慢性心理应激、下丘脑-垂体-肾上腺轴失调与动脉粥样硬化性心血管疾病：机制、效应大小及基于冥想干预的证据

结构化的批判 叙述性综述 人类和转化证据

目录

1. 摘要
2. 范围、来源与证据分级框架
3. 主要结论
4. 已确立的事实、可能的情况与未知的内容
5. 应激生理学：反应的架构
6. 皮质醇测量：基质、动力学与判读陷阱
7. 人类流行病学证据：应激暴露与动脉粥样硬化性心血管疾病
8. 皮质醇与动脉粥样硬化性心血管疾病：病因、中介因素还是标志物？
9. 机制一：自主神经与血流动力学
10. 机制 II： 炎症 与免疫
11. 机制三：内皮与血管
12. 机制四：代谢
13. 机制 V： 血栓形成 与急性诱发
14. 机制六：行为中介
15. 机制证据矩阵
16. 干预问题：分类学与主张阶梯
17. 基准综合
18. 论点 A：感知压力
19. 主张 B：皮质醇与 HPA 生理学
20. 主张 C：已证实 风险因素
21. 主张D： 动脉粥样硬化
22. 声称 E：硬终点临床事件
23. 证据质量、偏倚与不良反应
24. 带逐链路证据权重的集成模型
25. 临床与预防意义
26. 会改变这些结论的阈值
27. 尚未解答的主要研究问题
28. 结论：核心问题的直接解答
29. 本综述的局限性
30. 参考文献

1. 摘要

背景。. 慢性心理压力被广泛断言会导致心脏病，而冥想则被广泛断言能预防心脏病。这两种说法往往被重复提及，却鲜少受到审视。本综述的目的在于明确指出经同行评议的证据支持什么、什么在生物学上合理但在人类身上尚未证实，以及真正的疑问在何处。.

方法. 对通过MEDLINE/PubMed、Scopus、Embase以及Cochrane图书馆（CENTRAL和CDSR）检索（检索截至2026年8月20日）并结合参考文献与引文追溯所识别的同行评议文献进行的结构化批判性叙述性综述。所有定量论断均可追溯至原始来源或指定的 系统评价 尽可能在全文中使用 IEEE 风格的编号引文。这是 不 一项系统评价：未注册研究方案，未保留去重后的记录数量，且未应用正式的研究层面偏倚风险评估工具。证据根据明确标准被赋予作者评估的叙事置信度类别——强、中等、有限或无定论，每类均说明了降级依据，并在全文中区分了关于冥想的五个可分离的主张：（A）减轻感知压力；（B）改变皮质醇或下丘脑-垂体-肾上腺轴（HPA）生理机能；（C）改善既定的危险因素；（D）减缓可测量的动脉粥样硬化；（E）减少硬临床终点事件。.

调查结果. 对于几个前瞻性研究的心理社会应激结构，与新发疾病调整后的相对关联 冠心病 和缺血性 中风 通常在 1.2 至 1.3 的范围内，尽管估计值会因暴露定义和结局而异。这些是适度的关联，但在群体规模上却是不容忽视的。它们与报告的相对效应不具有直接的可比性 吸烟, 血压 还是累积的 载脂蛋白 B (载脂蛋白B)暴露，这些暴露在不同的暴露量表上进行测量，并由不同的设计建立；将它们相互排序需要本综述未尝试进行的校准绝对风险比较。病例对照估计值大大增加（INTERHEART：持续性一般压力的比值比为2.17，具有 人群归因危险度 （综合心理社会因素的比值约为32.5%），而病例交叉研究的估计值更高（比值比为2.44，针对 心肌梗死 （急性愤怒后的一小时内），但这些研究设计测量的是不同的东西，且不能与前瞻性风险互换。因果证据是中等的：它建立在时间顺序、生物学合理性、可重复的实验人类生理学以及一个具有里程碑意义的脑成像队列研究之上——而不是建立在 随机化, ，以及残差 混杂的 不能排除社会经济地位和行为因素的影响。. 基因仪器不属于这个列表。. 可用的 孟德尔随机化 仪器早晨血浆皮质醇，一个下游 生物标志物, ，并非心理压力；该问题在第 8.4 节的“皮质醇因果关系”部分进行了探讨，对于压力本身的因果地位没有任何分量。.

