2023 , Vol. 20 >Issue 04: 442 - 448
DOI: https://doi.org/10.3877/cma.j.issn.1672-6448.2023.04.011
基于超声造影的预测模型对甲状腺乳头状癌颈中央区淋巴结转移的诊断价值
Copy editor: 汪荣
收稿日期: 2021-08-12
网络出版日期: 2023-08-07
基金资助
中华人民共和国工业和信息化部项目(先进医疗设备应用技术评价:甲状腺成像体系超声评价规范)(2018MND102015)
版权
Diagnostic value of a predictive model based on contrast-enhanced ultrasound for prediction of central cervical lymph node metastasis in patients with thyroid papillary carcinoma
Received date: 2021-08-12
Online published: 2023-08-07
Copyright
建立基于超声造影(CEUS)的预测模型以评估甲状腺乳头状癌(PTC)患者颈部中央区淋巴结转移(CLNM)风险,并讨论该模型的诊断效能。
回顾性分析2017年1月至2019年12月于首都医科大学附属北京友谊医院确诊为单发PTC的患者108例。根据有无CLNM将其分为CLNM组(48例)与无CLNM组(60例),比较2组间一般临床资料、常规超声以及CEUS表现之间的差异。根据多因素Logistic回归分析建立预测模型,绘制ROC曲线评价模型及模型内各变量因素的诊断效能。
单因素分析结果显示,患者年龄、病灶最大径、是否接触被膜、增强程度以及增强均匀性在CLNM 组与无CLNM组之间差异具有统计学意义(P均<0.05)。多因素Logistic回归分析显示,年龄≤42岁(OR=4.783,P=0.002)、接触被膜(OR=4.833,P=0.006)、等或高增强(OR=3.696,P=0.013)以及不均匀增强(OR=5.264,P=0.004)是CLNM的独立危险因素。建立预测模型为:Y=-4.017+1.565×年龄≤42+1.576×病灶接触被膜+1.307×等或高增强+1.661×不均匀增强。该模型ROC曲线下面积(AUC)为0.800,敏感度及特异度分别为70.8%、78.3%;模型AUC大于模型内单一变量参数的AUC(P均<0.05)。
基于CEUS并联合常规超声及临床参数建立的预测模型,诊断效能较好,可用于评估PTC患者CLNM的风险。
关键词: 甲状腺乳头状癌; 中央区淋巴结转移; 超声造影; Logistic模型
陈启阳 , 刘玉江 , 刘金苹 , 谭小蕖 , 钱林学 , 胡向东 . 基于超声造影的预测模型对甲状腺乳头状癌颈中央区淋巴结转移的诊断价值[J]. 中华医学超声杂志(电子版), 2023 , 20(04) : 442 -448 . DOI: 10.3877/cma.j.issn.1672-6448.2023.04.011
To establish a predictive model based on contrast-enhanced ultrasound (CEUS) to evaluate the risk of central cervical lymph node metastasis (CLNM) in patients with thyroid papillary carcinoma (PTC), and to evaluate the diagnostic efficacy of the model.
A retrospective analysis was performed on 108 patients with single PTC who were examined by conventional ultrasonography and contrast-enhanced ultrasonography before operation at the Beijing Friendship Hospital of Capital Medical University from January 2017 to December 2019. The patients were divided into either a CLNM group or a non-CLNM group according to whether CLNM occurred, and the differences in clinical, conventional ultrasound, and CEUS characteristics between the two groups were compared. A predictive model was established according to multivariate regression analysis, and the diagnostic efficacy of the model was evaluated by receiver operating characteristic (ROC) curve analysis.
Univariate analysis indicated significant differences between the CLNM and non-CLNM groups in age, maximum tumor diameter, contact with capsule, degree of enhancement, and uniformity of enhancement (P<0.05). Multivariate logistic regression analysis showed that age≤42 (odds ratio [OR]=4.783, P=0.002), contact with capsule (OR=4.833, P=0.006), Hyper- or isoechoic enhancement (OR=3.696, P=0.013), and heterogeneous enhancement (OR=5.264, P=0.004) were independent risk factors for CLNM. ROC curve analysis showed that the area under the curve (AUC), sensitivity, and specificity of the predictive model were 0.800, 70.8%, and 78.3%, respectively. The AUC of the predictive model was greater than that of any single independent risk predictor (P<0.05).
The prediction model based on CEUS combined with conventional ultrasound and clinical factors has good diagnostic efficiency and can be used for the risk assessment of CLNM in PTC patients.
