Construction and analysis of prognostic model for esophageal cancer specific survival rate
- Author:
Lingmin ZHANG
1
Author Information
- Publication Type:Journal Article
- Keywords: Esophageal cancer; Prognostic model; ROC curve; Tumor-specific survival rate
- From: Journal of Xi'an Jiaotong University(Medical Sciences) 2020;41(4):606-611
- CountryChina
- Language:Chinese
- Abstract: Objective To analyze the factors affecting the prognosis of esophageal cancer and construct a reasonable prognostic model. Methods A total of 672 patients diagnosed with esophageal cancer in 2010 were selected as the study data. The factors affecting the specific survival rate of esophageal cancer were screened and modeled by LASSO COX regression analysis. The prediction was performed by R 3.5.3 software. The model was visually constructed and the value of the predictive model was analyzed. Results The 1, 3, and 5 years tumor-specific survival rates of 672 patients with esophageal cancer were 55.32% (95% CI: 51.56% to 59.08%), 30.07% (95% CI: 26.58% to 33.56%), and 25.04% (95% CI: 21.75% to 28.33%), respectively. LASSO regression analysis screened for 11 variables most relevant to prognosis, i.e., chemotherapy, primary site, tumor grade, T stage, N stage, M stage, radiotherapy sequence, surgery, tumor size, age, and ethnicity. Univariate and multivariate COX regression analyses showed that tumor grade, N stage, M stage, surgery and tumor size were independent risk factors for the prognosis of esophageal cancer. By visualizing the prognostic analysis, a Nomogram was constructed, and the consistency index C-index was 0.726 (95% CI: 0.703-0.749), which was significantly better than that of the 7th edition of AJCC's TNM staging system (C-index=0.654, 95% CI: 0.628-0.680). There was also good agreement between the predicted survival rates of the 1, 3 and 5 year calibration curves of this prognostic model and the actual survival rate. At the same time, the ROC curve results showed that if the patient's total score was greater than 15.570, the sensitivity and specificity were the best, the patient could be judged as a high-risk group. And the Nomogram prognostic model constructed in this paper had AUC=0.831, 95% CI: 0.801-0.859, and the prediction ability was significantly better than that of the traditional AJCC 7 version TNM staging system (AUC=0.759, 95% CI: 0.725-0.791). At the same time, a formula for predicting tumor-specific survival rate was constructed: CSS=5.988×10-5×points3+(-1.740×10-3)×points2+(-2.004×10-2)×points+0.685. Conclusion This study used LASSO regression to screen the variables that affected the prognosis of esophageal cancer and visually constructed relevant independent risk factors. It has high clinical value and is of great significance for the screening of high-risk population and the formulation of subsequent personalized diagnosis and treatment plans.
