Bristol's CAR-T Pause Was Three Months Old Before the Public Heard

Bristol Myers Squibb told regulators in early June. The rest of us found out this week, after an analyst note. Three months sat between those two facts, and that gap is the story of the week.

Also this week, a cardiovascular outcomes trial in more than eight thousand patients missed its endpoint and put a whole drug class in question. A mid-stage lupus readout took fifty five percent off a company in a day. And the FDA approved a cancer drug over its own advisers, while memos surfaced on a second approval it had granted over its own reviewers. What do you do with a week like that?

Welcome to The Pharma Closeout Week in Review for Sunday, September sixth. I'm Alex Mercer.

And I'm Maya Patel. It was a big week. Let's take it apart.

Novartis and Bristol Myers Squibb both paused their autoimmune CAR-T trials. Looking at the week, that is the one story that changed shape between Monday and Friday. Monday it was a safety story about three deaths. Friday it was a disclosure story about three months. If you heard none of it, here is the whole thing. Novartis halted its trials outside cancer, Bristol paused enrollment, and the people who needed to know found out on very different schedules.

Start with Novartis. Rap-cel is on hold across eight autoimmune and neuroscience indications, lupus, myasthenia gravis and multiple sclerosis among them, after three cases of immune effector cell-associated hemophagocytic syndrome. That is a rare and potentially life-threatening immune reaction. The holds were company-initiated on August twenty fourth. The cancer trials of the same therapy are still running, and Novartis says it is engaged with regulators.

Bristol didn't use that language.

No. Bristol paused enrollment in the autoimmune trials of zola-cel, its CD19-targeted CAR-T, and called the events transient and reversible. Out of an abundance of caution. It had already reported one case of the same syndrome in a Phase 1 study back in February. Trade press counted three deaths in CAR-T trials.

Has anyone put a mechanism on it?

Sami Corwin at William Blair has. The suggestion is that the rapid-manufacturing technology could be driving the increased cell expansion, and could be driving the reported toxicities with it. It came with a caveat. Other factors could be involved.

If Corwin is right, that is the uncomfortable part, because rapid manufacturing was supposed to be the thing that made autoimmune CAR-T scalable. Then Friday added what nobody had on Monday. STAT, citing the Wall Street Journal, puts Bristol's notifications to regulators, researchers and patient advocacy groups in early June. The broader confirmation followed reporting the week of August thirty first.

Three months. And told whom?

Bristol told regulators, researchers and advocacy groups, which is not nothing. The question isn't whether an agency was left in the dark. It's what an investigator running a lupus trial in July knew when he consented the next patient. Two things I'm watching, both on Bristol. Whether these company-initiated holds get formalized by regulators, and whether it starts the systemic sclerosis study it had planned for this month.

That is where I stop, Alex. In oncology, a patient accepts a life-threatening immune reaction as the price of a response, because the alternative is the disease. Now put a thirty year old with lupus in that chair, with decades of alternatives ahead of her. Nobody has said Bristol broke a rule here. But what did her consent form say in July, and who was supposed to tell her?

What happens to everybody else's program?

Five companies dropped and then recovered. Kyverna, Cabaletta, Allogene, CRISPR and Fate. BioPharma Dive walked the whole field on Wednesday, and the analysts split on exactly the right question. Roger Song at Jefferies points at manufacturing speed, and notes that autoimmune diseases have more activated immunity at baseline.

And the counter-case?

Phil Nadeau at TD Cowen makes it with a number. Cabaletta runs a nine-day process specifically to hold immune toxicity down, so he expects limited read-through to it. Corwin's own follow-up is not about manufacturing at all. It says greater physician awareness of the symptoms and the treatments will improve outcomes.

Which is three different bets on one signal, and they don't sit together. If the culprit is speed, Cabaletta's nine days is a moat and Autolus's traditional process is a moat. If the culprit is the patient's own immune system, nobody has a moat and the whole autoimmune thesis reprices. Kyverna is carrying that live, because its lead therapy is heading toward an approval decision while this is unresolved.

Kyverna's decision doesn't wait for the answer.

On the deal side, five transactions landed this week. Lilly agreed to buy Merida Biosciences for just under two point nine billion dollars if every milestone pays out, going after the autoantibodies behind Graves' disease. Roche licensed a preclinical trispecific from Simcere Zaiming for seventy five million dollars up front. And GSK paid HUTCHMED a hundred and ten million for a molecule that has never been dosed outside China.

Four of the five originated in Asia.

Two more. Novartis took options on Alteogen's subcutaneous platform out of Daejeon, worth up to three point two two three billion. And Gan and Lee licensed a GLP-1 to Menarini for sixty two million euros up front. HUTCHMED jumped sixteen percent in London on that upfront. GSK barely moved. Let's bring in Marcus Webb on this one.

And Lilly's own year?

How many have read out?

