Background: Serum allergen-specific immunoglobulins E (IgE) play a key role in allergy diagnosis along with clinical history and physical examination. Nowadays, allergen multiplex assays allow complex polyallergic cases to be solved as they assess up to 300 allergen-specific IgE. Recently, machine learning has emerged as a trending tool in medicine. The aim was to build a nationwide, open-access database to create an algorithm that could predict allergy diagnosis, severity, category (airborne, food, venom) and culprit allergens. Methods: A retrospective national database was created by the French Society of Allergology in collaboration with AllergoBioNet and the Health Data Hub. Collected data were de-identified patient profiles with five demographic items, twenty clinical items and sIgE results of one allergen multiplex assay. An international crowdsourced machine learning competition was hosted by the Trustii.io platform. Criteria for algorithm evaluation were the F-score (a measure of a model’s accuracy on a dataset) and external validation on patient profiles outside the database (80%-20%, respectively). Results: Data were collected from 4271 patient files. Two hundred and ninety-two data scientists competed with 3135 algorithms. The best F-scores were comprised between 78% and 80%. Models associated with the highest F-scores used gradient boosting classifiers such as LightGBM, CatBoost, XGBoost adapted for tabular datasets with categorical features. Conclusions: We report here the first artificial intelligence models applied to allergen multiplex arrays interpretation in a nationwide real-world database built to be open access. With F-scores close to 80%, the French Allergen Chip Challenge paves the way for a diagnostic prediction tool for practicing allergists.

Background: Oral immunotherapy (OIT) seems to be a promising treatment for hazelnut allergy but the available data is limited. Our objective was to evaluate the proportion of patients desensitised after hazelnut OIT, as well as to evaluate the proportion and type of adverse events (AEs) occurring during OIT and to identify clinical and biological factors predictive of OIT success. Methods: Single-center retrospective study, including pediatrics patients. Oral food challenge (OFC) was performed before and after OIT. For each patient, we collected clinical data and the level of hazelnut IgEs f17, nCor a 9, rCor a 14, at the time of the 2 OFCs. Results: After 12 months of OIT, 52.2% of children were desensitized. Those not desensitized increased their reactogenic dose by 6-fold. One-third presented an AE, mostly non-serious. They were significantly more likely to be desensitized if the levels of hazelnut IgE f17, nCor a 9, and rCor a 14 were initially low. Predictive thresholds of 16.3 kUA/L for hazelnut IgE f17 (p=0.0066) and 4.13 kUA/L for rCor a 14 (p=0.0447) were calculated with an NPV of 85%. Conclusion: Hazelnut OIT is effective in inducing tolerance in 52.2% of children. Children with lower specific IgEs (hazelnut f17, nCor a 9 and rCor a 14) were significantly more likely to achieve desensitization after 12 months of OIT. Predictive thresholds for f17 and rCor a 14 were defined above which the risks of failure of OIT are high.

Background: The aim of this study was to compare the technical and clinical effectiveness of two platforms (Phadia ImmunoCAP™ and Hycor NOVEOS) for the measurement of IgE specific for 10 food allergens. Methods: 289 patients, as part of allergy diagnosis or of their follow-up were included and tested for IgE specific for six food allergen extracts (egg white, cow’s milk, peanut, hazelnut, fish, shrimp) and four molecular allergens (Gal d 1, Bos d 8, Ara h 2, Cor a 14). Specific IgE measurements were carried out using the ImmunoCAP™ and NOVEOS methods. Food allergy diagnosis was established according to international guidelines. Results: A very good correlation (rho>0.9) was present between the two platforms, while specific IgE concentrations measured with NOVEOS were consistently lower (mean -15%) than with ImmunoCAP™. NOVEOS provided higher overall odd-ratios and relative risks for allergen extracts than ImmunoCAP™, but the difference was not significant. When all ten allergens were considered, NOVEOS provided better ROC curves (p=0.03) and thus, had a better ability to establish the true value. Finally, we found that the most discordant results were observed with hazelnut and peanut extracts, and were related to cross-reactive carbohydrate determinants on these two ImmunoCAP™. Conclusions: Specific IgE determination by either ImmunoCAP™ (odd-ratios of allergy = 25.1) or NOVEOS (odd-ratios of allergy = 33.0) is similarly highly informative on the risk of allergy in the selected population. The NOVEOS platform presents the advantage of being less affected by unwanted reactivity due to IgE specific for carbohydrate determinants, while requiring a ten-fold lower test sample volume.