皮质醇最好被理解为下丘脑-垂体-肾上腺（HPA）轴和糖皮质激素生理学的一个非特异性生物标志物以及一个在生物学上合理的介质。前瞻性研究和遗传学数据与心血管风险的微小因果贡献相符，但在通常的生理范围内因果关系仍不确定，而且皮质醇当然也不是心理压力的唯一传递者。 失调框架 比持续升高的循环皮质醇模型更具可辩护性。平缓的昼夜节律斜度、升高的傍晚皮质醇、改变的反应性，, 糖皮质激素受体抵抗, ，且改变的组织糖皮质激素代谢均与不良表型相关，尽管目前尚未确立单一的HPA表型作为决定性的心血管异常—— 元分析 在80项研究中，发现更平缓的斜率与总体较差的健康状况相关（r = 0.147），而随机效应心血管亚组则无统计学显著性 [78]。在白厅 II 号研究中，较平缓的昼夜斜率预示着心血管疾病死亡率（风险比 1.87，95% CI 1.32–2.64），而晨间皮质醇和 皮质醇觉醒反应 没有。在 LURIC 研究中，早晨血清皮质醇与心血管死亡率之间粗略的相关性（风险比为 1.32）在根据传统危险因素进行调整后便消失了（0.97）。.

冥想能有效减轻主观压力，并适度降低血压。目前尚无令人信服的、独立的、随机对照研究证据表明其能预防心肌梗死、中风或心血管死亡。 2024年Cochrane系统综述（81项随机试验，6,971名受试者，检索截止至2021年11月）发现，仅有两项对照研究提供了心血管临床事件数据：正念干预与无干预对照组（1项试验，110名受试者； RR 0.94，95% 置信区间 0.37–2.42；证据确定性极低）以及 超然冥想 与活性对照组相比（1项试验，201名参与者；RR 0.91，95%置信区间 0.56–1.49；证据确定性低）。 两项研究均未显示出可检测到的效果。其中唯一一项确定性中等收缩压估计值——超觉静坐与积极对照组的比较——也是其数值最小的：收缩压降低−2.33 mmHg。.

解释. 压力管理（包括冥想）是遵循指南的心血管预防的一种合理且低风险的辅助手段，主要用于改善血压、心理健康和行为自我调节。它不能替代 降脂, ，血压治疗、戒烟、血糖控制或体育活动。声称冥想“可以预防 心脏病发作,，” “逆转动脉粥样硬化，”或“使皮质醇正常化以防止 心血管疾病” 超出证据。.

2. 范围、来源与证据分级框架

2.1 解答的问题

本篇综述探讨了八个问题：

1. 慢性心理压力是否会增加动脉粥样硬化性心血管疾病（ASCVD）的风险？增加多少？
2. 该关系是因果关系而非受混杂因素影响的证据有多强？
3. 哪个生物学机制具有最强 人类 证据，与最强的动物或体外证据有何不同？
4. 皮质醇具体扮演什么角色——是原因、中介、标志物，还是某种组合？
5. 皮质醇升高本身具有致病性，还是下丘脑-垂体-肾上腺（HPA）轴失调是更准确的模型？
6. 压力和ASCVD之间的关联有多少是由血压和炎症介导的, 胰岛素抵抗, ，肥胖、睡眠和行为？
7. 冥想是否会改变这些指标——以及它处于证据等级的哪个级别？
8. 在今天，有什么是科学上有正当理由说得通的，又有什么构成了过度引申？

2.2 检索策略、信息源与入选标准

设计. 这是一个 结构化批判性叙事综述, ，而非系统评价。这一区别表述得十分明确，因为它限制了该评价可以主张的内容：它可以权衡, 年级, ，并对其识别出的证据进行解读，但不得断言该文献已被穷尽。.

数据库与日期. 检索了 MEDLINE/PubMed、Scopus、Embase 以及 Cochrane 图书馆（CENTRAL 和 Cochrane 系统评价数据库），检索的最终截止日期为 2026年8月20日. 使用 PubMed Central 获取全文，并且 不 被视为一个独立的书目数据库。参考文献列表和前向引文追溯（谷歌学术）被用于识别额外的同行评审研究，特别是针对科克伦检索（截止于2021年11月14日）未涵盖的2021年之后的冥想试验。.

