Interval estimation for operating characteristic of continuous biomarkers with controlled sensitivity or specificity

Area of Science:

• Biostatistics
• Medical Diagnostics
• Biomarker Evaluation

Background:

• Receiver Operating Characteristic (ROC) curves offer broad biomarker performance assessment.
• Clinical utility often requires operating at specific sensitivity or specificity levels.
• Existing interval estimation methods for diagnostic accuracy face challenges with variance estimation and erratic behaviors.

Purpose of the Study:

• To propose a novel statistical method for estimating diagnostic accuracy metrics at controlled sensitivity or specificity levels.
• To extend the superior performance of score intervals for binomial proportions to biomarker evaluation.
• To investigate both single-biomarker evaluation and two-biomarker comparisons.

Main Methods:

• Developed a novel extension of the score interval for biomarker problems.
• Utilized exact bootstrap methods to establish consistency of the bootstrap variance estimator.
• Investigated performance through extensive simulation studies.

Main Results:

• The proposed method demonstrates competitive performance compared to existing approaches.
• The bootstrap variance estimator's consistency was established.
• The method is applicable to both single and dual biomarker evaluations.

Conclusions:

• The novel score interval extension provides a robust approach for diagnostic accuracy estimation.
• The proposed methods offer improved statistical properties for biomarker evaluation.
• The approach is illustrated effectively using a prostate cancer diagnosis example.