Abstract:

Objective: To develop and validate a nomogram prediction model based on gingival carcinoma patient data from the Surveillance, Epidemiology, and End Results (SEER) database of the National Cancer Institute in the United States, and evaluate its effectiveness. Methods: The basic information, clinical information, and follow-up data of patients diagnosed with gingival carcinoma between 2004 and 2018 was downloaded from the SEER database, and the patients were randomly divided into a modeling group (70%) and a validation group (30%). Through univariate and multivariate Cox regression analysis, variables were screened to determine independent prognostic risk factors for gingival carcinoma patients, and a nomogram was constructed. The prediction model was evaluated from three perspectives: Discrimination, calibration, and clinical applicability. X-Tile software was used for risk stratification of the gingival carcinoma patients. Results: This study included 3 334 patients with gingival carcinoma, with 3-year, 5-year, and 10-year survival rates of 59.03%, 48.89%, and 30.09%, respectively. The concordance index (C-index) of the modeling group and validation group was 0.716 and 0.715, respectively; the receiver operating characteristic (ROC) curve results showed that the area under the curve (AUC) for predicting the 3-year, 5-year, and 10-year overall survival rates of gingival carcinoma patients in the modeling group was 0.758, 0.759, and 0.769, respectively. In the validation group, they were 0.746, 0.764, and 0.788, respectively. The C-index and AUC of the nomogram were higher than those of the traditional TNM staging system (P<0.001). The calibration curve results indicate that the nomogram model has good predictive accuracy. The decision curve analysis (DCA) results show that the nomogram has good clinical application value and is superior to the traditional TNM staging system.Patients were divided into low-risk group (<118.4 points), medium-risk group (118.4-213.6 points), and high-risk group (>213.6 points) based on the cutoff value of the total score in the nomogram. There was a statistically significant difference in survival rate among the three groups (P<0.001). Conclusion: Age, marital status, histological grading, T stage, N stage, surgical status, and radiotherapy status are factors affecting the survival of patients with gingival carcinoma. The nomogram prediction model constructed in this study can provide reference for predicting the prognosis of gingival carcinoma patients.

Key words: gingival carcinoma, risk factors, nomogram, SEER database, prognosis, prediction model

摘要：

目的：基于美国国家癌症研究所监测、流行病学和最终结果（Surveillance，Epidemiology，and End Results，SEER）数据库牙龈癌患者资料构建列线图预测模型，并评价其效果。方法：从SEER数据库下载2004—2018年诊断为牙龈癌患者的基本信息、临床信息及随访资料，将患者随机分为建模组（70%）和验证组（30%）。通过单因素和多因素Cox回归分析筛选变量，确定牙龈癌患者预后独立危险因素并构建列线图。从区分度、校准度及临床适用性3个角度评估预测模型，采用X-Tile软件对牙龈癌患者进行危险分层。结果：本研究共纳入3 334例牙龈癌症患者，患者3年、5年、10年生存率分别为59.03%、48.89%、30.09%。建模组和验证组一致性指数（concordance index，C-index）分别为0.716和0.715；受试者工作特征（receiver operating characteristic，ROC）曲线结果显示，建模组预测牙龈癌患者的3年、5年、10年总生存率的曲线下面积（area under the curve，AUC）为0.758、0.759、0.769，在验证组中分别为0.746、0.764、0.788，列线图预测模型的一致性指数和AUC均大于传统的TNM分期系统（P<0.001）。校准曲线结果表明列线图预测模型具有良好的预测准确性。临床决策曲线分析（decision curve analysis，DCA）结果显示列线图具有良好的临床应用价值，并优于传统TNM分期系统。根据列线图总分值截断值将患者分为低风险组（<118.4分）、中风险组（118.4~213.6分）和高风险组（>213.6分），3组生存率差异有统计学意义（P<0.001）。结论：年龄、婚姻状况、组织学分级、T分期、N分期、手术情况、放疗情况是影响牙龈癌患者生存的因素，本研究构建的列线图预测模型可以为预测牙龈癌患者预后提供参考。

关键词: 牙龈癌, 危险因素, 列线图, SEER数据库, 预后, 预测模型