Insight: This abstract proposes a novel framework for accelerating AMR mitigation and optimizing clinical trial efficiency through the widespread adoption of Decentralized Clinical Trials (DCTs)coupled with standardized Real-World Data (RWD) collection and analysis for Real-WorldEvidence (RWE). The main insight is that by strategically integrating these approaches, we canovercome the inherent shortcomings of traditional methods, enhance data interoperability, and foster a proactive response to AMR. Key Propositions: Two pivotal actions are proposed. First, harmonization of data collection andacquisition by the consistent application of CDISC-controlled terminologies across all data sources,including RWD, to ensure seamless integration and interoperability. Second, consolidation and global collaboration through centralized data repositories for RWD from diverse sources (e.g.,environmental, animal, human). This continuous feedback loop with global data streams will not only enhance RWD quality but also facilitate a comprehensive and real-time understanding of AMR trends.

The presentation will offer insights into the progress of the PhUSE working group ‘Supporting the Use of SEND for the Implementation of Virtual Control Groups’.

The use of virtual control groups in nonclinical studies is increasing due to low availability of some animal models as well as the demands imposed by the 4Rs: Replace, Refine, Reduce and Responsibility when conducting animal studies. Consequently, virtual control data is starting to appear in studies intended for regulatory submission where SEND is required.

CDISC SEND is the industry standard for electronic nonclinical data and feeds the data repositories from which virtual control data is extracted.
The PhUSE team has devised an approach for the inclusion of virtual control data in the study SEND dataset package. Using the existing SENDIG v3.1.1 standard, this proposal includes recommendations for incorporating virtual control animal data into the SEND domains, including trial design, demographics, exposure, and findings domains.

Behind every high-quality, CDISC-compliant SDTM dataset lies a team of data-driven heroes: Data Managers, Standards Managers, Clinical Programmers, and Biostatisticians, each bringing their own unique skillsets and expertise to the fight for clean, conformant data.

Our heroes may hail from different domains, but they share a common trait: a working knowledge of CDISC standards and familiarity with the tools of the trade. This foundational knowledge becomes their secret weapon to enabling smarter, more collaborative workflows. Understanding the strengths of each role unlocks opportunities to rethink traditional processes, distribute tasks more effectively, and innovate with workflows that leverage each team member’s unique superpowers.

This poster will present our SDTM Squad, outlining the signature skills each role possesses, supercharged by emerging contributions of AI. By highlighting the strengths of this dynamic team, we’ll demonstrate that with the right strategy and collaboration, regulatory compliance can be both faster and more heroic.

Preparing submission ready annotated Case Report Forms (CRFs) remains a time consuming, error prone step at study close. We developed a lightweight application that automates CRF annotation on Rave Medidata and Veeva EDC Platforms in alignment with FDA/CDISC expectations by combining three study inputs: (1) the blank CRF, (2) the SDTM mapping specification, and (3) the eDC/eCRF specification. The App parses the specifications, reconciles field to domain mappings, and programmatically places standardized annotations on the CRF, producing a navigable, submission ready PDF and an accompanying QC trace.

In internal use, the App has reduced manual formatting/annotation effort and improved consistency of domain/variable labeling and bookmarks. We will share examples, typical validation checks, common failure modes (e.g., label collisions and conditional field logic), and practical tips for embedding the tool into end of study workflows with CRO partners.

The use of AI in the generation of submission datasets of clinical trial results is currently a very hot topic, this although there still are few real applications.

In the last half year, we have been working on first possible implementations of AI for prediction/suggesting mappings for SDTM, and on predicting LOINC-CDISC mappings for LOINC codes not present yet in the by CDISC published LOINC-CDISC mappings. We also automatically generated CORE-YAML rules starting from ""CDISC Dataset Specializations"" and the CDISC-Library API. Also the use of AI for unit conversions (e.g. between ""conventional"" and ""SI"" units) has been investigated.

These are intended to complement features in our software tools that are mostly based on developed algorithms.
The poster will present the results of this work comparing results of AI, use of APIs, and algorithms.

Current clinical data submission relies on static "snapshots," causing significant bottlenecks due to manual mapping and disconnected workflows. This paper proposes a 2041 vision of a "Zero-Submission" ecosystem, where the traditional concept of filing becomes obsolete through continuous data integration. At the core of this evolution is an intelligent converter engine that enables real-time standardization from raw data to SDTM and automated derivation of ADaM datasets. By leveraging semantic AI and metadata-driven continuous pipelines, the system ensures perpetual data integrity and traceability via auto-generated data lineage without human intervention. This shift allows regulatory agencies to access live data streams for collaborative review, drastically accelerating drug approval cycles. Ultimately, the transition to a continuous regulatory ecosystem eliminates administrative burdens and ensures that life-saving therapies reach patients faster, redefining the relationship between industry and regulators through a foundation of transparent, real-time data flow.

In the complex landscape of clinical trials, maintaining data integrity is crucial. A dashboard has been introduced to streamline the process of identifying and addressing data structure discrepancies during study set up. This tool systematically monitors the data structure of a clinical trial with respect to SAE reporting to Pharmacovigilance, pinpointing issues such as missing fields, value range deviations, and categorization inconsistencies. By flagging these discrepancies, the dashboard enables swift corrective actions, which are essential for upholding the quality and compliance of clinical trial data used for SAE reporting to authorities. The dashboard can significantly enhance the efficiency of data structure review, reducing the time needed to manage discrepancies.

We built a Prototype which redefines protocol interpretation by fusing CDISC standards with generative AI in a deterministic, compliance-first architecture. Built on a vector database indexed with CDASHIG v2.3 and 3.0, the solution enables precise semantic retrieval of standard definitions and enforces real-time quality gates on trial design artifacts derived from clinical protocols. Unlike conventional automation, this model guarantees zero hallucination and full traceability—every output is anchored to protocol evidence and validated against CDISC Foundational Standards. The pipeline auto-generates Schedule of Activities, CRF specifications, and SDTM/CDASH mappings while embedding compliance checks at design time, reducing manual effort and accelerating database readiness. By combining AI-driven interpretation with standards-as-code, this approach demonstrates how connected ecosystems can deliver scalable, audit-ready digital transformation for clinical research—setting a new benchmark for reliability and regulatory alignment in next-generation trial design.

A clear data lineage is the cornerstone of improved data literacy that is critical for a streamlined and efficient CRF (Case Report Form) design-to-analysis workflow. TRACE, a data lineage dashboard, visualises how CRF items collected at site are transformed into SDTM variables interactively. By linking and mapping standard CRFs all the way to SDTM, TRACE fosters a shared data understanding amongst stakeholders across roles - data managers, programmers and biostatisticians – throughout the lifecycle of a study. This shared data understanding ensures that CRFs are designed to capture meaningful and analysable data. It additionally supports the development of new standards, minimising further maintenance and/or operational errors, as well as enhances adoption of standards.

In this poster, we will illustrate on the design and implementation elements of TRACE with examples.

At Cytel we initiated Dataset-JSON adoption by leveraging SAS macros, testing different processes, always with a focus on metadata extraction to ensure consistency.

To support review and quality control, we also developed an interactive Dataset-JSON Viewer, part of CDISC COSA Hackathons, that enables efficient visualization, exploration, and validation of CDISC Datasets-JSON. Integration of the open-source CDISC CORE engine within the viewer allows direct execution of CDISC Conformance Rules, providing immediate, rule-based feedback and supporting continuous data quality monitoring during dataset production.

With this poster we would like to share some lessons learnt from this practical experience, by providing concrete examples of Dataset-JSON conversion and considerations related to CORE outputs.

Consistent and correct ADaM specifications are essential for efficient development of analysis datasets. However, many issues with specifications are not found until the creation of the define.xml. This poster explains an automated macro designed to conduct early-stage quality control of ADaM specifications by checking cross-tab consistency. The macro examines relationships among datasets, variables, metadata, codelists, and methods. It identifies problems that could lead to errors or warnings in the define.xml. Examples include flagging codelists being defined but not being used, and vice versa; datasets calling undefined methods and methods defined without downstream use.

These checks enable self- and peer-quality control of ADaM specifications, without having to create a define.xml at the beginning of a study. This lowers rework later in the process, enhances the quality of specifications, and aids in the early detection of problems. This demonstrates how automation can improve metadata management and increase the effectiveness of ADaM development.

CRF data collection for Clinical Studies typically combine off-the-shelf global validated CRFs along-side trial-specific forms tailored to the unique requirements of each study. Once the line between global and study-specific forms is drawn, focus often shifts to the latter, while global standards are treated as immutable - after all, they’re “standard”, right? However, we’ve observed small deviations creeping into global CRFs during setup that, which becomes time‑consuming and costly to correct during implementation. Novo Nordisk has pledged to strengthen standardization to improve patient safety and study efficiency. To support this effort by easy identification of drifts, we propose a semi-automated dashboard with visual indicators that continuously monitor alignment with global standards throughout study setup. By detecting drift early, issues can be correct before they impact additional standard programs and become costly amendments. Think of it as a leak detector - keeping your study boat afloat and on course.

