Considerations in Sample Size, Type I Error, and Power for 2-Group Noninferiority Study Using Farrington-Manning Method

• Activity Number: 188
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Songtao Jiang* and Edmund McMullen and Hong Wang and Terry Liao
• Companies: Boston Scientific and Boston Scientific and Boston Scientific and Boston Scientific
• Keywords: Farrington-Manning method ; Wald method ; non-inferiority ; type I error ; power ; simulation

Abstract

The Farrington-Manning (FM) method is generally considered as a conservative method for sample size calculation in a binary outcome non-inferiority study with 1:1 test vs control allocation. Research [Rebekkah S. Dann, 2006] has shown that under certain selected scenarios the Wald method can be fairly more conservative with 2:1, 3:1, and 3:2 in test vs control allocation compared to the FM method. While the FM method maintains nominal type I error well with 1:1 test vs control allocation, the type I error increases as more sample allocated to the test group. To further understand the properties of the FM method in the medical device non-inferiority clinical study design, specifically to understand the conditions that both type I error and power are maintained, simulations are performed in assessing type I error, power, by varying unequal allocation in sample size ratio comparing Wald and FM Methods. Conclusions are made in terms of type I error, power, and sample size allocation ratio using a case study.