Predicting Cow's Milk Challenge Outcomes in Children: Multivariate Analysis of Clinical Predictors.

Introduction: Oral food challenges are the gold standard for diagnosis and reactivity thresholds but are resource-intensive and high-risk for reactions. Limited data on factors associated with increased risk of positive oral food challenges exist. We aimed to assess factors associated with positive oral food challenges and create a model to predict cow milk oral food challenge outcomes.

Methods: Children aged 5-18 being considered for cow's milk oral immunotherapy underwent a single-blind, placebo-controlled food challenge to cow's milk, with either positive (reaction) or negative (tolerance) outcomes. Initial factors recorded included sex, age, history of asthma, eczema, allergic rhinitis, prior epinephrine use for cow's milk-induced reactions, skin prick test size, serum levels of IgE antibodies to α-lactalbumin, β-lactoglobulin, and casein and log-transformed values. Stepwise backward multivariate Firth bias-reduced logistic regression was used to create the final model, and performance was assessed with ROC curves.

Results: 111 children underwent an oral food challenge, 103 patients reacted, and 8 tolerated the challenge. Univariate analysis showed skin prick test size, previous epinephrine use, history of asthma, and log-transformed α-lactalbumin, β-lactoglobulin, and casein were significantly associated with positive oral food challenge. The multivariate model included two factors: log-transformed casein (aOR 2.4; 95%CI 1.4-5.9; p<0.001) and previous epinephrine use (aOR 6.5; 95%CI 1.2-68.0; p=0.03). The final model showed good discriminatory performance (AUC 0.928 (95% CI 0.83 - 0.98). In comparison, a univariate model using only skin prick test (OR 1.44, 95%CI: 1.1-2.0; p=0.002) had worse discriminatory performance (AUC 0.83 (95% CI: 0.64-0.93).

Conclusion: The study suggests that logistic multivariate models, including log-transformed casein and previous epinephrine use, may help predict oral food challenge outcomes in pediatric patients. Future studies are needed to validate this with larger datasets.