第一部分 基于非靶向代谢组学的血管迷走性晕厥神经内分泌机制初探

背景：血管迷走性晕厥（vasovagal syncope，VVS）的病理生理机制复杂且尚未完全阐明，诸多研究表明，神经内分泌调节在VVS的发生中发挥重要作用，多种神经激素可能参与其中。然而，现有研究受限于传统检测技术所能检测物质的单一性，尚不清楚是否存在其他神经内分泌因子或相关通路参与VVS的发生。代谢组学是一种新兴的高通量技术，能够全面解析体内代谢物的动态变化，为深入探讨VVS发生过程中的多层次、多组织的应答反应提供了可能。

目的：本研究对VVS患者在晕厥发生前后的血浆样本进行非靶向代谢组学检测，旨在初步探索VVS发生过程中的系统性代谢改变及其相关通路，并为从整体水平探索VVS的病理生理机制提供新的视角与依据。

方法：本研究前瞻性入选了2023年2月至2024年2月期间，就诊于阜外医院心律失常二病区的VVS患者。所有患者均接受了直立倾斜试验（head-up tilt test，HUTT）检查，在HUTT静息仰卧位时采集了基线外周血样，HUTT阳性患者还采集了晕厥发作时的血样。采用超高效液相色谱-串联质谱技术，对参与者的血浆样本进行非靶向代谢组学检测。通过单变量分析以及主成分分析、偏最小二乘判别分析等多变量分析方法进行差异代谢物筛选，并进行代谢物鉴定及通路富集分析，明确其生物学功能及潜在意义。

结果：研究共纳入66例VVS患者，平均年龄为45.6 ± 17.0岁，其中女性占50%（33/66），平均晕厥次数为4次。HUTT阳性患者45例（68.2%），阴性患者21例（31.8%）。多变量分析结果表明，HUTT阳性及阴性患者基线、HUTT阳性患者晕厥发生前后，血浆代谢组谱均存在显著差异。在HUTT阳性与阴性患者的基线比较中，鉴定出62种差异性代谢物，其中28种上调、34种下调；此外，与基线相比，HUTT阳性患者在发生晕厥时有33种代谢物发生变化，其中32种上调、1种下调。进一步的通路富集分析发现，晕厥发生前后，共23条信号通路发生了显著变化，包括不饱和脂肪酸的生物合成、亚油酸的代谢、过氧化物酶体增殖物激活受体信号通路、醛固酮合成与分泌、脂肪酸的生物合成等通路。

结论：本研究发现，VVS患者在晕厥发作前后的血浆代谢组学特征具有显著差异；同时，HUTT阳性与阴性患者的代谢组学特征也呈现差异，提示VVS的发生可能伴随着全身性代谢改变。相关发现为深入理解VVS的病理生理机制及其发生过程中的神经内分泌变化提供了新的视角与依据。

第二部分 血清素合成与代谢通路在血管迷走性晕厥发病中的改变及意义

背景：神经内分泌因素可能密切参与血管迷走性晕厥（vasovagal syncope，VVS）的发生。血清素（5-hydroxytryptamine，即5-羟色胺）是一种重要的神经激素，广泛参与血流动力学及自主神经调节功能调控，但其在VVS发病中的作用尚不明确。

目的：本研究聚焦血清素合成与代谢通路，旨在探讨血清素合成与代谢相关关键物质在VVS发生中的变化及其潜在意义。

方法：本研究前瞻性纳入2023年2月至2024年2月期间，就诊于阜外医院心律失常二病区的VVS患者。所有患者均接受了直立倾斜试验（head-up tilt test，HUTT）检查，在HUTT静息仰卧位时采集基线外周血样，HUTT阳性患者还采集了晕厥发作时的血样，连续记录并分析了HUTT过程中血流动力学指标的变化。研究基于靶向代谢组学策略，通过液相色谱-串联质谱法定量检测了血浆及血小板中的血清素浓度，并测定血浆中血清素相关合成及代谢物质的水平，包括血清素主要合成原料色氨酸、血清素前体5-羟色氨酸（5-hydroxytryptophan，5-HTP）及其主要代谢物5-羟吲哚乙酸（5-hydroxyindoleacetic acid，5-HIAA）。比较HUTT阳性及阴性患者基线、及HUTT阳性患者晕厥前后这些物质的变化，并通过线性回归模型，进一步分析相关变化与晕厥临床特征及血流动力学指标之间的相关性。

