References for the TCE-IPDE working document
What to read, in what order, and what has actually been checked
No number behind the specification has been checked against its source. The teclistamab PK parameters in Section 8, the Friberg system values in Section 17 and the MCLA-117 dose table in Section 6 were transcribed, the last from an image of a poster. None can go into final work until it is checked.
Reading stands as follows. The first queue has five of six entries answered; GB261 is unread. The second queue, on intrapatient escalation in immunology, was assembled on 2026-08-31 from search results because no journal, registry or regulator site was reachable, so every entry is a prediction about a paper. The third, on T-cell engagers in healthy volunteers, was added on 2026-09-10 from a registry search and is in the same state.
The working document is Specification 2, which replaced Specification 1 on 2026-10-07, and the folder’s other documents are on the project index.
Status markers
- ❌ Not checked. No claim in the working document has been verified against this source. Where the working document characterizes it, the characterization comes from an abstract or a search result.
- ⚠️ Transcribed, unverified. A number, table or model structure in the working document came from this source but has not been checked against it.
- ✅ Checked. Verified against the source.
On a reading-queue entry, ✅ means the question posed against the entry has been answered, by reading the paper or by deciding it is out of scope, and the note on the entry says which. Five of the six in the first queue are ✅, and none of the fourteen in the second or the six in the third.
On a source entry below the queues, ✅ means the claim the specification draws from it has been checked against the source. None is ✅ yet. Change a marker when the check is done, not when the paper is opened.
How to read these
Read each paper against the question under its entry, not front to back. The general method is on the Reading Papers page: three passes, stopping at the pass that answers the question. An hour is the budget for a paper read this way.
Simon 1997 needed the full hour, spent as follows.
- Ten minutes on the abstract, the numbered definitions of designs 1 to 4, and the column headers of the simulation results tables. Which outcome metrics they chose to report answers the calendar-time question by itself.
- Forty-five minutes on two targets: the definition of design 1 and the rule for leaving the accelerated phase, which is the comparator question; and the results tables, for whether trial duration appears at all beside patients treated and patients treated at subtherapeutic doses.
- One conditional extra hour. If their dose-toxicity model carries a cumulative or carryover term, that is the carryover-toxicity entry’s question already parameterized. Otherwise do not reconstruct the model, because nothing downstream of it is being reused.
Duration turned out not to be among the reported outcomes, so the catch-up helps patients, not the clock hypothesis in Section 26 stays in the simulation, and Simon is cited for the direction rather than for a number. Entry 5 has the result.
Notes go one line per question, with an answer and the table or page it came from, plus anything that contradicts the specification. Until an entry has been read, its description comes from an abstract and is a prediction about what the paper says.
Reading queue
Read these before building Phase 1, in order; each makes the next easier to read. Entry 6 is still open, and it bears on the stopping rule.
1. Elmeliegy et al. 2024 — TCE dosing strategies and quantitative clinical pharmacology ✅
Clinical Pharmacology & Therapeutics. https://doi.org/10.1002/cpt.3361
Read the section on dose escalation. Talks about using small cohorts (n=1), large steps (3x), and starting around MABEL dosing but didn’t actually talk about intrapatient escalation. I don’t think this was really done for most of the TCEs that are approved.
2. Zhou et al. 2025 — modified MABEL starting dose for a solid-tumor TCE ✅
Clinical Pharmacology & Therapeutics. https://doi.org/10.1002/cpt.3431
Reported to have justified a 10-fold higher first-in-human starting dose and saved at least two escalation cohorts. It attacks the same gap from the other end, moving \(D_{\rm start}\) up rather than crossing the gap faster, so a generally applicable version shrinks \(R_{\rm predicted-gap}\) and the value of IPDE with it. The two strategies can be combined, and in the simulation a higher starting dose is a change to one parameter.
Decision not to read Decided not to read. This is a completely different topic. Related in that MABEL-based doses are a good approach and this IPDE can complement it.
3. Guo et al. 2025 — IP-CRM, intra-patient escalation inside a CRM ✅
Pharmaceutical Statistics. https://doi.org/10.1002/pst.2461
Places intrapatient escalation inside a continual reassessment method (CRM), with adaptively updated starting doses: the current successor to accelerated titration.
