A World Beyond Documentation: How CDISC European Interchange 2025 Revealed the Future of Connected Clinical Data »» Bollinside Programming Services

A World Beyond Documentation: How CDISC European Interchange 2025 Revealed the Future of Connected Clinical Data

As clinical data professionals gathered in Geneva for CDISC European Interchange 2025, the atmosphere carried a palpable sense of transformation. This wasn’t simply another standards conference filled with technical specifications and compliance checklists. Instead, presentations spanning three days underscored a fundamental paradigm shift: CDISC standards are evolving from static documentation requirements into dynamic, machine-executable infrastructure that enables true end-to-end automation.

The discourse revealed an industry at an inflection point—one where semantic interoperability, artificial intelligence, regulatory harmonisation and collaborative innovation are converging to reshape how clinical trials are designed, conducted and submitted. From 15-minute protocol-to-study builds to AI-assisted standards navigation, from real-world data integration to globally aligned electronic protocols, the future of clinical research is taking shape. And it’s more connected, intelligent and standardised than many anticipated.

The CDISC 360i Vision: From Fragmented Silos to Connected Ecosystems

The problem has been clear for years: clinical trial data flows through disconnected silos, with protocols interpreted manually into EDC specifications, raw data laboriously mapped to SDTM, analysis datasets derived through custom programming, and outputs recreated for each report. Each handoff introduces risk, delay and inconsistency. Each standard (CDASH, SDTM, ADaM, Analysis Results) has operated as an independent island with limited bridges between them.

CDISC 360i represents the industry’s most ambitious answer yet: a unified architecture where standards connect seamlessly from protocol design through regulatory submission. At its heart sits the Unified Study Definitions Model (USDM), providing machine-readable study definitions that can drive downstream automation. Surrounding it: Biomedical Concepts (BC’s) that provide semantic consistency, Dataset Specialisations that define how data should be structured, and Analysis Concepts that standardise analytical approaches.

The Geneva demonstrations made this tangible. Data4knowledge showed a legacy protocol converted to USDM, enriched with Biomedical Concepts, and transformed into a functioning study environment (complete with subjects enrolled and SDTM datasets generated) in just 15 minutes. What previously required weeks of specialist programming happened in the time it takes to grab coffee.

Multiple implementations are now live. Novo Nordisk presented their OpenStudyBuilder, which manages sponsor extensions to CDISC standards and supports USDM generation, all built on a knowledge graph foundation. GSK demonstrated their ARMADA (Analysis Results Metadata And Data Automation) framework, showing how analysis concepts enable write-once-read-many workflows across multiple outputs. OCS Life Sciences illustrated practical FHIR to SDTM mappings for Parkinson’s disease studies, demonstrating how Biomedical Concepts bridge real-world data sources with submission requirements.

The reported impacts are significant: organisations implementing connected metadata approaches cite 40-60% reductions in study startup time, elimination of redundant specification documents, and dramatic improvements in cross-study consistency. More importantly, the shift creates strategic capability rather than just operational efficiency. When protocols are machine-readable, when Biomedical Concepts define data semantically, when lineage is embedded rather than documented separately, organisations gain agility to respond to protocol amendments, regulatory questions and business needs with speed unimaginable in traditional approaches.

Yet the vision remains incomplete. Analysis Concepts are just beginning definition work (a working group formed in January 2025). Standards for advanced data types (multiomics, digital health technologies, medical imaging) are still maturing. And most crucially, the industry must solve the “people problem”: how to build the cross-functional understanding, governance structures and change management approaches needed to operationalise connected standards.

AI Integration: From Experimental to Operational

Artificial intelligence dominated discussions, but with a notably pragmatic tone. Gone were the breathless predictions of AI replacing programmers. Instead, presentations focused on specific, production-ready applications where AI accelerates human expertise rather than replacing it.

Standards navigation emerged as a high-value use case. AstraZeneca demonstrated an LLM powered system for searching their internal standards library, addressing a chronic pain point: programmers knowing standards exist somewhere but unable to find them efficiently. Their solution uses embeddings to retrieve relevant SDTM mappings based on natural language queries, dramatically reducing time spent hunting through documentation. SGS Pharma took this further, creating a chatbot that generates custom CDISC Open Rules in YAML format, complete with test data—converting what previously required deep technical expertise into a conversation.

