Establishment And Evaluation Of Recurrence Risk Model Of Graves’ Disease After Antithyroid Drugs Therapy

Objectives:To investigate the risk factors for recurrence of initial Graves’ disease(GD)after antithyroid drugs(ATD)withdrawal and establishing a risk prediction model to provide guidance for precise treatment of GD patients.Methods:Among recruited 308 patients with newly onset GD taking ATD from 2017 to 2019,170 patients who completed follow-up were enrolled and divided into relapse and remission group according to whether hyperthyroidism reoccurred within 2 years after ATD withdrawal in order to establish the discovery cohort.An internal validation cohort was constructed by repeating the sampling with bootstrap.Using Cox regression analysis to screen the risk factors and establish a prediction model.The differentiation and accuracy of the new model were evaluated and compared with the GREAT score.Results:Ninety GD patients(52.9%)relapsed within 2 years after ATD withdrawal.According to Cox regression analysis,family history of GD,youth(<30 years),severe goiter(degree Ⅱ-Ⅲ),high level of TRAB(≥13 mIU/ml),large thyroid volume(≥26.4 cm3)and low 25(OH)D(<14.7 ng/ml)were included in the Cox regression equation:h(t/X)=h0(t)exp(0.672 ×family history+0.405×age+0.491 xsevere goiter+0.808×TRAb+1.423×thyroid volume+0.579x25(OH)D).AUC in the discovery cohort and validation cohort were 0.857 and 0.892,and C-index were 0.777 and 0.761,respectively,which suggest that the new model had a high ability to distinguish high or low risk of recurrence after ATD withdrawal.As for the evaluation of model accuracy,in the discovery cohort,the predicted value was basically consistent with the actual observed value.In the validation cohort,the predicted value of the GD recurrence event was slightly higher than the actual observed value.The Brier scores in two cohorts were 0.144 and 0.136,indicating that the accuracy of the new model was fine.The AUC of the new model in both discovery and validation cohorts were greater than GREAT score,and the NRI increased significantly,which suggested that the prediction effect of the new model was better than GREAT score.Conclusions:The model can be used to evaluate the recurrence risk within 2 years for patients with newly onset GD after ATD withdrawal,which may facilitate clinicians to reasonably select treatment modalities in order to improve the remission rate of GD.