across therapeutic areas and phases, including non-standard phase 3 studies. End-to-end processing typically completes in approximately 10 minutes per study. This approach shows AI-driven automation can meaningfully improve efficiency, consistency, and standardization in clinical trial reporting workflows, supporting scalable and harmonized statistical deliverables across the drug development lifecycle while enabling statistical programmers to focus their expertise on higher-value analytical tasks.

The promise of AI for clinical trial statistical analysis is significant, yet most organizations struggle to move from compelling demos to production-grade capability. The barrier is rarely the model—it is the surrounding infrastructure and domain-specific context needed to make AI trustworthy, reproducible, and fit for regulated work. We share a proof-of-concept MVP of an AI-assisted statistical analysis platform, covering secure back-end infrastructure with human-in-the-loop review and traceable outputs, context engineering grounded in validated, version-controlled statistical modules, and a parallel validation framework. Throughout, the guiding principle is to augment, not replace, statisticians—who retain full accountability for interpretation and decisions.

Pfizer is modernizing its infrastructure to support an open-source and AI-enabled ecosystem for regulated clinical use cases. This session bridges the gap between clinical operations and statistical programming, focusing on the strategic business value of a next-generation Statistical Computing Environment (SCE). We will explore practical approaches to change management and orchestrating well-governed environments that seamlessly integrate R, Python, and Large Language Model (LLM) frameworks to accelerate clinical trial insights.

ANALEE is an AI-integrated EDA framework driven by metadata architecture that enables rapid deployment of interactive visualizations for structured clinical datasets. Leveraging Quarto's interactivity and metadata-driven schemas, we produced an easy-to-use point-and-click interface the framework delivers comprehensive descriptive analytics, advanced statistical methodologies, and extensibility for specialized modules customized for non-standard analyses for an audience beyond programming statisticians.

Assay data in clinical and translational research is fragmented across reports, spreadsheets, and databases, making reuse labor-intensive. AssayFinder AI, an agentic multi-agent platform, ingests reports and datasets, harmonizes metadata, and maps content to a Common Data Model, with human-in-the-loop validation and R-based statistical automation. Users query data in natural language, cutting preparation from days to minutes, while 21 CFR Part 11 and ALCOA+ alignment supports compliant, reproducible evidence reuse.

The promise of AI-assisted statistical analysis in clinical development is significant, yet deploying AI into a highly regulated production environment remains difficult. We describe a pragmatic, two-stage strategy—recognizing that the path to trustworthy AI runs through standardization, not around it. An iterative interim platform built on validated, reusable analytical modules delivers faster, reproducible analyses today while serving as a context-engineering engine: modularizing the statistical workflow into steps that must remain deterministic and those that allow flexibility, enabling early process and SOP development, and refining modules through user feedback. Throughout, the effort empowers, not replaces, statisticians.

This presentation will explore how ATLAS uses AI-enabled capabilities to help streamline trial operations, connect information across workflows, identify insights, and support proactive decision-making. The discussion will also highlight practical considerations for responsible AI adoption, including data quality, governance, integration, and human oversight.