Validation study of assays alternative to animal experiments from a statistical viewpoint

在替代动物实验的测定方法的验证研究中，有三个统计学问题需要解决。它们是（1）测定的可靠性和可行性，（2）待替代的体内试验的特征，以及（3）体外测定与体内试验的相关程度或可预测性。关于（1），本研究通过对验证研究收集的数据进行统计分析，为日本动物实验替代学会开展的验证研究做出了贡献，并通过数据分析得出结论，结晶紫含量测定法是最可靠的方法。关于（2），本研究提出了一种数据分析方法，称为最大对比法，用于评价化合物的毒性，并确定未观察到毒性的剂量。关于（3），本研究成功地确定了一组通过使用预测平方和作为标准来最佳预测Draize眼刺激评分的测定。

There are three statistical issues to be addressed in the validation study of assays alternative to animal experiments. They are (1) the reliability and feasibility of assays, (2) the features of in vivo tests to be alternated, and (3) the degree of correlation or the predictability of in vitro assays to in vivo tests. Regarding (1), this reserch contributed to the validation study sonducted by the Japanese Society of Alternatives to Animal Experiments by statistically analyzing the data collected by the validation study and induced the conclusion that the Crystal Violet Assay is the most reliable one through the data analysis. Regarding (2), this research proposed a method of data analysis, which is referred to as the maximum contrast method, to evaluate toxicity of chemical compounds and identify the dose with no observed toxicity. Regarding (3), this reserch succeeded in identifying a battery of assays optimally predict the Draize eye irritation scores by using the prediction sum of squares as the criterion.