The ICH E9(R1) estimand framework introduced a paradigm shift in defining clinical trial objectives, but current ADaM structures lack a standardized approach for representing estimands and intercurrent events. To address this gap, we investigate ADICE (ADaM Intercurrent Events and Estimand) as a new ADaM dataset designed to: a) Capture estimand definitions and associated strategies in a structured, CDISC-compliant format. b) Link intercurrent events to subjects and analysis populations for traceability and c) Integrate seamlessly with existing ADaM classes while maintaining metadata consistency.

This Poster/Presentation highlights the conceptual design of ADICE, its alignment with CDISC principles, and practical examples demonstrating how it simplifies programming and reporting for complex estimand strategies.

CDISC SDTM/ADaM standards provide a common framework for clinical data submission; however, therapeutic area and regulatory specific guidance can introduce additional complexity. For HIV-1 studies, FDA technical specifications describe reviewer expectations for select datasets, particularly for virology and efficacy endpoints. Implementing these expectations alongside CDISC standards requires careful interpretation and clear documentation.

This paper shares practical experience integrating FDA HIV 1 technical specifications within a CDISC aligned SDTM and ADaM model. The focus is on virology data, including the use of ADVR as a dedicated analysis dataset for HIV 1 RNA endpoints and its relationship to reviewer focused datasets such as ADEFFOUT. Common implementation challenges are discussed, including definition, visit selection, attribute differences and situations where deviations from standard ADaM structures were necessary. The presentation highlights practical approaches for maintaining CDISC compliance

Our industry is undergoing a transformative shift towards digital-first methodologies – one that requires re-thinking traditional processes. While most of us have spent the past decades using standards retrospectively to ensure compliance with regulatory submissions, this is no longer sufficient. The new digital era demands for standards to architect the future, as demonstrated by CDISC 360i. It's no longer a question of whether it can be done, but how to do it best. One approach is to start from scratch (greenfield approach), to then translate this into existing organizations that need a modernization roadmap to pivot from legacy document-centric processes to an interconnected metadata ecosystem.

In this presentation, I will explore the blueprint for this transformation, addressing both pathways (greenfield and modernization) and covering both mandatory and "innovation enablers" standards. This allows to envision a truly digital process providing high-fidelity data to power AI systems and get to submission faster.

OpenStudyBuilder is an open-source metadata and study definition repository supporting standards-based clinical trial automation. This presentation briefly reflects on key milestones and collaborations to date, including alignment with initiatives such as CDISC 360i and TransCelerate Digital Data Flow. The main focus, however, is on the road ahead: the evolving vision, strategic roadmap, and collaboration models designed to strengthen industry participation and shared ownership. Attendees will gain insight into how OpenStudyBuilder aims to further mature as a collaborative platform enabling scalable, standards-driven innovation.

Despite the availability of open-source tools for submission deliverables, activities like SDTM dataset, define.xml and Reviewer's Guide creation are often implemented as disconnected steps, limiting automation, reuse, and traceability.

This presentation demonstrates a metadata-driven approach aligned with CDISC360i, using a JSON specification connecting EDC build information to CDISC Library IG and CT standards. Reusable SDTM transformation algorithms from the COSA {sdtm.oak} R package are orchestrated based on standardised metadata, enabling automated and repeatable study code. Validation is performed using CDISC CORE to provide real-time compliance of the SDTM dataset JSON.

By using a single source of mapping metadata across the full CRT package, submission cycle times can drop dramatically with consistency and quality ensured across components.
This experience demonstrates how connected standards and CDISC360i principles can be realised in practice through metadata-driven, open-source workflows.

CDISC 360i is a multi-year initiative enabling end-to-end automation of the clinical data lifecycle, from digitized study design through analysis and reporting, using standards-based metadata and interoperable open-source tools. One significant opportunity is automating Define-XML generation and maintenance, which remains inefficient, manual, and error-prone.

This presentation highlights the 360i Define-XML Sub-team's Phase 1 accomplishments and Phase 2 plans. In Phase 1, we automated SDTM Define-XML generation by leveraging the Unified Study Definition Model (USDM), Biomedical Concepts, Dataset Specializations, and a new metadata specification model purpose-built for end-to-end automation. This model establishes the metadata structure needed for consistent, standards-driven generation of study-level artifacts.

In Phase 2, we are extending this approach to enable automated ADaM Define-XML creation through Analysis Concepts. The long-term objective is an automated, metadata-driven process for producing both SDTM and ADaM Define-XML, delivered as open-source software.

Clinical trials still rely on narrative protocols that must be manually interpreted to create EDC forms, monitoring plans, SAPs, analysis datasets, and CSRs. Each handoff introduces duplication, inconsistency, and avoidable delay. ICH M11 leverages the USDM structure to standardize how protocol data elements are represented, enabling direct automation of downstream trial processes.

This presentation demonstrates how a data-modeled M11 protocol can support a digital workflow spanning eCRF generation, direct extraction of EHR source data with PII redaction, and SDTM creation within a secure cloud workspace. Protocol-defined data requirements drive automated capture and validation, producing SDTM datasets with embedded provenance.

By replacing manual transcription at the clinical site and reducing the amount of source data verification through structured, automated data flow, this approach shortens timelines to database lock and strengthens confidence that what was designed, captured, analyzed, and submitted are one and the same.

Transforming Word/PDF protocols into the USDM and ICH M11 is challenging, manual, error-prone, and difficult to maintain. To overcome without requiring users to learn the USDM, we developed a digital-native solution compliant with USDM and ICH M11, enabling seamless DDF.

In Protocol AI, study design is captured in a structured format aligned with company templates and USDM requirements, allowing experts to work without changing their routines. The company’s protocol template is implemented as a one-time setup: it is uploaded, converted into a digital template, and automatically validated against ICH M11, then becomes available. AI or experts then use the USDM-formatted design data to generate protocol text.

Protocols is exported as USDM (Excel/JSON), MS Word, and PDF, reused for downstream documents (e.g., ICF, SAP), and pushed to operational systems (e.g., EDC). This approach eliminates transformation errors, user burden and manual checks while enabling rapid adoption of new updates, including USDM v4.0.

How many of us have stared at a Schedule of Activities and grappled with the exact intent or missed some important detail buried in footnotes?

The USDM and ICH M11 protocol template standard where both released in 2025. Used in combination, these standards bring the machine-readable study protocol within the grasp of clinical study teams, providing the foundation for more precise study design specifications.

This session will examine how USDM timeline construct can transform ambiguous SoA constructs into computable study logic and explore timelines patterns for handling complex study logic such as oncology cycles, AE handling and regular events handled outside the traditional face-to-face participant clinic visit.

Using open-source tools, we will load an M11 protocol directly into USDM, extract the Schedule of Activities, and expand the timeline logic that can be used to generate multiple downstream views of the study design focused on the needs of participants and site.

This presentation explores the integration of the Unified Study Data Model (USDM) including M11 support in OpenStudyBuilder, an open-source repository for metadata and study definitions.

The USDM and M11 standards streamline the exchange of study and protocol definitions, facilitating end-to-end automation across multiple systems. OpenStudyBuilder enhances these processes with its open interfaces and native data model, offering study exports in USDM and M11 formats as JSON or HTML file. This will now include support for narrative content, enabling us to fully support M11 document content in USDM.

The application also support M11 preview for easy overview supporting reviews and updates of structured M11 content managed within OpenStudyBuilder. Additionally, the API enables connectivity with other USDM-compliant systems.

We will discuss the challenges faced, and highlight the opportunities presented by adopting this innovative USDM standard.

While CDISC standards are mandatory for regulatory submissions, their adoption in academic Data Coordinating Centres (DCCs) remains inconsistent. In Phase I/II academic trials, the primary focus is IND maintenance and publication rather than FDA/EMA marketing authorisation, often rendering full CDISC implementation a perceived "resource luxury." However, the lack of standardisation creates significant hurdles for data sharing, meta-analysis, and ""future-proofing"" trials for industry transition.

This presentation explores the unique challenges faced by academia, including resource scarcity and fragmented workflows. We present a solution: an automated pipeline that integrates CDASH standards into CRF specifications at the design phase. By leveraging consistent metadata, this end-to-end workflow enables the automated generation of SDTM and ADaM datasets even in resource-constrained settings. We demonstrate that by lowering the barrier to entry through automation, academic institutions can achieve high-quality, interoperable data that enhances research reproducibility without overextending operational budgets.