表 1. 检索领域、概念字符串和资格标准。.

复习题	核心概念字符串（根据数据库语法进行调整）	合格的设计	主要除外责任
应激 → 动脉粥样硬化性心血管疾病结局	(“心理应激” OR “工作压力” 或 “工作压力” 或 “感知的压力” 或 孤独 或 “社会隔离” 或 “照顾者负担”）和（心肌梗死 或 “冠心病” 或 中风 或 “心血管死亡率” 或 动脉粥样硬化）	前瞻性队列, ，IPD荟萃分析，病例对照，病例交叉，系统评价	仅限横断面暴露-结局设计；仅限动物研究
皮质醇 / 下丘脑-垂体-肾上腺轴 → 动脉粥样硬化性心血管疾病结局	(皮质醇 OR “下丘脑-垂体-肾上腺轴” OR 糖皮质激素 OR “皮质醇觉醒反应” OR “昼夜节律斜率” OR “头发皮质醇”) AND (心血管 OR 冠心病 OR 中风 OR 死亡率 OR “冠状动脉) 动脉 钙”	前瞻性队列，孟德尔随机化，系统评价，自然实验（库欣）	病例报告；外源性类固醇药理学
机制途径	(stress OR cortisol OR catecholamine) AND (inflammation OR “flow-mediated dilation” OR endothelial OR haematopoiesis OR NLRP3 或 凝血 或 “胰岛素 ”抵抗”	人体实验、前瞻性机制队列研究、包含人体成分的转化研究	仅限动物研究，除非明确标明为此类研究
冥想 → 结果	(meditation OR mindfulness OR MBSR OR “transcendental meditation” OR “stress reduction”) AND (“blood pressure” OR cortisol OR “心率变异性” 或 炎症 或 “内中膜” 或 “心血管事件”）	随机对照试验; ；随机对照试验的系统评价和荟萃分析	非对照/单臂研究；未经同行评审的材料

资格. 纳入任何语言、任何地域、任何时间的同行评审人体研究。学会科学声明被引用为共识立场，并明确标注，绝不作为主要证据。非同行评审材料——机构新闻稿、杂志报道和组织评论——被排除在证据基础之外。若全文存在付费墙，则从已发表的摘要中获取定量数值，并在各个索引来源之间进行交叉核对；此类情况在第 29 节中予以标明。.

本篇评论所没有的内容。. 未前瞻性注册方案。筛选工作由单名审阅者独立完成，未经双人独立筛选。. 去重后的记录数量和研究筛选流程图未予保留，因此在此不予报告，亦未进行回顾性重建。. 正式工具（RoB 2、ROBINS-I、AMSTAR-2）未在研究层面应用；偏倚风险是通过对具名领域的叙述性评估来进行的。这些是实质性局限，第 2.5 节阐述了它们如何限制本综述允许提出的主张。.

2.3 作者评估的叙事置信度类别

使用了四个类别，这些类别应用于个别主权声索而非整个文献。. 这些是由作者评估的叙事置信度类别，而非 GRADE 评级, ，即使词语相同，它们也不能与 GRADE 确定性级别互换。.

分类	标准
强大	跨多种研究设计（包括人体实验数据或超大型前瞻性队列研究）的一致发现；在调整混杂因素后效应方向保持稳定；已在人体中证明了合理的机制
适度	具有一致的前瞻性关联或可重复的人体实验生理学证据，但受限于残余混杂、事件数较少、依赖单一研究或依赖于 替代终点
有限	小型、异质性或相互矛盾的人类研究；主要在动物或细胞系统中确立的机制；效果仅存在于亚组中
无定论	没有统计效力充足且低偏倚的研究来回答这个问题；现有的数据与一系列结论相容，包括无效果

为了使这些评估具有可审计性而非流于印象，表9中的每一个评分声明都带有明确的 降级依据, ，取自六个指定的领域： 偏倚风险，间接性，不精确性，不一致性，混杂, 以及 发表偏倚.

