Objective To investigate the predictive value of dynamic changes in thyrotropin receptor antibody （TRAb） titers combined with ultrasound elastography parameters for predicting recurrence risk in Graves' disease （GD）. Methods A total of 306 patients with GD admitted between January 2019 and November 2021 were enrolled. All patients received standard antithyroid drug treatment and were followed up. Dynamic changes in TRAb titers during treatment were monitored and the patients were divided into a continuously decreasing group and a dynamically increasing group based on the trend of change. Ultrasound elastography was used to measure thyroid volume， elasticity score， mean strain value （MEAN）， and blue area percentage （%AREA）. Independent risk factors for GD recurrence were identified via multivariate logistic regression analysis， and the predictive performance of indicators was evaluated using receiver operating characteristic （ROC） curves. Results A total of 14 cases were lost to follow-up， and 292 cases were eventually included to complete follow-up， including 103 cases in the dynamic increase group and 189 cases in the continuous decrease group. TRAb levels diverged significantly between groups starting at 12 months of treatment. The persistent-decrease group maintained a downward trend， remaining low at follow-up termination. In contrast， the dynamic-increase group showed a slow decline， reaching a nadir at 18 months before rising again， with sustained elevation during follow-up. By the end of follow-up， 87 patients （29.79%） experienced recurrence. The recurrence rate was significantly higher in the dynamic-increase group （66.02%， 68/103） than in the persistent-decrease group （10.05%， 19/189； χ² =99.831， P < 0.05）. Multivariate logistic regression identified TRAb titer elevation， elasticity score （3 ? 4）， elevated %AREA， and increased thyroid volume as independent risk factors for recurrence （P < 0.05）. ROC analysis showed area under the curve （AUC） values of 0.823 for TRAb dynamic elevation， 0.652 for elasticity score （3 ? 4）， 0.801 for %AREA elevation， and 0.724 for thyroid volume enlargement. The ultrasound-combined model achieved an AUC of 0.888， while the four-indicator combined model demonstrated the highest predictive accuracy （AUC = 0.913）. Conclusions Dynamic TRAb elevation and ultrasound elastography parameters are independent predictors of GD recurrence. Their combination enhances the accuracy of recurrence risk prediction.