食管鳞状细胞癌肿瘤病理长度的预后和预测意义

发布时间：2018-07-05 04:21

本文选题：食管癌 + 肿瘤长度　；参考：《浙江大学》2015年博士论文

【摘要】：目的：食管鳞状细胞癌肿瘤长度对预后的影响一直存在争议。本研究旨在探讨食管鳞状细胞癌肿瘤病理长度的预测和预后意义。并且选择理想的肿瘤长度临界值以获得准确的风险分层。方法：自2003年1月至2010年12月,1613例食管癌患者在我科行手术治疗。患者临床病理资料均从电子病历系统收集。利用Cox风险比例回归模型鞅残差(Martingale Residuals)分析获取肿瘤病理长度最佳临界值。探讨肿瘤病理长度对预后的影响及其与其他临床病理因素之间的相关性。利用基于赤池信息标准(Akaike Information Criterion, AIC)的Cox回归模型筛选临床病理变量以确立独立预后因子。各临床病理变量的预后预测准确度以Harrell一致性指数为指标(Harrell Concordance Index,C-index).建立数个Cox回归模型以评价临床病理变量与预后的关系。各模型采用自助重抽样法(bootstrap)作内部验证,行校准(validation)和鉴别(discrimination),并将模型以列线图(nomogram)方式表示以预测预后。时间依赖性接受者工作特征(time-dependent Receiver Operating Characteristic, time-dependentROC)也用于评价各模型的预测准确性。对于肿瘤病理长度是否增加TNM分期模型的预后准确性,使用生存资料的净重分类改善指数(Net Reclassification improvement, NRI)和整体鉴别指数(Integrated Discrimination Index, IDI)分析。结果：根据纳入标准,选择1435例经行根治术的食管鳞状细胞癌患者行回顾性分析。鞅残差分析显示肿瘤病理长度最佳临界值为4厘米。肿瘤病理长度与年龄、性别、肿瘤部位、T分期、N分期、以及切除淋巴结数目相关。肿瘤病理长度型厘米患者预后优于肿瘤病理长度4厘米患者(中位生存期,48月vs27月P0.001)。肿瘤分化程度、年龄、T分期、N分期、切除淋巴结数目、以及病理长度是食管鳞癌独立预后因子。肿瘤病理长度(C-index=58.1%)的预后准确性略低于N分期(C-index=67.1%)和T分期(C-index=60.5%)。在预测5年生存率方面,TNM分期模型预测准确度略小于TNM分期联合肿瘤病理长度模型(69.4%vs69.8%,P0.001)。各模型的校准图显示预测结果与实际结果一致性良好。类似结果也出现在时间依赖性接受者工作特征分析中。比较TNM分期模型与TNM分期加肿瘤病理长度模型,肿瘤病理长度未显示其能提高TNM分期预测准确性(NRI=0.001,P=0.637; IDI=0.046, P=0.498)。结论：肿瘤病理长度是食管鳞状细胞癌的独立预后因子。其最佳临界值为4厘米。但肿瘤病理长度并不提高现行TNM分期的预测准确性。
[Abstract]:Objective: the effect of tumor length on prognosis of esophageal squamous cell carcinoma has been controversial. The purpose of this study was to investigate the prognostic significance of pathological length of esophageal squamous cell carcinoma (ESCC). And select the ideal critical value of tumor length to obtain accurate risk stratification. Methods: from January 2003 to December 2010, 1613 patients with esophageal cancer underwent surgical treatment in our department. The clinical and pathological data of the patients were collected from the electronic medical record system. Martingale residuals (Martingale residuals) analysis was used to obtain the best critical value of pathological length of tumor. To investigate the influence of tumor pathological length on prognosis and its correlation with other clinicopathological factors. A Cox regression model based on Akaike Information criteria was used to screen clinicopathological variables to establish independent prognostic factors. The prognostic accuracy of all clinicopathological variables was determined by Harrell concordance index (Harrell concordance index). Several Cox regression models were established to evaluate the relationship between clinicopathologic variables and prognosis. The models were verified by self-help resampling method (bootstrap). The (validation) was calibrated and the (discrimination), was identified. The models were expressed as (nomogram) to predict the prognosis. Time-dependent receiver operating characteristic (time-dependent ROC) is also used to evaluate the prediction accuracy of each model. As to whether the pathological length of tumor increases the prognostic accuracy of TNM staging model, the net weight classification improvement index (NRI) and Integrated Discrimination Index (IDI) of survival data were used. Results: according to the inclusion criteria, 1435 patients with esophageal squamous cell carcinoma undergoing radical resection were retrospectively analyzed. Martingale residual analysis showed that the optimal critical value of pathological length was 4 cm. The pathological length of tumor was correlated with age, sex, T stage and N stage of tumor site, and the number of lymph nodes resected. The prognosis of patients with pathological length of tumor was better than that of patients with tumor length of 4 cm (median survival time, 48 months vs27 month, P0. 001). Tumor differentiation, age T stage and N stage, number of resected lymph nodes, and pathological length were independent prognostic factors of esophageal squamous cell carcinoma. The prognostic accuracy of tumor pathological length (58.1%) was slightly lower than that of N stage (67.1%) and T stage (60.5%). The prediction accuracy of TNM staging model was slightly less than that of TNM staging combined with tumor pathological length model (69.4 vs 69.8 / P0.001). The calibration diagrams of each model show that the predicted results are in good agreement with the actual results. Similar results were found in the analysis of work characteristics of time dependent recipients. Compared with TNM staging model and TNM staging plus tumor pathological length model, the tumor pathological length did not improve the accuracy of TNM staging (NRI0. 001, P0. 637, IDI 0. 046, P0. 498). Conclusion: tumor pathological length is an independent prognostic factor for esophageal squamous cell carcinoma. The optimum critical value is 4 cm. However, tumor pathological length does not improve the prediction accuracy of TNM staging.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：R735.1

【参考文献】