结果：共66名VVS患者（年龄45.6 ± 17.0岁，33名女性）进行了HUTT检查， 45例为HUTT阳性，21例为HUTT阴性。两组患者基线的血浆血清素和血小板血清素浓度未见显著差异，但HUTT阳性患者的基线血浆5-HTP水平显著高于HUTT阴性患者（3.9 ± 0.5 vs. 3.3 ± 0.4 ng/mL，P < 0.001），而基线血浆5-HIAA水平则显著低于HUTT阴性患者（10.3 ± 2.7 vs. 12.5 ± 4.3 ng/mL，P = 0.040）。晕厥发生后，血清素和色氨酸水平未见显著变化。但相较基线，晕厥时血浆5-HTP的浓度显著下降（3.6 ± 0.5 vs. 3.9 ± 0.5 ng/mL，P < 0.001），而血浆5-HIAA的浓度则显著上升（12.4 ± 3.6 vs. 10.3 ± 2.7 ng/mL，P < 0.001）。亚组分析结果显示，自发性VVS患者在晕厥时血清素水平较基线显著升高（15.9 ± 10.8 ng/mL vs. 11.8 ± 6.5 ng/mL，P = 0.037），且血清素变化幅度越大，倾斜触发晕厥的时间越短（R² = 0.38，P = 0.015）。线性回归分析发现，VVS发作过程中，血清素前体物质5-HTP的变化与心率、血压等变化存在一定的相关性。

结论：外周循环中血清素水平在VVS的发生中保持稳定，而血清素的前体物质5-HTP和主要代谢物5-HIAA则发生了显著变化，并且这些变化与VVS发作期间的血流动力学变化存在一定相关性。提示血清素合成与代谢途径的改变可能参与了VVS的发生，进一步充实了血清素能系统参与VVS发生的相关证据，并为VVS的神经内分泌机制提供了新的视角。

第三部分 基于代谢组学的血管迷走性晕厥的生物标志物筛选及诊断模型的建立

背景：血管迷走性晕厥（vasovagal syncope，VVS）由于临床表现的多样性及复杂的发病机制，其诊断仍面临诸多挑战。尽管已有研究报道了一些潜在的VVS生物标志物，但相关结果尚未系统化，且目前尚无基于生物标志物的VVS诊断模型。近年，代谢组学技术的迅速进步，为疾病的早期诊断及生物标志物筛选提供了新的思路和方法。

目的：本研究旨在通过非靶向代谢组学技术，筛选与VVS相关的代谢标志物，探索其在VVS发生中的潜在生物学作用，并基于这些生物标志物构建VVS的诊断模型。

方法：本研究前瞻性入选了2022年9月至2023年9月期间，就诊于阜外医院心律失常二病区的59例反复发作VVS（近6月发作＞2次）、且直立倾斜试验为阳性的患者，并按照年龄、性别、体质指数匹配了25名无晕厥病史的健康人作为健康对照组（healthy control，HC）。通过超高效液相色谱-串联质谱技术，对所有参与者的血浆样本进行非靶向代谢组学检测。采用t检验、差异倍数分析、正交偏最小二乘判别分析等方法筛选VVS组与HC组间的差异代谢物。为进一步筛选出能够区分VVS与HC的最优生物标志物组合，应用有监督的机器学习方法对所有筛选出的差异代谢物进行分析，识别VVS潜在的生物标志物并构建用于区分VVS与HC组的预测模型。通过受试者工作特征（receiver operating characteristic，ROC）曲线及其对应的曲线下面积（area under the curve，AUC）评估模型性能。

