When the Confirmatory Trial Fails: Sarepta, Traditional Approval, and the Hardest Submission a Writing Organisation Handles

The Confirmatory Study That Was Supposed to Settle It

Accelerated approval is a bargain. FDA grants marketing authorization on a surrogate or intermediate endpoint reasonably likely to predict clinical benefit, and the sponsor commits to a confirmatory trial that will verify that benefit. The confirmatory study is the other half of the deal — the evidence that is supposed to settle the question the surrogate left open.

For Sarepta's AMONDYS 45 and VYONDYS 53 — exon-skipping therapies for patients with Duchenne muscular dystrophy mutations amenable to exon 45 and exon 53 skipping — the ESSENCE confirmatory study did not meet statistical significance on its primary endpoint. Sarepta argues the totality includes favorable trends, a COVID-disruption analysis, real-world evidence, and long-term safety. FDA has now accepted sNDAs seeking to convert these accelerated approvals to traditional approval, with a target action date of February 28, 2027.

Whatever one thinks of the strategy, it creates one of the hardest submissions a regulatory writing organisation ever has to assemble: a filing that has to argue for conversion to traditional approval when the study designed to confirm benefit did not reach significance.

This is not a first submission, which introduces a program. It is not a resubmission on supportive existing data, where the argument is that what already exists is adequate. It is the rarer and harder case — a submission that has to make a totality-of-evidence benefit case after the confirmatory result came in short.

Why This Is Harder Than Any Adjacent Case

A first submission writes toward a clean hypothesis. A resubmission-on-existing-data — the uniQure and Regenxbio situations — writes to show that a package the agency already saw is adequate, into a posture that has become more receptive. Both are hard. Neither is this.

Here the central fact of the file is a confirmatory study that missed its primary endpoint. The reviewer knows it. The public knows it. The submission cannot minimize it, cannot bury it, and cannot pretend the accelerated-approval bargain was met when its verifying study did not reach significance. What the writing organisation can do is far narrower, and it is entirely a question of how the totality of evidence is assembled and framed: whether there is a coherent, honest, scientifically defensible case that clinical benefit is credible despite the confirmatory miss — and whether the submission makes that case without overreaching into claims the data cannot hold.

The writing cannot manufacture a benefit the evidence does not support. On a failed confirmatory trial, that limit is at its most exposed. But where a defensible case exists — in the original dystrophin-surrogate rationale, the functional trends from ESSENCE, the natural-history context, the real-world evidence accumulated since accelerated approval, and the long-term safety record — whether that case is visible and credible to the reviewer is a writing outcome, not a data outcome.

The Four Things the Submission Has to Do

Characterize the confirmatory miss directly and completely. The fastest way to lose a reviewer on a failed-confirmatory filing is to appear to be managing the failure — foregrounding the supportive data while the missed endpoint surfaces from the tables. The submission has to state the ESSENCE result plainly, explain what it did and did not show, and address the obvious questions — the endpoint, the 96-week timepoint, the COVID-disruption analysis if it is used, the statistical picture — before the reviewer raises them. Candor is not a concession here; it is the precondition for being believed on everything else. A COVID-disruption argument in particular has to be offered as a characterization of the data, not as an excuse for it, or it undercuts the credibility of the rest of the filing.

Connect the surrogate to the totality. The affirmative argument has to link the original dystrophin-surrogate rationale — the basis of the accelerated approvals — to the functional trends, the real-world evidence, and the safety record, assembled into a coherent whole. This is integrated-evidence writing at its most demanding, because the pieces have to explain why clinical benefit is credible when the single study meant to demonstrate it did not reach significance. The submission has to make the totality argument without implying the confirmatory result does not matter.

Address the accelerated-approval bargain head-on. The reviewer's structural question is unavoidable: the confirmatory study was the commitment, and it did not confirm — so why should these drugs convert to traditional approval. The submission that does not answer that question directly leaves the agency to answer it alone, and an agency answering it alone answers it toward the accelerated-approval framework's default remedies. The submission that proposes a defensible path — how the totality still supports clinical benefit, what additional evidence exists or is planned, and a benefit-risk argument specific to Duchenne's severity and unmet need — gives the agency something to weigh other than the binary.

Keep the clinical, statistical, and benefit-risk narratives telling one honest story. On a failed-confirmatory filing, the temptation to let the clinical overview lean optimistic while the statistics carry the miss is strongest, and the damage is greatest. A reviewer who finds the sections implying different conclusions about the same study stops trusting the submission entirely. Internal consistency is always discipline; here it is credibility itself. The submission also has to acknowledge the full agency history without becoming a document about the agency's inconsistency — the argument stays on the evidence.

The Disease Context Is Part of the Writing

Duchenne is a severe, progressive, fatal disease with limited options — and that context is legitimately part of the benefit-risk argument, not an appeal to sympathy. The writing organisation's task is to make the unmet-need and disease-severity case rigorously, as a factor the regulatory framework explicitly permits weighing, rather than as an emotional counterweight to a data problem. Done well, it is a defensible element of a benefit-risk narrative. Done poorly, it reads as an attempt to substitute urgency for evidence — and a reviewer reads the difference immediately.

The Wider Pattern

Failed confirmatory trials on accelerated-approval drugs are becoming more visible as the first large cohort of accelerated approvals reaches its verification milestones. Each one poses the same question: can the sponsor explain, honestly, why the totality still supports clinical benefit when the confirming study missed — or does the file collapse into either denial or capitulation. The sponsors that handle it are the ones whose writing organisations can hold candor and advocacy in the same document, building the totality-of-evidence case without pretending the confirmatory result away.

The Confirmatory Failure Is the News. The Writing Problem Is What the Submission Does Next.

The confirmatory failure is the news. The writing problem is what the submission does next.

Sarepta cannot write around ESSENCE. The missed primary endpoint is the central fact of the file.

The only credible route is to write through it: state the miss plainly, explain the COVID-disruption argument if used, connect the dystrophin-surrogate rationale to functional trends and real-world evidence, characterize safety, and propose a defensible benefit-risk position.

A failed confirmatory trial does not leave much room for rhetoric.

It leaves room only for disciplined totality-of-evidence writing.