References for the TCE dose-finding strategy working document

What to read, in what order, and what has actually been checked

The reading queue and source list for the one-page TCE dose-finding strategy, with a status marker on every entry recording whether the claim drawn from it has been verified against the source.

Nothing here has been read, and every marker is ❌. Entry 1 is the paper the whole project is derived from, and what the specification says about it comes from its abstract and from a machine-generated summary of the open-access text, fetched on 2026-09-11. A fetched summary is weaker evidence than an abstract, which is weaker than the paper. Every other entry below is a characterization written from a title or a search result. Treat each as a prediction about what its source says, and correct the entry when the source is read.

The working document is the specification, the artifact itself is the specimen page, 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 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 that entry has been answered, by reading the paper or by deciding it is out of scope, and the note on the entry says which. On a source entry below the queue, ✅ keeps the stricter meaning: the claim the specification draws from it has been checked against the source. None is ✅ yet.

Where to start

Two entries settle the project and the rest are context.

• Entry 1 first, with its supplement, and budget two hours rather than the one-hour default on the Reading Papers page. The deliverable is the framework’s two figures and the supplementary file, not the paper’s conclusion. Section 1 and Section 7 of the specification are written against a summary of this paper and are the work of replacing it.
• Entry 2 next, and it may take an hour. It is the completeness test for Block A of the decision inventory: a summary of what the approved T-cell engagers actually committed to, read to find the decision the inventory is missing.
• Entry 3 after that, for the same test applied to the first-in-human end of Block A, entries A1 to A5.
• Entry 4 when the carry-forward block is being filled for a real program. It is the constraint cited in Section 3 of the specification, and it should be read before that sentence is quoted to anyone.
• Entry 5 is optional, and only if the starting-dose row turns out to be contested on a real program.

Journal domains are frequently blocked from the container these notes are written in, so every link below may need a browser.

Reading queue

1. Sander et al. 2021 — the dose and regimen framework this narrows ❌

CPT: Pharmacometrics & Systems Pharmacology 10(11):1276–1280. doi:10.1002/psp4.12701 · PMC8592517

Sander O, Magnusson B, Ludwig I, et al. A framework to guide dose & regimen strategy for clinical drug development.

The question. Four, and the answers go into the specification rather than into this entry.

1. What is on Figure 1, page 1 of the framework? The specification drops the whole knowledge-collection half on the argument that it is a different exercise done once. Reading the page is how that argument is tested.
2. What is on Figure 2, page 2? The specification’s seven columns are a rewrite of it done without seeing it.
3. What does the supplementary file contain, and is it usable as a template directly?
4. How do teams rate relevance and criticality, on what scale, and does that rating do work the two status axes in Section 1 should be doing instead?

What the specification currently asserts about it, all of which is at risk: that it is two pages; that page 1 collects project and drug characteristics and page 2 condenses critical items, evaluates design options and summarizes the strategy; that teams rate relevance and criticality; and that it does not specify an owner or a revision cadence. The last of those is an absence claimed from a summary, which is the weakest kind of claim on this page.

2. Wang et al. 2026 — clinical pharmacology of TCEs across FDA approvals ❌

Clinical Pharmacology & Therapeutics. doi:10.1002/cpt.70020

Wang et al. Clinical Pharmacology Characterization of Bispecific T-Cell Engagers: A Summary Based on FDA Approvals.

The question. Is Block A of the inventory complete? Every approved T-cell engager committed to a starting dose, a step-up schedule, a premedication regimen, a monitoring requirement, and a rule for restarting after an interruption. A summary across approvals is the cheapest available test of whether the nine rows in Block A are the right nine. A row this paper describes that the inventory does not have is a defect in the inventory.

Second question, for Block C: how many of the approvals changed the dosing interval or added a reduced-intensity phase after the initial approval? That count is the argument for why C1 to C4 belong on the page from the start.

3. Graaf et al. 2025 — first-in-human dose selection for TCEs ❌

Clinical Pharmacology & Therapeutics. doi:10.1002/cpt.3487

Graaf et al. First-in-Human Dose Selection for T-Cell Engaging Bispecific Antibodies.

The question. Rows A1 to A5 are committed at Gate 0 with no human data, which the specification calls the gate with the worst ratio of decisions to evidence. What does this paper say the decision is actually made on, and does it name a dependency between the starting dose and the number of step-up doses that the inventory treats as separate rows?

4. FDA 2024 — Project Optimus dosage guidance ❌

Final guidance, August 2024, finalizing the January 2023 draft. FDA guidance page

Optimizing the Dosage of Human Prescription Drugs and Biological Products for the Treatment of Oncologic Diseases.

The question. Section 3 of the specification states that the guidance expects more than one dosage in a randomized comparison, and puts that on the page as a constraint rather than a decision. Does the guidance say that, and in what terms? What it actually requires, what it recommends, and where it leaves room are three different things, and the specification currently asserts the strongest reading.

5. Zhou et al. 2025 — MABEL-based starting dose for a solid tumour TCE ❌

Clinical Pharmacology & Therapeutics. doi:10.1002/cpt.3431 · PubMed 39285508

Zhou et al. An Innovative Approach to Optimize First-In-Human Minimal Anticipated Biological Effect Level Based Starting Dose in Oncology Trials for Bispecific T-Cell Engagers: Experience from A Solid Tumor Bispecific T-Cell Engager.

The question. One row, A1. Read only if the starting-dose row is contested on a real program, and then for what the alternative to a conventional MABEL is and what it costs.

6. Elmeliegy et al. 2024 — dosing strategies and quantitative clinical pharmacology for TCEs ❌

Clinical Pharmacology & Therapeutics. doi:10.1002/cpt.3361

Elmeliegy et al. Dosing Strategies and Quantitative Clinical Pharmacology for Bispecific T-Cell Engagers Development in Oncology.

The question. Context rather than a decision. Which of the twenty-one inventory rows does quantitative modelling already answer well, and which are decided on judgement because nothing can be modelled? The page is more useful on the second group, and this paper is the cheapest way to tell them apart.

Sources the specification’s claims come from

• Entry 1 is the source of every claim in Sections 1 and 7 about what Sander et al. proposed, and none of it has been checked. ❌
• Entry 4 is the source of the Project Optimus constraint in Section 3. ❌
• Nothing else in the specification is drawn from a source. Blocks A, B and C of the inventory were written from the initial notes and from general knowledge of approved T-cell engager labels, which is exactly the kind of recollection entry 2 exists to check.

Related, not duplicated here

The cytokine release syndrome literature belongs to the sibling project and is not repeated. See TCE-CRS-tolerance references for the epcoritamab repeated time-to-event model and the escalation-cohort sources behind rows A3 to A6, and the TCE-IPDE references for the starting-dose and intrapatient-escalation sources behind rows A1 and B2.

Not yet located

Two sources the specification implies exist and that have not been searched for.

• A published example of a dose strategy document in use. Section 7 argues that a living document with no owner goes stale, and cites nothing. If anyone has published what happened when a team ran the Sander framework on a real program, that is the entry that would settle it.
• A regulatory statement on re-priming after an interruption. Row A9 says every approved label carries such a rule and no program studies it. That is a claim about labels, and it is unchecked.