Biomedical Concept generation showcased AI’s potential for accelerating standards development. Lindus Health, working with CDISC, demonstrated using AI to extract potential BC’s from Therapeutic Area User Guides and match them to NCI terminology. Whilst acknowledging current limitations (AI lacks inherent understanding of what constitutes a well-formed Biomedical Concept), their “human-in-the-loop” approach shows promise for scaling what has been painfully manual work.

Protocol analysis and study setup automation featured prominently. Argenx presented their “Ask Sandy” tool, which uses LLMs to answer questions about data standards specifications, complete with source citations. Oracle demonstrated using AI to help navigate post-go-live amendments in automated study builds, addressing one of the thornier problems in modern trial management.

Yet every presenter emphasised guardrails. Formation Bio’s “Human in the Middle” paradigm articulated what became a conference-wide consensus: AI must be explainable, its outputs validated, and its actions auditable. FDA’s evolving guidance on AI/ML in drug development reinforces this principle—automation is welcome, but human expertise must remain at the heart of critical decisions, particularly in GxP contexts.

The practical approach emerging: AI as a productivity multiplier for knowledge workers, not a replacement. It excels at pattern matching, code generation, and information retrieval. It struggles with conceptual reasoning, regulatory nuance, and quality judgement. The organisations seeing value are those deploying AI for well-scoped tasks (generate draft mapping, suggest test data, retrieve similar past studies) whilst maintaining rigorous human review before anything enters production.

Real-World Data Integration: Standards Meet Reality

The collision between pristine trial protocols and messy real-world data created some of the conference’s most honest discussions. As External Control Arms (ECA’s) become mainstream (particularly in rare diseases where randomised controls are ethically or practically impossible), the industry confronts data that wasn’t designed for regulatory submission.

Cytel’s case studies illustrated the challenge vividly: combining pivotal trial data with natural history studies from different regions, each with inconsistent collection methods, missing variables, and misaligned terminology. Their solution required extensive data harmonisation, propensity score matching across incompatible covariate definitions, and creative interpretation of CDISC standards never designed for such heterogeneity. The programming took 4 to 5 hours per figure to run, and regulatory authorities requested access to blinded data requiring secure third-party execution environments.

Yet patterns for success are emerging. The CDISC RWD Lineage project, presented by OCS Life Sciences, provides frameworks for documenting data provenance from source through transformation to submission. Their Parkinson’s disease use case demonstrated FHIR to SDTM conversion using Biomedical Concepts as the semantic bridge, showing how standards can accommodate real-world sources without compromising submission quality.

Multiomics data presents perhaps the ultimate standardisation challenge. Novo Nordisk’s presentations on proteomics and transcriptomics highlighted the complexity: millions of data points per subject, computationally intensive preprocessing pipelines, rapidly evolving technologies, and file formats alien to traditional CDISC structures. Their solution: BioCompute Objects for computational provenance, “stack datasets” splitting domains by category (XOP for proteomics, XOT for transcriptomics), and close collaboration with ISO TC/215 Genomics Informatics standards.

The European Health Data Space (EHDS) provides regulatory impetus. Catherine Chronaki’s presentation on the International Patient Summary for Research (IPS+R) showed how Europe is building infrastructure to enable EHR data flow into research, with FHIR and CDISC standards converging to support both primary and secondary use.

The consensus: real-world data isn’t going away. Standards must flex to accommodate it whilst maintaining the rigour that makes submissions credible. The industry is learning (sometimes painfully) how to bridge pristine standardisation with imperfect reality.

Regulatory Alignment and Global Harmonisation

Perhaps the most encouraging theme: regulators and industry are converging on common technical infrastructure rather than diverging into regional silos.

ICH M11 (Clinical electronic Structured Harmonised Protocol) represents the flagship effort. Nick Halsey from EMA provided a comprehensive update: the template finalised, technical specifications under consultation, and FHIR Implementation Guides in development through HL7 Vulcan. The vision is compelling—globally consistent, structured protocols that are both human-readable and machine-executable, feeding directly into USDM and enabling automated study setup.

