The evaluation of imprecision in collaborative immunoassay quality-assessment programmes.

The calculation of within-laboratory imprecision in quality-assessment (QA) programmes normally involves combining data from different analyte concentrations to calculate an average standard deviation (SD) or coefficient of variation. However, for immunoassay neither of these parameters is concentration independent. This paper describes a method of calculating within-laboratory imprecision in QA programmes by assuming a linear relationship between SD and analyte concentration. This method is used in programmes conducted by the Australian Joint Working Party for Quality Control in Immunoassay to calculate imprecision at the limits of the reference range. Results from these programmes show that this method better represents the differences in imprecision between analytes, methods and laboratories than the calculation of a single imprecision parameter. The method is trivial for a computer and its robustness has been validated by Monte Carlo simulation. It is suggested that major differences in laboratory performance between different QA programmes may be due to inappropriate calculation of single imprecision parameters.