T-cell engagers in healthy volunteers
Finding the active dose before the first patient
A working note in the project on speeding up first-in-human T-cell engager studies: one of the approaches collected there, and the search that followed it. The registry and abstract search was run on 2026-09-10. Every source is characterized from a registry record, an abstract or a press release, none has been read, and every entry in the references carries ❌.
In brief
The idea. Give one or two doses of the T-cell engager (TCE) to healthy volunteers (HVs), climbing a dose ladder until the first sign of activity: a grade 1 cytokine release (fever), or peripheral B cells falling by half. That dose, found in weeks in a phase 1 unit, becomes the starting neighbourhood for patients, and the patient study begins where a conventional escalation would have finished.
It has been done. ✅ Three healthy volunteer studies of T-cell engagers are registered, all started between December 2025 and July 2026, all in Melbourne, all for autoimmune indications, all subcutaneous: an absolute bioavailability study of cizutamig, a BCMA x CD3 engager of Chinese origin; a first-in-human escalation of XmAb657, a CD19 x CD3 engager, that enrols healthy participants alongside patients; and a first-in-human study of CND319, a CD19/CD20 engager, with a single-ascending-dose part in healthy adults. None has posted a result. Section 2 has the table.
Where it applies. ⚠️ Autoimmune programmes, and only those. The three sponsors that dosed volunteers all built molecules with a tuned CD3 arm and an extended half-life, intended for one or two administrations, and the volunteer’s B-cell count is the same measurement as the patient’s. In oncology the target burden that drives cytokine release is absent in a volunteer, so the dose that produces fever in a volunteer says little about the dose that produces it in a patient with a tumour, and no oncology TCE has been given to volunteers. Section 4 works through the translation.
How much faster. Of the order of six months to a year against a conventional escalation in patients, under the assumptions in Section 5, which are there to be replaced with a programme’s own.
The Ouro question. ❌ Unresolved. A colleague recalled a Chinese healthy volunteer study of a TCE later acquired by Ouro Medicines. Ouro’s molecule is CM336, a BCMA x CD3 engager from Keymed Biosciences in Chengdu, licensed to Ouro in January 2025; Ouro was acquired by Gilead in 2026. No healthy volunteer study of CM336 or OM336 appears on ClinicalTrials.gov. The Chinese registries were not searched. The molecule that does have a registered volunteer study, cizutamig, is also a Chinese-origin BCMA x CD3 engager, from EpimAb in Shanghai by way of Vignette Bio and Candid Therapeutics, and the two may have been conflated. Section 2 has both, and the question to put back to the colleague is the trial identifier.
1. The idea, as posed
Start in healthy volunteers. Give one dose, or two, per volunteer, climbing between cohorts, and stop at the first dose that does either of two things: produces a grade 1 cytokine release syndrome, which in a volunteer means a fever, or depletes peripheral B cells by 50%. Call that the active dose. Then open the patient study with a starting dose and a step-up regimen already anchored to it.
The claim to test is about speed. A phase 1 unit recruits volunteers from a standing pool, doses a cohort within days of the previous cohort’s readout, and reads a single-dose cytokine and B-cell response inside a week. A conventional oncology escalation waits on accrual of patients with the right disease at the right stage, a 28-day dose-limiting-toxicity window, and a cohort review. If the volunteer ladder climbs three times faster, the gap between the starting dose and the active dose is crossed before the first patient is enrolled.
2. What has been done
Registered healthy volunteer studies of T-cell engagers
| Study | Molecule | Target | Design | Volunteers | Start | Status | Marker |
|---|---|---|---|---|---|---|---|
| NCT07236411 | Cizutamig (Candid; EMB-06, EpimAb) | BCMA x CD3, 2+2 FIT-Ig | Randomized, double-blind, single IV or SC dose against placebo; absolute bioavailability | 24 healthy adults, all arms | 2025-12-04 | Completed 2026-07-14, no results posted | ❌ |
| NCT07284797 | XmAb657 (Xencor) | CD19 x CD3, 2+1, extended half-life | Open-label first-in-human dose escalation, SC | Healthy adults as the first cohort, then myositis, systemic sclerosis, Sjögren’s | 2026-02-24 | Recruiting; primary completion Oct 2028 | ❌ |
| NCT07712939 | CND319 (Candid) | CD19/CD20 x CD3 | Open-label first-in-human; single ascending dose in volunteers, multiple ascending dose in patients | Healthy adults 18–65 in Part 1 SAD; rheumatoid arthritis and Sjögren’s in MAD and expansion | 2026-07-20 | Recruiting; primary completion Mar 2028 | ❌ |
Three things the table shows. All three studies are recent: the earliest started in December 2025. All three are for autoimmune indications, and none is an oncology TCE. All three molecules were engineered for the setting: a CD3 arm with reduced affinity, a long half-life, subcutaneous administration, and a single dose or short course as the intended regimen.
