Related Experiment Videos
An optimal three-stage design for phase II clinical trials
L G Ensign1, E A Gehan, D S Kamen
1Department of Biomathematics, University of Texas M.D. Anderson Cancer Center, Houston 77030.
Statistics in Medicine
|September 15, 1994
Summary
This study introduces a three-stage clinical trial design for cancer therapeutics, optimizing sample size and early stopping rules for efficacy testing. The new design improves efficiency, especially for treatments with low response rates.
Area of Science:
- Biostatistics
- Clinical Trial Design
- Cancer Therapeutics
Background:
- Phase II cancer clinical trials typically use single-arm studies to assess experimental treatment (E) promise.
- Binary endpoints (response/no response) with response probability (p) are common efficacy criteria.
Purpose of the Study:
- Propose a three-stage optimal design for testing hypotheses H0: p ≤ p0 versus H1: p ≥ p1.
- Determine if an experimental treatment merits further testing based on efficacy.
Main Methods:
- Combines earlier proposals by Gehan and Simon for a three-stage design.
- Allows early rejection of H1 (treatment inefficacy) at any stage based on treatment failures.
- Acceptance of H1 (treatment efficacy) is only possible at the final stage.
Main Results:
- The design minimizes expected sample size when p = p0, with a minimum stage 1 sample size of 5.
- Offers greatest utility for treatments with small true response rates, enabling early stopping for failures.
- Compared to Simon's two-stage design, it allows earlier stopping with 0 successes and has a smaller expected sample size under H0.
Conclusions:
- The proposed three-stage design enhances early stopping capabilities and optimizes sample size in phase II cancer trials.
- It is particularly beneficial when early identification of non-promising treatments is crucial.
- The design provides a statistically sound and efficient alternative to existing two-stage methods.