Real-world evidence (RWE)
Also known as: RWE
Real-world evidence (RWE) is clinical evidence about how medical products work in practice, drawn from real-world data such as electronic health records, insurance claims, registries, and wearable devices, rather than from tightly controlled clinical trials.
Regulators and payers increasingly use it to support decisions. While RWE can reveal how treatments work in broader populations, it's prone to bias, and a well-designed study is what separates insight from coincidence.
Real-world evidence (RWE), explained at three levels
At home
Think of a car crash test versus years of insurance claims. The crash test, like a clinical trial, is controlled and precise but artificial. Claims data shows how cars actually perform on real roads with real drivers. Real-world evidence does the same for medicines: it analyzes health records, claims, and registries to see how treatments work for everyday patients after approval.
At work
Real-world evidence is clinical evidence drawn from analyzing real-world data, such as electronic health records, claims, registries, and wearables, rather than clinical trials. Regulators use it for safety monitoring and sometimes approvals; payers use it to judge value. For drug and device companies, it can support label expansions and pricing deals, if the study design is credible.
At home
Real-world evidence starts with data collected during ordinary care: prescriptions filled, diagnoses recorded, hospital stays, and lab results. Researchers link these records and compare people who took a treatment with similar people who did not. Because the treatment was not randomly assigned, the groups may differ in hidden ways, so results are less certain than a trial's. In exchange, they include older adults, pregnant people, and those with several conditions whom trials often exclude.
At work
Real-world data comes from EHRs, claims, registries, and devices, each with gaps: claims lack lab values, and EHRs miss care delivered elsewhere. Common data models such as OMOP standardize structure so one analysis can run across many databases. Uses include post-market safety surveillance, external control arms for rare disease studies, label expansions, comparative effectiveness research, and outcomes-based agreements. In Canada, provincial administrative data held by bodies like ICES in Ontario support much of this work.
At home
The core problem is confounding. People who choose a treatment may differ from those who do not: healthier, wealthier, or more careful about follow-up. A drug can then look effective simply because healthier people took it, called the healthy user effect. Statistical methods adjust for measured differences, but not unmeasured ones. While good methods help a lot, in my view understanding confounding is the single most important skill in reading RWE.
At work
Rigorous RWE designs the analysis as if it were a trial, called target trial emulation: define eligibility, treatment strategies, time zero, and outcomes before looking at results. Methods include active comparators, propensity scores, and sensitivity analyses for unmeasured confounding, with protocols registered in advance. The FDA, EMA, Health Canada, and Canada's Drug Agency have published RWE guidance. Canada's provincial data holdings are rich but fragmented, so access agreements often take longer than the analysis.
Real-world examples
At home
Sophie read that a new blood thinner seemed safer than an older one in real-world studies of older adults. Her pharmacist explained that those studies included people like her grandmother, who would have been excluded from the original trials.
At work
Thomas is a health economist at a Canadian drug agency. When a rare disease drug had only a small trial, his team used a provincial registry to compare outcomes with similar untreated patients, helping inform a coverage decision with conditions.
How it compares
| feature | Real-world evidence (RWE) | Randomized controlled trial |
|---|---|---|
| Setting | Routine care | Controlled study |
| Population | Broad, diverse | Selected by criteria |
| Bias risk | Higher | Lower due to randomization |
| Speed and cost | Faster, cheaper | Slower, expensive |
Trials show whether a treatment can work. RWE shows how it works in everyday life.
For analysts
RWE studies can be designed to reach almost any conclusion. I'd ask whether the protocol was registered before analysis and whether the comparison group is genuinely similar, before trusting any real-world claim.
History & origin
Researchers have long used observational data, such as cancer registries, to study treatments and safety. Large databases from electronic records and insurance claims expanded what was possible in the 2000s. The US 21st Century Cures Act of 2016 directed the FDA to develop an RWE framework, which it published in 2018, and the European Medicines Agency launched the DARWIN EU data network in 2022. During COVID-19, real-world studies tracked vaccine effectiveness across millions of people, often faster than trials could. Health Canada and Canada's Drug Agency have published guidance on using RWE in regulatory and reimbursement decisions. In my opinion, RWE is a powerful complement to trials, not a replacement, and claims otherwise usually come from someone selling data.