Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics.
Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics.
复制标题
DOI：
10.1371/journal.pmed.0050165
复制
发表时间：
2008-08-05
期刊：
影响因子：
15.8
通讯作者：
Maas AI
中科院分区：
文献类型：
--
作者：
Steyerberg EW;Mushkudiani N;Perel P;Butcher I;Lu J;McHugh GS;Murray GD;Marmarou A;Roberts I;Habbema JD;Maas AI
创伤性脑损伤（TBI）是导致死亡和残疾的主要原因。对入院时预后的可靠预测具有重要的临床意义。我们的目标是开发具有现成的传统和新型预测因子的预后模型。前瞻性收集的个体患者资料分析了11项研究。我们考虑了入院时可用的logistic回归模型预测因子，根据格拉斯哥结局量表预测损伤后6个月的死亡率和不良结局。在8,509名重度或中度TBI患者中建立了预后模型，通过依次省略11项研究进行交叉验证。外部验证来自最近医学研究委员会重大头部损伤后皮质类固醇随机化（MRC CRASH）试验的6681名患者。我们发现，年龄、运动评分、瞳孔反应性和CT特征（包括外伤性蛛网膜下腔出血）是影响预后的最重要因素。交叉验证时，结合年龄、运动评分和瞳孔反应性的预后模型的受试者工作特征曲线下面积（AUC）在0.66至0.84之间。考虑到CT特征、继发性损伤（低血压和缺氧）以及实验室参数（血糖和血红蛋白），该性能可以得到改善（AUC增加约0.05）。外部验证表明，该模型判别能力良好（AUC为0.80）。结果系统地比预期的差，但在CRASH试验中来自高收入国家的1588例患者中，结果较预期的差。使用基线特征的预后模型可以充分区分TBI后6个月预后好坏的患者，特别是在考虑CT和实验室结果以及传统预测因素的情况下。模型预测可以支持临床实践和研究，包括随机对照试验的设计和分析。Ewout Steyerberg和他的同事利用入院时的数据描述了一个预测外伤性脑损伤结果的预后模型。外伤性脑损伤（TBI）在世界范围内引起了大量的发病率和死亡率。例如，根据疾病控制中心的数据，每年大约有140万美国人会遭受tbi（头部受伤）。其中，110万人将接受急诊治疗并出院，23.5万人将住院，5万人将死亡。发展中国家的疾病负担要高得多，交通事故等创伤性脑损伤的病因发生率较高，治疗可能较少。考虑到治疗创伤性脑损伤所需的资源，一个非常有用的研究工具将是在入院时准确预测特定损伤的结果的能力。目前，诸如格拉斯哥昏迷评分（Glasgow Coma Scale）之类的评分对预测损伤后24小时的预后有用，但对损伤前的预后无效。预测模型之所以有用，有几个原因。在临床上，它们帮助医生和病人决定治疗方案。它们在比较不同患者组的结果和计划随机对照试验的研究中也很有用。在过去几年中，IMPACT研究小组使用了一个大型数据库，其中包括1984年至1997年间进行的8项随机对照试验和3项观察性研究的数据，该研究是IMPACT研究小组所做的众多分析之一。还有其他正在进行的研究也在寻求开发新的预后模型；最近的一项研究发表在《英国医学杂志》（BMJ）上，该研究小组包括《公共科学图书馆医学》（PLoS Medicine）论文的主要作者。作者分析了从数据库中包含的11项研究中收集的个体患者的前瞻性数据，并导出模型来预测8,509名重度或中度TBI患者损伤后6个月的死亡率和不良预后。他们发现，年龄、运动评分、瞳孔反应性和CT扫描特征（包括外伤性蛛网膜下腔出血）是影响预后的最重要因素。核心预后模型可由年龄、运动评分和瞳孔反应性综合得出。结合CT特征、继发问题（低血压和缺氧）以及实验室血糖和血红蛋白的测量，可以得到更好的评分。然后对这些分数进行测试，看看它们对另一组患者的预测结果有多好——6681名患者来自最近医学研究委员会重大头部损伤后皮质类固醇随机化（MRC CRASH）试验。在这篇论文中，作者表明，利用入院时收集的特征作为常规护理的一部分来建立预后模型是可能的，该模型可以区分TBI后6个月预后良好和不良的患者，特别是如果将CT扫描和实验室结果添加到基本模型中。这篇论文必须与其他研究一起考虑，尤其是上面提到的最近发表在BMJ杂志上的论文（MRC CRASH Trial合作者[2008]预测创伤性脑损伤后的预后：基于国际大队列患者的实用预后模型）。英国医学杂志336:425-429。《英国医学杂志》的研究提出了一套相似但略有不同的模型，特别关注发展中国家的患者；在这种情况下，使用CRASH试验中的患者来生成模型，使用IMPACT数据库中的患者来验证模型的一个变体。不幸的是，在最初的审查过程中，这篇相关的论文没有向我们披露；然而，在《公共科学图书馆·医学》随后对这篇手稿的审议中，我们了解到了这一点。在与审稿人讨论后，我们认为PLoS Medicine论文中描述的模型与另一篇论文中报道的模型有很大不同，因此继续发表这篇论文。然而，理想情况下，这两组模型应该同时被审查和发表，这样读者就可以很容易地根据彼此来评估两组不同模型的各自优点和价值。然而，同样发表在PLoS Medicine上的一篇Perspective文章讨论了这两组模型（见下文）。请通过本摘要的在线版本http://dx.doi.org/10.1371/journal.pmed.0050165访问这些网站。Andrews和Young在PLoS Medicine Perspective的一篇文章中进一步讨论了这篇论文和上面提到的BMJ论文。TBI Impact网站提供了一个工具来计算本文中描述的分数。CRASH试验用于验证这里提到的分数，有一个网站来解释试验及其结果。R软件用于预后分析。MedlinePlus百科全书有关于头部损伤的信息，世卫组织关于神经创伤的网站从全球角度讨论了头部损伤，疾病预防控制中心的国家伤害预防和控制中心提供了美国头部损伤的统计数据和预防建议
Traumatic brain injury (TBI) is a leading cause of death and disability. A reliable prediction of outcome on admission is of great clinical relevance. We aimed to develop prognostic models with readily available traditional and novel predictors. Prospectively collected individual patient data were analyzed from 11 studies. We considered predictors available at admission in logistic regression models to predict mortality and unfavorable outcome according to the Glasgow Outcome Scale at 6 mo after injury. Prognostic models were developed in 8,509 patients with severe or moderate TBI, with cross-validation by omission of each of the 11 studies in turn. External validation was on 6,681 patients from the recent Medical Research Council Corticosteroid Randomisation