Aligning Test and Reference Methods by Simultaneous Correction of Proportional and Constant Bias in Comparative...

Area of Science:

• Biostatistics
• Diagnostic accuracy
• Biomarker analysis

Background:

• Constant and proportional biases impact diagnostic accuracy.
• Current methods address these biases independently, reducing precision.
• Accurate bias correction is crucial for reliable diagnostic test comparisons.

Purpose of the Study:

• Introduce a unified regression-based method for correcting both constant and proportional biases.
• Enhance precision and reliability in diagnostic accuracy assessments.
• Provide a foundation for reproducible recalibration across diagnostic platforms.

Main Methods:

• Developed a unified regression-based correction technique.
• Applied linear and log-transformed regression adjustments.
• Validated the method using simulated polymerase chain reaction viral load data and Bland-Altman plots.

Main Results:

• The unified correction successfully eliminated bias, evidenced by regression line confidence intervals including ideal values (slope=1, intercept=0).
• Residual analysis confirmed the model's linearity assumptions.
• Demonstrated effective bias correction in simulated viral load data.

Conclusions:

• The unified correction method improves the reliability of comparative diagnostic studies.
• Offers a reproducible framework for recalibrating diagnostic platforms.
• Significantly enhances the accuracy and precision of diagnostic test evaluations.