On July 11, 2025, the FDA turned Ultragenyx away. The document that arrived was a complete response letter, the agency’s way of declining to approve an application as it stands, and by the FDA’s own account it did not fault the science. The clinical reviewers had called the trial data robust. What the letter flagged was chemistry, manufacturing, and controls: inspection findings and paperwork at the plant that makes the drug. The molecule was fine. The factory was the problem.
Fourteen months later, on September 17, 2026, the same agency approved that same therapy under the name Fayuvi, the first treatment ever cleared in the United States for Sanfilippo syndrome type A. The science the FDA had praised in the summer of 2025 did not change in between. What changed was the manufacturing file, and, when the list price finally landed, the number attached to a single dose: $3.95 million. That makes Fayuvi the second-most-expensive medicine ever sold in the country, and it goes to market on the strength of 17 treated children measured against a comparison group the company built itself.
The disease earns the drama
Sanfilippo type A, known in the literature as mucopolysaccharidosis IIIA, is one of the cruelest diagnoses in pediatric medicine, and nothing here is meant to soften that. A missing enzyme lets a sugar called heparan sulfate build up in the cells of the brain. Children develop close to normally for a year or two, then begin to slide: first the language goes, then the sleep, then the behavior, then the ability to walk, and finally the recognition of the parents standing over the bed. Most do not survive their teens. Until this approval there was nothing to offer but management of the symptoms while the disease did its work.
So when the FDA calls Fayuvi a “significant scientific milestone,” and when Ultragenyx chief executive Emil Kakkis calls it “an important first step” toward reaching families, the words are doing honest work. A one-time infusion that delivers a working copy of the missing gene to the nervous system is exactly what this field has chased for two decades. The question a reader deserves is not whether that is remarkable. It is what, precisely, the agency saw before it signed off, and how firm the ground under it is.
Seventeen children, and a control group the sponsor built
The approval rests on a single study, Transpher A, open-label and single-arm. No placebo group, no randomization. Seventeen treated children make up the analysis population, and their outcomes were measured against a natural-history cohort of 27 untreated children with the same disease, a comparison group the sponsor assembled from the historical record rather than enrolled alongside the treated ones.
On the top-line measure the result is striking. Treated children scored an average of 23.5 points higher on the Bayley-III cognitive scale than the untreated children over the window from 24 to 60 months of age, with a p-value below 0.0001. Spinal-fluid heparan sulfate, the sugar that poisons the brain in this disease, fell and stayed down. Follow-up in some children runs nearly eight years, and across that stretch the treated children kept gaining cognitive ground while the comparison children did not.
That contrast is the whole case for the drug. But notice what it is and is not. A single-arm trial read against a historical control is the weakest design the FDA will accept, precisely because the two groups were never on the same footing to begin with, and because a sponsor that gets to pick its own comparison cohort has every incentive to pick favorably. The Bayley-III is a developmental test for toddlers. A raw-score gap between ages two and five is an early signal worth having, but what it captures is early cognitive development, not the outcome families live toward: whether a child keeps speaking and walking and knowing their own mother into adolescence. That takes years the trial has only begun to log.
None of this is a reason to have withheld the drug. For a uniformly fatal disease with no alternative, demanding a placebo arm, randomizing dying children to nothing, would be indefensible, and the agency was right not to ask for it. The honest description is narrower and less flattering than the press release: a real effect, measured against a comparison group the maker built, on a childhood cognitive test, over a window that ends before the disease’s worst years begin. It is strong evidence for its class. It is not the settled certainty the word milestone invites you to feel.
The fine print, and the number
Fayuvi is delivered by an AAV9 vector, the kind of gene-therapy carrier built to hand its cargo to a cell without stitching itself into the DNA. Built to. The FDA’s notice warns that, as with other AAV gene therapies, the delivered genetic material could still integrate into the genome and, over the long term, could raise the risk of tumor development, the kind of risk that by definition eight years of follow-up cannot yet have ruled out. The trial’s nearer-term numbers are blunter: liver-enzyme elevations in 85 percent of patients, vomiting in 67 percent, fever in 41 percent, alongside a separate warning for thrombotic microangiopathy.
Then there is the price. At $3.95 million, Fayuvi trails only Orchard’s Lenmeldy, at $4.25 million, on the list of the most expensive medicines ever sold in the United States. No family pays that in cash. The bill lands on insurers and, for a pediatric genetic disease, disproportionately on Medicaid, which is to say the public. Ultragenyx has told investors it can push production toward “biologics-like levels,” and analysts at Jefferies model peak sales around $250 million against a treatable population the company itself puts at 3,000 to 5,000 children in the markets it can reach. The arithmetic of ultra-rare gene therapy is not subtle: a small number of patients, a very large number on each one, and a public payer that cannot say no to a dying child.
Sanfilippo gene therapy has lived in single-digit cohorts for years, a handful of children tested at a time and followed slowly. A separate program in the type B form of the disease has only just published five-year data on five patients. Seventeen treated children is, by the standards of this corner of medicine, a large trial. That tells you how thin the ground is under every claim in the field, not how thick it is under this one.
Fayuvi is a first, and for the parents of a child with Sanfilippo A it is the only door that has ever existed. That is true and it matters. What is also true, and absent from every triumphant version of the announcement, is that the agency turned this exact molecule away in the summer of 2025 over how it was manufactured, that the efficacy case is a single-arm trial read against a comparison group its maker assembled, that the endpoint is a toddler’s cognitive score with the adolescent years still ahead of it, and that the whole thing lists at nearly $4 million a dose.
A complete response letter is not a verdict on the science, and this one, by the FDA’s own telling, never was. The agency liked the data in July of 2025. It took another fourteen months and a corrected manufacturing file to get the drug to the children it had already been shown to help, and not one day of that review touched the number Ultragenyx would charge for it: $3.95 million, set by the company, owed by the public, and questioned by no one with the standing to say no.
Sources
- FDA – Approves First Gene Therapy for Pediatric Patients with Sanfilippo Syndrome Type A (Sept 2026)
- Ultragenyx – Approval of FAYUVI, Transpher A data, safety, and price (Sept 17, 2026)
- BioSpace – Ultragenyx wins first Sanfilippo approval, priced near $4M, after the July 2025 manufacturing rejection
- Ultragenyx – Complete Response Letter for UX111, CMC observations with clinical data called robust (July 11, 2025)
- MedPage Today – Gene Therapy for Sanfilippo Syndrome Gets FDA Nod
- Molecular Genetics and Metabolism – Five-year follow-up after gene therapy with ABO-101 in five children with Sanfilippo syndrome type B (2026)