Insertional oncogenesis is no longer theory: AAV sponsors should reopen their malignancy files

The word that changes the file

FDA's Center for Biologics Evaluation and Research (CBER) has opened an investigation into tumour reports with Zolgensma. Its wording matters. CBER cites confirmed AAV vector insertional oncogenesis involving GLI1. It says it is assessing whether regulatory action is needed after postmarketing malignancy reports. It also says the neoplasia risk applies to both onasemnogene abeparvovec products, Zolgensma and Itvisma.

For years, integration of AAV vectors into host DNA was a footnote in gene therapy risk discussions: a low-probability, long-horizon concern, usually handled with a line in a risk management plan and a long-term follow-up commitment. A regulator using the word 'confirmed' moves the issue from the theoretical section of the file into the signal section. Any QPPV with an AAV product on the market or in late development should treat that shift as relevant to their own portfolio, whatever their vector or indication.

What we do not know, and why that argues for acting

The records do not say how many malignancy reports FDA has received, in which patients, after what latency, or what the tumour types were. They do not say what regulatory action, if any, will follow. CBER has said only that it is assessing. I will not speculate on those points, and neither should anyone building a business case from this announcement.

But the gaps are the argument. Malignancy is a late-appearing, low-frequency event. Spontaneous reporting rarely captures it well, and a causal mechanism is rarely proven from reports alone. Something confirmed at molecular level, in a product used postmarketing, tells us that existing surveillance was able to find the link. Whether other sponsors' systems could do the same is a question they should answer before a regulator asks it.

The strongest counter-argument

The best objection is about scope. The records describe one active substance, in two products, and a specific gene, GLI1. Nothing in them says other AAV vectors, capsids, promoters or doses carry the same risk. Class-wide alarm on one product's findings could push sponsors into reflexive label changes, unnecessary patient anxiety and consumption of scarce PV capacity. Regulators generally want evidence before extending a risk to a class, and the Zolgensma investigation is still open.

That is fair, and I do not argue for class labelling. I argue for class-level questions. Asking whether our data could show integration-linked tumours costs a review. Not asking costs us the chance to discover the answer on our own terms. A sponsor that has already examined its malignancy cases, and can say what it did and found, is in a much better position than one whose first analysis is a response to a regulator's request.

What a serious response looks like

First, a retrospective look at all malignancy and neoplasm reports for the product, including those coded as unrelated or coincidental, with attention to time to onset and tumour site. Second, a check of whether follow-up procedures ask for tumour tissue or molecular analysis when a malignancy is reported. A case that is closed as 'unlikely related' without tissue characterisation cannot test an integration hypothesis. Third, a look at whether long-term follow-up registries and PSUR or periodic reporting sections are designed to catch late events, rather than only meeting a commitment's letter.

Finally, set triggers in advance. Decide what pattern of malignancy reports would prompt escalation to the regulator or a change to the risk management plan, and document the reasoning. The point is to act on a defensible, pre-agreed threshold rather than improvise under pressure.

Recommendation

Do not wait for CBER's conclusion. Its outcome will tell us about onasemnogene abeparvovec. The way to learn whether it tells us anything about your product is to run your own review while the picture is still open. Put it on the next safety governance agenda, assign an owner, and keep the paper trail.