Two procedures for checking agreement between twin analyzers (Abbott Architect ci 8200) were tested in 23 blood and 7 urine parameters (10,160 paired results from 7,882 blood and 2,278 urine tests). Two protocols were compared. In protocol 1, acceptance criterion is based on standard-deviation originated either from French recommendations (Société française de biologie clinique) or from within subject biological variation. In protocol 2, acceptance criterion is based on values of expanded uncertainty of measurements calculated according to SH GTA 04. Percentages of comparisons refused were significantly different (p < 0.05) and varied from 0.0 to 18.6% (median: 0.45%) and from 0.0 to 8.75% (median: 1.2%). For 9 blood parameters (sodium, total CO2 , total proteins, calcium, urea, LDH, amylase, lipase and troponin I) and 4 urine parameters (sodium, protein, glucose, creatinine) a significant difference was observed between protocol 1 and 2. For the majority of tests, evaluation criterion based on within run standard deviation (protocol 1) is less stringent than protocol 2 based on expanded uncertainly of measurements. The use of expanded uncertainty as comparability criterion seems to be an interesting approach, especially for assays presenting wide theoretical physiopathology changes.
Keywords: SH REF 14; assay uncertainty; standard deviation; twin analyzer comparison.