The cut point is an important parameter for immunogenicity assay validation and critical to immunogenicity assessment in clinical trials. FDA (2019) recommends using a statistical approach to derive cut point, with an appropriate outlier removal procedure. In general, the industry follows the methods described in Shankar et al. (2008) and Zhang et al. (2013) among others to determine cut point. Outlier removal is a necessary step during the cut point determination exercise to reduce potential false negative classifications. However, the widely used statistical outlier removal method, namely, Tukey's box-plot method (1.5 times inter-quartile range, IQR), is often found to be overly conservative in the sense that it removes too many “outliers”. Tukey's box-plot method can be used to flag potential outliers for further investigation, however, it is not a hypothesis testing based statistical method. Removing these suspected “outliers” will lead to lower cut point which might confound immunogenicity assessment due to the presence of many low false positives. Besides, the very nature of assay analytical variability has a non-negligible adverse impact on the reliability of ADA classification in terms of false positive and false negative, demanding as large as possible contribution from biological variability relative to analytical variability. A new outlier removal procedure, which takes into account the relative magnitude between biological variability and analytical variability within the sample population, is proposed and statistically justified. After sequential removal of analytical and biological outliers, a 5% false positive rate and 1% false positive rate in screening and confirmatory assays, respectively, are still targeted without increasing potential false negatives. Internal data shows that this practice has minimal impact on assay sensitivity and has the advantage of selecting true positive samples. It is shown that the new procedure is more appropriate for cut point determination.

切点是免疫原性测定验证的重要参数，对于临床试验中的免疫原性评估至关重要。FDA（2019）建议使用统计方法得出切点，并采用适当的异常值去除程序。一般而言，该行业遵循Shankar等人（1989年）中描述的方法。（2008）和Zhang等。（2013年）确定切入点。在切点确定过程中，离群值去除是减少潜在的假阴性分类的必要步骤。但是，通常会发现广泛使用的统计离群值消除方法（即Tukey的箱形图方法（四分位数间距的1.5倍，IQR））过于保守，因为它消除了太多的“离群值”。Tukey的箱线图方法可用于标记潜在的异常值，以供进一步研究，它不是基于假设检验的统计方法。删除这些可疑的“异常值”将导致降低切点，由于存在许多低假阳性，这可能会混淆免疫原性评估。此外，就假阳性和假阴性而言，化验分析变异性的本质对ADA分类的可靠性具有不可忽略的不利影响，要求生物学变异性对分析变异性的贡献尽可能大。提出了一种新的离群值去除程序，该程序考虑了样本群体内生物学变异性和分析变异性之间的相对大小，并在统计学上证明了其合理性。在顺序去除分析和生物学异常值之后，在筛选和确认试验中，仍分别以5％的假阳性率和1％的假阳性率作为目标，而不会增加潜在的假阴性。内部数据表明，这种做法对测定灵敏度的影响最小，并且具有选择真正阳性样品的优势。结果表明，新程序更适合确定切点。