Objective: To investigate the predictive value of echocardiographic quantitative parameters combined with serum creatine kinase(CK) and creatine kinase-myocardial band(CK-MB) for early weaning failure from extracorporeal membrane oxygenation(ECMO) in patients with acute myocardial infarction(AMI), based on machine learning(ML) models. Methods: A total of 245 AMI patients(training set) admitted to our hospital from December 2022 to May 2024 were retrospectively enrolled and divided into a weaning failure group(n=78) and a successful weaning group(n=167) according to their early ECMO weaning outcomes. Additionally, 105 AMI patients admitted from June 2024 to March 2025 were selected as a validation set, with a training-to-validation ratio of 7∶3. General clinical data and laboratory findings were collected. Independent predictors of early weaning failure were identified using univariate analysis, least absolute shrinkage and selection operator(LASSO) regression, and multivariate Logistic regression. Logistic regression(LR), back-propagation neural network(BPNN), and random forest(RF) models were constructed. The predictive performance, calibration, and clinical net benefit of the combined model and the clinical-indicators-only model for early weaning failure were evaluated using the area under the receiver operating characteristic curve(AUC), confusion matrix metrics(accuracy, precision, recall, F1 score), calibration curves, and decision curves, followed by external validation in the validation set. Results: The early ECMO weaning failure rates in the training and validation sets were 31.84% and 31.43%, respectively. Age, duration of ECMO support, multivessel disease, left ventricular ejection fraction(LVEF), high-sensitivity troponin Ⅰ and CK-MB were identified as independent predictors of early weaning failure from ECMO in AMI patients(all P<0.05). The RF-based combined model and the clinical-indicators-only model showed superior AUC, confusion matrix metrics, calibration and clinical net benefit compared with the other models in both the training and validation sets. Specifically, the RF combined model achieved AUCs of 0.915(95%CI:0.879-0.952) in the training set and 0.828(95%CI:0.749-0.908) in the validation set, along with better confusion matrix metrics than the clinical-indicators-only model(training set AUC:0.838, 95%CI:0.787-0.888; validation set AUC:0.741, 95%CI:0.644-0.837). The independent predictors in the RF combined model, ranked by importance from highest to lowest, were CK-MB>high-sensitivity troponin Ⅰ>LVEF>duration of ECMO support>age>multivessel disease. Conclusion: The RF model demonstrates superior performance in predicting early ECMO weaning failure risk in AMI patients. Moreover, the combined model incorporating LVEF and serum CK-MB has better predictive performance than the clinical-indicators-only model. |
[1] 闫小菊,李萌,刘帅,等.急性心肌梗死患者基质γ-羧基谷氨酸蛋白对经皮冠状动脉介入治疗术后心血管事件的预测[J].中华老年心脑血管病杂志,2025,27(8):986-990.
[2] EZAD S M,RYAN M,DONKER D W,et al.Unloading the left ventricle in venoarterial ECMO:in whom,when,and how?[J].Circulation,2023,147(16):1237-1250.
[3] RAJSIC S,TREML B,JADZIC D,et al.Extracorporeal membrane oxygenation for cardiogenic shock:a meta-analysis of mortality and complications[J].Ann Intensive Care,2022,12(1):93.
[4] WANG H,CEN Z,LIU X,et al.Use of extracorporeal membrane oxygenation in adult trauma patients with refractory acute cardiopulmonary failure:guideline from the Chinese society of extracorporeal life support 2025[J].Crit Care,2025,29(1):334.
[5] CHARBONNEAU F,CHAHINIAN K,BEBAWI E,et al.Parameters associated with successful weaning of veno-arterial extracorporeal membrane oxygenation:a systematic review[J].Crit Care,2022,26(1):375.
[6] 郭建红,喻红霞,李丹.超声心动图联合H-FABP、cTnI及CK-MB在AMI诊断、病情评估中的应用[J].分子诊断与治疗杂志,2022,14(9):1476-1479,1483.
[7] 刘洪岩,王方明,亓华新,等.心肌酶谱、补体C1q及糖化血红蛋白变异指数与急性心肌梗死患者经皮冠状动脉介入治疗术后心力衰竭的关系[J].中国临床医生杂志,2025,53(8):941-944.
