Optimal predictive probability designs for randomized biomarker-guided oncology trials.

Area of Science:

• Oncology
• Clinical Trial Design
• Biostatistics

Background:

• Biomarker-targeted cancer therapies are advancing rapidly, with many receiving accelerated approval based on single-arm phase II trials.
• However, the absence of randomization in these trials limits the assessment of comparative efficacy against standard chemotherapies, especially for specific biomarker subpopulations.
• Historical control data often lacks the specificity needed for accurate comparisons, highlighting a need for robust trial designs.

Purpose of the Study:

• To propose and evaluate novel randomized designs for early-phase biomarker-guided oncology clinical trials.
• To enhance the efficiency of these trials by incorporating futility monitoring for multiple biomarker subpopulations.
• To provide a framework for more informative early-phase studies that can better inform later-stage development.

Main Methods:

• Three distinct randomized designs were proposed for biomarker-guided early-phase oncology trials.
• The optimal efficiency predictive probability method was employed for futility monitoring across biomarker subpopulations.
• Candidate designs were selected based on stringent statistical criteria (Type I error between 0.05-0.1, power ≥ 0.8).
• A simulation study, using data from a real-world atezolizumab trial in urothelial carcinoma, evaluated the operating characteristics of the proposed designs.

Main Results:

• The proposed enrichment design demonstrated the smallest average sample sizes (101.0 under null, 218.0 under alternative).
• The stratified control arm design required average sample sizes of 144.8 (null) and 213.8 (alternative).
• The pooled control arm design enrolled an average of 113.2 (null) and 159.6 (alternative) patients.
• These designs achieved average sample size reductions of 23-48% under the alternative and 47-64% under the null compared to a similar real-world trial.

Conclusions:

• Randomized designs incorporating futility stopping can significantly enhance the efficiency of early-phase biomarker-guided oncology trials.
• These innovative designs allow for the efficient collection of more comprehensive data on both treatment and control arms.
• The findings suggest that smaller phase II trials are feasible, providing valuable information to guide subsequent phase III studies.