甲状腺乳头状癌被膜侵犯患者中央区淋巴结转移风险因素列线图预测模型的建立及验证

To establish and validate a nomogram prediction model for the risk factors of central lymph node metastasis in patients with capsular invasion of papillary thyroid carcinoma

目的探讨甲状腺乳头状癌(PTC)伴被膜侵犯时中央区淋巴结转移(centrallymphnodemetastasis,CLNM)的相关因素,并构建临床列线图预测模型,为临床诊治提供理论依据。方法回顾性分析2020年10月~2023年10月保定市第一中心医院普外科收治的PTC侵犯被膜患者的临床资料,包括性别、年龄、体重指数(BMI)、纵横比、肿瘤位置、多灶性、微钙化、合并桥本甲状腺炎(Hashimoto thyroiditis,HT)和肿瘤直径。根据是否出现CLNM分为正常组(107例)和转移组(108例),对数据进行单因素及多因素分析,构建可视化列线图预测模型,并绘制受试者工作特征(ROC)曲线评估模型诊断效能。采用重复抽样1000次的Bootstrap检验对列线图模型进行内部验证。一致性指数(consistency index,C-index)和校准曲线来表示模型的预测性能和预测准确度。最后绘制临床决策曲线(DCA)判断模型的临床应用能力。结果共纳入215例PTC伴被膜侵犯患者,其中108例(50.23%)发生CLNM。单因素分析显示CLNM发生与患者肿瘤直径、纵横比>1、肿瘤位于下极、多灶性和合并HT有关(P<0.05)。多因素回归分析显示,肿瘤直径、纵横比>1、肿瘤位于下极和多灶性是CLNM的独立危险因素(OR=1.401、1.875、2.291、2.303,P<0.05),合并HT是CLNM的保护因素(OR=0.501,P<0.05)。基于以上危险因素构建PTC伴被膜侵犯患者发生CLNM的列线图预测模型,ROC曲线显示:曲线下面积AUC为0.859,95%CI:0.792~0.925,约登指数为0.734,敏感度为0.878,特异度为0.856,模型有较高的预测价值。内部验证C-index=0.83(95%CI:0.748~0.959)。校正曲线显示预测值靠近理想曲线,有较好一致性,在DCA曲线中该模型在临床效能表现良好。结论当PTC伴被膜侵犯时,肿瘤较大、纵横比>1、肿瘤位于下极、多灶性提示发生CLNM的风险较高,而合并HT是发生CLNM的保护因素。基于以上危险因素构建的列线图模型有较高的区分度和校准度,有助于临床医师术前评估和术中探查,以便尽早对高危患者进行预防和治疗。

OBJECTIVE To investigate the related factors of central lymph node metastasis(CLNM)in papillary thyroid carcinoma(PTC)with capsular invasion,and to construct a clinical nomogram prediction model.Its purpose is to provide theoretical basis for clinical diagnosis and treatment.METHODS The clinical data of PTC patients with capsule invasion admitted to the Department of General Surgery,Baoding First Central Hospital from October,2020 to October,2023 were retrospectively analyzed.The data included gender,age,body mass index(BMI),aspect ratio,tumor location,multifocality,microcalcification,Hashimoto thyroiditis(HT)and tumor diameter.According to the presence or absence of CLNM,the patients were divided into the normal group(107 cases)and the metastasis group(108 cases).Univariate and multivariate analysis of the data were performed to construct a visual nomogram prediction model,and the receiver operating characteristic(ROC)curve was drawn to evaluate the diagnostic efficacy of the model.The nomogram model was internally verified using a Bootstrap test with 1000 repeated samples.Consistency index(C-index)and calibration curve were used to describe the prediction performance and prediction accuracy of the model.Finally,the clinical decision curve(DCA)was drawn to determine the clinical application ability of the model.RESULTS A total of 215 PTC patients with capsular invasion were included,of whom 108(50.23%)had CLNM.Univariate analysis showed that the occurrence of CLNM was associated with tumor diameter,aspect ratio>1,tumor located in the lower pole,multifocality,and HT(P<0.05).Multivariate regression analysis showed that tumor diameter,aspect ratio>1,tumor located in the lower pole and multifocus were independent risk factors for CLNM(OR=1.401,1.875,2.291,2.303,P<0.05),and HT was a protective factor for CLNM(OR=0.501,P<0.05).Based on the above risk factors,a nomogram prediction model for CLNM in patients with PTC with capsule invasion was constructed.The ROC curves showed that the area under the curve(AUC)was 0.859(95%CI:0.792-0.925,Yoden Index was 0.734,the sensitivity was 0.878,a specificity was 0.856),and the model had higher predictive value.Internal validation consistency index(C-index)was 0.83(95%CI,0.748 to 0.959).The calibration curve showed that the predictive value was close to the ideal curve,and it had good consistency.The DCA curve showed that the model had good clinical efficacy.CONCLUSION Larger tumor size,aspect ratio>1,tumor located in the lower pole and multifocality suggest higher risk of CLNM in PTC with capsular invasion,while HT is a protective factor for CLNM.The nomogram model based on the above risk factors has high discrimination and calibration,which is helpful for clinicians in preoperative evaluation and intraoperative exploration,so that high-risk patients can be prevented and treated as soon as possible.