The FDA’s AAV Approval Playbook Has Gaps That Could Define the Next Decade of Gene Therapy

Picture the regulatory affairs lead at a mid-size gene therapy sponsor sitting across from her team in January 2027. She has a Phase 2 AAV program for a rare CNS disorder, a compelling mechanism, and a dataset that looks, on paper, every bit as strong as the one Novartis carried into the FDA for Zolgensma’s BLA. She opens, and she looks for the precedent that will anchor her submission strategy. What she finds instead is a patchwork of case-by-case decisions, post-hoc manufacturing commitments, and long-term follow-up obligations that were defined after approval, not before. The map she needs does not exist. What she is holding is a collection of footprints.

The FDA has approved several AAV-based gene therapies, including Zolgensma in 2019, Luxturna (voretigene neparvovec) for inherited retinal disease before it, and Hemgenix (etranacogene dezaparvovec-drlb), CSL Behring’s hemophilia B therapy approved November 22, 2022. Each approval expanded the evidence base for what the agency will accept. Each also exposed, in the public record, exactly where the FDA improvised. For sponsors now designing pivotal programs, those improvisations are not just historical curiosities. They are the operational terrain.

The Manufacturing Standard That Moves

Start with chemistry, manufacturing, and controls, because that is where most AAV programs break down before they ever reach a clinical endpoint discussion. The FDA’s Center for Biologics Evaluation and Research published a draft guidance in September 2024, “Frequently Asked Questions: Developing Potential Cellular and Gene Therapy Products,” under docket FDA-2024-D-4311, consolidating recurring questions from CGT developers about exactly these issues. The fact that CBER needed to publish a FAQ consolidation six years after Zolgensma’s approval tells you something about how clearly the original expectations were communicated.

What the approved AAV programs revealed about CMC is both instructive and uncomfortable. Zolgensma’s approval came with manufacturing commitments that Novartis was required to fulfill post-approval, meaning the FDA accepted a level of CMC uncertainty at licensure that it would likely scrutinize more heavily today. The agency’s posture has tightened, but it has tightened through ad hoc enforcement rather than through published, prospective standards. Sponsors reading the 2024 draft guidance will find answers to questions they already encountered in pre-IND meetings. They will not find a consistent vector potency specification framework or a harmonized approach to impurity thresholds across AAV serotypes.

That gap has real consequences. A sponsor developing an AAV9-based CNS program cannot point to Zolgensma’s BLA and extract a potency specification template, because Zolgensma’s specifications were negotiated in the context of its specific manufacturing process and patient population. Each new program must rebuild that negotiation from scratch, with a moving agency target.

What the Pivotal Trials Actually Established

Zolgensma’s pivotal data came from the START and STR1VE-US studies. The START trial enrolled a small number of patients with Type 1 SMA and demonstrated functional motor improvements that were striking against the natural history of the disease. STR1VE-US expanded the confirmatory evidence. Those enrollment numbers, small even by rare disease standards, are worth pausing on. For any other therapeutic modality, a 15-patient pivotal dataset would invite a Complete Response Letter before the submission was fully docketed. The FDA accepted it because the unmet need was catastrophic and the effect size was large enough that chance confounding was implausible.

But the precedent this sets is frequently misread by sponsors. The lesson drawn in many regulatory strategy sessions is “small trials can work for rare disease gene therapy.” The more precise lesson is that small trials can work when the disease is rapidly fatal without treatment, when the endpoint is unambiguous functional rescue, and when the manufacturing process is sufficiently characterized to assure dose consistency. Strip away any one of those three conditions and the FDA’s flexibility evaporates. Sponsors developing AAV programs for less acute indications, where natural history unfolds over years and functional endpoints require patient-reported instrumentation, face a fundamentally different evidentiary bar even though they are building on the same genetic therapy platform.

Hemgenix introduced another layer of complexity. Its November 2022 approval was supported by a pivotal single-arm study in adults with hemophilia B measuring factor IX activity levels as the primary endpoint. CSL Behring and uniQure demonstrated durable factor IX expression, but the approval also required a robust post-marketing commitment for long-term efficacy and safety surveillance. That commitment reflects an uncomfortable regulatory truth: the FDA approved Hemgenix knowing that the five-year post-treatment observation period referenced in its own January 2020 guidance for AAV vectors would not be complete for years.

The agency’s January 2020 final guidance, “Long Term Follow-Up After Administration of Human Gene Therapy Products,” specifies five years of post-treatment observation for AAV vectors. For integration risk assessment, the guidance references extended observation periods beyond five years for some product categories. Approving products before that window closes means the FDA is making benefit-risk judgments under uncertainty it has explicitly quantified and then chosen to carry forward into the market. That is a defensible posture for severe, life-threatening disease. Sponsors need to understand it as a posture, not a precedent that generalizes cleanly.

The Integration Signal Nobody Wants to Talk About

Reports have emerged linking Zolgensma to hepatocellular tumors in recipients, with evidence suggesting the AAV vector had integrated into the genome near oncogenic loci in at least some cases. Novartis disclosed these findings, and the FDA reviewed them. The agency did not pull the product. What regulators concluded, reasonably, is that the benefit-risk balance for a child dying of Type 1 SMA still favors treatment. But those two cases fundamentally changed what the FDA will expect from new AAV programs in terms of integration site analysis, biodistribution characterization, and long-term registry commitments.

