心脏再同步化治疗患者临床预后风险评分系统的构建与验证

摘要

Objective: To create and validate a scoring system for predicting clinical prognosis in patients with cardiac resynchronization therapy (CRT). Methods: A cohort of 367 consecutive patients received CRT in our hospital from 2010-01 to 2015-12 were enrolled. The endpoint follow-up events were all-cause death including heart transplantation and heart failure re-admission. The patients were randomly categorized into 2 groups: Modeling group, to develop HEAL scoring system,n=300 and Veriifcation group, to validate HEAL model,n=67. HEAL system was established by Cox proportional hazards regression model, discrimination between HEAL and EARRN scoring systems was evaluated by AUC of ROC, HEAL calibration was assessed by Hosmer-Lemeshow test and clinical endpoint evaluation by 2 scoring systems were compared by Kaplan-Meier method. Results: Modeling group analysis indicated that hs-CRP (HR=1.137, 95% CI 1.072-1.205,P<0.001), big endothelin-1 (HR=1.934, 95% CI 1.066-3.507,P=0.03), left atrial diameter (HR=1.045, 95% CI 1.007-1.084,P=0.02) and NYHA IV (HR=2.583, 95% CI 1.331-5.013,P=0.005) were the independent risk factors of adverse prognosis in CRT patients. Based on β partial regression coefifcient, HEAL scoring system was established to classify the patient's risk levels: low risk<4, moderate risk 4-10 and high risk>10. AUC for risk classification in Modeling group and Verification group were 0.719(95% CI 0.629-0.809) and 0.708 (95% CI 0.539-0.878), HEAL can well distinguish clinical prognosis in patients at different risk levels (log-rank test showed in Modeling groupP<0.001 and in Veriifcation groupP=0.002); Hosmer-Lemeshow test presented good calibration,P=0.952. All 367 patients were respectively evaluated by HEAL and EARRN scoring systems, HEAL had the better discrimination than EARRN as AUC 0.763 (95% CI 0.692-0.833) vs AUC 0.602 (95% CI 0.517-0.687). Conclusion: HEAL scoring system can effectively predict adverse prognosis in CRT patients, it had the better discrimination than EARRN system and was valuable to distinguish high risk patients in clinical practice.%目的:创建并验证心脏再同步化治疗(CRT)患者临床预后的风险评分系统.方法:连续纳入2010-01至2015-12于我院首次接受CRT的患者367例.随访终点事件为全因死亡(包括心脏移植)和心力衰竭再住院.随机选取300例患者为建模组构建风险评分系统,其余67例为验证组进行验证.采用Cox风险比例回归模型建立评分系统;受试者工作特征(ROC)曲线下面积(AUC)评价并对比HEAL评分与EARRN评分的区分度;Hosmer-Lemeshow法评价拟合优度;Kaplan-Meier法比较不同评分患者的临床终点.结果:建模组分析显示,高敏C反应蛋白(HR=1.137,95%CI:1.072~1.205,P<0.001)、大内皮素-1(HR=1.934,95%CI:1.066~3.507,P=0.03)、左心房前后径(HR=1.045,95%CI:1.007~1.084,P=0.02)、纽约心脏协会(NYHA)心功能Ⅳ级(HR=2.583,95%CI:1.331~5.013,P=0.005)是CRT患者不良预后的独立危险因素.依据危险因素β偏回归系数建立HEAL评分,根据分值划定患者危险分级:<4分为低危,4~10分为中危,>10分为高危.低、中、高危分级在建模组和验证组的ROC曲线下面积分别为0.719(95%CI:0.629~0.809)和0.708(95%CI:0.539~0.878),该评分可良好地区分不同危险分级患者的临床预后(Log-rank检验,建模组P<0.001,验证组P=0.002).Hosmer-Lemeshow拟合优度较好(P=0.952).对367例患者分别采用HEAL评分与EARRN评分,发现HEAL评分(AUC:0.763,95%CI:0.692~0.833)较EARRN评分(AUC:0.602,95%CI:0.517~0.687)区分度更高.结论:HEAL评分能有效地预测CRT患者不良预后,区分度优于EARRN评分,对于识别高危患者具有临床实践价值.