The technical architecture is coming together: ICH M11 defines what should be in the protocol, CDISC USDM defines how to model it, and HL7 FHIR defines how to exchange it. HL7 Vulcan connectathons in Dallas, Atlanta and Madrid demonstrated progressively more sophisticated scenarios—from basic protocol elements to complex Schedule of Activities with timing and biomedical concepts fully defined.

FDA engagement accelerated notably. The precision FDA pilot, presented as part of the PRISM initiative, will test electronic M11 protocol submission workflows. The goal: sponsors submit structured protocols through open-source portals, FDA receives them in standardised format enabling automated dashboards and analysis—reducing approval timelines and reviewer burden simultaneously.

FDA Business Rules integration with CDISC Open Rules addresses a longstanding pain point. Previously, sponsors received validation feedback in narrative form, requiring interpretation and custom implementation. Now, executable rules capture FDA expectations in machine-readable format (YAML), can be run by anyone using the open-source CORE engine, and increasingly align with CDISC conformance rules. The Research Collaboration Agreement between CDISC and FDA CDER/CBER formalises this partnership.

Real-Time Oncology Review (RTOR) drives practical standardisation. Multiple presentations addressed how RTOR requirements (submission of top-line results and datasets immediately after database lock) necessitate high-quality, standards-compliant data from day one. The OCE/OOD Safety Team Standard Data Requests provide specific variable lists and derivation guidance, and organisations like SGS have built RTOR-compliant pipelines as their default approach.

Globally: Japan’s user network grows despite language barriers, EHDS creates Europe-wide infrastructure, and TransCelerate coordinates cross-company approaches. The trajectory is towards harmonised global standards, not fragmentation.

Automation at Scale: The Operational Reality

Theory became practice across numerous presentations demonstrating production-grade automation built on standards foundations.

Novo Nordisk’s 100% standardisation approach represents perhaps the most comprehensive implementation. Their journey: complete item-level standardisation in EDC using Biomedical Concepts, defined through both CDISC standards and internal extensions. Every EDC item embeds its SDTM target, mapping logic, and lineage metadata. The result: automated SDTM generation for 95-97% of variables, with persistent keys linking every SDTM record back to source. Their broader statistics: 70%+ ADaM automation, 60%+ reduction in study setup time, and elimination of the “lost in translation” errors that plague manual interpretation chains.

What enabled this? Moving standardisation as far left as possible—defining items correctly once rather than fixing mapping problems later. Breaking standards into modular, granular elements rather than monolithic forms. Establishing strong governance whilst maintaining flexibility through sponsor-defined Biomedical Concepts. And fundamentally rethinking the EDC design principle: instead of “design for sites,” design for end-to-end data flow with site usability as a constraint to optimise.

CDISC Open Rules advanced from pilot to production. Multiple organisations demonstrated integration into validation workflows—some using Pinnacle21’s validation engine, others building custom pipelines. SGS showed their automated validation toolkit that reads ADNCA datasets, executes CORE rules, generates conformance reports, and writes validated outputs to version-controlled repositories—all orchestrated through SAS/Python/Git integration.

Analysis Results Standard (ARS) implementations multiplied. CDISC’s new eTFL Portal, developed with Clymb Clinical, provides downloadable packages containing ADaM datasets, ARS metadata, analysis results datasets, and TFL displays—all standards-compliant and ready to use. AstraZeneca’s AZSOL (Standard Output Library) demonstrated how template-based TFLs, combined with automation tools like MOSAIC Biometrics, enable instant monitoring of standards adherence through traffic-light dashboards.

Metadata-driven approaches permeated presentations. Merck KGaA’s “Library Navigator” repurposed Define-XML to create browsable documentation for their entire standards library—CRF forms, SDTM specifications, completion guidelines—all hyperlinked and navigable like a website. OCS Life Sciences showed metadata-driven PK NCA workflows using the ADaM IG for Non-Compartmental Analysis, eliminating the manual specification documents that previously consumed weeks.

Collaboration as Competitive Advantage

The industry’s recognition of pre-competitive collaboration reached new maturity—moving from nice-to-have to strategic necessity.

The Vaccines Industry Standards Group (VISG), comprising seven major manufacturers, exemplified this shift. Médéric Celle (Sanofi) and Estella Sani (GSK) described monthly meetings where competitors openly share regulatory feedback, discuss implementation approaches, and align on interpretations of guidance documents. The value proposition: avoid redundant work responding to the same FDA comments, achieve consistency that benefits regulators and sponsors alike, and speak with unified voice when engaging agencies.

