A nomogram based on baseline predictors for Graves' disease relapse after antithyroid drug withdrawal: a retrospective study

To develop and validate a practical prediction model for relapse of Graves' disease within one year after discontinuation of antithyroid drugs, using routinely available baseline characteristics, and to compare its performance with an existing score.

This single-center retrospective cohort study enrolled 654 adults with Graves' disease who achieved remission after at least 12 months of antithyroid drug therapy and subsequently discontinued treatment. The cohort was randomly split into training (70%) and testing (30%) sets. Predictors were selected by least absolute shrinkage and selection operator (LASSO) regression, and a multivariable logistic regression model was built on multiply imputed data. Model performance was assessed by discrimination, calibration, and decision curve analysis and was benchmarked against the Graves Recurrent Events After Therapy (GREAT) score. A nomogram was constructed to visualize individualized risk.

Five independent baseline predictors emerged: younger age, male sex, goiter grade 2-3, higher free thyroxine, and higher thyrotropin receptor antibody. The model demonstrated good discrimination, with an area under the receiver operating characteristic curve of 0.84 in the testing set, which showed discrimination comparable to that of the GREAT score (0.82; DeLong P = 0.31). Across quartiles of predicted probability, the observed relapse rate increased from 23.4% to 97.9% (test for trend p < 0.001). Calibration was satisfactory, and decision curve analysis indicated net clinical benefit across relevant threshold probabilities. The nomogram translates these five variables into a single probability of relapse. An interactive online calculator was concurrently developed to facilitate clinical use.

Readily accessible clinical and laboratory parameters can identify patients at elevated risk of early relapse after antithyroid drug withdrawal for Graves' disease. The proposed nomogram offers a simple, quantitative tool to support personalized treatment discussions and shared decision-making, although external validation in diverse populations is necessary before clinical implementation.