Published online: December 19, 2017
The rapid increase in incidence of Eosinophilic Esophagitis (EoE) in the pediatric and adult populations emphasizes the urgent need to improve the current diagnostic strategies for this allergy of the upper gastrointestinal tract. Currently, the gold standard of EoE diagnosis is based on the quantification of tissue eosinophils in biopsies (>15 eosinophils per high power field) in combination with assessment of clinical symptomatology. One of the challenges with this approach in clinical practice is the large inter-patient variability of disease presentation and the lack of predictable markers for therapy response. Therefore, it has become imperative to develop novel diagnostic strategies to allow for a more detailed understanding of the EoE pathology in the individual patient. Individualized EoE diagnosis should further facilitate the development of outcome oriented personalized treatment options.
In an article recently published in The Journal of Allergy & Clinical Immunology (JACI), Sallis and colleagues described a medical algorithm designed to diagnose patients with EoE based on the analysis of individual inflammatory transcript profiles from esophageal biopsies. Supervised machine learning techniques were used to establish a modularly designed algorithm as a tool for primary diagnosis of EoE. An expansion of the program with a secondary analysis loop allowed for the diagnosis of an EoE patient subpopulation.
Based on a disease-specific probability score (p(EoE)), the medical algorithm was able to accurately diagnose EoE with a sensitivity of 91% and a specificity of 93%. The primary diagnosis of equivocal patients that could not be diagnosed or were misdiagnosed with their first biopsy was improved by 84.6%. Using the secondary analysis loop, the authors developed a composite score to evaluate the allergic status of EoE patients. With this score, the authors defined an EoE subpopulation with elevated esophageal IgE production that showed signs of increased Th2-type esophageal inflammation and a strong contribution of mast cells to the phenotype.
Machine learning strategies are state of the art methods to use large patient data sets in clinical practice. This novel medical algorithm can be applied for primary diagnosis of EoE and, with the integration of additional analysis loops, its utility can be expanded to the diagnosis of EoE subpopulations. Going forward, the EoE algorithm will facilitate the discovery of predictive markers for therapy response, which will help to make informed decisions with regards to the treatment of choice for individual patients, thus, this algorithm proves an important first step towards establishing diagnostic decision trees that allow for individualized precision medicine in EoE patients.
The Journal of Allergy and Clinical Immunology (JACI) is an official scientific journal of the AAAAI, and is the most-cited journal in the field of allergy and clinical immunology.