Unbiased estimation of biomarker panel performance when combining training and testing data in a group sequential...

Area of Science:

• Biostatistics
• Biomarker Discovery
• Clinical Trial Design

Background:

• Sjögren's Syndrome diagnosis can be challenging in symptomatic populations.
• Development of accurate and efficient screening tests is crucial for early identification.
• Combining multiple biomarkers may improve diagnostic accuracy over single markers.

Purpose of the Study:

• To propose a methodology for combining multiple biomarkers for Sjögren's Syndrome screening.
• To evaluate biomarker panel performance using a two-stage group sequential design.
• To develop an unbiased estimation method for biomarker combination rules and diagnostic performance.

Main Methods:

• A two-stage group sequential design to evaluate biomarker panels.
• Nonparametric conditional resampling algorithm for unbiased estimation.
• Incorporation and evaluation of the Copas and Corbett correction for bias.
• Extensive simulation studies to assess finite sample performance.

Main Results:

• The proposed methodology allows for efficient specimen conservation.
• The nonparametric resampling algorithm provides unbiased estimates of biomarker combination rules.
• The study provides guidelines for designing similar biomarker evaluation studies.
• The methods were successfully applied in the National Cancer Institute's Early Detection Network Urinary PCA3 Evaluation Trial.

Conclusions:

• The developed two-stage design and statistical methods offer an efficient approach for biomarker panel evaluation.
• This methodology can aid in the development of screening tests for diseases like Sjögren's Syndrome.
• The study contributes to robust statistical approaches for biomarker-based diagnostic test development.