Biomarker Combination Based on the Youden Index With and Without Gold Standard.

Area of Science:

• Biostatistics
• Medical Diagnostics
• Bioinformatics

Background:

• Clinical practice often involves multiple biomarkers for disease diagnosis.
• Combining biomarkers can enhance diagnostic accuracy.
• Existing methods for biomarker combination have limitations, such as assuming a perfect gold standard or specific data distributions.

Purpose of the Study:

• To propose a two-stage method for optimal linear combination and cutoff value determination of multiple biomarkers.
• To improve diagnostic accuracy in practical settings where reference tests may be imperfect.
• To develop a robust method applicable to various diagnostic scenarios.

Main Methods:

• A two-stage approach is proposed: first, maximizing an approximated empirical area under the ROC curve (AUC) to estimate optimal linear coefficients.
• Second, maximizing the empirical Youden index to determine the optimal cutoff point for classification.
• The method is grounded in the semiparametric single index model.

Main Results:

• The proposed estimators for linear coefficients, cutoff point, and Youden index are shown to be consistent under regularity conditions.
• The method demonstrates applicability even when the reference standard is imperfect.
• Performance was validated through simulations and application to a Chinese medicine diagnostic scale.

Conclusions:

• The developed two-stage method provides a robust approach for combining multiple biomarkers and establishing optimal cutoff values.
• This method enhances diagnostic accuracy and is practical for real-world clinical settings, including those with imperfect reference standards.
• The approach is versatile and demonstrated effective in a real-world application for developing a diagnostic scale.