≥55岁甲状腺乳头状癌患者中央区淋巴结转移的临床预测模型构建与验证

Construction and validation of a clinical prediction model for central lymph node metastasis in patients with high age-risk papillary thyroid cancer

目的分析≥55岁甲状腺乳头状癌(papillary thyroid cancer,PTC)患者发生中央区淋巴结转移(central lymph nodem etastasis,CLNM)的危险因素,构建有预测价值的列线图预测模型。方法法回顾性纳人2019年11月至2022年2月郑州大学第一附属医院甲状腺外一科病区由有甲状腺专科手术经验的高年资医师实施的颈部低领横切口开放性手术治疗的≥55岁PTC患者共406例,收集患者包括人口学特征、疾病特征、实验室检测指标在内的关键临床资料,采用单因素和多因素Logistic回归分析筛选危险因素并构建临床预测模型与列线图。结果研究收集406例≥55岁PTC患者,纳人建模组285例,验证组121例。建模组285例患者建模中筛选出CLNM独立危险因素,建立预测模型Logistic(P)=-2.70552+0.73882×性别(男)+0.38321×肿瘤最大径+1.02489x双侧累及(是)+0.99038x临床明显的颈部淋巴结(是)+3.15004x局部侵犯(是)。预测模型建模组受试者工作特征曲线(receiver operating characteristic curve,ROC)对应最大Youden指数的截断值为0.190,模型R0C曲线下面积(area under the curve,AUC)为0.769(95%CI:0.707~0.832),预测灵敏度82.6%、特异度63.0%、准确度67.7%。应用验证组121例验证模型,并绘制R0C曲线,诊断灵敏度66.7%、特异度74.5%、准确度72.7%,AUC为0.760(95%CI:0.663~0.856)。建模组和验证组进行Hosmer-Lemeshow检验结果提示拟合效果较好(P=0.221、0.706)。建模组和验证组决策曲线分析(DCA)结果证实模型在不同的概率阈值下具有明显的临床决策价值,建模组与验证组模型的表现均优于传统决策策略,模型具有良好的普适性和可靠性。结论性别、肿瘤最大径、肿瘤双侧腺叶累及、临床明显的颈部淋巴结及局部侵犯是≥55岁PTC患者CLNM的独立预测因素,基于上述危险因素构建的临床风险列线图预测模型具有较好的预测性能。

Objective To analyze the risk factors for central lymph node metastasis(CLNM)in patients with papillary thyroid cancer(PTC)aged 55 years and above,and to construct a predictive model with columnar graph.Methods This retrospective study included 406 PTC patients aged 55 and above,treated at the First Affili-ated Hospital of Zhengzhou University from Nov.2019 to Feb.2022.Data on demographic characteristics,disease features,and laboratory test results were collected.Univariate and multivariate logistic regression analyses were used to identify risk factors for CLNM and develop a clinical prediction model and nomogram.Results The study involved 406 patients,divided into a modeling group(285 patients)and a validation group(121 patients).The predictive model identified independent risk factors for CLNM.In the modeling group,the model demonstrated a ROC AUC of 0.769,with 82.6%sensitivity,63.0%specificity,and 67.7%accuracy.The validation group showed 66.7%sensitivity,74.5%specificity,and 72.7%accuracy,with an AUC of 0.760.Hosmer-Lemeshow tests indicated good fit in both groups.Decision curve analysis confirmed the model's clinical decision-making value,showing better performance than traditional strategies and good generalizability and reliability.Conclusions Sex,maximum tumor diameter,bilateral involvement of thyroid lobes,clinically evident cervical lymph nodes,and local invasion are independent predictive factors for CLNM in patients over 55 with papillary thyroid carcinoma(PTC).A clinical risk stratification nomogram model based on these risk factors demonstrates good predictive performance.