Abstract
BACKGROUND: Differentiated thyroid carcinoma (DTC) originates from abnormal follicular cells and accounts for approximately 90-95% of thyroid malignancies. The diagnosis of radioiodine refractory DTC (RR-DTC) is based on clinical evolution and iodine uptake characteristics rather than pathological characteristics. Thus, it takes a long time to become apparent, and the definition of RR-DTC covers multiple aspects. We aimed to analyze the clinical and molecular imaging characteristics of patients with RR-DTC and identify independent predictors to develop an RR-DTC scoring system and a simple nomogram for predicting the probability of RR-DTC. We reviewed the data of 404 patients with metastatic DTC who underwent both post-RAI WB therapy scintigraphy and (18)F-fluorodeoxyglucose ((18)F-FDG) positron emission tomography/computed tomography. Data on the clinical features and molecular characteristics of RR-DTC and non-RR-DTC cases were obtained from medical records. We screened for predictors using univariate analyses, obtained independent predictors through multivariate analyses, and then established a scoring system and a simple nomogram for predicting RR-DTC according to the corresponding odds ratio (OR) values. RESULTS: Diagnosis at age ≥ 48 years (OR, 1.037; 95% confidence interval [CI], 1.007-1.069), recurrence between the operation and iodine-131 treatment (OR, 7.362; 95% CI 2.388-22.698), uptake of (18)F-FDG (OR, 39.534; 95% CI 18.590-84.076), and the metastasis site (OR, 4.365; 95% CI 1.593-11.965) were highly independently associated with RR-DTC. We established a scoring system for predicting RR-DTC, showing that the area under the receiver operating characteristic curve (AUC) with a cutoff value of 10 points (AUC = 0.898) had a higher discernibility than any other single independent predictor. The risk factors of RR-DTC in nomogram modeling include diagnosis at age ≥ 48 years, recurrence between the operation and iodine-131 treatment, uptake of (18)F-FDG, and the site of metastasis. The concordance index (c-Index) of the nomogram was 0.9. CONCLUSIONS: We demonstrated that a predictive model based on four factors has a good ability to predict RR-DTC. An index score ≥ 10 points was found to be the optimal index point for predicting RR-DTC. Moreover, this nomogram model has good predictive ability and stability. This model may help establish an active surveillance or appropriate treatment strategy for RR-DTC cases.