基于动脉自旋标记灌注参数和临床病理特征的列线图预测中晚期鼻咽癌放化疗疗效

目的探讨基于动脉自旋标记(ASL)MRI灌注参数和临床病理特征构建的列线图对中晚期鼻咽癌(ANPC,Ⅲ和Ⅳ期)放化疗疗效的预测价值。方法前瞻性纳入2018年6月至2021年1月江南大学附属医院经病理证实的70例ANPC患者,在放化疗前行鼻咽MR平扫、ASL和增强扫描,放化疗疗程结束后1周内行常规MR复查。记录放化疗前的ASL灌注参数肿瘤血流量(TBF)和临床病理特征,并在T1WI图像上测量肿瘤最大径(MD)。根据实体瘤疗效评价标准将患者分为放化疗有效组(48例)和无效组(22例)。采用独立样本t检验比较有效组与无效组间TBF、年龄、MD的差异,以χ²检验比较两组间性别、临床分期、病理类型的差异;使用二元逻辑回归分析分别构建临床病理模型和TBF、临床病理联合模型,并建立联合模型的列线图。采用受试者操作特征(ROC)曲线获得模型的诊断效能,采用DeLong法比较模型间曲线下面积(AUC)的差异。建立列线图的校准曲线,并获得一致性指数(C指数)。结果放化疗有效组和无效组的TBF分别为(113±9)、(97±14)ml·100 g^-1·min^-1,差异有统计学意义(t=5.17,P<0.001),放化疗有效组的MD值小于无效组,差异有统计学意义(t=-2.24,P=0.028)。2组间临床分期和病理类型差异均有统计学意义(χ²值分别为12.21、12.95,P均<0.001)。通过逻辑回归分析纳入3个独立预测因子,包括TBF(OR值7.749)、临床分期(OR值0.129)及病理类型(OR值5.228)。TBF模型预测放化疗疗效的AUC为0.843,灵敏度为87.5%,特异度为72.7%;临床病理模型的AUC为0.822,灵敏度为80.2%,特异度为59.1%;联合模型列线图的AUC为0.893,灵敏度为81.2%,特异度为90.9%。联合模型列线图与TBF模型的AUC比较差异无统计学意义(Z=1.23,P=0.215),但高于临床病理模型的AUC,差异有统计学意义(Z=2.47,P=0.031)。校准曲线显示联合模型列线图预测值与临床实际观察值间具有较好的一致性,C指数为0.892。结论TBF、临床分期和病理类型是ANPC患者放化疗疗效的独立预测因子,基于此3项因素构建的列线图在预测放化疗疗效方面具有较高的效能。

Nomogram for predicting the response to chemoradiotherapy in advanced nasopharyngeal carcinoma based on arterial spin-labeled perfusion parameters and clinicopathological features

Objective To explore the value of nomogram based on arterial spin labeling(ASL)MRI perfusion parameters and clinicopathological features in predicting the response to chemoradiotherapy(CRT)in advanced nasopharyngeal carcinoma(ANPC,stageⅢandⅣ).Methods From June 2018 to January 2021,70 patients with ANPC confirmed by pathology were prospectively enrolled in Affiliated Hospital of Jiangnan University.Nasopharyngeal MRI plain scan,ASL and contrast-enhanced scan were performed before CRT,and routine MRI re-examination was performed within 1 week after the end of CRT.The pre-CRT perfusion parameter tumor blood flow(TBF)from ASL and clinicopathological features were recorded,and the maximum diameter(MD)of the tumor on T1WI images was measured.The patients were divided into CRT effective group(48 cases)and ineffective group(22 cases)according to the response evaluation criteria in solid tumors.The independent sample t test was used to compare the differences of TBF,age and MD between effective group and ineffective group.Theχ² test was used to compare the differences of gender,clinical stage and pathological type between the 2 groups.Using binary logistic regression analysis,clinicopathological model and TBF combined clinicopathological model were constructed,and the nomogram of combined model was constructed.The diagnostic efficacy of the models was obtained by receiver operating characteristic(ROC)curve analysis,and the area under the ROC curves(AUC)of the 3 models were compared by DeLong method.The calibration curve for the nomogram was generated,and the concordance index(C index)was acquired.Results The TBF of the effective group and the ineffective group were(113±9)and(97±14)ml·100 g^-1·min^-1,with a statistical difference(t=5.17,P<0.001).The MD value of the effective group was smaller than that of the ineffective group,with a statistical difference(t=-2.24,P=0.028).There were statistical differences in clinical stage and pathological type between the 2 groups(χ² values were 12.21 and 12.95,respectively,both P<0.001).Three independent predictors,including TBF(OR=7.749),clinical stage(OR=0.129)and pathological type(OR=5.228),were included in logistic regression analysis.The AUC,sensitivity and specificity of TBF model in predicting the response to CRT were 0.843,87.5%and 72.7%,of clinicopathological model were 0.822,80.2%and 59.1%,of the nomogram model were 0.893,81.2%and 90.9%.There was no statistical difference of AUC between the nomogram model and TBF model(Z=1.23,P=0.215).However,the AUC of the nomogram model was greater than that of the clinicopathological model(Z=2.47,P=0.031).The calibration curve showed that there was a good concordance index(C index=0.892)between the predicted value of nomogram and the actual clinical observation value.Conclusion TBF,clinical stage and pathological type are independent predictors of the response to CRT in ANPC patients,and the nomogram based on these three factors has a good ability in predicting the response to CRT.