The U.S. Food and Drug Administration formally announced Thursday that it is investigating a serious risk of malignancy linked to both of Novartis’s approved gene therapies for spinal muscular atrophy — Zolgensma and Itvisma — after molecular analysis of tumor tissue in Zolgensma patients confirmed a specific causal mechanism: the therapy’s viral vector DNA integrating into the genome near a known cancer-driving gene called GLI1 and switching it on.

The FDA’s action, published November 6 by its Center for Biologics Evaluation and Research (CBER), marks the first time a U.S. regulator has formally characterized a commercial AAV gene therapy as having caused insertional oncogenesis in humans — moving the risk from “theoretical,” the language still on Zolgensma’s product label, to confirmed and under active investigation.

The agency says it will “expeditiously” revise the prescribing information for both products.

The Confirmed Finding: What the FDA’s Own Investigation Shows

The safety communication identifies nine total tumor reports in patients previously treated with Zolgensma since its first U.S. approval in May 2019. Of those nine, three cases specifically involved the GLI1 oncogene — Glioma-Associated Oncogene Homolog 1, a transcription factor in the Sonic hedgehog signaling pathway that drives cell proliferation when abnormally overexpressed.

The GLI1 gene is active during fetal and early postnatal development and is normally silenced thereafter. When activated in a mature cell by an external mechanism — including, the FDA now confirms, by the insertion of gene therapy vector DNA — it can initiate uncontrolled cell division.

In two of the three GLI1-linked cases, investigators were able to recover direct forensic evidence from tumor tissue: Zolgensma’s AAV9 vector DNA had integrated into the patient’s genome, with the promoter region of the therapeutic construct inserted immediately upstream of or within the GLI1 gene itself. This promoter element — a potent synthetic sequence designed to drive strong, constitutive gene expression — effectively became a molecular switch permanently turned to “on” for a gene that had no business being expressed.

The mechanism driving GLI1 overexpression in the third case remains under investigation.

The tumors reported across all nine Zolgensma cases have involved the central and peripheral nervous systems, spinal and paraspinal regions, and the kidney. Novartis has stated that the two patients with confirmed causal insertional events received treatment for their tumors and are continuing to be monitored; it has not disclosed tumor types, patient ages, or outcomes beyond that statement.

Itvisma Brought Under the Risk Umbrella — Despite Zero Reported Cases

The FDA’s decision to extend its investigation to Itvisma (onasemnogene abeparvovec-brve), approved just eleven months ago for patients aged two and older, may be the most consequential aspect of its November 6 communication.

No tumor cases have been reported in Itvisma patients. The FDA’s rationale for including it is mechanistic: Itvisma and Zolgensma share the same AAV9 viral vector and the same therapeutic transgene construct. The only difference between them is the route of administration — Zolgensma is delivered intravenously; Itvisma is delivered intrathecally, by direct injection into the cerebrospinal fluid surrounding the spinal cord, targeting older patients for whom IV delivery would not achieve sufficient CNS penetration.

That shared architecture means Itvisma carries the same capacity for vector DNA integration events, using the same CAG-class promoter that the FDA’s investigation has identified as the structural element driving oncogene activation in the confirmed cases. The agency has determined it cannot treat the absence of reported cases as evidence of absence of risk in a therapy approved less than a year ago with a known multi-year latency to tumor presentation.

Itvisma was approved November 24, 2025, based on Phase III STEER and Phase IIIb STRENGTH clinical study results. It was the first gene replacement therapy for SMA approved for use in adults.

A Convergent Mechanism Across Three Independent Programs

The Zolgensma cases do not stand alone. The FDA’s communication explicitly cites two peer-reviewed reports published this year in the New England Journal of Medicine describing analogous integration events in entirely separate gene therapy programs — using different viral capsids, different diseases, and different delivery routes, but the same class of synthetic promoter.

Program 1 — FBX-101 (Krabbe disease): A case report by Vo et al. in NEJM (October 15, 2026) described a patient who developed a paraspinal neoplasm after receiving FBX-101, a gene therapy that uses an AAVrh.10 capsid vector carrying a CAG promoter. Molecular analysis of the tumor identified integration of the vector’s CAG promoter-splice donor sequence directly into the GLI1 gene — generating a chimeric transcript that drove full-length GLI1 expression. The same oncogene. The same mechanism. A different company, disease, and capsid.

Program 2 — RGX-111 (MPS I/Hurler syndrome): A case report by Ahrens-Nicklas et al. in NEJM (June 4, 2026) described a five-year-old who developed a neuroepithelial tumor four years after receiving RGX-111, REGENXBIO’s AAV9-based intracisternal therapy for Hurler syndrome. Sequencing of the tumor tissue revealed rearranged AAV vector DNA integrated into the PLAG1 proto-oncogene — a developmental transcription factor closely related in function to GLI1, also normally silenced after birth, and also reactivated by promoter insertion into a proliferating cell. That case prompted the FDA to place clinical holds on both RGX-111 and RGX-121 (REGENXBIO’s MPS II program) in January 2026.

