Partitioning biochemical reference data into subgroups: comparison of existing methods

Area of Science:

• Clinical Biochemistry
• Biostatistics
• Reference Interval Determination

Background:

• Accurate partitioning of biochemical reference data into subgroups is crucial for clinical interpretation.
• Existing methods for subgrouping reference data have limitations in their statistical approaches.
• The goal is to establish reliable group-specific reference intervals when appropriate.

Purpose of the Study:

• To compare the efficacy of four distinct methods for partitioning biochemical reference data into subgroups.
• To identify the strengths and weaknesses of each method in determining appropriate reference intervals.
• To evaluate advancements in methods for subgrouping reference data.

Main Methods:

• Comparison of Sinton et al. and Ichihara and Kawai methods based on quotient of differences and variances.
• Evaluation of Harris and Boyd method focusing on maintaining outlier proportions (ideal 2.5%).
• Assessment of a "new method" with variations designed to overcome drawbacks of previous approaches.

Main Results:

• Sinton et al. criterion may be too stringent, favoring common intervals over group-specific ones.
• Ichihara and Kawai method can yield inconsistent results due to reliance on variance.
• Harris and Boyd method shows poor correlation with desired outlier proportions.
• The "new method" demonstrates improvements by addressing identified weaknesses.

Conclusions:

• Methods relying solely on distribution differences (Sinton, Ichihara & Kawai) are less reliable for reference interval partitioning.
• The principle of maintaining ideal outlier proportions (Harris & Boyd, new method) is a significant improvement.
• The "new method" offers a more robust and accurate approach to subgrouping biochemical reference data.