Exploring Clinical Determinants and Machine Learning–Based Prediction of Disease Outcomes in Graves Ophthalmopathy

Background Graves ophthalmopathy (GO) exhibits heterogeneous clinical behavior, and predicting disease severity and course remains challenging. This study aimed to evaluate clinical determinants of GO using conventional statistical analyses and machine learning (ML) approaches to improve prognostic prediction. Methods Medical records of 153 patients with GO were retrospectively reviewed. Demographic characteristics, clinical findings, thyroid function, autoantibody levels, imaging features, and smoking status were analyzed. Patients were classified according to disease severity, activity, onset pattern, and tissue predominance. In addition to classical statistical tests, logistic regression, random forest, support vector classifier (SVC), and k-nearest neighbors (k-NN) algorithms were trained and evaluated using accuracy, F1-score, and AUC metrics. Results Male sex and active smoking were significantly associated with higher disease severity and activity. Muscle-predominant GO was associated with older age, male sex, smoking, higher severity, and higher activity. No significant differences were observed among clinical subgroups regarding thyroid hormone levels or autoantibody titers. Among ML models, the SVC showed the best performance for predicting severity (AUC = 0.818); Random Forest best estimated tissue predominance (AUC = 0.662), while logistic regression showed the highest accuracy for onset prediction. Smoking was the strongest predictor of EUGOGO-based severity, whereas age was most influential in NOSPECS-based assessments. Conclusion Demographic and clinical factors, particularly smoking, sex, age, and muscle involvement, appear to be stronger determinants of GO course than thyroid-related biochemical parameters. Machine learning approaches demonstrated meaningful discriminatory capability for predicting complex clinical outcomes and may contribute to future risk stratification frameworks.