A Competing Risk Model Nomogram to Predict the Long-Term Prognosis of Lung Carcinoid.

ABSTRACT

BACKGROUND: The prediction of long-term, cancer-specific survival of lung carcinoid remains controversial. We aimed to build a prognostic model by using competing-risk analysis to predict the long-term, cancer-specific survival of lung carcinoid patients. METHODS: Patients were retrospectively enrolled from the SEER database, and clinicopathological data were collected. Univariable and multivariable competing-risk analyses were conducted to identify prognostic factors. A competing-risk model and a nomogram were developed by using independent prognostic factors. The model was assessed by using concordance index and calibration curves. RESULTS: A total of 2496 patients were enrolled, of which 267 (10.7%) died of diagnosed carcinoma; 316 (12.7%) died because of other reasons. The 5-year, 10-year, and 15-year cancer-specific survival of carcinoid patients were 91.35%, 86.60%, and 84.39%, respectively. Multivariable analysis demonstrated that increasing age, male, larger tumor size, higher N stage, M1, atypical carcinoid, and undergoing no surgery were independent risk factors. A competing-risk model based on the risk factors and a corresponding nomogram were developed. Concordance index of the developed model for 5-year, 10-year, and 15-year were 0.891, 0.856, 0.836 respectively in the training cohort and 0.876, 0.841, 0.819 respectively in the validation cohort after bootstrap adjustment. The calibration curves of 5-year, 10-year, and 15-year showed good agreement. CONCLUSIONS: Increasing age, male, larger tumor size, higher N stage, M1, atypical carcinoid, and undergoing no surgery were independent risk factors. A competing risk model of excellent performance in predicting long-term survival was developed, and a nomogram was established.