Health organizations, including the WHO, call for greater inclusion of pregnant women in clinical research to address historic evidence gaps. Traceability and CDISC adherence are uniquely challenging when deriving datasets for pregnancy trials and trials where participants become pregnant. Establishing CDISC standards for pregnancy data is important to closing evidence gaps and reducing delays in treatment guidelines and regulatory approval.
This presentation focuses on strategies for deriving ADaMs for trials with pregnancies, built from initial framework in the HIV Therapeutic Area User Guide v1.0 and Pediatrics User Guide v1.0. Use cases are from IMPAACT Network trials dedicated to enrolling pregnant women and their infants, and include infants from subsequent pregnancies as associated persons. Examples cover guidance on supporting variables, strategies for multiple pregnancies and gestations, and recommendations for pregnancy occurrence and family-level ADaM Other datasets. Using CDISC best practices produces transparent pregnancy data, supporting the need for better research representation.

The xShare initiative advances the EHDS vision enabling European citizens to seamlessly and securely share their health data across borders. Collaboration among CDISC, HL7, LOINC, SNOMED, IEEE, IHE, CEN working with ISO, ICD 11 and NCI EVS has accelerated the evolution of the International Patient Summary into IPS+, expanding its ability to represent healthcare and research concepts needed for modern, efficient data driven care.

Through aligned models, terminologies, and implementation approaches, SDOs have created a computable foundation supporting both clinical and research use while preserving the IPS mission of safe, effective cross border care.

The xShare button operationalizes IPS+, giving citizens a simple, privacy preserving way to authorize standardized data exchange for care and research. This presentation outlines the collaborative process, technical extensions, governance, and innovative solutions from the xShare open calls, and also the expected impact on interoperability, patient empowerment, and integrated clinical research and healthcare ecosystems across Europe.

PROTECT-CHILD is a European Union-funded research and innovation project aiming to enhance evidence-based decision making and quality of life for pediatric transplant patients across Europe by improving data accessibility.

By securely integrating clinical, genomic, and real-world health data, the project enables advanced data-driven insights to support personalized transplant care. PROTECT-CHILD develops a privacy-preserving, interoperable data infrastructure aligned with EHDS, EOSC, and GDPR enabling federated analytics across harmonized datasets.

This presentation will outline the opportunity to align and harmonize applicable global standards (including HL7 FHIR resources, OMOP and CDISC’s Transplant TAUG as well as the Unified Digital Protocol, with the Protect-Child data model, using AI mapping methods to connect real-world data, (i.e. clinical research and healthcare data)), the collaborative process and the expected impacts on interoperability, data accessibility, patient empowerment on healthcare ecosystems across Europe in alignment with the EHDS.

In 2017, we launched an ambitious initiative to use AI/ML for classifying and extracting information from millions of TMF documents to enable migration to a TMF reference model–based system. We partnered with a vendor having working AI tool and set bold expectations: 100% automation and 98% accuracy. Reality fell short. While ~70% automation was attainable, overwhelming numbers of false positives undermined the credibility of the process, forcing us to abandon that approach and reset the strategy, expectations and switching gears to blending automated business rules, ML/AI, manual cleanup. This story is about compromises and corner-cutting to achieve the goal: successful migration of large volume of documents, data transformation while keeping business continuity for ongoing studies. Our experience underscores a critical lesson: AI is powerful, not magical. Its success hinges less on technical ambition and more on realistic expectations, sharp problem framing, and using AI only where it truly adds value.

Imagine running a Michelin‑star kitchen without a sous‑chef—chaos and missed details. Trial Master File (TMF) operations are no different. AI is your digital sous‑chef, handling the mise‑en‑place so you can focus on the artistry of quality and compliance. It chops through tedious tasks like document classification, metadata extraction, and folder placement, transforming TMFs from static archives into dynamic systems.

This session shows how AI clears clutter and preps your “ingredients”: checking email tone before you plate communication, turning meeting notes into a ready‑to‑serve task list, and sprinkling calendar reminders where needed. Then comes the entrée—predictive intelligence that flags missing signatures and site risks before they spoil timelines.

Just as a sous‑chef needs fresh ingredients, AI thrives on quality data and retraining. Attendees leave with practical prompts and governance recipes to compress 20‑step workflows into one or two well‑seasoned AI requests—delivering speed, accuracy, and continuous improvement. Bon appétit!

Traditional TMF risk assessments vary widely and often rely on inconsistent interviews or retrospective TMF metrics. This session introduces a simple, AI-guided, question-only TMF Risk Assessment Framework that standardizes how study teams evaluate TMF risk without accessing or ingesting any TMF data. Using structured prompts and a rules-based risk model, the approach produces a clear risk profile, narrative justification, and recommended mitigations aligned with ICH E6(R3). We will present the underlying risk model, sample question flows, and scoring logic. Attendees will learn a practical, low-risk method to improve TMF oversight that can be applied across any sponsor, CRO, or eTMF system.

AI is everywhere in the conversation — but where is it in the actual TMF? The hype suggests intelligent systems that classify, check quality, and surface insights automatically. The reality is more nuanced: most organizations are still working out what AI can do for them, what it requires from them, and what they're not yet comfortable letting it touch.

This panel brings together technology leadership and sponsor practitioners from organizations of different sizes to share practical experiences: what's delivering value, what's still on the wishlist, and where compliance and data quality concerns are shaping adoption.

We'll explore where AI is genuinely helping today, what foundations need to be in place first, and how organizational size affects the opportunity. Attendees will leave with an honest view of where the industry stands and a clearer sense of what ""ready"" actually looks like.

Panelists:

• Traci Wendler
• Aaron Grant
• Harpreet Lachhar

As our industry experiences significant initiatives such as ICH-E6(R3), Risk based approach and TMF model overhaul, it is increasingly important to effectively integrating new initiatives and changes at the local level. Local communities have emerged as pivotal a platform for facilitating this transition, offering unique opportunities for collaboration, knowledge-sharing and practical problem-solving. In Denmark and Japan, established communities have demonstrated the value of localized engagement and the communities are fostering innovation and supporting adaptation to evolving industry standards. Italy is currently in the process of establishing similar network, drawing on L&L from other regions. 

This panel will examine the distinct advantages these communities provide, such as enhanced communication and tailored support, while also addressing the challenges inherent in their development and sustainability. By exploring these cases, panelists aim to highlight best practices and identify strategies to overcome obstacles, ultimately promoting effective community-driven change within the industry.

The term soft power is most often associated with diplomacy and geopolitics, yet its core principles can also be applied to TMF management to enhance quality and compliance. Soft power relies on co-option rather than coercion; drawing on cultural appeal, influence, partnership, and shared values. These concepts may seem distant from a world dominated by metrics, KPIs, and regulatory requirements, but they can be powerful tools for engaging stakeholders.

This session explores how the principles of soft power can transform TMF and stakeholder management. Through real-world examples, it will illustrate how collaboration, attentive support, and consistent encouragement foster stronger engagement, higher-quality documentation, and healthier TMFs.

Soft power alone, however, is not sufficient. Its counterpart—hard power—uses data, reporting, and regulation to enforce compliance. The discussion will conclude by demonstrating how a balanced combination of both approaches, known as smart power, can create a more effective and sustainable TMF management strategy.

As organizations adopt digital Trial Master File (TMF) models aligned with the TMF Reference Model, many struggle to achieve consistent use across trial teams despite significant technology investment. This session shares practical lessons from regulated life-sciences implementations where digital TMF platforms were supported by structured change management and embedded, in-application enablement.

The presentation explores common adoption challenges such as unclear ownership, inconsistent filing behaviors, and reliance on one-time training. It demonstrates how role-based digital guidance, contextual support at the point of work, and usage insights can reinforce correct behaviors and improve TMF quality over time. The session also highlights how intelligent automation and targeted governance help reduce manual effort while maintaining inspection readiness.

Attendees will gain practical insights into how people, process, and technology must be aligned to move TMF from a compliance exercise to a trusted operational asset, supporting both regulatory expectations and day-to-day trial execution.

Keeping People in the Loop: Empowering TMF Teams Through AI and Automation

As AI and automation transform the TMF landscape, our focus is on empowering people, not just improving technology. At Regeneron, automating repetitive tasks—like workflow initiation, data entry, and routine checks—freed our TMF team to take on more meaningful roles, such as managing unique study TMFs and providing adhoc support for functional areas. Through targeted upskilling, proactive change management, and a culture that values experimentation, automation became a tool for growth. This presentation shares how we identified automation opportunities, coached associates for expanded capabilities, and fostered a mindset that embraces technology as an enabler.

The results: greater engagement, stronger TMF ownership, and a team ready to innovate. This is a people-driven transformation, redefining what it means to be a TMF professional in an AI-enabled world, offering attendees actionable ideas for building their own human-centered automation strategies.