结果：59例VVS患者的平均年龄为43.8 ± 16.2岁，其中女性30例（占50.8%）；HC组平均年龄43.1 ± 12岁，女性14例（占56.0%）。VVS 组和 HC 组的代谢组学特征存在显著差异，共鉴定出 289 个上调、39 个下调的差异代谢物。通路富集分析结果表明，这些差异代谢物主要富集在苯丙氨酸代谢、咖啡因代谢以及缬氨酸、亮氨酸和异亮氨酸的生物合成通路。通过随机森林算法成功筛选出 7种代谢物作为VVS的诊断标志物，包括磷脂酰肌醇（14:1/14:1）、磷脂酰肌醇（18:1/18:1）、磷脂酰肌醇（16:0/16:0）、溶血磷脂酰胆碱（20:4）、溶血磷脂酰胆碱（22:5）、溶血磷脂酰胆碱（22:6）和亚麻油酸。基于这7种代谢物，分别使用三种机器学习算法构建VVS诊断模型：随机森林模型（AUC 0.933；95% CI 0.836-1），支持向量机模型（AUC 0.958；95%CI 0.886-1），线性判别分析模型（AUC 0.916；95% CI 0.791-1），均显示出对VVS良好的预测性能。

结论：本研究通过非靶向代谢组学分析发现，VVS患者与健康个体之间在代谢谱上存在显著差异。此外，基于机器学习筛选的7种生物标志物及其组合构建的诊断模型，显示出对VVS较高的预测价值，有助于为VVS的诊断提供新的策略，并加深对其发病机制的理解。

第四部分  基于简易临床指标的心源性晕厥诊断模型的开发和验证

背景：心源性晕厥相较于非心源性晕厥患者，具有更高的死亡率或其他不良事件风险。因此，晕厥管理的关键环节之一是早期识别这类患者。建立精准的心源性晕厥早期识别体系，具有重要临床价值。

目的：本研究旨在开发一种基于简易临床指标的心源性晕厥诊断模型，通过分析一组易于获取的临床参数，预测诊断为心源性晕厥的概率，从而提升对该疾病的早期识别能力。

方法：本研究回顾性分析了2021年1月至2022年6月期间，于阜外医院住院诊治的晕厥患者。收集其临床资料，将所有病因明确的心源性和非心源性晕厥（包括反射性晕厥或直立性低血压性晕厥）患者纳入分析，并随机将患者按7：3比例分成训练集和验证集。使用最小绝对收缩和选择算子（least absolute shrinkage and selection operator，LASSO）回归对初始纳入的36个临床变量进行特征筛选，随后使用多因素Logistic回归分析确定心源性晕厥的独立预测因子，并基于这些特征构建区分心源性晕厥和非心源性晕厥的诊断模型。在训练集及验证集中，分别通过受试者工作特征（receiver operating characteristic，ROC）分析、校准曲线分析及决策曲线分析等评估模型性能。

结果：本研究共纳入919例晕厥患者，其中877例（年龄57.2 ± 15.0岁，女性358例）病因明确的晕厥患者被纳入诊断模型的开发，包含心源性晕厥患者522例和非心源性晕厥患者355例。在36个候选临床变量中，采用LASSO回归与Logistic回归联合筛选出5个心源性晕厥的独立预测因子，包括体质指数（OR 1.088；95%CI 1.022-1.158；P = 0.008）、胸部前驱症状（OR 5.251；95%CI 3.326-8.288；P < 0.001）、N末端B型脑钠肽前体的对数值（Log NT-proBNP）（OR 1.463；95%CI 1.240-1.727；P < 0.001）、左室射血分数（OR 0.940；95%CI 0.908-0.973；P < 0.001）以及心电图异常（OR 6.171；95%CI 3.966-9.600；P < 0.001）。基于这5个变量构建的诊断模型，在训练集和验证集中，ROC曲线下面积（AUC）分别为0.873（95% CI 0.845-0.902）和0.856（95% CI 0.809-0.903），表现出对心源性晕厥较高的识别准确性，经Bootstrap方法内部验证表明该模型具有良好的稳健性。

结论：本研究成功开发并验证了一种新的心源性晕厥诊断模型，该模型表现出对心源性以及非心源性晕厥优异的区分能力，有助于临床医生提高晕厥诊断的效率，促进心源性晕厥的早期识别与及时干预。

Part I

Preliminary Exploration of the Neuroendocrine Mechanisms of Vasovagal Syncope Based on Untargeted Metabolomics

Background: The pathogenesis of vasovagal syncope (VVS) is complex and not fully understood. Numerous studies have suggested that neuroendocrine regulation plays a significant role in VVS, with multiple neurohormones potentially involved. However, existing research is limited by traditional detection technologies that identify only a narrow range of substances, and it remains unclear whether other neuroendocrine factors or related pathways contribute to the development of VVS. Metabolomics, an emerging high-throughput technology, offers a promising approach to comprehensively analyze the dynamic changes in metabolites, thereby providing an opportunity to explore the multi-layered, multi-tissue responses during the occurrence of VVS.

