This study aimed to develop an autoimmune thyroid disease (AITD) risk prediction model for patients with vitiligo based on readily available characteristics.
A retrospective analysis was conducted on the clinical characteristics, demographics, skin lesions, and laboratory test results of patients with vitiligo. To develop a model to predict the risk of AITD, the Least Absolute Shrinkage and Selection Operator (LASSO) method was used to optimize feature selection, and logistic regression analysis was used to select further features. The C-index, Hosmer–Lemeshow test, and decision curve analysis were used to evaluate the calibration, discrimination ability and clinical utility of the model. Internally, the model was verified using bootstrapping; externally, two independent cohorts were used to confirm model accuracy.
Sex, vitiligo type, family history of AITD, family history of other autoimmune disease, thyroid nodules or tumors, negative emotions, skin involvement exceeding 5% of body surface area, and positive immune serology (IgA, IgG, IgM, C3, and C4) were predictors of AITD in the prediction nomogram. The model showed good calibration and discrimination (C-index: 0.746; 95% confidence interval: 0.701–0.792). The accuracy of this predictive model was 74.6%.In both internal validation (a C-index of 1000 times) and external validation, the C-index outperformed (0.732, 0.869, and 0.777). The decision curve showed that the AITD nomogram had a good guiding role in clinical practice.
The novel AITD nomogram effectively evaluated the risk of AITD in patients with vitiligo.