Receiver operating characteristics curve analysis of cutaneous skin test reactions to predict hay fever and asthma symptoms in the Normative Aging Study.

Abstract

We used receiver operating characteristic (ROC) curves to determine optimal cutoff values and to evaluate the ability of ragweed, mixed grasses, mixed trees, and house-dust antigens to predict symptoms of hay fever and asthma. Between 1984 and 1993, 1386 members of the Normative Aging Study cohort had at least one examination that included evaluation of cutaneous hypersensitivity to common aeroallergens by skin prick testing. Standard questionnaires based on the American Thoracic Society DLD-78 questionnaire obtained information on respiratory symptoms, illness, and smoking habits. Ragweed was the best univariate predictor for both hay fever and asthma symptoms: the area under the ROC curve for hay fever was 0.71 (95%) CI 0.67-0.74); for asthma, it was 0.65 (95% CI 0.55-0.75). For hay fever, dichotomizing the average reaction to four antigens at 2 mm gave a sensitivity of 56% and a specificity of 83%; at 3 mm, the sensitivity was 38% and specificity 92%. For asthma, dichotomizing the skin test reaction at 2 mm gave a sensitivity ranging from 30% for reaction to mixed grasses to 56% for the ragweed antigen reaction. Although in this older population skin test reactivities did not predict current asthma symptoms with high sensitivity, our results support the use of cutoff values of 2 or 3 mm as commonly used in studies of hay fever symptoms.