The diagnostic performance of FDA-cleared latex specific IgE immunoassays has been reported using manufacturers' derived cutoffs (JACI, 103:925-930, 1999). We extend these findings using receiver operating characteristic (ROC) curves and evaluate the repeatability of the assays at low specific IgE concentrations for latex and other antigens. Sera from 311 subjects (131 latex puncture skin test [PST] positive and 180 PST negative) were pair-analyzed for latex-specific IgE antibodies in the AlaSTAT microplate (Ala). Hycor HY-TECH RAST (HY) and Pharmacia-UpJohn CAP Systems (CAP). Accuracy was evaluated by ROC analysis. Reproducibility was evaluated by repeatedly testing (3 times, 58 sera; Ala, latex antigen, sera from above) or 86 different sera, 2 or 3 times (CAP, varied antigens). The areas under the ROC curves (AUCs)+/- standard error (SEM) were: 0.858 +/- 0.024, 0.869 +/- 0.024 and 0.924 +/- 0.017. respectively for Ala, CAP and HY. The HY system had a greater (P <0.05) AUC based on PST than that observed for Ala and CAP. Cutoffs yielding maximal diagnostic efficiencies were <.35 kU(A)/L for CAP (87.1%) and Ala (88.1%) and 0.11 kU/L for HY (88.7%) (P=NS). Ala repeatability studies showed 66% of re-runs (38/58) yielded equivalent Ala positive (pos) or negative (neg) status, while 20/58 (34%. P <0.0001) yielded one or more repeated results where pos/neg status were discordant. The mean value of Ala pos results where replicates showed discordant neg results was 0.44 +/- 0.02 kU/L. The Ala inter-assay coefficient of variation (CV) was 4.0% +/- 0.5 1%. For the CAP system, 61.6% (53/86) of re-runs yielded equivalent CAP pos or neg results; 38.4% of sera (33/86, P <0.0001) yielded one or more repeated result(s) where pos/neg status were discordant. The mean value of CAP pos results where replicates showed discordant neg results was 0.41 0.06 kU(A)/L. The CAP inter-assay CV was 13.1% +/- 0.4%. These data indicate that ROC analysis is beneficial in increasing the accuracy of commercial assays for anti-latex IgE antibody by defining a more optimal positive threshold cutoff for some tests. Caution should be exercised when modifying cutoff levels which are different from those suggested by the manufacturers' and the basis of their FDA-clearance. Some commercial tests suffer from significant imprecision at levels close to their recommended positive cutoffs, yielding 35-38% of samples with neg or pos discrepancies upon repeated testing. This may explain the wide disparity in reported seroprevalence rates for latex allergy and discordant data for individual sera between commercial assays, as well as the low sensitivity of these assays compared to PST. These data underscore the need for a well-characterized PST reagent for the diagnosis of latex allergy.