2.4 构成整篇综述的两个区别

首先：暴露-结局层级。. 连接证据 感知压力 到 生物标志物 不是将生物标志物与 风险因素, ，这不是将风险因素联系起来的证据 动脉粥样硬化, ，这不是将动脉粥样硬化联系起来的证据 硬终点指标. 每个链接都必须单独建立。文献经常跨这些层级借用可信度。.

第二：时间尺度。. 长达数十年的慢性暴露与在几分钟内起作用的急性触发是具有不同证据基础的不同现象（图 4）。在急性愤怒发作后的一小时内，心肌梗死的优势比（odds ratio）为 2.44，这是关于 易损斑块 跨越临床阈值；这并不是关于如何 斑块 形成。混淆它们会使压力的表面致动脉粥样硬化效应放大数倍。.

2.5 本综述可以声称和不能声称的内容

由于该检索是有结构但非系统的，本综述采用了以下自我设定的限制条件，第 22 至 28 节的编写均遵循了这些限制：

1. 不得提出无条件缺勤主张。. 表明不存在相关试验或研究的陈述表述为“we identified no…”，并附有检索日期。.
2. 不保证内容详尽无遗。. 这个词 全面的 不用于描述该搜索。.
3. 怀疑论的对称性. 应用于冥想声明的证据标准同样适用于压力因果关系和皮质醇因果关系的声明。如果该综述拒绝从干预试验中的生物标志物变化推断事件减少，它同样拒绝从观察性队列研究中的生物标志物变化推断斑块进展。.
4. 成绩是批改者的判断。. 它们作为结构化推理提供，并陈述了降级理由，而不是作为正式的确定性评级。.

3. 关键结论

1. 前瞻性关联是真实存在的，但程度较小。. 对几个常被研究的心理社会结构（工作压力、感知压力、孤独感、社会隔离）进行的调整后的前瞻性估计，通常落在近似的 2–1.3 冠心病和缺血性中风发病率的区间，尽管个别估计值在两个方向上都超出此区间。在仅基于已发表文献的综合分析、调整不完全的分析以及病例对照研究文献中，1.4至1.6左右的估计值更为常见，不应将其作为前瞻性估计值来呈现。.
2. 病例对照和病例交叉估计值更大，且含义有所不同。. INTERHEART 报告的比值比为 17 (99% 置信区间 1.84–2.55) 对于持续的总体压力以及大约的群体归因风险 32.5% 针对综合心理社会因素——而不仅仅是针对长期压力——1]; INTERHEART 病例交叉分析显示 OR 2.44（99% 置信区间 2.06–2.89） 在愤怒或情绪激动后一小时内发生的心肌梗死9]。两者都具有信息量；都不是前瞻性发病率估计值。.
3. 发表偏倚已被证实，而非假设。. 在 IPD-Work 联合研究中，工作紧张感带来的风险比（hazard ratio）为 43 (1.15–1.77) 与已发表的队列相比 1.16 (1.02–1.32) 在未发表的研究中——这提供了异常直接的证据，表明发表状态与更大的报告效应有关，也是赋予较低数值更大权重的理由 [3].
4. 皮质醇并非全部真相，甚至可能不是最主要的因素。. 循环皮质醇与心血管结局的关系在各项研究中并不一致，其结果模式追踪了 测量窗口 比它追踪生物学特征的契合度更高（图 2 和图 5）。.
5. HPA轴功能失调是一个比持续性高皮质醇血症更具可辩护性的框架 ——但“失调”不应被归结为一种单一的强制性表型。昼夜节律变平、深夜皮质醇升高、反应性改变以及糖皮质激素受体抵抗都在不良环境中被报道过，但它们的心血管预后意义并不统一，且依据的是心血管事件较少的人群队列。现有证据不支持这样一种普遍模型：即慢性心理压力可靠地导致循环皮质醇持续升高。.
6. 炎症和造血信号传导是人类最具连贯性的机制通路之一——甚至比皮质醇更甚。. 在包含22例事件的单一影像学队列中，研究证实了“杏仁核 → 骨髓 → 动脉炎症 → 事件”这一通路；本综述认为，这是目前可获得的最具整合性的脑-动脉人类数据集之一——其中骨髓活动介导了杏仁核–动脉炎症关系的46%，而动脉炎症则介导了杏仁核–事件关系的39% [11].
7. 行为是因果系统的一部分，而不是干扰变量。. 由于慢性痛苦会导致吸烟、缺乏运动、睡眠不好以及不遵医嘱，通过统计学方法将这些因素调整掉可能会低估压力的总体影响——尽管这并不能就此隔离出直接的生物学效应，而生物学效应需要正式的中介分析。.
8. 冥想的证据从A类主张到E类主张急剧减弱 （图 8）。感知压力：中等。皮质醇：有限到中等，且主要局限于高危样本。血压：中等，在使用主动对照组后比通常报道的要小。动脉粥样硬化：有限且无定论——早期的一项阳性 cIMT 试验脱落率很高，而随后在不同人群中进行的一项随机试验未显示出组间 cIMT 受益。硬终点事件：无定论。.
9. 比较对象的选择对答案有实质性影响。. 在Cochrane系统综述中，每项大型血压估计值要么具有I² ≥ 87%，要么纳入的受试者少于150人；唯一一项确定性中等水平的收缩压估计值为−2.33 mmHg（图7）。.
10. 研究者效忠偏见是该文献中一个重要的潜在偏见来源, ，特别是在超觉静息研究中，唯一报告出硬终点获益的试验均源自单一附属机构的研究者网络。这是需要独立重复研究的理由，本身并不证明研究者的倾向性产生了所报告的效果。.

