Bayesian Sample Size Calculations for a Non-Inferiority Test of Two Proportions in Clinical Trials

Contemp Clin Trials. 2008 Jul;29(4):507-16. doi: 10.1016/j.cct.2007.12.001. Epub 2007 Dec 23.

Abstract

In the process of clinical trials and health-care evaluation, Bayesian approaches have increasingly become the center of attention. In this article, sample size calculations for a non-inferiority test of two independent binomial proportions in a clinical trial are considered in a Bayesian framework. The hybrid Neyman-Pearson-Bayesian (hNPB) probability, the conditionally Bayesian (cB) probability and the unconditionally Bayesian (uB) probability are formulated through a conjugate normal analysis. The sample sizes are calculated based on formulas where normal prior distributions are assumed, and are compared with the Neyman-Pearson (NP) sample size. Our results show that the sample size based on the hNPB probability allows us to critically evaluate the appropriateness of the NP sample size. It is suggested that the sample size calculated based on the cB probability formula is smaller than the NP sample size.

Publication types

  • Research Support, Non-U.S. Gov't
  • Review

MeSH terms

  • Bayes Theorem*
  • Clinical Trials as Topic / methods*
  • Clinical Trials as Topic / statistics & numerical data
  • Humans
  • Models, Statistical
  • Probability
  • Sample Size*