Assessment of bias with emphasis on method comparison

Area of Science:

• Clinical Chemistry
• Laboratory Medicine
• Analytical Science

Background:

• Bias is a critical performance metric in laboratory methods, distinct from accuracy.
• Method comparison studies are essential for evaluating new analytical techniques against established ones.

Purpose of the Study:

• To detail the procedures and data handling for bias assessment in method comparison studies.
• To provide guidance on interpreting results and establishing acceptable bias criteria.

Main Methods:

• Specimen testing using both candidate and existing methods.
• Statistical analysis including difference plots, linear regression (Deming, Passing-Bablok models), and log transformations.
• Evaluation of linearity and recovery as indicators of potential bias.

Main Results:

• Correlation is insufficient for bias assessment; difference plots are more informative.
• Appropriate statistical models and criteria are necessary for robust bias evaluation.
• Linearity and recovery tests serve as early warnings for concealed bias.

Conclusions:

• A comprehensive approach considering all steps is required to accept or reject a new method based on bias assessment.
• Method comparison studies, when properly conducted and analyzed, are vital for ensuring reliable laboratory diagnostics.