Clinicopathological nomograms model to predict long-term overall survival and cancer-specific survival in skin cutaneous melanoma: a population-based study
Lingkai Su,Jukun Song,Zhuhui Duan,Chen Zhu,Juxiang Peng,Xiongfei Sun,Wei Yu,Xinhai Yin,Jianguo Zhu,Hui Chen
DOI: https://doi.org/10.18632/oncotarget.23863
2015-01-01
Oncotarget
Abstract:// Lingkai Su 1 , Jukun Song 2 , Zhuhui Duan 1 , Chen Zhu 3 , Juxiang Peng 3 , Xiongfei Sun 4 , Wei Yu 5 , Xinhai Yin 6 , Jianguo Zhu 6 and Hui Chen 1 1 Affiliated Hospital of Stomatology, Medical College, Zhejiang University, Hangzhou, P.R. China 2 Department of Oral and Maxillofacial Surgery, Guizhou Provincial People’s Hospital, Guizhou, China 3 Guiyang Hospital of Stomatology, Medical College, Guiyang, P.R.China 4 The Second Clinical Medical College of Jinan University, Guizhou, China 5 Renmin Hospital, Hubei University of Medicine, Hubei, China 6 Department of Urology, Guizhou Provincial People’s Hospital, Guizhou, China Correspondence to: Hui Chen, email: huiczju66@zju.edu.cn Keywords: skin cutaneous melanoma; nomograms Received: July 04, 2017 Accepted: October 30, 2017 Published: January 02, 2018 ABSTRACT Skin cutaneous melanoma (SCM) is one of the most aggressive skin cancers with a high mortality rate and the fastest growing global incidence rate of all malignancies. The study was aimed to develop clinical nomograms to predict long-term overall survival and cancer-specific survival (CSS) in patients with SCM. Data of patients diagnosed between 2003 and 2013 were retrieved from the Surveillance, Epidemiology, and End Results. The patients were randomly divided into the Training (70%, n = 22,101) and Validation (30%, n = 9,479) cohorts. The probability of CSS and death from other causes was calculated by the competing risk regression model. Out of 31,580 patients, 4865 died from skin cutaneous melanoma and 2215 died from other causes. The 3- and 5-year probabilities of overall death were 0.671 and 0.865 in the training cohorts, respectively. The 3- and 5-year probabilities of specific death were 0.410 and 0.506 in the training cohorts, respectively. Univariate and multivariate analyses were employed to choose the independent prognosis variable for OS and CSS. Nomogram models predicting the overall survival and CSS were established according to 14 clinicopathologic characteristics (age at the diagnosis, race, sex, tumor location, tumor histology, TNM stage, Breslow thickness, Clark level, tumor ulceration, tumor size, radiotherapy, and surgery of primary site), with high concordance indexes in both internal validation (0.860 for OS and 0.901 for CSS) and external validation (0.859 for OS and 0.904 for CSS). The nomogram model could accurately estimate probabilities of OS and CSS for patients with SCM. The established nomograms could help clinicians in screening patients with high risk of SCM and facilitate individualized treatment.