Hoh Brian L, Sistrom Christopher L, Firment Christopher S, Fautheree Gregory L, Velat Gregory J, Whiting Jobyna H, Reavey-Cantwell John F, Lewis Stephen B
Department of Neurological Surgery, University of Florida College of Medicine, Gainesville, Florida 32610-0265, USA.
Neurosurgery. 2007 Oct;61(4):716-22; discussion 722-3. doi: 10.1227/01.NEU.0000298899.77097.BF.
Determining factors predictive of the natural risk of rupture of cerebral aneurysms is difficult because of the need to control for confounding variables. We studied factors associated with rupture in a study model of patients with multiple cerebral aneurysms, one aneurysm that had ruptured and one or more that had not, in which each patient served as their own internal control.
We collected aneurysm location, one-dimensional measurements, and two-dimensional indices from the computed tomographic angiograms of patients in the proposed study model and compared ruptured versus unruptured aneurysms. Bivariate statistics were supplemented with multivariable logistic regression analysis to model ruptured status. A total of 40 candidate models were evaluated for predictive power and fit with Wald scoring, Cox and Snell R2, Hosmer and Lemeshow tests, case classification counting, and residual analysis to determine which of the computed tomographic angiographic measurements or indices were jointly associated with and predictive of aneurysm rupture.
Thirty patients with 67 aneurysms (30 ruptured, 37 unruptured) were studied. Maximum diameter, height, maximum width, bulge height, parent artery diameter, aspect ratio, bottleneck factor, and aneurysm/parent artery ratio were significantly (P < 0.05) associated with ruptured aneurysms on bivariate analysis. When best subsets and stepwise multivariable logistic regression was performed, bottleneck factor (odds ratio = 1.25, confidence interval = 1.11-1.41 for every 0.1 increase) and height-width ratio (odds ratio = 1.23, confidence interval = 1.03-1.47 for every 0.1 increase) were the only measures that were significantly predictive of rupture.
In a case-control study of patients with multiple cerebral aneurysms, increased bottleneck factor and height-width ratio were consistently associated with rupture.
由于需要控制混杂变量，确定预测脑动脉瘤自然破裂风险的因素具有一定难度。我们在一个研究模型中，对患有多个脑动脉瘤的患者进行了研究，该模型中有一个已破裂的动脉瘤和一个或多个未破裂的动脉瘤，每个患者自身作为内部对照。
我们从拟议研究模型中患者的计算机断层血管造影图像中收集动脉瘤位置、一维测量值和二维指标，并比较破裂与未破裂的动脉瘤。双变量统计辅以多变量逻辑回归分析以模拟破裂状态。总共评估了40个候选模型的预测能力，并通过Wald评分、Cox和Snell R2、Hosmer和Lemeshow检验、病例分类计数以及残差分析来确定哪些计算机断层血管造影测量值或指标与动脉瘤破裂共同相关并具有预测性。
对30例患者的67个动脉瘤（30个破裂，37个未破裂）进行了研究。在双变量分析中，最大直径、高度、最大宽度、凸起高度、母动脉直径、纵横比、瓶颈因子和动脉瘤/母动脉比与破裂的动脉瘤显著相关（P < 0.05）。当进行最佳子集和逐步多变量逻辑回归分析时，瓶颈因子（每增加0.1，优势比 = 1.25，置信区间 = 1.11 - 1.
