Sizing studies

How many patients a cohort needs to characterize dose-response

Index of the documents in the sizing-studies project: how many patients a Phase 1/2 expansion cohort needs to characterize a dose-response or biomarker-response relationship, and the references it draws the sample size criteria from.

Sizing a cohort to reject a null hypothesis that two doses are the same is usually the wrong question in early oncology. Sizing Studies works out the more useful one: how many patients are needed to characterize a dose-response or biomarker-response relationship to a stated precision, using Frank Harrell’s rule of thumb and Richard Riley’s four-question framework, worked against both a binary and a continuous endpoint.

Documents

• Sizing Studies for Dose-Response Characterization. The four sizing approaches, worked binary and continuous examples with real code, and why a continuous endpoint is usually the practical choice in oncology.
• References. The Riley and Harrell papers the sizing criteria come from, and a marker on each recording whether the page’s use of it has been checked against the source.

Data

• emax_tumor_size.rds, in data/synthetic/dose-response. 61 patients: average daily dose before the first tumor assessment, and percent change in tumor size at that assessment. Used in Section 4’s Emax fit.