Objective: This study aimed to perform untargeted metabolomic profiling of plasma samples collected from VVS patients before and after syncope episodes. The primary objective is to explore the systemic metabolic changes and related pathways during VVS. Additionally, this study sought to provide novel insights and perspectives for understanding the pathophysiological mechanisms of VVS from a holistic level.

Methods: This study prospectively enrolled VVS patients from the second department of the arrhythmia center at Fuwai Hospital between February 2023 and February 2024. All patients underwent the head-up tilt test (HUTT). Peripheral blood samples were collected while patients were in the resting supine position during the HUTT, and for HUTT-positive patients, additional blood samples were collected immediately during the occurrence of syncope. The plasma samples were analyzed using ultra-high-performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS) for untargeted metabolomic profiling. Differential metabolites were screened using univariate analysis, principal component analysis (PCA), partial least squares discriminant analysis (PLS-DA), and other multivariate analysis methods. Metabolite identification and pathway enrichment analysis were then conducted to explore their biological functions and potential significance.

Results: A total of 66 VVS patients were included in the study, with a mean age of 45.6 ± 17.0 years. 50% (33/66) of the patients were female, and the mean number of syncope episodes was 4. Of these, 45 patients (68.2%) were HUTT-positive (HUTT+), and 21 (31.8%) were HUTT-negative (HUTT-). Multivariate analysis revealed significant differences in plasma metabolic profiles at baseline between HUTT+ and HUTT- patients, as well as before and after syncope in HUTT+ patients. A comparison of baseline profiles between HUTT+ and HUTT- patients identified 62 differential metabolites, with 28 metabolites upregulated and 34 downregulated. Furthermore, in HUTT+ patients, 33 metabolites showed significant changes during syncope, with 32 upregulated and 1 downregulated compared to baseline. Pathway enrichment analysis identified 23 significantly altered signaling pathways, including pathways related to biosynthesis of unsaturated fatty acids, linoleic acid metabolism, PPAR signaling pathway, aldosterone synthesis and secretion, and fatty acid biosynthesis.

Conclusions: This study indicated significant differences in the plasma metabolomic profiles of VVS patients before and after syncope. Furthermore, metabolic characteristics also differed between HUTT+ and HUTT- patients, suggesting that the occurrence of VVS may be associated with systemic metabolic alterations. These findings provide new perspectives and evidence for a deeper understanding of the pathophysiology of VVS and the neuroendocrine changes involved in its development.

Part II

Changes and Implications of Serotonin Synthesis and Metabolism Pathways in the Pathogenesis of Vasovagal Syncope

Background: Neuroendocrine factors are believed to play a significant role in the pathophysiology of vasovagal syncope (VVS). Serotonin (5-hydroxytryptamine, 5-HT) is a crucial neurotransmitter involved in the regulation of hemodynamics and autonomic functions. However, its role in the pathogenesis of VVS remains unclear.

Objective: This study focused on the serotonin synthesis and metabolism pathway, aiming to investigate the changes in key serotonin-related substances in VVS and their potential implications.

Methods: This study prospectively enrolled VVS patients from the second department of the arrhythmia center at Fuwai Hospital between February 2023 and February 2024. All participants underwent the head-up tilt test (HUTT), with baseline peripheral blood samples collected in the supine position. In HUTT-positive patients, blood samples were also collected immediately during the occurrence of syncope. Hemodynamic parameters such as blood pressure and heart rate were continuously recorded and analyzed. A targeted metabolomics approach using liquid chromatography-tandem mass spectrometry (LC-MS/MS) was employed to quantify serotonin concentrations in plasma and platelets. The levels of serotonin-related synthesis and metabolites, including tryptophan, 5-hydroxytryptophan (5-HTP), and 5-hydroxyindoleacetic acid (5-HIAA), were also measured. The changes in these substances were compared between HUTT-positive (HUTT+) and HUTT-negative (HUTT-) patients, as well as before and after syncope in HUTT+ patients. Linear regression models were used to analyze the correlations between these changes and clinical characteristics and hemodynamic indices during syncope.

