Abstract

Background: A reference interval (RI) is usually derived from a data set obtained cross-sectionally without consideration of the long-term analytical coefficient of variation (CV) (CV_A) or within-individual CV (CV_I).

Methods: A total of 135 healthy volunteers, 22–59 years of age, working in eight clinical laboratories in Japan participated in a study to determine RIs for 26 commonly measured biochemical analytes. Blood samples were collected monthly for 1 year and measured in each laboratory. Common quality control (QC) specimens were measured monthly. RIs were derived parametrically using a modified Box-Cox power transformation method. Subjects with abnormal results for test items other than the one for derivation were excluded.

Results: Although there was no statistically significant between-month variation, unacceptably large between-laboratory differences were observed in QC test results. Thus, all values were transformed to those of the main laboratory based on cross-check test results. CV_I values computed by two-level nested analysis of variance were generally greater than those commonly cited in the literature. Average RIs derived from the monthly results agreed well with those derived from the entire results. However, CVs of lower and upper limits of the monthly RIs were greater than CV_A.

Conclusions: To derive reliable RIs, it is important to consider long-term variability of CV_I and CV_A.

Clin Chem Lab Med 2008;46:691–8.