Transformed ROC Curve for Biomarker Evaluation.

Area of Science:

• Biostatistics
• Biomedical Data Analysis
• Diagnostic Accuracy

Background:

• Conventional Area Under the ROC Curve (AUC) has limitations in fully assessing diagnostic accuracy for certain non-standard biomarkers.
• Non-standard biomarkers may require advanced methods for accurate evaluation.

Purpose of the Study:

• Introduce a transformed ROC curve and transformed AUC (TAUC) to address limitations of conventional AUC.
• Demonstrate TAUC's ability to relate improper biomarkers to proper ones via non-monotone transformation.
• Develop and validate nonparametric estimation methods for TAUC and the transformation.

Main Methods:

• Introduced a novel transformed ROC curve and transformed AUC (TAUC).
• Developed nonparametric estimation techniques for the non-monotone transformation and TAUC.
• Established consistency and asymptotic normality for the proposed estimators.
• Conducted extensive simulation studies and analyzed real biomedical data.

Main Results:

• TAUC effectively relates improper biomarkers to proper ones through non-monotone transformation.
• The proposed nonparametric estimation methods for TAUC are consistent and asymptotically normal.
• Simulation studies confirmed the performance of the TAUC method.
• Case studies demonstrated TAUC's utility in identifying important biomarkers missed by traditional methods.

Conclusions:

• TAUC offers a valuable extension to AUC for enhanced diagnostic accuracy assessment.
• The proposed nonparametric methods provide reliable estimation for TAUC and transformations.
• TAUC facilitates the identification of significant biomarkers potentially overlooked by conventional screening.