A note on testing families of hypotheses using graphical procedures

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmaceutical Research

Background:

• Confirmatory clinical trials require strict control of the familywise error rate when testing multiple hypotheses.
• Existing multiple testing procedures are often designed for single families of hypotheses, limiting their application in complex trials.
• Graphical methods have been developed for single or multiple families of hypotheses, aiding visualization and performance.

Purpose of the Study:

• To present a general algorithm for calculating adjusted p-values in sequentially rejective graphical test procedures.
• To extend the applicability of graphical methods to scenarios involving grouped families of hypotheses.
• To provide a flexible framework for addressing multiplicity in confirmatory clinical trials with multiple objectives.

Main Methods:

• Development of a general algorithm for adjusted p-value calculation.
• Application to sequentially rejective graphical test procedures.
• Consideration of grouped families of hypotheses where individual tests may not be graphical.

Main Results:

• The proposed algorithm enables adjusted p-value calculation for a broader range of multiple testing scenarios.
• It accommodates grouped families of hypotheses, offering flexibility beyond single-family structures.
• The method is applicable even when internal test procedures within families are not graphical.

Conclusions:

• The developed algorithm enhances statistical rigor in drug development by providing a versatile tool for multiple testing.
• This approach supports confirmatory trials with complex objectives by managing multiplicity across grouped hypothesis families.
• The findings contribute to the literature on graphical methods and adjusted p-value calculations in pharmaceutical statistics.