Not many, and that is exactly why the option structure matters, Maya. Novartis did not buy Alteogen's platform. It bought options on it, across multiple products, and it pays only if it exercises. Alteogen collects that three point two two three billion only if every option is exercised and every milestone is hit. That is a ceiling, not a cheque.

That still isn't proof, though.

Then look at the buyers who did commit cash. More than a hundred licensing deals with Chinese biotechs since the start of last year, the quarterly count peaking in the first three months of this year, and AstraZeneca, AbbVie and Roche each writing upfronts of at least five hundred and fifty million dollars. Those are not option premiums. That is committed cash.

The trispecific Roche bought is preclinical. Simcere says this is its sixth out-licensing deal, more than six point one billion in aggregate potential across Roche, AbbVie, Ipsen and Boehringer, and Alteogen just signed its fourth of the year. What does the fourth buyer own that the first three don't? You're not buying an advantage. You're buying access on the same terms as the people bidding against you.

I'm waiting for the first dose.

Then watch the one that has never been dosed outside China. HUTCHMED runs that global Phase 1 itself, starting in the back half of this year, and nobody has shown which tumours carry both of its targets.

On the regulatory and clinical side, put the week in two columns. Two Phase 3 trials failed and so did a mid-stage study. Novartis's pelacarsen missed its primary endpoint in more than eight thousand patients with established cardiovascular disease, and it missed while lowering lipoprotein A. Alumis missed every primary and secondary objective in a mid-stage lupus study and lost fifty five percent of its value in a day. It trades around ten dollars now, after starting this year on psoriasis data.

And the fallback?

A prespecified high interferon subgroup that Alumis itself says was under-represented in the trial. How do you rescue a program on a subgroup the trial under-represented? Ultragenyx's Phase 3 of apazunersen in Angelman syndrome missed on cognition and on its key secondary. Its share price was cut in half, and the same release announced a program review and what the company called significant expense reductions, which is a company telling you it is cutting spending after a failed Phase 3.

Two wins, and both won on safety.

Remibrutinib first. Novartis's drug beat Sanofi's Aubagio on relapse rate across nearly two thousand patients in relapsing MS, with zero cases meeting Hy's Law. Hold that number against the class. The FDA cited six Hy's Law cases when it rejected Sanofi's tolebrutinib, and Roche's fenebrutinib has reported one. Zero, six, one. Those three numbers are liver signals.

And the second win?

AbbVie's etentamig cleared both primary endpoints in myeloma. Seventy four percent response against forty five point seven percent for investigator's choice, a hazard ratio of point four zero, in three hundred ninety three triple-class exposed patients with a median of three prior lines. AbbVie built the announcement around cytokine release syndrome in twenty eight percent, almost all of it mild, none of it severe.

Then the agency column. On Friday the FDA granted accelerated approval to AstraZeneca's Etcamah in advanced hormone receptor positive, HER2 negative breast cancer, a median of sixteen months of progression free survival against nine point two. The same advisers had voted six to three against that application in April. The label carries a boxed warning on heart rhythm risk with certain other medicines, and the FDA authorized Guardant360 as the companion test alongside it.

And the confirmatory obligation is still open. The agency wants proof that switching before confirmed progression actually delivers clinical benefit, and that is the same burden pelacarsen failed to carry in a different organ. Lowering a marker is not the same as preventing the event, and approving earlier is not the same as helping sooner. Pelacarsen answered the first half of that on Friday. Etcamah has not answered the second.

Two more from the agency, and the first one is strange. uniQure filed its Huntington's gene therapy on Wednesday and asked for priority review. The filing rests on an external control, and in February the commissioner at the time said publicly that the same data showed no benefit. The agency reversed on the same dataset. On Thursday, Ionis's Zanvastro became the first treatment ever approved for Alexander disease, more than two weeks ahead of the September twenty second date Tuesday's calendar carried.

Ionis owns both ends of that.

Pelacarsen is Ionis partnered with Novartis, and it failed in more than eight thousand patients. Zanvastro, the first medicine ever approved for a disease that had none, is also an Ionis drug. Forty nine patients in that pivotal trial, walking speed held steady over sixty one weeks, a label covering all ages, and in a substudy of four children under two, motor skills improved.

Pelacarsen had been lowering lipoprotein A by as much as eighty percent at twenty four weeks, and the reductions in the failed trial were in that same range. It hit its pharmacology target and still did not reduce cardiovascular events. BioPharma Dive came back to it yesterday, on a Saturday, because the question is no longer about one molecule.

What are the analysts actually asking?

Myles Minter at William Blair now reads meaningful risk into the other ongoing trials in the class. Dennis Ding at Jefferies wants the correlation between baseline lipoprotein A, the size of the reduction, and any clinical benefit. If there is no relationship between how much came out and who did better, the problem was never the drug and never the dose. It was the target, and every trial in this class is powered on it.

Under the radar this week, the memos came out on an FDA approval its own review team wanted to reject. The memos are public now. Maya, this one is yours.

