When Real-World Evidence Can Replace — or Supplement — Your Clinical Trial

The FDA's December 2025 guidance on real-world evidence (RWE) for medical devices is not just a regulatory update. It is a signal that the evidentiary landscape for device submissions is shifting — and sponsors who understand that shift will be better positioned to design smarter, faster, and less burdensome pathways to market.

The question we hear most often from our clients is not "what is RWE?" They know what it is. The question is: when can we actually use it? When does real-world data (RWD) rise to the level of valid scientific evidence that FDA will accept in place of — or alongside — a traditional clinical study?

The answer is more nuanced than a checklist, but it is more accessible than many sponsors realize.

The Core Principle: Relevant and Reliable

The FDA's updated guidance makes one thing clear from the outset: RWD is not inherently inferior to data from a traditional randomized trial. What matters is whether the data are relevant and reliable for the specific regulatory question at hand.

Relevance asks whether your data source captures the right patients, the right device exposures, the right outcomes, and does so across a time horizon that reflects actual clinical experience. Reliability asks whether the data was collected consistently, with appropriate quality controls, and in a manner that supports defensible causal inference.

If you can demonstrate both — and document them rigorously — FDA has opened the door to using RWE as primary clinical evidence, not just supplementary color.

Where RWE Can Carry the Submission

The 2025 guidance explicitly lists contexts in which RWE may serve as the primary basis for regulatory action. These are not hypothetical. FDA's own case examples illustrate approvals and clearances built substantially on RWD.

Original PMA approvals. In one example drawn from the guidance, an implanted device received original PMA approval using data from an overseas registry — hundreds of patients, more than two years of follow-up — compared against prospectively defined performance goals derived from a meta-analysis of published literature. No traditional IDE study. The keys: devices were clearly identified by their unique device identifier, the registry data were generalizable to U.S. patients, and data capture methods ensured completeness and consistency across sites.

Expanded indications. A PMA panel-track supplement used RWD from a sponsor's patient database linked to Medicare claims as the primary safety evidence for a new patient population. When you already have a marketed device with real-world uptake, longitudinal claims and registry data can characterize safety in a far larger patient population than any trial you could realistically enroll — and can surface rare events that prospective studies are almost never powered to detect.

External control groups. Single-arm studies comparing to a registry-derived control group can support approval when the registry captures outcome definitions and covariates comparable to those in the interventional arm — particularly when propensity score methods are applied to account for baseline differences.

De Novo and 510(k) submissions. The guidance also describes a De Novo granted in part on EHR data from over 500 patients that confirmed device performance and explored subgroup effects that the prospective single-arm study could not address. RWE resolved uncertainty that remained after the traditional study — a pattern FDA is receptive to.

Where RWE Supplements But Doesn't Replace

Not every submission can be built on RWD alone, and it is important to be honest about where the limitations lie.

Confounding is the central challenge. Unlike a randomized trial, observational data cannot guarantee that treated and untreated patients are comparable at baseline. Propensity score methods, directed acyclic graphs, and sensitivity analyses can reduce — but not eliminate — this uncertainty. FDA is sophisticated on this point, and submissions that do not demonstrate rigorous confounding control are unlikely to receive favorable review.

Outcome capture is often incomplete. Administrative claims data are powerful for capturing procedures, hospitalizations, and mortality. They are weaker for capturing functional outcomes, patient-reported measures, or device-specific performance metrics that are not coded. If your primary endpoint is a nuanced clinical assessment, you may need a hybrid design: protocol-driven data collection for short-term outcomes, with RWD supporting longer-term follow-up.

Operational definitions require validation. The guidance places significant emphasis on the distinction between a conceptual definition — what you intend to measure — and an operational definition — the specific codes, algorithms, and timeframes used to identify that construct in your dataset. Small differences in operational definitions can produce meaningfully different results. FDA expects sponsors to validate these definitions, particularly for primary endpoints, often through chart abstraction against a reference standard.

The Protocol Requirement Is Non-Negotiable

One of the clearest messages in the 2025 guidance is that a prespecified, finalized protocol is required before you look at outcome data — regardless of whether your study uses extant RWD or prospectively collected data. This is not a formality.

Sponsors who analyze RWD, observe promising patterns, and then draft a protocol to match those findings are engaging in a form of data dredging that FDA's reviewers are trained to identify. The guidance states explicitly that during protocol development, study teams should not have access to outcome measure results. Protocol amendments must be dated, time-stamped, and justified.

For most clients, this means the statistical analysis plan — including the propensity score model, the primary and secondary endpoints, and all sensitivity analyses — must be locked before any outcome-related data are accessed. BSC builds this discipline into our engagement model from day one.

A Framework for Making the Decision

When a client asks whether RWE can support their submission, we work through a structured set of questions:

1. What is the regulatory question? A PMA for a high-risk device with no predicate has a different evidentiary threshold than a 510(k) or a post-approval study requirement. Clarity on the decision matters before evaluating data sources.

2. Does a sufficient RWD source exist? This means not just data volume, but data completeness — longitudinality, UDI capture or equivalent device identification, covariate availability, and continuity of care. Many sponsors are surprised to learn that a registry they participate in does not capture the variables needed for their specific study question.

3. Can you control for confounding? This is where biostatistical rigor becomes decisive. If the treated and comparator populations differ substantially on key prognostic factors — and the data to adjust for those factors are not available — RWE may not be sufficient.

4. What is FDA's prior experience with this data source? The guidance encourages sponsors to document prior use of the same RWD source for similar populations. A data source with published validation studies and prior regulatory acceptance is substantially more credible than a novel or proprietary database with no track record.

5. Would a hybrid design resolve the remaining uncertainty? In many cases, the strongest submission combines a targeted prospective element — focused on the primary effectiveness endpoint — with RWD supporting longer-term safety follow-up, generalizability, and subgroup analyses. This design can satisfy FDA's evidentiary standards while substantially reducing the time and cost of evidence generation.

The Strategic Opportunity

The 2025 FDA RWE guidance represents a meaningful expansion of the evidentiary toolkit available to medical device sponsors. For companies with access to registries, linked claims data, or device-generated data streams, this guidance provides a clear framework for transforming existing data into regulatory evidence — potentially replacing years of traditional clinical follow-up.

But the opportunity is only realized by sponsors who approach RWD with the same methodological rigor they would apply to a prospective trial: prespecified protocols, validated outcome definitions, transparent confounding control, and honest documentation of data limitations.

The question is not whether real-world evidence is appropriate for your submission. The question is whether you are building the infrastructure — statistical, operational, and regulatory — to use it effectively.

Biomedical Statistical Consulting LLC (BSC®) has provided biostatistical and clinical study design support for FDA medical device submissions since 1986. For questions about RWE strategy, study design, or statistical analysis plans, contact us at biomedstat.com.