The Food and Drug Administration said on Nov. 6 that it is investigating a serious risk of malignancy linked to Zolgensma, Novartis's one-time gene therapy for spinal muscular atrophy, after the company reported nine tumors in patients treated since the therapy's approval in 2019. Two of the cases carry the strongest evidence: vector DNA had inserted near or inside GLI1, a known cancer gene, and the agency described them as confirmed product-related insertional oncogenesis.
The numbers are small. More than 6,000 patients have received Zolgensma worldwide, according to Novartis, and nine tumors in a population that size is a rate low enough that the FDA still says the benefits of Zolgensma and its intrathecal sibling, Itvisma, "continue to outweigh their potential risks" for approved uses.
The Zolgensma episode is not only a test of one drug. It is a test of the machinery that is supposed to catch rare, delayed harms from a class of therapies whose central promise is that a single treatment can replace years of chronic care. That machinery depends on long follow-up, clear labels, and companies that disclose what they know when they know it. The first two are now under review. The third was tested by the way these cases became public.
What the nine reports show
Novartis disclosed the cases on Nov. 5, after SMA Europe and other patient groups pressed for information. Two of the nine had been reported earlier, in 2023 and 2025. The other seven were previously unknown, STAT reported. The FDA's notice says the tumors occurred in the kidney, the spinal and paraspinal region, and the central and peripheral nervous systems.
Three of the nine tumors have been tested, according to Cure SMA, and one appears unrelated to Zolgensma, BioPharma Dive reported. The FDA says two cases showed vector DNA integrated into the patient's genome, with the construct's promoter inserted upstream of or within GLI1. In a third, the mechanism of GLI1 overexpression remains under investigation.
GLI1 is not a random address. It is a gene in the hedgehog signaling pathway, and the FDA's notice links the finding to two other AAV programs. In one, a 3-year-old boy with Krabbe disease developed a paraspinal tumor after treatment with FBX-101, and testing found a GLI1 insertion. In the other, a 5-year-old boy with a lysosomal storage disorder developed a neuroepithelial tumor after treatment with RGX-111, and testing found a PLAG1 insertion. Both were reported in the New England Journal of Medicine this year, and the FDA says it is assessing tumor risk across other AAV gene-therapy programs as well.
Zolgensma's label has carried a warning since approval: there is a "theoretical risk of tumorigenicity" from integration of AAV vector DNA into the genome. The label also says the clinical relevance of individual integration events is unknown. The word theoretical was accurate when the label was written. AAV vectors are built to persist mostly as episomes, separate from the cell's chromosomes. Integration was known to happen at low frequency, but it was not the mechanism the field expected to cause cancer.
How a vector becomes an oncogene
A gene therapy vector is a delivery vehicle. Zolgensma uses a non-replicating AAV9 capsid to carry a working copy of SMN1 into motor neurons. The therapeutic payload is the gene. The part that matters in the tumor cases is the vector's promoter, a stretch of DNA that switches the gene on. If the promoter lands near a growth-control gene, it can drive that gene instead.
The FBX-101 case, reported in the New England Journal of Medicine in October, shows how specific the mechanism can be. The boy received an AAVrh.10 vector carrying GALC. Twenty-six months later, imaging found a paraspinal mass. Sequencing showed a clonal insertion in GLI1 intron 1. The inserted fragment was not the full therapeutic vector. It was a subgenomic piece that included a promoter and a splice donor but lacked the GALC transgene and its polyadenylation signal. The result was a chimeric transcript that drove full-length GLI1.
That detail is uncomfortable for the field because it suggests the cancer risk may depend less on the therapeutic gene than on the design of the vector and the manufacturing process that produced it. The FBX-101 authors flagged concern about a promoter-only species arising during manufacturing. The FDA's notice does not go that far, but it does say it is evaluating the risk across other AAV programs, not only Zolgensma.
AAV gene therapy has been sold on a simple idea: the vector delivers a gene and then gets out of the way. The FBX-101 and RGX-111 cases show that the vector can leave a piece of itself behind in a place that matters. The risk is still rare, but it is not zero, and it is no longer purely theoretical.
The disclosure problem
The scientific finding is the easier half of this story. The harder half is how it reached the public.
Seven of the nine cases were not publicly known before this week. They surfaced through patient advocates and social media, according to STAT and BioPharma Dive, after Novartis told neurologists earlier in the week. SMA Europe said it learned of the cases through confidential channels and told Novartis that the community needed timely public communication. Novartis says it notified health authorities after each case and that patient safety is its priority.