4. 已确定的、可能的和未知的

已确立（强有力的证据）。. 前瞻性研究和病例对照研究表明，几种心理社会应激结构与心血管事件相关，而人类影像学和机制研究则为特定的中间途径提供了支持证据。急性精神应激可产生可重复的、短暂的损害 内皮功能 在健康成年人和冠心病患者中。急性情绪应激可以触发 心脏事件 在已经患有疾病的人群中。病理性皮质醇增多症——库欣综合征——会导致 高血压, ，糖脂代谢异常，内脏脂肪过多，血脂异常，以及过早的心血管死亡。冥想可以降低感知到的压力并适度降低血压。在人类中，慢性压力与白细胞生物学改变有关。.

很可能（中等证据）。. 普通慢性压力对心血管风险的因果贡献。（关于压力诱发斑块的直接人类证据） 发生或进展 明显弱于临床事件、血压通路、内皮反应和急性触发的证据，其等级为“有限至中等”。杏仁核-骨髓-动脉轴。炎症作为中介因素。昼夜皮质醇失调作为心血管死亡率的预测指标。皮质醇作为 possible causal contributor — genetic evidence is compatible with a small effect but is statistically inconclusive. Behavioural mediation is probably important, but the fraction of the total association attributable to behavioural pathways is not established.

Limited or uncertain. Whether meditation meaningfully changes cortisol outside at-risk populations. Whether meditation slows atherosclerosis. The independent clinical value of any circulating cortisol measure for cardiovascular risk stratification. Whether the specific NLRP3/IL-1β axis mediates psychological-stress effects in humans, as distinct from mediating atherosclerosis generally. The proportional contribution of direct neuroendocrine injury versus indirect behavioural pathways.

Inconclusive. Whether any stress-reduction intervention reduces hard cardiovascular events. We identified no adequately powered, low-bias, independently conducted trial in our search to 20 August 2026. This is the central gap in the field.

5. Stress Physiology: The Architecture of the Response

5.1 Central appraisal

Psychological stress depends not simply on an external input but on the brain’s appraisal of threat, demand, predictability, and coping resources. Threat cues are processed in distributed cortico-limbic circuitry — prefrontal cortex, amygdala, hippocampus, insula — in which prefrontal, limbic, insular and hypothalamic networks jointly regulate threat appraisal and response, with prefrontal regulation able to modulate amygdalar output and contribute to terminating the response once a threat resolves. Chronic psychological adversity is associated with a shift in this balance toward sustained amygdalar activity relative to prefrontal regulation [11,37]. From the amygdala and the hypothalamic paraventricular nucleus, two effector arms descend.

