The weaker drug passed. The stronger one failed. Same FDA week.

Biotech & Longevity 4 min 18 sources

In one week, an FDA advisory panel endorsed Replimune's twice-rejected melanoma drug, which was tested without a control group, and voted down Capricor's Duchenne heart therapy, which ran a full randomised trial. Plus GSK's cost cuts, a J&J cell-therapy bet, another Novo Nordisk stumble, and an Ebola vaccine only a charity would fund.

Key takeaways

An FDA panel backed Replimune's melanoma drug, tested without a control group, then rejected Capricor's Duchenne therapy, which ran a full randomised trial - the weaker evidence advanced and the stronger evidence stalled.
Big pharma kept reshaping itself: GSK cut costs to fund late-stage drugs and AI, J&J bet 2.6 billion dollars on in-body cell therapy, and Novo Nordisk stumbled again trying to grow past obesity.
An Ebola vaccine and a syphilis drug shortage show the same gap: the treatments the world most needs are often the ones no market will pay to make.

Two panels, opposite calls

In the same week, the same kind of FDA advisory panel looked at two drugs and went opposite ways - and, oddly, backed the one with the weaker evidence.

On Thursday, outside experts convened by the FDA voted 10 to 3 that Replimune has enough data for the agency to review RP1, an engineered virus for advanced melanoma [1]. The vote is a turnaround. The FDA had declined to approve RP1 twice, most recently in April, because the company leaned on a single-arm study - one with no comparison group. Without a comparison, there is no way to know what would have happened without the drug [2]. An advisory committee (an “adcomm”) is a panel of outside doctors and scientists that recommends; the FDA makes the final call. Replimune wants accelerated approval for RP1 alongside Bristol Myers Squibb’s Opdivo, in melanoma patients whose tumours grew despite earlier treatment [3]. Its case rests partly on a survival analysis suggesting patients who responded to RP1 lived longer than those who did not. That is a comparison of responders to non-responders, which the FDA’s own scientists warned can flatter a drug [4].

The day before, a different panel voted 9 to 3 against Capricor Therapeutics. It found the company had not shown “substantial evidence” that its cell therapy, deramiocel, works for the heart damage of Duchenne muscular dystrophy [5]. Duchenne is an inherited disease that wastes muscle; the heart failure it causes is the leading cause of death for these patients [7]. And here is the twist: Capricor ran the stronger study - HOPE-3, a phase 3 trial that was randomised, double-blind and placebo-controlled, the gold standard [6]. Reviewers argued the drug missed its original goals, and that the benefit the company pointed to appeared only after it changed its statistical plan [7].

So the drug with the weaker trial advanced, and the drug with the stronger trial stalled. The difference was not just the data. It was how each panel weighed the two ways it could be wrong.

The pharma reshuffle

GSK said it will cut costs by about 1.9 billion pounds, shedding jobs to pay for faster, late-stage drug development and a new research centre in Cambridge [8]. In the same stretch it signed a deal worth up to 110 million dollars with UK biotech Relation Therapeutics [9]. Relation will build datasets and train AI models that hunt for new drug targets.

Johnson & Johnson took an option to buy Sail for nearly 2.6 billion dollars [10]. The startup works on “in vivo” CAR-T - cell therapy made inside the body rather than in a lab - and J&J is betting the approach can treat autoimmune disease. AstraZeneca beat profit forecasts and held to its goal of 80 billion dollars in annual revenue by 2030 [11].

Novo Nordisk’s rough patch

The maker of Ozempic and Wegovy keeps struggling to grow beyond weight loss. Its experimental drug ziltivekimab - an antibody aimed at inflammation - failed to cut major cardiovascular events in a late-stage trial of patients with heart and kidney disease [12]. It was the second recent setback in its push past obesity [13].

Others are finding new uses for the same drug class. Altimmune said its next-generation GLP-1 - the family behind Ozempic - sharply curbed heavy drinking in a mid-stage trial for alcohol use disorder, a result it called unequivocal [14]. Early data, one trial - but a sign these drugs may reach well past weight.

The gene-editing frontier, and its cost

Researchers reported using CRISPR base editing - a precise way to rewrite single DNA letters - inside the body to blunt the protein that drives Huntington’s disease [15]. Huntington’s is a fatal inherited brain disorder with no cure. It is early laboratory work, not a treatment.

The frontier has a darker side too. A Chinese university opened an investigation into the death of a young woman after an experimental gene-editing procedure [16]. It is a reminder that moving fast on the human body carries a real cost when the guardrails slip.

