The following is a summary of “Reproducibility and Generalizability of the Web-Based Tool to Predict Lamina Propria Fibrosis in Eosinophilic Esophagitis,” published in the July 2023 issue of the Pediatric Gastroenterology and Nutrition by Hiremath et al.

Researchers conducted an examination of the reproducibility and generalizability of the newly developed web-based model for predicting lamina propria fibrosis (LPF) in esophageal biopsies with insufficient lamina propria (LP) from patients diagnosed with eosinophilic esophagitis (EoE) using a separate dataset consisting of 183 cases. 

For the grade and stage scores of LPF, the AUC (area under the curve) for the predictive model was 0.77 (0.69–0.84) and 0.75 (0.67–0.82), and its accuracy was 78% and 72%, respectively. The model performance metrics exhibited similarities to those of the original model. A notable positive correlation was observed between the predictive probability of the models and the grade and stage of LPF evaluated by a pathologist (grade: r² = 0.48, P < 0.001; and stage: r² = 0.39, P < 0.001). 

These findings provide evidence for the reliability and applicability of the online-based model in predicting the occurrence of low-grade dysplasia in esophageal biopsies with insufficient lamina propria in eosinophilic esophagitis. Further investigations are necessary to enhance the web-based prognostic models to offer predictive probabilities for the sub-scores of LPF severity.

Source: journals.lww.com/jpgn/Abstract/2023/07000/Reproducibility_and_Generalizability_of_the.14.aspx