Answer while reading. I read the introduction, I think this method is more complex and not what we’re doing in immunology studies (CRM). I mean maybe that’s worth considering, but that would be a future step. This is future work, and it now sits in Section 27b of the specification.
4. Falke et al. 2025 — carryover toxicity with intra-patient dose escalation ✅
Statistics in Medicine. doi:10.1002/sim.70059 · PMC12098223 · PubMed 40405047. Journal and authors identified from a search rather than from the paper.
Treats the confound in Section 16 methodologically: how to attribute an event in a patient who has already received every lower dose on the ladder.
Answer while reading. I decided to defer this till later. I’m not trying to come up with a quantitative safety model here.
5. Simon et al. 1997 — accelerated titration designs ✅
J Natl Cancer Inst 89(15):1138–1147. https://doi.org/10.1093/jnci/89.15.1138
The origin of the mechanism. Designs 2 to 4 combine single-patient cohorts in the low-dose region with intrapatient escalation, and the paper defines what an efficient conventional comparator looks like. Read last of the five.
Methods
Definitions: - Conventional Design = Fibonacci Method. dose increments are 100, 67, 50, 40%, and uses a 3+3-like approach - Accelerated = 1 patient cohorts when safe (instead of 3) until DLT or 2 G2 tox - Intrapatient Escalation = If tox is G0-1, then next dose can increase 100%
Four models
- Single Steps + Conventional Design
- Single Steps + Accelerated until AE in first cycle, then Conventional
- Double Steps + Accelerated until AE in first cycle, then Conventional
- Double Steps + Accelerated until AE in any cycle, then Conventional
A. No intrapatient escalation B. With intrapatient escalation.
Results
Full numbers with the page each came from are in the reading notes. The short version:
Fewer patients does not mean a shorter trial. Cohorts are tabulated separately from patients as a duration proxy (Fig. 3, p.1143). Model 2 enrols 15 fewer patients than model 1 and saves no cohorts; model 1 needs slightly fewer, since model 2 overshoots and spends cohorts de-escalating (p.1142). The saving appears only with double steps (p.1142).
| Model | Patients, mean | Patients, median | Cohorts vs model 1 |
|---|---|---|---|
| 1 | 39.9 | 36.7 | baseline |
| 2 | 24.4 | 21.8 | no advantage |
| 3 | 20.7 | 19.3 | substantial saving |
| 4 | 21.2 | 19.1 | substantial saving |
Intrapatient escalation (B against A) reduces undertreated patients and changes nothing else measured.
| Metric | 3A, no IPDE | 3B, with IPDE | Source |
|---|---|---|---|
| Patients, mean | not reported | 20.7 | 3B p.1141 |
| Cohorts | not reported | Fig. 3 only | p.1143 |
| Worst toxicity grade 0–1 | 6.5 | 3.9 | 3A p.1144, 3B p.1142 |
| Grade 3 | 5.7 | 6.8 | 3A p.1143, 3B p.1142 |
| Grade 4 | 3.2 | 4.3 | 3A p.1143, 3B p.1142 |
On 1A against 1B, where no acceleration is involved, intrapatient escalation has “no effect on the number of patients or number of cohorts” and moves undertreated patients from 23.3 to 19.3 (p.1143).
No figure or table splits A from B. On patients and cohorts the comparison is the phrase “little or no effect” (p.1143), so the catch-up helps patients, not the clock hypothesis in Section 26 is supported in direction, with no published number to cite in place of the simulation. The toxicity grades are the only numbers given for both A and B.
The comparator is 3A. The finding that carries over is that time is bought with dose step size rather than with who escalates.
6. Song et al. 2023 — GB261 first-in-human dose escalation ❌
Blood 2023;142(Supplement 1):1719. Trial NCT04923048. Abstract · Poster
Song Y, et al. GB261, an Fc-Function Enabled and CD3 Affinity De-Tuned CD20/CD3 Bispecific Antibody, Demonstrated a Highly Advantageous Safety/Efficacy Balance in an Ongoing First-in-Human Dose-Escalation Study in Patients with Relapsed/Refractory Non-Hodgkin Lymphoma.
Read for the intrapatient stopping rule. The escalation ran accelerated titration followed by 3+3, across flat and step-up regimens at doses from 1 mg to 300 mg in 47 patients with relapsed or refractory B-cell non-Hodgkin lymphoma (B-NHL): the shape of the scout design, in the same target class.
