Two-stage biomarker panel study and estimation allowing early termination for futility

Area of Science:

• Biostatistics
• Biomarker Discovery
• Cancer Research

Background:

• Technological advancements offer numerous biomarkers for chronic diseases like cancer.
• Many potential biomarkers fail clinical validation due to insufficient performance.
• Group sequential designs can improve cost-effectiveness in biomarker studies using biobanks.

Purpose of the Study:

• To adapt a 2-stage group sequential design for validating a panel of biomarkers.
• To develop conditional estimators for clinical performance using updated risk models.
• To evaluate the proposed design and estimators against traditional fixed-sample and existing 2-stage designs.

Main Methods:

• Adaptation of a 2-stage group sequential design for biomarker panels.
• Development of conditional estimators using updated risk models and resampling procedures.
• Analytical derivation of estimators under a multivariate binormal distribution assumption.

Main Results:

• The proposed 2-stage design reduces sample size for unpromising biomarker panels.
• The design controls the rejection rate while increasing efficiency for promising panels.
• The updated risk model enhances statistical efficiency compared to designs that do not update.

Conclusions:

• The proposed 2-stage design and conditional estimators offer an efficient approach for biomarker panel validation.
• This adaptive strategy is particularly beneficial for cost-effectiveness in biomarker studies.
• The methods were successfully applied to a high-grade prostate cancer biomarker panel study.