The FDA’s new draft gene therapy guidance landed this morning, and the phrase that stopped me appears on page after page: established platform knowledge. Sponsors are invited to lean on it. It will, the agency promises, “raise our collective efficiency while maintaining the highest standards.”

Ten months ago, this same FDA revoked Sarepta Therapeutics’ AAVrh74 platform technology designation. Three patients treated with Sarepta’s gene therapy products had died of acute liver failure. The platform had not held.

So when the agency tells me that on June 2, 2026 it has issued draft guidance titled Leveraging Prior Knowledge in the Development of Human Gene Therapy Products Incorporating Genome Editing, I want you to see the collision plainly. The load-bearing concept in this document is the exact bet the FDA itself just had to walk away from.

Why “platform” is the load-bearing word

An AAV gene therapy works like this: you package a corrected gene inside an engineered virus, an adeno-associated virus, inject it intravenously, and the virus delivers the genetic cargo into the patient’s cells. The viral capsid is the platform. Swap the cargo for a different disease and you can, in theory, skip a chunk of development for the next program. That is what “platform knowledge” means in practice. That is what Sarepta was selling.

AAV has one feature that makes the platform conversation more dangerous than it sounds. Once a viral vector is in your bloodstream and your immune system has reacted to it, you cannot un-dose it. There is no taper. There is no retitration. The dose is the experiment. Which is why platform safety has to be earned the hard way, body by body, before any regulator gets to wave a hand and say we already know how this behaves.

The legal distinction, and why it does not save the argument

The FDA will tell you, fairly, that the June 2 guidance distinguishes its use of “platform knowledge” from the statutory platform technology designation that Sarepta lost. Different program, different bar. The new document is technically scoped to human gene therapy products incorporating genome editing, with a note that some recommendations may also apply to AAV vector products. CRISPR-style edits, base editors, prime editors. Sarepta’s Elevidys is AAV gene replacement, not gene editing.

Fine. Read what the guidance actually invites sponsors to leverage, then, and see if the difference reassures you. Internal company platform knowledge across an applicant’s portfolio. FDA master files from prior submissions. Chemistry-manufacturing-controls data from related products. Nonclinical and clinical results from earlier programs. Peer-reviewed literature. Consortium data-sharing initiatives. The logic is identical across statutes: we have learned things from one program, let us count them as known in the next.

The question is whether the FDA can identify the kind of platform learning that holds across programs and the kind that does not, after the agency itself just revoked a designation built on the same assumption.

Who is at the agency right now

The two senior drug-review centers are both being run by acting directors. Vinay Prasad announced his departure from CBER in March 2026 and exited at the end of April, after the Sarepta episode and the Elevidys decisions it forced. Marty Makary, the commissioner, left. Tracy Beth Høeg, the acting CDER chief, walked out behind them. Katherine Szarama, the next acting CBER director, lasted three weeks before she was replaced on May 20.

Her successor is Karim Mikhail, an ex-Merck executive who spent more than two decades at the company before serving as president and CEO of Amarin Corporation. Mikhail joined the FDA in 2025 as a senior advisor leading what the agency politely calls “U.S. competitiveness reforms.” Read competitiveness as accelerating approvals.

Mikhail’s pull-quote on the new guidance: “Today’s action reflects the FDA’s commitment to get safe and effective cell and gene therapies to patients faster.” His deputy at the Office of Therapeutic Products, Vijay Kumar, offered the line that always demands a second read: “Leveraging prior knowledge does not mean lowering the bar; it means raising our collective efficiency while maintaining the highest standards.”

Whenever an agency official tells you a new pathway does not lower the bar, the prior question is whether the bar was holding up the floor it was supposed to.

How we got here

The Plausible Mechanism Pathway, released February 23, 2026, was the marquee project of Commissioner Makary and his then-CBER director Prasad. The two of them previewed it in the New England Journal of Medicine in November 2025. For individualized therapies targeting ultra-rare mutations, the framework said, a plausible biological mechanism plus surrogate endpoints could substitute for a traditional showing of clinical benefit.

The June 2 guidance is the operational follow-on. Where the February document said here is a flexible path to approval, the June document tells sponsors how to fill it. The cost of a new gene therapy submission goes down. The companies who benefit most are the ones with prior platform knowledge to reuse: portfolio depth, master files at FDA, consortium memberships, and the resources to monetize all of it. The small developer chasing a mutation eleven patients share still gets a politely worded INTERACT meeting and the same biology to solve.

The public comment period runs 90 days from Federal Register publication.

The Elevidys receipts

If you want to know whether the bar was holding before any of this, read the FDA’s July 18, 2025 action on Sarepta. Three patient deaths potentially related to AAVrh74-based gene therapy products. Acute liver failure as the apparent mechanism. The agency requested Sarepta voluntarily suspend U.S. distribution of Elevidys and placed the phase 3 ENVISION trial on clinical hold along with the rest of Sarepta’s AAVrh74 program.

And most pointedly, the FDA revoked Sarepta’s AAVrh74 platform technology designation, writing that “the preliminary evidence is insufficient to demonstrate that AAVrh74 Platform Technology has the potential to be incorporated in, or utilized by, more than one drug without an adverse effect on safety.”

That sentence is the counterargument to the June 2 guidance, written by the same agency, less than a year apart. The legal mechanism is technically different. The underlying bet is the same.

What I would watch

I want this technology to work. Sickle cell, Duchenne, lysosomal storage diseases, the inborn errors that kill kids before they finish elementary school. A working gene-editing fix would be one of the great medical advances of our lifetime, and rare-disease families have been waiting their whole lives. They are right to demand speed. That is exactly why platform claims need brutal scrutiny. The kids who get these shots cannot un-take them.

Here is what I will be watching. Which company files first under the new guidance, and who their reviewer at CBER is. Whether the platform knowledge they invoke includes the acute-liver-failure signal tied to AAVrh74, or whether the agency lets sponsors carry forward the convenient findings and quietly leave the inconvenient ones in the prior program’s file. If safety failures count as leveraged knowledge, the bar held. If they do not, the agency lowered the bar on its own way out the door.

Sources

  1. FDA – Draft Guidance: Leveraging Prior Knowledge in the Development of Human Gene Therapy Products Incorporating Genome Editing (2026)
  2. FDA – Press announcement: FDA Issues Draft Guidance to Help Accelerate Cell and Gene Therapies for Patients (2026)
  3. FDA – Press announcement: FDA Requests Sarepta Therapeutics Suspend Distribution of Elevidys (2025)
  4. STAT News – FDA unveils rules for bespoke gene therapies: the Plausible Mechanism Pathway (2026)
  5. AABB – Szarama Departs CBER; Karim Mikhail Named Acting Director (2026)
  6. FDA – Karim Mikhail leadership profile
  7. Fierce Pharma – After Makary departure, acting CDER chief Høeg and CBER’s Szarama exit (2026)