The question to answer from the source. What rule ended the accelerated phase and started the 3+3. Whether that rule was a toxicity event in the Simon sense, a pharmacodynamic threshold, or both, and whether the patient who triggered it stopped escalating alone or every patient stopped. The abstract states the two phases and the dose range without stating the rule, so the answer is in the poster or the protocol.
A second reason to read it. CD20+ cells were undetectable 24 hours after infusion across the dose range studied. Where that holds, blood B-cell count does not separate dose levels for this class, and a depletion threshold can serve as a stopping trigger but not as a dose-ranging endpoint. Section 10 of the specification calibrates \(EC_{50}\) so that the threshold does separate adjacent dose levels, and GB261 tests that assumption.
Neither host is reachable from the container these notes were written in, so both links need a browser.
Reading queue: intrapatient escalation in immunology
Searched 2026-08-31, oncology excluded. Immunology does escalate within patients, and three of the studies found escalate on a biomarker rather than on toxicity. In the autoimmune cytopenias, dose finding by intrapatient escalation is established practice: rilzabrutinib’s phase 1–2 in immune thrombocytopenia is an adaptive dose-finding study whose escalation mechanism is within-patient, and its sibling study in pemphigus vulgaris gates the step on target engagement. Antigen-specific immunotherapy supplies a second family, where the within-patient ladder is a fixed titration and one first-in-human study is registered as a within-subject dose-escalating design.
Nothing in this queue has been read. The container could not reach clinicaltrials.gov, PubMed, PMC, Nature, Wiley or ema.europa.eu, so every entry is a characterization from a search result and carries ❌. Several quote a search summary rather than the paper or the registry record, and those are marked where they appear.
The direction of the ladder is the difference
| Study | Indication | Ladder | What governs the step |
|---|---|---|---|
| Rilzabrutinib phase 1–2 | Immune thrombocytopenia | 200 mg QD to 400 mg BID over 24 weeks | Platelet response, reviewed every 28 days |
| PRN1008 phase 2 | Pemphigus vulgaris | 400 mg QD to 400 BID to 600 BID | Clinical response and BTK occupancy |
| BMS-986004 | Immune thrombocytopenia | One step, at day 50 | A platelet threshold not reached |
| ATX-GD-59 first-in-human | Graves’ hyperthyroidism | 25, 50, 100, 400, 800 µg, 10 doses at 2-week intervals | Prespecified titration |
| ATX-MS-1467 | Relapsing multiple sclerosis | The same five doses at 14-day intervals | Prespecified titration |
| Nexvax2 phase 1 | Celiac disease | 3 µg to 900 µg over nine weeks | Prespecified, amended after an adverse event |
Every biomarker-gated study in the table escalates until the patient responds, and stops climbing when the marker moves the right way. Scout-and-backfill also stops when the marker moves, but there the response signals that the programme has found its dose. The mechanism is the same with the purpose reversed, which makes the rilzabrutinib and PRN1008 escalation rules the closest published relatives of the Section 13 rule. Read them before that rule is finalized.
The search did not find the combination the specification proposes: a scout whose ladder exists to nominate a dose for the programme, with a separate cohort confirming it. The stopping rule is still what is new. The prior art is not oncology’s alone.
Where to start
Entries 7 and 8 first, since they are the two escalation rules to read the Section 13 rule against. Entry 12 next, because it is the only published measurement anywhere of how much a within-patient ladder raises the tolerated dose, which is the Section 3 confound with a number on it. Then 13 and 17, which decide what the T-cell-engager field already does. Entry 20 is the objection the design has to survive.
7. Kuter et al. 2022 — rilzabrutinib in immune thrombocytopenia ❌
New England Journal of Medicine 386:1421–1431. https://doi.org/10.1056/NEJMoa2110297 · NCT03395210
An international, adaptive, open-label, dose-finding phase 1–2 study in 60 previously treated patients, reported as using intrapatient dose escalation with a 3+3, over 24 weeks. Starting doses 200 mg once daily, 400 mg once daily, 300 mg twice daily and 400 mg twice daily. Escalation was permitted every 28 days on the investigator’s judgement to improve response, with a data and safety monitoring committee reviewing safety and efficacy before any patient escalated. The primary endpoints were safety and platelet response, defined as two consecutive counts at or above 50×10⁹/L with a rise of at least 20×10⁹/L from baseline.
