A ROC-based test for evaluating the group difference with an application to neonatal audiology screening

Area of Science:

• Biostatistics
• Medical Diagnostics
• Statistical Modeling

Background:

• Accurate comparison of two samples is crucial in medical research and diagnostics.
• Existing methods for sample comparison may lack power, particularly with non-normal biomarker data.
• Receiver Operating Characteristic (ROC) curve analysis is a standard tool for evaluating diagnostic tests.

Purpose of the Study:

• To propose a novel and powerful statistical method for comparing two samples.
• To enhance diagnostic accuracy by utilizing ROC curve model parameters.
• To provide a robust statistical test applicable to non-normal biomarker data.

Main Methods:

• Inference drawn from Receiver Operating Characteristic (ROC) curve model parameters.
• Estimation of ROC parameters using a linear model framework on empirical sensitivities and specificities.
• Development of a comprehensive statistic based on the Cauchy combination.
• Implementation of an efficient one-layer wild permutation procedure for P-value calculation.

Main Results:

• The proposed method provides a more powerful test than existing methods in several situations.
• The Cauchy combination statistic demonstrates effectiveness across all considered scenarios.
• The method is particularly advantageous when underlying continuous biomarker results are non-normal.
• The approach was illustrated using a neonatal audiology diagnostic example.

Conclusions:

• The novel ROC curve parameter-based method offers a powerful and versatile tool for comparing two samples.
• This method enhances statistical power, especially in scenarios with non-normal biomarker distributions.
• The proposed approach, including the Cauchy combination statistic and permutation procedure, provides a robust framework for diagnostic accuracy assessment.