What none of the registry records shows is the readout the idea depends on. The registered primary outcomes are adverse events (XmAb657, CND319) and absolute bioavailability (cizutamig); the secondary outcomes are serum concentration and anti-drug antibodies. B-cell counts and cytokines will have been measured and are not registered as endpoints, so what the volunteer cohorts found will have to come from a presentation. Xencor has said it expects to report on the XmAb657 study in the second half of 2026.
What is known about the molecules from patients
Cizutamig had been given to 40 patients with relapsed or refractory multiple myeloma at weekly doses from 0.2 to 300 mg before the volunteer study, with cytokine release syndrome at grade 1 in 18% and grade 2 in 8%, median onset one day, and no ICANS; and to one patient with refractory rheumatoid arthritis, with no cytokine release (entry 24). By January 2026 the sponsor reported 80 patients dosed, half with autoimmune disease, mild cytokine release in fewer than 20%, and no ICANS. So the volunteer study came after the dose range was known from patients, which is the reverse of the idea in Section 1. It was run to measure bioavailability, and it establishes that a volunteer can be dosed, not that the active dose was found that way.
XmAb657 is the case that matches the idea: a first-in-human escalation whose first cohort is healthy participants. In cynomolgus monkeys a single dose suppressed B cells to below 5 cells/µL for at least 42 days, in lymph node and marrow as well as blood, with no clinical signs of cytokine release, and a half-life near 15 days (entry 25). Whether the volunteer cohort climbs to a B-cell-depleting dose, or stops at a dose chosen to be subactive, is not stated in the registry record and is the first thing to find out when the study reports.
The Ouro Medicines molecule
CM336, gamgertamig, is a BCMA x CD3 engager developed by Keymed Biosciences in Chengdu, in a phase 1/2 in multiple myeloma in China since 2022 and in autoimmune cytopenias, bullous disease, Sjögren’s and several plasma cell disorders since 2025. Keymed licensed rights outside Greater China to Platina Medicines, a unit of Ouro Medicines, in January 2025; the molecule was renamed OM336; Gilead announced the acquisition of Ouro in March 2026 and completed it in June. The registered OM336 studies are in autoimmune cytopenias (United States and Australia), Sjögren’s and myositis (New Zealand), and sensitized kidney transplant candidates (Austria and Germany). None enrols healthy volunteers, and the one CM336 record that carries the healthy-volunteer flag, NCT06745687, is a study in high-risk smouldering myeloma whose flag is a registry error.
If a Chinese healthy volunteer study of CM336 exists it would be on the Center for Drug Evaluation registry (chinadrugtrials.org.cn) or the Chinese Clinical Trial Registry, neither of which was searched. The question for the colleague is the trial identifier or the molecule’s code, since the recollection may equally be of cizutamig.
What was searched
ClinicalTrials.gov, by intervention terms (bispecific, trispecific, T-cell engager, CD3) restricted to studies accepting healthy volunteers; Europe PMC for published reports of a CD3-engaging bispecific in healthy subjects, which returned none; and press releases and meeting abstracts for the three molecules found. Not searched: the Chinese registries, the EU Clinical Trials register, the Australian register (ANZCTR), and the PAGE and ACoP abstract archives.
3. What a volunteer study can measure
The readouts, in the order they arrive after a single dose.
| Hours after dose | Readout | What it says |
|---|---|---|
| 2–6 | Lymphocyte count falls (T-cell margination) | CD3 was engaged. The earliest pharmacodynamic signal, and present at doses well below the active one. |
| 2–24 | IL-6, IFN-γ, TNF in serum | The magnitude of T-cell activation. The quantity a grade 1 CRS is the clinical shadow of. |
| 6–48 | Fever | Grade 1 cytokine release syndrome, the first of the two stopping signals in Section 1. |
| 24–72 | Peripheral CD19+ B cells | Depletion, the second stopping signal. Depth at nadir and, on later visits, the duration. |
| Days to weeks | Serum concentration, anti-drug antibodies | PK, and whether a second dose is even interpretable. |
The two stopping signals will not arrive together. A CD3 arm tuned down, which all three molecules in Section 2 have, is tuned to deplete without releasing cytokine. In a volunteer that means B cells may fall by half, or to zero, at a dose that produces no fever at all, and the first fever may come a log higher or never. The idea as posed treats either signal as “active”; the design has to say which one, because they name different doses. For an autoimmune programme the answer is depletion, since depletion is the goal and the fever is the cost, and the fever dose then becomes an upper bound the patient study need not approach.
