Development and validation of a nomogram for predicting non-response to tocilizumab in thyroid eye disease: a retrospective study

PURPOSE: This study aims to integrate single photon emission computed tomography/computed tomography (SPECT/CT) parameters with other biomarkers to develop a predictive model for assessing the risk of non-response to tocilizumab (TCZ) therapy in patients with thyroid eye disease (TED). METHODS: Univariate Cox regression, least absolute shrinkage and selection operator (LASSO) method, and multivariate Cox regression were used to identify predictors of non-response to TCZ and construct a nomogram. The model’s performance was validated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). RESULTS: In this retrospective study, 195 patients with moderate-to-severe TED who received subcutaneous TCZ were included. The nomogram included four predictors: intraocular pressure (IOP), thyroglobulin (TG) level, maximum standardized uptake value (SUVmax) of extraocular muscles (EOMs), and SUVmax of the lacrimal glands. In the training set, the model achieved 1-, 2-, and 3-month area under the ROC curve (AUC) values of 0.787 (95% CI: 0.681–0.893), 0.747 (95% CI: 0.641–0.853), and 0.834 (95% CI: 0.734–0.934), respectively. In the validation set, the corresponding AUC values were 0.761 (95% CI: 0.567–0.955), 0.799 (95% CI: 0.663–0.934), and 0.833 (95% CI: 0.655-1). Calibration curves demonstrated good agreement between predicted and observed probabilities, while DCA confirmed that both the nomogram and individual predictors provided varying degrees of clinical net benefit. CONCLUSIONS: This study successfully integrated SPECT/CT parameters with other biomarkers to develop a nomogram for predicting non-response to TCZ in moderate-to-severe TED. The model, incorporating four predictive factors, may help personalize treatment strategies and improve clinical decision-making for TED patients.