5.2 The sympathetic-adrenal-medullary arm

The SAM axis responds within seconds. Sympathetic outflow and adrenal medullary secretion raise circulating epinephrine and norepinephrine, increasing 心率, contractility, and peripheral vascular resistance while parasympathetic (vagal) tone is withdrawn. This is often reflected in changes in vagally mediated heart-rate variability, though HRV is also influenced by respiration, posture and recording conditions and is not a one-to-one readout of 迷走神经张力. Beyond haemodynamics, catecholamine signalling acts directly on immune and haematopoietic compartments — a point developed in Section 10.

5.3 The hypothalamic-pituitary-adrenal arm

The HPA axis operates over minutes to hours. Paraventricular neurons release corticotropin-releasing hormone and 精氨酸 vasopressin into the hypophyseal portal circulation; CRH stimulates pituitary corticotropes to secrete adrenocorticotropic hormone; ACTH binds melanocortin-2 receptors in the adrenal zona fasciculata, driving conversion of 胆固醇 to cortisol. Cortisol then exerts negative feedback at both hypothalamic and pituitary levels.

Acutely, this is adaptive. Cortisol mobilizes energy substrates, enhances vascular reactivity to catecholamines, and restrains the immune response. The pathological question is never whether these systems activate — they must — but whether repeated activation, failed recovery, altered circadian timing, or altered receptor responsiveness generates sustained allostatic burden.

5.4 Receptor biology: two receptors, not one

Cortisol binds two intracellular nuclear receptors with markedly different affinities:

• 这 mineralocorticoid receptor (MR, Type I) has high affinity and is substantially occupied at basal cortisol concentrations.
• 这 glucocorticoid receptor (GR, Type II) has lower affinity and is engaged principally during stress-induced peaks and the morning surge.

This two-receptor architecture matters clinically. Under sustained glucocorticoid excess, renal MR occupancy by cortisol — normally limited by pre-receptor metabolism, below — has been proposed to promote sodium retention, volume expansion and hypertension [30]. This is one described route by which glucocorticoid excess may become a haemodynamic problem.

5.5 Pre-receptor metabolism: the tissue-level amplifier

Circulating cortisol is a poor proxy for the cortisol a given tissue actually experiences, because two enzymes modulate local exposure independently of plasma concentration:

• 11β-HSD1, abundant in liver and visceral adipose tissue, regenerates active cortisol from inactive cortisone. It amplifies intracellular glucocorticoid signalling, gluconeogenesis, and 内脏脂肪 deposition even when plasma cortisol is normal [30].
• 11β-HSD2, expressed in aldosterone-sensitive distal nephron, inactivates cortisol to cortisone, protecting the renal MR from occupancy by cortisol [30].

The practical implication is that a normal serum cortisol does not exclude tissue-level glucocorticoid excess. This is a real limitation on any clinical strategy built around measuring circulating cortisol.

5.6 The circadian architecture

Under unstressed conditions cortisol follows a pronounced circadian rhythm driven by the suprachiasmatic nucleus: a sharp rise peaking roughly 30–45 minutes after waking (the 皮质醇觉醒反应, CAR), a steep decline across the day, and a nadir near midnight (Figure 1A). The rhythm — its amplitude, its slope, and the timing of its nadir — carries information that a single value cannot.

5.7 Why “chronic stress produces high cortisol” is wrong

This is the single most important correction in this review. Chronic stress does 不 reliably produce persistently elevated cortisol. Depending on chronicity, timing, developmental history, and receptor sensitivity, it can produce hypercortisolism, hypocortisolism, a blunted CAR, an exaggerated CAR, or a flatter diurnal slope with elevated late-day or nadir cortisol — phenotypes associated with adverse health outcomes in several cohorts, but without uniformly consistent cardiovascular associations in meta-analysis [78].

Three further observations complete the picture:

Blunted reactivity is not necessarily resilience. Meta-analysis of adverse childhood experiences finds blunted rather than exaggerated cortisol and cardiovascular stress reactivity [40]. Low reactivity in that context reflects dysregulation, not robustness. In 梅萨, blunted diastolic blood-pressure reactivity to acute stress predicted premature 全因死亡率 [42]. Blunted reactivity is therefore not evidence of resilience; but the prognostic implications vary by physiological measure and population, and these data do not establish a general U-shaped rule across HPA and cardiovascular responses.

