Patients vs. Healthy Volunteers in First-in-Human Trials of CAR-T and T-Cell Engagers for Autoimmune Disease

Patients or healthy volunteers? First-in-human trials of CAR-T and T-cell engagers in autoimmune disease

Short answer

Every early clinical program we identified for CAR-T cells or CD3-based bispecific T-cell engagers (TCEs) in lupus, myasthenia gravis (MG) and systemic sclerosis (SSc) dosed patients, not healthy volunteers. Most restricted entry to patients with active, severe or treatment-refractory disease. An early five-patient series treated life-threatening, multi-drug-refractory SLE under a compassionate-use program ⁴²⁵. The registered phase 1 CAR-T and TCE trials in these indications all specify disease populations ^{153160161115121318390201208}. We found no published or registered study in which a healthy volunteer received a CAR-T product or a TCE for an autoimmune indication ^97319201208.

No FDA, EMA or national guidance document we reviewed addresses autoimmune CAR-T or TCE population selection by name. The patient-only approach follows from general cell and gene therapy, ATMP and first-in-human (FIH) guidance:

• FDA states that for most cellular and gene therapy trials "the benefit-risk profile is not acceptable for healthy volunteers" ²⁶⁴.
• EMA states that investigational ATMP trials are "usually conducted in patients, not healthy volunteers" ²⁴¹.
• Health Canada says healthy-volunteer FIH trials of cell therapies "may be unethical" ⁴¹.
• Japan's regenerative medicine GCP guidance defaults to "patients with the disease or condition for which the investigational product is intended" ³⁷⁰.
• For bispecific antibodies, FDA says healthy volunteers "may not be appropriate" because of immunogenicity and toxicity ²⁵⁴.

What the early trials actually enrolled

CAR-T in systemic lupus erythematosus

• An early five-patient series. Five patients (median age 22) had life-threatening SLE, a median SLEDAI of 16, and disease refractory to several immunosuppressants. They received autologous anti-CD19 CAR-T cells (1×10^6/kg) after fludarabine and cyclophosphamide lymphodepletion under a compassionate-use program ⁴²⁵.
• First formal phase 1 data. A single-arm, multicentre phase 1 trial treated 13 patients with biopsy-confirmed class III to V lupus nephritis using a BCMA-CD19 compound CAR-T. One underdosed patient was recorded as compassionate use ⁴³⁰.
• Registered phase 1 programs consistently require refractory or active disease. Examples:
  • Kyverna's KYV-101 in refractory lupus nephritis (KYSA-1, KYSA-3) ¹⁶¹¹⁶⁰
  • CHOP's CART19 in adolescents and young adults with SLE ¹⁵¹
  • Multiple Chinese investigator-initiated CD19, BCMA/CD19 and allogeneic CAR-T studies in refractory SLE ^{148155156158167171}
• Allogeneic products. Some programs use universal CAR-T derived from healthy donors. The donor is the cell source only; recipients are still patients with refractory SLE ¹⁵⁶³¹⁹.

CAR-T in myasthenia gravis

• Descartes-08 is an autologous BCMA-directed mRNA CAR-T. It was studied in a phase 1b/2a trial (MG-001, NCT04146051) that enrolled 14 adults with generalized MG and an MG-ADL score of 6 or higher. Phase 1b required MGFA class III to IV disease ⁴¹⁷. It was then tested in a randomized, placebo-controlled phase 2b trial in 26 patients ⁴²⁰.
• No lymphodepletion. Descartes-08 was given without lymphodepletion chemotherapy, including outpatient dosing in phase 2b ⁴¹⁷⁴¹⁹. This lowers procedural risk, but the sponsor still enrolled only patients.
• Other phase 1 MG programs also enroll only patients, usually with refractory disease:
  • CD19 CAR-T in refractory MG (Zhejiang University) ¹¹⁵
  • BCMA CAR-T in refractory generalized MG with MGFA and MG-ADL/QMG thresholds ¹¹⁹¹²⁴
  • Arcellx's anito-cel dose escalation in generalized MG ¹²¹
  • Novartis's YTB323 in treatment-resistant generalized MG ¹¹⁸
  • A pediatric Descartes-08 study in juvenile MG ¹²⁸

CAR-T in systemic sclerosis

• Industry phase 1/2 programs enroll adults with SSc. Kyverna's KYV-101 enrolls adults with SSc ³. Cabaletta's RESET-SSc (CABA-201) requires active disease ¹⁸.
• CASTLE basket trial. This phase 1/2a trial treated 9 treatment-resistant SSc patients with zorpocabtagene autoleucel after cyclophosphamide and fludarabine lymphodepletion ³⁴⁰.
• Pediatric and basket studies. A pediatric study targets childhood-onset refractory SSc ⁶⁷. European basket studies enroll severe, treatment-failure systemic autoimmune disease ²⁰¹⁶.
• Not every phase 1 study is first-in-human. The French CELL-ATTACK 1 study describes itself as "not first-in-human" ¹⁶. For regulators, "first in autoimmune disease" and "first in human" can be different questions.

