Model Based TDP

Model-based assessment of a biomarker response criterion

Comparing an observed 9-of-10 responder rule with model-based predictions across doses, for a criterion of at least 80% biomarker decline in 90% of patients.

Can a model fitted across three dose cohorts of ten patients make a better dose decision than requiring nine observed responders in a cohort, where response means at least an 80% biomarker decline?

Published methods support retaining continuous biomarker measurements and integrating information across patients and doses. The closest decision framework is probability of pharmacological success (PoPS). The size of the benefit for this particular criterion still depends on the response distribution, the model assumptions and the decision’s required certainty.

Documents

• Working specification. The response definition, decisions, analysis comparisons and simulation program for an integrated pharmacometric approach.
• Literature review. Which published approaches answer this question, what their comparisons establish, and where the evidence stops.
• Worked comparison. Exact calculations for the 9-of-10 rule and a reproducible simulation separating continuous-data gains from pooling across doses.
• References. Ranked reading queue, search scope and source-by-source verification notes.

The initial calculations use a single biomarker assessment. Longitudinal and pharmacokinetic/pharmacodynamic comparisons are specified for the next stage. The activity does not establish a clinical efficacy or safety threshold.

Related Activities

• Sizing studies: precision and sample size for dose-response characterization.
• Logistic regression for dose finding: fitting binary dose-response relationships.
• ITP-PK-Platelet: a disease-specific investigation of the added value of integrated modeling.