16 Points Up, 7.6 Down: The $3.95M Math Behind the First Sanfilippo Gene Therapy
On September 17, the FDA approved Fayuvi, the first treatment for Sanfilippo syndrome type A, a fatal childhood brain disease, at $3.95 million per one-time infusion. The headline number is a 23.5-point cognitive advantage. The decomposition is more interesting, and so is the business model: peak sales of $200–250 million implies roughly 57 treated children a year, worldwide.
Fifty-seven.
That is roughly how many children per year, worldwide, J.P. Morgan’s peak-sales estimate says will receive the most expensive gene therapy ever approved for a brain disease. Not fifty-seven thousand. Fifty-seven.
On September 17, 2026, the FDA granted standard full approval to Fayuvi (rebisufligene etisparvovec-hopf), Ultragenyx Pharmaceutical’s one-time intravenous gene therapy for Sanfilippo syndrome type A, also called mucopolysaccharidosis type IIIA, making it the first treatment of any kind for a disease that had previously offered families nothing beyond supportive care (FDA news release). No treatment had ever altered the disease’s course. Children with Sanfilippo type A inherit two broken copies of the SGSH gene, cannot manufacture the enzyme sulfamidase, and accumulate heparan sulfate in the brain until cognition, language, and motor function unravel in early childhood, with a median life expectancy of about fifteen years.
Fayuvi uses an adeno-associated virus serotype 9 vector to deliver a working SGSH gene, dosed once at 3.0 × 1013 vector genomes per kilogram, at a U.S. list price of $3.95 million. That makes it the second-priciest gene therapy on record, behind Orchard Therapeutics’ Lenmeldy at $4.25 million (Reuters). Ultragenyx shares rose 13 percent to $14.50 on the news, one day after touching a 52-week low of $12.73.
Now the number everyone quoted, and the one nobody decomposed.
In the pivotal Transpher A trial, seventeen treated children with a mean age of 21.8 months at treatment were compared against twenty-seven untreated children drawn from a natural-history cohort, with efficacy defined as the mean change in Bayley-III Cognitive raw score, a zero-to-ninety-one scale, measured between the ages of twenty-four and sixty months. Treated group: +16.0 points. Untreated group: −7.6. Difference: 23.5 points, with a 95 percent confidence interval of 17.2 to 29.9 and a p-value below 0.0001 (Optum Rx approval brief).
Read that decomposition again. Treated children did not merely stop declining; they gained sixteen points, about 17.6 percent of the scale’s full range, while the comparison group lost 7.6 points, about 8.4 percent of the range, producing a total swing of roughly twenty-six percent of the scale over three years. “Halted regression” undersells the result by half. Their trajectory bent upward rather than flattening, and long-term follow-up extending to nearly eight years found that cerebrospinal-fluid heparan sulfate remained reduced across the treated age groups (Voice of Healthcare).
That is the medicine, and here is the business, which looks like nothing else in pharma. J.P. Morgan estimates peak worldwide sales of $200–250 million, and taking the midpoint of $225 million and dividing by $3.95 million per dose yields about fifty-seven treated patients per year at peak, globally. Set against that, Ultragenyx’s own figure of 3,000–5,000 patients in “commercially accessible geographies,” multiplied by the list price, produces $11.9–19.8 billion, a number that will never materialize because most prevalent patients are already past the label’s “preserved neurodevelopmental function” requirement and because list price is not net price. My honest model is fifty-some children a year, each worth several million dollars of gross revenue, with no refills, ever.
| One-time gene therapy | Indication | U.S. list price |
|---|---|---|
| Lenmeldy (Orchard) | Metachromatic leukodystrophy | $4.25M |
| Fayuvi (Ultragenyx) | Sanfilippo syndrome type A | $3.95M |
| Genglycos (Ultragenyx) | Glycogen storage disease type Ia | $2.7M |
| Casgevy (Vertex) | Sickle cell / beta thalassemia | $2.2M |
Incident math explains why the patient count stays small, because published epidemiology puts MPS IIIA’s lifetime risk at birth between effectively zero and 1.62 per hundred thousand live births (Zelei et al. systematic review). Against roughly 3.6 million U.S. births a year, the upper bound is about fifty-eight new type-A cases annually; European-derived estimates center on fifteen to forty. But the label demands preserved neurodevelopmental function, and without universal newborn screening, diagnosis usually follows symptom onset while developmental slowing begins in the second year of life, which means the treatable incident pool is a fraction of an already tiny number, plausibly ten to thirty children per year in the United States. A prospective natural-history study warned explicitly that for optimal outcomes, treatment may need to precede steep cognitive decline and brain atrophy (Shapiro et al.). At list price, that is a theoretical $40–120 million in annual U.S. revenue before any payer discount, which is also, not coincidentally, the scale at which a thirteen percent stock pop on a fifteen-dollar stock starts to look rational.
