Validation of the efficacy of nutritional protocols for suspected cow's milk protein allergy in IPEX syndrome driven by multimodal artificial intelligence

To implement individualized nutritional therapy for immune dysregulation, polyendocrinopathy, enteropathy, X-linked (IPEX) syndrome children with suspected cow's milk protein allergy and explore the clinical effects of nutritional intervention via artificial intelligence methods. A child diagnosed with IPEX syndrome received personalized nutritional intervention: oral short peptide formula showed poor efficacy in treating inflammatory bowel disease (IBD). Considering the possibility of secondary cow's milk protein allergy, the formula was adjusted to an extensively hydrolyzed formula on Day 31st day; blood glucose was controlled using the carbohydrate counting method. The efficacy of nutritional intervention was double-verified by the Bayesian Online Change Point Detection (BOCD) algorithm and the piecewise linear regression model. The analysis method centered on the McNemar test combined with the Generalized Estimating Equation (GEE) model was used to detect the efficacy of blood glucose control. After nutritional intervention, the blood glucose compliance rate was higher than that before the intervention. A significant inflection point in gastrointestinal symptoms appeared between days 32 to 35, the nutritional status was improved, and the therapeutic effect was verified. Clinical nutritionist improve gastrointestinal symptoms and nutritional status in children with IPEX syndrome through personalized nutrition plans, promote disease stability, and lay the foundation for subsequent hematopoietic stem cell transplantation. When biological sample verification is unavailable, artificial intelligence (AI) provides certain guidance for diagnosis and efficacy assessment.