Glucocorticoid receptor resistance decouples cortisol from its effect. In chronically stressed caregivers, salivary cortisol profiles can resemble those of controls while monocytes show diminished glucocorticoid-responsive transcription, enhanced NF-κB-related transcription, and higher inflammatory markers [14,15]. Multiple molecular mechanisms have been proposed to alter GR sensitivity, including changes in receptor expression, impaired ligand binding and nuclear translocation, co-chaperone regulation of receptor turnover, and epigenetic modification at loci such as FKBP5. Several of these details derive primarily from mechanistic and translational work rather than from direct human cardiovascular experiments [30], and they are listed here as candidate mechanisms rather than as demonstrated steps in a human pathway.

This resolves the central paradox. Cortisol is anti-inflammatory, yet chronic stress is pro-inflammatory. One mechanism that may partly reconcile this is reduced glucocorticoid sensitivity in selected immune-cell populations — the target tissue listening less well. It is a candidate resolution, not a demonstrated universal property of chronic stress. On this account inflammation proceeds not despite adequate cortisol but because adequate cortisol is less effectively transduced — while catecholamine- and sympathetically-driven pro-inflammatory haematopoiesis pushes in the same direction. Chronic inflammation in stressed populations is, however, multifactorial: sleep, adiposity, metabolic state, infection and behaviour all contribute, and glucocorticoid resistance is one contributor among several.

The accurate formulation is therefore stress → HPA-axis dysregulation, not stress → high cortisol.

图 1. Diurnal cortisol pattern may be more informative than a single morning level for cardiovascular prognosis. Panel A is a schematic of characteristic phenotypes, not measured data. Panel B reports observed hazard ratios; in Whitehall II, morning cortisol and the cortisol awakening response did not predict mortality.

6. Measuring Cortisol: Matrices, Kinetics, and Interpretive Traps

Part of the apparent inconsistency in the cortisol–ASCVD literature reflects differences in sampling window and matrix rather than biological disagreement — though confounding, disease state, medication, population heterogeneity and analytic choices also contribute [41] (Figure 2).

Figure 2. Cortisol biomarkers measure different things over different timescales. A single morning serum cortisol captures one time point in a system with a pronounced circadian rhythm; part of the inconsistency in this literature reflects that mismatch, alongside genuine biological heterogeneity.

Table 2. Cortisol measurement matrices.

Matrix	Window captured	它测量什么	Principal strengths	Principal limitations
Serum / plasma, single draw	Single time point	Total cortisol (bound + free)	Widely available; standardized assays	Acutely sensitive to venepuncture stress and sampling time; influenced by corticosteroid-binding globulin; a poor index of chronic exposure
Salivary, multi-sample	Hours to a day	Free, biologically active cortisol	Non-invasive; enables CAR and diurnal slope; repeatable	Adherence-dependent; sensitive to waking-time accuracy, food, and smoking
24-hour urinary free cortisol	1 day	Integrated daily free-cortisol output	Integrates across the day; unaffected by single-timepoint noise	Collection errors common; affected by renal function; one day may not represent chronic state
Hair cortisol / cortisone	Weeks to months (conventionally ≈1 cm of proximal scalp hair ≈ 1 month, with substantial biological and assay variability)	Cumulative systemic exposure	A useful longer-term integrated glucocorticoid biomarker, although standardization and clinical interpretation remain incomplete	Affected by hair biology, colour, washing and cosmetic treatment; assay heterogeneity; clinical scaling not established

Three interpretive consequences follow.

First, a single morning serum cortisol is a weak marker of chronic stress [41], and studies using it should not be treated as testing the same hypothesis as studies using diurnal salivary profiles or hair glucocorticoids. One well-conducted 横断面研究 found no association between circulating plasma cortisol and coronary or peripheral arterial disease [29] — a finding that constrains the clinical utility of spot cortisol rather than refuting the broader stress hypothesis.

Second, the shape of the reported effect tracks the matrix. In Figure 5, the largest point estimates come from integrated-exposure measures in small cohorts with few events; the most precise estimates come from pooled single-timepoint measures with modest effects. This pattern is compatible with measurement differences and small-study imprecision, including potential small-study inflation or winner’s-curse effects, but does not by itself identify the true causal effect size.

