颅内动脉瘤血管内栓塞术后复发风险预测评分模型的建立与验证

Establishment and validation of a scoring model for predicting the recurrence risk after endovascular embolization of intracranial aneurysms

目的建立一个综合、简易、有效的复发风险预测评分模型以评估颅内动脉瘤血管内栓塞术后复发可能性大小,为手术方案的选择及术后处理提供指导意义。方法回顾性纳入长海医院神经外科2012年5月至2014年5月接受血管内栓塞治疗的颅内动脉瘤患者434例(共441个动脉瘤)并作为建模组,模型建立后,前瞻性纳入2015年1月至6月接受血管内栓塞治疗的颅内动脉瘤患者109例(共109个动脉瘤)并作为验证组。在建模队列中,依据前期动脉瘤复发危险因素的多因素Logistic回归分析结果建立复发风险预测评分模型;在验证队列中对模型进行验证。根据建模组评分模型受试者工作特征(ROC)曲线最佳cut-off值将评分表分为复发低危和复发高危。将复发风险预测评分模型与北美复发风险分层评分(ARSS)模型和Raymond分级进行比较。结果多因素Logistic回归分析显示,纳入评分并最终建立复发风险预测评分模型的3个因素为非支架辅助栓塞(1分)、Raymond分级≥Ⅱ级(1分)及动脉瘤大小[动脉瘤>25 mm(3分),动脉瘤10~25 mm(1分),动脉瘤<10 mm(0分)]。验证提示该评分体系具有较高的预测价值(AUC=0.738,95%CI:0.641~0.834,P<0.05)和拟合优度(Hosmer-Lemeshowχ2=2.109,P=0.146);将评分模型进一步分为复发低危(0~1分)和复发高危(2~5分),其敏感度为72.73%(48/66),特异度为68.80%(258/375)。动脉瘤复发风险预测评分模型的预测能力与ARSS评分相似(χ2=0.54,P=0.462),并且优于Raymond分级(χ2=15.10,P<0.01)。结论该研究所构建的简易动脉瘤复发风险预测评分模型可准确预测动脉瘤复发,但尚需开展多中心大样本的前瞻性研究以进一步验证。

Objective To establish a comprehensive, simple, and effective scoring model for predicting the recurrence risk after endovascular embolization of intraeranial aneurysms in order to assess the possibility of recurrence and to provide guidance for the selection of surgical protocols and postoperative management. Methods From May 2012 to May 2014,434 patients （441 aneurysms） with intracranial aneurysm treated with endovascular embolization at the Department of Neurosurgery, Changhai Hospital, the Second Military Medical University were enrolled retrospectively, and they were used as a modeling group. After modeling, 109 patients （109 aneurysms） were used as a validation group. In the modeling cohort, a predictive scoring model of recurrence risk was established according to the results of multivariate logistic regression analysis ; the model was validated in the validation cohort. According to the scoring model of the modeling group ,the scoring table of best cut-off value of the receiver operating characteristic （ROC） curves was divided into a low-risk and a high-risk of recurrence. The recurrence risk score model was compared with the North America aneurysm reeanalization stratification scale （ARSS） model,and Raymond grade. Results Multivariate logistic regression analysis showed that the 3 factors included in the scores and finally, a established scoring model of recurrence risk prediction were non-stent assisted embolization （ 1 point ） , Raymond grade t〉 11 （ 1 point）,and the size of aneurysm （ aneurysm 〉 25 mm [ 3 points） ], aneurysm 10 -25 mm [ 1 point ], and aneurysm 〈 10 mm[0 point] ）. The validation indicated that the scoring system had higher predictive value （ AUC = 0.738,95% CI O. 641 -0.834 ,P 〈 0.05 ） and goodness of fit （ Hosmer-Lemeshow X^2 = 2. 109 ,P = 0. 146）. The scoring table was further divided into the low-risk recurrence （0-1 point） and high-risk re- currence （2 -5 points）, its sensitivity was 72.73% （48/66） and specificity was 68.80% （258/375）. The predictive ability of the aneurysm recurrence risk score model was similar to that of the ARSS score （X^2 =0.54,P=0. 462）,and it was better than the Raymond grade （X^2 = 15. 10,P 〈0.01）. Conclusion The established simple aneurysm recurrence risk predicting score model in this study may accurately predict the recurrence of aneurysms, however, a multicenter, large sample prospective study is needed for further validation.