The FDA Doesn’t Have a GLP-1 Problem. It Has an Evidence Problem.

On March 19, 2026, Nature Medicine published a trial emulation study that should be sitting on every regulatory affairs director’s desk by now. The study examined 174,678 patients with type 1 diabetes across electronic health record databases and found that initiation of GLP-1 receptor agonists was associated with a meaningfully reduced risk of major adverse cardiovascular events and end-stage kidney disease. No randomization. No placebo arm. No CONSORT diagram. Just a methodology sophisticated enough to approximate one — and a patient population large enough to make the results difficult to dismiss.

The clinical signal matters. But the regulatory question it detonates is more consequential: when observational evidence at this scale points toward clear benefit in a population that cannot realistically be randomized, what does the FDA actually owe sponsors — and patients — in terms of a defined evidentiary pathway to label expansion?

The FDA’s posture on real-world evidence is understandable. It is also incomplete. And the gap between those two things is where patients with type 1 diabetes are currently living.

The Emulation Isn’t the Problem

Trial emulation as a methodology is not new, and it is not fringe. The framework — developed largely through the work of Miguel Hernán and colleagues at Harvard — uses observational data to approximate the structure of a hypothetical randomized trial: defined eligibility criteria, a specified time zero, pre-specified outcomes, and active attempts to control for confounding through inverse probability weighting or g-estimation. The Nature Medicine study applied this framework to EHR data from 174,678 type 1 diabetes patients, a sample size that dwarfs most Phase 3 cardiovascular outcomes trials by an order of magnitude. The EMPA-REG OUTCOME trial, which formed the basis of empagliflozin’s cardiovascular indication, enrolled 7,020 patients. The LEADER trial for liraglutide enrolled 9,340. Scale alone does not validate methodology, but it does raise the bar for what it means to dismiss the signal.

The FDA acknowledged trial emulation as a legitimate analytical approach in its 2023 framework for real-world evidence in regulatory submissions, specifically noting that observational studies designed to emulate RCTs can reduce confounding bias when pre-registered and transparently reported. The 2021 FDA guidance on real-world data and real-world evidence for regulatory decision-making further outlined conditions under which RWE could support label expansions — including when traditional RCTs are “not feasible or not ethical.” Type 1 diabetes is precisely that population. GLP-1RAs are currently approved for type 2 diabetes; the glycemic physiology in type 1 differs substantially, and designing a powered, placebo-controlled cardiovascular outcomes trial in type 1 patients on background insulin therapy, with all attendant hypoglycemia risks and extended follow-up requirements, would take a decade and cost well over $500 million. The FDA knows this. Sponsors know this. The question is whether anyone is willing to act on it.

But the credibility of the methodology does not automatically translate into a regulatory submission strategy. And this is where the Nature Medicine findings expose a structural gap the FDA has been orbiting without closing.

The Precedent the FDA Is Ignoring

Consider what the agency has already accepted in adjacent spaces. In 2022, the FDA cleared the use of digital biomarker data from wearable devices to support labeling claims — a decision that rested heavily on retrospective data and algorithm validation rather than prospective RCT confirmation. In oncology, the agency has granted accelerated approvals on single-arm trials with response rate endpoints for years, with post-market confirmatory trials sometimes arriving years later, or not at all. The 2022 Oncology Center of Excellence review identified over 40 accelerated approvals that had been on the market for more than three years without confirmatory trial completion. The FDA tolerated that epistemic gap in oncology because the unmet need was visible and the alternative — withholding potentially active agents — was politically and ethically untenable.

The unmet need in type 1 diabetes cardiovascular outcomes is equally visible. Patients with type 1 diabetes have a two- to four-fold higher risk of major cardiovascular events compared to age-matched individuals without diabetes. They are currently excluded from the GLP-1RA label expansions that type 2 patients have benefited from since the LEADER and SUSTAIN-6 trials reshaped prescribing practice beginning in 2016. And the Nature Medicine emulation study, with its 174,678-patient base, is providing the kind of signal strength that, in oncology, would already be generating advisory committee meetings.

So why is cardiovascular outcomes in type 1 diabetes being held to a stricter standard than tumor shrinkage in metastatic lung cancer? The counterintuitive answer is that the FDA’s RWE framework is actually more conservative in chronic disease settings precisely because it can be. In oncology, the urgency overrides the epistemological caution. In metabolic disease, regulators have the luxury of demanding more — and they are using it, to the detriment of patients who have been waiting on the wrong side of a label boundary for a decade.

What Sponsors Must Do Before the FDA Does It for Them

The Nature Medicine study was not sponsored by Novo Nordisk or Eli Lilly. It was an independent academic emulation, which means neither company can claim it as proprietary data in a submission — but both can reference it as external evidence of biological plausibility and population-level signal. That distinction matters operationally. Under FDA’s 2023 RWE framework, sponsors are permitted to submit RWE packages that include independently generated studies alongside sponsor-conducted analyses, provided the data provenance is transparent and the analytical methodology is pre-specified or reconstructable. The 174,678-patient emulation qualifies on both dimensions.

What Novo Nordisk and Lilly should be doing right now — and what their regulatory affairs teams are almost certainly already modeling — is constructing a Type B meeting request to CDER’s Division of Diabetes, Lipid Disorders, and Obesity asking explicitly: will a trial emulation study of this design, combined with a sponsor-conducted sensitivity analysis on a separate EHR cohort, constitute sufficient evidence to support a label expansion for GLP-1RAs in type 1 diabetes for cardiovascular risk reduction? The FDA’s answer to that question, whether yes, no, or not without additional confirmatory data, would create a precedent that defines the RWE pathway for every chronic disease indication behind it.

The agency has the tools. The 2021 RWE guidance exists. The 2023 framework exists. The Prescription Drug User Fee Act VII commitments include explicit language about advancing RWE standards for regulatory decision-making. What does not yet exist is a documented, division-specific standard for when trial emulation evidence is sufficient for a primary label expansion — not a supplemental claim, not an accelerated approval placeholder, but a full indication based on observational data that was designed with randomized-trial discipline from the outset.

If the FDA continues to treat that question as unanswerable, sponsors will continue designing expensive, underpowered trials to generate regulatory cover rather than scientific insight. Novo Nordisk’s Phase 3 program for semaglutide in type 1 diabetes — SUSTAIN FORTE — explored glycemic outcomes but was not powered for cardiovascular endpoints. The gap between what that trial measured and what the Nature Medicine emulation now suggests is precisely the distance the FDA’s RWE framework was designed to close. The framework exists. The evidence exists. The 174,678 patients’ worth of signal exists.

The FDA must now decide whether its own guidance means what it says — or whether “feasibility” is a threshold that only applies when the agency is ready to apply it.