Patient groups are not asking for a different risk calculation. They are asking for enough information to weigh the risk with their care teams. SMA Europe said families should not have to piece together a potential safety concern from "social media, rumours or incomplete reports." The UK's national multidisciplinary team reviewed the three tumors that have been tested, found two related to Zolgensma and one unrelated, and recommended no change to routine care. Muscular Dystrophy UK repeated that guidance.
Both things can be true. The clinical guidance can be stable while the information environment is not. A family that has already made the decision to treat, or is making it now, needs to know that the risk label has changed in substance even if the recommended care has not.
What the FDA is weighing
The FDA's notice is short and careful. The agency says it is evaluating the adequacy of the current prescribing information and the long-term follow-up studies, and whether regulatory action is warranted. It is also looking at the risk across other AAV programs. The label already carries a boxed warning for acute liver failure, updated after two deaths in 2022, so the precedent for revising it exists. The question is whether the label can keep calling the risk theoretical now that the mechanism has been confirmed in patients.
Itvisma matters to that question. The intrathecal version was approved in November 2025 for patients age 2 and older. It uses the same vector and construct, and the FDA says the tumor risk determination applies to it as well, even though no neoplasia cases have been reported after Itvisma to date. That extends the surveillance question from infants to older children and adults, a population that will live with any delayed risk for decades.
The regulatory options are not mysterious. The FDA can revise the label, require longer or more structured follow-up, convene an advisory committee, or take no action beyond monitoring. It can also ask for a class-wide review. What it cannot do is make the underlying biology go away. The agency's own notice cites two other AAV programs with integration-associated tumors, which means the Zolgensma cases are part of a pattern that extends beyond one company.
What families and clinicians should do
For families, the FDA's practical advice is narrow. Patients treated with Zolgensma or Itvisma should be monitored for tumor formation, and clinicians should report a tumor or another suspected reaction to Novartis or the FDA's MedWatch program. The agency did not tell patients to stop treatment. Cure SMA and Muscular Dystrophy UK advised families to contact their SMA care team rather than change care plans on the basis of the reports alone.
The calculus is not abstract. Zolgensma is approved for infants with SMA under 2, a population in which untreated disease can progress quickly. The therapy is given once, and the alternatives, nusinersen and risdiplam, are chronic treatments with their own burdens. A rare delayed cancer risk does not cancel the benefit of a treatment that can preserve motor function. It does belong in the conversation, and it is now a confirmed risk rather than a hypothetical one.
For clinicians, the harder task is the one the label has always asked of them: keep looking. A tumor that appears years after an infusion may not be connected in the chart. The FDA's notice is a reminder to ask, and to report. Postmarketing surveillance is not a passive database. It is a system that depends on someone noticing an unusual case and filing it.
The next test
The Zolgensma story will be told in two ways. One is a safety story about a rare side effect of a powerful therapy. The other is a story about whether gene therapy can govern itself as it moves from a handful of approved products to a broad platform.
The field's pitch is that one treatment can replace a lifetime of care. That pitch carries an obligation that ordinary drug regulation does not: the follow-up has to last long enough to catch harms that appear after the trial ends, and the information has to reach the families who are living with the treatment. The FDA says it is evaluating both the label and the follow-up studies. The patient groups are asking for the same thing.
The nine reports are rare. That is the reason to build a surveillance system sensitive enough to find them, not a reason to wave them away. The word theoretical did a lot of work in that label. It is now doing less.
Primary sources
- U.S. Food and Drug Administration for the Nov. 6, 2026 safety communication on Zolgensma and Itvisma, including the nine reports, the GLI1 findings, the benefit-risk assessment, and the follow-up review.
- New England Journal of Medicine for the October 2026 correspondence on the FBX-101 paraspinal neoplasm and the May 2026 report on the RGX-111 neuroepithelial tumor.
- BioPharma Dive for Novartis's Nov. 5 disclosure, the case counts, the Cure SMA assessment, and the 6,000-patient exposure figure.
- STAT for the reporting on how the seven previously unknown cases surfaced and the earlier 2023 and 2025 case reports.
- Muscular Dystrophy UK for the UK national multidisciplinary team's review of the three tested tumors and its guidance to families.
- Novartis for the Zolgensma prescribing information, section 5.6 on AAV vector integration and tumorigenicity risk.
- U.S. Food and Drug Administration for the November 2025 approval of Itvisma and its indication for patients age 2 and older.