Evaluations of Parallelism Testing Methods Using ROC Analysis

Abstract

Parallelism is often a prerequisite in relative potency determination using bioassays. It involves testing the similarity between a pair of dose–response curves of reference standard and the test sample. Parallelism testing methods currently in use include p-value-based significance tests and interval-based equivalence tests. Recently the conceptual and empirical properties of these tests have been assessed and compared by researchers. However, the evaluations are incomplete due to the lack of a common framework for comparing the two methods. Based on receiver operating characteristic (ROC) curve analysis, we propose that the area under the ROC curve (AUC) be used as an overall figure of merit for parallelism tests. A simulation study is conducted to compare the performance of significance and equivalence tests. We demonstrate that the equivalence test outperforms the significance test if the equivalence limits are properly selected. A method for determining the optimal equivalence limits is suggested. Such optimal equivalence limits potentially make the application of the equivalence test to parallelism testing both practical and feasible.

Keywords:

• Area under the curve (AUC)
• Bioassay
• Consumers’ risk
• Parallelism
• Producers’ risk
• Relative potency

Acknowledgments

We thank the referees and associate editor for their comments that have helped improve the article.