2019 年第 2 期 第 14 卷

基于危险因素构建的Logistic回归模型对甲状腺乳头状癌颈部淋巴结转移的预测价值研究

Prediction for cervical lymphatic metastasis in papillary thyroid carcinoma patients using Logistic regression model

作者：赵跃刘春蓉张桐硕余恋雨黄景昊王宁孟震吴凤云

英文作者：

单位：300162天津，武装警察部队后勤学院附属医院甲状腺和乳腺外科（赵跃、黄景昊、王宁、孟震、吴凤云），病理科(刘春蓉)，检验科（张桐硕）；300162天津，武装警察部队后勤学院学员二大队九队（余恋雨）

英文单位：

关键词：甲状腺乳头状癌；颈部淋巴结转移；Logistic回归模型

英文关键词：

• 摘要：

• 【摘要】目的    通过分析甲状腺乳头状癌(PTC)患者颈部淋巴结转移的危险因素，构建预测颈部淋巴结转移的Logistic回归模型，探讨该模型对PTC患者颈部淋巴结转移的预测价值。方法    回顾性选取2015年1月至2017年12月在武装警察部队后勤学院附属医院初次行甲状腺手术且术后病理证实为PTC的患者共103例，根据术后病理结果有无颈部淋巴结转移将患者分为转移组（43例）和非转移组（60例）。单因素分析颈部淋巴结转移的相关危险因素，建立预测PTC颈部淋巴结转移风险的Logistic回归模型，并绘制受试者工作特征(ROC)曲线评价其预测效能。结果    单因素分析结果显示，PTC患者年龄、性别、体重指数、颈部淋巴结肿大、微钙化和游离三碘甲状腺原氨酸水平与颈部淋巴结转移有关，差异均有统计学意义（均P＜0.05）。多因素Logistic回归分析结果显示，体重指数（X3）、结节微钙化（X4）和颈部淋巴结肿大(X5)是PTC患者发生颈部淋巴结转移的独立危险因素（比值比=2.744、3.208、5.754，95%置信区间：1.014～7.427、1.187～8.668、1.560～21.230，均P＜0.05），得到回归方程：Y=-6.401+1.010X3+1.166X4+1.750X5。ROC曲线下面积为0.781（95%置信区间：0.693～0.869，P＜0.01），当约登指数为最大值时，该模型鉴别PTC颈部淋巴结转移的最佳截点值为-3.933，敏感度为81.4%，特异度为66.7%。结论    体重指数、结节微钙化和颈部淋巴结肿大是PTC患者发生颈部淋巴结转移的独立危险因素，本研究构建的Logistic回归模型对PTC患者颈部淋巴结转移有较好的预测价值。

• 【Abstract】Objective    To analyze the risk factors of cervical lymph node metastasis in patients with thyroid papillary carcinoma(PTC), create a Logistic regression model for lymphatic metastasis prediction and estimate the predictive value. Methods    Clinical data of 103 patients who had PTC resection confirmed by postoperative pathology from January 2015 to December 2017 in Affiliated Hospital of Logistics University of Chinese People′s Armed Police Forces were retrospectively analyzed. According to postoperative pathological results, 43 patients had cervical lymph node metastasis and 60 patients had no metastasis. Risk factors of metastasis were analyzed by single-factor method and Logistic regression. Prediction efficiency of the Logistic regression model for lymphatic metastasis was analyzed by receiver operating characteristic(ROC) curve. Results    Single-factor analysis showed that age, gender, body mass index, lymphadenopathy, microcalcification and free triiodothyronine were related to cervical lymphatic metastasis(all P＜0.05). Multi-factor logistic regression showed that body mass index(X3), lymphadenopathy(X4) and microcalcification(X5) were independent risk factors of cervical lymphatic metastasis（OR=2.744, 3.208, 5.754; 95% confidence interval： 1.014-7.427, 1.187-8.668, 1.560-21.230; all P＜0.05）. Logistic regression equation was obtained, Y=-6.401+1.010X3+1.166X4+1.750X5. Optimal cut-off point of the equation was -3.933 estimated by ROC curve and the area under curve was 0.781（95% confidence interval： 0.693-0.869， P＜0.01）; the sensitivity was 81.4%; the specificity was 66.7%. Conclusions    Body mass index, microcalcification and lymphadenopathy were independent risk factors of cervical lymphatic metastasis in PTC patients. Logistic regression model shows a good predictive value for cervical lymphatic metastasis.