Estimation of bias between 2 analytical methods at clinically important ranges

Area of Science:

• Clinical Chemistry
• Biostatistics
• Laboratory Medicine

Background:

• Accurate bias estimation is crucial for comparing analytical methods in clinical diagnostics.
• Existing methods like NCCLS EP9-A2 may not be suitable for all laboratory applications, especially those involving count data.

Purpose of the Study:

• To introduce a novel statistical approach for estimating bias between two analytical methods across different clinical ranges.
• To provide a flexible and robust framework for bias assessment applicable to various laboratory techniques.

Main Methods:

• The approach models replicated data using maximum likelihood estimation.
• It accounts for repeatability and trueness bias, partitioning the latter into constant and proportional components.
• The method supports normal, Poisson, and binomial distributions, suitable for particle counting methods.

Main Results:

• The approach provides a full spectrum of statistical inference, including confidence intervals and hypothesis testing.
• Estimates are practically interpretable and can be related to clinical decision points.
• The method is demonstrated to be a viable alternative to the NCCLS EP9-A2 approach.

Conclusions:

• The proposed method offers a comprehensive and adaptable tool for analytical bias estimation in clinical laboratories.
• It is particularly recommended for hematology and other particle-counting applications where standard methods may be insufficient.
• This approach enhances the reliability of inter-method comparisons and supports informed clinical decision-making.