Interval estimation of binomial proportion in clinical trials with a two-stage design

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Pharmacological Research

Background:

• Phase II clinical trials commonly use two-stage designs to efficiently evaluate drug efficacy.
• These designs involve sequential binomial responses to determine if a drug warrants further investigation.
• Avoiding ineffective treatments is a primary goal in early-phase drug development.

Purpose of the Study:

• To construct and evaluate interval estimators for response probability in two-stage Phase II clinical trials.
• To compare the performance of asymptotic (Wald, score) and exact (Clopper-Pearson, Sterne) interval estimators.
• To provide recommendations for interval estimation methods in Simon's optimal and minimax designs.

Main Methods:

• Development of four interval estimators: Wald, score, Clopper-Pearson, and Sterne.
• Utilizing two-stage binomial responses, with the first stage following a truncated binomial distribution.
• Performance evaluation based on mean actual coverage probability and expected interval width.

Main Results:

• Comparison of interval estimators revealed varying performance characteristics.
• The score interval demonstrated favorable performance across evaluated metrics.
• The truncated binomial distribution was applied to the first-stage response.

Conclusions:

• The score interval is recommended for response probability estimation in two-stage Phase II clinical trials.
• This recommendation applies to both Simon's optimal and minimax two-stage designs.
• The study provides valuable insights for optimizing clinical trial design and analysis.