摘要
目的:探讨射血分数轻度降低的心力衰竭(HFmrEF)全因死亡的影响因素,通过长期随访数据构建列线图模型,用于预测患者1年、2年、3年期的全因死亡率.方法:顺序入选2014年4月 2019年3月山西省3所三级甲等医院诊治的慢性心力衰竭且射血分数为41％～49％的患者1 148例进行随访,以全因死亡为终点事件,随访截止日期为2022年4月1日.将所有研究对象按7∶3比例随机分为训练集和验证集,训练集用于模型的构建,验证集用于模型性能的评估.利用Cox回归分析患者全因死亡的影响因素,R4.3.1用于构建列线图预测模型.采用一致性指数(C-index)、受试者工作特征曲线下面积(AUC)和校正曲线评价模型的区分度和预测性能,通过临床决策曲线(DCA)评估模型的临床潜在应用价值.根据ROC曲线确定的最佳截断阈值对患者进行死亡风险分层,并采用Kaplan-Meier曲线比较高风险组和低风险组患者之间的生存差异.结果:多因素Cox回归分析显示,住院费用全自付(HR=4.722,95％CI:2.544～8.765)、入院NYHA心功能Ⅳ级(HR=2.982,95％CI:1.507～5.898)、N 末端脑钠肽前体(对数值)[lg(NT-proBNP)]水平升高(HR=2.360,95％CI:1.414～3.938)、合并心房颤动(HR=2.321,95％CI:1.419～3.797)增加患者全因死亡风险;高估算肾小球滤过率(eG-FR)水平(HR=0.984,95％CI:0.973～0.995)、服用血管紧张素转换酶抑制剂/血管紧张素Ⅱ受体拮抗剂(ACEI/ARB)药物(HR=0.320,95％CI:0.191～0.535)及接受经皮冠状动脉介入治疗/冠状动脉旁路移植术(PCI/CABG)治疗(HR=0.503,95％CI:0.264～0.958)降低患者全因死亡风险.依此构建的列线图模型经测试集验证,C-index为0.839,1年、2年、3年生存期预测模型的AUC分别为0.864、0.860、0.857.校正曲线和DCA曲线结果显示,模型预测效果和实际生存情况拟合度较好,具有较好的临床适用性.风险分层能够有效区分高、低危患者的预后.结论:基于医保类型、NYHA心功能分级、NT-proBNP、eGFR、心房颤动、ACEI/ARB类药物及PCI/CABG治疗7个因素构建的列线图预测模型有助于高风险HFmrEF患者早期识别与治疗决策指导.
Abstract
Objective:To explore the effect factors of the death of heart failure with mildly reduced ejection fraction(HFmrEF),and to construct a nomogram model based on long-term follow-up data to predict the all-cause mortality of patients at 1 year,2 years and 3 years.Methods:A total of 1148 patients with chronic heart failure and ejection fraction of 41％-49％who were diagnosed and treated in three tertiary-level A hospitals in Shanxi Province from April 2014 to March 2019 were sequentially enrolled for follow-up,with all-cause death as the end-point event,and the deadline for follow-up was April 1,2022.All study subjects were randomly divided into a training set and a validation set in a 7∶3 ratio;the training set was used for model construction,and the validation set was used for model performance evaluation.Cox regression was used to analyze the factors influencing all-cause mortality in patients,and R 4.3.1 was used to construct the column-line graph prediction model.The con-sistency index(C-index),the area under the working curve(AUC)of subjects,and calibration curves were used to evaluate the discriminatory and predictive performance of the model,and the clinical potential application value of the model was assessed by the clinical decision curve(DCA).Based on the optimal cut-off threshold determined by the ROC curve,the patients were stratified by death risk,and the Kaplan-Meier curves were used to compare the survival differences between the patients in the high-risk and low-risk groups.Results:Multifactorial Cox regres-sion analysis showed that self-paying(HR=4.722,95％CI:2.544 to 8.765),admission the New York Heart Association class Ⅳ(HR=2.982,95％CI:1.507 to 5.898),elevated of N-terminal brain natriuretic peptide pre-cursor(Logarithmic value)[lg(NT-proBNP)]levels(HR=2.360 with 95％CI:1.414 to 3.938),and combined atrial fibrillation(HR=2.321,95％CI:1.419 to 3.797)increased the risk of all-cause mortality in patients;high estimated glomerular filtration rate(eGFR)levels(HR=0.984,95％CI:0.973 to 0.995),and the use of angio-tensin-converting enzyme inhibitors or angiotensin receptor antagonist(ACEI/ARB)medications(HR=0.320,95％CI:0.191～0.535)and receiving percutaneous coronary intervention or coronary artery bypass grafting(PCI/CABG)treatments(HR=0.503,95％CI:0.264～0.958)reduced the risk of all-cause mortality in patients.The C-index of the column-line graph model constructed accordingly was 0.839,and the AUCs of the 1-,2-,and 3-year survival prediction models were 0.864,0.860,and 0.857,respectively,and the calibration curves suggested that the model prediction effect was basically in line with the actual survival situation,and the DC A curves indica-ted that the model had good clinical applicability.Risk stratification can effectively distinguish the prognosis of high-and low-risk patients.Conclusion:Based on these 7 factors(medical insurance type,NYHA classification,NT-proBNP,eGFR,atrial fibrillation,ACEI/ARB drugs,and PCI/CABG treatment),the nomogram is helpful for the early identification and treatment decision-making guidance of high-risk HFmrEF patients.
NETL
NSTL
万方数据