Final reflection paper sets direction for regulatory use of RWD in non-interventional studies

The European Medicines Agency (EMA) has published the final version of its ‘Reflection paper on the use of real-world data (RWD) in non-interventional studies (NIS) to generate real-world evidence (RWE) for regulatory purposes’. The paper provides detailed methodological guidance for stakeholders involved in the design, conduct and assessment of NIS intended to support regulatory decision-making in the EU.

Developed by the Methodology Working Party (MWP) and adopted by the Committee for Medicinal Products for Human Use (CHMP) on March 17, 2025, the paper outlines key principles for generating reliable RWE from RWD, and complements the broader HMA EMA roadmap for regulatory guidance on RWE. As stated in the roadmap,

“The objective of the reflection paper was to discuss methodological aspects of NIS using RWD to generate RWE for regulatory purposes. The main scope of the reflection paper was design, conduct and analysis of NIS using RWD.”

The paper builds on a multi-stage development process, including a public consultation held between May 3 and August 31, 2024, which received nearly 700 comments from stakeholders across patient organizations, industry, academia, public bodies, and professional societies.

Scope and purpose

The reflection paper is intended for a broad range of stakeholders involved in NIS using RWD, including marketing authorization holders, applicants, regulatory authorities, health technology assessment bodies, payers, academic researchers, data holders, and organizations representing patients and healthcare professionals. It applies to studies conducted both within and outside the EU.

The paper focuses on the use of RWD to evaluate medicines in real-world clinical settings. Data sources may include electronic health records, insurance claims, registries, prescribing and dispensing data, patient-reported outcomes, and data from wearable devices. It distinguishes between studies with descriptive objectives, which aim to characterize populations or patterns of treatment, and those with causal objectives, which seek to estimate the effects of medical interventions on outcomes.

“Given the large amount of information that NIS using RWD can generate for regulatory purposes, it is important to understand their limitations as well as how some of these limitations could be overcome or mitigated to increase the reliability of the evidence,” the paper states.

Design and feasibility planning

The design of a study should be driven by a clearly defined research question and an assessment of whether RWD are appropriate for meeting the study objectives. The EMA explain that, “it is the Marketing Authorisation Holders (MAHs) and Applicant’s responsibility to justify that the use of RWD is appropriate and feasible to meet the pre-defined study objectives.”

A feasibility assessment is recommended before finalizing the protocol. This should consider study population characteristics, sample size, inclusion and exclusion criteria, data completeness, ethical requirements and the reliability and relevance of the proposed data sources.

Approaches to bias and confounding

Through the paper, the EMA provides extensive guidance on identifying and minimizing selection bias, information bias and time-related bias in NIS designs. Key recommendations include using incident user designs, validating algorithms, and aligning the timing of eligibility, exposure, and follow-up periods.

For studies with causal objectives, the target trial emulation (TTE) framework is recommended to strengthen internal validity. The paper explains that TTE, “provides a structured and coherent framework for the design of NIS with a causal objective,” and can improve the internal validity of results.

To assess residual confounding, the use of negative or positive control exposures and outcomes is encouraged. Potential effect modification should also be explored through stratified and sensitivity analyses, with comparisons to the target population to inform generalizability.

Legal and regulatory requirements

The paper confirms that all applicable EU legislation must be followed, including regulations on clinical trials, pharmacovigilance, and data protection under GDPR and EUDPR. It also references related EMA guidance documents, such as the Guideline on registry-based studies and the Data Quality Framework for EU medicines regulation.

Early dialogue with regulators is advised to determine whether an RWD based NIS is relevant and feasible for a specific regulatory application. The EMA emphasize that:

“The relevance of a NIS using RWD to generate RWE for a specific application can only be determined on a case-by-case basis.”

Governance, transparency, and data quality

The EMA underscores the importance of good governance and transparency throughout the study lifecycle. Study conduct should follow the ENCePP Code of Conduct and ISPE’s Good Pharmacoepidemiology Practices. Stakeholders are encouraged to register their studies and data sources in the HMA EMA catalogues, and to make protocols, study reports and statistical codes publicly available.

Data quality is addressed in terms of reliability and relevance. The agency recommends using established frameworks, such as the HMA EMA Data Quality Framework, to evaluate whether a dataset is suitable for regulatory use. When multiple data sources are involved, efforts should be made to assess heterogeneity and manage its impact on results.

Statistical considerations

Statistical analysis plans should be finalized before dataset preparation and should include all model assumptions. The paper encourages the use of effect estimation and confidence intervals rather than reliance on hypothesis testing. It also notes that, “an integrated evaluation is essential for any conclusion based on results,” considering both statistical precision and potential bias.

Time-dependent exposures, treatment switches and missing data require particular attention. Analyses should be aligned with established regulatory frameworks, including the ICH E9 (R1) addendum on estimands and sensitivity analysis in clinical trials to the guideline on statistical principles for clinical trials. Stratified and sensitivity analyses should be pre-specified where possible, and supplementary analyses may be used to explore data limitations or support interpretation.