K-Means Clustering Reveals Long-Term Thyrotropin Receptor Antibody Patterns in Graves' Disease: Insights from a...

Background/Objectives: We aimed to explore long-term trajectories of thyroid-stimulating hormone receptor antibody (TRAb) in patients with Graves' disease (GD) and to identify key factors associated with TRAb normalization. We also investigated whether these trajectories correlate with Graves' orbitopathy (GO) comorbidity. Methods: We retrospectively reviewed 403 patients with GD who had an initial TRAb level ≥ 1.5 IU/L between 2010 and 2021, monitoring their TRAb levels for at least 3 years. K-means clustering was performed to categorize patients into distinct TRAb change patterns (A, B, C, D). We employed a Cox regression-based time-to-event model, expressing results as "Survival ratio" rather than the conventional Hazard ratio, to reflect the proportion of patients achieving TRAb normalization over time. Key variables included age, sex, initial TRAb, and GO comorbidity. Results: Four unique TRAb patterns emerged, differing primarily in baseline TRAb levels, duration of GD, and treatment approaches. Pattern A demonstrated the highest TRAb normalization rate (96%), whereas Patterns B (80%), C (29%), and D (13%) showed lower probabilities. Regrouping into A vs. BCD further emphasized the distinct normalization profile of Pattern A. A higher "Survival ratio" was observed in female patients and those with baseline TRAb < 6.14 IU/L. In contrast, patients whose TRAb levels were ≥6.14 IU/L frequently exhibited persistently elevated values over a decade. GO comorbidity did not significantly differ among the four patterns. Conclusions: K-means clustering revealed four unique TRAb change patterns in GD, with baseline TRAb (stratified by the median of 6.14 IU/L) and sex emerging as significant predictors of normalization. These findings highlight the importance of early TRAb monitoring and tailored therapeutic strategies, particularly for those with persistently elevated TRAb levels.