Sponsors face inconsistent eCOA vendor outputs and instrument variability that slow SDTM delivery. At argenx, we are building a streamlined process that transforms raw eCOA data into SDTM domains (QS, FT, RS) using (1) vendor/data-stream-specific DTA templates to allow the delivery of the data always in the same format, (2) concept dictionaries for QRS controlled terminology, and (3) a mapping engine in LSAF that transforms the data and performs basic quality checks. This design enables the reuse across instruments and CROs while maintaining traceability to the source. Our goal is to improve consistency and reduce the cycle time.

In our presentation, we will share our governance model, provide examples of mappings and how it is created, discuss lessons learned during the process, and explain strategies for implementing a 360i-style automation.

In 2025, Novo Nordisk underwent a series of process reengineering amid the rollout of a new data infrastructure and achieved a new record: SDTM available at FPFV+1 day. By leveraging CDISC standards (CDASH and SDTM), biomedical concept and controlled terminologies, we managed to attain 100% standardisation with full data lineage and end-to-end automation, from EDC to SDTM.

In this presentation, we will share what we have learned – both successes and challenges – in the past year and discuss the strategy for operationalising the implementation across therapeutic areas and studies.

SDTM datasets are standardized by structure, yet complete and consistent representation of source data often requires additional review. In the absence of end-to-end automation, we present S3C, a lightweight oversight tool that conducts predefined RAW-To-SDTM spot checks and real-time comparisons to assess data inclusion and transformation fidelity. Rather than duplicating conformance engines, S3C targets completeness and consistency, accelerating triage while preserving expert review for study-specific derivations.

S3C is powered by crowd-sourced, machine-readable YAML rules that encode reusable checks and mappings. This community-driven catalogue enables rapid iteration, transparency, and portability across studies, while supporting decentralized contributions from domain experts within our TA-Area. Visualizations highlight domain and variable-level differences; SQL-style filters enable targeted subsetting; a side-by-side viewer aligns records and supports filtered exports for traceable sharing.

This pragmatic approach complements existing validation practices, reduces repetitive review effort, and offers an adaptable framework for oversight of vendor delivered SDTM across diverse study contexts.

CDISC Open Rules is redefining how data conformance validation is performed across the industry. Significant progress has been achieved on both the rules and the CDISC Open Rules Engine (CORE), with major refactoring efforts focused on improving performance, scalability, and long-term sustainability. This presentation will provide a comprehensive update on the current state of the initiative: from the rule authoring process and completion status to the engine and the technical enhancements implemented so far. The roadmap ahead will outline the plans for expanding rule coverage, governance, and innovations that will enable broad adoption through open-source collaboration. Discover how these developments will impact your validation workflows and learn how you can contribute your expertise to help shape the future of CDISC Open Rules as the open-source reference implementation for conformance rules.

The CDISC CORE application plays a critical role in validating clinical study data against regulatory standards, yet traditional deployment models limit its scalability and operational efficiency. Current machine based deployment poses challenges especially from validation perspective for closed systems like SAS Life Science Analytics Framework (LSAF). This paper presents a practical, end-to-end deployment strategy that transforms CDISC CORE from a regular machine–based application into a cloud-native REST API hosted in cloud environment like Azure Container Apps. It walks thru end-to-end process from initial deployment on an virtual machine using Python, followed by containerization using an optimized Docker image to deployment in Azure Cloud. Along the way, we discuss common challenges encountered—such as dependency compatibility, memory constraints, and runtime configuration—and the architectural decisions used to address them. The resulting solution enables a stable, scalable, and reproducible REST API endpoint for CDISC CORE, suitable for enterprise and regulatory workflows.

CDISC dataset validation is a critical requirement for regulatory submissions in clinical trials. CORE provide standardized compliance checks; however, study teams often require additional custom validations to address protocol-specific data quality needs. We developed an R Shiny–based validation framework that integrates execution of CDISC Open Rules using core.exe with study-specific custom rules. CORE is executed directly from R for standards-based validation, while custom rules are categorized as simple or complex. Simple custom rules, including dataset existence, variable existence, outlier detection, and dependency checks, are configured and executed dynamically through the Shiny dashboard. Complex custom rules involving multi-variable logic or cross-domain dependencies are defined and implemented using validated R scripts. Validation results from CDISC Open Rules and custom rules are consolidated into a single report. Visualization of dataset status by errors and warnings improves interpretability and supports efficient, transparent data review while maintaining alignment with CDISC standards.

The imminent release of the CDISC SDTMIG v4.0 / SDTM v3.0 represents a pivotal moment for SDTM standards. As a major version update, it introduces significant changes that sponsors should begin preparing for now. Key updates include support for Multiple Subject Instances (MSI), the transition from SUPPQUAL to Non-Standard Variables (NSVs), and metadata restructuring aligned with SDTM Model v3.0. The –BLFL variable has been removed and Sections 1–4, which provide guidance across SDTM domains, have been reorganized for improved clarity and usability. Variables are now organized into Variable Groups, enhancing structure and interpretation. New domains such as Event Adjudication (EA) and Gastrointestinal Findings (GI) further expand the ability to handle more types of data. Also noteworthy are updates to the Protocol Deviations (DV) domain and specimen-based Findings domains. This paper will explore what’s new in SDTMIG v4.0 and provide practical guidance to help sponsors navigate this era of transition.

Cytogenetic results are central to diagnosis, risk assessment, and tracking treatment response in leukemia and lymphoma. Using the SDTM Genetics Findings (GF) domain standardizes these data but is challenging across karyotypes, FISH panels, fusion transcripts, copy-number changes, and multi-specimen longitudinal workflows. We defined practical GF mapping rules and partnered with clinical teams to map each result to GF parameters, aligning test methods and controlled terminology with CDISC codelists to establish a standard approach for hematology studies. Integrating GF improves data quality, interoperability, cross-study comparability, and reviewer efficiency in hematology trials.

In some cases, participants use a diary to record their dosing information each day. Managing this data within the EC presents several challenges. We plan to tackle some of these issues and explain our approach in the SDTM.

A key challenge was the lack of visit-specific details for diary entries. This made it difficult to match scheduled and actual records. We also dealt with incomplete data on drug intake. In titration cases, planned doses were logged weekly as single records, while administered doses were recorded daily at each time point. Additionally, the protocol used different units for planned and actual doses, which required careful handling.
These are a few challenges we faced. We want to discuss how we addressed these issues in the EC and EX during the presentation.

Placeholders have long served as markers of expected documentation, yet their growing automation exposes both their value and their vulnerability. This session explores how TMF RM v4 and emerging technologies are reshaping the concept of expectedness, moving from static lists toward dynamic, metadata-driven models. We’ll examine the regulatory basis for placeholders, common missteps that create a false sense of completeness, and strategies for maintaining integrity as automation increases. Attendees will leave with a practical framework for balancing efficiency with accuracy—ensuring placeholders continue to support quality, not obscure it, as TMF management enters an era of intelligent expectedness.

In the world of Trial Master File (TMF) management, many organizations are stuck in “auto mode,” snapping study-level metrics without adjusting the focus. But just like a photographer who only zooms in on one subject, this narrow view can miss the bigger picture. To truly capture operational excellence, we need to switch lenses—zooming out to a portfolio-level perspective that reveals patterns, inefficiencies, and opportunities hiding in plain sight.

This session will consider how organisations can reframe their TMF strategy by developing a wide-angle view across studies. We’ll explore how portfolio-level insights can expose recurring quality issues, streamline resources, and sharpen inspection readiness. With real-world snapshots and practical tips, you’ll learn how to use metrics to tell a cohesive story, align cross-functional teams, and bring your TMF into sharper focus. It’s time to stop cropping out the context—let’s develop a clearer picture of TMF success.

Document timeliness has long been considered a standard for assessing overall TMF quality, but is it truly the best measure? While timeliness provides some insight, there are significant caveats that must be addressed to ensure accuracy. For example, certain documents may be several years old, which raises questions about how timeliness should be calculated, should these documents be marked as overdue, or excluded from the overall metric?

Evaluating timeliness often leads to many “whys” and “whats,” making it a complex indicator. Perhaps it’s time to shift our focus toward milestone completeness, where defining which documents are expected to be filed in the eTMF according to established TMF milestones. This approach could provide a more meaningful measure of quality, where the focus is on having the documents when they are need, rather than simply meeting timeliness compliance.

Panelists:

• Chris Jones, Novartis

Bridging the gap between the TMF and the ISF and understanding the key components that are solely in the ISF vs. the TMF, as well as key structure updates that you need to know about to help you be ready for your sites using the ISF structure!

Update on ISF Initiative from the CDISC subteam.

Plans for TMFSMV1 and ISF Version 2.0

The disconnect between Investigator Site Files and sponsor TMF continues to create inefficiencies, inconsistencies and compliance risks in clinical trials. Fragmented systems, manual processes and limited visibility often hinder timely reconciliation and effective oversight.

