I would like to make several corrections to the paper titled ‘Statistical methods for assessing interchangeability of biosimilars’ by Chow, Yang, Starr and Chiu, Statistics in Medicine.
In Section 4.2, the symbols for estimator should be used for clarity in the steps illustrating how to derive totality biosimilarity index, that is,
- Step 1.
Obtain , the biosimilarity index for the ith domain;
- Step 2.
Define the totality biosimilarity index as , where wi is the weight for the ith domain, where i = 1, … ,k (number of domains); and
- Step 3.
Claim biosimilarity if the 95% confidence lower bound of is greater than a prespecified value .
In Section 4.3, the switching index should be defined as , the smallest order of the biosimilarity indices. Thus, probability density function of the defined switching index is given by
And the expected value and the variance of SI can be given by
where denoted as the second raw moment of SI.
When the observations are dependent and/or from different distribution, the nonparametric bounds as obtained in  is
where sample size n = 4.
In Section 4.4, the first sentence should have read, ‘A similar idea can be applied to develop ___SI_alternating index (AI) under an appropriate study design.’
The authors apologize for the oversight.