The drug no market wants to make

The week’s quietest story is about a gap the market leaves open. Merck and the charity Wellcome are gearing up to manufacture doses of an experimental vaccine against the Bundibugyo strain of Ebola behind a fast-spreading outbreak [17]. The effort is backed by up to 8.5 million dollars from the epidemic-preparedness group CEPI. And in the United States, syphilis is spreading while researchers face federal constraints and a shortage of the cheap, decades-old drug that treats it [18]. Neither an Ebola vaccine nor a penicillin refill promises much profit - which is exactly why they need someone other than the market to step in.

02 · Lesson · why it matters

The two ways to be wrong, and who pays for each one

Every gate can fail two ways - wave through something bad, or block something good - and no one can shut off both at once.

Two panels, one week, opposite calls

In one week, the same kind of FDA advisory panel looked at two drugs and split in opposite directions. It endorsed Replimune’s melanoma drug, which had been rejected twice and never tested against a comparison group. The day before, it rejected Capricor’s therapy for Duchenne muscular dystrophy - even though that one ran the stronger study, a full randomised, placebo-controlled trial.

Read that again. The drug with the weaker evidence advanced. The drug with the stronger evidence stalled. If a panel just graded the science, that ordering makes no sense. Something other than the raw quality of the data was doing the deciding.

A gate can be wrong two ways

Every gatekeeper faces the same shape of problem, and it has nothing to do with medicine. There are two ways to be wrong, not one.

You can let through something that should have been stopped - approve a drug that does not work. Call that a false yes. Or you can block something that should have passed - reject a drug that would have helped. Call that a false no.

Here is the hard part. You cannot drive both to zero at the same time. Tighten the gate to catch every dud, and you will also turn away some real cures. Loosen it so no good drug is ever denied, and some worthless ones slip through with them. A smoke alarm set to catch every fire will shriek at burnt toast; set it to ignore the toast and it will sleep through some fires. There is no setting that does neither.

The panel was choosing, not just measuring

Once you see the two errors, the week stops being a contradiction and becomes a choice made twice.

For advanced melanoma patients whose tumours had already grown through other treatments, the panel leaned toward the false yes. Better to risk approving a drug that might not work than to deny a dying person their last option. The weak evidence did not suddenly get strong. The panel decided which mistake it could live with.

For Capricor’s drug, it went the other way. Reviewers saw a trial that missed its original goals, with the benefit appearing only after the company changed how it counted. That raised the odds this was a false yes in disguise - and the panel refused to make it. Same committee, same week, opposite tolerance for the same kind of error.

The bar is a buried decision

Now the part that hides in plain sight. The rules that decide these cases were themselves a choice about which error to fear more.

The demand for a control group - the reason Replimune was turned down twice before - is not a neutral fact of science. It is a rule that says: we would rather wrongly block a good drug than wrongly wave through a bad one. That default was built by people. It then poses as simply how rigor works. And it serves some and costs others at the same time. It protects future patients from paying for duds. It also makes desperate patients wait, or go without, while the proof is gathered.

Both of those are true at once. A strict gate is not the villain and a loose gate is not the hero. Each just moves the harm to a different doorstep.

The two errors land on different people

That is the piece worth carrying. A false yes and a false no do not fall on the same person.

A wrongly approved drug spreads a thin harm across many: future patients who take something useless, an insurance system that pays for it, the slow erosion of trust when the approval unravels. A wrongly rejected drug lands as a heavy harm on a few: the specific people who needed it now, many of whom will not be alive when a better-run trial finally settles the question.

And the people bearing one error are almost never the ones bearing the other - nor the ones setting the bar. The dying patient who wants the loose gate is not in the room where the strict rule is written. The future patient protected by that rule has no idea they were ever at risk. You are somewhere in this web too. On a bad day you are the person begging the gate to open; on another you are the one it quietly protects, and never thanks.

What the seat can’t see

None of this is unique to the FDA. A loan officer, a border agent, a hiring panel, a jury, a spam filter - each stands at a gate that can fail two ways. And each has quietly decided whose mistake is more acceptable. The setting always looks like plain good judgment from the inside.

Seeing the trade-off does not tell you where the bar belongs. It cannot; that depends on which error you would rather suffer, and reasonable people weigh the two harms differently. What it does is make “obviously they got it wrong” harder to say with a straight face. The person on the other side of the gate is trading one error against another, and paying for it - often in a currency you cannot see from your seat.

03 · Lab · your turn

You Are the Gate

Set an approval bar and feel the two ways a gatekeeper can be wrong trade off, with the least-harm setting shifting as the stakes of each error change.

04 · Hope · carry this

A panel that can be pulled toward mercy for the dying and toward caution for the vulnerable is not broken - it is people trying to weigh two real harms honestly. That we argue so openly about which mistake we can bear is itself a kind of progress.