F. Deblay et al., RECEIVER OPERATING CHARACTERISTIC ANALYSIS - A USEFUL METHOD FOR A COMPARISON OF THE CLINICAL RELEVANCE OF 2 IN-VITRO IGE TESTS, Journal of allergy and clinical immunology, 92(2), 1993, pp. 255-263
Background: The objectives of this study were to demonstrate the usefu
lness of Receiver Operating Characteristic (ROC) analysis in the defin
ition of a positive threshold in the comparison of two in vitro tests
for detecting IgE antibodies and to determine the clinical relevance o
f specific IgE tests according to the sensitivity and specificity dete
rmined by ROC analysis. Methods: Specific IgE levels were compared wit
h the use of both MATRIX (Abbott Laboratories, Chicago, Ill.) and PHAD
EZYM (Pharmacia, Uppsala, Sweden) tests in 109 patients. 86 allergic p
atients and 23 control subjects. All allergic patients had a clinical
history of allergy to mite, cat, grass pollen, or birch pollen and pos
itive skin prick test results to the suspected allergen. Using various
levels of specific IgE, we determined the points of the ROC curves wi
th the best ratio of true positives to false positives for four aeroal
lergens (i. e., mite, cat, grass pollen, and birch pollen). Results: E
xcept for mite allergen there was no correspondence between the positi
ve threshold values determined by ROC analysis with MATRIX and PHADEZY
M tests. In terms of class, when ROC analysis was used, the threshold
was class III with the MATRIX test and class II with PHADEZYM test for
cat, grass pollen, and birch pollen and class III for mite. According
to the threshold value, sensitivity and specificity for MATRIX and PH
ADEZYM tests were similar except for cat, for which the PHADEZYM test
had a lower sensitivity (0. 71) than the MATRIX test (0.90). With mite
allergen the predictive value of only one positive MATRIX test result
was 0.96. The specificities of MATRIX and PHADEZYM tests were 100% in
both nonatopic and atopic control groups Conclusions: We conclude tha
t ROC analysis is a reliable method for comparing IgE in vitro tests a
nd that it emphasizes the lack of correspondence between arbitrary uni
ts given for each in vitro test. ROC analysis appears to be a good met
hod for helping to standardize IgE in vitro tests.