Adjustment for measurement error in evaluating diagnostic biomarkers by using an internal reliability sample.

Area of Science:

• Biostatistics
• Biomarker Discovery
• Diagnostic Accuracy

Background:

• Biomarkers are crucial for disease diagnosis but susceptible to measurement error.
• Lab conditions and intra-subject variability introduce inaccuracies in biomarker readings.
• Existing methods may not adequately correct for this inherent measurement error.

Purpose of the Study:

• To develop a parametric bias-correction approach for biomarker diagnostic measures.
• To accurately estimate sensitivity, specificity, Youden index, predictive values, and likelihood ratios.
• To address the impact of measurement error on biomarker performance evaluation.

Main Methods:

• Utilized an internal reliability sample of the biomarker.
• Developed a likelihood-based parametric bias-correction method.
• Derived asymptotic properties, including consistency and normal distribution of estimators.
• Proposed confidence intervals and receiver operating characteristic (ROC) curve confidence bands.

Main Results:

• The proposed method effectively removes bias caused by biomarker measurement error.
• Outperformed the naive approach (ignoring error) in both point and interval estimation.
• Demonstrated consistency and asymptotic normality of the bias-corrected estimators.
• Analyzed bias in naive estimates, showing potential for underestimation or anti-conservative bias.

Conclusions:

• A parametric bias-correction approach using reliability samples accurately evaluates biomarker performance.
• The method provides reliable estimates for various diagnostic measures, crucial for clinical application.
• Recommends collecting reliability samples during biomarker discovery for robust performance assessment, as illustrated in an Alzheimer's disease study.