Objective To investigate the value of serum uric acid for prognosis prediction of traumatic brain injury by discrimination and calibration.Methods A total of 330 patients with traumatic brain injury (GCS 3~14) admitted to the First Affiliated Hospital of Soochow University between Nov.2010 and Oct.2012 were enrolled.Patients were divided into a survival group (GOS 2~5) and a death group (GOS 1).Logistic regression models were created with serum uric acid, blood glucose and serum uric acid + blood glucose, respectively.Discrimination and calibration were evaluated by ROC curve, calibration curve and Hosmer-Lemeshow test.Results AUCs of serum uric acid, blood glucose and uric acid+blood glucose were 0.7178, 0.7809 and 0.8134, respectively.The p values of Hosmer-Lemeshow test for serum uric acid, blood glucose, serum uric acid + blood glucose were 0.1508, 0.3366 and 0.8179, respectively.Calibration curves of them were close to the line of perfect prediction.Conclusion Hyperuricemia in the early stage is a risk factor for patients of TBI, and it could be used as a novel predictor owing to its good discrimination and calibration for prognosis prediction of TBI.%目的 通过分析血尿酸对创伤性脑损伤(TBI)死亡预测的区分度与校准度评价其对TBI预后判断的价值.方法 回顾性收集我院2010-11~2012-10收治的330例TBI患者病历资料:男性233例(70.6%),年龄(48.6±18.1)岁.病例入选标准:明确诊断为TBI,受伤后24 h以内有血尿酸数据,GCS评分3~14分,除外年龄<14岁、孕妇、既往有自身免疫性疾病、痛风病史及服用别嘌呤醇患者.血糖及血尿酸数据为入院次日早晨空腹静脉血测得,同时记录年龄、瞳孔反应和GCS评分等临床数据,电话随访伤后6个月格拉斯哥预后评分(GOS),按是否在随访期内死亡将所有患者分为死亡组(GOS 1分)和生存组(GOS 2~5分).分别以血尿酸、血糖、血尿酸+血糖为自变量,建立Logistic回归模型预测TBI死亡风险,采用ROC曲线下面积、校准曲线及Hosmer-Lemeshow拟合优度检验,比较血尿酸与血糖对TBI死亡预测的区分度和校准度.结果 两组的性别构成差异无统计学意义(P=0.066),死亡组患者表现为高龄、瞳孔反应差、GCS评分低和血尿酸水平高,与生存组比较差异有统计学意义(P<0.05).血糖AUC(0.7809, 95%CI 0.71986~0.84192)略高于血尿酸(0.7178, 95%CI 0.64997~0.78562),但差异无统计学意义(P=0.147);血尿酸+血糖模型的AUC提高到0.8134, 95%CI 0.75683~0.87002.Hosmer-Lemeshow拟合优度检验,得到血尿酸、血糖预测模型对应的P值为0.1508 vs.0.3366(均>0.05).血尿酸+血糖模型Hosmer-Lemeshow拟合优度检验的P值提高到0.8179.校准曲线显示,血尿酸和血糖的校准曲线均靠近45度线,两者联合的校准曲线进一步贴近45度线.结论 高尿酸是TBI患者预后不良的危险因素,对死亡预测具有较高的区分度和校准度,是一个简单、有效的TBI诊疗决策参考指标.
展开▼
