基于MRI影像学特征联合肿瘤标志物建立直肠癌壁外血管侵犯模型的术前预测效能分析

Preoperative predictive performance of the model based on MRI features combined with tumor markers for extramural vascular invasion in rectal cancer

目的探讨基于磁共振成像(MRI)影像学特征联合肿瘤标志物建立直肠癌壁外血管侵犯(EMVI)模型的术前预测效能。方法选取2021年1月—2024年1月阜阳市人民医院收治的107例直肠癌患者,所有患者行病理检查、MRI检查,术后根据直肠癌是否发生EMVI分为阴性组和阳性组。比较两组临床资料、MRI影像学特征、肿瘤标志物,分析直肠癌发生EMVI的影响因素。分析基于MRI影像学特征联合肿瘤标志物建立的Logistic回归模型对直肠癌EMVI阳性的预测价值。结果107例直肠癌患者中,34例发生EMVI。两组年龄、直肠癌家族史、性别、肿瘤分化程度、体质量指数比较,差异均无统计学意义(P>0.05)。阳性组T3、4分期、环周切缘阳性率、基于高分辨率磁共振成像的EMVI(mrEMVI)评分>2分占比均高于阴性组(P<0.05)。阳性组血清癌胚抗原(CEA)、糖类抗原199(CA199)水平均高于阴性组(P<0.05)。多因素一般Logistic回归分析结果显示:影像T分期[O^R=5.703(95%CI:2.507,12.973)]、环周切缘[O^R=3.554(95%CI:1.562,8.084)]、mrEMVI评分[O^R=3.777(95%CI:1.660,8.593)]、血清CEA[O^R=4.104(95%CI:1.804,9.336)]、CA199[O^R=4.559(95%CI:2.004,10.370)]为直肠癌患者发生EMVI的影响因素(P<0.05)。基于MRI影像学特征联合肿瘤标志物构建Logistic回归模型,Log(P)=-4.573+1.741影像T分期+1.268环周切缘+1.329mrEMVI评分+1.412血清CEA+1.517血清CA199,拟合效果良好(P<0.05)。受试者工作特征曲线结果显示,该模型术前预测直肠癌EMVI阳性的敏感性为91.18%(95%CI:0.752,0.977)、特异性为89.04%(95%CI:0.790,0.948),曲线下面积为0.912(95%CI:0.857,0.964)(P<0.05)。结论基于影像T分期、环周切缘、mrEMVI评分、血清CEA、CA199水平联合建立的Logistic回归模型用于术前预测直肠癌EMVI状态具有良好的价值。

Objective To evaluate the preoperative predictive performance of the model based on magnetic resonance imaging(MRI)features and tumor markers for extramural vascular invasion(EMVI)in rectal cancer.Methods The 107 patients with rectal cancer admitted to our hospital from January 2021 to January 2024 were selected.All patients underwent pathological examination and MRI scans,and were divided into negative group and positive group according to the occurrence of EMVI.The clinical data,MRI features and tumor markers were compared between the positive group and the negative group,and the influencing factors for EMVI in rectal cancer were analyzed. The predictive value of the Logistic regression model based on MRI features combined with tumor markers for EMVI in rectal cancer was assessed. Results Among the 107 patients with rectal cancer, 34 cases developed EMVI. There was no difference in age, family history of rectal cancer, sex composition, degree of tumor differentiation and body mass index (BMI) between the two groups (P > 0.05). The proportion of stage T3 and T4 tumors, the rate of positive circumferential resection margin, and the proportion of MRI-detected EMVI (mrEMVI) score > 2 in the positive group were all higher than those in the negative group (P < 0.05). The serum levels of carcinoembryonic antigen (CEA) and carbohydrate antigen 199 (CA199) in the positive group were higher than those in the negative group (P < 0.05). Multivariable Logistic regression analysis showed that the T stage [O^R= 5.703 (95% CI: 2.507, 12.973) ], circumferential resection margin [O^R = 3.554 (95% CI: 1.562, 8.084) ], mrEMVI scores [O^R =3.777 (95% CI: 1.660, 8.593) ], and serum levels of CEA [O^R = 4.104 (95% CI: 1.804, 9.336) ] and CA199 [O^R = 4.559 (95% CI: 2.004, 10.370) ] were factors affecting the occurrence of EMVI in rectal cancer (P < 0.05). The Logistic regression model was constructed based on MRI features and tumor markers as Log (P) = -4.573 + 1.741 × T stage + 1.268 × circumferential resection margin + 1.329 × mrEMVI score + 1.412 × serum CEA + 1.517 × serum CA199 and was well-fitted (P < 0.05). The receiver operating characteristic (ROC) curve analysis showed that the sensitivity of the regression model in predicting EMVI in rectal cancer before surgery was 91.18% (95% CI: 0.752, 0.977), with a specificity of 89.04% (95% CI: 0.790, 0.948), and an area under the curve of 0.912 (95% CI: 0.857, 0.964) (P < 0.05). Conclusion The Logistic regression model based on the T stage, circumferential resection margin, mrEMVI scores, and serum levels of CEA and CA199 is of great value for preoperative prediction of EMVI in rectal cancer.