Which strategy helps control Type I error when multiple analyses are planned in a trial?

• A. Increasing the sample size without adjusting alpha.
• B. Hierarchical testing, Bonferroni, alpha-spending, and gatekeeping.
• C. Ignoring secondary endpoints.
• D. Removing the primary endpoint.

Answer: (B)

When multiple analyses are planned in a trial, the chance of finding at least one false positive increases. To keep the overall false-positive rate at the intended level (the family-wise Type I error), you use strategies that allocate or constrain alpha across tests and analyses.

The approach that includes hierarchical testing, Bonferroni, alpha-spending, and gatekeeping is designed to do exactly that. Bonferroni divides the overall alpha among the individual tests, making each test more stringent. Hierarchical testing sets an order where the primary endpoint must be significant before secondary endpoints are tested, preserving the overall error rate. Alpha-spending uses a predefined spending function to allocate portions of alpha across interim analyses or multiple looks, controlling the cumulative Type I error. Gatekeeping imposes a sequence across endpoints or populations, only allowing downstream tests to proceed if upstream ones are significant, again protecting the error rate. Together, these methods provide robust control of false positives when multiple analyses are planned.

Increasing sample size without adjusting alpha doesn’t inherently control Type I error across multiple analyses. Ignoring secondary endpoints ignores the multiple-testing problem altogether. Removing the primary endpoint sidesteps the original hypothesis rather than addressing error control in the presence of multiple analyses.