One volunteer per dose, or several. Volunteers are not scarce, so there is no reason to climb the ladder within one person. A cohort of four to six per dose level, with a sentinel pair dosed a day ahead of the rest, is the ordinary phase 1 shape and gives a between-subject estimate of the response at each dose, which one person climbing cannot.
A second dose in the same volunteer. Give a priming dose, wait a week, give a higher dose, and compare the cytokine response with a cohort that received the higher dose naive. That is a direct measurement of the tolerance effect, a primed T-cell pool releasing less cytokine at the next dose, in a system with no tumour and no disease, and it is the measurement a step-up regimen for patients would be designed from. The confound is that by the second dose the volunteer’s blood B cells are gone, so a lower cytokine response could mean primed T cells or an absent target. Tissue B cells persist longer than blood B cells and are the majority of the pool, which is why the comparison is still informative, but the two explanations have to be separated by the B-cell measurement, not assumed away.
4. Translation from volunteer to patient
The active dose found in a volunteer is useful exactly to the extent that the patient’s response at the same dose is predictable from it. Three things differ.
Target burden. Cytokine release scales with the number of target cells the engager finds. A volunteer’s peripheral B-cell count is in the ordinary range; an autoimmune patient’s is usually similar, unless prior rituximab has emptied it, and the plasma cell compartment that a BCMA engager targets is larger in a patient with active humoral disease. A patient with a haematological malignancy carries orders of magnitude more target. So the fever dose in a volunteer is a lower bound for the fever dose in a low-burden patient and an overestimate, by an unknown factor, for a patient with tumour. The B-cell depletion dose translates better, since depletion of a normal-sized compartment is the same event in both.
T-cell fitness. An autoimmune patient on glucocorticoids, mycophenolate or methotrexate has a suppressed T-cell response; a volunteer does not. The volunteer dose may over-predict activity in a treated patient, in the direction that makes the patient study start too low rather than too high, which is the safe direction and the slow one.
Duration. A volunteer given a long-half-life engager loses peripheral B cells for weeks to months (42 days or more in the primate study, entry 25). That is the cost the volunteer bears for no benefit, and it is the reason the three registered studies are all for indications where a single administration is the intended regimen. The ethics review will have weighed it, and the consent form will say how, and neither document has been read.
Where the translation holds. Autoimmune indications, B-cell or plasma-cell targets, engineered CD3 arms, a depletion readout rather than a fever readout. That is the set of conditions under which the three registered studies were run, and it is the set under which the idea in Section 1 works. Outside it, the volunteer dose is a MABEL with human data behind it, which is still an improvement on a MABEL without.
5. Where it applies, and how much time it saves
A volunteer study replaces a prediction of the active dose from primate data with a measurement in humans. The residual uncertainty is the volunteer-to-patient translation of Section 4, a factor of a few rather than of tens.
- ✅ B-cell-depleting TCE, autoimmune indication. Run the volunteer study. The patient study starts one or two steps below the volunteer depletion dose, with a step-up regimen designed from the volunteer cytokine data.
- ❌ Solid-tumour or haematological TCE. No volunteer has been dosed with an oncology TCE, the target burden argument in Section 4 says a volunteer readout would not translate, and the ethics of depleting a healthy person’s B cells for a lymphoma programme are worse than for an autoimmune one.
How much time. Assumptions, not measurements, and the reader should change them.
| Quantity | Conventional 3+3 in patients | Volunteer single ascending dose |
|---|---|---|
| Recruitment per cohort | Weeks, disease-dependent | Days, from a standing pool |
| Observation before the dose decision | 28-day DLT window | 7 days: CRS window closed, B-cell nadir passed |
| Cohort cadence | 6–8 weeks | 1–2 weeks |
| Dose levels to cross a 3-log gap at 3x steps | 6–7 | 6–7 |
| Time to the active dose | 9–14 months | 2–4 months |
The saving is of the order of six months to a year. A saving under about eight weeks, one conventional cohort cycle plus its accrual, would sit inside the noise of ordinary trial execution; this one does not, if the cadence assumptions hold, and the cadence is the row to check first against a real unit’s calendar.
6. What to do with this
- Ask the colleague for the trial identifier behind the Ouro recollection, and search the Chinese registries if it is not forthcoming.
- Watch for the XmAb657 volunteer cohort results, expected in the second half of 2026. What dose the volunteers reached, what their B cells did, and whether anyone had a fever answers most of Section 3.
- Read entry 20, the EMA first-in-human guideline, for what it says about healthy volunteers and immunomodulators, since that is the document an ethics committee reads.
- Put the speed table against a real phase 1 unit’s cohort calendar and a real programme’s escalation history, so that the assumptions become two measured columns.