Abstract：
BACKGROUND: Hyperglycemia in the acute phase of intracerebral hemorrhage (ICH) has been associated with poor functional outcomes, however all interventions to lower glucose have yielded neutral or negative results. We attempt an explanation of the causal role of hyperglycemia in ΙCH outcome using generalized structural equation modeling.
METHODS: Consecutive primary ICH patients admitted to an academic hospital between 2007 and 2018 were identified. Patients with missing baseline or follow up CT scans and without 90 day follow up status were excluded. We constructed a causal model accounting for pre-defined markers of ICH severity to evaluate the association between mean 72 h glucose and 90 day functional outcome measured by modified Rankin Scale, dichotomized as favorable ≤2 or unfavorable >2.
RESULTS: Primary analyses included 410 patients (70.4 ± 13.8years, 43 % female). Mean 72 h glucose was 137.5 ± 33.4mg/dl and 102 (25 %) patients were diabetic. On univariable analysis, higher glucose levels were negatively correlated with favorable outcome (p < 0.0001). However in the structural equation model, this relationship was significantly attenuated (p = 0.06) after accounting for the causal effect of diabetes (p < 0.0001), hematoma volume (p < 0.0001), intraventricular extension (p = 0.01) and Glasgow coma scale (p = 0.001) on glucose levels. On secondary analyses stratifying by diagnosis of diabetes, higher glucose levels were negatively correlated with favorable outcome in patients without diabetes (p = 0.04), but not in patients with diabetes (p = 0.35).
CONCLUSIONS: Hyperglycemia may be a downstream effect of other markers of ICH severity, particularly among patients without diabetes, suggesting a possible explanation for the limited evidence of glucose lowering interventions in outcome.
METHODS: Consecutive primary ICH patients admitted to an academic hospital between 2007 and 2018 were identified. Patients with missing baseline or follow up CT scans and without 90 day follow up status were excluded. We constructed a causal model accounting for pre-defined markers of ICH severity to evaluate the association between mean 72 h glucose and 90 day functional outcome measured by modified Rankin Scale, dichotomized as favorable ≤2 or unfavorable >2.
RESULTS: Primary analyses included 410 patients (70.4 ± 13.8years, 43 % female). Mean 72 h glucose was 137.5 ± 33.4mg/dl and 102 (25 %) patients were diabetic. On univariable analysis, higher glucose levels were negatively correlated with favorable outcome (p < 0.0001). However in the structural equation model, this relationship was significantly attenuated (p = 0.06) after accounting for the causal effect of diabetes (p < 0.0001), hematoma volume (p < 0.0001), intraventricular extension (p = 0.01) and Glasgow coma scale (p = 0.001) on glucose levels. On secondary analyses stratifying by diagnosis of diabetes, higher glucose levels were negatively correlated with favorable outcome in patients without diabetes (p = 0.04), but not in patients with diabetes (p = 0.35).
CONCLUSIONS: Hyperglycemia may be a downstream effect of other markers of ICH severity, particularly among patients without diabetes, suggesting a possible explanation for the limited evidence of glucose lowering interventions in outcome.
摘要：
背景：脑出血（ICH）急性期的高血糖与不良的功能结局有关，然而,所有降低血糖的干预措施均产生中性或阴性结果.我们尝试使用广义结构方程模型来解释高血糖在ICH结果中的因果作用。
方法：确定了2007年至2018年期间在一家学术医院收治的连续原发性ICH患者。排除缺少基线或随访CT扫描且无90天随访状态的患者。我们构建了一个考虑ICH严重程度的预定义指标的因果模型，以评估平均72小时血糖与通过改良Rankin量表测量的90天功能结果之间的关联。二分为有利的≤2或不利的>2。
结果：主要分析包括410例患者（70.4±13.8岁，43%女性）。平均72小时血糖为137.5±33.4mg/dl，102例（25％）患者患有糖尿病。关于单变量分析，较高的血糖水平与良好的结局呈负相关(p<0.0001).然而，在结构方程模型中，在考虑了糖尿病的因果效应（p<0.0001）后，这种关系显着减弱（p=0.06），血肿体积（p<0.0001），葡萄糖水平的脑室内扩展（p=0.01）和格拉斯哥昏迷量表（p=0.001）。在通过诊断糖尿病进行分层的二级分析中，在非糖尿病患者中，较高的血糖水平与良好的预后呈负相关（p=0.04），但不是糖尿病患者（p=0.35）。
结论：高血糖可能是其他ICH严重程度标志物的下游效应，特别是在没有糖尿病的患者中，为降糖干预结局的有限证据提供可能的解释。
方法：确定了2007年至2018年期间在一家学术医院收治的连续原发性ICH患者。排除缺少基线或随访CT扫描且无90天随访状态的患者。我们构建了一个考虑ICH严重程度的预定义指标的因果模型，以评估平均72小时血糖与通过改良Rankin量表测量的90天功能结果之间的关联。二分为有利的≤2或不利的>2。
结果：主要分析包括410例患者（70.4±13.8岁，43%女性）。平均72小时血糖为137.5±33.4mg/dl，102例（25％）患者患有糖尿病。关于单变量分析，较高的血糖水平与良好的结局呈负相关(p<0.0001).然而，在结构方程模型中，在考虑了糖尿病的因果效应（p<0.0001）后，这种关系显着减弱（p=0.06），血肿体积（p<0.0001），葡萄糖水平的脑室内扩展（p=0.01）和格拉斯哥昏迷量表（p=0.001）。在通过诊断糖尿病进行分层的二级分析中，在非糖尿病患者中，较高的血糖水平与良好的预后呈负相关（p=0.04），但不是糖尿病患者（p=0.35）。
结论：高血糖可能是其他ICH严重程度标志物的下游效应，特别是在没有糖尿病的患者中，为降糖干预结局的有限证据提供可能的解释。
