Percutaneous coronary intervention (PCI) is the main treatment for acute coronary syndrome (ACS), but some patients may experience recurrent major cardiovascular events (MACE) after the treatment. Recent studies have shown that lysophospholipid metabolites such as lysophosphatidylcholine (LPC) and lysophosphatidic acid (LPA), which are important components of oxidized low-density lipoprotein and low-density lipoprotein, can promote the formation and rupture of atherosclerotic plaques. However, it remains unclear whether lysophospholipid metabolites can be used to predict the occurrence of MACE following PCI in patients with ACS.

To investigate the predictive value of lysophospholipids for MACE following PCI in patients with ACS.

The study included patients with ACS who underwent PCI at Beijing Anzhen Hospital, Capital Medical University from June 2017 to September 2019. Baseline data of the patients were collected, and targeted metabolomics was performed to detect phospholipids and lysophospholipids. Patients were followed up at 1, 3, 6, 9, and 12 months post-enrollment, and then every 6 months thereafter, through outpatient visits and telephone consultations to record the occurrence of MACE. Principal component analysis (PCA) score plots were used to analyze the metabolic profiles and inter-group distributions of lysophospholipids between the non-MACE and MACE groups. A partial least squares-discriminant analysis (PLS-DA) with variable importance in projection (VIP) plots was utilized to assess the differential metabolites of lysophospholipids between the groups. The importance of each phospholipid and lysophospholipid metabolite was ranked using the random forest accuracy decrease diagram. Monte Carlo cross-validation was applied to construct the receiver operating characteristic (ROC) curve for the multivariable random forest models composed of different numbers of metabolites, and the area under the curve (AUC) was calculated to select key lysophospholipid metabolites associated with MACE. The accuracy of the predictive models was assessed using permutation tests.

Participants included 212 cases of patients, with an average follow-up of 3 years. Patients were divided into the MACE group (n=29) and the non-MACE group (n=183) based on whether MACE occurred during the follow-up period. There was no statistically significant difference in baseline data between the two groups (P>0.05). The PCA score plot revealed a distinct distribution of MACE and non-MACE group samples, indicating significant differences in the lysophospholipid metabolite profiles. Integrated analysis identified significant changes in LPA, oxidized lysophosphatidylcholine (LPC-O), LPE, LPI, and LPS. Monte Carlo cross-validation was used to construct a predictive model for MACE using the top 13 ranked lipid metabolites [mainly including 10 lysophospholipid metabolites (LPA 16: 0, LPA 18: 1, LPC-O 16: 0, LPC 16: 0, LPG 18: 2, LPC 18: 0, LPE 20: 3, LPE 22: 6, LPG 18: 1, LPS 20: 4) and 3 phospholipid metabolites (PA 16: 0-18: 0, PA 16: 0-20: 4, PA 16: 0-18: 1) ]. The ROC curve indicated that the model had an AUC of 0.934 (95%CI=0.793-0.998) for predicting MACE.

Abnormal expression of lysophospholipids metabolites in preoperative serum is closely correlated with the risk of recurrent MACE following PCI in patients with ACS. This exhibits significant predictive efficacy and clinical value for MACE following PCI in patients with ACS.

经皮冠状动脉介入治疗（PCI）是急性冠脉综合征（ACS）的主要治疗方法，但术后有部分患者再发主要心血管不良事件（MACE）。目前多项研究表明，作为氧化低密度脂蛋白和低密度脂蛋白的重要组成部分，溶血磷脂酰胆碱（LPC）、溶血磷脂酸（LPA）等溶血磷脂类脂质代谢物可促进动脉粥样硬化斑块形成和破裂，但溶血磷脂类脂质代谢物是否能用于ACS患者PCI治疗后再发MACE的预测尚不明确。

利用靶向代谢组学和机器学习模型，探究磷脂及溶血磷脂类代谢小分子在接受PCI治疗的ACS患者术后发生MACE的预测价值。

纳入2017年6月—2019年9月就诊于首都医科大学附属北京安贞医院接受PCI治疗的ACS患者作为研究对象，收集患者的基线资料，靶向代谢组学检测磷脂及溶血磷脂类脂质代谢物。自患者入组后1、3、6、9个月和12个月以及其后每6个月通过门诊及电话随访患者MACE发生情况。采用主成分分析（PCA）得分图对非MACE组和MACE组进行溶血磷脂代谢轮廓和组间分布分析，基于偏最小二乘法判别分析（PLS-DA）的差异权重贡献值（VIP）图谱，评估代谢组间差异溶血磷脂类代谢小分子。利用随机森林准确度下降值图对每种磷脂和溶血磷脂类代谢小分子的变量重要性进行排序，使用蒙特卡洛交叉验证对不同数量代谢物所构成的多变量随机森林模型绘制受试者工作特征（ROC）曲线风险预测模型并计算ROC曲线下面积（AUC），筛选与MACE事件相关的关键溶血磷脂类代谢物，并使用置换检验对其所构成的预测模型准确度进行评估。

共纳入患者212例，平均随访时间为3年。根据患者在随访期间是否发生MACE将其分为MACE组（n=29）与非MACE组（n=183），两组患者基线资料比较，差异无统计学意义（P>0.05）。PCA得分图结果发现MACE组与非MACE组患者样本在得分图中的分布位置明显不同，两组患者溶血磷脂类代谢产物谱存在显著差异。利用整合分析图进行差异代谢物筛选，结果发现LPA、氧化溶血磷脂酰胆碱（LPC-O）、溶血磷脂酰乙醇胺（LPE）、溶血磷脂酰肌醇（LPI）、溶血磷脂酰丝氨酸（LPS）变化显著。蒙特卡洛交叉验证对不同数量代谢物所构成的多变量随机森林模型纳入排序前13位的脂质代谢小分子[主要包含10个溶血磷脂类代谢小分子（LPA 16：0、LPA 18：1、LPC-O 16：0、LPC 16：0、LPG 18：2、LPC 18：0、LPE 20：3、LPE 22：6、LPG 18：1、LPS 20：4）和3种磷脂代谢小分子（PA 16：0-18：0、PA 16：0-20：4、PA 16：0-18：1）]构建MACE的预测模型，ROC曲线结果显示模型预测MACE的AUC=0.934（95%CI=0.793~0.998）。

接受PCI手术的ACS患者术前血清中多种溶血磷脂类代谢小分子的异常表达与再发MACE事件的风险密切相关，在预测ACS患者PCI术后MACE中具有显著的预警作用和临床应用价值。