Bispecific T-cell engagers

• A-319 (CD19×CD3). A phase 1 study in adults aged 18 to 60 with active/refractory SLE and positive autoantibodies. Healthy volunteers are not eligible ³⁹⁰³⁹³. Twelve active SLE patients have been reported from the intravenous arm, described as first in disease ⁵⁸.
• Other registered phase 1 TCE studies also enroll only patients:
  • GB261 (CD20×CD3) in refractory seropositive SLE ²⁰¹
  • F182112 (BCMA×CD3) in SLE ²⁰⁰
  • CM336 (BCMA×CD3) in autoimmune bullous disease ²⁰³
  • Alnuctamab (LATTE) in moderate-to-severe refractory SLE, described as the first study of that molecule in SLE rather than a first-in-human study ²⁰⁸
• Repurposed oncology TCEs. Early autoimmune use has come through compassionate or named-patient access in refractory patients, not formal FIH trials:
  • Teclistamab in treatment-refractory SSc ⁶⁴ and in an 18-patient series (10 SSc) with a median of five prior therapies ⁶²
  • Low-dose blinatumomab in multidrug-resistant rheumatoid arthritis ⁷³⁶⁷
• A true FIH example. ABO2203, a lipid-nanoparticle mRNA encoding a CD19-targeting TCE, had its first-in-human study in three patients with refractory secondary immune thrombocytopenia ⁸³.

Caveat on the registry data

Each registered CAR-T and TCE trial cited here sets the ClinicalTrials.gov healthy-volunteers flag to No, and each trial's stated population and eligibility text restricts entry to patients ²⁰¹²⁰⁸. We also checked for CD3-bispecific studies in healthy participants for autoimmune indications. One immunology program (GSK3888130B, NCT05131971) dosed healthy participants, but the record does not identify it as a T-cell engager ²⁹⁵²⁹⁶.

What regulators say about the population choice

FDA

FDA's early-phase CGT guidance defines healthy volunteers as people who "do not have the disease or condition of interest." It limits their use to products with short duration of action or a well-characterized class safety profile ²⁶⁴. Autologous CAR-T cells, which persist and expand after infusion, do not fit either exception.

The same guidance cautions against assuming the sickest patients are the right FIH population. Limited physiologic reserve can make toxicity harder to tolerate, and disease-related events and concomitant therapies can confound interpretation. The sponsor must justify the chosen population in the IND ²⁶⁴²⁶⁵. If eligibility depends on failure of available therapy, the protocol should document that each patient's options were adequately evaluated ²⁶⁵²⁶⁷. This matters for autoimmune programs, where "refractory" is the most common entry criterion in the registered trials ^153160115390.

For CAR-T outside oncology, FDA has not published population-specific recommendations. The 2024 CAR-T guidance directs non-oncology sponsors to pre-IND engagement with OTP ¹⁴⁴. Its oncology stopping-rule examples, such as more than two Grade 4 CRS events in an FIH CAR-T study, show the level of safety oversight FDA expects ¹⁴⁵.

For TCEs, the 2021 bispecific guidance pairs its caution on healthy volunteers with a recommendation to consider a MABEL-based FIH dose for agonistic constructs ²⁵⁴. FDA's 2023 immunotoxicity guidance adds that for immune-stimulating biopharmaceuticals, a MABEL or pharmacologically active dose may be more appropriate than a NOAEL-based dose ²⁵⁵. FDA's June 2026 draft on QSP-based MABEL dosing tells sponsors to choose the most conservative starting dose when estimates disagree. It allows a starting dose above absolute MABEL in life-threatening disease with justification and safeguards such as staggered enrollment ²⁵⁹.

EMA and the EU

EMA's revised FIH guideline sets no default population. It lists the factors that decide it, including whether toxicity can be justified in healthy volunteers and whether the target is present in healthy people as well as patients ²²⁸. The 2007 version named CD3 and CD28 "super-agonists," and cytokine cascades with potential for uncontrolled amplification, as modes of action needing special precautions ²²⁵.

For FIH trials that start in patients, the recommended starting dose is one expected to have minimal pharmacological effect while reflecting disease severity. EMA advises a single-ascending-dose design in the first patient cohort where healthy people and patients may differ ²²².

The 2025 investigational ATMP guideline is the most direct EU statement. It says trials are usually conducted in patients and that the protocol must justify the population. Sponsors should consider disease burden, ability to tolerate adverse events, and prior and concomitant therapies ²⁴¹. Staggered dosing, waiting periods and stopping rules from the general FIH guideline still apply to ATMPs ²⁴⁰²³³.

