甲状腺微小乳头状癌中央区颈部淋巴结转移风险评分系统 的初步构建
| 作者: |
1马文卿,
1周平,
1梁永平,
1田双明,
1赵永锋,
1刘稳刚
1 中南大学湘雅三医院 超声科，湖南 长沙 410013 |
| 通讯: |
周平
Email: zhouping1000@hotmail.com |
| DOI: | 10.3978/.2018.06.015 |
| 基金: | 湖南省科技厅科技计划资助项目（2014SK3245）。 |
摘要
目的：探讨通过甲状腺微小乳头状癌（PTMC）结节的超声图像特征与患者临床资料构建中央区颈部淋巴结转移的预测模型。
方法：收集 121 例手术与病理确诊为 PTMC 患者的临床资料及其 166 个结节的超声图像，通过统计学方法分析找出中央区颈部淋巴结转移相关因素，分别建立中央区颈部淋巴结转移风险预测的 Logistic评分模型与 Additive 评分模型。
结果：统计结果显示，年龄、结节多灶性、结节大小、结节与甲状腺包膜的关系、超声造影结节周边增强情况与中央区颈部淋巴结转移风险有关（均 P<0.05）；将以上因素进行赋值评分；所得总分以ROC 曲线下面积（AUC）评价，当 Logistic 模型得分为 6.5 分时，AUC 为 0.964、敏感性 98.4%、特异性 74.1%；当 Additive 模型得分为 33.5 分时，AUC 为 0.928、敏感性 88.9%、特异性 84.5%；Hosmer-Lemeshow 拟合优度检验示 Logistic 模型拟合较好。
结论：所构建的 Logistic 模型对 PTMC 中央区颈部淋巴结转移有较好的预测价值，可为 PTMC 治疗方案的选择提供量化依据，其分值≥ 7 分时，需高度怀疑存在中央区颈部淋巴结转移的可能，反之，转移可能性较小。
关键词：
甲状腺肿瘤；癌，乳头状；淋巴转移；超声检查；危险性评估
方法：收集 121 例手术与病理确诊为 PTMC 患者的临床资料及其 166 个结节的超声图像，通过统计学方法分析找出中央区颈部淋巴结转移相关因素，分别建立中央区颈部淋巴结转移风险预测的 Logistic评分模型与 Additive 评分模型。
结果：统计结果显示，年龄、结节多灶性、结节大小、结节与甲状腺包膜的关系、超声造影结节周边增强情况与中央区颈部淋巴结转移风险有关（均 P<0.05）；将以上因素进行赋值评分；所得总分以ROC 曲线下面积（AUC）评价，当 Logistic 模型得分为 6.5 分时，AUC 为 0.964、敏感性 98.4%、特异性 74.1%；当 Additive 模型得分为 33.5 分时，AUC 为 0.928、敏感性 88.9%、特异性 84.5%；Hosmer-Lemeshow 拟合优度检验示 Logistic 模型拟合较好。
结论：所构建的 Logistic 模型对 PTMC 中央区颈部淋巴结转移有较好的预测价值，可为 PTMC 治疗方案的选择提供量化依据，其分值≥ 7 分时，需高度怀疑存在中央区颈部淋巴结转移的可能，反之，转移可能性较小。
Preliminary construction of risk scoring system for estimation of central cervical lymph node metastasis in papillary thyroid microcarcinoma
CorrespondingAuthor:ZHOU Ping Email: zhouping1000@hotmail.com
Abstract
Objective: To investigate the construction of a predictive model for estimation of central cervical lymph node metastasis of papillary thyroid microcarcinoma (PTMC) based on the ultrasonographic features of the nodules and clinical data of the patients.
Methods: Th e clinical data of 121 patients with PTMC that was confi rmed by surgical and pathological fi ndings and their ultrasonic images of 166 thyroid nodules were collected. The relevant factors for central cervical lymph node metastasis were analyzed and picked up by statistical methods, and then, the Logistic and Additive scoring models for estimating risk of central cervical lymph node metastasis were established, respectively.
Results: The results of statistical analyses showed that age, multifocal lesion, nodule size, relationship between nodule and thyroid capsule, and imaging feature of perinodular enhancement were closely related to the risk of central cervical lymph node metastasis (all P<0.05). These factors were numerically assigned and scored. According to the total score evaluated by the area under the ROC curve (AUC), when the score derived from the Logistic model reached 6.5, the AUC was 0.964 with a sensitivity of 98.4% and a specificity of 74.1%; when the score derived from Additive model reached 33.5, the AUC was 0.928 with a sensitivity of 88.9% and a specificity of 84.5%. Hosmer-Lemeshow goodness of fit test indicated that Logistic model had a better fit.
Conclusion: The constructed Logistic model has certain predictive value for estimating risk of central cervical lymph node metastasis in PTMC, and it can provide a quantitative basis for the treatment plan selection of PTMC. When its score is equal to or greater than 7, the possibility of central cervical lymph node metastasis should be highly suspected.
Keywords：
Thyroid Neoplasms; Carcinoma
Papillary; Lymphatic Metastasis; Ultrasonography; Risk Assessment
Methods: Th e clinical data of 121 patients with PTMC that was confi rmed by surgical and pathological fi ndings and their ultrasonic images of 166 thyroid nodules were collected. The relevant factors for central cervical lymph node metastasis were analyzed and picked up by statistical methods, and then, the Logistic and Additive scoring models for estimating risk of central cervical lymph node metastasis were established, respectively.
Results: The results of statistical analyses showed that age, multifocal lesion, nodule size, relationship between nodule and thyroid capsule, and imaging feature of perinodular enhancement were closely related to the risk of central cervical lymph node metastasis (all P<0.05). These factors were numerically assigned and scored. According to the total score evaluated by the area under the ROC curve (AUC), when the score derived from the Logistic model reached 6.5, the AUC was 0.964 with a sensitivity of 98.4% and a specificity of 74.1%; when the score derived from Additive model reached 33.5, the AUC was 0.928 with a sensitivity of 88.9% and a specificity of 84.5%. Hosmer-Lemeshow goodness of fit test indicated that Logistic model had a better fit.
Conclusion: The constructed Logistic model has certain predictive value for estimating risk of central cervical lymph node metastasis in PTMC, and it can provide a quantitative basis for the treatment plan selection of PTMC. When its score is equal to or greater than 7, the possibility of central cervical lymph node metastasis should be highly suspected.