How HEOR Teams Build Evidence That Sticks

By John Oncea, Chief Editor, Clinical Tech Leader

Jessica Jalbert, a card-carrying epidemiologist in the Health Economics and Outcomes Research team at Regeneron, has watched the potential and impact of real-world evidence grow.

Historically, HEOR’s role was to generate evidence that demonstrated a treatment’s unmet need and value differentiation by evaluating disease burden, drug utilization, and the effectiveness of standard of care and novel therapies.

That has changed, and for Jalbert, the shift traces to the 21st Century Cures Act, followed by the FDA’s 2018 Real-World Evidence Framework, which opened the door to using real-world evidence in support of regulatory decision-making. RWE itself isn’t new to drug development; it has been used for decades to assess and monitor the safety of new treatments. What’s new is applying RWE specifically to demonstrate effectiveness in ways that can shape a regulator’s decision to approve a new treatment.

“There’s been a whole momentum shift,” Jalbert says. “There is now acceptance that real-world evidence about a drug’s effectiveness, when derived from relevant and reliable real-world data and by leveraging appropriate methods, can be used to support and inform decisions by regulators and payers to approve or reimburse a treatment, respectively.”

From Support Function To Decision Engine

The vocabulary Jalbert uses for her work is precise in a way that matters. She draws a clean line between real-world data – patient health and care-delivery information drawn from sources like administrative claims, electronic health records, and registries, generally collected outside the context of a clinical trial – and real-world evidence, the inferences generated when real-world data are analyzed to answer a defined question.

That distinction isn’t semantic. It’s operational. Getting robust, valid, reproducible inferences requires a deep understanding of epidemiologic study design and analytic methods, plus a deep understanding of the data itself, since the data are generally not collected for research purposes. Anyone who can code and has access to the data can generate an inference. Whether that inference is valid is a different question entirely.

Regeneron’s HEOR team includes a center of excellence (CoE) in epidemiology and RWE, one that Jalbert leads. The CoE is a multidisciplinary team comprising epidemiologists and biostatisticians, real-world data, and analytics scientists with expertise in the complexities of real-world data, and engineers specializing in AI and advanced analytics. The CoE is focused on producing RWE at the highest level of quality. This structure isn’t common across the industry and reflects Regeneron’s commitment to patients and to the underlying science.

Where Real-World Evidence Earns Its Place In The Arc

Jalbert acknowledges that well-conducted randomized controlled trials sit at the top of the causal evidence hierarchy. “RCTs are the gold standard for evaluating if a treatment works or not; if a trial can be done, it should be,” she says plainly. That clarity matters, because there is sometimes the misperception that RWE is an alternative to conducting trials rather than a complement to them. Jalbert resists that framing.

Real-world evidence adds distinct value at every stage of the product life cycle, but not in the same way at every stage. RWE can characterize a patient population and unmet need, including medication utilization and the natural history of a disease, which can be useful for informing trial design, planning, and execution.

At approval, not all questions about a treatment’s safety and effectiveness have been answered. How will the risks and benefits observed in a controlled clinical study translate into actual practice? In populations not generally included in clinical trials – such as older patients – and in situations where conducting comparative trials is not feasible or ethical, or when standard of care is changing rapidly or is heterogeneous, RWE may be the only approach for assessing comparative safety or effectiveness questions.

The Credibility Problem

The strength of the evidence does not depend on the volume of available real-world data but on the quality of that data, the selection of an appropriate data source, and the rigor of the study design and analytic technique used to generate it.

Unlike data collected under controlled trial conditions, real-world data is generated as a byproduct of clinical care. That origin creates inherent limitations: missing values, inconsistent coding practices, and gaps in longitudinal records are routine rather than exceptional. Selecting a data source that is genuinely fit-for-purpose for the question at hand – and understanding where its coverage breaks down – is itself a methodological decision with direct consequences for the credibility of any evidence derived from it.

For Jalbert, the abundance of RWD is not the primary challenge. The real challenge is ensuring that the evidence generated from the data is credible and, when appropriate, decision-grade. That credibility is built through pre-specification, methodological rigor, and transparency – defining and locking in protocols and analysis plans before the analysis of outcomes, using appropriate study design and analytical approaches, and interpreting findings in a manner consistent with the strengths and limitations of the underlying design. As she notes, “hypothesis-generating findings should inform future research, not be overstated as causal evidence.”

Sponsors can approach regulators with real-world evidence generated opportunistically, but Jalbert is clear that RWE for regulatory decision-making requires planning and early engagement with regulators. For decision-grade RWE, the expectation is that sponsors explain why a clinical trial cannot be conducted and demonstrate that the real-world data are fit-for-purpose for answering the specific question at hand. Sharing the intent to conduct an RWE study before it begins, obtaining FDA feedback on the protocol and statistical analysis plan, and documenting any deviations from that plan as the study proceeds are all critical steps. Registering the protocol publicly is also encouraged, though not required.

Jalbert notes that these same standards – identifying fit-for-purpose data and pre-specifying analyses – are the basis for principled RWE research. They’re practices HEOR and other RWE-generating teams should be applying across their work.

Designing Evidence For More Than One Audience

Evidence needs can overlap among regulators, payers, and clinicians, but Jalbert cautions against assuming that each audience necessarily requires a separate study. When the research question is relevant to all three, a rigorous study may inform the decisions of multiple stakeholders.

The direction of that use is not always interchangeable. Evidence developed for regulatory decision-making may also be useful to payers and clinicians when it addresses questions relevant to them. A study conducted for publication can provide valuable scientific insights, but it cannot automatically be repurposed for regulatory decision-making because the context of use and the evidentiary requirements are stricter. Across audiences, usefulness still depends on a clearly defined question, an appropriate design, transparency about the population and methods, and conclusions that stay within the study’s limitations.

“Every study has to be understood in the context of its limitations,” Jalbert says. “We need to be careful not to overstate the conclusions and implications of the work.” Rigorous evidence can support firm conclusions, but those conclusions must be based on what the data and results can support. A well-executed observational study can be meaningful for decision-making while remaining clear about residual confounding. Furthermore, it is important to distinguish between a hypothesis-generating study and one designed for causal inference.

That restraint, matching the claim to the question, design, and intended use, strengthens the credibility of the evidence.

What Comes Next

When Jalbert looks ahead at what the industry needs to change most, she points to the need to focus on investing in rigorous methods rather than chasing fast answers. “There’s a perception that evidence should be generated at the click of a button,” she says. “I’d love to see stakeholders understand that quality RWE takes time. Design and analytic decisions need to be weighed and justified.”

That doesn’t mean Jalbert sees technology as an obstacle. She’s optimistic about how generative AI can contribute positively to the pace and quality of RWE research. “It can help us generate RWE more quickly and more efficiently,” she says, “but there are still limitations to the technology.”

In practice, she sees its clearest near-term value in aiding coding and analysis, in the extraction of concepts from unstructured data, and in helping draft queries and protocols. What doesn’t change is accountability: every task remains the responsibility of a human researcher. The future of RWE will be shaped not by replacing human expertise but by augmenting it with technologies that enable researchers to work faster and more efficiently to generate high-quality, actionable evidence.

Jessica Jalbert is an epidemiologist in the Health Economics and Outcomes Research team at Regeneron. Her public comments and professional perspective informed this article. The views and opinions expressed in this article are those of the contributor and should not be attributed to any organization with which the contributor is employed.