Autoimmune CAR-T Safety Faces Its First Serious Test

Autoimmune CAR-T safety is no longer an assumption. In late August 2026, Novartis paused eight immunology and neuroscience trials of rapcabtagene autoleucel (rap-cel) after three fatal cases of a severe inflammatory complication, and Bristol Myers Squibb separately halted enrollment in autoimmune studies of zolacabtagene autoleucel (zola-cel).

The pauses are a safety shock to the “immune reset” thesis, not a collapse of it. CAR-T therapy, proven in B-cell cancers, has been moving fast into lupus, myositis, multiple sclerosis, and related diseases on the promise that a one-time treatment can produce durable, drug-free remission. Early efficacy signals still look encouraging. What has changed is the assumption that the inflammatory toxicities seen in oncology would stay mild once the same platform moved into autoimmune patients.

What happened

Novartis halted screening, randomization, and dosing across eight early- and mid-stage immunology and neuroscience trials on 24 August 2026. The trigger was three serious cases of immune effector cell-associated hemophagocytic syndrome (IEC-HS), and complications from those events led to fatal outcomes. Novartis is working with regulators and independent data-monitoring committees. Its two oncology trials in lymphoma and leukemia, including high-risk large B-cell lymphoma, are unaffected.

Ad

The paused studies include the Phase 2 AUTOGRAPH program in systemic lupus erythematosus and lupus nephritis, systemic sclerosis, ANCA-associated vasculitis, and idiopathic inflammatory myopathies, plus Phase 1/2 trials in rheumatoid arthritis, Sjögren’s disease, generalized myasthenia gravis, relapsing MS, and non-active progressive MS.

Rap-cel (YTB323) uses the same CAR construct as Kymriah (tisagenlecleucel) but is engineered for faster, simpler manufacturing.

Bristol Myers Squibb paused enrollment in autoimmune studies of zola-cel after routine safety surveillance picked up what it described as transient and reversible inflammatory events. The company reported no deaths tied to this pause and framed the move as voluntary, out of an abundance of caution, with a stated aim to finish the review and resume as quickly as possible. Zola-cel (BMS-986353) is built on the construct used in Breyanzi (lisocabtagene maraleucel). One earlier IEC-HS case was described in February Phase 1 results.

Why IEC-HS matters for autoimmune CAR-T safety

IEC-HS, also described as an HLH-like syndrome, is a rare and severe hyperinflammatory complication of immune-effector therapies including CAR-T and some T-cell-engaging bispecifics. The engineered cells expand quickly and drive a cytokine cascade that can damage organs. It is a known, labeled risk in oncology CAR-T, but risk tolerance is lower in autoimmune disease: these patients are not facing immediate cancer mortality, and effective chronic therapies already exist.

People with active autoimmunity already have primed inflammatory circuitry, so a CAR-T expansion burst may be harder to tolerate than it is in many oncology settings. How quickly the syndrome is recognized and treated also matters. Novartis has said IEC-HS is a known CAR-T risk and that it is working on earlier identification and management.

Developer positions after the holds

Shares in Kyverna, Cabaletta Bio, Allogene, CRISPR Therapeutics, and Fate Therapeutics fell on the news and then recovered as investors parsed differences in process and construct. Peer developers using more conventional manufacturing argue the read-through is limited.

Kyverna said on 1 September 2026 that miv-cel “has a distinct construct design and is produced using a well-established, validated manufacturing process,” that it has seen no safety findings that would lead it to pause any program, and that it remains on its stiff person syndrome filing timeline.

Why manufacturing became the debate

Standard autologous CAR-T production can take several weeks: leukapheresis, shipment to a specialized plant, genetic modification, expansion, quality release, and return. Novartis and BMS have invested in faster processes that grow modified cells more quickly.

Analysts at Jefferies, William Blair, and elsewhere have raised the possibility that faster expansion kinetics produce a steeper post-infusion cell rise and a larger inflammatory pulse. TD Cowen’s Phil Nadeau argued Cabaletta’s nine-day standardized process was built to “minimize immune related toxicities” and that protocol differences limit read-through. Jefferies’ Roger Song noted that autoimmune patients have “more activated immunity at baseline,” which could amplify any manufacturing-driven burst.

This remains a hypothesis, not a finding. Alternatives include chance clustering, indication-specific biology, lymphodepletion intensity, CD28 versus 4-1BB costimulation, and how quickly IEC-HS was recognized and treated.

What changes next

Safety packages will now be scrutinized for IEC-HS specifically, alongside the established risks of cytokine release syndrome and immune effector cell-associated neurotoxicity syndrome. Expect more protocol language on early labs, exclusion of recent infection or hyperinflammation, and predefined rescue algorithms. The manufacturing story has become a safety story: companies with conventional or tightly controlled processes will keep drawing that contrast, while rapid-process programs will need to show comparable expansion kinetics and monitoring. William Blair has argued the big-pharma pauses cut near-term competitive crowding, while also noting that greater physician awareness of IEC-HS should improve management over time.

For Kyverna, the nearest-term regulatory test, any later IEC-HS case would be highly visible. Cabaletta’s late-stage myositis program and lupus work remain on the field’s critical path. Clinicians and trial sites operate with structurally lower risk tolerance in autoimmunity than in relapsed or refractory cancer, and enrollment may slow even in unpaused programs as investigators and ethics boards digest the Novartis deaths. If a first approval proceeds, labeling and monitoring expectations may be shaped by this episode even for an applicant with a clean database, and the value case for a one-time cell therapy will have to price in rare but catastrophic inflammatory risk, monitoring cost, and specialized-center delivery.

What to watch

The open questions are whether Novartis finds a shared feature among the three fatal cases, whether BMS restarts quickly with only protocol tweaks, and whether Kyverna’s Q4 2026 filing draws FDA questions on class-wide IEC-HS risk. Other CD19 autoimmune programs at Autolus, Allogene, CRISPR, and Fate may disclose protocol amendments or further IEC-HS cases, and regulators could add holds beyond the companies’ voluntary actions.

Bottom line

The field is splitting, at least in investor and protocol language, into rapid-process Big Pharma CD19 programs and conventional-process or allogeneic specialists. Until Novartis publishes a root-cause analysis and peers accumulate larger safety datasets, every autoimmune CAR-T readout will be read first for IEC-HS, then for efficacy.