The nearest published relative of this project’s design: an autoimmune disease, a dose-finding study, escalation inside the patient, and a blood-count threshold deciding each step.
The question. What the escalation rule was. “Investigator’s judgement” and a 3+3 are two different mechanisms and the paper reports both, so find out which one moved a patient up. Whether the dose-response conclusion used the within-patient data or only the starting doses. Whether the 28-day cycle was set by the platelet turnover time, which is the \(W_{\rm decision}\) question in Section 15 answered from the biology of the marker rather than from a safety committee’s habit.
8. PRN1008 in pemphigus vulgaris ❌
NCT02704429, phase 2 open-label, the same molecule as entry 7.
Reported as a starting dose of 400 mg twice daily with intra-patient dose escalation to a maximum of 600 mg twice daily, based on clinical response and BTK occupancy. Part B is reported as starting at 400 mg once daily, escalating to 400 mg twice daily at or after week 3 for insufficient clinical response, and again to 600 mg twice daily at or after week 5.
This is the only study found in which a target-engagement measurement gates a within-patient dose step. Bruton tyrosine kinase occupancy plays the part that peripheral B-cell count plays in Section 13.
The question. How occupancy entered the rule: a numeric threshold, a cohort-level check, or a supporting observation beside the clinical response. What happened when occupancy was high and response was absent, which is the case that decides whether the marker can stop a ladder on its own. Section 13 assumes it can.
9. BMS-986004 in immune thrombocytopenia ❌
Anti-CD40L domain antibody. NCT02273960
A single conditional within-patient step, reported as taken after the response phase. A subject receives the next higher dose if a clinically significant response has not been achieved on the last two consecutive platelet counts at days 43 and 50, where the response is a count at or above 50,000/mm³ with a rise of at least 20,000/mm³ from baseline and no bleeding, and if safety is favourable. The subject then stays at the higher dose for the rest of the study.
The rule is numeric, prespecified and published in the protocol, which is what Section 13’s rule needs to be. Copy its shape.
The question. Why one step rather than a ladder, and what the protocol says about a subject who responds at the lower dose. Also whether the two consecutive counts requirement was there to defend against the false brake described in Section 13, which would be a precedent for the confirmatory-sample question that section leaves open.
10. ATX-GD-59 in Graves’ hyperthyroidism ❌
Thyrotropin-receptor peptide immunotherapy. Pearce et al., phase 1. NCT02973802 · PMC6648194
Registered as a phase 1, first-in-human, open-label within-subject dose-escalating study in about 12 subjects. Upward titration over five dose levels of 25, 50, 100, 400 and 800 µg, then five further injections at the top dose, ten injections in all at two-week intervals. Ten subjects received all ten doses and five had free triiodothyronine inside the reference interval by week 18.
A first-in-human study in an autoimmune disease whose escalation design is within-subject rather than between-cohort. The specification should stop describing that as an oncology practice.
The question. What the first-in-human rationale was, given that the whole dose range is traversed inside every subject including the first. What the starting dose was set from, and whether a subject could stop early. The design has no scouting purpose, so the entry decides how far the precedent reaches.
11. ATX-MS-1467 in relapsing multiple sclerosis ❌
Myelin basic protein peptide immunotherapy. Neurology 2018 · NCT01973491
The same ladder as entry 10 in a different disease: 25, 50, 100, 400 and 800 µg at 14-day intervals over 8 weeks, then four further 800 µg doses, in 43 participants across intradermal and subcutaneous cohorts. An earlier first-in-human study in six people with secondary progressive disease used six injections on the same ascending scheme. Slow titration with a longer full-dose period is reported to be associated with the reduction in gadolinium-enhancing lesions.
The question. Whether the titration exists for tolerance induction, so that the ladder is part of the mechanism, or for safety, so that it is a step-up regimen. If the former, this family is not a precedent for scouting at all and should be cited only for the fact that within-patient escalation is routine in immunology.