The common thread: All three programs — Zolgensma, FBX-101, and RGX-111 — used strong synthetic promoters derived from the CAG sequence (CMV enhancer fused to chicken β-actin promoter and rabbit β-globin splice acceptor). In both the FBX-101 and RGX-111 cases, investigators found that the integrating DNA was not the full therapeutic construct but a truncated, rearranged subgenomic fragment that had lost the therapeutic gene itself — but retained the promoter-enhancer and splice donor sequences from the 5′ end of the construct. Those are precisely the elements capable of activating a downstream oncogene, with no therapeutic benefit remaining.

This convergence carries a specific implication the FDA’s communication states only obliquely but which the three-program pattern makes explicit: the risk criterion for class-wide surveillance may not be “uses an AAV vector” but more specifically “uses a CAG-class synthetic promoter in an AAV vector delivered to dividing or progenitor cell populations.” The FDA has announced it is monitoring all similar AAV vector-based gene therapy programs — a signal that any pipeline therapy sharing these structural features should expect enhanced scrutiny.

Why Infants May Face a Higher Structural Risk

One dimension of these cases that has drawn attention from researchers is the age of the patient populations receiving the highest-risk doses.

Zolgensma’s approved indication is pediatric patients under two years of age, who receive a single high-dose intravenous infusion. Infant livers and central nervous systems contain actively dividing progenitor cells. When an AAV vector DNA molecule integrates into a dividing progenitor — a rare event, but not a theoretical one — that cell passes the integration on to all of its progeny. The result is a large clonal population carrying the insertion, amplifying the probability that one of those cells may acquire additional mutations sufficient to drive tumor formation.

Adult patients receiving AAV therapies at comparable vector doses are exposed to more post-mitotic (non-dividing) cells, which limits this clonal amplification mechanism even if integration occurs. Itvisma’s target population — patients aged two years and older, with less actively dividing CNS progenitor pools than infants — may face a different exposure profile, though the FDA has declined to draw any public distinction between the two therapies’ risk profiles at this stage.

What the FDA Is Doing — and What It Has Not Done

The agency has announced four specific actions:

1. Label revision: The existing prescribing information language describing only “a theoretical risk of tumorigenicity” is under expedited review for revision. No timeline has been given; the FDA said only that it is evaluating the adequacy of the current information.
2. Long-term follow-up scrutiny: Novartis operates long-term safety studies under its pharmacovigilance obligations. The FDA is reviewing whether those studies are adequately designed to detect and characterize late-onset oncogenic events — specifically whether they are capturing the full treated population and whether surveillance intervals are sufficient given the documented multi-year latency to tumor presentation.
3. Enhanced monitoring guidance: Healthcare providers have been instructed to monitor patients receiving either therapy for signs of tumor formation and to report any suspected adverse reactions or new tumor diagnoses. Reports can be made to Novartis Gene Therapies at 1-833-828-3947 or directly to FDA via MedWatch at 1-800-FDA-1088.
4. Broader class surveillance: The FDA has stated it is “closely monitoring the risk of neoplasia, including malignancy, in other similar AAV vector-based gene therapy programs” — language that signals an active, formal review extending beyond these two Novartis products.

What the FDA has not done: No recall, market withdrawal, or clinical hold has been issued for Zolgensma or Itvisma. The FDA has explicitly maintained that the overall benefits of both therapies continue to outweigh their risks for their approved indications. The agency cannot place a “clinical hold” on a commercially approved product — that instrument applies only to ongoing clinical trials. The regulatory options available for commercial products short of market withdrawal include label changes, a Risk Evaluation and Mitigation Strategy (REMS) imposing prescribing restrictions and monitoring requirements, and mandated post-approval studies. The FDA has not yet announced which, if any, of these it intends to pursue beyond the label revision currently underway.

Why the Benefit-Risk Calculation Remains Complex

This safety signal does not arise in a clinical vacuum. SMA Type 1 — the most severe form of the disease, affecting children diagnosed in infancy — carried a mortality rate of approximately 90% before age two without treatment. Zolgensma’s approval in 2019 represented a genuine and documented survival advance: clinical trial data demonstrated durable improvements in motor milestone achievement and survival over more than five years of follow-up.

That context makes the benefit-risk calculation one of the most difficult in current pediatric medicine. Withholding a therapy with a confirmed curative effect from an infant with SMA Type 1 carries near-certain near-term mortality. Administering it now carries a confirmed but still-rare risk of a late-onset tumor with an estimated latency of one to five or more years.

Physicians and families making this decision after November 6, 2026, do so with one important difference from those who made it on November 5: the FDA has formally confirmed that the risk is real and mechanism-specific, not theoretical — and has named the molecular target.

The European Medicines Agency (EMA), which had been reviewing Itvisma under its Joint Clinical Assessment process and was expected to issue a formal opinion in mid-2026, had not announced a formal response to the FDA’s safety communication as of publication.

For Patients and Families: What to Know Now

If your child received Zolgensma: The FDA is not recommending discontinuation of ongoing care, and no changes to routine pharmacovigilance have been issued that change standard clinical management. The agency and advocacy groups are recommending that families contact their treating neurologist to discuss long-term MRI monitoring for spinal, paraspinal, and CNS lesions — the specific tumor locations identified in confirmed cases. The latency window for tumor development based on published cases ranges from under two years to over four years post-infusion.

