Germanova Olga, Galati Giuseppe, Germanov Andrey, Shchukin Yurii, Syunyakov Timur, Biondi-Zoccai Giuseppe
International Center for Education and Research in Cardiovascular Pathology and Cardiovisualization, Samara State Medical University, Samara, Russia -
International Center for Education and Research in Cardiovascular Pathology and Cardiovisualization, Samara State Medical University, Samara, Russia.
Minerva Cardiol Angiol. 2025 Feb;73(1):86-94. doi: 10.23736/S2724-5683.24.06549-9. Epub 2024 Oct 8.
The aim of this study was the creation of an optimal model for predicting arterial vascular complications in patients with extrasystolic arrhythmia.
A single-center prospective study was performed with involving 634 patients with supraventricular or ventricular extrasystoles (ES) of 700 or more per 24 hours. The control group consisted of 106 people with ES less than 700 per 24 hours. The main and control groups were initially equivalent in anthropometric criteria and concomitant pathology. The list of examinations included laboratory methods (including lipid profile, coagulograms), as well as instrumental studies (transthoracic and/or transesophageal echocardiography (EchoCG), Doppler ultrasound of the brachiocephalic arteries and arteries of the lower extremities, 24-hours ECG monitoring, according to the indications - computed tomography or magnetic resonance imaging of the brain, coronary angiography, stress echocardiography. Prospective observation of patients performed for 1 year after the initial examination. Combined end points: development of arterial vascular complications - stroke, myocardial infarction, distal arterial embolism of other locations. We studied the data on identified complications. Next, we built models for predicting complications in various ways: Decision Tree; Bootstrap Forest; Boosted Tree; Neural Boosted; Support Vector Machines; Fit Stepwise; Nominal Logistic; Generalized Regression Lasso; Generalized Regression Forward Selection; Generalized Regression Pruned Forward Selection; Generalized Regression Elastic Net; Generalized Regression Ridge. To assess the quality of the models and compare them we used cross-validation with 30 replications.
The highest profit values with minimal values of false positive results were obtained for the Bootstrap Forest model. Basing on this model, we created arterial vascular complications predictive score in extrasystolic arrhythmia "EX-prognosis" that included the following parameters: atheroma type III in carotid arteries - 3 points; age 69+ years old - 2 points; ES appearing before transmitral blood flow peak in cardiac cycle 700 and more per 24 hours - 1 point; carotid arteries stenosis, non-significant - 1 point. If total number is 3 and more points, the risk of arterial vascular complications within 1 year is high.
We recommend to use the scale "EX-prognosis" in the clinical practice. For a quick assessment of the total risk, it is optimal to implement the risk14.exe program - calculator - developed by us for a personal computer, based on this scale.
本研究的目的是创建一个预测早搏性心律失常患者动脉血管并发症的最佳模型。
进行了一项单中心前瞻性研究，纳入634例每24小时室上性或室性早搏（ES）达700次及以上的患者。对照组由106例每24小时ES少于700次的患者组成。主要组和对照组在人体测量标准和伴随疾病方面最初是相当的。检查项目包括实验室方法（包括血脂谱、凝血图）以及仪器检查（经胸和/或经食管超声心动图（EchoCG）、头臂动脉和下肢动脉的多普勒超声、24小时心电图监测，根据指征进行脑部计算机断层扫描或磁共振成像、冠状动脉造影、负荷超声心动图）。在初次检查后对患者进行为期1年的前瞻性观察。联合终点：动脉血管并发症的发生——中风、心肌梗死、其他部位的远端动脉栓塞。我们研究了已确定并发症的数据。接下来，我们以各种方式构建并发症预测模型：决策树；自助森林；增强树；神经增强；支持向量机；逐步拟合；名义逻辑回归；广义回归套索；广义回归向前选择；广义回归修剪向前选择；广义回归弹性网；广义回归岭回归。为了评估模型的质量并进行比较，我们使用了30次重复的交叉验证。
自助森林模型获得了最高的收益值和最低的假阳性结果值。基于该模型，我们创建了早搏性心律失常的动脉血管并发症预测评分“EX-预后”，其包括以下参数：颈动脉III型动脉粥样硬化——3分；年龄69岁及以上——2分；每24小时700次及以上早搏出现在心动周期二尖瓣血流峰值之前——1分；颈动脉非显著性狭窄——1分。如果总分达到3分及以上，则1年内发生动脉血管并发症的风险较高。
我们建议在临床实践中使用“EX-预后”量表。为了快速评估总体风险，基于该量表，最佳做法是在个人计算机上运行我们开发的risk14.exe程序——计算器。
