基于列线图模型构建的分化型甲状腺癌患者淋巴结转移的风险评估

目的:探讨分化型甲状腺癌（differentiated thyroid carcinoma,DTC）术后患者颈部淋巴结转移(lymphatic node metastasis,LNM)的独立预测因素，并构建列线图模型对其LNM风险进行个性化预测，以期为后续131I治疗提供可靠依据。方法:回顾性分析2020年1月至2022年1月本院行甲状腺全切术+颈部淋巴结清扫术，首次131I治疗的332例DTC患者资料，依据131I治疗后全身显像（Rx-WBS）及SPECT/CT断层融合显像结果，将患者分为无转移组（M0）和颈部淋巴结转移组（M1）。纳入相关评价指标:性别、年龄、肿瘤直径、包膜侵犯、肿瘤N分期及血小板计数（P）、中性粒细胞计数（N）、淋巴细胞计数（L）、刺激性甲状腺球蛋白（preablative stimulated Tg,ps-Tg）、促甲状腺激素（TSH）。采用logistic多因素分析确定DTC术后LNM的独立预测因素；运用R3.6.1软件整合所有预测因素，根据多因素分析所得系数值绘制列线图模型，采用受试者工作特征（ROC）曲线下面积（AUC）评价列线图模型的区分度，并采用bootstrap法绘制校准图对模型的性能进行验证。结果:多因素分析结果显示，性别（P＜0.001），年龄（P=0.044），ps-Tg（P＜0.001），包膜侵犯（P＜0.001）、L计数（P=0.028）是LNM的独立预测因素。整合上述5个因素的列线图模型预测LNM风险的区分度为0.713（AUC=0.713，95%CI 0.706～0.720），模型预测的LNM风险与实际LNM发生率具有良好的一致性，预测LNM风险的误差均在10%以内。结论:性别、年龄、ps-Tg、包膜侵犯、L计数是DTC术后患者LNM的独立预测因素，基于上述指标构建的LNM风险预测模型有助于指导DTC术后患者个性化131I治疗方案的制定。

LNM risk assessment of DTC patients based on a constructed nomogram model

Objective:In order to determine the first treatment dose of 131I after differentiated thyroid carcinoma (DTC) postoperation,a nomogram model was established to explore the independent predictors of lymphatic node metastasis (LNM) in DTC.Methods:332 DTC patients data,with whole thyroid resected and lymphatic node clearance operation in Tangdu Hospital from January 2020 to January 2022,were divided into metastasis-free group (M0) and cervical lymph node metastasis group (M1),according to the results of systemic imaging (Rx-WBS) and SPECT/CT fusion imaging after thyroid 131I treatment.The possible related indicators including gender,age,tumor diameter,capsule invasion,tumor N stage and platelet count (P),neutrophil count (N),lymphocyte count (L),preablative stimulated Tg (ps-Tg)and thyroid stimulating hormone (TSH) were concluded.Logistic multivariate analysis was used to determine the independent predictors of LNM after DTC postoperation.A nomogram model was developed by R3.6.1 to integrate all predictors based on multivariate analysis,evaluated according to the receiver operating characteristic curve,and then validated by the bootstrap calibration method.Results:The results of multivariate analysis showed that gender (P＜0.001),age (P=0.044),ps-Tg (P＜0.001),capsule invasion (P＜0.001),and L count (P=0.028) were independent predictors of LNM respectively.The AUC of the nomogram model integrating the above five factors achieved 0.713 (95%CI 0.706~0.720).And the model prediction of LNM risk was consistent within the actual incidence,with prediction errors less than 10%.Conclusion:Gender,age,ps-Tg,capsule invasion,and L count were independent predictors of LNM after DTC operation.The nomogram prediction model based on the multi-indicators is capable to guide personalized 131I treatment after DTC operation.