A Bayesian adaptive blinded sample size adjustment method for risk differences

Andrew Montgomery Hartley
  • Pharmaceutical Statistics, September 2015, Wiley
  • DOI: 10.1002/pst.1708

New Adaptive Sample Size Adjustment Method for Risk Differences

What is it about?

In a clinical trial comparing 2 risks (proportions), such as the proportions of cures for 2 antibiotics, the sample size is usually selected to provide a targeted statistical power, assuming pre-specified values of those risks. In most such trials, however, the risks are not known, so that the power associated with any sample size is also not known. In those cases, a blinded sample size adjustment (BSSA) can be performed mid-trial, in which the data collected to date are used to adjust assumptions about the risks. However, existing BSSA methods are not effective when the risk difference is highly uncertain. This paper introduces a novel BSSA method for such cases, which demonstrably improves the ability either to achieve the targeted power, or to maximize the overall net benefit of the trial, through sample size selection.

Why is it important?

Previously developed BSSA methods for risk differences are ineffective. The proposed method enhances the ability to adjust sample size to meet scientific and commercial objectives.

The following have contributed to this page: Dr Andrew M Hartley

In partnership with: