基于FNA-Tg和血清Tg的二元Logit回归模型对甲状腺癌淋巴结转移或复发的判定价值

Value of Binary Logit Regression Model Based on FNA-Tg and Serum Tg for Determining Lymph Node Metastasis or Recurrence of Thyroid Cancer

目的 通过二元Logit回归分析患者淋巴结细针细胞学穿刺洗脱液甲状腺球蛋白测定值(FNATg值)、血清甲状腺球蛋白值(血清Tg值)及其比值,建立对甲状腺癌淋巴结转移、复发情况的预测模型。方法 (1)收集术前为分化型甲状腺癌伴可疑淋巴结转移患者64例。(2)通过ROC曲线初步评估FNA-Tg值、血清Tg值及其比值对淋巴结良恶性鉴别的诊断价值。(3)使用二元Logit回归分析进一步评价上述3项指标的诊断价值并建立预测模型。结果 (1)ROC曲线:淋巴结FNA-Tg对应的AUC值为0.998(95%CI 99.27%~100.34%),切点值为0.82;血清Tg对应的AUC值为0.824(95%CI 70.64%~94.14%),切点值为18.27;淋巴结FNA-Tg/血清Tg对应的AUC值为1.000(95%CI 100.00%~100.00%),切点值为0.461。(2)二元Logit回归全进入法3项自变量同时存在,二元Logit回归全进入法3项自变量同时存在,FNA-Tg值、血清Tg值及其比值3项P值均> 0.05,模型3项自变量同时存在对淋巴结良恶性无诊断价值;二元Logit回归全进入法淋巴结FNATg单一变量。淋巴结FNA-Tg可以解释88%淋巴结良恶性的变化,P=0.038 <0.05,回归系数2.68,OR值14.587,预测模型公式为:In(p/1-p)=-4.122+2.680*淋巴结FNA-Tg,似然比检验P=0.000 <0.001,AIC值14.386,BIC值18.704,总体预测准确率为95.31%;二元Logit回归全进入法血清Tg单一变量,血清Tg能解释淋巴结良恶性27.8%的变化原因,P=0.000 <0.001,回归系数0.164,OR值1.179,预测模型公式为:In(p/1-p)=-2.466+0.164*血清Tg,似然比检验P=0.000 <0.001,AIC值67.33,BIC值71.648,总体预测准确率为75%。结论 术前FNA-Tg、血清Tg及其比值在判定甲状腺结节良恶性有很好的诊断价值。基于FNA-Tg、血清Tg的二元Logit回归模型公式能较好的预测甲状腺癌患者淋巴结转移、复发的情况。

Objective To establish a prediction model for lymph node metastasis and recurrence of thyroid cancer by analyzing patients’ lymph node FNA-Tg values, serum Tg values, and their ratios by binary logit regression. Methods 1. 64 patients with preoperative differentiated thyroid cancer with suspected lymph node metastasis were collected. 2. The diagnostic value of FNA-Tg value, serum Tg value, and their ratios for the differentiation of benign and malignant lymph nodes were initially evaluated by the ROC curve. 3. The diagnostic value of the above three indices was further evaluated by binary logit regression analysis and a prediction model was established. Results 1. ROC curve:the AUC value corresponding to lymph node FNA-Tg was 0.998(95% CI99.27%-100.34%),with a cut point value of 0.82;the AUC value corresponding to serum Tg was 0.824(95% CI70.64%-94.14%),with a cut point value of 18.27;the AUC value corresponding to lymph node FNA-Tg/serum Tg The AUC value of FNA-Tg/serum Tg in lymph nodes was 1.000(95% CI 100.00%-100.00%),and the cut point value was 0.461. 2. Binary logit regression all-entry method with the simultaneous presence of three independent variables:i) Binary logit regression all-entry method with the simultaneous presence of three independent variables:FNA-Tg value,serum Tg value and their ratios all had P values > 0.05,and the model The simultaneous presence of three independent variables has no diagnostic value for benign and malignant lymph nodes;ii) Binary Logit regression all-entry method lymph node FNA-Tg single variable results:lymph node FNA-Tg could explain 88% of the variation in benign and malignant lymph nodes, P = 0.038 < 0.05, regression coefficient 2.68, OR value14.587,prediction model formula was:In(p/1-p) =-4.122 + 2.680* lymph node FNA-Tg,likelihood ratio test P = 0.000 < 0.001,AIC value 14.386,BIC value 18.704,overall prediction accuracy of 95.31%;iii) Binary Logit regression all-entry method serum Tg single variable results:serum Tg explained 27.8% of the variation in benign and malignant lymph nodes, P = 0.000 < 0.001, regression coefficient 0.164, OR 1.179, prediction model formula:In(p/1-p) =-2.466 + 0.164*serum Tg,likelihood ratio test P = 0.000 < 0.001,AIC value 67.33,BIC value 71.648,overall prediction accuracy 75%. Conclusions 1. Preoperative FNA-Tg,serum Tg and their ratios have good diagnostic value in determining the benignity and malignancy of thyroid nodules. 2. The binary logit regression model(stepwise method) formula based on serum Tg can better predict the benignity and malignancy of thyroid nodules.