[8] 张燕茹,李志娟,罗世宁.机器学习在心肌梗死诊疗中的应用进展[J/OL].中国循证心血管医学杂志,2025,(7):887-890[2025-08-19].https://link.cnki.net/urlid/11.5719.R.20250815.1038.054.
[9] RILEY R D,ENSOR J,SNELL K I E,et al.Calculating the sample size required for developing a clinical prediction model[J].BMJ,2020,368:m441.
[10] 张洁,席寅,王亚,等.《成人体外膜肺氧合技术操作规范(2024年版)》解读:ECMO的撤离[J].中华重症医学电子杂志,2025,11(1):60-64.
[11] 陈春红.急性心肌梗死的诊断与治疗[M].石家庄:河北科学技术出版社,2011:1021-1028.
[12] LORUSSO R,SHEKAR K,MACLAREN G,et al.ELSO interim guidelines for venoarterial extracorporeal membrane oxygenation in adult cardiac patients[J].ASAIO J,2021,67(8):827-844.
[13] 赵欣然,徐以康,吴桂平,等.急性心肌梗死冠脉多支病变的危险因素及风险预测模型研究进展[J].医药前沿,2025,15(5):22-25.
[14] ZAKIEV V,VOROBYEVA N,MALAYA I,et al.The association of myocardial infarction history and geriatric syndromes in the elderly:data from the cross-sectional study EVKALIPT[J].J Clin Med,2024,13(21):6420.
[15] ABDULRAHMAN M,MAKKI M,BENTALEB M,et al.Current role of extracorporeal membrane oxygenation for the management of trauma patients:indications and results[J].World J Crit Care Med,2025,14(1):96694.
[16] SHIN J,KANG E M,LEE S H,et al.Prognostic implication of venoarterial extracorporeal membrane oxygenation in acute myocardial infarction-related cardiogenic shock[J].J Intensive Care,2025,13(1):38.
[17] TAVAZZI G,COLOMBO C N J,KLERSY C,et al.Echocardiographic parameters for weaning from extracorporeal membrane oxygenation-the role of longitudinal function and cardiac time intervals[J].Eur Heart J Cardiovasc Imaging,2025,26(2):359-367.
[18] KUMAR S,GRIFFITH N,WALTER D,et al.Characterization of myocardial injury with high-sensitivity troponin[J].Tex Heart Inst J,2023,50(6):e238108.
[19] GOYAL A.CK-MB kinetics as a superior prognostic indicator in primary PCI:moving beyond TIMI flow grade 3[J].JACC Asia,2025,5(5):660-662.
[20] SOLEIMANI H,NAJDAGHI S,DAVANI D N,et al.Predicting in-hospital mortality in patients with acute myocardial infarction:a comparison of machine learning approaches[J].Clin Cardiol,2025,48(4):e70124.
[21] YANG Y,TANG J,MA L,et al.A systematic comparison of short-term and long-term mortality prediction in acute myocardial infarction using machine learning models[J].BMC Med Inform Decis Mak,2025,25(1):208.
[22] WANG N,WANG M,ZHOU Y,et al.Sequential data-based patient similarity framework for patient outcome prediction:algorithm development[J].J Med Internet Res,2022,24(1):e30720.
[23] SUZUKI S,KIMURA K,UEHARA T,et al.Validation of predictors of successful weaning from mechanical circulatory support and histological features in lymphocytic fulminant myocarditis[J].Open Heart,2025,12(1):e003224.
[24] SAELMANS A,SEINEN T,PERA V,et al.Implementation and updating of clinical prediction models:a systematic review[J].Mayo Clin Proc Digit Health,2025,3(3):100228.
[25] TEODORO D,NADERI N,YAZDANI A,et al.A scoping review of artificial intelligence applications in clinical trial risk assessment[J].npj Digit Med,2025,8:486.
[26] SHANG H,WANG J,CONG Y,et al.Risk prediction models for heart failure after myocardial infarction:a systematic review[J].Medicine,2025,104(30):e43480. |
