Precision and accuracy of commercial laboratories' ability to classify positive and/or negative allergen-specific IgE results

Background:Accurate and reliable evaluation of the presence or absence of allergen-specific IgE is important in the differential diagnosis of allergic disease. A variety of different commercial tests are available for this purpose. There are few data available to judge how the results of these different tests compare with one another in everyday use. Objective: To examine prospectively the extent of comparability among specific IgE results from different laboratories. Methods: Six diagnostic laboratories employing five different methods to assay specific IgE were selected, Aliquots from 26 serum samples that contained variable levels of IgE specific to 17 common aeroallergens were sent in triplicate to each study laboratory during a 6-week time period. Results were reported numerically and by class scores and then compared by examining their concordance using Kendall's W nonparametric statistical test. In addition, cut-off values were compared by a best agreement analysis using reported results. Reproducibility was determined using precision profiles based upon the coefficient of variation among triplicates for each allergen across the range of reported results. Results: In all, 7,813 tests were analyzed. Concordance among different assays in commercial use with one exception was not good. This was particularly true around the cut-off region where most assays demonstrated high imprecision. The Pharmacia CAP System used by two different laboratories demonstrated highly comparable results with good precision. Some assays were reproducible but not accurate. Others were neither reproducible nor accurate. Conclusions: The results of this study indicate that not all commercial laboratories/assays for specific IgE provide reproducible and accurate data. Significant potential for misdiagnosis was detected for some reported results. Methods were identified that do give sensitive, accurate, and reproducible results.