Predicting food allergy: The value of patient history reinforced.

Abstract

EAACI guidelines emphasize the importance of patient history in diagnosing food allergy (FA) and the need for studies investigating its value using standardized allergy-focused questionnaires. To determine the contribution of reaction characteristics, allergic comorbidities and demographics to prediction of FA in individuals experiencing food-related adverse reactions. Adult and school-age participants in the standardized EuroPrevall population surveys, with self-reported FA, were included. Penalized multivariable regression was used to assess the association of patient history determinants with "probable" FA, defined as a food-specific case history supported by relevant IgE sensitization. In adults (N = 844), reproducibility of reaction (OR 1.35 [95% CI 1.29-1.41]), oral allergy symptoms (OAS) (4.46 [4.19-4.75]), allergic rhinitis (AR) comorbidity (2.82 [2.68-2.95]), asthma comorbidity (1.38 [1.30-1.46]) and male sex (1.50 [1.41-1.59]) were positively associated with probable FA. Gastrointestinal symptoms (0.88 [0.85-0.91]) made probable FA less likely. The AUC of a model combining all selected predictors was 0.85 after cross-validation. In children (N = 670), OAS (2.26 [2.09-2.44]) and AR comorbidity (1.47 [CI 1.39-1.55]) contributed most to prediction of probable FA, with a combined cross-validation-based AUC of 0.73. When focusing on plant foods, the dominant source of FA in adults, the pediatric model also included gastrointestinal symptoms (inverse association), and the AUC increased to 0.81. In both adults and school-age children from the general population, reporting of OAS and of AR comorbidity appear to be the strongest predictors of probable FA. Patient history particularly allows for good discrimination between presence and absence of probable plant FA.