This presentation explores how AI-enabled technology can help bridge the gap between ISF and sponsor TMF by supporting intelligent document classification, metadata harmonization and automated quality checks across organizational boundaries. AI-driven tools can assist in identifying missing, outdated or inconsistent documents, highlighting discrepancies between ISF and TMF while preserving clear roles, accountability and human oversight.

Through practical examples, we demonstrate how combining AI, standardized workflows and secure collaboration platforms improves timeliness, reduces manual effort and supports risk-based monitoring. Special focus is placed on regulatory compliance, validation, data integrity and site usability. When applied responsibly, AI-enhanced technology can transform ISF–TMF alignment from a reactive reconciliation exercise into a continuous, inspection-ready process.ncy and confidence in TMF health.

Successful TMF management is built on partnerships where Sponsors and CROs combine their unique expertise to achieve shared goals. By aligning early and respecting each organization’s strengths, teams can ensure inspection readiness while optimizing timelines and resources.
Collaborative strategies include:

•    Early engagement: Joint TMF planning during study start-up to align expectations and processes
•    Mutual respect: Sponsors bring institutional knowledge and oversight; CROs contribute proven systems and operational efficiency
•    Shared decision-making: Both parties evaluate customizations to balance value with simplicity
•    Leveraging combined expertise: Integrating CRO best practices with Sponsor-specific requirements for a tailored approach
•    Open communication: Regular touchpoints and transparent escalation pathways to maintain alignment
•    Flexible governance: Frameworks that honor standardization while accommodating legitimate study-specific needs

This session highlights real-world examples where collaboration—not compromise—turned TMF management into a strategic advantage, demonstrating that alignment benefits both organizations and elevates study quality.

The CDISC Open Rules project continues evolving, creating new opportunities for smarter data validation. The introduction of JSONata as a rule-writing language marks a significant step forward. This addition reflects a broader evolution from rules on tabular formats to rules on more complex, object-based structures such as USDM, enabling richer validation logic. Building on earlier experiences with custom rules for external data transfers and AI-generated YAML rules, we are now applying JSONata for non-tabular structures like Define.xml. In this session, I will share our experience and show how combining CORE’s capabilities with custom logic can accelerate adoption and improve compliance. Attendees will gain practical insights into writing JSONata rules, integrating custom rules, and envisioning the role of AI in future rule development.

CORE is not only a great tool for validating submission datasets for SDTM, SEND and ADaM using the CDISC and FDA rules, but it also allows organizations to develop their own "custom" rules.

For example, a CRO may have QC rulesets for each of its sponsor customers. Or a sponsor company has specific quality requirements that go beyond the CDISC and FDA rules, maybe even depending on the study or study type.

The presentation will show how such rules can be developed in YAML, either using a local installation of the CDISC "Rule Editor", or an open-source tool that we developed that directly communicates with the the CORE engine, and that has the advantage it can be tested on own datasets in either SAS-XPT or Dataset-JSON format.

Once internally approved, these "custom" rules can then be executed in CORE, either on a single-rule basis, or as a "custom local standard".

When working with standards, one must integrate a multiplicity of information sources: SDTM, IG(s), Controlled Terminology, codetables, CT Relationships, and the interpretation of the text in the standards. But the availability of these standards in a computer-readable form is currently limited: within CDISC sources, we mainly have the CDISC Library, which provides tables, variables, codelists, and CORE rules.

We present a richer model of the standards, which would make it possible to simplify the CORE rules by embedding more information directly into the structure of the model itself. We also propose a simpler and more explicit model for specifying and implementing these CORE rules.

Sometimes, in a clinical study, there may be data collected on persons who are not the subjects of the study. How to handle such data in SDTM is described in the SDTM Implementation Guide for Associated Person (SDTMIG-AP). For ADaM, however, there is no comparable guidance available, except for one use case, which is slightly hidden in a Therapeutic Area User Guide (TAUG).

Building from this use case, this presentation will explore the different roles associated persons data can play in a study analysis, aided by examples for the different scenarios. Best practices for handling associated persons data in ADaM will be suggested under consideration of the ADaM fundamental principles, existing ADaM rules and general naming conventions.

A question also addressed in this presentation is where best practices like these could be found in the future ADaM documents ecosystem.

Managing subjects with multiple randomizations or screening attempts poses challenges for traceability and analysis. To address this complexity, Chiesi and SGS collaboratively implemented ADaM.ADPL (Participation-Level Analysis Dataset) providing a robust solution to ensure clarity and consistency in data handling. Here, ADPL provides one record per participation per subject, ensuring transparency while maintaining analysis integrity. A structured, well-documented process was established to guide programming teams on data selection and handling scenarios. This included decisions on selection of SUBJIDs in scope of analysis at SDTM level and the use of SDTM.XM for non-analysis SUBJIDs. ADPL serves as the source for ADaM.ADSL, enabling consistent derivation of subject-level data while excluding non-analysis participations from the analysis. This approach impacts downstream deliverables such as define.xml, SDRG, and ADRG, ensuring compliance and clarity in submissions. This presentation demonstrates how robust documentation and clear strategies can streamline the implementation of CDISC standards for complex enrollment patterns.

Pharmacokinetic Non-Compartmental Analysis (NCA) is a key step in evaluating drug exposure and disposition. This work summarizes and compares several studies implementations of ADaM datasets supporting NCA, covering the complete data flow from relevant eCRF actual data and bioanalytical analysis results to finalized ADaM datasets. It highlights the main challenges and best practices observed in dataset design, derivation, and traceability, with particular focus on alignment with the ADaM Implementation Guide for NCA.

The way of handling preclinical data aligns with SEND standardization. Starting from raw data collected by different laboratories, arriving at a submission package standardized and ready for FDA review. This work explains all the processes that end with the SEND package, with raw data, preclinical dashboard, and data standardization. The principal aim of our process is to help scientists make decisions using live data from all studies linked by a common compound.

The application of CDISC SDTM standards to medical device clinical studies remains limited compared to the pharmaceutical domain, despite increasing regulatory pressure under the EU MDR. Our work describes a practical, sponsor-driven approach to implementing SDTM in medical device trials, highlighting challenges, implementation strategies, and opportunities for innovation. Using real-world experience, we illustrate data flows from raw Excel/CSV sources to ODM generation and SDTM mapping, including adaptations of standard domains, sponsor-defined variables, and the use of the GF domain for emerging data types such as genetic information. The benefits observed include improved data quality, cross-study consistency, regulatory readiness, and long-term data reusability. The work also discusses how standardized, structured datasets can enable secondary use of legacy trials and provide a foundation for future AI-driven insights. This experience demonstrates how SDTM adoption can support more efficient, scalable, and innovation-ready clinical development in the medical device landscape.

Biomarker data is fundamental for understanding disease mechanisms, assessing treatment responses, and facilitating future exploratory research. Clinical protocols commonly include the following single sentence: "Biomarker data will be collected for exploratory research." Yet this seemingly simple phrase represents a sophisticated operational and scientific framework involving data generation, coordination across multiple vendors, and complex data processing workflows encompassing ingestion, standardization, analysis, and interpretation.
At argenx, we are co-creating a dedicated biomarker data team and robust process to support these data processing workflows.
This presentation aims to outline strategies and challenges identified in our journey. Focus is put on data acquisition (i.e. tailored Data Transfer strategies to accommodate diverse capabilities of our vendors), data transformation and standardization (i.e. bridging scientific needs when they are ahead of CDISC SDTM), and data analysis (ADaM). We will conclude by outlining future directions to overcome current challenges and exploring opportunities for future automatization of the process.

ICH E6(R3) is now in full force across regulatory agencies, and with it comes new ways of working and thinking about clinical trials. Many of us have been flirting with risk-based approaches for our TMFs, but now with “risk” appearing 89 times over 86 pages, it’s time to really commit.

In this session, we’ll explore 5 methods for building a risk-based TMF approach, from determining which records to collect to leveraging the metrics your teams produce. We’ll cover how these methods apply across different study types, the role technology plays in enabling risk-based decisions, and the evolving relationship between sponsors and CROs. We’ll discuss implementation strategies, pros and cons, and a critical reality: while risk-based approaches can reduce workload and boost team confidence when executed well, they can actually increase burden when they fail. We’ll show you how to tip the odds in your favour.

Acquisitions often create complex challenges for TMF integration teams: inconsistent processes, regulatory deadlines, and records dispersed across multiple systems. Effective migration requires early assessment of study status, document location, and metadata to guide transfer methods and risk-based strategies. These strategies prioritize high-risk studies for early migration, apply risk ratings to critical milestones, and focus on quality checks of essential documents such as regulatory submissions and patient safety records. Additional measures include metadata-driven mapping, automated audits checks, and real-time dashboards to monitor progress and prevent gaps. Collaboration between the divesting organization and internal teams is essential to maintain compliance and data integrity. This session will share lessons learned from a recent acquisition, demonstrating how risk-based approaches, deliberate planning and targeted QC transformed a complex migration into a controlled, successful outcome. Implications for future work include expanding automation, refining predictive risk models, and developing industry-wide standards to streamline TMF migrations.