If your child recently received Itvisma: No tumor cases have been reported in Itvisma patients. The FDA’s extension of scrutiny to Itvisma is prospective and mechanistic. Families should discuss the FDA’s communication with their treating physician and ensure they are enrolled in Novartis’s post-marketing surveillance programs.

If you are currently weighing Zolgensma for a newly diagnosed SMA infant: The FDA has not changed its approval or restricted access. Any decision should be made in consultation with a neuromuscular disease specialist who can walk through the current confirmed tumor data, the absolute risk estimates for the specific SMA type, and the alternative approved therapies — including nusinersen (Spinraza) and risdiplam (Evrysdi), both of which carry different mechanism-based risk profiles.

The FDA’s full safety communication, including references to the published NEJM case reports, is available on CBER’s safety page.

Frequently Asked Questions

What is insertional oncogenesis, and why was it considered a theoretical risk before now?

Gene therapies based on AAV vectors deliver therapeutic DNA into cells. While the DNA mostly persists in an episomal form — floating in the nucleus without integrating into chromosomes — integration events do occasionally occur. When an integrated fragment lands near a gene that regulates cell growth, the strong promoter elements embedded in the therapeutic construct can activate that gene. This mechanism was the central failure in first-generation retroviral gene therapy trials in the early 2000s, most notably the French SCID-X1 trial in which four of nine treated children developed T-cell leukemia after vector DNA integrated near the LMO2 proto-oncogene. AAV vectors were then established as lower-risk alternatives because they integrate far less frequently than retroviruses and lack the same enhancer architecture. The risk was acknowledged in labeling as “theoretical” because no cases had been confirmed at scale in an approved product — until now.

What is GLI1 and why does its activation cause cancer?

GLI1 (Glioma-Associated Oncogene Homolog 1) is a transcription factor in the Sonic hedgehog signaling pathway. It regulates genes that drive cell proliferation during development and is normally active in embryonic tissue before being silenced after birth. Its constitutive overexpression — activation that persists without normal regulatory shutoff — drives uncontrolled cell division. GLI1 amplification and overexpression is documented in glioblastoma multiforme, medulloblastoma, basal cell carcinoma, rhabdomyosarcoma, and several gastrointestinal cancers. When a gene therapy vector integrates a potent synthetic promoter (the CAG element) upstream of or within GLI1, it converts a postnatally silenced gene into one that is constitutively expressed — turning on a proliferative signal in tissue where no such signal is supposed to be active.

What is the CAG promoter and why does it appear in all three affected programs?

The CAG promoter is a synthetic hybrid constructed from the cytomegalovirus (CMV) immediate-early enhancer fused to the chicken β-actin promoter and the rabbit β-globin splice acceptor. It became a standard component of AAV gene therapy vectors because it drives strong, ubiquitous transgene expression across diverse cell types — which is exactly what a gene therapy for a systemic disease needs. Its strength is also its structural risk: when a fragment retaining the CAG sequence integrates near a developmental proto-oncogene, it creates constitutive expression of that gene at precisely the level it was designed to maintain. The FDA’s class-wide surveillance signal implies that the question it is now investigating is whether the CAG promoter carries a specific structural affinity for developmental transcription factor loci, and whether pipeline therapies using the same element require redesign or additional safety studies.

Is there a risk to adults who received Itvisma?

Itvisma was approved for patients aged two years and older, teenagers, and adults — expanding gene replacement to age groups previously ineligible for Zolgensma. No tumor cases have been reported in Itvisma patients. Adult patients in general have a lower proportion of actively dividing progenitor cells than infants, which may theoretically limit the clonal expansion of integration events — but the FDA has drawn no distinction between populations in its safety communication and has explicitly stated that Itvisma is under investigation.

Why isn’t the FDA pulling these drugs from the market?

FDA market withdrawal is reserved for cases in which the harm conclusively outweighs the benefit for the approved population. For SMA Type 1, withholding Zolgensma means near-certain mortality before age two. The FDA’s current assessment is that the confirmed tumor risk — real but rare, with nine reports across more than 3,700 commercially treated patients — does not reach the threshold that would justify removing the only approved curative gene replacement option for the disease’s most severe form. That benefit-risk balance may change if additional cases are identified, if the long-term safety studies prove inadequate to characterize the true incidence, or if the label revision triggers clinical behavior changes significant enough to alter the treated population.

What was different about how these cases came to light?

The initial disclosure on November 5 came from Novartis itself, reportedly following parent-driven social media posts about tumor diagnoses in Zolgensma-treated children. The FDA’s formal safety communication on November 6 — the subject of this article — represents the regulator’s own independent characterization of the evidence. That sequence is itself under scrutiny: the drug’s label still described the tumor risk as “theoretical” through November 5, 2026 — more than seven years after approval and with nine tumor reports in Novartis’s own pharmacovigilance system. The FDA has not announced whether it is reviewing the adequacy of post-marketing reporting timelines in this case.