Performance of four regression frameworks with varying precision profiles in simulated reference material...

Area of Science:

• Clinical Chemistry
• Biostatistics
• Measurement Science

Background:

• Reference material (RM) commutability assessment is crucial for accurate clinical diagnostics.
• Traditional regression methods (OLS, Deming) assume constant variance, which often doesn't hold true for measurement procedures.
• Flexible regression frameworks like quantile regression and GAMLSS can model non-constant variance.

Purpose of the Study:

• To evaluate the performance of different regression frameworks for assessing RM commutability under various imprecision profiles.
• To determine the impact of clinical sample (CS) number and RM concentration relative to CS on commutability assessment outcomes.
• To identify the most appropriate statistical methods for RM commutability assessment based on measurement variance characteristics.

Main Methods:

• Simulated four distinct imprecision profiles (constant, proportional, mixed variance).
• Assessed four regression frameworks: OLS, Deming, quantile regression, and GAMLSS.
• Varied the number of CS, data transformations, and RM concentration relative to CS.
• Evaluated framework performance by the proportion of false commutability rejections and compared with nominal probability coverage.

Main Results:

• For simple variance profiles, Deming regression (with or without log transformation) was most efficient.
• For mixed variance profiles, GAMLSS with smoothing techniques proved more suitable.
• Increasing CS number and strategic RM placement improved accuracy in mixed variance scenarios.
• U-shaped coefficient of variation profiles posed challenges even for flexible models.

Conclusions:

• Measurement procedure variance profiles and RM location relative to CS concentration significantly impact the false rejection rate in commutability assessments.
• The choice of regression framework should be guided by the specific variance characteristics of the measurement procedure.
• Careful consideration of these factors is essential for reliable RM commutability evaluations.