The growing acceptance of real-world data (RWD) and real-world evidence (RWE) has expanded the horizon for how we generate insights to support decision-making across the product lifecycle. However, while the value of RWD is increasingly recognized by regulators, payers, and safety bodies, the design and methodological rigor of these studies must be carefully tailored to their intended use. Whether the goal is to support label expansion, strengthen value dossiers, or enhance safety monitoring, each application requires a distinct evidence strategy and level of robustness.
In the context of label expansion, RWD is increasingly being used to complement clinical trial evidence or, in rare cases, stand alone in submissions to regulatory authorities. This requires adherence to high standards of data quality, confounding control, and transparency to meet regulatory expectations, particularly when aiming for indication extensions or population expansion.
For value dossiers, especially those supporting health technology assessments (HTAs) and payer negotiations, RWD plays a vital role in demonstrating real-world effectiveness, healthcare resource utilization, and burden of illness. The methodological expectations in this space focus on relevance to local practice, representativeness of local populations, and transparency in cost modeling inputs.
Meanwhile, in safety monitoring, RWD studies offer the unique advantage of continuous observation over time in broad patient populations. These studies often prioritize rapid signal detection and long-term risk characterization, which places emphasis on data completeness, temporal alignment, and statistical power for rare events.
Despite the common use of RWD across these domains, the methodological choices must align with the evidence expectations of each stakeholder. Key decisions, such as cohort definitions, comparator selection, confounding adjustment methods, and endpoint operationalization, can have vastly different implications depending on the application. A "one size fits all" approach to RWD design not only limits impact, but can also undermine the credibility and utility of the findings.
In this article, I explore key methodological considerations in RWD study design across different intended uses, and share best practices for ensuring your study is fit-for-purpose, whether you are pursuing regulatory engagement, market access value, or post-marketing safety insights.
1. Label Expansion: Meeting Regulatory Expectations
When RWD is used to support label expansion, whether for a new indication, broader population, or new formulation, regulators expect the same scientific rigor as traditional clinical trials. Key methodological considerations include:
- Clear definition of treatment exposure and population: Precision is critical, especially when mimicking inclusion/exclusion criteria from prior RCTs.
- Robust comparator selection: External controls or historical comparators must be justified, well-matched, and relevant. Regulatory-quality evidence often centers on selecting active comparators with similar treatment intent. Comparator cohorts must be carefully constructed using similar eligibility criteria and baseline covariates.
- Control for confounding: Use of advanced methods is often necessary.
- Transparency and reproducibility: Full protocol registration, pre-specified analysis plans, and sensitivity analyses strengthen credibility.
- Fit-for-purpose data: High-quality, longitudinal data with validated outcomes and minimal missingness is critical.
Typical Data Sources: Integrated electronic medical records (EMRs), linked claims-EMR databases, disease registries, or multicenter healthcare networks with validated data quality controls.
2. Value Dossiers: Demonstrating Real-World Impact
RWD in value dossiers is used to contextualize clinical trial results, support budget impact models, and demonstrate cost-effectiveness. Key best practices include:
- Local Data Requirements: Many HTA agencies strongly prefer or require data sourced from their own country, healthcare system or region in which pricing/reimbursement decisions are made. For example, HAS (France) and NICE (UK) explicitly favor local data over international datasets.
- Local Standards of Care: Comparator selection must reflect real-world standards of care within the jurisdiction (e.g., UK, France, Germany), recognizing that therapeutic pathways and reimbursement rules vary widely.
- Relevant Outcomes: HTA agencies and payers are interested in pragmatic, patient-centered, and economic endpoints such as hospitalization rates, medication adherence, time to treatment failure, quality-adjusted life years (QALYs), and healthcare costs.
- Subgroup Analyses: Economic models often require understanding how treatment impacts specific patient segments. Subgroup data can reveal heterogeneity in value that informs budget impact models.
- Resource utilization data: Robust capture of medical costs, service use, and care pathways is essential to feed economic models.
- Modeling Inputs: Real-world inputs into cost-effectiveness models (e.g., transition probabilities, duration of effect, utility values) must be robust and reflect clinical reality.
- Transparency in assumptions: Clearly state modeling assumptions, extrapolation techniques, and justify parameter choices with evidence.
Typical Data Sources: National claims data, administrative healthcare databases, registry-EMR linkages, national health surveys with economic indicators, and payer-held datasets.
3. Safety Monitoring: Supporting Lifecycle Risk Management
Post-authorization safety studies (PASS) using RWD allow sponsors and regulators to monitor risks in broader, more diverse populations. Key methodological priorities include:
- Temporal precision: Accurate capture of timing between exposure and adverse events is crucial to assess causality.
- Signal detection methods: Employ epidemiologic techniques suitable for low event rates or rare outcomes.
- Data linkage: Where appropriate, link claims, electronic health records, or registries to enhance completeness of safety data.
- Ongoing analysis plans: Incorporate interim analysis, periodic safety updates, and automated monitoring tools where feasible.
- Timeliness: Safety monitoring requires near-real-time or periodic updates, particularly for high-risk medications or newly launched products. Infrastructure for routine surveillance is essential.
- Exposure Assessment: Understanding dose, duration, switching patterns, and co-medications is crucial. Misclassification of exposure is a key risk.
- Outcome Validation: Adverse events often require validation via chart reviews, physician adjudication, or algorithm refinement.
Typical Data Sources: Spontaneous reporting systems (e.g., FAERS, EudraVigilance), longitudinal claims databases, EMRs, Sentinel-like systems, and disease-specific registries.
Conclusion: Elevating RWD to Decision-Grade Evidence
As the boundaries between clinical research, regulatory science, and market access continue to blur, the role of RWD is expanding, and so are the expectations. Whether the goal is to secure a label expansion, demonstrate real-world value to payers, or monitor long-term safety, RWD must be designed with intentionality, transparency, and scientific rigor.
A well-executed RWD study is not just a dataset with statistical outputs; it’s a strategic asset that speaks the language of regulators, payers, and clinicians. But for it to deliver that value, it must be tailored to its intended use, with design choices that reflect the target audience. Misalignment between the question, the data, and the method can lead to irrelevant or misleading analyses.
As agencies like the FDA, EMA, and HTA bodies refine their guidance on the acceptable use of RWE, sponsors and researchers must embrace fit-for-purpose design principles. This means investing early in methodological planning, choosing data sources wisely, and proactively addressing limitations.
Ultimately, the success of RWD studies lies in their credibility and relevance, two qualities that can only be achieved when methodological decisions are shaped by the purpose they serve.
