Continuous endpoints in Bayesian two-stage designs

ABSTRACT

Single-arm two-stage designs for phase II of clinical trials typically focus on a binary endpoint obtained by dichotomizing an underlying continuous measure of treatment efficacy. To avoid the resulting loss of information, we propose a two-stage design based on a Bayesian predictive approach that directly uses the original continuous endpoint. Numerical results are provided with reference to phase II cancer trials aimed at assessing tumor shrinking effect of an experimental treatment.

KEYWORDS:

• Analysis prior
• Design prior
• Sample size determination
• Skew normal distribution

Acknowledgments

We would like to thank Ludovico Piccinato for his valuable comments to the previous version of this article. We are also grateful to the Associate Editor and the anonymous reviewers for their helpful suggestions.