Point estimation in adaptive enrichment designs.

Area of Science:

• Clinical Trials
• Biostatistics
• Pharmaceutical Research

Background:

• Adaptive enrichment designs allow selecting patient subgroups or the full population based on interim data.
• These designs are valuable for therapies with potentially differential efficacy across patient groups.
• Selection based on efficacy data can introduce bias into standard estimation methods.

Purpose of the Study:

• To address bias in adaptive enrichment designs.
• To present and evaluate alternative statistical estimators for two-stage designs with a single subgroup.
• To identify methods that effectively reduce bias in subgroup selection.

Main Methods:

• Investigated six alternative estimators for two-stage adaptive enrichment designs.
• Conducted a simulation study to assess estimator performance across various scenarios.
• Focused on bias reduction when subgroup selection is based on interim efficacy data.

Main Results:

• Identified a combined estimator (uniformly minimum variance conditionally unbiased estimator + conditional moment estimator) as most effective.
• This combined method demonstrated consistent bias reduction across diverse simulation scenarios.
• Evaluated the practical application of these estimators using a clinical trial example.

Conclusions:

• The proposed combined estimator offers a robust solution for mitigating bias in adaptive enrichment trials.
• Accurate estimation is crucial for reliable efficacy conclusions in subgroup-selected trials.
• These methods enhance the validity of findings from adaptive clinical trial designs.