Objectives: Interpatient distribution data for lipase (Roche Cobas® assay) showed an unexpected data gap, where no results were reported. This gap occurred beginning at a point just above the assay's primary measurement range (i.e., above the cutoff (300U/L) for automated repeat-on-dilution). Calculation or other errors within the automated dilution process were ruled out. Linearity of assay results was investigated.
Design and methods: Linearity of experimental sample dilution series data was assessed by correlation coefficient, intercept, and constancy of slope.
Results: Dilution experiment data demonstrated a discontinuity of results between 300 and 400U/L consistent with the observed gap in patient data. Although data within the presumed linear range of the assay had a high linear correlation coefficient (r2>0.99), a non-zero intercept and progressively variable slope were inconsistent with linearity. Although the assay was assessed as linear by the College of American Pathology linearity survey, survey data also demonstrated non-linearity for this assay when analyzed for slopes and intercept.
Conclusions: Non-linearity in the presumed linear range of an assay can produce gaps in patient data above a repeat-on-dilution cutoff. As in this instance, CAP linearity surveys may not identify certain forms of non-linearity.
Keywords: Linearity; Lipase; Proficiency testing; Quality assurance.
Copyright © 2015 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.