Within-subject biological variation estimates using an indirect data mining strategy. Spanish multicenter pilot study...

Area of Science:

• Clinical Chemistry
• Laboratory Medicine
• Biostatistics

Background:

• Traditional direct methods for estimating biological variation (BV) have inherent limitations.
• Indirect methods offer an alternative approach to overcome these limitations.
• Existing indirect methods require refinement for accurate BV estimation.

Purpose of the Study:

• To present a novel indirect method for estimating within-subject biological variation (CV_I) and confidence intervals (CI).
• To evaluate the performance of the new method using data from a multicenter pilot study.
• To compare the obtained CV_I estimates with established gold standards, specifically the EFLM-BVD CV_I estimates.

Main Methods:

• A multicenter pilot study was conducted over 18 months involving 3 Spanish hospitals.
• Data for 7 measurands were collected from patients aged 18-75 years with multiple determinations.
• Four strategies were employed: coefficient of variation ratio (rCoeV), and bootstrap methods (OS1, RS2, RS3), with RS2 and RS3 utilizing symmetric reference change value (RCV) for database cleaning.

Main Results:

• The RS2 and RS3 strategies demonstrated the best correlation for CV_I estimates against the EFLM-BVD gold standard.
• RS3, which combined RCV and outlier removal, showed strong performance.
• The rCoeV and OS1 strategies resulted in an overestimation of CV_I values.

Conclusions:

• The developed indirect method, incorporating symmetric RCV for target population selection, yields robust CV_I estimates.
• The obtained CV_I estimates exhibit good correlation with the EFLM-BVD database.
• This strategy effectively addresses limitations of direct methods, including challenges in calculating confidence intervals.