Why Real-World Evidence Can No Longer Wait Until After Approval

The evidence model in drug development has been quietly reorganizing for years. Real-world evidence (RWE), long treated as a post-market footnote, has been moving upstream, informing trial design, endpoint selection, and regulatory submissions. The clearest signal yet came in February, when the FDA formalized a new default for marketing authorization: one adequate and well-controlled trial, supported by confirmatory evidence, with RWE explicitly named as eligible to fulfill that confirmatory role. The bar has shifted—and it has shifted in RWE’s favor.

But the reorganization predates the announcement and extends beyond the approval stage. For decades, clinical evidence followed a predictable sequence: randomized controlled trials (RCTs) at the top, regulatory submissions next, and RWE only after approval. That linear model is giving way to something more continuous: a self-reinforcing system in which RWE runs alongside traditional trial data from early development through post-market surveillance, interrogating and sharpening findings in real time, not retrospectively.

In this article, I will examine what that transition demands of pharma organizations, particularly around data infrastructure, longitudinal evidence, and cross-source synthesis.

The problem with the linear hierarchy

The traditional evidence hierarchy served its purpose when the dominant question was straightforward: Does this molecule work? Under controlled conditions, RCTs answered that with statistical rigor, while RWE was reserved for after approval, capturing effects and interactions that trials could not anticipate.

By the time real-world data flags a problem under the old model, hundreds of millions in resources may already be committed, programs well advanced, and regulatory packages in motion. The cost of late discovery is not just financial. It is clinical. Whether RWE needs to wait until after approval at all is a question that regulatory precedents have already begun to answer. It does not.

The second opinion in action

When RWE is positioned as a second opinion rather than a post-market report, it raises questions in near real time: Are the patients entering the trial representative of those who will eventually use the drug? Do early signals align with what observational data from similar patients already show? Are there safety patterns in routine care data that warrant closer examination?

The FDA's Real-World Evidence Program and EMA guidance both reflect an acknowledgment that real-world data can inform, though not replace, trial evidence during active development. What is consistently underestimated is that this is as much an operating model challenge as a methodological one—and the harder problem to solve. Real-world data must be accessible and interpretable on a timeline that is useful during trial conduct, not months after a database lock, when course corrections are no longer possible.

The data challenge

For RWE to function as a concurrent validation layer, the underlying infrastructure should be built for continuity rather than snapshots. Most clinical data systems are designed for discrete events, such as a trial start, an interim analysis, or a submission. However, real-world data does not work that way. Electronic health records (EHRs) accumulate daily, claims data reflects utilization in near real time, and patient-reported data arrives asynchronously, requiring infrastructure capable of ingesting and analyzing these streams on an ongoing basis. Cloud-based platforms with standardized data models are increasingly the foundation for this kind of integration.

Volume alone does not equal value. A dataset with broad coverage but limited longitudinal follow-up answers different questions than one with years of continuous patient history. Assembling that depth requires deliberate data partnerships, rigorous curation, and honest clarity about where coverage gaps may distort or silently invalidate the picture.

The organizational challenge

Even organizations that solve the data problem will struggle if they treat RWE as a parallel workstream rather than an integrated function. Claims data, EHRs, registries, and patient-reported outcomes each illuminate distinct aspects of a patient’s experience. Drawing a coherent signal from that landscape requires simultaneous reasoning across sources, reconciling their differing structures, populations, and inherent biases without flattening the distinctions that make each valuable. Methods such as target trial emulation and federated analysis are making that more feasible, though both require expertise that many organizations are still developing.

The organizations furthest along share a common trait. They have invested in unified data environments where clinical trial data and real-world data are managed as complementary inputs into a single evidentiary framework, not siloed assets that teams scramble to reconcile at the last minute.

The evidence model worth building toward

When the concurrent validation model is working, the nature of the regulatory conversation changes. An organization that has run RWE concurrently arrives at the approval stage with evidence already stress-tested from multiple directions, not evidence assembled under pressure.

That is a fundamentally different starting position: earlier signal detection rather than late-stage course corrections, endpoint choices validated against real-world outcomes, and safety profiles continuously assessed rather than reconstructed from periodic reviews.

Organizations that treat this regulatory inflection point as a mandate to build will find themselves with development programs that are not just more efficient, but more defensible, more credible, and better positioned for the scrutiny that follows approval. This is what leadership in the second-opinion era of drug development looks like.

Christopher P. Boone, PhD, group vice president of research services, Oracle Health and Life Sciences