On 29 July an FDA advisory committee voted nine to three that Capricor had not shown substantial evidence that deramiocel is effective for cardiomyopathy in patients with Duchenne muscular dystrophy. The next day the same advisory committee voted ten to three that the efficacy results from Replimune's IGNYTE study are evaluable and clinically meaningful. Both companies had arrived behind an FDA briefing document that publicly took their efficacy case apart, and both had watched their share price fall when it was posted. Endpoints' Max Bayer summarised the week as "two days, two similar briefing documents and two companies with diving share prices."
The interesting part is not that the agency was sceptical of both. It is that the same starting position produced opposite results, twenty-four hours apart, in front of the same committee.
Start with what the briefing document now is. It is a public analysis, written by the reviewers, of why a sponsor's evidence may not support what the sponsor says it supports, released days before the sponsor gets to respond. For Capricor, FDA said the therapy lacked efficacy and the stock fell. For Replimune, FDA staff said the pivotal trial had failed to prove RP1 works and the stock fell. Acting CBER director Karim Mikhail opened the Capricor meeting by saying he stood by those documents and the agency's analysis, which removed any doubt about whether a briefing document is a working draft or a position. It is worth noting who said it: an acting director choosing to put his weight behind the review staff's written analysis rather than leave room between himself and it.
What the Capricor meeting actually turned on is worth sitting with, because it is not a disagreement about biology. FDA reviewers said that after the randomised, double-blind portion of HOPE-3 had completed, and during the open-label extension, changes were made to the pre-specified statistical analysis plan, generating at least two further versions. On the agency's reading the trial failed and the efficacy claims are an artefact of those revisions. Capricor disputes that interpretation. Set the merits aside for a moment and look at the shape of the argument: a regulatory decision in a fatal childhood disease came down to which version of a statistical analysis plan counts, and when each version came into being. That is a documentation question wearing a statistics costume, and anyone who has watched an analysis change late knows the recomputation is the easy part. The hard part is establishing which version was the plan, and proving it from the record rather than from memory. It is the argument for treating the record rather than the file as the thing you trust, made by a company that is now having to win it in public.
Which is why the sponsor's briefing book stopped being a summary of the programme some time ago. It is a rebuttal, filed in public, to an argument the reviewer has already made and published. That is a different document with a different job, and the way most organisations staff it has not caught up.
Consider what it has to do. It has to answer a specific written analysis rather than present a general case. It has to do that without conceding the analytical frame, because a rebuttal that accepts the agency's framing and then argues about the numbers inside it has already lost. It has to be legible to advisers who are not the review division, who may not know the programme, and who will read it alongside patient testimony. And it has to survive being read by investors, competitors and journalists on the day it lands. A clinical study report is longer and harder to get right, but it is read by people paid to read it carefully. The briefing book is read once, quickly, by people deciding something.
I do not know why the two votes diverged, and neither does anyone working from the tallies. Fierce Biotech described the Capricor meeting as marked by a statistical dispute; BioSpace called it chaotic. Advisers also spent time on upper limb function, where several thought Capricor's case was stronger though still unproven, which suggests the discussion moved onto ground the sponsor had not built the vote around. The transcripts will say more about that than the numbers do.
What I am confident about is narrower. The agency's document sets the terms. It does not settle them.
That is worth saying plainly, because we said something close to the opposite here in June. Writing about Replimune's third BLA after two complete response letters, the framing was that the agency stands by its rejections and the writing task is to work within a position FDA has taken and maintained. Six weeks later advisers looked at the agency's maintained position on that exact product and went the other way. The June piece was not wrong about how FDA behaves across review cycles. It was too quick to treat the agency's stated position as the fixed point in the room, and July is a fair correction.
There is a second thing in the Capricor vote that has gone unremarked. The question the panel answered was whether there was substantial evidence of effectiveness, which is the statutory standard rather than a casual phrase. On 24 June FDA reissued its draft guidance on demonstrating substantial evidence of effectiveness for human drug and biological products, in the agency's words to focus on generating rigorous scientific evidence in the most efficient manner, and it clarifies that one adequate and well-controlled investigation together with confirmatory evidence may satisfy the standard. Comments close on 22 September. So a panel applied a statutory standard while the agency is revising its guidance on how that standard can be shown, and the revision points toward accepting less rather than more. Sponsors who have seen an appeal turn on those same words will recognise how much weight the phrasing carries. It is the same problem described last week in a different register.
The practical consequence is a resourcing question rather than a matter of craft. If the briefing book is where a programme's argument is made in public against a published counter-argument, it should be drafted early by the people who understand the evidence best, and rehearsed against the strongest version of the agency's case rather than the version the team hopes to get. In most companies it is assembled late, by whoever has capacity, out of material built for another audience. That was defensible when the briefing document was a courtesy summary. It stopped being defensible once the agency began publishing its own analysis first and the market started pricing it the same afternoon.
Neither vote binds FDA, and both programmes had agency decisions still ahead of them. What has changed is what the meeting is. The sponsor now walks into an argument between two documents, and did not write the one that goes first.