When the Regulator Becomes the Risk
A $3.2 million drug, dying children, and what happens when science loses to pressure
The Elevidys story
There is a moment in the Elevidys story that tells you everything you need to know about the state of pharmaceutical regulation. It is June 2024. Sarepta Therapeutics’ gene therapy for Duchenne muscular dystrophy has just failed its pivotal Phase 3 trial. The FDA’s own clinical reviewers, pharmacology team, and statisticians have all concluded that the drug does not work — or at least that the evidence does not establish that it does. And yet Peter Marks, Director of the FDA’s Centre for Biologics Evaluation and Research (CBER), overrules all three of them, grants full approval, and expands the label to cover non-ambulatory patients — those whose disease has already progressed to the point where they can no longer walk.
The drug costs $3.2 million per patient. Within a year, teenagers are dying from acute liver failure after receiving it.
This is not a story about one rogue regulator. It is a story about a system that has been engineered — piece by piece, over many years — to make this kind of outcome not just possible, but predictable.
A pattern hiding in plain sight
Elevidys was not the first time Sarepta had navigated this particular route to market. In 2016, the company’s exon-skipping therapy eteplirsen was approved for DMD by Janet Woodcock at the FDA, again over the objections of her own review staff, again on the basis of a surrogate endpoint rather than demonstrated clinical benefit. The agency’s advisory committee had voted against approval. The chief medical officer at the time resigned in protest.
Sarepta had learnt something important from that episode: the FDA could be moved. Patient advocacy groups, wielding the moral weight of dying children and the political optics of denying them hope, combined with a company willing to fight publicly and loudly, could shift the regulatory outcome. The scientific consensus inside the agency was not, it turned out, the deciding factor.
By the time Elevidys arrived — first as an accelerated approval in June 2023, then as a full approval in June 2024 — the playbook was well established.
The accelerated approval problem
Accelerated approval was designed as a compassionate pathway. It allows drugs to be approved on the basis of surrogate endpoints — biomarkers or measures that are thought likely to predict clinical benefit — on the condition that confirmatory trials are run and real-world benefit is subsequently demonstrated. It was a reasonable idea, born from the AIDS crisis of the 1980s and the urgent need to get promising therapies to dying patients faster.
What it has become, in too many cases, is a mechanism for getting drugs to market on thin evidence and keeping them there even when the confirmatory data disappoint. Elevidys was approved on the basis of micro-dystrophin expression — the ability of the therapy to produce a shortened but functional form of the dystrophin protein. When the Phase 3 EMBARK trial results landed, the primary endpoint — a composite functional motor score — was not met. The drug produced the protein. Whether that helped patients move, breathe, or live longer remained unproven.
Peter Marks had the authority, and the obligation, to take that evidence seriously. Instead, he converted the accelerated approval to full approval and broadened the patient population. His own agency’s scientists had told him not to.
Three deaths, and a reckoning
The consequences arrived in 2025. In March, a teenage non-ambulatory boy died from acute liver failure less than two months after receiving Elevidys. In June, a second teenage non-ambulatory patient died the same way. In July, a third death — this time a 51-year-old patient in a clinical trial using the same AAVrh74 viral vector that underpins Elevidys. Sarepta’s share price collapsed 42% on the second death.
Liver toxicity is a known risk with adeno-associated virus gene therapies. It was in Elevidys’ prescribing information. What was not in the label — and what the deaths made unavoidable — was a black box warning for acute liver failure and death.
Vinay Prasad, who had succeeded Marks as CBER Director in May 2025, was unsparing in his assessment. Marks, he wrote on X, had “overturned 3 reviewers to approve a gene therapy that seems to be killing children and blowing their livers up.” The language was intemperate, perhaps, but the underlying charge was accurate.
On 18 July 2025, the FDA placed a clinical hold on Sarepta’s limb-girdle trials, revoked the company’s platform technology designation, and requested a voluntary suspension of all Elevidys shipments — including for ambulatory patients, where the deaths had not occurred. Sarepta refused. The company argued, not unreasonably, that the safety signal was in non-ambulatory patients and that ambulatory patients were being denied a therapy that might benefit them. Three days later, under pressure, Sarepta relented — citing the need to maintain a “productive and positive working relationship with FDA.”
That phrase deserves a moment’s reflection. A company that had initially defied a federal regulator’s safety request ultimately complied not because the science demanded it, but because the commercial relationship required it. The agency, meanwhile, was being torn apart from the outside: rare disease advocates, political figures, and media voices mounted an intense campaign against Prasad for being too restrictive. He resigned. Then returned ten days later. Commissioner Makary subsequently resigned too.
What this tells us
It would be convenient to lay all of this at the feet of Peter Marks. He made poor decisions, repeatedly, in the face of clear scientific advice. But the system that allowed him to do so — that in fact incentivised him to do so — is the more important subject.
The accelerated approval pathway creates perverse incentives. Companies invest heavily in patient advocacy, knowing it moves regulators. Regulators, in turn, face enormous public pressure to approve therapies for conditions where no alternatives exist, and enormous reputational risk if they are seen to deny hope to desperate families. The scientific staff who do the careful work of reviewing evidence find themselves overruled by senior officials who are navigating political, commercial, and media dynamics that have nothing to do with the data in front of them.
This is financialisation applied to regulation. The same forces that have hollowed out pharmaceutical manufacturing — the prioritisation of speed, return on investment, and market positioning over slow, careful, evidence-based development — have now reached deep into the regulatory apparatus itself. The FDA is not immune to the industry’s pathologies. It is shaped by them.
A different playbook
Peter Marks, now outside the agency, has been calling for a new regulatory framework for gene therapies — one based on process validation rather than product-by-product clinical packages. There is something to be said for the idea. But it is worth noting that the man proposing a new system is the same man whose decisions under the old one led to three deaths and a crisis of institutional credibility.
The structural reforms needed go further than any one approval pathway. They require insulating scientific reviewers from political override. They require post-approval surveillance that is robust enough to catch safety signals before they become fatalities. They require honest reckoning with the limits of surrogate endpoints — and with the commercial interests that make surrogate endpoints so attractive to everyone in the room except the patient receiving the therapy.
The Elevidys story is not finished. Sarepta’s AAVrh74 vector has since received a new FDA platform technology designation. The regulatory rehabilitation is already underway.
The children who died cannot be rehabilitated. They deserved better from a system that was supposed to protect them.
Good article, shows the real difficulties of pushing the boundaries.
Is the answer to let the patient decide? But, to do that the scientific pros and cons need to presented in an unbiased form and in a language that is acceptable to the patient.
Taking the view that nothing is safe the patient needs to consider if they want quality of life or quantity, and only the patient can decide that. The question would be Is it worth the risk, as with all drugs. Only the patient can decide that.
The term approved by FDA is perhaps misleading in that it infers too much: no drug is safe they all carry risk of harm, yet some are approved.
Of course at $3.2 million it is an academic decision, until you can get funding.
I read last week that Vinay Prasad is no longer head of biologics at FDA. It sounds like he was removed for political reasons. I read he blocked a number of gene therapies from going to market including a gene therapy for flu from Moderna. As a result Pharma has had a challenging year financially because as you mentioned there’s nothing new coming down the pipeline. It sounds like the next individual who heads up CBER will likely be more accommodating to Pharma. An mRNA vaccine for the flu sounds pretty scary when you consider the damage from the Covid shots.