12. Goel et al. 2017 — Nexvax2 in celiac disease ❌
EBioMedicine. PubMed 29191561 · ScienceDirect · NCT02528799
Randomized, double-blind, placebo-controlled phase 1 testing whether stepwise dose escalation attenuates the first-dose effect. Fixed dosing without escalation is reported to limit the maximum tolerated dose to 150 µg; with escalation, cohort 3 received 3, 9, 30, 60, 90, 150, 300, 450, 600, 750 and then eight doses of 900 µg over nine weeks. The escalation was amended mid-study to add 3 µg and 9 µg steps after a participant withdrew with gastrointestinal events at 30 and 60 µg.
Read this one for Section 3. The specification’s central caution is that a patient who reached a dose up a ladder differs from a patient who received it directly, and this is the only study found that measures the size of that difference: a sixfold rise in the tolerated dose, in an autoimmune indication, with the ladder as the only change. Section 14 makes the same argument for cytokine release and calls the net effect ambiguous.
The question. Whether the 150 µg and 900 µg figures are tolerability limits in the same sense, and what the escalation bought at the doses both regimens reached. If a ladder raises the tolerated dose sixfold, the backfill non-confirmation Section 7 expects is real, and the confirmation penalty is larger than the specification assumes.
13. de Graaf et al. 2025 — FIH dose selection for TCE bispecifics ❌
Clinical Pharmacology & Therapeutics. https://doi.org/10.1002/cpt.3487 · PubMed 39690707
Also listed below under “Related, not yet placed”. A search summary of this paper returns the sentence that intra-patient dose escalation may be permitted to maximize collection of information at relevant doses and minimize patient exposure to sub-optimal doses, with individual patient doses increased to the highest cleared dose level tolerated by completed cohorts. That is Design B, stated as acceptable practice. The sentence comes from a search engine and the same wording appears in oncology protocols, so it may not be de Graaf’s.
The question. Whether that sentence, or anything like it, is in the paper. If it is, Section 6 cites it for catch-up escalation instead of resting the case on Simon 1997. Also whether the paper treats non-oncology indications.
14. Blinatumomab in refractory systemic sclerosis ❌
Case series, five patients. PubMed 41792022 · ScienceDirect
Reported as a 14-day continuous infusion at 9 µg/day for 7 days escalated to 28 µg/day for 7 days. A separate report of the same drug in the same indication describes four cycles at 9, 9, 28 and 28 µg/day. A 3-fold within-patient increase in a CD19 T-cell engager in an autoimmune population, at one rung.
The question. Whether the step to 28 µg/day was scheduled or conditional on what happened at 9 µg/day, and what would have held a patient at 9.
15. Bucci et al. 2024 — blinatumomab in refractory rheumatoid arthritis ❌
Nature Medicine. https://doi.org/10.1038/s41591-024-02964-1 · PubMed 38671240
Six patients under compassionate use, 2 to 3 cycles of 96-hour continuous infusion, cumulative doses reported elsewhere as 77 to 189 µg. B cells depleted and disease activity fell in all six. The spread of cumulative doses suggests the dose differed between patients or between cycles.
The question. Whether a patient’s dose rose across cycles, and on what rule.
16. Bucci et al. 2025 — BCMA T-cell engager in refractory autoimmune disease ❌
New England Journal of Medicine. https://www.nejm.org/doi/full/10.1056/NEJMc2506740
Ten patients across systemic sclerosis, myositis, IgG4-related disease, Sjögren’s, Graves’ disease and rheumatoid arthritis. Teclistamab at 0.06, 0.3 and 1.5 mg/kg on days 1, 3 and 5, then a single 1.5 mg/kg dose four weeks later. Eight of ten in drug-free remission at a median six months.
The question. Whether the step-up was compressed against the myeloma label, which places these doses on days 1, 4 and 8, and what justified it. Section 15 sets \(W_{\rm decision}\) at 7 days partly on the argument that approved products escalate within a patient every 2 to 4 days. If an autoimmune population tolerates a 2-day interval, the weekly scout ladder is more conservative than Section 15 claims.
17. CLN-978 first-in-human, systemic lupus and Sjögren’s disease ❌
CD19xCD3 T-cell engager, subcutaneous. NCT06613360 (SLE) · NCT07041099 (Sjögren’s)
Reported as a modified single-ascending-dose escalation over target doses of 10, 20, 30 and 45 µg with at least three patients per cohort, where every cohort above 10 µg receives a 10 µg step-up on day 1 and the cohort target on day 8. Deep, dose-dependent B-cell depletion is reported, with B cells unquantifiable in half the patients dosed at 20 µg or above.