Risk-based quality management is a core expectation in clinical research, yet Trial Master File (TMF) review practices often remain static, focusing on the same documents at fixed intervals regardless of evolving study risk. This approach can overemphasize completeness while missing early signals related to oversight, execution, and compliance.

This session presents a data-driven framework for advancing risk-based TMF review beyond periodic completeness checks. Building on established risk signals, the approach integrates organizational TMF risk, study-level risk identified in key plans and KPIs, and site-level indicators to dynamically prioritize review focus. Rather than reviewing the same artifacts repeatedly, TMF reviews are targeted toward areas where risk is greatest and where evidence of planned oversight should be present.

Attendees will gain insight into how risk signals can be translated into actionable TMF review strategies, improving traceability between identified risks and TMF evidence, strengthening inspection readiness, and enabling more adaptive, risk-aligned oversight.

As Zone Lead for Zone 10 (Data Management) during the transition to TMF Standard v1.0, I led an initiative defined by strong volunteer engagement and cross-functional cooperation. Collaboration with the triage committee supported alignment with governance principles throughout decision-making. Our work began by harmonizing terminology across organizations to ensure a common language as the basis for subsequent decisions. A major achievement was distinguishing data management from system-related activities, enabling partnership with the CSV group to create a dedicated space for system validation records and optimize record types. We invested significant effort in restructuring Zone 10 by reorganizing record groups, removing misleading definitions, eliminating outdated record types, and fostering open discussions on accountability for zone ownership. These efforts positioned Zone 10 for a streamlined, future-ready TMF framework that promotes consistency, efficiency, and collaboration. While this work is still in progress, additional insights will be available by the congress in May.

The CDISC Open Rules Engine (CORE), initially designed to validate CDISC-standard datasets such as SDTM and ADaM, offers a flexible framework for automated data quality checks. This paper presents CORE adaptation for the electronic Trial Master File (eTMF) Reference Model to address persistent challenges in manual document review. At Enovalife, we customized CORE to perform automated quality control across the eTMF, detecting issues such as incomplete metadata, misfiling, duplicates, non-standard naming, etc. By integrating Sponsor-specific rules, the system produces rapid, consistent, and comprehensive conformance reports. This approach significantly improves efficiency, standardization, and accuracy, reducing manual workload and compliance risks. Beyond ensuring regulatory alignment, the customized CORE enables continuous, proactive quality oversight—transforming eTMF review from a periodic compliance activity into a strategic process that enhances operational performance and inspection readiness across the clinical trial lifecycle.

As organisations adopt metadata-driven Statistical Computing Environments (SCEs), the next frontier for CDISC standards is capturing analytical provenance. While CDISC foundational standards, Define-XML, and the evolving Analysis Results Standard (ARS) provide robust traceability from data collection to analysis results, they do not describe how outputs are produced—nor the workflows, code versions, environments, and quality checks involved.

This presentation proposes a CDISC-aligned provenance layer, extending ARS and supporting the CDISC360i vision of end-to-end automation. The model formalizes heterogeneous SCE audit-trail elements (workflow steps, execution metadata, environment details, reviewer actions, and AI-assisted components) and links them to specific ARS outputs.

Standardizing provenance would bridge human and machine interpretability, support automated audit readiness, and enable reproducible evidence generation across tools and vendors. This will also expand CDISC’s mission from data traceability to evidence traceability in an increasingly AI-assisted analytical landscape, with a design extendable to other clinical systems.

Efficient and compliant database setup is critical to ensure high-quality clinical trial outputs and meeting regulatory expectations. At SGS, we leverage open-source technologies to streamline processes from study start-up through submission, focusing on standards- driven automation rather than traditional manual approaches. Using Python, we have developed reusable libraries for SDTM conversion and form mapping, significantly reducing programming efforts and accelerating timelines. These standardized components, fully aligned with CDISC requirements, ensure consistency, scalability, and reusability across studies. This presentation, will share our experience with using Python for conversion setup as well as integrating Azure DevOps and Git for version control, enabling traceability and continuous improvement. Attendees will gain practical insights into the flexibility of open-source tools driving automation and efficiency throughout the clinical data lifecycle.

CodeX accelerates creation of high-quality metadata when new concepts for clinical trials are not yet defined. The tool integrates with the OpenStudyBuilder API to view existing concepts to avoid redundancy and to save new concepts directly in the library. Starting from preliminary inputs from subject-matter experts, it identifies concepts by searching CDISC Biomedical Concepts, CDISC Controlled Terminology, and the NCI Thesaurus. When a match exists, the preferred name and definition are combined with business rules and SDTM domain-specific rules to auto-generate SDTM metadata, Schedule of Activities name, and ADaM PARAM/PARAMCD. If no match exists, AI agents suggest a draft name and definition based on the content in the dictionaries. By replacing complex written naming conventions with guided automation, CodeX accelerates and improves metadata creation in a process that is easy to learn and update as standards evolve. It reduces redundancy resulting in faster trial metadata setup and reduced standards maintenance.

This presentation introduces a modern approach to specifying AnalysisData Model (ADaM) datasets using YAML files instead of a traditional Excel-centered approach. The {mighty.metadata} R package provides thetooling to work with these YAML specifications, offering an open source solution for ADaM dataset specification and documentation.This approach delivers two key advantages. First, YAML's readable syntax makes specifications easy to work with for both humans, machines, and AIagents. This enables downstream automation processes, including define.xml and analysis script generation. Second, version control enables teams to track changes, work together effectively, and maintain complete specification history. The presentation will demonstrate how to use YAML and the{mighty.metadata} package to specify ADaM datasets. I will walk through the process of defining a dataset, its columns, and associated BDS parameters using YAML syntax and {mighty.metadata} functionality.

Estimands have transformed the way clinical trials define and interpret treatment effects, ensuring alignment between objectives, design, and analysis. This presentation explores the foundational principles of Estimands, drawing on key insights from PHUSE papers and industry guidance.

We will demonstrate practical applications using ADaM datasets, including ADEFF and ADICE, showcasing examples that illustrate how Estimands can be implemented effectively to address intercurrent events and improve clarity in trial results.

Looking ahead, we will discuss emerging challenges, complex summaries, and opportunities for further refinement of Estimand strategies.
The goal is to gain a deeper understanding of current best practices and future directions, ensuring everyone is equipped to apply Estimands with confidence and precision in an evolving regulatory and scientific landscape.

ADaMIG v1.2 introduced the concept of Pre-ADSL, an ADaM OTHER dataset intended for use when certain ADSL variables are better derived in other ADaM datasets. Our collaborator adopted this concept and formalized it as ADCORE in their ADaM standard.
We share our experience from a study where ADCORE was the first created ADaM domain and included all standard ADSL variables except those based on efficacy endpoints, which were derived using downstream efficacy ADaMs. We reflect on past studies where ADCORE was not used but, in hindsight, we feel it would have improved workflow and traceability, and we present a case where its use was limited - specifically, when ADSL analysis set flags needed early availability, but derivations depended on complex relative dose intensity calculations. This presentation aims to provide a balanced perspective on integrating ADCORE into ADaM workflows and when its use may - or may not - be beneficial.

CDISC biomedical concepts combine all the features of an assessment. For example, the glucose biomedical concept includes the potential food/fast state, the methodology, the specimen, the units and more. However, when instantiated for a specific study, not all potential properties are allowed. Features like the assessment method, vital signs position, and lab specimens are usually fixed.
To ensure consistency between studies and to improve efficiency, it is key to store standard biomedical concept configurations in a standards repository. This allows systems and users to pick up the right configuration automatically and adapt if needed. We will showcase this with our new open-source tool in which the biomedical concept instantiations are retrieved from the CDISC library via its API, can then be configured, and finally exported in USDM format enabling direct reuse in clinical study designs.

CDISC standards define the semantics of clinical data, but the industry still lacks an agreed structural model for how metadata itself should be organized, validated, and related. Most organizations rely on templates, spreadsheets, or tool-specific models to represent SDTM, ADaM, and Controlled Terminology, leading to inconsistencies and challenges with reuse, governance, and automation.

This presentation introduces a schema-first approach to clinical metadata: a method that defines the structure, relationships, and rules governing metadata before creating study-level specifications or building tools. By modelling the “metadata-of-metadata” layer—domains, variables, codelists, terms, value-level metadata, and their relationships—organizations gain a consistent foundation that supports quality, versioning, and interoperability across studies and systems.

The talk demonstrates why this structural layer is essential for scaling metadata reuse, and how it complements, rather than replaces, CDISC standards by providing a reliable implementation framework for metadata-driven processes and automation.

