Inflammatory Endotypes of Chronic Adenoiditis and Their Impact on Persistent Middle Ear Dysfunction: A 2-Year Retrospective Translational Study Integrating Clustering and Machine Learning Approaches

<i>Background and Objectives</i>: Chronic adenoiditis is a major contributor to persistent middle ear dysfunction (PMED) in children; however, clinical evolution varies considerably despite similar anatomical obstruction. This study aimed to identify inflammatory endotypes of chronic adenoiditis using unsupervised clustering and to evaluate their association with PMED through mechanistic and predictive modeling. <i>Materials and Methods</i>: A retrospective cohort of 236 children (3-12 years) with chronic adenoiditis and otitis media with effusion was analyzed. Clinical, endoscopic, audiological, and hematologic inflammatory parameters (eosinophils, NLR, ELR, CRP, IgE) were included. K-means clustering identified inflammatory endotypes. Associations with PMED at six months were evaluated using multivariate logistic regression and mediation analysis. Predictive performance was compared using logistic regression, random forest, and gradient boosting models, with SHAP-based interpretability and decision curve analysis. <i>Results</i>: Three distinct endotypes were identified: eosinophilic (28%), neutrophilic (41%), and fibrotic-obstructive (31%). PMED occurred in 44% of the fibrotic endotype compared with 22% in the eosinophilic group (<i>p</i> < 0.001). In multivariate analysis, the fibrotic endotype independently predicted PMED (OR = 3.48, 95% CI 1.92-6.31), alongside PTA > 30 dB (OR = 2.91) and NLR > 3.5 (OR = 2.36). Mediation analysis showed that hearing impairment accounted for 34% of the effect of anatomical obstruction on persistence. Gradient boosting achieved superior discrimination (AUC = 0.90) and demonstrated the highest net clinical benefit. <i>Conclusions</i>: Chronic adenoiditis comprises biologically distinct inflammatory endotypes with differential risk of persistent middle ear dysfunction. Integrating inflammatory profiling with machine learning enhances mechanistic understanding and risk stratification, supporting precision-based management in pediatric otorhinolaryngology.