Design A with one step-up, running now, in the population Section 22 calls the immune-reset archetype. It is the comparator the simulation argues against, and a realistic one.
The question. Three. The between-cohort review window, which is the largest single driver of the conventional timeline in Section 15. Whether the protocol lets an enrolled patient move up to a cleared dose level, which is Design B. And what the fixed 10 µg step-up implies for Section 7, which assumes a step-up proportional to the target rather than fixed across cohorts. MCLA-117 fixes its prefix from cohort 9, so two programmes now disagree with the base-case rule.
18. Teplizumab, the 14-day ascending regimen ❌
Anti-CD3 in type 1 diabetes. Reported as 51, 103, 206 and 413 µg/m² on days 1 to 4, then 826 µg/m² on days 5 to 14, cumulative 9,034 µg/m².
The oldest within-patient ladder in immunology: about 16-fold over four days, fixed rather than conditional. Its 2-fold daily steps sit close to Simon’s double steps and far below the \(F_{\rm quiet}=10\) of Section 13.
The question. What set the ramp, and whether the cytokine-release argument for it is the priming mechanism Section 14 invokes.
19. Individualized build-up: venom, aeroallergen and oral immunotherapy ❌
Rush and cluster build-up protocols compared · AR101 peanut oral immunotherapy, PALISADE
Where immunology escalates within a patient to that patient’s own limit rather than to a programme dose. Rush venom immunotherapy is reported as 0.02 µg to a 100 µg maintenance dose in 15 increments over three days; AR101’s initial escalation runs 0.5 mg to 6 mg of peanut protein in one day at 20 to 30-minute intervals, and a patient not tolerating 3 mg leaves the study. A systemic reaction sends the patient back down the ladder rather than stopping the programme.
The question. How the field handles the confound in Section 16, attributing a reaction in a patient who has already received every lower dose. Also what a stopping-dose choice buys: rush protocols are reported to have been compared at stopping doses of 1%, 4% and 10% of maintenance, which is the Section 13 brake question in another vocabulary.
20. EMA 2017 first-in-human guideline ❌
Guideline on strategies to identify and mitigate risks for first-in-human and early clinical trials (EMEA/CHMP/SWP/28367/07 Rev. 1). Host unreachable from the container.
A search summary reports that including the same subject across multiple cohorts is possible but must be scientifically justified in the protocol, and that re-enrolment into a higher dose cohort requires an appropriately defined washout period. It also reports that a dose level may not be repeated once a stopping rule has been met.
Read as written, a washout requirement is incompatible with a weekly ladder in a molecule whose effective half-life is about 8 days (Section 7). Entry 10 is a registered first-in-human study that escalates within the subject with no washout, so either the constraint does not reach a patient population or it is read more loosely than the summary suggests.
The question. What the guideline says, whether the passage covers patient trials or only healthy volunteers, and how ATX-GD-59 was authorized against it.
What was searched and did not turn up
Recorded so the search is not repeated. No study was found in which within-patient escalation exists to nominate a dose for the programme rather than to treat the patient climbing, and none in which a separate cohort then confirms an interpretable regimen. No accelerated titration design was found in an autoimmune indication. Registry search was not possible, so the negative covers the published literature and search results, not protocols.
Cat-PAD is a deliberate counterexample: the peptide was designed to be given at therapeutic doses from the first administration, to avoid the dose escalation conventional immunotherapy requires. Where the first-dose reaction can be engineered out, the ladder disappears.
The methodological papers on intrapatient escalation are motivated by oncology or by rare disease rather than by immunology: Guo et al. 2025 in the first queue, AIDE (Zhou et al. 2023, Pharmaceutical Statistics, https://doi.org/10.1002/pst.2272), the patient-retreat PRIDE scheme for rare diseases, and TWICEBEE for multi-cycle immunotherapy. They belong with Section 27b and none is a v1 dependency.
