A Patient Died in BMS’s MS CAR-T Trial. No One Can Say Why.

A patient with highly active relapsing multiple sclerosis died 44 days after being infused with zolacabtagene autoleucel, Bristol Myers Squibb’s investigational CD19-directed CAR-T therapy, in the Phase 1 Breakfree-2 study. The disclosed sequence is specific: late-onset motor symptoms that were explicitly not ICANS, intermittent corticosteroids and intravenous immunoglobulin that failed to resolve them, high-dose chemotherapy, then septic pneumonia, then death the following day.

Whether the therapy caused it is unresolved, and BMS says so plainly. “The patient had a late onset side effect, a series of other complications and was heavily pretreated prior to CAR T, confounding our ability to determine the root cause of the issue,” a company spokesperson said.

That is the honest version of the record. A patient is dead, the sponsor cannot say why, and the structure that was supposed to make that answerable is the thing worth examining.

This is a modality signal, not a company scandal

The temptation is to read the death as a BMS failure. The sequence of events does not support that, and the real story is larger.

In September, Novartis halted eight trials of its own CAR-T candidate after three deaths, and BMS paused zola-cel shortly afterward. Both pauses were sponsor-initiated. Neither company has reported a regulatory clinical hold, which matters: these were companies stopping themselves, not an agency stopping them.

BMS attributes its own pause to an increase in ICANS events in the pivotal Breakfree-SLE lupus study, says it did not see the same break in trend in the Phase 1 MS study, and notes the MS death occurred after the pause decision had already been taken. On BMS’s account these are two separate signals rather than one pattern, and nothing in the public record contradicts that.

The surrounding numbers deserve the same precision. Across the zola-cel autoimmune programs, 190 patients have now been treated in Phase 1 with no other Grade 5 event. In the MS cohorts specifically, cytokine release syndrome occurred in 42.4% of patients and ICANS in 7.6%, most Grade 1 or 2 and resolving in a median of two days. That is a toxicity profile a reviewer can argue about. It is not a program in freefall.

The question the pause does not answer

What two sponsors pausing inside the same quarter does establish is that the field moved into autoimmune disease faster than its safety understanding did.

BMS’s approved CAR-T products, Abecma and Breyanzi, already carry FDA boxed warnings for T cell malignancies following treatment in oncology indications. The agency documented those signals in its labeling updates. The platform entered autoimmune disease with that knowledge on the table.

Autoimmune patients are not oncology patients. Their immune systems are dysregulated, they are often on chronic immunosuppression, and the conditioning regimens required for CAR-T engraftment carry infection risks that differ materially from what a lymphoma patient presents. Septic pneumonia is the kind of outcome that intensive lymphodepletion in an already-compromised immune environment can drive. That is a hypothesis about this death, not a finding, and BMS has not endorsed it.

Which is exactly why the preclinical question is worth asking out loud: what did the nonclinical package for an autoimmune indication predict about infection risk during lymphodepletion, and was it built on disease-specific work or carried over from oncology? A sponsor that builds an autoimmune CAR-T IND on oncology-derived preclinical data alone, without model work accounting for the altered immune milieu, is running a thinner experiment than the Phase 1 label implies. Whether BMS did that is not established in the public record, and it should be.

The FDA’s own posture makes the question sharper. Its current thinking on autoimmune CAR-T was articulated through an editorial in the Annals of Internal Medicine rather than formal guidance, promising a “flexible regulatory approach” with case-by-case review of nonclinical requirements. The agency explicitly acknowledged that CAR-T preclinical testing is complicated by the biological complexity of the modality. That acknowledgment is honest. It also leaves the threshold for “enough characterization” undefined at precisely the moment a dozen sponsors are racing into the same space.

What the record should show next

BMS says it is working with the FDA to restart all zola-cel studies, targeting resumed enrollment in early 2027. That restart is the document worth watching.

Sponsors are required under IND safety reporting regulations to report unexpected fatal serious adverse events to FDA on an expedited basis. The question is not whether BMS filed, it is what the agency required before letting enrollment resume, and whether any of that becomes visible. The T cell malignancy signal from Abecma and Breyanzi produced labeling updates the field could read and act on. A death and two sponsor-initiated pauses inside one quarter should produce something comparably legible.

If instead the FDA processes these events quietly, inside individual IND files, the flexible posture it promised for autoimmune CAR-T becomes harder to distinguish from regulatory silence. Not because the agency did nothing, but because nobody outside the IND can tell.

For sponsors with autoimmune CAR-T programs in dose escalation right now, the operational read is narrower and more useful than any verdict on BMS. Convene your DSMB, pull your infection-monitoring protocol, and verify that your stopping rules were written for your patient population rather than borrowed from an oncology predecessor study. A death from septic pneumonia after lymphodepletion in an autoimmune patient is a specific enough event to design against, whatever its final adjudication.

The adjudication may never come. BMS has already said the confounders are too tangled to isolate a root cause, and that is a defensible clinical position rather than an evasive one. But “we cannot determine why this patient died” is a reasonable answer only once, and only if the next programs are built so the question is answerable next time.

References

1. Pharmaceutical Executive, “Bristol Myers Squibb Discloses Patient Death in CAR-T Multiple Sclerosis Trial”
2. BioSpace, “BMS Reveals Death From Septic Pneumonia of Multiple Sclerosis Patient in CAR T Trial”
3. BioSpace, “Novartis Pauses CAR T Studies Due to 3 Deaths; BMS Follows Suit After Safety Events”
4. BioPharm International, “BMS Reports Death in Phase 1 Zolacabtagene Autoleucel MS Study”
5. FDA, Boxed Warning for T Cell Malignancies Following BCMA- or CD19-Directed CAR-T
6. FDA, T Cell Malignancy Safety Communication documentation
7. Annals of Internal Medicine, FDA editorial on CAR-T regulatory approach for autoimmune and rheumatic conditions