Real-World Evidence (RWE)

Real-World Evidence (RWE) is clinical evidence about the usage, benefits, or risks of a medical product, derived from the analysis of Real-World Data (RWD) — data relating to patient health status and care delivery that is collected outside the controlled structure of a randomized controlled trial (RCT). The two terms are related but not synonymous: RWD is the raw material (EHRs, claims, registries, patient-generated data); RWE is the clinical conclusion drawn from analyzing it once it has been assessed as fit for the question at hand.

Why RWE exists as a distinct FDA evidence category

Section 3022 of the 21st Century Cures Act (Public Law 114-255, enacted December 13, 2016) directed the FDA to establish a program to evaluate the potential use of RWE to help support the approval of a new indication for an already-approved drug, and to help support or satisfy post-approval study requirements. FDA responded with the Framework for FDA’s Real-World Evidence Program (December 2018), followed by a series of guidance documents on data standards and specific RWD source types, including EHR and claims data and patient registries. This gave RWE a defined regulatory pathway, separate from — not a replacement for — the traditional RCT pathway that remains the default standard for establishing efficacy.

How RWE differs from traditional RCT evidence

The distinction is about data provenance and study design, not data quality in the abstract:

• Collection context. RCT data are purpose-collected under a controlled protocol with prospectively defined endpoints, a case report form, and monitoring. RWD is collected as a byproduct of routine care or other real-world processes — clinical encounters, insurance claims, registry enrollment, device use — and only secondarily analyzed for research purposes.
• Randomization and blinding. RCTs isolate a treatment effect by randomizing patients to arms and often blinding assessment. Most RWE study designs (retrospective database analyses, prospective registries, external control arms) are observational and must instead rely on statistical methods to address confounding and selection bias, often within a formal target trial emulation framework that specifies, up front, the hypothetical trial the analysis is standing in for.
• Population. RCTs typically enroll a narrower, protocol-defined population under inclusion/exclusion criteria. RWD reflects the broader, more heterogeneous population actually receiving a product in practice, which is part of RWE’s value for post-marketing safety and effectiveness questions — but also part of why confounding is harder to control.

FDA’s own framing treats RWE fitness as a two-part test: whether the underlying RWD is fit for use (relevant to the question, reliable in its collection and provenance, and sufficiently complete) and whether the study design built on that data adequately addresses bias and confounding. Data that fails either test remains real-world data without becoming regulatory-grade real-world evidence.

Where FDA has accepted RWE

FDA’s Framework and subsequent guidance describe RWE use cases including: supporting a new indication for an already-approved drug, satisfying post-approval (Phase IV) study commitments, informing pragmatic trial design, and — particularly in oncology and rare disease, where a concurrent randomized control arm may be impractical or unethical — constructing an external control arm from curated registry, EHR, or claims data. RWE is used far more often to answer questions about long-term safety, comparative effectiveness, and treatment patterns after approval than to establish initial efficacy for a novel product, where RCT evidence remains the norm.

How RWD/RWE differs from digital twins and synthetic control arms

Real-world data is not the same as a simulated patient. RWD/RWE build regulatory evidence by curating and analyzing data collected from actual patients outside the trial — EHRs, claims, and registries — and then applying study-design safeguards (fitness-for-use assessment, bias and confounding adjustment) to that real data. A digital twin is a model-generated projection of an individual patient’s likely trajectory, typically used to augment or partially replace a concurrent control arm rather than to serve as the underlying evidentiary basis for a regulatory claim. FDA engages with both approaches but treats them as distinct methodological categories with different validation expectations. See CASRAI’s guide to digital twins in clinical trials for how synthetic control arms compare to RWD-derived external control arms in practice.

Related terms

See also non-interventional study, intent-to-treat vs. per-protocol analysis, and clinical outcome assessment (COA) for related distinctions in how clinical evidence is designed, analyzed, and categorized for regulatory purposes.