Adaptive randomized phase II design for biomarker threshold selection and independent evaluation

Area of Science:

• Clinical Trial Design
• Biostatistics
• Pharmacogenomics

Background:

• Predictive biomarkers for experimental cancer therapies often lack validated thresholds at phase II trial initiation.
• Retrospective biomarker validation can lead to over-accrual of non-benefiting patients, resulting in underpowered studies and suboptimal care.

Purpose of the Study:

• Propose an adaptive randomized phase II study design for prospective biomarker threshold identification.
• Incorporate early futility stopping, mid-trial accrual restriction to marker-positive subjects, and final evaluation in the identified benefit population.

Main Methods:

• Utilize interim analysis to decide on trial continuation, futility stopping, or adaptive accrual based on biomarker promise.
• Perform final efficacy analyses in the target population identified at interim analysis.
• Conduct simulation studies to assess error rates, power, and sample size efficiency.

Main Results:

• The adaptive design effectively identifies predictive biomarkers and restricts accrual to patients most likely to benefit.
• Type I and II error rates are controlled through candidate threshold prevalence restriction and optimized interim analysis timing.
• Demonstrates favorable aspects including controlled error rates, adequate power, and reduced average sample size.

Conclusions:

• The proposed adaptive design offers a solution for identifying and validating continuous biomarkers within randomized phase II trials.
• Facilitates personalized medicine by ensuring experimental treatments are evaluated in the patient subgroups most likely to respond.