结直肠癌隐匿性腹膜转移风险预测模型的建立和验证  

Development and validation of a risk prediction model for occult peritoneal metastasis in patients with colorectal cancer

作　　者：钟攀益 张子龙 窦荣章 陶浩冉 王首超 胡应朝 杨朝纲[1,2,3,4] 王舒艺 熊斌[1,2,3,4] Zhong Panyi;Zhang Zilong;Dou Rongzhang;Tao Haoran;Wang Shouchao;Hu Yingchao;Yang Chaogang;Wang Shuyi;Xiong Bin(Department of Gastrointestinal Surgery,Zhongnan Hospital of Wuhan University,Wuhan 430071,China;Department of Gastric and Colorectal Surgical Oncology,Zhongnan Hospital of Wuhan University,Wuhan 430071,China;Hubei Key Laboratory of Tumor Biological Behaviors,Wuhan 430071,China;Hubei Cancer Clinical Study Center,Wuhan 430071,China;Department of Gastrointestinal Surgery,Jingzhou Central Hospital of Yangtze University,Jingzhou 434020,Hubei Province,China;Department of Radiology,Zhongnan Hospital of Wuhan University,Wuhan 430071,China)

机构地区：[1]武汉大学中南医院胃肠外科,武汉430071 [2]肿瘤生物学行为湖北省重点实验室,武汉430071 [3]湖北省肿瘤医学临床研究中心,武汉430071 [4]武汉市腹膜癌临床医学中心,武汉430071 [5]长江大学荆州市中心医院胃肠外科,湖北荆州434020 [6]武汉大学中南医院放射影像科,武汉430071

出　　处：《首都医科大学学报》2023年第1期115-125,共11页Journal of Capital Medical University

基　　金：国家自然科学基金(82173330);湖北省自然科学基金(2019CFB809)。

摘　　要：目的利用血液学参数和临床病理特征建立用于术前预测结直肠癌(colorectal cancer,CRC)隐匿性腹膜转移的风险预测模型并进行验证。方法回顾性地收集2015年7月至2021年7月在武汉大学中南医院收治并行手术治疗的710例CRC患者的资料,基于手术时间按3∶1比例分为训练集和内部验证集,并收集2019年7月到2021年7月在长江大学荆州市中心医院的193例CRC患者作为独立的外部验证集。通过套索-逻辑回归算法(the least absolute shrinkage and selective operator-Logistic,LASSO-Logistic)筛选出的预测因子建立风险预测模型,建立术前预测CRC隐匿性腹膜转移的诺模图。从区分度、校准能力和临床净收益等方面验证诺模图的性能和临床效益。结果从血液学参数和临床病理特征中筛选出6个预测因子:肿瘤病理类型、浸润深度、影像学腹水、糖类抗原125(carbohydrate antigen 125,CA125)、糖类抗原199(carbohydrate antigen 199,CA199)和D-二聚体。建立了结合血液学参数和临床病理特征的模型,并基于模型构建了用于术前预测CRC隐匿性腹膜转移的诺模图。诺模图在训练集、内部验证集和外部验证集的受试者工作特征曲线下面积分别为:0.956(95%CI:0.936~0.975)、0.891(95%CI:0.857~0.925)和0.901(95%CI:0.860~0.942),校准曲线和临床决策分析曲线验证了模型较好的校准能力和临床净收益。在训练集中,采用约登指数获得分组的最佳截断值为206.6,此时的灵敏度为94.3%,特异度为86.6%,Kappa值为0.753,说明模型预测结果与实际情况具有较高的真实性和一致性,模型具有较好的临床预测性能。结论结合血液学参数和临床病理特征构建了预测CRC隐匿性腹膜转移的风险预测模型,其具有良好的区分度,校准能力和临床净收益,能为CRC隐匿性腹膜转移的诊断和治疗决策提供参考。Objective We aimed to develop and validate a risk prediction model for occult peritoneal metastasis of colorectal cancer(CRC)by using hematologic parameters and clinicopathological features.Methods The data of 710 CRC patients who underwent surgery in Zhongnan Hospital of Wuhan University from July 2015 to July 2021 were retrospectively collected and divided into training set and internal validation set according to the ratio of 3∶1 based on the operation times.A total of 193 CRC patients in Jingzhou Central Hospital of Yangtze University from July 2019 to July 2021 were collected as an independent external validation set.A prediction model was established based on the predictors selected by the least absolute shrinkage and selective operator-Logistic(LASSO-Logistic)algorithm.Based on the model,the nomogram was developed for preoperative prediction of occult peritoneal metastasis of CRC.The performance and clinical benefit of nomogram were validated from the aspects of discrimination,calibration ability and clinical net benefit.Results Six predictors were screened from hematologic parameters and clinicopathological features:tumor pathological type,depth of invasion,imaging ascites,carbohydrate antigen 125(CA125),carbohydrate antigen 199(CA199)and D-dimer.The model combining hematological parameter and clinicopathological feature was established.Based on the combination model,a nomogram was constructed for preoperative prediction of occult peritoneal metastasis in CRC.The area under the receiver operating characteristic curve(AUC)of the nomogram were 0.956(95%CI:0.936-0.975),0.891(95%CI:0.857-0.925)and 0.901(95%CI:0.860-0.942)in the training set,internal validation set and external validation set,respectively.The calibration curve and clinical decision analysis curve verified the good calibration ability and clinical net benefit of the model.In the training set,the Youden index was used to obtain the best cut-off value of 206.6,the sensitivity was 94.3%,the specificity was 86.6%,and the Kappa value was 0.753,ind

关 键 词：结直肠癌 隐匿性腹膜转移 风险预测模型 诺模图 术前预测 

分 类 号：R735.3[医药卫生—肿瘤]