after Significant Head Injury (MRC CRASH) trial. We found that the strongest predictors of outcome were age, motor score, pupillary reactivity, and CT characteristics, including the presence of traumatic subarachnoid hemorrhage. A prognostic model that combined age, motor score, and pupillary reactivity had an area under the receiver operating characteristic curve (AUC) between 0.66 and 0.84 at cross-validation. This performance could be improved (AUC increased by approximately 0.05) by considering CT characteristics, secondary insults (hypotension and hypoxia), and laboratory parameters (glucose and hemoglobin). External validation confirmed that the discriminative ability of the model was adequate (AUC 0.80). Outcomes were systematically worse than predicted, but less so in 1,588 patients who were from high-income countries in the CRASH trial. Prognostic models using baseline characteristics provide adequate discrimination between patients with good and poor 6 mo outcomes after TBI, especially if CT and laboratory findings are considered in addition to traditional predictors. The model predictions may support clinical practice and research, including the design and analysis of randomized controlled trials. Ewout Steyerberg and colleagues describe a prognostic model for the prediction of outcome of traumatic brain injury using data available on admission. Traumatic brain injury (TBI) causes a large amount of morbidity and mortality worldwide. According to the Centers for Disease Control, for example, about 1.4 million Americans will sustain a TBI—a head injury—each year. Of these, 1.1 million will be treated and released from an emergency department, 235,000 will be hospitalized, and 50,000 will die. The burden of disease is much higher in the developing world, where the causes of TBI such as traffic accidents occur at higher rates and treatment may be less available. Given the resources required to treat TBI, a very useful research tool would be the ability to accurately predict on admission to hospital what the outcome of a given injury might be. Currently, scores such as the Glasgow Coma Scale are useful to predict outcome 24 h after the injury but not before. Prognostic models are useful for several reasons. Clinically, they help doctors and patients make decisions about treatment. They are also useful in research studies that compare outcomes in different groups of patients and when planning randomized controlled trials. The study presented here is one of a number of analyses done by the IMPACT research group over the past several years using a large database that includes data from eight randomized controlled trials and three observational studies conducted between 1984 and 1997. There are other ongoing studies that also seek to develop new prognostic models; one such recent study was published in BMJ by a group involving the lead author of the PLoS Medicine paper described here. The authors analyzed data that had been collected prospectively on individual patients from the 11 studies included in the database and derived models to predict mortality and unfavorable outcome at 6 mo after injury for the 8,509 patients with severe or moderate TBI. They found that the strongest predictors of outcome were age, motor score, pupillary reactivity, and characteristics on the CT scan, including