A combined model of BCVA, TRAb, and NLR predicts response to intravenous methylprednisolone in dysthyroid optic neuropathy

Background High-dose intravenous methylprednisolone (HD-IVMP) is the first-line treatment for dysthyroid optic neuropathy (DON). However, a significant proportion of patients exhibit a poor response, risking irreversible vision loss when surgical decompression is delayed. Identifying reliable predictors for failure of HD-IVMP is crucial for optimizing treatment strategies. This retrospective study aimed to identify risk factors associated with a poor response to HD-IVMP in DON patients, develop a predictive model for clinical decision-making, and assist in determining the optimal timing of orbital decompression surgery. Methods A retrospective cohort study was conducted involving 44 patients (44 eyes) diagnosed with DON and treated with HD-IVMP. Basic characteristics, clinical features, orbital computed tomography (CT) findings, and laboratory parameters were analyzed. Treatment response was defined as complete visual recovery without requiring surgical decompression within 6 months. Least absolute shrinkage and selection operator (LASSO) and multivariable logistic regression analyses were used to screen and identify independent predictive factors. The predictive performance of the model was evaluated by plotting the receiver operating characteristic (ROC) curve. Additionally, decision curve analysis (DCA), net reclassification index (NRI), integrated discrimination improvement (IDI), and bootstrap internal validation were used to evaluate the clinical utility of the model. Results At the 6-month follow-up, 20 of the 44 (42.5%) eyes achieved complete visual recovery and consequently avoided surgical decompression. A multivariable analysis identified three independent predictors of non-response to HD-IVMP: poorer baseline best-corrected visual acuity (BCVA; odds ratio OR = 1.43, 95% CI: 1.05–1.95, p = 0.04), higher thyroid-stimulating hormone receptor antibody (TRAb) levels (95% CI: 1.02–2.61, p = 0.04), and elevated NLR (OR = 1.30, 95% CI: 1.04–1.63, p = 0.02). A combined predictive model integrating these factors demonstrated superior performance (AUC = 0.944) with a sensitivity of 88% and a specificity of 85%, exceeding any single predictor. The internal validation also confirmed the model’s robust clinical utility. Conclusion This study establishes that baseline BCVA, TRAb, and NLR are significant and readily available risk factors for failure of HD-IVMP in DON. The developed combination model provides clinicians with a valuable tool for the early identification of patients who are unlikely to benefit from medical therapy alone, thereby facilitating timely surgical intervention to prevent permanent visual damage and improve overall management outcomes. External validation of this model is required in future prospective studies.