Congress created a tool in 2022, through the Food and Drug Omnibus Reform Act, called platform technology designation. The idea was to speed rare-disease drug development: if a company uses a well-understood technology across several programs, it can reuse data about that technology to support multiple applications, rather than proving the same thing from scratch each time. The FDA granted Sarepta such a designation for its AAVrh74 viral vector, the delivery system shared by its approved Duchenne therapy Elevidys and a set of investigational gene therapies for limb-girdle muscular dystrophy.
Then the logic ran in reverse. After three deaths from acute liver failure, two in Elevidys-treated Duchenne patients and one in an LGMD trial participant, all linked to therapies using that same vector, the FDA placed a clinical hold across all of Sarepta's LGMD trials and revoked the platform designation. In a STAT First Opinion essay, an LGMD patient named Kat Bryant Knudson argues the agency turned the tool inside out, using the platform concept to halt everything at once while restoring access only case by case, leaving patients like her stranded. For people with a progressive disease, she writes, time is not neutral: every month of delay means irreversible muscle loss.
Her frustration is understandable and her underlying point is real, but the specific inconsistency she identifies is more defensible than it appears, and seeing why matters, because the genuine problem lies somewhere adjacent to where she locates it, and the difference changes what reform would actually help.
The apparent inconsistency
The grievance is easy to state. The FDA treated the programs as one thing when it wanted to stop them, invoking the shared vector to justify a blanket hold across every LGMD trial, and then treated them as separate things when it came to restarting, applying careful, disease-by-disease benefit-risk judgment that restored access for one disease state while leaving the LGMD patients on hold. Used the connection to halt broadly; used the distinctions to restore narrowly. Framed that way, it looks like the agency reached for whichever logic stopped treatment, an asymmetry that falls on the patients left waiting.
If that were an accurate description of arbitrary inconsistency, it would be a serious charge. But the asymmetry between how safety and benefit evidence behave is not arbitrary. It reflects a real difference in the two kinds of evidence, and once that difference is clear, halting broadly while restoring narrowly stops looking like a contradiction and starts looking like the logically correct response to the situation.
Why safety aggregates but benefit does not
Here is the distinction that dissolves most of the apparent inconsistency. Safety risk and clinical benefit have different logical structures when a component is shared across programs.
The deaths appear to have been caused by acute liver failure, a known and serious danger of high-dose AAV gene therapy, and critically, the suspected culprit is the vector itself, the AAVrh74 delivery system common to all the programs. If the shared vector is what damages the liver, then that risk is a property of the vector, not of any one disease program, which means a death caused by the vector in one program genuinely is evidence about the vector's risk in every program that uses it. The harm travels with the shared component. So halting broadly on a shared safety signal is not overreach; it is the correct inference. When the thing that killed patients is the thing all the programs have in common, pausing all of them while you investigate is exactly what caution requires.
Benefit works the opposite way. The benefit of a gene therapy is not a property of the vector; it is a property of what the therapy does for a specific disease in specific patients. Elevidys has an established, FDA-reviewed benefit in certain Duchenne patients. The LGMD programs are earlier in development, with less-proven benefit, in different patient populations with different baseline risks. The shared vector does not make their benefits equivalent, because benefit does not aggregate across diseases the way vector risk does. So when the FDA weighs whether to restore access, it has to do so disease by disease, because the risk is now shared and demonstrated while the benefit is specific and, for the LGMD programs, less established.
Put those together and the pattern the patient calls inconsistent is coherent: aggregate the risk, because the risk is shared, and disaggregate the benefit, because the benefit is not. Halt broadly, restore narrowly. It is not the agency gaming the logic. It is the agency correctly recognizing that a shared cause of harm and a disease-specific benefit call for different scopes of judgment.
Where the real grievance survives
None of that makes the patient's frustration wrong. It relocates it. The valid version of her complaint is not that the FDA was inconsistent between halting and restoring. It is that the process carries a deep, systematic bias toward inaction, one that shows up in exactly these situations and weighs against progressive-disease patients specifically.
The bias comes from an asymmetry in what gets counted. When a therapy causes a death, that death is visible, attributable, and lands squarely on the agency that allowed the therapy: a name, a case, a clear line from decision to harm. When a hold causes harm, the irreversible muscle loss a progressive-disease patient suffers during the months of delay, that harm is diffuse, invisible, statistical, and attributed to no one. Nobody writes down that a specific patient lost a specific increment of function because a trial was paused for a specific number of months. The death from acting is counted; the decline from waiting is not. And a process that counts one kind of harm vividly and the other kind not at all will systematically err toward whichever choice avoids the counted harm, which is inaction. Regulators are not being cowardly when this happens; they are responding rationally to an accounting system that records their visible mistakes and ignores their invisible ones.
