Multidimensional risk prediction model for infant atopic dermatitis: A prospective cohort study

Objective: This study aimed to construct a multi-dimensional risk prediction model for atopic dermatitis (AD) by integrating the maternal-fetal immune axis, genetic risk factors, and environmental exposures. Methods: The study prospectively enrolled 503 full-term newborns, with parental allergic history collected via questionnaire, maternal/cord blood biomarkers (IL-4, IL-13, IL-31, IL-33, IgE, TSLP) quantified by ELISA, and dust mite exposure dynamically assessed through quarterly standardized sampling. A 1-year follow-up was conducted to assess AD incidence in the neonatal cohort. Variables were screened via univariate analysis, LASSO regression and multivariable logistic regression to construct a nomogram model, with performance evaluated by ROC curve, calibration curve, Hosmer-Lemeshow test, triple cross-validation (repeated 10-fold, leave-one-out and bootstrap), and decision curve analysis. Results: A total of 456 infants were finally included (106 infants in the AD group and 350 infants in the non-AD group). Through a multi-stage screening process, 6 risk factors were identified, including cord blood IgE and TSLP, maternal blood IL-4 and IL-33, mother with allergic history, and dust mite exposure levels; subsequently, a predictive model was constructed based on these factors. Upon evaluation, the model showed good discriminatory ability, calibration degree, robustness, and clinical applicability. Conclusions: Cord blood IgE and TSLP, maternal blood IL-4 and IL-33, mother with allergic history, and dust mite exposure levels have shown good predictive value for AD. However, multicenter studies will be required to verify the universality of the model.