References for Model Based TDP

Reading order, search scope and checked source claims

Annotated sources for model-based biomarker response decisions, with explicit distinctions between checked full-text passages, abstract-only evidence and outstanding reading.

The literature review gives the synthesis. The specification defines the comparison, and the worked example contains independently implemented calculations.

Status Markers

• ❌ Not checked. The proposed claim or methodological detail has not been verified against the source.
• ⚠️ Transcribed, unverified. A numerical result or model detail has been copied but not checked at its source location.
• ✅ Checked. The specific claim recorded below was verified against the stated source material. The access field distinguishes full-text passages, indexed full text and abstracts. This does not mean that the study was independently reproduced.

Reading-queue markers refer to the question posed there. Source markers refer only to the narrower claim recorded in that source entry. An abstract can verify a paper’s stated purpose while leaving its implementation audit open. No paper’s simulations were independently reproduced in this review.

How to Read These

Start with the population target and the decision being evaluated. For each comparison, check which data the comparator discards, which assumptions the model adds, and whether the result concerns precision, hypothesis-test power or dose-selection accuracy. A sample size ratio transfers only if those features transfer.

Reading Queue

1. Zhou 2020; Chen 2023 — does the proposed criterion already have a name? ✅ PoPS is the closest match. Read the definition and simulation construction first. The framework question is answered.
2. Suissa 1991; Peacock 2012 — how does continuous estimation recover a responder probability? ❌ The abstracts establish the prior art. Obtain and audit the full statistical derivations before reusing their published variance formulas. The worked example uses its own derivation.
3. Wason and Seaman 2013; McMenamin et al. 2018 — what happens with composite endpoints and small samples? ✅ The relevant methods, results and discussion passages were checked. Implementation of a composite endpoint here remains future work.
4. Karlsson 2013; Zandvliet 2010 — what supports an added PMX benefit? ✅ The comparators and outcome measures establish useful but indirect precedents. Reproducing their simulation code is outside this review.
5. Pinheiro 2014 — which shape-uncertainty method suits three dose levels? ❌ Scope was checked in the abstract; the implementation and small sample behavior need a separate assessment before use.

Search Scope

Search performed on 22 September 2026 using web-indexed PubMed/PMC records, publisher pages and author or institutional repositories, with backward citation searching from the responder-analysis and PoPS papers. Searches covered:

• target decision profile, target dose profile, TDP pharmacometrics, biomarker inhibition and model-based dose selection.
• probability of pharmacological success, population coverage and probability of target attainment.
• binary methods for continuous outcomes, distributional approach, augmented binary, responder endpoints and small-sample corrections.
• Pharmacometric proof-of-concept power comparisons, model-based dose selection and dose-response model uncertainty.
• Recent work on the cost of dichotomization, including the 2026 paper below.

Included primary methods papers and author-reported applications that preserve a response probability, evaluate dose decisions, or compare the information used by analyses. Included a methods overview to trace prior art. General endorsements of modeling, individual therapeutic drug monitoring and unrelated biomarker-selection designs were not used as evidence for this criterion.

This is a focused review rather than a systematic database review: there was no exported database search, duplicate screening or exhaustive citation count. Some PMC direct requests returned browser checks; publisher versions or indexed source passages were used where indicated. No reviewed source was found to establish the exact three-by-ten, 90%/80% head-to-head comparison. This is a bounded search finding, not evidence that none exists.

Sources

1. Zhou, Graff and Chen 2020 — PoPS Definition ✅

Zhou X, Graff O, Chen C. Quantifying the probability of pharmacological success to inform compound progression decisions. PLOS ONE 15:e0240234. Article and DOI.

Access and location: Full text; Introduction, Materials and methods, Simulation setup and Results. Checked claim: Population pharmacological coverage and knowledge uncertainty are combined in a decision probability. Applicability: Direct conceptual match to a biomarker target in a specified fraction of patients. Limit: Simulation-based development applications, without a repeated-trial comparison against 9/10 responders.

2. Chen et al. 2023 — PoPS Applications ✅

Chen C, Zhou X, Lavezzi SM, Arshad U, Sharma R. Concept and application of the probability of pharmacological success (PoPS) as a decision tool in drug development: a position paper. Journal of Translational Medicine 21:17. Full text.

Access and location: Full text; equations and implementation sections, Cases 1–4. Checked claim: The framework distinguishes population variability from estimation and translational uncertainty. Applicability: Guides the nested simulation. Limit: Anonymized decision cases; the antibiotic case explicitly omitted several uncertainty components, so it should not be cited as an example of propagating all uncertainty.

3. Suissa 1991 — Single Continuous Outcome ✅

Suissa S. Binary methods for continuous outcomes: a parametric alternative. Journal of Clinical Epidemiology 44:241–248. PubMed abstract · DOI.

Access and location: Abstract only. Checked claim: Normal-distribution estimation of threshold probabilities, with a stated heavier-tail sensitivity. Applicability: Earliest directly relevant predecessor found. Limit: Full-text equations and numerical examples remain unchecked.

4. Peacock et al. 2012 — Distributional Approach ✅

Peacock JL, Sauzet O, Ewings SM, Kerry SM. Dichotomising continuous data while retaining statistical power using a distributional approach. Statistics in Medicine 31:3089–3103. Publisher abstract.

Access and location: Publisher abstract; bibliographic details also checked against the authors’ institutional record. Checked claim: Continuous-data modeling can produce proportion differences and intervals under distributional assumptions. Applicability: Terminology and method family. Limit: No numerical saving from this paper is transferred here; the full methods have not been audited.