The FDA’s guidance on biodistribution studies specifies that quantitative PCR assays used in preclinical biodistribution must meet defined sensitivity thresholds, with particular attention to germline transmission and gonadal tissue analysis. What the Zolgensma tumor cases added to that framework is a heightened expectation around integration site sequencing in preclinical programs, including for predominantly non-integrating AAV vectors, an expectation that now appears in pre-IND feedback even though no final guidance document codifies it with specificity for gene therapy programs broadly; a 2026 FDA draft guidance addresses integration site analysis for genome-editing products specifically, but no equivalent final rule covers the wider AAV context.

This is the structural problem the approved AAV programs collectively expose: the most important regulatory expectations are communicated through meeting feedback, complete response letters, and post-approval commitments rather than through guidance documents that sponsors can use to build prospective protocols. The FDA is effectively publishing its AAV policy in installments, using each approval and each post-market safety signal as the vehicle. Sponsors who read only the guidance documents are operating on outdated intelligence.

Which raises a question that the September 2024 FAQ guidance does not fully answer: how should a sponsor design a Phase 2 AAV trial today to anticipate the manufacturing, safety, and follow-up standards that the FDA will apply at BLA review two or three years from now?

Building the Protocol the FDA Isn’t Describing

The operational answer requires treating the three approved AAV programs not as templates but as stress tests. Run your protocol against each approval and ask what the FDA required that your current design does not address. For manufacturing, the Zolgensma precedent argues for locking your vector potency assay before Phase 2 begins, not before Phase 3. CMC changes between Phase 2 and Phase 3 have triggered comparability exercises that caused significant program delays. The 2024 CBER FAQ guidance warns explicitly that manufacturing changes after dose selection can compromise the ability to bridge clinical data.

For long-term follow-up, the Hemgenix approval argues for embedding your five-year observational registry infrastructure into the pivotal trial design from day one, with a pre-specified statistical analysis plan for the post-approval data. FDA reviewers have flagged programs that treat long-term follow-up as a post-approval problem to solve later. When the reviewer asks in a Type B meeting how you plan to maintain five-year follow-up in a rare disease population where patients may change institutions, lose insurance coverage, or simply disengage, the answer “we will figure that out after approval” is not an answer the agency accepts anymore.

For safety monitoring, the Zolgensma tumor cases argue for designing integration site analysis into your preclinical program now, before the FDA formalizes the requirement in guidance. Sponsors who wait for the guidance will find themselves retrofitting studies that should have been prospective. The FDA has signaled in pre-IND feedback across multiple CNS AAV programs that it wants integration site distribution data from relevant animal models with adequate sample sizes, using next-generation sequencing methods sensitive enough to detect low-frequency events. The agency has referenced informally a sensitivity floor requiring detection of low-frequency integration events, though no published guidance document codifies a specific numeric threshold.

None of this is published in a single document the sponsor can cite. It lives in the institutional memory of regulatory affairs teams that have been through AAV pre-IND meetings in 2024 and 2025, and in the inspection record of manufacturers who have learned what the agency expects by receiving 483 observations after the fact. The GMP enforcement environment for cell and gene therapy manufacturing has tightened visibly since 2023, with FDA inspections at AAV facilities producing findings around environmental monitoring, raw material qualification, and in-process testing that reflect standards beyond what the current guidance documents specify explicitly.

The regulatory affairs lead staring at the Zolgensma approval summary in January 2027 is not looking at a roadmap. She is looking at the FDA’s field notes from a decade of improvisation, written in the language of post-approval commitments and case-by-case decisions. The sponsors who build their Phase 2 protocols around what those field notes actually say, rather than what the guidance documents formally require, are the ones whose BLAs will not trigger a complete response cycle over issues the agency has been signaling for years. The integration site sequencing she runs in her preclinical program this quarter may be the data point that makes the difference. The registry infrastructure she builds now may be the operational detail that determines whether her approval is a footnote or a precedent.

References

1. Nature Reviews Drug Discovery, “Regulatory lessons from FDA review of approved AAV gene therapies”
2. FDA, “FDA Approves Innovative Gene Therapy to Treat Pediatric Patients with Spinal Muscular Atrophy” (Zolgensma approval, May 24, 2019)
3. Drugs.com, “Hemgenix Approval History” (etranacogene dezaparvovec-drlb, November 22, 2022)
4. AllSci, “FDA Draft Guidance: Frequently Asked Questions, Developing Potential Cellular and Gene Therapy Products (FDA-2024-D-4311, September 2024)”
5. FDA, “Long Term Follow-Up After Administration of Human Gene Therapy Products; Guidance for Industry” (January 30, 2020)
6. BioPharma Dive, “Zolgensma tumor reports: Novartis discloses AAV vector integration in cancer cases (2023, 2025)”
7. Halloran Consulting Group, “Implementing FDA Guidance on Long-Term Follow-Up for Gene Editing” (2022)
8. RehabPub, “FDA Approves Zolgensma Gene Therapy: START and STR1VE-US Trial Data Summary”