Reading queue: T-cell engagers in healthy volunteers
Searched 2026-09-10 for T-cell engagers in healthy volunteers. Three registered volunteer studies, all autoimmune, all since December 2025, none with a result posted. Every entry is a registry record, a meeting abstract or a press release, and every entry carries ❌. Europe PMC returned no published report of a CD3-engaging bispecific dosed in healthy subjects. The Chinese registries, the EU register and ANZCTR were not searched.
21. NCT07236411 — cizutamig absolute bioavailability in healthy adults ❌
Candid Therapeutics. https://clinicaltrials.gov/study/NCT07236411
A randomized, double-blind, single-dose study, intravenous or subcutaneous against placebo, in 24 healthy adults at Nucleus Network, Melbourne. Started 2025-12-04, completed 2026-07-14. Primary outcome absolute bioavailability through eight weeks; secondary outcomes safety, PK and anti-drug antibodies. Cizutamig is EMB-06, a 2+2 BCMA x CD3 engager on EpimAb’s FIT-Ig platform, licensed outside Greater China to Vignette Bio and then Candid.
The question. What single dose the volunteers received, relative to the 0.2 to 300 mg weekly range already given to myeloma patients (entry 24), and what their B cells, free light chains and cytokines did. None is a registered outcome, so the answer is in a presentation not yet found. Free light chains are the readout Specification 2 would use for a BCMA engager.
22. NCT07284797 — XmAb657 first-in-human in healthy participants and autoimmune patients ❌
Xencor. https://clinicaltrials.gov/study/NCT07284797
An open-label subcutaneous dose escalation, 60 participants estimated, Melbourne, started 2026-02-24, primary completion October 2028. Healthy adults are the first cohort, followed by idiopathic inflammatory myopathy, systemic sclerosis and Sjögren’s disease. Primary outcome adverse events through week 52; secondary serum concentration. The sponsor has said it will report progress in the second half of 2026.
The question. Whether the volunteer cohort escalates to a B-cell-depleting dose or stops at a dose chosen to be subactive. This is the study whose design matches the idea in Section 1 of the volunteer note, and the one whose readout decides whether the idea works.
23. NCT07712939 — CND319 first-in-human, SAD in volunteers and MAD in patients ❌
Candid Therapeutics. https://clinicaltrials.gov/study/NCT07712939
Open-label, 105 participants estimated, Melbourne, started 2026-07-20, primary completion March 2028. Part 1 single ascending dose in healthy adults aged 18 to 65; multiple ascending dose and Part 2 expansion in rheumatoid arthritis and Sjögren’s. CND319 is a CD19/CD20 x CD3 engager. Primary outcomes safety through week 12 (SAD) or 24; secondary PK and anti-drug antibodies.
The question. The same as entry 22, and whether the patient MAD starting dose was chosen from the volunteer SAD, which would be the idea applied.
24. Grieshaber-Bouyer et al., ACR Convergence 2025, abstract 2641 — cizutamig, preclinical to clinical translation ❌
Forty relapsed or refractory myeloma patients at 0.2 to 300 mg weekly, cytokine release grade 1 in 18% and grade 2 in 8%, median onset one day, no ICANS; one refractory rheumatoid arthritis patient with near-complete plasma cell depletion in marrow and lymph node at three weeks and no cytokine release. The patient dose range that the volunteer study in entry 21 was run against.
The question. At what dose the first fever appeared in patients, for comparison with whatever the volunteers show.
25. Bernett et al., ACR Convergence 2025, abstract 0010 — XmAb657 in cynomolgus monkeys ❌
A single intravenous or subcutaneous dose suppressed B cells below 5 cells/µL for at least 42 days, with marked depletion in lymph node and marrow, no clinical signs of cytokine release, and a half-life near 15 days. The primate prediction that the volunteer cohort of entry 22 tests.
The question. The dose that did it, in mg/kg, since that is the number the volunteer starting dose was derived from.
26. Suntharalingam et al. 2006 — cytokine storm in the TGN1412 phase 1 trial ❌
New England Journal of Medicine 355(10):1018–1028. https://doi.org/10.1056/NEJMoa063842 · PMID 16908486
Six healthy volunteers, a CD28 superagonist, a dose 500 times below the primate no-effect level, and multi-organ failure within hours. It is why an ethics committee scrutinizes a T-cell agonist in healthy volunteers hardest, and the case behind entry 20’s guideline.