Results: A total of 66 VVS patients (45.6 ± 17.0 years, 33 females) were included in the study, of whom 45 were HUTT+ and 21 were HUTT-. No significant differences were observed in baseline plasma serotonin and platelet serotonin concentrations between the two groups. However, baseline plasma 5-HTP levels were significantly higher in HUTT+ patients compared to HUTT- patients (3.9 ± 0.5 vs. 3.3 ± 0.4 ng/mL, P < 0.001), while baseline plasma 5-HIAA levels were significantly lower in HUTT+ patients (10.3 ± 2.7 vs. 12.5 ± 4.3 ng/mL, P = 0.040). No significant changes were observed in plasma serotonin, platelet serotonin, or plasma tryptophan levels before and after syncope. However, plasma 5-HTP concentrations significantly decreased during syncope compared to baseline (3.6 ± 0.5 vs. 3.9 ± 0.5 ng/mL, P < 0.001), while plasma 5-HIAA concentrations significantly increased (12.4 ± 3.6 vs. 10.3 ± 2.7 ng/mL, P < 0.001). Subgroup analysis showed that spontaneous VVS patients exhibited a significant increase in plasma serotonin levels during syncope compared to baseline (15.9 ± 10.8 vs. 11.8 ± 6.5 ng/mL, P = 0.037), and the greater the serotonin change, the shorter the time to syncope onset during HUTT (R² = 0.38, P = 0.015). The linear regression analysis revealed some certain correlations between the changes in the serotonin precursor 5-HTP and variations in heart rate and blood pressure during VVS episodes.

Conclusions: Peripheral serotonin levels remained stable during VVS episode, while significant changes were observed in the serotonin precursor (5-HTP) and the major metabolite (5-HIAA), which correlated with hemodynamic changes during syncope. These findings suggest that alterations in serotonin synthesis and metabolism may contribute to the pathogenesis of VVS, providing new insights into the neuroendocrine mechanisms involved in VVS.

Part III

Biomarker Identification and Diagnostic Model Establishment for Vasovagal Syncope Based on Metabolomics

Background: Vasovagal syncope (VVS) presents diagnostic challenges due to its diverse clinical manifestations and complex pathophysiological mechanisms. Although several potential biomarkers for VVS have been identified, these findings have not yet been systematically organized, and no biomarker-based diagnostic model for VVS is currently available. Recently, the fast development of metabolomics technique offers new approaches for the early diagnosis of VVS and the discovery of novel biomarkers.

Objective: This study aimed to identify metabolic biomarkers associated with VVS using untargeted metabolomics, explore their potential biological roles in the pathogenesis of VVS, and develop a diagnostic model for VVS based on these biomarkers.

Methods: This study prospectively enrolled 59 patients with recurrent VVS (＞2 episodes in the past six months) and a positive head-up tilt test, who visited the second department of the arrhythmia center at Fuwai Hospital from September 2022 to September 2023. A healthy control (HC) group consisted of 25 healthy individuals matched by age, sex, and body mass index. Plasma samples from all participants were analyzed using ultra-high-performance liquid chromatography coupled with tandem mass spectrometry (UHPLC-MS/MS) for untargeted metabolomics. The differential metabolites between the VVS and HC groups were identified using t-tests, fold change analysis, and orthogonal partial least squares discriminant analysis (OPLS-DA). To further select the optimal combination of biomarkers to distinguish VVS from HC, supervised machine learning (ML) methods were employed to analyze the differential metabolites and build predictive models. Model performance was assessed using receiver operating characteristic (ROC) curves and their corresponding area under the curve (AUC).

