Background: This study aimed to elucidate the relationship between thyroid-related parameters and the prognosis of Graves' disease (GD). Methods: This retrospective cohort study included patients newly diagnosed with GD between January 2009 and December 2010. We applied multivariable Cox proportional hazard models to estimate the hazard ratio (HR) and 95% confidence intervals (CI) for the relationships between the parameters at anti-thyroid drug (ATD) discontinuation and GD relapses. We also developed prediction models for GD prognosis based on parameters at ATD initiation and discontinuation using logistic regression and machine learning algorithm (gradient-boosting machine GBM). Results: The median observation period was 11.3 years. In 1,508 patients who discontinued ATDs after the first course, 1,073 remained in remission, whereas 435 relapsed within a median of 20 months (interquartile range, 10.0 to 49.7). In the Cox model, we found that patients with younger age, higher thyroid stimulating hormone (TSH) receptor antibody (TRAb), and lower TSH levels at ATD discontinuation were more likely to relapse than those without (per 5-year increase in age: HR, 0.92 95% CI, 0.88 to 0.95; per 1-unit increase in TSH: HR, 0.86 95% CI, 0.79 to 0.94; per 1-unit increase in TRAb: HR, 1.15 95% CI, 1.06 to 1.24). Both logistic regression and GBM demonstrated similar predictive performance (area under the curve, 0.61 to 0.63), showing that age and parameters at both ATD initiation and discontinuation contribute to GD relapse. Conclusion: TRAb levels at ATD discontinuation may contribute to the prognosis of GD, but they are not sufficient on their own to predict relapse.
Suzuki et al. (Fri,) studied this question.