Construction and clinical application of a predictive model for the efficacy of ¹³¹I therapy in refractory Graves' disease in children and adolescents.

Abstract

Background: Radioactive iodine (RAI) is a crucial treatment for refractory pediatric and adolescent Graves' disease (GD) cases characterized by poor response to antithyroid drug (ATD), recurrence, adverse effects, or excessive thyroid enlargement. Compared to adults, pediatric and adolescent patients currently lack effective efficacy prediction models tailored to this specific population. Furthermore, the applicability of existing adult dose calculation standards to children remains controversial. This study aims to develop a predictive model and explore the cutoff values of key variables influencing treatment efficacy, thereby providing a new adjunctive tool for the clinical management of pediatric and adolescent GD patients.

Methods: A total of 95 children and adolescents with GD who received RAI therapy at the Department of Nuclear Medicine, the Second Affiliated Hospital of Nanchang University, between January 2020 and May 2025 were initially screened. Following the exclusion of 13 patients who had undergone prior thyroid surgery, received non-initial RAI treatment, or lacked regular follow-up, 83 patients (aged 9-19 years; 58 females) were finally enrolled. Based on thyroid function status 6 months post-treatment, patients were categorized into the cured group (Clinical Cure and Hypothyroidism) and the uncured group (Ineffective and Partial Improvement). Univariate analysis and LASSO regression were employed for preliminary variable screening to identify predictive indicators. A multivariate logistic regression model was subsequently constructed based on these selected variables. Model performance was evaluated using the area under the curve (AUC), calibration curves, and decision curve analysis (DCA), with internal validation performed via Bootstrap (2000 resamplings). The optimal cutoff values were determined using the Youden index.

Results: The proposed dose per gram of thyroid tissue was a protective factor (OR = 1.048, 95% CI (1.01-1.09), P = 0.018); Thyroid weight (OR = 0.956, 95% CI (0.924-0.983), P = 0.004) and TRAb (OR = 0.914, 95% CI (0.861-0.963), P = 0.001) were risk factors. The model achieved an AUC of 0.833 on the original dataset, and an adjusted AUC of 0.810, After model establishment, further analysis of key variables revealed significantly reduced cure rates in patients with thyroid weight > 55.52 g, TRAb > 13.8 IU/L, and the proposed dose per gram of thyroid tissue < 97.5 µCi/g (3.61 MBq/g).

Conclusion: A model incorporating the proposed dose per gram of thyroid tissue, TRAb, and thyroid weight effectively predicts treatment efficacy. For high-risk patients (thyroid weight > 55.52 g, TRAb > 13.8 IU/L), increasing the proposed dose (≥ 97.5 µCi/g or 3.61 MBq/g) is recommended to enhance treatment outcomes.