Bayesian Sample Size Determination based on Power Prior in Biosimilar Studies

• 作 者: WU Kejian PAN Haitao ZHAO Qingbo LI Chanjuan WANG Ling XIA Jielai
• 机 构: Department of Mathematics, Fourth Military Medical University, Xi'an Shannxi 710032, China
• 出 处: 2017年中国卫生统计学学术年会
• 发表日期: 2017年10期

　　Objective We consider a Bayesian approach of sample size determination (SSD) for the design of biosimilarity trials in biosimilar studies.We are concerned with the use of historical data based on the power prior.Methods We propose a new method of calculation,by which we determine in advance the estimated difference between the posterior mean and the power function or the highest posterior density (HPD) interval.We illustrate how the different weight parameter δ of power prior and the calibrated power prior (CPP) can be fruitfully employed to measure the portion of historical information needed in the current study.Results When the historical data D0 are congruent with the reference arm data DR,the δ of the CPP is close to 1 to determine a considerable increase in the sample size required.When the D0 are not congruent with DR,the result of the δ obtained by the CPP are close to the results of frequentist method,i.e.δ=0.Conclusion The numerical results of the CPP has desirable operating characteristics.The proposed Bayesian SSD approach can reduce the sample size that is required for the new experiment appropriately by borrowing the useful historical information.