Why Governance May Become the New Competitive Advantage in RWE
The publication of the European Medicines Agency’s new Data Quality Framework for EU medicine regulation: application to Real-World Data (RWD) marks an important moment in the evolution of Real-World Evidence (RWE) research in Europe. While much of the discussion around RWE has traditionally focused on analytics, study design, and causal inference, one section of the framework deserves particular attention for a different reason. Chapter 4, dedicated to the “characterisation of systems and processes underpinning RWD,” moves the regulatory focus upstream, toward the operational environments in which data are generated, transformed, and maintained.
At first glance, this may appear to be a highly technical topic, relevant mainly to data engineers or compliance specialists. In reality, Chapter 4 reflects a broader transformation in how regulators think about evidence generation. The EMA is effectively stating that the credibility of RWE does not depend solely on the statistical properties of a dataset, but also on the reliability and transparency of the systems producing that data. This is a significant conceptual change, and one that may have long-term implications for pharmaceutical companies, technology vendors, healthcare databases, and AI-driven research platforms.
Historically, many RWE initiatives concentrated on evaluating the quality of data once it had already been extracted and harmonised. Researchers examined completeness, coding accuracy, consistency, and plausibility, often treating datasets as relatively static assets ready for analysis. However, observational healthcare data emerge from complex operational systems involving hospitals, physicians, reimbursement systems, registries, laboratories, software vendors, and data transformation pipelines. Every step in that chain introduces potential variability, bias, or error.
The EMA framework recognises this reality explicitly. Chapter 4 emphasises that understanding how data are collected and processed is just as important as analysing the data themselves. In practice, this means regulators increasingly expect organisations to demonstrate traceability across the entire data lifecycle. Questions that were once considered secondary are becoming central to regulatory confidence: Who originally entered the data? For what purpose? How were records transformed during extraction and harmonisation? What governance procedures exist for correcting errors or managing updates? Can historical changes be reconstructed through audit trails?
This approach resembles the logic long applied in regulated manufacturing environments. In pharmaceutical production, quality is not assessed only by inspecting the final product; it is also ensured through validated systems, documented procedures, and controlled operational processes. Chapter 4 suggests that a similar philosophy is now being extended to RWD systems. The implication is significant because it reframes RWE credibility as an organisational capability rather than simply an analytical outcome.
One of the most interesting aspects of the framework is its implicit introduction of operational maturity as a determinant of data reliability. Not all RWD sources are equal, and the EMA appears to acknowledge that healthcare data environments exist along a spectrum of governance maturity. Some databases operate with highly standardised procedures, comprehensive metadata documentation, robust quality management systems, and sophisticated audit capabilities. Others rely on fragmented documentation, inconsistent coding practices, or poorly characterised transformation pipelines.
Rather than presenting quality as a binary concept, the framework encourages a more nuanced assessment of whether a system is sufficiently mature for a specific regulatory purpose. This is particularly important because RWD infrastructures are intrinsically heterogeneous. Electronic health records (EHRs), disease registries, insurance claims databases, and federated research networks all operate under different technical and organisational conditions. A rigid one-size-fits-all model would likely be impractical. By focusing on characterisation and transparency instead of imposing absolute uniformity, the EMA adopts a more realistic and scalable regulatory approach.
The consequences for industry may be substantial. For years, competitive advantage in the RWE landscape was often associated with database size or geographic coverage. Organisations promoted the number of patients available for analysis or the breadth of longitudinal follow-up. Chapter 4 introduces a different dimension of value: demonstrable governance quality. In the future, sponsors selecting data partners may increasingly evaluate not only the scale of a dataset, but also the operational robustness of the environment behind it.
This could reshape relationships between pharmaceutical companies and data providers. Vendors capable of demonstrating strong governance frameworks, documented Extract, Transform and Load (ETL) processes, version control, metadata transparency, and reproducible workflows may become more attractive partners for regulatory-grade research. Conversely, organisations unable to provide clear evidence of process reliability may struggle to support high-impact regulatory submissions, regardless of dataset size.
The framework also has important implications for artificial intelligence (AI) and advanced analytics. Although Chapter 4 does not explicitly focus on AI, its principles align closely with emerging concerns around algorithmic transparency and trustworthy machine learning (ML). AI systems are highly sensitive to upstream data quality issues. Biases introduced during clinical documentation, coding harmonisation, linkage procedures, or transformation workflows can propagate into predictive models and potentially distort their outputs.
In this context, system characterisation becomes essential not only for traditional epidemiology but also for AI governance. Regulators and stakeholders increasingly recognise that explainable algorithms alone are insufficient if the provenance of the underlying data remains opaque. A sophisticated ML model trained on poorly characterised data pipelines may still produce unreliable evidence. By emphasising traceability, auditability, and operational transparency, the EMA framework indirectly establishes foundational requirements for trustworthy AI-driven RWE.
There is also a broader regulatory trend visible behind Chapter 4. Across Europe, regulators are increasingly converging around principles of lifecycle oversight, reproducibility, and institutional accountability. Similar concepts appear in discussions surrounding AI regulation, digital health technologies, and data governance more generally. The EMA framework therefore reflects more than a narrow technical update; it signals a deeper cultural evolution in evidence generation.
Importantly, this evolution does not imply that RWD must replicate the rigid structures of clinical trials. Observational data will always retain certain limitations because they originate from routine care rather than controlled experimental environments. However, the framework suggests that regulators now expect organisations to understand and document those limitations systematically rather than treating them as unavoidable ambiguities.
Ultimately, Chapter 4 may become one of the most influential sections of the EMA Data Quality Framework precisely because it moves the conversation beyond datasets themselves. It asks organisations to demonstrate that their data ecosystems are understandable, governable, and trustworthy. In doing so, it elevates operational transparency into a core component of scientific credibility.
As RWE continues to gain importance in medicines regulation, this shift may redefine what it means for evidence to be considered “fit for purpose.” The future of regulatory-grade RWE may depend not only on methodological sophistication or analytical innovation, but also on the maturity of the systems and processes quietly operating behind the scenes.
Reference:
European Medicines Agency (2026) Data quality framework for EU medicines regulation: application to real-world data. Amsterdam: European Medicines Agency.
