Quantifying and comparing the accuracy of binary biomarkers when predicting a failure time outcome

Area of Science:

• Biostatistics
• Survival Analysis
• Biomarker Validation

Background:

• Traditional predictive values assess binary outcomes, lacking standard methods for survival data.
• Accurate prediction of failure times is crucial in many medical and biological fields.
• Existing methods for evaluating biomarker accuracy in survival analysis are limited.

Purpose of the Study:

• To propose and validate extended predictive values for quantifying biomarker accuracy with censored survival data.
• To develop methods for rigorously comparing the predictive accuracy of multiple biomarkers.
• To provide a statistical framework for estimating and testing differences in biomarker predictive accuracy.

Main Methods:

• Extension of traditional positive and negative predictive values to handle censored survival data.
• Development of a marginal regression framework for estimating predictive accuracy.
• Statistical methods for comparing and testing the significance of differences in predictive accuracy between biomarkers.

Main Results:

• The proposed extended predictive values provide a standardized approach for assessing biomarker accuracy in survival analysis.
• The marginal regression framework allows for reliable estimation of predictive accuracy differences.
• Statistical tests enable rigorous comparison of biomarker performance.

Conclusions:

• The extended predictive values offer a robust and interpretable method for evaluating biomarkers predicting failure times.
• This framework facilitates more accurate biomarker selection and validation in survival studies.
• The proposed methods enhance the quantitative assessment of predictive accuracy in censored survival data.