2018 年 39 巻 1 号 p. 9-22
In a relative potency assessment, it is necessary to make assumptions about the similarities between substances and their dose-response profile. For example, in a parallel line bioassay which uses the dose-response data within the linear response range, we need to demonstrate that the dose-response slopes of the study substances are approximately parallel. When using multiple animals for testing, it is also crucial to confirm that this parallelism exists not only for the averages but also within each animal (Uehara et al. 2016a).
Meanwhile, when applying a linear mixed effect model to the analysis of parallel line assays, the between-substance difference of the slopes can be treated as a random effect. Thus, under a balanced assay design, we can derive an efficient score test to assess the quantile of the slope difference (McCulloch et al. 2008), which enables us to determine whether the majority of animals have their slope difference within the acceptable range.
We applied this approach to the assessment of intrasubject parallelism with the intention of ameliorating the conservatism of our previous method (Uehara et al. 2016b). We present an example that uses the proposed method, along with the results of simulation studies.