Dissociation Between Symptom Burden and Structural Pathology in Patients Undergoing Upper Gastrointestinal Endoscopy: A Retrospective Single-Center Study Using Latent Class Analysis

Background and Objectives: Upper gastrointestinal symptoms often correspond poorly with endoscopic and histological pathology. We aimed to identify lifestyle and psychosocial phenotypes and examine their clinical features, objective findings, and symptom–pathology concordance. Materials and Methods: This retrospective single-center study included adults undergoing outpatient upper gastrointestinal endoscopy at a tertiary gastroenterology center in Latvia between 1 May and 31 December 2025. Latent class analysis was applied to questionnaire-derived lifestyle, behavioral, anthropometric, and psychosocial indicators. Classes were subsequently characterized using clinical features, endoscopic findings, and histology. Additional analyses assessed potential confounding by age and sex and exploratory predictive discrimination. Results: The cohort included 2192 patients, of whom 1755 had histology. A three-class solution had the lowest mean Bayesian Information Criterion, with substantial agreement across 20 imputed datasets. Class 1 (46.9%) combined high psychosocial burden with comparatively low substance exposure. Class 2 (12.0%) combined greater substance exposure with high psychosocial burden and the highest anxiety prevalence. Class 3 (41.0%) exhibited lower psychosocial burden, less prolonged sitting, fewer patients with short sleep duration, and the highest combined prevalence of overweight and obesity. The clinical feature composite showed limited diagnostic performance, and correlations between clinical feature burden and pathology burdens were negligible (Spearman’s ρ = −0.012 to 0.026; all p > 0.05). Clinical feature-dominant discordance was most frequent in Class 2 (25.5%), whereas pathology-dominant discordance was most frequent in Class 3 (29.7%). Eleven clinical-feature associations remained significant after age and sex adjustment and false discovery rate correction; no overall associations with individual endoscopic or histological findings met this threshold. Exploratory ROC analyses showed limited discrimination. Conclusions: The phenotypes demonstrated distinct patterns of symptom–pathology dissociation. Lifestyle-based phenotyping provides a framework for characterizing this heterogeneity, while its clinical application requires validation in independent populations.