IMPACT和CRASH模型对创伤性颅脑损伤患者预后评估价值的比较研究

Prognostic Value of IMPACT and CRASH Models in the Assessment of Patients with Traumatic Brain Injury:a Comparative Study

背景国际颅脑损伤预后临床测试研究(IMPACT)和重型颅脑损伤后皮质类固醇的随机化研究(CRASH模型是国际上具有影响力的创伤性颅脑损伤(TBI)预后预测模型,需要继续开发,完善和持续的外部验证,以确保对不同环境的普适性。目的同时在中国TBI人群中进行验证IMPACT和CRASH模型的预后评估价值并进行比较。方法选取2017—2019年在郑州大学附属郑州中心医院内接受治疗的TBI患者210例为研究对象,收集纳入患者的基本信息。随访观察患者14 d存活情况和6个月格拉斯哥预后评分(GOS),随访截止时间为2021年6月,终止事件为中途失访。绘制受试者工作特征曲线(ROC曲线)评估IMPACT和CRASH模型对TBI患者预后的预测效能,计算ROC曲线下面积(AUC)。采用Brier评分评价模型的校准度。结果患者平均年龄(54.0±17.4)岁,分别绘制IMPACT模型与CRASH模型预测TBI患者预后的ROC曲线,结果显示IMPACT核心模型、CT模型、实验室模型预测TBI患者6个月GOS预后不良的AUC分别为0.807(95%CI=0.747~0.866,P<0.001)、0.843(95%CI=0.789~0.897,P<0.001)、0.845(95%CI=0.793~0.897,P<0.001),Brier评分分别为0.179、0.164、0.161;IMPACT核心模型、CT模型、实验室模型预测TBI患者6个月死亡的AUC分别为0.868(95%CI=0.816~0.919,P<0.001)、0.896(95%CI=0.851~0.941,P<0.001)、0.892(95%CI=0.850~0.935,P<0.001),Brier评分分别为0.151、0.144、0.136。CRASH基本模型、CT模型预测TBI患者6个月GOS预后不良的AUC分别为0.747(95%CI=0.682~0.813,P<0.001)、0.766(95%CI=0.703~0.829,P<0.001),Brier评分分别为0.306、0.308;CRASH基本模型、CT模型预测TBI患者14 d死亡的AUC分别为0.791(95%CI=0.723~0.860,P<0.001)、0.797(95%CI=0.728~0.865,P<0.001);Brier评分分别为0.348、0.383。结论对于TBI患者的预后,IMPACT模型整体较CRASH模型显示出较好的预测能力。

Background International Mission on Prognosis and Analysis of Clinical Trial(IMPACT)and Corticosteroid Randomisation after Significant Head Injury(CRASH)are internationally influential prognostic models for traumatic brain injury(TBI),which need to be continuously developed,improved and continuously verified externally to ensure generalizability to different settings.Objective To verify and compare the prognostic evaluation values of IMPACT and CRASH simultaneously in TBI population in China.Methods A total of 210 TBI patients treated in Zhengzhou Central Hospital Affiliated to Zhengzhou University from 2017 to 2019 were retrospectively selected as the study objects,and the basic information of the included patients was collected.The 14-day survival and 6-month Glasgow prognostic score(GOS)of the patients were followed up until June 2021,with a termination event of loss of follow-up.Receiver operating characteristic(ROC)curve was plotted to evaluate the predictive efficacy of IMPACT and CRASH models in TBI patients,and the area under ROC curve(AUC)was calculated.Brier score was used to evaluate the calibration degree of the model.Results The average age of the patients was(54.0±17.4)years.The ROC curves of the IMPACT model and CRASH model to predict the prognosis of TBI patients were plotted,respectively,and the results showed that the AUCs of IMPACT core model,CT model and laboratory model to predict the poor prognosis of GOS in TBI patients at 6 months was 0.807(95%CI=0.747-0.866,P<0.001),0.843(95%CI=0.789-0.897,P<0.001),0.845(95%CI=0.793-0.897,P<0.001),Brier scores were 0.179,0.164,0.161,respectively;the AUCs of IMPACT core model,CT model and laboratory model predicting 6-month death in TBI patients were 0.868(95%CI=0.816-0.919,P<0.001),0.896(95%CI=0.851-0.941,P<0.001),0.892(95%CI=0.850-0.935,P<0.001),and Brier scores were 0.151,0.144 and 0.136,respectively.The AUCs of CRASH basic model and CT model to predict the poor prognosis of GOS in TBI patients at 6 months was 0.747(95%CI=0.682-0.813,P<0.001)and 0.766(95%CI=0.703-0.829,P<0.001);Brier scores were 0.306 and 0.308,respectively.The AUCs of CRASH basic model and CT model for predicting 14-day death of TBI patients were 0.791(95%CI=0.723-0.860,P<0.001)and 0.797(95%CI=0.728-0.865,P<0.001),the Brier scores were 0.348 and 0.383,respectively.Conclusion For the prognosis of TBI patients,the IMPACT model showed better overall predictive capacity than the CRASH model.