Real-World Evidence (RWE): how to generate regulatory-grade insights from Real-World Data (RWD)
Real-world data (RWD)—such as electronic health records, claims, registries, and data from digital health technologies—has become a strategic asset across the lifecycle of medicines and medical devices. The challenge is not access to data, but the ability to transform RWD into trustworthy real-world evidence (RWE) that can support decisions by regulators, payers, clinicians, and patients.
Recent international work highlights an urgent need for transparent, generally accepted processes that build trust in the evidence-generation pathway—especially as regulators increasingly consider RWE for effectiveness and safety questions beyond what traditional randomized controlled trials (RCTs) can answer.
Why RWE matters (and why RCTs are not always enough)
RCTs remain the gold standard for demonstrating efficacy, but they can be limited when studying rare adverse events, long-term outcomes, or real-world effectiveness in populations under-represented in trials. This creates an “efficacy–effectiveness gap”, where outcomes observed under controlled trial conditions differ from those seen in routine care. The growing use of RWE is, in part, a response to this gap—so decisions reflect how treatments perform in the real world.
From RWD to RWE: definitions that align stakeholders
Regulators define RWD as routinely collected data about patient health status and/or healthcare delivery from multiple sources (EHRs, claims, registries, digital technologies). RWE is the clinical evidence derived from analyzing RWD. These definitions are broadly consistent across agencies and are central to modern guidance.
Where RWE adds value across the product lifecycle
RWE can support decisions from early development through post-marketing, for example:
- Early development: understanding patient populations, care pathways, and unmet need
- Development and registration: contextualizing outcomes, supporting external comparators where appropriate, demonstrating representativeness
- Lifecycle management: long-term safety, effectiveness in routine practice, utilization, adherence, and outcomes in subgroups
International consensus work emphasizes that RWE is most powerful when it complements—rather than replaces—clinical trials, extending knowledge into settings and populations where trials are limited.
Step 1: Start with the right question (and define the “decision”)
Regulatory-grade RWE starts with clarity: What decision must be supported, for whom, and at what point in the lifecycle? A well-formed research question drives everything that follows: data selection, design, analysis, and reporting. This is why modern guidance increasingly distinguishes between descriptive studies (e.g., treatment patterns, burden of disease) and causal studies (estimating treatment effects).
Step 2: Choose fit-for-purpose data (relevance + reliability)
Not all databases are suitable for all questions. Fitness-for-purpose depends on whether the data contain valid measures of exposures, outcomes, and key covariates (including confounders), plus sufficient population size, representativeness, and follow-up. When one source is insufficient, using multiple sources (including federated systems) can increase sample size, broaden representativeness, and extend follow-up—provided that harmonization and governance are robust.
Regulators and health authorities have also invested in distributed/federated real-world data networks to enable analytics at scale (e.g., Sentinel in the US, CNODES in Canada, MID-NET in Japan, DARWIN EU in Europe). These networks rely on coordinating centers, data quality and governance models, and common data models.
Step 3: Design for validity (target trial emulation, estimands, and bias control)
The biggest credibility risks in observational RWE are bias and confounding. Modern pharmacoepidemiology increasingly uses “target trial emulation” to design RWD studies as if they were randomized trials—explicitly defining eligibility criteria, time zero, follow-up, endpoints, and analysis strategies. This approach clarifies critical choices (like index date and exposure definition) and helps prevent avoidable errors.
Additional frameworks strengthen causal interpretation, including:
- Estimands, to precisely define the treatment effect of interest and handle intercurrent events
- Bias prevention, including explicit safeguards against immortal time bias and other timing-related errors
Recent RWE reviews emphasize that using these approaches (alone or in combination) improves robustness and transparency, and is increasingly promoted in health technology assessment frameworks as well.
Step 4: Plan and report transparently (so evidence is trusted)
Trust is not only about methods—it is about process. A consistent theme across international reports is the need for structured planning, protocol discipline, and reproducible reporting. Tools and templates such as STaRT-RWE (for planning/reporting implementation) and HARPER (harmonized protocol template) support standardization and transparency.
Evidence-generation programs also increasingly emphasize feasibility assessment frameworks to determine whether a data source can credibly answer a question before a study begins—reducing waste and improving decision readiness.
Step 5: Address ethics, privacy, and governance from day one
Because RWD involves sensitive health information, ethics and governance are central to sustainable RWE. Key issues include appropriate consent approaches, privacy and data protection, transparency about data processing, and de-identification/pseudonymization strategies that protect individuals while enabling clinically meaningful analyses. Global harmonization remains a priority.
RWE in clinical development: “Hope is not a strategy”
A recurring operational lesson from industry experience: relying on expert opinion or historical patterns—without current, representative point-of-care evidence about the indicated population—can lead to suboptimal trial design, delayed programs, and missed opportunities. A practical blueprint is to integrate RWE generation early and phase it alongside development investment: characterize the disease and population, understand treatment patterns and monitoring, quantify outcomes/natural history, and then support trial-similar contextual cohorts.
The future: data science, privacy-preserving analytics, and fairness
RWD is messy, heterogeneous, and multi-modal, which is why data science methods are becoming increasingly important across the RWE pipeline—from preprocessing and record linkage to causal inference and reliability checks. The field is also advancing privacy-preserving approaches (e.g., differential privacy, federated learning, secure computation) and placing stronger emphasis on algorithmic bias and fairness, so RWE does not amplify existing health inequities.
Conclusion: what “regulatory-grade RWE” really requires
Regulatory-grade RWE is not a single method or dataset. It is a disciplined end-to-end approach: start from the decision, select fit-for-purpose data, design for validity (often via target trial thinking), execute transparently using structured protocols and reporting templates, and embed governance and privacy safeguards from the beginning. As guidance converges globally, organizations that institutionalize these practices will generate more credible evidence—and make better, faster decisions that ultimately improve public health.
FAQ (SEO)
What is the difference between RWD and RWE?
RWD is routinely collected health/healthcare data; RWE is the clinical evidence produced by analyzing RWD.
Can RWE replace randomized clinical trials?
Only in limited circumstances; most often RWE complements RCTs by extending evidence to real-world populations, long-term follow-up, and uncommon outcomes.
What makes RWE “trustworthy” for regulators and payers?
Fit-for-purpose data, rigorous design to minimize bias, and transparent planning/reporting (e.g., structured templates and reproducible protocols).
