深部白质DTI全局指标对ICU重型颅脑创伤患者1年预后的预测价值:MRI-COMA与CENTER-TBI联合研究
本研究利用两个多中心队列(29个中心),将MRI-COMA队列拆分为训练集(50例,2006年至2014年中期入组)与测试集(140例,2014年起临床常规随访),测试集再与CENTER-TBI队列56例ICU患者(2014–2020年入组)合并验证。以伤后7–35天深部白质部分各向异性与平均弥散率及受伤时年龄构建支持向量分类器,按1年GOSE 1–3为不良预后、GOSE 4–8为良好预后进行二分类。训练集ROC曲线下面积0.93;合并验证集预测不良预后敏感度49%、特异度96.8%,预测良好预后敏感度58.5%、特异度97.1%。作者认为该模型以适合ICU个体化决策的特异度,识别出约一半不良预后患者和约三分之二良好预后患者。
要点
- 模型输入为伤后7–35天深部白质部分各向异性、平均弥散率及受伤时年龄,输出1年GOSE二分类(1–3不良 vs 4–8良好)。
- 训练集(MRI-COMA-Train,50例)ROC曲线下面积0.93。
- 合并验证集(MRI-COMA-Test 140例+CENTER-TBI 56例)预测不良预后:敏感度49%、特异度96.8%。
- 合并验证集预测良好预后:敏感度58.5%、特异度97.1%。
- 模型特异度高而敏感度有限,约一半不良预后患者和约三分之二良好预后患者被识别。
适用人群与场景
- ICU重型颅脑创伤患者,伤后7–35天完成MRI弥散指标测量。
- 两个多中心队列共29个中心:MRI-COMA队列(NCT00577954)拆分为MRI-COMA-Train(50例,2006年至2014年中期入组)与MRI-COMA-Test(140例,2014年起临床常规随访),后者与CENTER-TBI队列(NCT02210221)56例ICU患者(2014–2020年入组)合并验证。
诊断标准与关键阈值
- 1年格拉斯哥结局扩展量表(GOSE)评分:GOSE 1–3为不良预后(UFO),GOSE 4–8为良好预后(FO)。
- MRI测量时间窗为伤后第7天至第35天。
- 模型纳入指标:深部白质部分各向异性、平均弥散率及受伤时年龄。
研究设计与关键结果
- 研究类型:基于两个多中心队列(29个中心)的预后模型开发与验证研究。
- 训练集MRI-COMA-Train(50例)ROC曲线下面积0.93。
- 合并验证集(MRI-COMA-Test 140例+CENTER-TBI 56例)预测不良预后:敏感度49%,特异度96.8%。
- 合并验证集预测良好预后:敏感度58.5%,特异度97.1%。
- 作者结论:模型以适合ICU个体化决策的特异度,识别出约一半1年不良预后患者和约三分之二良好预后患者。
证据类型、等级与局限
- 证据类型:多中心队列研究(预后模型开发与验证)。
- 局限:原文未报告局限性。
对床旁实践的改变
- 作者认为该模型可在ICU个体化决策中提供客观信息,其特异度与个体化决策过程相适应。
- 模型在合并验证集中识别约一半不良预后患者和约三分之二良好预后患者。
提炼说明
- 本条依据论文摘要整理;摘要未涉及「处置流程与药物剂量」「红旗征与不能漏的情况」「指南或共识出处」,故未列出这些栏目;全文请点「阅读原文」。
英文原文(摘要)· 点击展开
PURPOSE: A reliable tool for outcome prognostication in severe traumatic brain injury (TBI) would improve intensive care unit (ICU) decision-making process by providing objective information to caregivers and family. This study aimed at designing a new classification score based on magnetic resonance (MR) diffusion metrics measured in the deep white matter between day 7 and day 35 after TBI to predict 1-year clinical outcome.
METHODS: Two multicenter cohorts (29 centers) were used. MRI-COMA cohort (NCT00577954) was split into MRI-COMA-Train (50 patients enrolled between 2006 and mid-2014) and MRI-COMA-Test (140 patients followed up in clinical routine from 2014) sub-cohorts. These latter patients were pooled with 56 ICU patients (enrolled from 2014 to 2020) from CENTER-TBI cohort (NCT02210221). Patients were dichotomised depending on their 1-year Glasgow outcome scale extended (GOSE) score: GOSE 1-3, unfavorable outcome (UFO); GOSE 4-8, favorable outcome (FO). A support vector classifier incorporating fractional anisotropy and mean diffusivity measured in deep white matter, and age at the time of injury was developed to predict whether the patients would be either UFO or FO.
RESULTS: The model achieved an area under the ROC curve of 0.93 on MRI-COMA-Train training dataset, and 49% sensitivity for 96.8% specificity in predicting UFO and 58.5% sensitivity for 97.1% specificity in predicting FO on the pooled MRI-COMA-Test and CENTER-TBI validation datasets.
CONCLUSION: The model successfully identified, with a specificity compatible with a personalized decision-making process in ICU, one in two patients who had an unfavorable outcome at 1 year after the injury, and two-thirds of the patients who experienced a favorable outcome.
参考来源
- Puybasset L, Perlbarg V, Unrug J, et al. Prognostic value of global deep white matter DTI metrics for 1-year outcome prediction in ICU traumatic brain injury patients: an MRI-COMA and CENTER-TBI combined study. Intensive Care Med. 2022. PMID: 34904191. doi:10.1007/s00134-021-06583-z
供执业医师抢救与诊疗参考,不替代临床判断;剂量请结合体重、肝肾功能和本院药品规格复核,以指南原文为准。