SPECIALTY TOPICS · GUIDE
Real-World Evidence: What It Can and Cannot Show
What regulators mean by real-world data and real-world evidence, how it differs from a randomized trial, and what a 32-trial comparison found about when the two agree.
Content last updated . This page has not been medically reviewed.
Back to Emerging Wellness ScienceA phrase that gets used loosely
"Real-world evidence" sounds like proof from everyday life. The term has a narrower meaning, and the difference matters when a clinic, company or article uses it. This page explains the term, what it can answer, where it falls short and how it compares with a randomized trial. It is education only and is not advice about any treatment.What the words mean
The U.S. Food and Drug Administration (FDA) defines real-world data as data about patient health status or the delivery of health care that are routinely collected from a variety of sources, for example electronic health records, medical claims, product or disease registries and digital health technologies [1]. Real-world evidence is the clinical evidence about the usage and potential benefits or risks of a medical product that comes from analysis of real-world data [1].So real-world evidence is the result of a planned analysis, not the data themselves. By this definition, a collection of patient testimonials or a clinic's selected success stories is not real-world evidence. That is our plain reading of the definition, not an FDA statement.
Randomized trial versus real-world study
| Randomized controlled trial | Real-world study | |
|---|---|---|
| Who gets the treatment | Assigned by chance, which balances prognostic factors on average [2] | Decided in everyday care, often influenced by patient factors [2] |
| Main strength | Can estimate cause and effect [2] | Uses routinely collected information from many people and settings [1] |
| Main weakness | Results come from the trial's own setting and participants, which may not match everyday care | Confounding and selection bias are particular concerns [2] |
The Cochrane Handbook says randomization aims to balance prognostic factors across groups and so prevent confounding, and that biases due to confounding or selection of participants are specific, or particularly important, for non-randomized studies [2]. See Observational Studies and Confounding and Efficacy vs Real-World Effectiveness.
When do the two agree?
Researchers tested this by repeating the design of 32 randomized trials of medicines using insurance-claims databases, with the plans registered in advance [3]. The trials were selected because they were feasible to emulate. Across the pairs, the results correlated at 0.82 (95% confidence interval 0.64 to 0.91); 75% agreed on statistical significance, 66% agreed on the estimate and 75% on the standardized difference [3]. When the database study could closely copy the trial's design and measurements (16 trials), agreement was higher (correlation 0.93, 95% CI 0.79 to 0.97); when it could not (16 trials), agreement was weaker (0.53, 95% CI 0.00 to 0.83) [3]. The authors concluded that real-world studies can reach similar conclusions as trials when design and measurements can be closely emulated, but this may be difficult, and differences from chance, emulation and residual confounding are hard to disentangle [3].The lesson is not "real-world evidence is unreliable" or "reliable," but that its value depends on design and on whether the question suits the data.
Questions to ask about a real-world claim
- Which data source: records, claims, a registry, a wearable?
- Who was compared with whom, and how were the groups made comparable?
- Was the analysis planned and registered in advance?
- What outcome was measured, and was it recorded the same way in every group?
- Does the claim name a product and a use, or is it a general success story?
Related: Digital Biomarkers and Patient-Reported Outcomes are two kinds of measurements that real-world studies often rely on.
Frequently asked questions
Is real-world evidence the same as an observational study?
Not exactly. The FDA defines it by its source (routinely collected data) and says it is derived from analysis of those data [1]. Many real-world studies are observational, and some use randomization inside routine care [1].
Can real-world evidence replace a randomized trial?
Sometimes it can support a regulatory decision, and the FDA is developing frameworks for that [1]. In a 32-trial comparison, agreement varied from strong to weak depending on how closely the design could be copied [3].
Is a clinic's patient data real-world evidence?
Not by the FDA definition unless it comes from analysis of routinely collected data designed for the question [1]. Offering a treatment does not establish that it is approved or effective.
Does real-world evidence say anything about safety?
Yes. The FDA has a long history of using real-world data to monitor postmarket safety of approved drugs [1].
More details: evidence, limits and cautions
The details behind this page: cautions, limits of the evidence and where sources disagree. Open any note to read it.
How regulators use it
References
- U.S. Food and Drug Administration. Real-World Evidence (content current as of 06/03/2026). Read 2026-10-22.
- Reeves BC, Deeks JJ, Higgins JPT, Shea B, Tugwell P, Wells GA. Chapter 24: Including non-randomized studies on intervention effects. Cochrane Handbook for Systematic Reviews of Interventions, version 6.5. Read 2026-10-22.
- Wang SV, et al. Emulation of Randomized Clinical Trials With Nonrandomized Database Analyses: Results of 32 Clinical Trials. JAMA 2023. Read 2026-10-22.