Cut-off values for total serum immunoglobulin E between non-atopic and atopic children in north-west Croatia.

BACKGROUND: The aim of this study was to determine cut-off values for total serum immunoglobulin E (IgE) between non-atopic and atopic children with respiratory symptoms. Children of 0-16 years of age were evaluated for respiratory symptoms of >4-week duration. METHODS: Children were divided into two groups: non-atopic children (n=3355) who were non-IgE-sensitized with undetectable allergen-specific IgE (<0.35 kIU(A)/L), and atopic children (n=4620) who were sensitized to > or =1 allergens (specific IgE > or =0.35 kIU(A)/L). Upper and lower centiles were determined and cut-off values calculated using receiver operating characteristic (ROC) analysis. RESULTS: Serum total IgE increased with age in both groups, although at a variable level and rate, and reached a plateau at 9 and 10 years in non-atopic and atopic children, respectively. Atopic children had on average 14-fold higher serum total IgE compared to non-atopic children. In both groups, the median was lower than the corresponding mean and the distribution skewness was always positive (group I, 0.87; group II, 0.91). In almost all age groups, the 95th percentile for non-atopic children corresponded to the calculated cut-off values, whereas the 10th percentile for atopic children corresponded to the respective cut-off values only until the age of 8 years, after which greater differences between the cut-off values and the 10th percentile were recorded. Cut-off values for total serum IgE in children up to 16 years were determined with diagnostic sensitivity, specificity and area under the ROC curve of 96%, 91% and 0.950, respectively. CONCLUSIONS: The 95th percentile for total IgE in non-atopic children and the 10th percentile in atopic children could be taken as cut-off values in children up to 8 years of age, after which significant percentile discrepancies between non-atopic and atopic children were recorded. Since atopic subjects show a more irregular centile distribution, cut-off values are best determined by ROC analysis.