Drug Evaluation Committee Current Status and Future Issues of Multiplicity Problems in Clinical Trials An Explanation for Non-Statisticians
June 2020
In statistical hypothesis testing for drug development, there are two types of errors: the error of concluding that a test drug is “effective” when it is not, and the error of concluding that a test drug is “ineffective” when it is actually effective.While relatively simple multiple comparison correction methods, such as the Bonferroni method, were traditionally used, the increasing complexity of hypothesis structures and the growing demand for methods with higher statistical power have led to the proposal of multiple comparison correction methods—such as the gatekeeping method—that accommodate more complex hypothesis structures.As methods for correcting for multiple comparisons have become more sophisticated, a deep understanding of these methods is necessary to select the appropriate one that aligns with the trial objectives and hypothesis structure, requiring a high level of expertise.
Discussions regarding valid hypothesis structures aligned with the study objectives and the accompanying decision-making frameworks are critical issues that involve not only those responsible for statistical analysis but also many other relevant functions.
This paper is intended to provide an explanation of the issues of multiplicity associated with hypothesis testing to pharmaceutical development professionals who are not specialists in statistics, and we hope you will find it useful.
Japan Pharmaceutical Manufacturers Association, Drug Evaluation Committee, Data Science Subcommittee
FY2019 Task Force 4: Team for Adjusting for Multiplicity in Clinical Trials