Individualized survival prediction in medullary thyroid carcinoma using a least absolute shrinkage and selection operator-Cox-based nomogram

ObjectiveMedullary thyroid carcinoma is a rare malignancy with heterogenous outcomes, and the existing American Joint Committee on Cancer staging system does not adequately incorporate patient- or treatment-specific prognostic factors.MethodsPatients with medullary thyroid carcinoma were identified from the Surveillance, Epidemiology, and End Results database and randomly assigned to derivation and validation groups. Five prognostic approaches, including Cox regression, least absolute shrinkage and selection operator-Cox, CoxBoost, random survival forest, and extreme gradient boosting, were evaluated using Harrell's concordance index. Least absolute shrinkage and selection operator-Cox, which demonstrates optimal performance and simplicity, was selected to develop a nomogram to predict 3- and 5-year mortality.ResultsA total of 4157 patients were included, with 2938 assigned to the derivation group and 1219 to the validation group. The least absolute shrinkage and selection operator-Cox model demonstrated strong predictive performance, yielding Harrell's concordance indices of 0.813 and 0.852, both exceeding those of American Joint Committee on Cancer staging (0.730 and 0.722). Calibration plots showed close agreement between predicted and observed mortality, and decision curve analysis revealed that the nomogram provided a higher net benefit compared with strategies of treating all or no patients.ConclusionA nomogram integrating demographic, clinical, and treatment factors was developed to predict individual survival in medullary thyroid carcinoma, providing complementary information to American Joint Committee on Cancer staging and facilitating personalized risk stratification.