Control of familywise errors in multiple endpoint assessments via stepwise permutation tests

Area of Science:

• Statistics
• Biostatistics
• Clinical Trials

Background:

• Multiple endpoint assessments in clinical research require robust statistical methods to control for Type I error rates.
• Existing multiple comparison procedures, such as Bonferroni, Holm, and Rom, have limitations in power and applicability under certain data conditions.

Purpose of the Study:

• To introduce and evaluate novel permutation-based sequentially rejective multiple comparison procedures.
• To compare the statistical power of these new methods against established procedures like Holm, Rom, and Bonferroni.
• To demonstrate the practical application of these methods in analyzing real-world clinical data.

Main Methods:

• Development of permutation-based sequentially rejective multiple comparison procedures.
• Utilizing Monte Carlo simulations to assess and compare the power of different multiple comparison methods.
• Application of the developed methods to visual field data from optic neuritis patients.

Main Results:

• The newly devised permutation-based methods demonstrated superior or comparable power to traditional methods (Holm, Rom, Bonferroni) in various scenarios.
• The enhanced power was particularly evident when dealing with a large number of endpoints.
• Performance remained strong even with significantly correlated data or questionable distributional assumptions.

Conclusions:

• Permutation-based sequentially rejective procedures are valuable tools for multiple endpoint assessments, especially in complex datasets.
• These methods offer a powerful alternative when standard statistical assumptions are violated.
• The study highlights the utility of these novel approaches in clinical research, exemplified by optic neuritis visual field analysis.