Resampling approach for determination of the method for reference interval calculation in clinical laboratory...

Area of Science:

• Clinical laboratory science
• Biostatistics
• Medical diagnostics

Background:

• Reference intervals (RI) are essential for interpreting laboratory test results.
• Current methods for RI determination, primarily parametric and nonparametric, are often chosen arbitrarily.
• A robust comparison of these methods is needed to ensure accurate clinical interpretation.

Purpose of the Study:

• To demonstrate the utility of a resampling approach for comparing RI determination techniques.
• To identify the most suitable method for calculating RI based on data distribution.
• To provide guidance for selecting appropriate RI calculation procedures.

Main Methods:

• Applied parametric, transformed parametric, and quantile-based bootstrapping methods to calculate RI.
• Utilized random samples from complement factor B observations (n=81) and a simulated normal population.
• Employed a resampling strategy to compare the performance of different RI calculation methods.

Main Results:

• Differences in RI between methods can exceed 20%.
• The transformed parametric method proved optimal for non-normally distributed complement factor B data, yielding unbiased RI with minimal confidence limits and interquartile ranges.
• For normally distributed data, parametric methods performed best; quantile-based bootstrapping showed bias at low sample sizes, and the transformed parametric method produced significant bias.

Conclusions:

• A resampling approach offers a valuable tool for comparing RI calculation methods.
• The transformed parametric method is recommended for non-normally distributed data like complement factor B.
• The study provides an algorithm to aid researchers in selecting the appropriate RI determination method.