Why does this work? Standards implementation is genuinely pre-competitive—everyone must comply with the same requirements, and inconsistency benefits no one. The challenges are common (outdated guidance, inconsistent feedback, limited documentation), resources are constrained across all organisations, and regulatory relationships improve when industry presents coherent approaches.

CDISC’s own model evolved towards collaboration. The Biomedical Concepts curation team now includes volunteers from industry actively contributing definitions. The Analysis Concepts working group formed in January 2025 with representatives from pharma, CRO’s, and academia. CDISC Open Rules operates as truly open-source, with community contributions welcome and vendor implementations encouraged.

Academic-industry bridges strengthened. Prometrika demonstrated mapping REDCap data (used extensively in academic research) to SDTM, addressing the gap where 65.9% of academic research organisations surveyed had no CDISC implementation. Wakayama Medical University described establishing a CDISC team in Japan’s academic context, emphasising mutual support over isolation.

TransCelerate’s Digital Data Flow (DDF) partnership with CDISC exemplified cross-organisation alignment at scale—multiple companies jointly developing USDM specifications, reference implementations, and validation rules rather than each building proprietary solutions.

The insight: in a domain as technically complex and regulated as clinical trials, shared infrastructure and collective problem-solving create more value than proprietary differentiation on standards compliance.

The Human Challenge: Change Management and Skills

Technology proved easier than people—a theme echoing through implementation stories.

The skills gap is substantial and growing. Survey data presented showed only 34% of organisations have adequate CDISC expertise. CRO’s face particular challenges: Cytel’s Angelo Tinazzi described “governing the ungovernable”—balancing diverse sponsor requirements against regulatory expectations whilst building subject-matter expert capacity across hundreds of programmers. Their solution: lightweight but structured governance, ticketing systems for transparency, targeted SME support at critical study phases, and continuous knowledge sharing.

Training acceleration is underway but insufficient. CDISC launched 12 new courses in the past year, including the first Biomedical Concepts training delivered at this conference. Yet accessibility remains limited for academic institutions and smaller organisations. The Japan CDISC User Group’s growth demonstrates demand—SDTM team attendance jumped from routine meetings to 45 participants when hands-on learning opportunities emerged.

Organisational change management is make-or-break. Multiple presentations emphasised that technology implementation fails without stakeholder engagement, clear communication of benefits, robust support mechanisms (communities of practice, accessible subject-matter experts, updated training), and cultures where resistance is expected and constructively addressed.

The CRO dilemma encapsulates the challenge: how to maintain standardisation whilst accommodating legitimate sponsor-specific needs? How to build deep expertise when staff turnover is high? How to justify investment in standards when billability pressures dominate? Successful CRO’s are establishing dedicated standards functions, creating reusable intellectual property, and demonstrating that standards expertise reduces rework and accelerates delivery—making it profitable, not just compliant.

The consensus: organisations underestimate the human capital investment required. Standards transformation is a multi-year journey requiring sustained commitment, not a one-time project. Leadership must champion the change, fund the capability building, and celebrate the incremental wins that maintain momentum.

Emerging Frontiers and Future Directions

Several nascent areas signal where the industry heads next.

Analysis Concepts emerged as the critical missing link between USDM and Analysis Results Standard. The working group, co-led by CDISC and including volunteers from across industry, is tackling fundamental questions: What constitutes an analysis concept? How does it differ from derivation concepts? How should analysis methods, populations, and timing be standardised? Kirsten Langendorf (data4knowledge) presented early thinking: structured, reusable definitions of statistical approaches that enable automation, reduce ambiguity, and support regulatory transparency. The vision: statisticians select and configure analysis concepts just as data managers select Biomedical Concepts, creating machine-readable analysis plans.

Digital Health Technologies (DHT) demand new frameworks. CDISC’s DHT initiative, presented alongside DiME (Digital Medicine Society) collaboration, addresses wearables, continuous monitoring, and sensor data—sources generating orders of magnitude more data points than traditional eCRF’s. The approach: align CDISC Biomedical Concepts with DiME’s Library of Digital Endpoints and Core Measures, create implementation guidance for common scenarios (CGM, physical activity, sleep, heart rate), and develop provisional concepts for community feedback.

