Comparison of two (data mining) indirect approaches for between-subject biological variation determination

Area of Science:

• Clinical chemistry
• Biostatistics
• Laboratory medicine

Background:

• Between-subject biological variation (CVg) is crucial for analytical performance specifications, delta checks, and index of individuality calculations.
• Accurate estimation of CVg is essential for robust laboratory practice.

Purpose of the Study:

• To compare the performance of two indirect, data mining approaches for deriving between-subject biological variation (CVg).
• To evaluate the impact of various simulation parameters on the accuracy of CVg estimation.

Main Methods:

• Numerical simulations were used to compare the Expected Mean Squares (EMS) method against the Harris and Fraser approach.
• Parameters varied included mean difference (d), within-subject coefficient of variation (CVi), pathological fraction (f), and CVi increase (e).
• Performance was assessed by the relative fractional error of the recovered CVg compared to the true value.

Main Results:

• Both EMS and Harris and Fraser methods showed similar reliable performance, with a high percentage of conditions within ±0.20 fractional error for both normal and lognormal distributions.
• Performance was better when using calculated CVi compared to presumptive CVi.
• CVg estimation accuracy improved with decreasing values of f, d, and CVi.

Conclusions:

• The two statistical approaches (EMS and Harris and Fraser) demonstrated reliable performance for estimating CVg under the examined simulation conditions.
• These findings support the utility of these indirect methods in laboratory practice.