Point estimation for adaptive trial designs II: Practical considerations and guidance

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Regulatory Science

Background:

• Adaptive clinical trials allow design modifications during the study, but conventional treatment effect estimates can be biased.
• Recent FDA guidance emphasizes the need for unbiased estimates in adaptive designs, yet practical application remains limited.
• This article, Part II of a series, focuses on bias in point estimation for adaptive trials.

Approach:

• Reviews the impact of bias on standard estimators and its negative consequences.
• Examines current practices for reporting point estimates in adaptive trials.
• Illustrates the computation of various estimators using a real adaptive trial example and a simulation study.

Key Points:

• While average values of different estimators may be similar, individual trial results can show significant differences.
• The choice of estimator can notably impact the estimated treatment effect for a specific trial realization.
• Bias in point estimation should be addressed throughout the adaptive design lifecycle, with a prespecified estimation strategy.

Conclusions:

• Proposes guidelines for selecting and reporting point estimates in adaptive trials.
• Recommends the use of unbiased or bias-reduced estimators when available.
• Emphasizes the importance of prespecifying the estimation strategy in the statistical analysis plan for adaptive designs.