Properties of multiple intersection-union tests for multiple endpoints in combination therapy trials.

Area of Science:

• Biostatistics
• Clinical Trials
• Pharmacology

Background:

• Combination drug trials often involve multiple endpoints, necessitating robust statistical methods.
• Controlling the familywise error rate (FWER) is crucial for valid inferences in such studies.
• Existing methods like Bonferroni-Holm may lack power in intersection-union testing scenarios.

Purpose of the Study:

• To evaluate and compare statistical methods for controlling FWER in intersection-union tests for combination drug trials.
• To develop and assess heuristics for FWER control tailored to the intersection-union setting.
• To provide practical recommendations for selecting appropriate statistical tests in clinical trial settings.

Main Methods:

• The study employed closed testing procedures to control the familywise error rate (FWER) in the strong sense.
• Methods evaluated include Bonferroni-Holm, Simes-Hommel, and Resampling-Based approaches.
• A simulation study specifically designed for intersection-union settings was used to assess performance.

Main Results:

• Both Resampling-Based and Simes-Hommel methods demonstrated superior performance over Bonferroni-Holm.
• Simulation results indicated that the choice between Simes-Hommel and Resampling-Based depends on the specific alternative hypothesis.
• The Simes-Hommel method was found to be simpler and generally possessed good statistical power.

Conclusions:

• The Simes-Hommel intersection-union test is recommended for its balance of simplicity and power in controlling FWER.
• The study provides valuable insights for statisticians designing and analyzing combination drug trials with multiple endpoints.
• The practical application of these methods was demonstrated using real-world data from an asthma therapy clinical trial.