In global clinical research, common data standards such as CDISC are essential for consistency, interoperability, and trust. Yet as standards are adopted across regions, well-intentioned local enhancements can gradually dilute their original intent, leading to inconsistency and reduced comparability.

This presentation explores the importance of respecting original specifications while still enabling innovation, drawing on real-world experiences and emphasizing the roles of governance, education, and shared understanding across cultures and disciplines.
To make this concrete, we conclude with an unexpected analogy: the original tiramisù recipe. Like a data standard, it has a precise specification, one that excludes alcohol. Nevertheless, international reinterpretations, especially popular among Belgian friends, often introduce rum or amaretto while keeping the same name.

In closing, we take a position in defense of tradition: in SDTM, non-standard variables should coexist via Supplemental Qualifiers rather than being mixed into standard domains, optional additions remain external, while the core stays intact.

The implementation of the European Clinical Trials Regulation (EU CTR) was intended to create a cleaner canvas for harmonized and streamlined clinical trial application submissions across the EU. While this new palette has brought brighter strokes of transparency and centralized processes, it has splattered unexpected challenges across the Trial Master File (TMF) landscape. Under EU CTR, certain records have shifted in scope, format, and availability, leaving TMF teams to fill in the blanks like a paint-by-number puzzle—striving for compliance and completeness without a fully defined outline.

Come see how the TMF community is mixing colors, layering strokes, and assembling a new masterpiece to resolve EU CTR record conundrums and bring clarity to the evolving TMF record management canvas.

The transition to ICH E6(R3) fundamentally redefines compliance, moving from static, document-centric checklists to dynamic, risk-proportionate record oversight. This shift creates critical vulnerabilities for organizations relying on legacy TMF models that cannot validate the "fitness-for-purpose" of modern, decentralized data systems.

This paper presents a strategic transition framework rooted in Quality by Design (QbD). The solution details a methodology for auditing system architecture against R3's Data Governance standards and redefining Standard Operating Procedures (SOPs) to distinguish "Essential Documents" from the broader "Essential Records." comprehensive vendor oversight models are also introduced to manage the data lifecycle across fragmented clinical ecosystems.

By prioritizing Critical-to-Quality (CtQ) factors over administrative completeness, this risk-based strategy ensures organizations achieve sustained inspection readiness. The result is a resilient TMF model capable of supporting complex trial designs while reducing non-value-added documentation burdens.

The electronic Trial Master File (eTMF) is the authoritative repository for essential clinical trial documents. Global trials and cloud based eTMF platforms create opportunities for productivity and collaboration — but also raise regulatory, privacy and national security concerns when personal data and regulated documents traverse or are accessible from jurisdictions with differing laws and threat profiles. The U.S. Cybersecurity Information Sharing Act (CISA) and recent national security approaches to protecting patient data from foreign adversaries (notably China) have created new expectations for how sponsors control access, storage location and transfer of trial related data. At the same time ICH guidance on re use of clinical data and data integrity expectations require that any controls preserve data provenance, audit trails and quality for regulatory inspection and secondary use. 

We will present GSK’s approach to securing eTMF systems to meet these obligations and highlight challenges or remaining gaps and development priorities.

Potential future standards, if correctly defined and implemented, have the potential to change the game for TMF Health and completeness. Study Management has its own data flows that include EDC, IXRS, RBQM, X11 and document-based processes in RIM and Medical Writing. An eTMF system must sit as part of the data flow between these systems and both electronic and document processes at each Clinical Site in order to be truly effective. For the TMF documents to be as contemporaneous as possible, the archive must be aware of the multitude of milestones and events that have occurred to ensure automated workflows are triggered on time. Since these systems are often multi-vendor, the right standards that promote clean integrations between heterogenous systems can make all the difference.

In this discussion, the audience will hear about challenges with different standards, and on-the-ground experiences with integrations between clinical systems.

This presentation explores the evolution of the Electronic Trial Master File (eTMF) from a passive repository to a dynamic study management tool. By examining the synergy between eTMFs and complementary clinical systems, we identify how integrated data flows can streamline oversight. We will analyse the critical gaps in current data exchange and compare the architectural benefits of single-vendor platforms against the flexibility of multi-vendor ecosystems. Central to this transition is the role of standardised APIs (Application Programming Interfaces) and the Exchange Mechanism Specification (EMS) in fostering interoperability. Attendees will gain insights into how standardisation facilitates seamless data movement, allowing the TMF to serve as a real-time indicator of study health rather than a retrospective archive. Ultimately, the session defines a roadmap for leveraging interoperability standards to enhance compliance and operational efficiency across the clinical trial landscape.

At the conclusion of a trial, there are two main considerations when it comes to the TMF. The first is whether these essential records are complete and accurate, the second is where these will be retained.

Retention regulations (EU CTR), requires sponsors to ensure records are available, accessible and usable for +25 years. It’s therefore crucial to ensure any quality issues are identified and resolved and a retention strategy is in place.

Hedian Records Management’s Sarah Hitching and Arkivum’s Chris Sigley will provide guidance on how to approach both of these in line with industry good practice.

Firstly, Sarah will explore the why, what, when and how of quality checking the TMF. Chris will then explore how these records should be retained.

Both will provide practical guidance on how to approach these challenges, including how a risk-based approach is essential for managing the TMF at the end of a trial.

Panelists: 

• Sarah Hitching, Hedian
• Traci Wendler, Genmab
• Cristina Iannaccone, SGS

The Digital Data Flow (DDF) Initiative is transforming clinical trials by enabling fully digital workflows starting with protocol digitization. This foundation paves the way for automated, dynamic, and higher‑quality study setup, boosting efficiency across drug development. With USDM v4.0 and the M11 Guideline, Clinical Implementation Template, and Technical Specification, organizations now have a harmonized framework for digitizing protocol content—but many still struggle with how to begin practical implementation.

Building on earlier USDM explorations—including NLP and LLM‑based extraction (Phuse EU Connect 2024, ML02)—we collaborated with data4knowledge (d4k) in 2025 to prototype a USDM‑aligned Schedule of Activities (SoA) based on the argenx Protocol Template. A digital SoA enables major efficiencies, from smoother IRB/EC submissions and better patient‑engagement tools to automated data agreements and rapid eProtocol or concept‑document generation. This presentation shares practical steps, lessons learned, and key achievements from our “Trial Participant” journey POC to accelerate USDM adoption.

Clinical protocol SoA content varies widely. Visit names, procedures, timepoints, and windows are inconsistent, obscuring intended design and complicating reuse by downstream consumers and across studies. “One SoA for All” introduces a layered abstraction converting bespoke tables into standards-aligned, digitized content:

•    A structural layer normalizes the temporal backbone across study types. 
•    An attribute layer codifies timing semantics - visit windows, cycle/day patterns, repeat rules, and epoch transitions - making them explicit and machine-interpretable. 
•    A value-set layer applies controlled terminology, removing ambiguous labels by providing consistent terminologies for SOA activities

SoA standardization paired with automation enables multiple efficiencies across clinical research stakeholders. The harmonized SoA aligns with USDM enabling digitized study design, facilitating protocol authoring, accelerating start-up, reducing amendments, and enabling automation at scale for Sponsors/CROs. Service providers will receive clear and consistent study design requirements. Clinical sites will have clear visit expectations, reduced interpretation burden, and fewer protocol deviations. 

Evinova applies the Unified Study Definitions Model (USDM) to transform unstructured protocol PDFs into USDM-compliant JSON via AI extraction, creating a machine-readable source of truth. This standards-first approach advances Digital Data Flow with explicit semantics for arms, visits, procedures, and endpoints.
In collaboration with a large biopharma partner, we piloted USDM-driven extraction and design in the Study Design and Planning workspace. Cross-functional teams co-authored executable designs and surfaced feasibility, site planning, and procedure burden early. Early outcomes across multiple studies indicate 40–60% faster study build times, reduced rework from clearer semantics, and better cross-functional alignment.

We will present non-identifiable evidence during the conference session, focusing on observed efficiency gains, quality improvements, and lessons learned from the collaboration. USDM reduces ambiguity, strengthens change control, and supports traceable transformations through standards-based automation and open-source integration, improving data quality, auditability, and the connection from protocol intent to operational reality.

Industry initiatives like Digital Data Flow (DDF) and standards such as CDISC USDM aim to streamline clinical trial data transformation, yet sponsors often implement these differently, creating interoperability challenges. A rigid approach cannot accommodate these variations. This paper introduces a Bring Your Own Model (BYOM) concept that enables a plug-and-play framework for managing diverse data models and flows while maintaining compliance and traceability. BYOM allows study teams to incorporate proprietary or extended models alongside standard ones, apply version control, and preserve metadata lineage from protocol through SDTM, ADaM, and TFL outputs. It is also helpful in adoption of evolving standards such as USDM and ARS. The approach supports write-once/reuse-many principles, integrates with digital protocol assets, and leverages AI-assisted mapping under governance for transparency. We will present architectural considerations and examples demonstrating how BYOM addresses the industry need for flexibility, accelerates study build timelines, and supports evolving regulatory and interoperability standards.

