Development and Validation of Prognostic Models in Patients With Stage IV Thyroid Cancer Undergoing Surgical Treatment

Abstract

Background

The low survival rate among patients with advanced thyroid cancer implies a dearth of effective treatment strategies. Research on advanced thyroid cancer has focused on targeted drugs, while neglecting the prognostic prediction for patients with advanced thyroid cancer who are undergoing surgical treatment. Consequently, further investigation into the prognostic implications of surgery for patients with advanced thyroid cancer is of great significance.

Methods

This study retrieved and screened all stage IV thyroid cancer patients between 2010 and 2020 in the SEER database, and finally included 670 patients in the study. Independent risk factors, which were used to develop a nomogram model predicting prognosis, were screened using univariate and multivariate Cox regression analyses. Furthermore, the log-rank test was used to compare differences in the Kaplan-Meier curves of overall survival (OS) and cancer-specific death (CSS) in different risk factor groups.

Result

Survival analysis was performed based on the OS and TCSS of the surgical and non-surgical groups, and the Kaplan-Meier curves were statistically significant. Cox multivariate analysis showed statistical significance in grade, T stage, surgery, and radiotherapy. When statistically significant variables were incorporated into the construction of the prediction model, the C-index of the training group was 0.857, and the areas under the ROC curve of the modeling group were 0.927, 0.925, and 0.919, respectively. The areas under the ROC curve for the validation group were 0.934, 0.937, and 0.944, respectively. The calibration curves and DCA curves showed excellent prediction effect in the prediction model.

Conclusion

The prognostic prediction model established in this study demonstrated higher predictive accuracy.