That is the true asymmetry in the LGMD case, and it is genuinely unfair to the patients, but it is a different unfairness than the one the op-ed names. It is not that the platform logic was applied in two directions. It is that the cost of delay, which for an irreversible progressive disease is real and mounting, does not appear on the ledger the FDA is effectively keeping, so it gets underweighted against the highly visible cost of a treatment-related death. When the patient writes that time is not neutral, she is naming precisely the harm the process fails to count.
What this does and does not justify
Holding both of these together, the safety logic and the accounting bias, points toward a specific and limited conclusion rather than a blanket verdict. It does not justify restoring access to the LGMD programs regardless of the safety signal. Three deaths from acute liver failure tied to a shared vector are not a bureaucratic pretext; they are a real, mechanistic danger, and the desperation of the patient population does not make the vector any safer. Nor does it justify abandoning the disease-by-disease benefit-risk approach, which is the correct way to handle a shared risk against specific benefits. The uncomfortable part of the honest analysis is that the LGMD programs may be on hold not because the agency is being inconsistent but because their benefit-risk genuinely looks worse than Elevidys's: the same now-demonstrated vector risk, weighed against benefit evidence that is thinner and earlier. That is a defensible reason to move more cautiously on them specifically, and it is not one the patient's framing fully engages.
What the analysis does justify is a demand that the invisible cost be made visible and counted. The reform worth wanting is not looser safety oversight; it is a benefit-risk process that explicitly weighs the harm of delay for irreversible progressive diseases, and that moves through the disaggregated re-adjudication with urgency proportionate to how fast the untreated disease takes function away. A hold that is correct in principle can still be wrong in duration, and the way it becomes wrong is by treating the clock as free. If the FDA counted the muscle lost during each additional month of hold with the same seriousness it counts a treatment-related death, the disaggregated benefit-risk math would not change its structure, but it would change its pace, and pace is much of what the patient is actually asking for.
How to read it
The LGMD case is a genuine collision between two legitimate things: a real safety signal that warrants caution, and a real cost of delay that warrants urgency. It is not a simple story of a heartless agency versus wronged patients, and it is not a simple story of prudent regulators versus impatient advocates. Both the FDA's caution and the patient's frustration are grounded, which is what makes the case hard rather than obvious.
The clarifying correction is that the inconsistency the op-ed identifies, halting broadly and restoring narrowly, is mostly not an inconsistency at all, because safety risk from a shared vector genuinely aggregates while disease-specific benefit genuinely does not. The valid grievance underneath it is deeper and harder to fix: a regulatory process that vividly counts the harm of acting and fails to count the harm of waiting will always tilt against patients with progressive diseases, for whom waiting is itself a serious, irreversible harm. The answer is not to stop taking three deaths seriously. It is to start taking the cost of delay seriously enough to put it on the same ledger, so that when the agency weighs whether and how fast to restore access, the muscle these patients are losing while they wait is counted as something, rather than as nothing. That would not lower the safety bar. It would only stop pretending that time, for these patients, is free.
Primary sources
- STAT's First Opinion essay by Kat Bryant Knudson for the patient's account of living with LGMD, the argument that the FDA is misapplying platform technology designation created by FDORA 2022, the claim that the agency used the platform to halt all programs while restoring access only for one disease state, the observation that LGMD 2E/R4 patients remain on hold, and the framing that for progressive diseases time is not neutral and delay means irreversible loss.
- Fierce Biotech and CGTLive for the FDA's July 2025 clinical hold across all of Sarepta's investigational LGMD trials and revocation of the AAVrh74 platform technology designation, the three deaths potentially related to the products, the detail that two occurred in Elevidys-treated Duchenne patients and one in a 51-year-old LGMD 2D/R3 patient receiving SRP-9004, and that all three deaths appear caused by acute liver failure.
- TipRanks/TheFly and Sarepta releases for the scope of the hold across SRP-9003, SRP-9004, SRP-6004, and SRP-9005, the completed enrollment and dosing in the SRP-9003 Phase 3 trial, and the company's intent to seek an accelerated-approval pathway after the hold is lifted.
- CGTLive for Sarepta's original platform technology designation for the rAAVrh74 vector and the FDA's description of the designation's purpose.
- Clinical Trials Arena, Patient Worthy, and Pharmacy Times for context on the LGMD treatment landscape, the absence of approved therapies, BridgeBio's competing oral candidate, and the progressive nature of the disease with respiratory failure as the primary cause of mortality.