Results: The average age of the 59 VVS patients was 43.8 ± 16.2 years, with 30 (50.8%) females, while the average age of the HC group was 43.1 ± 12 years, with 14 (56.0%) females. Metabolomic profiling revealed significant differences between the VVS and HC groups, identifying 289 upregulated and 39 downregulated differential metabolites. Pathway enrichment analysis indicated that these metabolites were primarily enriched in phenylalanine metabolism, caffeine metabolism, and the biosynthesis of valine, leucine, and isoleucine. The random forest algorithm successfully identified 7 metabolites as diagnostic biomarkers for VVS, including phosphatidylinositol (14:1/14:1), phosphatidylinositol (18:1/18:1), phosphatidylinositol (16:0/16:0), lysophosphatidylcholine (20:4), lysophosphatidylcholine (22:5), lysophosphatidylcholine (22:6), and linoleic acid. Based on these 7 metabolites, three ML models were constructed: random forest model (AUC 0.933; 95% CI 0.836–1), support vector machine model (AUC 0.958; 95% CI 0.886–1), and linear discriminant analysis model (AUC 0.916; 95% CI 0.791–1). All ML models demonstrated excellent predictive performance for VVS.

Conclusions: This study reveals significant metabolomics differences between VVS patients and healthy individuals through untargeted metabolomics analysis. Furthermore, the diagnostic models constructed using seven ML-identified biomarkers show high predictive value for VVS. These findings provide new strategies for VVS diagnosis and contribute to a deeper understanding of its pathogenesis.

Part IV

Development and Validation of a Diagnostic Model for Cardiac Syncope Based on Easily Accessible Clinical Indicators

Background: Cardiac syncope is associated with a higher risk of mortality and adverse outcomes compared to non-cardiac syncope. Therefore, early identification of patients with cardiac syncope is critical in syncope management. The establishment of a diagnostic model for the early identification of such patients is of considerable clinical importance.

Objective: This study aimed to develop a diagnostic model for cardiac syncope based on simple clinical indicators. By analyzing a set of easily accessible clinical parameters, the model sought to predict the likelihood of a diagnosis of cardiac syncope, thereby enhancing early detection capabilities.

Methods: This retrospective study analyzed syncope patients admitted to Fuwai Hospital between January 2021 and June 2022. Clinical data were collected, and patients with well-defined causes of syncope were included. Both cardiac and non-cardiac syncope cases (including reflex syncope and orthostatic hypotension-induced syncope) were analyzed. Patients were randomly divided into two groups: a training set and a validation set, with a 7:3 ratio. Feature selection was conducted using the least absolute shrinkage and selection operator (LASSO) regression on 36 candidate clinical variables. Subsequently, multivariable logistic regression analysis was performed to identify independent predictors of cardiac syncope, and a diagnostic model distinguishing cardiac from non-cardiac syncope was developed. Model performance was evaluated in the training and validation sets using receiver operating characteristic (ROC) analysis, calibration curves, and decision curve analysis.

Results: A total of 919 syncope patients were included, with 877 cases (age 57.2 ± 15.0 years, 358 females) having well-defined causes of syncope. Among them, 522 were cardiac syncope patients and 355 were non-cardiac syncope patients. From the 36 initial clinical variables, LASSO and logistic regression jointly identified five independent predictors of cardiac syncope: body mass index (OR 1.088; 95% CI 1.022-1.158; P = 0.008), chest symptoms preceding syncope (OR 5.251; 95% CI 3.326-8.288; P < 0.001), logarithmic NT-proBNP (OR 1.463; 95% CI 1.240-1.727; P < 0.001), left ventricular ejection fraction (OR 0.940; 95% CI 0.908-0.973; P < 0.001), and abnormal electrocardiogram (OR 6.171; 95% CI 3.966-9.600; P < 0.001). The diagnostic model based on these five variables achieved an area under the ROC curve (AUC) of 0.873 (95% CI 0.845-0.902) in the training set and 0.856 (95% CI 0.809-0.903) in the validation set, demonstrating high accuracy in identifying cardiac syncope. Internal validation using Bootstrap method confirmed the robustness of the model.

Conclusions: This study successfully developed and validated a new diagnostic model for cardiac syncope. The model demonstrated excellent discriminatory ability between cardiac and non-cardiac syncope, helping physicians improve diagnostic efficiency and promoting early identification and prompt intervention for cardiac syncope.