基于临床多模态超声特征的移植肾排异反应预测模型构建

Construction of predictive model for renal allograft rejection based on clinical multimodal ultrasound features

目的联合多模态超声特征和临床资料构建列线图预测移植肾排异反应发生状态。方法回顾性分析2021年1月至2022年3月在东部战区总医院经肾活检明确病变类型的102例移植肾患者的超声图像特征及临床资料,依据Banff移植肾病理学诊断标准(2017版)将患者分为排异组与肾病组。应用多因素Logistic回归筛选与排异反应发生状态相关的独立预测因素,基于独立预测因素建立回归模型并绘制列线图。采用Bootstrap法对列线图进行内部验证,并绘制ROC曲线、校准曲线评估列线图诊断效能。结果血尿素氮浓度、肾主动脉阻力指数、绝对达峰时间、皮质回声是排异反应发生的独立预测因素(OR=1.073、1.078、0.843、0.205,均P<0.05)。基于血尿素氮浓度、肾主动脉阻力指数、绝对达峰时间、皮质回声所构建的排异反应发生状态预测模型,AUC为0.814(95%CI=0.722~0.905),根据约登指数确定截断值为0.67(对应总分约为157分),内部验证(AUC=0.788)及校准曲线均显示列线图具有较高的预测能力。结论基于多模态超声特征及临床资料构建的移植肾患者排异反应发生状态预测模型,可用于指导移植肾功能不全患者的个体化治疗。

Objective To construct a nomogram for predicting the occurrence of renal allograft rejection based on the combination of multimodal ultrasound features and clinical data.Methods The ultrasound findings and clinical characteristics of 102 patients with transplanted kidneys who underwent renal biopsy in the General Hospital of Eastern Theater Command from January 2021 to March 2022 were analyzed retrospectively.Patients were divided into rejection group and nephropathy group according to Banff transplant kidney pathological diagnostic criteria(2017 edition).Multivariate Logistic regression was used to screen independent predictors related to the status of rejection,and nomograms were drawn based on the independent predictors.The internal validation of the nomogram was carried out by Bootstrap method,and the ROC curve and calibration curve were utilized to evaluate the diagnostic efficacy of the nomogram.Results Blood urea nitrogen concentration,renal aortic resistance index,absolute time to peak and cortical echo were independent predictors of rejection(OR=1.073,1.078,0.843,0.205;all P<0.05).Incorporating blood urea nitrogen concentration,renal aortic resistance index,absolute peak time and cortical echo,the nomogram was constructed.The AUC of the predictive model was 0.814(95%CI=0.722-0.905)and the cutoff value was 0.67(corresponding to a total score of about 157 points).Both internal verification(AUC=0.788)and calibration curve demonstrated the clinical usefulness of the nomogram.Conclusions The nomogram for predicting the occurrence of rejection in renal allograft patients based on multimodal ultrasound features and clinical data can guide the individualized treatment of patients with renal dysfunction.