Ultragenyx’s pricing defense is that lifetime supportive care for a Sanfilippo child can approach $8 million, making $3.95 million look like a bargain at roughly half the cost. That $8 million is the company’s estimate, not an independent cost study, and it assumes the full disease course with maximal care. Payer arithmetic runs harsher: $3.95 million leaves the building on day one, while the benefit accrues over years that durability data have not yet covered. Somebody has to underwrite that timing mismatch. “The science is beautiful” is not an accepted form of collateral.
One more line item belongs in the model.
It may be the most lucrative one. Both of Ultragenyx’s 2026 gene-therapy approvals, Genglycos in August and Fayuvi in September, arrived with sellable priority review vouchers, which analyst Sami Corwin of William Blair said should significantly bolster the company’s balance sheet (BioPharma Dive). Vouchers of this type have historically changed hands for on the order of $100 million each, so at fifty-seven patients a year, a single voucher sale can rival multiple years of Fayuvi revenue. This drug treats children; the voucher treats the balance sheet.
Context on the regulatory path. The FDA rejected Fayuvi in July 2025 over manufacturing issues rather than the clinical data, Ultragenyx resubmitted in January 2026 seeking accelerated approval, and the agency instead granted standard full approval two days ahead of the September 19 action date. Full approval on a single-arm, open-label study with an external control is the FDA at its most flexible, and it tells you exactly how seriously the agency weighs “uniformly fatal, no alternative” in pediatric rare disease.
The Strongest Case Against This Article
State the skeptic’s case at full strength, because it is strong: seventeen treated children, no randomization, no placebo, no blinding, measured against twenty-seven historical controls: that is the trial design the FDA exists to reject, and in any common disease it would have been rejected outright. External controls flatter, because natural-history cohorts can differ in supportive care, diagnostic era, and ascertainment, and a 23.5-point gap computed against a declining historical group is the rosiest possible framing of the counterfactual. Its label’s “preserved neurodevelopmental function” is doing quiet work as well, since the trial children were treated at a mean age of 21.8 months or were older children with a cognitive developmental quotient of at least sixty, so the result may not generalize to the later-diagnosed children who make up much of the prevalent population. On safety, the label carries hepatotoxicity, thrombocytopenia, thrombotic microangiopathy, infusion reactions, and a theoretical malignancy risk from vector-genome integration, plus a corticosteroid course that starts before infusion and runs at least eight weeks in toddlers. And on price, $3.95 million for evidence this thin sets a precedent that every subsequent ultra-rare therapy will cite for the next decade. If you believe small, uncontrolled trials should not support near-$4-million price tags, this approval is the counterexample you will be arguing about for years.
What This Does Not Prove
Here are the honest bounds. Durability is the biggest open question, because the longest follow-up runs about eight years, AAV-delivered transgenes are known to wane over time, and nobody has lifetime data on children treated this young. Generalizability is narrow by design: the label leaves “preserved neurodevelopmental function” to physicians and families, so real-world outcomes in later-diagnosed children remain unknown. Underneath the incident math sits poor-quality epidemiology, with the systematic review itself concluding that higher-quality data are needed, which is why the U.S. estimate spans fifteen to fifty-eight new cases a year. All revenue arithmetic here uses the $3.95 million list price; net prices after payer negotiation are undisclosed and will certainly be lower. That $8 million lifetime-care figure is Ultragenyx’s, not an independent study. And the voucher math is directional: vouchers were issued, but no sale price has been disclosed.
What You Can Do
Four moves, depending on who you are.
If you are a parent of a toddler showing developmental slowing plus new behavioral changes, push for genetic and metabolic testing early and explicitly ask about MPS III, because the treatment window is “preserved neurodevelopment” and every month of diagnostic delay can close it. If you are a pediatrician, remember that Sanfilippo is not on universal newborn screening, which makes the eighteen- and twenty-four-month developmental checks the de facto screening program, and a child regressing after normal early milestones deserves a workup rather than reassurance. If your family is facing a diagnosis, Ultragenyx’s UltraCare program at 888-756-8657 coordinates access, treatment runs through qualified U.S. gene-therapy centers with Nationwide Children’s among them, and shipping to centers was expected thirty to sixty days post-approval. That network is coming online now. And if you make policy, the case for adding MPS IIIA to newborn screening panels just became dramatically stronger, because a therapy that works only before decline makes early detection the binding constraint on its value. Screen first, then treat, because the current order is backwards.
The Bottom Line
A one-time infusion now offers children with Sanfilippo type A something that did not exist six weeks ago: a chance to gain cognitive ground instead of losing it, with treated children up sixteen points on a ninety-one-point scale while untreated children fell 7.6. Its $3.95 million price is real, the evidence base is seventeen children and a historical comparison, and the business runs on roughly one patient a week, worldwide. It is simultaneously a genuine medical milestone and a case study in ultra-rare economics, where the drug is priced for the fifty-seventh patient as much as the first, and what happens next is not a scientific question but a logistical one. Whether newborn screening and fast diagnosis can find children while their neurodevelopment is still worth preserving will decide how much of that milestone reaches patients. The therapy arrived, but the system that delivers it on time has not.