the presence of traumatic subarachnoid hemorrhage. A core prognostic model could be derived from the combination of age, motor score, and pupillary reactivity. A better score could be obtained by adding CT characteristics, secondary problems (hypotension and hypoxia), and laboratory measurements of glucose and hemoglobin. The scores were then tested to see how well they predicted outcome in a different group of patients—6,681 patients from the recent Medical Research Council Corticosteroid Randomisation after Significant Head Injury (MRC CRASH) trial. In this paper the authors show that it is possible to produce prognostic models using characteristics collected on admission as part of routine care that can discriminate between patients with good and poor outcomes 6 mo after TBI, especially if the results from CT scans and laboratory findings are added to basic models. This paper has to be considered together with other studies, especially the paper mentioned above, which was recently published in the BMJ (MRC CRASH Trial Collaborators [2008] Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. BMJ 336: 425–429.). The BMJ study presented a set of similar, but subtly different models, with specific focus on patients in developing countries; in that case, the patients in the CRASH trial were used to produce the models, and the patients in the IMPACT database were used to verify one variant of the models. Unfortunately this related paper was not disclosed to us during the initial review process; however, during PLoS Medicine's subsequent consideration of this manuscript we learned of it. After discussion with the reviewers, we took the decision that the models described in the PLoS Medicine paper are sufficiently different from those reported in the other paper and as such proceeded with publication of the paper. Ideally, however, these two sets of models would have been reviewed and published side by side, so that readers could easily evaluate the respective merits and value of the two different sets of models in the light of each other. The two sets of models are, however, discussed in a Perspective article also published in PLoS Medicine (see below). Please access these Web sites via the online version of this summary at http://dx.doi.org/10.1371/journal.pmed.0050165. This paper and the BMJ paper mentioned above are discussed further in a PLoS Medicine Perspective article by Andrews and Young The TBI Impact site provides a tool to calculate the scores described in this paper The CRASH trial, which is used to validate the scores mentioned here, has a Web site explaining the trial and its results The R software, which was used for the prognostic analyses, is freely available The MedlinePlus encyclopedia has information on head injury The WHO site on neurotrauma discusses head injury from a global perspective The CDC's National Center for Injury Prevention and Control gives statistics on head injury in the US and advice on prevention
登录
查看更多内容
影响因子：
3.3
作者：
Enders, Craig K.
通讯作者：
Enders, Craig K.
影响因子：
4.2
作者：
Marmarou, Anthony;Lu, Juan;Maas, Andrew I. R.
通讯作者：
Maas, Andrew I. R.
影响因子：
4.8
作者：
Hernández, AV;Steyerberg, EW;Maas, AIR
通讯作者：
Maas, AIR
影响因子：
4.8
作者：
Maas, AIR;Hukkelhoven, CWPM;Steyerberg, EW
通讯作者：
Steyerberg, EW
影响因子：
39.2
作者：
Justice, AC;Covinsky, KE;Berlin, JA
通讯作者：
Berlin, JA