Real-World Evidence (RWE)

Real-world evidence is the analysis of real-world data, meaning information generated by ordinary healthcare rather than by a protocol: electronic health records, insurance claims, dispensing records and disease registries. Most such analyses are observational, comparing those who received a treatment with those who did not. Pragmatic trials sit between the categories, keeping randomisation while relaxing eligibility and delivering the intervention through routine care.

The 21st Century Cures Act of 2016 directed the FDA to formalise how real-world evidence can support regulatory decisions, and in practice it is accepted most readily for safety, for label expansions, and as an external control where a randomised comparison is infeasible, as in rare disease. Registry and claims data are also where post-marketing safety signals for widely used drugs surface first, because the exposed population is far larger than any trial.

Its value is coverage of exactly what trials exclude: elderly patients, multiple comorbidities, polypharmacy, long horizons and imperfect adherence. The cost is confounding by indication, because prescribing is a decision made for reasons that also predict outcome. Propensity scores and matching adjust only for variables that were recorded, and the important ones in these databases frequently are not.

The characteristic overreach is reporting an association from a claims database as an effect, complete with a causal verb. Immortal time bias and the healthy-adherer effect produce sizeable spurious benefits here with some regularity. In the other direction, real-world evidence cannot rescue an untested compound: an online tally of self-reported grey-market experiences is anecdote, and calling it real-world data does not move it up the hierarchy.