The effect of estimating prevalences on the population-wise error rate

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Clinical trials often evaluate treatments across multiple target populations.
• Controlling the population-wise error rate (PWER) is crucial for accurate efficacy assessment.
• PWER represents the probability of exposing a future patient to an ineffective treatment.

Purpose of the Study:

• To investigate methods for controlling the population-wise error rate (PWER) in multi-population clinical trials.
• To assess the performance of maximum-likelihood estimators for PWER when population prevalences are unknown.
• To analyze the relationship between expected PWER and study-specific PWER.

Main Methods:

• Simulations were conducted to evaluate the proposed estimation method.
• The study considered up to eight overlapping populations.
• Maximum-likelihood estimation under a multinomial distribution was employed.
• The maximum strata-wise family wise error rate was also analyzed.

Main Results:

• Using maximum-likelihood estimators did not substantially inflate the true population-wise error rate.
• The expected PWER was almost perfectly controlled across simulations.
• Study-specific PWER values varied within a narrow range.
• The maximum strata-wise family wise error rate was generally bounded by twice the significance level.

Conclusions:

• Maximum-likelihood estimation provides a viable approach to control PWER in multi-population trials, even with unknown prevalences.
• The proposed method offers reliable control over expected PWER.
• Understanding the distinction between expected and study-specific PWER is important for trial interpretation.