National regulators

• MHRA points sponsors to the CHMP first-in-human guideline for FIH risk factors ⁴³¹. It gives patient-first trials their own label ("first time in patient") for cases where healthy-volunteer dosing would be unethical or impossible ³⁹⁹. FIH and ATMP trials are excluded from the UK's automatic authorisation route ³⁹⁵. Higher-risk FIH trials can be referred to the CTBVEAG ⁴⁰⁶.
• Health Canada cites the inherent risks of cell administration, including tumour formation, immunogenicity, inflammatory events and biodistribution. It expects longer-than-normal follow-up matched to the patient population ⁴¹⁴²⁴⁵.
• Japan makes patient enrollment the default for regenerative medicine products, including gene-modified cell products ³⁷⁰. Its gene therapy guidance frames residual unknown risk as acceptable for seriously ill patients with limited conventional options ³⁶¹.
• Germany's PEI confirms that the hospital exemption covers individual-patient ATMP use ³¹⁰. In the pages reviewed, it does not describe specific approvals for autoimmune CAR-T ³⁰⁹³¹⁰.

Why patient-only enrollment is the norm

The guidance and trial data give four connected reasons.

1. Persistence and irreversibility. FDA's core objection to healthy volunteers in CGT trials is the risk of extended or permanent effects ²⁶⁴²⁷⁰. Autologous and allogeneic CAR-T products expand and persist, so their exposure cannot be treated as a time-limited drug exposure.
2. The pharmacology is the toxicity. EMA lists B-cell depletion, alongside CRS and ICANS, as an important adverse reaction of CAR-T ²³⁷. The intended effect in autoimmune disease is deep B-cell or plasma-cell depletion. A healthy volunteer would take on the main adverse effect with no possible benefit.
3. Procedure-related risk. Most autoimmune CAR-T protocols use fludarabine and cyclophosphamide lymphodepletion ⁴²⁵³⁴⁰. EMA requires that apheresis and lymphodepletion risks be assessed as part of the therapeutic intervention and captured in the safety database ²⁴¹²³⁹. Descartes-08's lymphodepletion-free design reduces this burden, but it was still tested only in MG patients ⁴¹⁷.
4. Informative data. EMA's FIH guideline weighs whether the target and PK/PD behaviour in healthy people would represent the patient population ²²⁸³⁹⁴. B-cell reconstitution, autoantibody decline and disease activity can only be measured in patients.

For TCEs the reasoning is less absolute. FDA frames healthy-volunteer unsuitability as product-specific ("may not be appropriate") ²⁵⁴, and EMA's FIH guideline is explicitly case by case ²²⁸. In practice, every autoimmune TCE study we found enrolled patients, and the earliest access to repurposed oncology TCEs was compassionate use in refractory disease ^{3902012086473}.

Practical implications for sponsors

• Justify the population explicitly. FDA expects the IND to explain why the chosen patients offer the best balance of risk, potential benefit and interpretable data. The most severe patients are not automatically the right choice ²⁶⁴²⁶⁵. EMA expects the same in the CTA ²⁴¹.
• Engage early on non-oncology CAR-T. FDA's CAR-T guidance does not cover autoimmune indications and points sponsors to pre-IND discussion with OTP ¹⁴⁴.
• Document "refractory" rigorously. When eligibility depends on failed standard therapy, protocols should record how each patient's alternatives were evaluated ²⁶⁷.
• Size the TCE starting dose conservatively. Use MABEL or a minimally pharmacologically active dose, supported by cytokine-release data ²⁵⁴²⁵⁵. Consider a single-ascending-dose first patient cohort ²²²³⁹⁷.
• Build ATMP-grade safeguards. Plan staggered dosing, waiting periods and stopping rules ²⁴⁰. Use predefined CRS/ICANS management algorithms and long-term follow-up ²³⁷⁴⁵.
• Plan for national layers. These include UK expert-advisory referral for higher-risk FIH trials ⁴⁰⁶, Japan's patient default for regenerative products ³⁷⁰, and Germany's hospital-exemption route for individual-patient ATMP use ³¹⁰.

Evidence gaps

• None of the regulatory documents reviewed discusses autoimmune CAR-T or TCE population selection by name. The positions above apply general CGT, ATMP, bispecific and FIH guidance to this setting.
• This article does not draw on agency-specific feedback on individual programs, such as pre-IND minutes, EMA scientific advice or CTA assessment reports.
• The only published SSc CAR-T data cited here are from the CASTLE basket trial ³⁴⁰. Earlier compassionate-use SSc case reports are not covered.