Using a dose-finding benchmark to quantify the loss incurred by dichotomization in Phase II dose-ranging studies

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Pharmaceutical Research

Background:

• Continuous clinical trial endpoints are often dichotomized for analytical simplicity, despite potential information loss.
• Existing research quantifies information loss in parallel-group and two-stage Phase II trials, but not specifically in dose-ranging trials.
• Larger sample sizes are generally required for dichotomized endpoints compared to continuous ones to maintain statistical accuracy.

Purpose of the Study:

• To propose and apply a novel approach for estimating information losses due to endpoint dichotomization in Phase II dose-ranging trials.
• To quantify the impact of dichotomization on required sample sizes in dose-ranging studies.
• To provide a clinical setting-dependent estimation method, independent of specific dose-ranging designs.

Main Methods:

• Utilized a nonparametric optimal benchmark for dose-finding trials to establish an upper bound on performance.
• Applied the benchmark to assess Phase II dose-ranging trial designs and quantify dichotomization losses.
• Used parameters from real-world clinical trials across diverse therapeutic areas for realistic scenario analysis.

Main Results:

• The ratio of sample sizes needed for continuous versus binary endpoints typically ranged from 70% to 75%.
• In certain scenarios, the required sample size for binary endpoints could be as low as 50% of that for continuous endpoints.
• The proposed method provides a flexible way to estimate these losses without needing the specific dose-ranging design.

Conclusions:

• Endpoint dichotomization in Phase II dose-ranging trials leads to significant information loss, necessitating larger sample sizes.
• The developed benchmark approach effectively quantifies these losses, offering valuable insights for trial design.
• These findings highlight the trade-offs between analytical simplicity and statistical efficiency when choosing clinical endpoints.