5. Wason and Seaman 2013 — Augmented Binary Method ✅

Wason JMS, Seaman SR. Using continuous data on tumour measurements to improve inference in phase II cancer studies. Statistics in Medicine 32:4639–4650. Full text · DOI.

Access and location: Full text; Sections 3.4–3.5 and Discussion. Checked claim: Relative performance depends on the failure mechanism; the original implementation carries a sample-size caution. Applicability: Retains a composite response definition while using its component data. Limit: Original parameterization is too elaborate to assume suitable for ten patients per arm.

6. McMenamin, Berglind and Wason 2018 — Small Samples ✅

McMenamin M, Berglind A, Wason JMS. Improving the analysis of composite endpoints in rare disease trials. Orphanet Journal of Rare Diseases 13:81. Full text.

Access and location: Full text; Methods, Results on log-odds and probability scales, Discussion. Checked claim: Small-sample corrections change error control, which differs by effect scale. Applicability: Justifies explicit calibration. Limit: Resampling used 30–80 total patients split between two arms, not three cohorts of ten at a 90% target.

7. McMenamin et al. 2021 — Software and MUSE Example ✅

McMenamin M, Grayling MJ, Berglind A, Wason JMS. Increasing power in the analysis of responder endpoints in rheumatology: a software tutorial. BMC Rheumatology 5:54. Full text.

Access and location: Full text; Results, Sample size determination and Figure 4 description. Checked claim: The reported 50-versus-135 planning comparison uses a specified risk difference, 80% power and one-sided 5% testing. Applicability: Concrete software-supported example. Limit: Disease- and endpoint-specific, without independent reproduction here. The review uses the explicit sample-size calculation rather than reconstructing savings from other prose about interval widths.

8. Karlsson et al. 2013 — Pharmacometric Power Comparisons ✅

Karlsson KE, Vong C, Bergstrand M, Jonsson EN, Karlsson MO. Comparisons of Analysis Methods for Proof-of-Concept Trials. CPT: Pharmacometrics & Systems Pharmacology 2:e23. Source · DOI.

Access and location: Indexed PMC full-text passages; Abstract, Discussion and Methods. Direct page access returned a browser check. Checked claim: The reported gains compare drug-effect tests using different amounts of longitudinal information. Applicability: Supports including a longitudinal comparator. Limit: Correct-model simulations and proof-of-concept testing do not establish population-target attainment.

9. Zandvliet et al. 2010 — Model-Based Dose Selection ✅

Zandvliet AS, Karlsson MO, Schellens JHM, Copalu W, Beijnen JH, Huitema ADR. Two-stage model-based clinical trial design to optimize phase I development of novel anticancer agents. Investigational New Drugs 28:61–75; online 2009. Full text.

Access and location: Publisher full text; Summary, Methods and Results. Checked claim: Post-hoc model-based dose recommendations improved precision in the simulated myelosuppression setting. Applicability: Dose-decision performance as an outcome. Limit: Different endpoint, regimens and trial design; its numerical gain is not applied here.

10. Pinheiro et al. 2014 — Dose-Response Shape Uncertainty ✅

Pinheiro J, Bornkamp B, Glimm E, Bretz F. Model-based dose finding under model uncertainty using general parametric models. Statistics in Medicine 33:1646–1661; online 2013. PubMed · DOI.

Access and location: PubMed and author-preprint abstracts. Checked claim: Generalized MCP-Mod addresses multiple candidate shapes and broader endpoint models. Applicability: Shape uncertainty in M4. Limit: Full methods and finite-sample performance remain unchecked; trend detection alone does not test the population-coverage criterion.

11. van Zwet, Harrell and Senn 2026 — Information Loss ✅

van Zwet EW, Harrell FE Jr, Senn SJ. An Empirical Assessment of the Cost of Dichotomization of the Outcome of Clinical Trials. Statistics in Medicine 45:e70402. Full text.

Access and location: Publisher full text; Sections 3–5; first published 5 February 2026. Checked claim: Normal-distribution calculations recover response probabilities from continuous measurements and quantify information loss. Applicability: Current synthesis and worked example. Limit: The empirical comparison spans different trials and cannot isolate a causal benefit of one analysis method.

12. Wason, McMenamin and Dodd 2020 — Methods Overview ✅

Wason J, McMenamin M, Dodd S. Analysis of responder-based endpoints: improving power through utilising continuous components. Trials 21:427. Full text.

Access and location: Publisher full text and PubMed abstract. Checked claim: Connects distributional and augmented-binary methods across clinical endpoint definitions. Applicability: Citation discovery and orientation. Limit: An overview, not an independent estimate of the proposed design’s operating characteristics.

Further Reading Not Used for Quantitative Claims

• ❌ Lei et al. 2017. Comparison of Dichotomized and Distributional Approaches in Rare Event Clinical Trial Design: a Fixed Bayesian Design. PubMed, DOI. Abstract screened; normal mixtures may inform the refractory-subgroup sensitivity analysis. Full-text comparison not audited.
• ❌ Uryniak et al. 2011. Responder Analyses—A PhRMA Position Paper. DOI. Background position identified in citation searches. Publisher access failed; no quantitative claim here relies on it.
• ❌ Sauzet et al. 2015. Dichotomisation using a distributional approach when the outcome is skewed. PMC. Identified for the nonnormal extension; methods not yet checked.