Multidimensional Predictive Model for Assessing Clinical Activity in Thyroid Eye Disease

Yang Li^1,2,3Guang-Hong Zhang^1,2,4Man Tian^1,2,3Chuan Hua^1,2,3,5Jian-Ping Zhai^1,2,6Yan- Qiong He^1,2,6*Xin-He Zuo^1,2,3,5*

• ¹Thyroid Center of Hubei Provincial Hospital of Traditional Chinese Medicine, Wuhan, China
• ²Affiliated Hospital of Hubei University of Chinese Medicine, Wuhan, China
• ³Hubei Shi zhen Laboratory, Wuhan, China
• ⁴Department of ophthalmology, Hubei Provincial Hospital of Traditional Chinese Medicine, Wuhan, China
• ⁵Department of Hepatology, Hubei Key Laboratory of The Theory and Application Research of Liver and Kidney in Traditional Chinese Medicine, Hubei Provincial Hospital of Traditional Chinese Medicine, Wuhan, China
• ⁶Department of Nuclear Medicine, Hubei Provincial Hospital of Traditional Chinese Medicine, Wuhan, China

Objective: Thyroid Eye Disease (TED) is an autoimmune disorder with complex inflammatory activity that remains challenging to assess accurately. Current method, mainly the Clinical Activity Score (CAS), exhibits limitations in objectivity and comprehensiveness. This study aimed to develop a multidimensional predictive model integrating clinical parameters, SPECT/CT imaging data, and serum biomarkers, to improve TED activity evaluation.Methods: This retrospective research included 36 TED patients (72 eyes) diagnosed by EUGOGO criteria who underwent SPECT/CT examination. The Clinical Activity Score (CAS) was used to evaluate inflammatory activity. Variables with significant associations with CAS-defined activity were identified using univariate analysis, and Bayesian shrinkage regression (BSR) and the least absolute shrinkage and selection operator (LASSO) were utilized for variable selection in the primary cohort. Predictive models were constructed and evaluated using receiver operating characteristic (ROC) curves (internally validated via five-fold cross-validation), decision curve analysis (DCA), and calibration curves.Results: Five predictive models were constructed. The comprehensive Model 4, combining clinical, imaging (EX, maximal SPECT/CT uptake ratio [URmax]), and serum biomarkers (TRAb, RBC), achieved superior diagnostic accuracy (AUC: 91.18%; sensitivity: 0.91; specificity: 0.86).Model 5, retaining variables significant in univariate and multivariate analyses, demonstrated robust performance (AUC: 85.97%) with superior stability during cross-validation (ROC mean: 0.8417).Key predictors included male sex (OR = 11.74), TRAb levels, EX, URmax, and RBC count.SPECT/CT-derived URmax correlated strongly with disease activity, while serum biomarkers complemented imaging limitations.Conclusion: Multidimensional integration of clinical, imaging, and biomarker data significantly enhances TED activity evaluation compared to single-modality approaches. The multidimensional model offers superior diagnostic accuracy, addressing the limitations of conventional methods. These findings advocate for a holistic approach in TED management.