Multiomics integration challenges continue. Beyond Novo Nordisk’s proteomics/transcriptomics work, presentations highlighted metabolomics, epigenomics, and microbiomics, each with unique data volumes, preprocessing pipelines, and technical dependencies. The industry is learning that tabular SDTM structures may not suffice; solutions will likely combine traditional domains for high-level results with external files (stored with rich BioCompute metadata) for raw/intermediate data.

Infrastructure evolution accelerates: cloud-native architectures for standards repositories, graph databases for connected metadata (OpenStudyBuilder uses Neo4j), proliferating APIs (CDISC Library API, Dataset-JSON API, USDM API), and long-term digital preservation strategies addressing 25+ year retention requirements under EU CTR and other regulations.

Advanced analytics increasingly intersect standards: AI/ML model metadata, medical imaging provenance, neuroscience-specific data types (EEG, polysomnography), and the estimands framework requiring precise specification of intercurrent events and strategies.

The frontier is broad and active—standardisation is expanding, not contracting, in scope.

Practical Lessons and Actionable Insights

What should organisations do Monday morning?

For Leadership

Recognise standards as strategic capability, not compliance overhead. Organisations with mature standards infrastructure achieve 40-60% faster study startup.
Commit to multi-year transformation with sustained funding. Quick wins exist, but comprehensive change takes 3 to 5 years.
Invest in change management at least as heavily as technology. People and processes, not tools, determine success.
Participate in pre-competitive collaboration—join CDISC working groups, industry consortia, and user networks.

For Standards Teams

Establish governance before scaling. Lightweight structures (ticketing systems, clear approval workflows, regular forums) prevent chaos as complexity grows.
Engage stakeholders continuously, not just at milestones. Monthly touchpoints with programmers, data managers, and statisticians maintain alignment.
Build communities of practice internally. Experts isolated in standards teams create bottlenecks; distributed expertise scales.
Document decisions and rationale, not just specifications. Future teams need to understand the “why” behind choices.

For Programmers and Data Managers

Learn standards connections, not just individual pieces. Understanding how Biomedical Concepts link CDASH to SDTM to ADaM multiplies effectiveness.
Engage with CDISC volunteer opportunities—working groups, user networks, conference presentations. The cross-company learning accelerates growth.
Think end-to-end, not just your domain. The most valuable contributors see how their work enables (or blocks) downstream processes.

For Vendors and CRO’s

Build standards into products from inception, not as afterthoughts. Retro-fitting compliance is expensive and imperfect.
Provide configuration over customisation. Flexible, metadata-driven solutions serve diverse clients without forking codebases.
Invest in education services. Clients need implementation support and change management as much as software.

Quick Wins Anyone Can Pursue

Implement CDISC Open Rules for validation (free, open-source)
Adopt Dataset-JSON for internal workflows (future-proof, tool-agnostic)
Create standards library using Biomedical Concepts framework
Establish ticketing system for standards questions (transparency and knowledge capture)
Join CDISC user network for your region (free, immediate access to expertise)

Synthesis and Outlook

CDISC European Interchange 2025 marks an inflection point: standards have evolved from documentation requirements to executable infrastructure. The convergence of semantic interoperability (Biomedical Concepts), digital protocols (USDM/M11), regulatory alignment (FDA, EMA, ICH harmonisation), and enabling technologies (AI, cloud platforms, APIs) creates unprecedented opportunity for transformation.

Yet technology alone delivers nothing. Every successful implementation story from Geneva emphasised the same prerequisites:

Strategic commitment treating standards as competitive advantage, not compliance burden
Investment in people through sustained training, communities of practice, and career development
Collaborative mindset embracing pre-competitive cooperation and open-source contribution
Balanced innovation maintaining regulatory trust whilst accelerating technical progress
Long-term perspective recognising transformation as a multi-year journey requiring patience and persistence

The path forward is clear but demanding. Organisations that harmonise innovation with integrity, automation with human expertise, and efficiency with quality will define the next era of clinical research. The evidence from Geneva is compelling: the future of clinical data is connected, intelligent, and standardised.

The journey has begun. Will you lead, follow, or be left behind?

Further Resources

See more Trends & Insights