To the Editor, Primary eosinophilic gastrointestinal disorders (EGIDs) are a spectrum of emerging inflammatory diseases, which may involve any part of the gastrointestinal (GI) tract and lead to a pathological… Click to show full abstract
To the Editor, Primary eosinophilic gastrointestinal disorders (EGIDs) are a spectrum of emerging inflammatory diseases, which may involve any part of the gastrointestinal (GI) tract and lead to a pathological eosinophilic mucosal infiltration.1,2 Based on the anatomical site of the eosinophil inflammation, EGIDs are classified into eosinophilic esophagitis (EoE) and nonesophageal EGIDs. There is increasing interest in EGIDs heterogeneity related to clinical presentation, comorbidities, natural history, and response to therapies. To date, no studies stratifying pediatric patients with EGIDs into clinical phenotypes with a datadriven approach have been published. This study aimed to characterize EGIDs heterogeneity by performing cluster analysis on a cohort of children and adolescents followed at the Pediatric Center for Eosinophilic Gastrointestinal Disorders (CPED) in Pavia, Italy, using an extensive pediatric primary care database from our University Hospital. Diagnosis of EoE was based on the finding of ≥15 eosinophils/ highpower field (HPF) in at least one esophageal biopsy.3 As there are no consensus guidelines for the diagnosis of nonesophageal EGIDs, pathology reports were reviewed based on the pathological cutoffs proposed by Collins et al.4 All patients with a secondary cause of pathological eosinophilic inflammation (ie, inflammatory bowel diseases, parasite infections, intestinal vasculitis, and malignancies) of the GI tract were excluded. Data collected from enrolled EGIDs patients included demographics (date of birth, age at diagnosis, gender, and ethnicity), early life history (gestational age, birth weight, delivery mode, neonatal intensive care unit [NICU] admission, exclusive breastfeeding for the first three months of life, and bronchiolitis), early environmental tobacco smoke (ETS) exposure, medical history of coexisting atopic (allergic rhinitis, asthma, atopic dermatitis, and food allergy), and nonatopic diseases (congenital and genetic diseases, autoimmune diseases, connective tissue disorders, neuropsychiatric disorders, and recurrent respiratory infections), and symptoms at the time of diagnosis. All patients underwent skin prick tests for foods (milk, egg, soy, rice, codfish, shrimp, almond, hazelnut, walnut, peanut, tomato, kiwi, and peach) and inhalant allergens (dust mites, grass, birch, hazel, molds, cat, dog, mugwort, and ragweed). Laboratory data included serum total immunoglobulin E (IgE) and peripheral blood eosinophil count. According to the current validated reference score (EREFS), endoscopic findings were reported, and esophageal mucosa was considered pathological when the EREFS total score was ≥2. For nonesophageal EGIDs, endoscopic findings were deemed abnormal when macroscopic alterations (mucosal hyperemia, edema, erosions, or nodular lymphoid hyperplasia) were reported. Data about treatments initiated at the time of diagnosis were also collected, including medications (corticosteroids and protonpump inhibitors [PPIs]) and food elimination diets. All data were extracted from electronic medical records (OrmawebTM and FenixTM, Software). Every patient identifier (name and surname) was replaced with a specific numeric code. The Ethics Committee approved this study (protocol number 2021.311/457, GOLDEN study, identifier: NCT05219903). A total of 31 variables were used as input for the clustering algorithm. Gower's general dissimilarity coefficient was used to compute the distance matrix since the input variables were of mixed types (numerical and categorical).5 Then, we applied the “partitioning around medoids” algorithm, increasing the number of candidate clusters from one to five.6 The Silhouette statistic was used to determine the optimal number of clusters (the larger, the better).7 For each input variable, we used the Fisher's exact test (categorical variables) or the KruskalWallis test (numerical variables) to perform pairwise comparisons between the clusters. Holm's method was used to adjust the pvalues for multiple comparisons. The statistical analyses were performed through R version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was set at p < 0.05. The study population comprised 60 patients (73% males); 38 (63%) subjects had EoE and 22 (37%) had nonesophageal EGIDs (Table S1). According to the Silhouette statistics, the best data partition comprised three clusters. Pairwise association tests identified 13 distinctive features (Table 1, Figure 1). Cluster 1 (38%): EoE diagnosis with pathological endoscopic findings (EREFS > 2), common allergic rhinitis and allergic sensitization, especially to dust mites, grass and hazel tree, and peanut (Table S2), epigastric/abdominal pain without diarrhea, topical corticosteroid therapy and PPIs use, and infrequent NICU admission. Cluster 2 (27%): nonesophageal EGIDs diagnosis with normal endoscopic findings, diarrhea, rare gastroesophageal reflux disease (GERD), and infrequent PPIs use. Cluster 3 (35%): EoE diagnosis with pathological endoscopic findings (EREFS > 2), GERD, ETS exposure, NICU admission, PPIs use, infrequent allergic sensitization, infrequent abdominal pain and diarrhea, infrequent corticosteroid use.
               
Click one of the above tabs to view related content.