The question. What the paper and the subsequent guidance say distinguishes a CD28 superagonist from a target-dependent CD3 engager with a tuned arm, since that distinction is what the three sponsors in entries 21 to 23 argued.
Not found, and where it was not looked for
No healthy volunteer study of CM336 or OM336 (Keymed; Ouro Medicines, acquired by Gilead in 2026) is registered on ClinicalTrials.gov. The record NCT06745687, which carries the healthy-volunteer flag, is a study in high-risk smouldering myeloma and the flag is an error. The Chinese Center for Drug Evaluation registry and the Chinese Clinical Trial Registry were not searched, and a colleague’s recollection of a Chinese volunteer study of the Ouro molecule is unresolved until they are, or until the colleague supplies an identifier.
Sources the specification’s numbers come from
Retrieve and verify these rather than relying on the transcription in the working document.
| Status | Source | Taken from it | Used in |
|---|---|---|---|
| ⚠️ | Teclistamab population PK model | SC absorption, two-compartment disposition, time-dependent clearance, and the typical parameter estimates | Section 8 |
| ⚠️ | Friberg semimechanistic myelosuppression model | Model structure, \(MTT \approx 116\)–\(125\) h, \(\gamma \approx 0.17\) | Section 17 |
| ❌ | Mosunetuzumab translational TCE modeling | Precedent for T-cell-mediated B-cell killing as a PD output | Section 9 |
| ❌ | Clinical pharmacology and dose optimization of T-cell engagers | Historical FIH escalation, step-up dosing, safety considerations | Sections 6, 14 |
| ❌ | Teclistamab ICANS clinical experience | Incidence and onset distribution claims | Section 16 |
Sources behind Specification 2
The FLC numbers and claims in Specification 2. Each was taken from a search result or a secondary summary on 2026-10-07, and none has been read at source.
| Status | Source | Taken from it | Used in |
|---|---|---|---|
| ⚠️ | Katzmann et al. 2002, Clin Chem, FLC reference intervals | Kappa 3.3–19.4 mg/L, lambda 5.7–26.3 mg/L, ratio 0.26–1.65 | Sections 4, 11, 22 |
| ⚠️ | Serum FLC half-life; primary source not located. Quoted in a high-cutoff haemodialysis kinetics paper among others | Kappa 2–4 h, lambda 3–6 h, renal clearance, longer in renal failure | Section 9 |
| ⚠️ | Kumar et al. 2016, IMWG consensus response criteria, Lancet Oncology; not retrieved | Measurable disease by FLC: involved FLC ≥ 100 mg/L with an abnormal ratio | Section 22 |
| ❌ | Teclistamab sBCMA and exposure, ASCO abstract | Baseline sBCMA not a significant covariate on teclistamab exposure | Section 8 |
| ❌ | sBCMA during teclistamab and talquetamab, EHA 2021 | sBCMA falls in responders | Section 9 |
| ❌ | Early FLC reduction predicts outcome in intact-immunoglobulin myeloma | Precedent for an early FLC fall as a readout of plasma-cell kill | Section 9 |
Not yet looked for: the within-person variability of serial FLC measurements, which sets whether \(\theta_{\rm move}\) needs a confirmatory sample (Section 13); the disagreement between FLC assays from different manufacturers (Section 11); and any estimate of plasma-cell recovery after a BCMA engager, which sets \(t_{1/2,P}\) (Section 9).
The MCLA-117 worked example
The Cycle 1 escalation scheme in Section 6 is transcribed from an image of the poster’s dose table, not from the poster itself. Every number in those two tables and in the figure needs checking before it is cited. The study is MCLA-117 (tepoditamab), a CLEC12A x CD3 bispecific, in relapsed or refractory AML.
| Status | Source | Link |
|---|---|---|
| ⚠️ | EHA 2020 e-poster, MCLA-117 CL01 | https://merus.nl/app/uploads/2020/06/EHA-2020-MCLA-117-CL01-e-poster.pdf |
| ❌ | EHA library record | https://library.ehaweb.org/eha/2020/eha25th/294456/ |
| ❌ | Trial registration, NCT03038230 | https://clinicaltrials.gov/study/NCT03038230 |
The study protocol has not been located. Registered protocol documents, where posted, appear on the registration page above.