Statistical Considerations in Demonstrating CMC Analytical Similarity for a Biosimilar Product

Demonstration of analytical similarity between a reference product and a biosimilar product is required as part of the biosimilarity approval process. A statistical two one-sided test (TOST) based on the difference of means proposed in the literature and recommended by the FDA for Tier 1 quality attributes is demonstrated to have a type I error rate greater than the specified level, which cannot be corrected by increasing sample size. In this paper, an alternative TOST based on the effect size is demonstrated to maintain the desired type I error rate, and in some situations, provide greater power than the TOST based on the mean difference. Results are demonstrated both analytically and with computer simulations. An example with calculations is provided in the paper.

生物类似药获批审评流程中，需证明参比制剂与生物类似药间的分析相似性。现有文献提出并经美国食品药品监督管理局（FDA）推荐的、基于均值差的统计双向单侧检验（Two One-Sided Test，TOST），在应用于一级质量属性（Tier 1 quality attributes）时，其I类错误率超出预设水准，且无法通过增加样本量进行校正。本文提出一种基于效应量的备选双向单侧检验，该方法可维持预设的I类错误率，在部分场景下的检验效能还优于基于均值差的TOST。本文通过解析推导与计算机模拟两种方式验证了上述结论，并附上带完整计算过程的示例案例。