Newly released memos show Asha Das, who directs the office of clinical evaluation at the agency's biologics center, overruling her own reviewers to grant accelerated approval to Replimune's Tudriqev. What the reviewers said reframes it. This would have been the third rejection for this drug. A complete response letter in July of last year said the package was insufficient to conclude substantial evidence of effectiveness. There was a second rejection in April. And in August the review team still had Replimune failing to address the deficiencies named in those letters.

And Endpoints calls this the second one?

It published the memos on Wednesday and reports this as the second consecutive high-profile override at that center. The approval letter is dated August sixth. I can accept one override as a judgment call, because the system is built to allow judgment calls. Two in a row, written down where anyone can read them, is something I would start calling a posture. I can't see the file behind that first one, and I'm not going to pretend the sentence isn't there either.

I'm not with you on the posture, Maya. Endpoints named Replimune, named Asha Das, named the August sixth letter. Nobody has named the first override. Not the drug, not the sponsor, not the date. That is one clause in one story, and I wouldn't build a filing strategy on it. Look at what that memo leans on. The advisory committee voted ten to three in favor of this drug. Das sided with an outside expert panel against her own internal reviewers, which that system is built to permit.

Then argue the evidence instead of the process. The approval rests on a single arm study. A hundred and forty patients enrolled, ninety one evaluated, a twenty four percent response rate, median duration of response fourteen point one months. The confirmatory trial, IGNYTE-3, is still enrolling, with topline expected in September of 2030. That is four years between this approval and the evidence meant to justify it. Now set it beside Etcamah, the same accelerated pathway, where the agency overruled its outside advisers instead of its own reviewers.

September twenty thirty. That one lands. What is genuinely new here is not the override, though. It is that Replimune's competitors can now read what the review division recommended and what the office director did instead, and that changes what a recommendation is worth inside the building. And uniQure walked an application into that same center on Wednesday. If priority review is granted, that puts an action date in late April or early May of 2027.

Twenty thirty is a long wait.

Quick note before we look ahead. If The Pharma Closeout is how you close out your pharma week, follow the show on Spotify or Apple Podcasts. Now, the week ahead, and it starts quiet. Friday brings an FDA decision on Telix's Pixclara, a brain imaging agent, and that is the only dated action between tomorrow and Friday. Then it thickens. September nineteenth carries two rare disease decisions, Ultragenyx in Sanfilippo type A and IntraBio's levacetylleucine in ataxia telangiectasia.

No date on the pelacarsen data?

None. An upcoming congress, Novartis says.

Then Amgen and Lilly are writing next year's budgets blind. Amgen's olpasiran took lipoprotein A down more than ninety five percent in mid-stage testing, which is more than pelacarsen ever managed, and after this week nobody can say whether lowering it further helps at all. Lilly and Silence Therapeutics are advancing their own gene-silencing approaches into the same question.

What about Glasgow?

September twenty fifth matters most to me. AbbVie takes the full CERVINO data to a plenary there, and Glasgow is where we find out whether the cytokine release grading table supports the outpatient claim AbbVie is building the launch on. October brings the full remibrutinib results at MSToronto. Disability progression was a positive trend at three months and only nominally significant at six, and that is the number that decides how broad the label gets.

Four documents have no date on them at all. uniQure's sixty day filing acceptance letter, which sets the entire 2027 calendar. Novartis's pelacarsen presentation. Alumis still plans its psoriasis filing by year end, and on my read that filing is now most of the company's story. And Kyverna's approval decision, which is where we find out whether this week was two companies or a whole field.

The thing I'll carry into Monday has no date on it either. Three months sat between the people Bristol told in June and the rest of us, and the science was identical on both ends of that gap. That is what I'll be thinking about tomorrow morning. Enjoy the rest of your Sunday.

And that wraps our week in review for Sunday, September sixth. Cell therapy's biggest idea picked up a disclosure problem. Two Phase 3 trials and a mid-stage study failed, one putting a whole drug class in question. The agency cleared one medicine over its advisers, and memos surfaced on another. And five deals landed that Maya and I still don't agree about. Follow the show on Apple Podcasts, Spotify, or wherever you listen, and a quick rating or share helps other listeners find us. You'll start every weekday with this. **Tags:** Novartis, Bristol Myers Squibb, CAR-T, rap-cel, zola-cel, Lilly, Merida, Roche, Simcere, GSK, HUTCHMED, Alteogen, Menarini, pelacarsen, Alumis, Ultragenyx, remibrutinib, AbbVie, etentamig, AstraZeneca, Etcamah, uniQure, Ionis, Zanvastro, Replimune, Telix, Kyverna, Cabaletta, Guardant360 **Sources:** the week's aired scripts (2026-08-31 STUDIO v2, 09-01, 09-02, 09-03, 09-04 v5) and `research_output.md` / `week_synthesis.md` from scheduled run 34062572761, plus the second enrichment round inserted for this take. Every figure, date and attribution in this script is carried from one of those places; see `MANUAL_TAKE_GATES.md` for the fact-audit record.