While Study Data Tabulation Model (SDTM) remains a core requirement for regulatory submissions, mapping data from multiple sources remains a largely manual and resource-intensive process that requires extensive expertise, quality control, and consistent application of standards across studies. While automation initiatives already exist, many fall short in handling nuanced cases or integrating with existing workflows.

This presentation addresses key challenges in SDTM generation, including variability in mapping decisions, limited reuse of standards, increasing metadata complexity, and traceability for regulatory review. We propose an approach where we train AI in advanced SDTM implementation scenarios and sponsor-specific nuances. This methodology, grounded on SDTM implementation guidance, controlled terminology, and metadata patterns, generates mappings and its rationales, provides confidence indicators and facilitates review by documenting mapping decisions.

This session will provide a view on improving the intelligence of the SDTM generation process by enhancing human expertise, while preserving sponsor-defined standards and maintaining regulatory traceability.

Generative AI approaches to SDTM mapping often focus on prediction accuracy but lack integration with conformance validation, leaving specification errors undetected until data collection. We describe a closed-loop framework that combines generative AI mapping with automated CDISC CORE validation using synthetic data. The proposed methodology extracts field metadata from CRFs as well as non-EDC data such as data transfer specifications and generates SDTM variable mappings using large language models grounded in SDTMIG specifications. The system validates mappings proactively by generating synthetic SDTM datasets and executing CDISC CORE conformance rules. Validation failures trigger an iterative refinement loop in which the AI corrects mappings based on specific CDISC CORE rule violations. Early results demonstrate the detection of logic-based derivation errors, context-dependent Value Level Metadata violations, and cross-variable inconsistencies (e.g., mismatched Test Codes and Units). This approach demonstrates the feasibility of shifting CDISC compliance early into the specification process.

Novo Nordisk is targeting a 40% reduction in programmer-to-study ratio for ADaM programming. Process analysis revealed that faster ADaM programming alone would not reach this goal; validation, standardization, and bookkeeping would also need optimization. Traditional generative AI approaches writing ADaM code do not solve this.

Instead, we are implementing the open-source “mighty” framework: pre-validated derivation-level building blocks combined with machine-readable study specifications to deterministically auto-generate validated ADaM programs and define.xml. Reusable components eliminate redundant validation while generating programs and define.xml from a single source minimizes bookkeeping.

Additionally, the atomic architecture enables targeted AI integration, with assistants handling specific, bounded tasks where they excel rather than attempting open-ended code generation.

In our talk, we present both the technical architecture of this open-source framework, as well as early implementation experiences from a flagship project.

Regulatory submissions are hindered by manual, siloed workflows that produce source data ill-suited for SDTM, leading to late-stage rework, double programming across SDTM–ADaM–TLF, and weak traceability. We present an LLM-enabled, end-to-end solution that streamlines clinical data capture and standards implementation. Part 1 standardizes eCRF design by aligning study requirements with CDISC Biomedical Concepts and approved data elements, leveraging the CDISC library and organizational repositories to ensure seamless raw-to-SDTM conversion. Part 2 automates standards-driven transformation from raw data to SDTM, SDTM to ADaM, and TLF generation, with automatic, auditable code creation. The platform establishes continuous traceability from protocol, SAP, and TLF shells to datasets and outputs, providing verifiable evidence that study requirements are met. This approach reduces manual effort, improves data quality, and enhances compliance and confidence in regulatory submissions.

Advance strategy and planning are key to a successful regulatory submission meeting next generation submission timelines. A customized submission plan considering quality, data dependencies, and standards, while meeting tight timelines of each delivery is critical in a readout to quickly deliver CSR TLFs, Narratives, ISS, CRT packages, and more for submission.

In the CVRM Therapeutic Area in AstraZeneca, we recently executed a drug project submission containing
•    Readout of multiple clinical studies, including a pivotal study with multiple data locks, and including ISS
•    Considerations and decisions for data adherence/documentation based on CDISC and regulatory guidelines
•    Submission to multiple health authorities: FDA, EMA, and others.
 

This presentation provides a deep dive into the planning and execution of a drug project, from first Phase III FSI to regulatory submission, focusing on key success factors and lessons learned across study conduct, strategy, and planning for successful and rapid regulatory submissions.

This presentation outlines a strategic framework for adapting an eData Submission package originally prepared for one regulatory authority to meet the requirements of another. The case study focuses on transforming validated datasets and metadata, compliant with FDA standards, into a submission aligned with Japanese regulatory expectations (pMDA). Key challenges include differences in validation engines, metadata documentation (e.g., ARM), and management of findings. The approach combined a structured gap analysis, remediation planning, and cross-functional coordination to ensure inspection readiness. Emphasis is placed on regulatory intelligence, meaning the strategic collection and analysis of regulatory information to harmonize requirements across agencies in different countries (e.g., FDA vs. pMDA). Additionally, flexible governance ensured processes were adapted without rebuilding from scratch, enabling reuse of validated components while meeting each authority’s expectations. This experience offers practical insights for Sponsors navigating global submissions, leveraging existing assets efficiently while maintaining compliance.

In 2023 we presented at PHUSE EU an experience with CBER review expectations for vaccine submissions, highlighting unanticipated requests for retrospective changes to SDTM and ADaM going far beyond data standards and FDA requirements.

Since then, industry initiatives presented at CDISC EU 2025 have emerged to improve shared understanding of CBER requirements and to facilitate cross-sponsor exchange of regulatory feedback and additional data requests.

This presentation provides an updated perspective based on three additional years of interaction with CBER. We describe the volume and nature of retrospective CBER requests affecting SDTM and ADaM structure, in some cases violating the standards, and downstream TFLs. In several cases, changes were requested after CSRs had been finalized, requiring substantial rework, re-validation, and regeneration of TFLs late in the review cycle.

With this presentation we would like to share of these requests and lessons learned, aiming to support sponsors and ongoing CDISC industry collaboration.

The FDA’s recent Federal Register Notice on the potential adoption of CDISC Dataset-JSON v1.1 signals a major shift in how clinical data will be exchanged and reviewed. Dataset-JSON is positioned as the successor to SAS XPT, offering a structured, machine readable standard with richer metadata and native compatibility with modern analytics languages including R and Python. This presentation covers why the transition is inevitable, what the new format enables, and how organizations should prepare now. Drawing on lessons from the PHUSE and CDISC joint pilot and informed by William Qubeck’s leadership of the original define.xml initiative, the session outlines the operational, technical, and regulatory impacts of moving to Dataset-JSON. Key topics include metadata alignment, integration with legacy pipelines, automation, traceability, and validation under GxP. Attendees gain a pragmatic readiness framework to apply immediately as regulators define the future of submission standards and organizations modernize roadmaps and systems.

Customer (TBD) and Veeva Systems (Jason Methia, VP, TMF Strategy) share how a standardized TMF foundation creates consistent, scalable processes across studies, partners, and regions. Learn how a standard-first approach enables TMF teams to move beyond manual oversight to adaptive compliance, where core TMF best practices help ensure trial documentation is compliant and inspection-ready.

Learn how to: 
- Standardize core metadata and document lifecycles to enable seamless data exchange and automation.
- Implement adaptive compliance to proactively maintain inspection readiness. 
- Balance industry standards with flexible configurations that meet unique operational needs.

In my experience, some clinical sponsors are just now catching up to the TMF process requirements of 2022, while CDISC is proposing a monumental change. My understanding is that european sponsors are even more likely to still have paper TMFs than in the United States. In this panel I would ask CDISC TMF leaders practical questions that I will gather from small biotech/emerging pharma companies that are only vaguely aware of the transition to the TMF Standard and currently have no transition plan. I'll ask questions like:

• Realistically, when should emerging biotechs start preparing for implementation of the TMF Standard and why
• Who should be trained
• How is CDISC preparing to roll out training
• How can people get involved in the roll out
• Has there been a poll on who in the industry is aware of the change and how they are receiving it?
• What challenges do you anticipate in the uptake of this change in the Clinical Trials community and what can we do as a community to help.

This presentation will give an overview of the different activities that are being undertaken or that are planned as part of the TMF Reference Model Standardization. A presentation of the Standardization roadmap will be followed by presentations of the different workstreams including DDF/ICH M11 Alignment, metadata model development, tools development and controlled terminology.

Panelists:

• Dr. Yuki Ando, PMDA
• Eftychia-Eirini Psarelli, EMA
• Dr. Torsten Stemmler, BfArM
• FDA Representative TBA
• EMA Representative TBA

Sujit and Jamie will announce the dates and location for the 2027 CDISC Europe Interchange, as well as the winners for Best Presentation and Best Poster for both the CDISC and TMF tracks.