Objective To evaluate the efficacy and safety of tocilizumab (TCZ) in treating active, moderate-to-severe, glucocorticoid-resistant thyroid eye disease (TED), and to develop a nomogram to predict the response to TCZ treatment.Methods This study involving 59 patients with active, moderate-to-severe, glucocorticoid-resistant TED from three medical institutions in China. The primary outcome was the treatment response rate at 12 weeks and 24 weeks, as evaluated by the Clinical Activity Score (CAS). Secondary outcomes included proptosis parameters and the thickness of extraocular muscles measured by orbital magnetic resonance imaging (MRI), diplopia, serum thyroid-stimulating receptor antibody (TRAb) and interleukin-6 (IL-6) levels, and adverse events. Factors influencing the treatment response were identified by multivariate logistic regression, and a nomogram was then created to predict the success or failure of TCZ treatment.Results An absolute response rate of 59.32% (35/59) was detected at 12 weeks and 73.08% (38/52) at 24 weeks after TCZ treatment. CAS significantly decreased from 4 at baseline to 2 following TCZ treatment (p < .001). Compared to that at baseline, the exophthalmometric value was significantly alleviated in both eyes at 12 weeks of TCZ treatment (right eye 22.08 ± 3.09 mm vs. 20.40 ± 2.77 mm, left eye, 22.09 ± 3.26 mm vs. 20.52 ± 2.59 mm, both p < .001). At 24 weeks, a significant decrease of diplopia grade (by one or more grades) from baseline was seen in 45.95% (17/37) of the affected eyes, and a complete resolution was observed in 9 TED patients (p < .05). No serious adverse events were reported during the study period. The multivariate logistic regression analysis showed that smoking history, baseline CAS, proptosis, and free thyroxine (FT4) levels were risk factors for a poor response to TCZ treatment (p < .05). The receiver operating characteristic (ROC) curve analysis demonstrated that the nomogram, created based on these risk factors, generated an area under the curve (AUC) of 0.801 (95% CI: 0.659–0.943) in the training set and 0.792 (95% CI: 0.558–1) in the validation set. Furthermore, the Hosmer-Lemeshow calibration curve showed that the C-statistic in ROC, Brier score, χ2 value and p value of the nomogram were 0.801, 0.171, 6.125 and 0.633 in the training set, and 0.778, 0.167, 6.353 and 0.608 in the validation set, respectively.Conclusion TCZ treatment appears to inhibit clinical activity and proptosis in patients with active, moderate-to-severe, glucocorticoid-resistant TED, with a high safety profile. A nomogram incorporating smoking history, baseline CAS, proptosis, and FT4 levels can help predict the response to TCZ treatment.
Yu et al. (2026) studied this question.