A surrogate is only as good as its assay: write the measurement case first

Two filings, one dependency

Two filings this week show the same dependency. Larimar Therapeutics has completed its rolling BLA for nomlabofusp in Friedreich's ataxia and is seeking FDA accelerated approval on skin frataxin as a possible novel surrogate endpoint. The company expects acceptance of the filing in early 2027, and the Phase 3 FORWARD-FA confirmatory trial is still enrolling. Spruce Biosciences, meanwhile, says it still intends to seek accelerated approval of tralesinidase alfa for MPS IIIB in December. The records do not say what evidence that filing will rest on, so I will not guess.

On the same day, EMA released for consultation a draft guideline on predictive biomarker assays across the medicinal product lifecycle, setting out expectations for assays used to select patients for treatment.

The position

Regulatory affairs teams should stop treating the biomarker measurement as a development detail that sits beside the regulatory strategy. When approval depends on a surrogate, or on an assay that decides who is treated, the measurement is part of the claim. A sponsor that cannot show how the number was generated, why it can be trusted and how it will stay trustworthy after approval has an exposed dossier, however good the biology.

I want to be precise about the limits of the evidence. EMA's draft concerns assays that select patients. Larimar's skin frataxin is proposed as a surrogate endpoint, which is a different job. The records do not show EMA addressing surrogate endpoints, and they do not show FDA's view of skin frataxin. The link I am drawing is my inference: regulators on both sides of the Atlantic are scrutinising the measurement itself, and a lifecycle expectation for one kind of assay is a reasonable preview of the questions others will face.

The word 'lifecycle' in the EMA title matters. An assay validated for a pivotal study can still change afterwards, through a new site, a new platform or a new sample type, and each change is a comparability question. Accelerated pathways sharpen the point, because the confirmatory trial then runs in parallel with the marketed product. If the surrogate is measured one way in the filing and another way in confirmation, the link between them gets harder to defend.

The strongest counter-argument

The best objection is about sequence and cost. The EMA text is a draft, not a requirement. FDA accelerated approval is decided case by case, and no one yet knows whether the agency will accept skin frataxin. Building a lifecycle-grade assay file for a surrogate that may be rejected looks like spending ahead of the decision. A confirmatory trial that is still enrolling is also, in effect, the safety net: if the surrogate fails to predict benefit, the trial says so.

That argument is fair as far as it goes, but it undervalues what the measurement file does before any decision. It is the evidence a regulator will use to judge whether the surrogate is credible at all. It is also what lets a sponsor change a site, vendor or platform without reopening the question. A confirmatory trial tells you whether the bet paid off. It does not repair a surrogate whose measurement history cannot be reconstructed. The cost of writing the case early is modest next to the cost of explaining gaps later.

What to do

Read the EMA draft now, even if your product is not a selection-driven therapy, and decide whether to comment during consultation; check the closing date on the EMA page, as it is not in the records. Assign one owner for biomarker and assay evidence across clinical, CMC and regulatory functions. Make comparability for any assay change a planned activity rather than a reaction.

— The VigiEye regulatory desk