We sought to assess the value and reliability of serologic testing for predicting clinical food allergy in a population-based cohort of infants with atopic dermatitis (AD). Infants 3-18 months of age, recruited from the general population, were followed quarterly for 3 years and carefully evaluated for evidence of immediate reactions to foods. Specific serum IgE levels for six foods were assayed at 3-5 years. Parents were interviewed at each visit regarding past/current immediate food-specific reactions involving skin, gut or respiratory systems. Data were entered into Excel for calculations of performance characteristics. Nine of the 40 patients (23%) who completed 3 years of follow-up had reactions to one or more foods. Reactions occurred in 5, 11 and 18% to milk, peanut and egg ingestion, respectively. In contrast, 30% of food-specific serum IgE tests were above normal. Predictive reliability of tests was generally low unless values were in the high range for peanut and milk. Conversely, egg allergy could be seen across a nearly full-spectrum of IgE values, making prediction highly unreliable. We conclude that physician and patient misinterpretation of the relevance and reliability of allergy testing may misdirect proper prevention and therapy of AD.