|
Abstract:
|
The choice of similarity or inferiority margins is a key design element for biosimilar and non-inferiority studies. There are three major issues with the current practice of margin specifications: The subjectivity in the process of determining the margin, the data variability not considered or only implicitly expressed, and the concern of biocreep controversy. In this paper, an average inferiority measure (AIM) is introduced as a tool for specification of margins to address these issues and the standardized margins derived from AIM reflect naturally the variability of data. AIM is applied to the normal, binary and the survival data. A general theorem is obtained to establish the asymptotic normality of a test statistic for the non-inferiority hypothesis defined by the standardized margin. We also propose an additional requirement on the treatment effect difference for concluding non-inferiority to address the concern of biocreep and show that this additional requirement does not reduce the power when the treatment and control are true equivalent.
|