表1 CLNM组与无CLNM组PTC患者的一般资料比较[例(%)] |
资料 | CLNM组(n=48) | 无CLNM组(n=60) | χ2值 | P值 |
---|---|---|---|---|
性别 | 3.953 | <0.001 | ||
男 | 18(37.5) | 4(6.70) | ||
女 | 30(62.5) | 56(93.3) | ||
年龄 | 3.750 | <0.001 | ||
≤42岁 | 35(72.9) | 22(36.7) | ||
>42岁 | 13(27.1) | 38(63.3) | ||
病灶最大径 | 3.620 | <0.001 | ||
<0.95 cm | 16(33.3) | 41(68.3) | ||
≥0.95 cm | 32(66.7) | 19(31.7) | ||
位置 | 0.372 | 0.923 | ||
左叶 | 21(43.8) | 28(46.7) | ||
右叶 | 26(54.2) | 31(51.7) | ||
峡部 | 1(2.1) | 1(1.7) |
注:CLNM为颈中央区淋巴结转移;PTC为甲状腺乳头状癌 |
表2 CLNM组与无CLNM组PTC患者常规超声特征比较[例(%)] |
超声特征 | CLNM组(n=48) | 无CLNM组(n=60) | χ2值 | P值 |
---|---|---|---|---|
形态 | 1.114 | 0.265 | ||
规则 | 15(31.3) | 25(41.7) | ||
不规则 | 33(68.8) | 35(58.3) | ||
纵横比 | 1.571 | 0.116 | ||
>1 | 24(50.0) | 39(65.0) | ||
≤1 | 24(50.0) | 21(35.0) | ||
有无钙化 | 1.359 | 0.174 | ||
有钙化 | 43(89.6) | 48(80.0) | ||
无钙化 | 5(10.4) | 12(20.0) | ||
回声 | 1.401 | 0.161 | ||
低回声 | 42(87.5) | 57(95.0) | ||
等回声或高回声 | 6(12.5) | 3(5.0) | ||
是否与被膜接触 | 2.136 | 0.033 | ||
接触 | 40(83.3) | 39(65.0) | ||
不接触 | 8(16.7) | 21(35.0) |
注:CLNM为颈中央区淋巴结转移;PTC为甲状腺乳头状癌 |
表3 CLNM组与无CLNM组PTC患者CEUS特征比较[例(%)] |
超声造影特征 | CLNM组(n=48) | 无CLNM组(n=60) | χ2值 | P值 |
---|---|---|---|---|
增强程度 | 3.168 | 0.005 | ||
低增强 | 19(39.6) | 42(70.0) | ||
等或高增强 | 29(60.4) | 18(30.0) | ||
增强方式 | 0.747 | 0.455 | ||
向心性增强 | 27(56.3) | 38(63.3) | ||
弥散性增强 | 21(43.8) | 22(36.7) | ||
增强均匀性 | 2.131 | 0.033 | ||
均匀增强 | 10(20.8) | 24(40.0) | ||
不均匀增强 | 38(79.2) | 36(60.0) | ||
被膜增强是否连续 | 1.401 | 0.161 | ||
是 | 18(37.5) | 15(25.0) | ||
否 | 30(62.5) | 45(75.0) |
注:CLNM为颈中央区淋巴结转移;PTC为甲状腺乳头状癌 |
表4 预测模型各因素变量名及赋值表 |
因素 | 变量名 | 赋值说明 |
---|---|---|
年龄 | X1 | >42岁=0,≤42岁=1 |
是否接触被膜 | X2 | 不接触=0,接触=1 |
增强程度 | X3 | 低增强=0,等或高增强=1 |
增强均匀性 颈中央区淋巴结转移 | X4 Y | 均匀增强=0,不均匀增强=1 转移=0,未转移=1 |
表5 PTC患者CLNM的多因素Logistic回归分析 |
因素 | 回归系数 | 标准误差 | 优势比(OR值) | P值 | 95%CI |
---|---|---|---|---|---|
年龄 | 1.565 | 0.528 | 3.696 | 0.013 | 1.749~13.070 |
是否接触被膜 | 1.576 | 0.576 | 4.833 | 0.006 | 1.562~14.955 |
增强程度 | 1.307 | 0.528 | 3.696 | 0.013 | 1.312~10.410 |
增强均匀性 | 1.661 | 0.573 | 5.264 | 0.004 | 1.397~12.123 |
常量 | -4.017 | 0.894 | 0.018 | <0.001 |
注:CLNM为颈中央区淋巴结转移;PTC为甲状腺乳头状癌;CI为置信区间 |
表6 预测模型以及模型内各变量参数对PTC患者CLNM的诊断效能比较 |
参数 | AUC(95%CI) | P值 | SEN(%) | SPE(%) | PPV(%) | NPV(%) |
---|---|---|---|---|---|---|
预测模型 | 0.800(0.717~0.883) | - | 70.8 | 78.3 | 72.3 | 77.0 |
年龄≤42岁 | 0.681(0.579~0.783) | 0.008 | 72.9 | 63.3 | 61.4 | 74.5 |
接触被膜 | 0.592(0.485~0.699) | <0.001 | 83.3 | 35.0 | 50.6 | 72.4 |
等或高增强 | 0.652(0.547~0.757) | 0.005 | 60.4 | 70.0 | 61.7 | 68.8 |
不均匀增强 | 0.696(0.489~0.703) | <0.001 | 79.2 | 40.0 | 51.4 | 70.6 |
注:PTC为甲状腺乳头状癌;CLNM为中央区淋巴结转移;CI为置信区间;AUC为ROC曲线下面积;SEN为敏感度;SPE为特异度;PPV为阳性预测值;NPV为阴性预测值;P值为预测模型与模型内各单一因素AUC比较所得 |
1 |
|
2 |
|
3 |
|
4 |
|
5 |
|
6 |
|
7 |
|
8 |
|
9 |
|
10 |
|
11 |
|
12 |
|
13 |
|
14 |
|
15 |
|
16 |
|
17 |
|
18 |
|
19 |
|
20 |
|
21 |
|
/
〈 |
|
〉 |