Real World Evidence (RWE)

Real world evidence is clinical evidence about the usage and potential benefits or risks of a medical product or intervention derived from the analysis of real world data collected outside of conventional randomized controlled trials. RWE is generated by applying appropriate research methods to real world data sources such as electronic health records, claims databases, registries, and patient generated data. These definitions follow the US FDA, which defines RWD as data relating to patient health status and/or the delivery of health care routinely collected from a variety of sources, and RWE as clinical evidence about the usage and potential benefits or risks of a medical product derived from analysis of RWD; the EMA's definition of RWD is similar but explicitly limits it to sources other than traditional clinical trials. Under the FDA's 2018 framework, RWE can come from observational studies and also from randomized designs that use RWD, such as pragmatic trials that collect outcomes from routine records.

The distinction between RWD (the raw data) and RWE (the evidence generated from analyzing that data) is important because evidence quality depends not only on the data source but also on the analytical methods applied. RWE has become increasingly important in regulatory decision making, health technology assessment, and clinical practice. In the United States, the 21st Century Cures Act of 2016 directed the FDA to evaluate the use of RWE to help support approval of new indications for approved drugs and to help support or satisfy post approval study requirements; the FDA published the framework for its RWE program in 2018. Health systems, payers, and guideline developers use RWE to assess comparative effectiveness, monitor safety signals, and evaluate the generalizability of trial findings to routine practice.

Generating RWE involves several steps. First, a clearly defined research question is formulated, ideally with a protocol specifying the study design, population, exposure, comparator, outcomes, and analytical plan. Second, appropriate RWD sources are identified and accessed, with attention to data quality, completeness, and representativeness. Third, the study is designed using methods appropriate for observational data, such as cohort, case control, or self controlled designs.

The analytical phase addresses the fundamental challenge that treatment assignment in real world practice is not random. Methods for reducing confounding include propensity score matching or weighting, instrumental variables, and regression discontinuity; negative control analyses help detect residual confounding. Sensitivity analyses assessing the impact of unmeasured confounding (such as the E value) strengthen the credibility of the findings.

Target trial emulation is an increasingly popular framework for designing RWE studies. It involves specifying the hypothetical randomized trial that would ideally answer the research question and then using observational data to emulate each element of that trial (eligibility criteria, treatment strategies, assignment procedures, the start and end of follow up, outcomes, causal contrasts, and analysis plan). This approach helps researchers avoid common biases in observational research, such as immortal time bias and prevalent user bias, by aligning eligibility, treatment assignment and the start of follow up (time zero), as a randomized trial does. It does not remove confounding by itself: an emulated trial can still be biased by confounders that were not measured.

RWE can complement trial evidence in several ways: confirming that trial efficacy translates to real world effectiveness, identifying subgroups that benefit most or least, detecting rare adverse events that trials were not powered to find, and providing evidence for populations excluded from trials. However, observational RWE has inherent limitations, including the inability to control for unmeasured confounders, potential measurement error in routine data, and the possibility that observed associations reflect prescribing patterns rather than causal effects.

Use RWE when the research question concerns comparative effectiveness, safety, or healthcare utilization in routine practice; when randomized trial evidence is unavailable, insufficient, or not generalizable to the population of interest; or when regulatory or payer decision making requires supplementary evidence. RWE is particularly valuable for post marketing surveillance, pragmatic comparative effectiveness research, and health technology assessment. It should not substitute for randomized trials when a trial is feasible and the question concerns initial efficacy in a defined population.

A common error is conflating RWD with RWE, using the terms interchangeably when RWE specifically requires the application of rigorous analytical methods to RWD. Another mistake is designing an RWE study without a protocol, allowing the flexible nature of observational research to enable data driven analytical decisions that inflate false positive rates. Researchers sometimes overstate the causal implications of RWE without discussing the limitations of observational methods. Finally, selecting only RWD sources that are convenient rather than those most appropriate for the research question can introduce selection bias and limit the relevance of the findings.

A pharmaceutical company seeks to expand the indication for an antiplatelet drug to include peripheral artery disease. Earlier randomized trials of the drug in coronary disease included too few patients with peripheral artery disease to settle the question, and a dedicated trial would require five years and 15,000 patients, so the company supplements the trial evidence with an RWE study using claims data from 200,000 patients diagnosed with peripheral artery disease. Using a new user, active comparator design, which compares patients starting the drug with patients starting another antiplatelet drug, and propensity score weighting, the study compares major adverse limb events between the two groups over a median of 2.8 years. The results show a 22% lower hazard of major adverse limb events with the drug (HR 0.78, 95% CI: 0.69 to 0.88), supporting the expanded indication. In this hypothetical case, the FDA accepts the study as supplementary evidence alongside the existing trial data, not as a replacement for randomized evidence.