Objective  To construct nomogram models based on the Surveillance, Epidemiology, and End Results (SEER) database for predicting the prognosis of patients with papillary thyroid carcinoma (PTC), follicular thyroid carcinoma (FTC), and medullary thyroid carcinoma (MTC), and to evaluate their predictive value for patient prognosis.

Methods  A retrospective cohort study was conducted. The clinicopathological and follow-up data of patients diagnosed with PTC, FTC, and MTC from 2004 to 2015 were obtained from the SEER database. A total of 101476 patients with thyroid cancer, including 78618 females (77.5%) and 22858 males (22.5%) with a median age of 48 years (range: 18–90 years), were included. Among these patients, 93942 had PTC, 5699 had FTC, and 1835 had MTC. Through stratified random sampling, the patients in each of the three groups were divided into a training set and a validation set at a ratio of 7∶3. The PTC group had 65762 patients in the training set and 28180 in the validation set; the FTC group had 3991 patients in the training set and 1708 in the validation set; and the MTC group had 1287 patients in the training set and 548 in the validation set. Univariate and multivariate Cox proportional hazards regression models were used to screen independent prognostic factors for patients with thyroid cancer, and nomogram models based on these factors were constructed. Time-dependent receiver operating characteristic (ROC) curves, decision curve analysis (DCA), and calibration curves were used to evaluate the predictive performance of the nomogram models.

Results  Univariate and multivariate Cox proportional hazards regression analyses showed that age, sex, marital status, race, annual household income, time from diagnosis to treatment, maximum tumor diameter, extrathyroidal extension, American Joint Committee on Cancer (AJCC) stage, T stage, N stage, M stage, surgical method, and number of lymph nodes removed were independent prognostic factors for overall survival (OS) and cancer-specific survival (CSS) in patients with PTC. Nonprimary site surgery was also an independent prognostic factor for OS in patients with PTC. Age, sex, marital status, annual household income, maximum tumor diameter, AJCC stage, T stage, N stage, M stage, surgical method, number of lymph nodes removed, and nonprimary site surgery were independent prognostic factors for OS in patients with FTC. Age, marital status, AJCC stage, N stage, M stage, surgical method, number of lymph nodes removed, and nonprimary site surgery were independent prognostic factors for CSS in patients with FTC. Age, marital status, time from diagnosis to treatment, maximum tumor diameter, extrathyroidal extension, AJCC stage, T stage, M stage, and surgical method were independent prognostic factors for OS in patients with MTC. Age, maximum tumor diameter, extrathyroidal extension, AJCC stage, T stage, and M stage were independent prognostic factors for CSS in patients with MTC. The areas under the ROC curves of the constructed nomogram models based on these independent prognostic factors for predicting OS and CSS in patients with PTC, FTC, and MTC in the training and validation sets ranged from 0.791 to 0.958, with all values close to or greater than 0.8, indicating good predictive performance. DCA and calibration curves suggested that the nomogram models had good clinical applicability and satisfactory calibration.

Conclusion  The constructed nomogram models based on clinicopathological features from the SEER database can effectively predict OS and CSS in patients with thyroid cancer and may provide a reference for prognostic evaluation and individualized treatment.