Third, cortisone may outperform cortisol in hair. In the Lifelines cohort, hair cortisone — but not hair cortisol — was independently associated with incident cardiovascular disease, most strongly in participants under 60 (odds ratio 4.21, 95% CI 1.91–9.07 per log₁₀ unit) [20]. The finding is biologically interesting and requires replication before any clinical inference.

7. Human Epidemiologic Evidence: Stress Exposures and ASCVD

Figure 3 displays the principal estimates. This section explains what each contributes and where each is weak.

Figure 3. Psychosocial exposures and atherosclerotic cardiovascular outcomes. Study design determines interpretation. Many adjusted prospective estimates for CHD and ischaemic-stroke outcomes fall in the approximate 1.2–1.4 range, although individual prospective estimates lie outside that interval. INTERHEART and INTERSTROKE intervals are 99% CI as reported.

7.1 Case-control evidence: INTERHEART and INTERSTROKE

INTERHEART enrolled 11,119 cases of first myocardial infarction and 13,648 controls across 52 countries. Permanent general stress carried an odds ratio of 2.17 (99% CI 1.84–2.55); several periods of stress, 1.45; financial stress, 1.33; stressful life events, 1.48; depression, 1.55. A high internal locus of control was protective (0.68). Separately, the combined psychosocial factor — not permanent general stress alone — carried a population-attributable risk of approximately 32.5% for myocardial infarction, and the pattern was consistent across regions, sexes, and ethnic groups. The association persisted after adjustment for income and education [1].

INTERSTROKE, using the same architecture for acute stroke, found psychosocial stress at odds ratio 1.30 (99% CI 1.06–1.60) and depression at 1.35 (99% CI 1.10–1.66) [2]. Note that both studies report 99% rather than 95% 置信区间 by design; the intervals are therefore wider than they would otherwise appear.

The limitation is structural and unavoidable. In a case-control study of myocardial infarction, stress exposure is ascertained 在……之后 the event, from people who have just had a heart attack and are searching for an explanation. Post-event ascertainment creates potential for differential recall that could inflate associations, although its direction and magnitude cannot simply be assumed. INTERHEART’s consistency across 52 countries strengthens generalizability, but cross-regional consistency does not eliminate differential recall. A well-designed case-control study can validly estimate an exposure odds ratio, and with incidence-density sampling that odds ratio estimates an incidence-rate ratio. What the design does not do is directly measure population incidence, and it remains vulnerable to differential retrospective exposure ascertainment. This is why the 2.17 figure, though frequently quoted, should not anchor the reader’s sense of magnitude.

7.2 Occupational strain

这 IPD-Work consortium pooled individual participant data from 13 European cohorts: 197,473 initially disease-free workers, 2,358 incident coronary events, mean 7.5 years of follow-up. Job strain — high psychological demand combined with low decision latitude — carried a hazard ratio of 1.23 (95% CI 1.10–1.37), with a population-attributable risk of roughly 3.4% [3].

Two features of this analysis are more instructive than the headline number.

First, a point of precision usually lost in secondary citation: the headline hazard ratio of 1.23 is adjusted for sex and age only. The consortium reported separately that the association was also present after adjustment for socioeconomic status and for lifestyle and conventional risk factors, and in analyses excluding events occurring in the first three years (HR 1.31, 1.15–1.48) and five years (HR 1.30, 1.13–1.50) of follow-up to address 反向因果关系. The effect is robust across specifications, but the widely quoted 1.23 should not be described as fully risk-factor-adjusted.

Second, and more valuable, the consortium compared previously published with previously unpublished cohorts. The hazard ratio in previously published cohorts was 1.43 (1.15–1.77); in previously unpublished cohorts it was 1.16 (1.02–1.32). This provides unusually direct evidence that publication status was associated with effect-size differences within a single consortium — though the contrast may also reflect factors correlated with publication status rather than publication bias alone. It is also specific to the job-strain corpus and should not be assumed to hold at the same magnitude in other stress literatures. It may nonetheless help explain part of the discrepancy between reviews reporting 1.2–1.4 and those reporting 1.4–1.6 — the latter are, in substantial part, reading the published literature at face value.

The opposite criticism has also been made, and should be recorded. Published correspondence argued that the constituent cohorts carry unacknowledged biases toward the null — single-occasion exposure measurement, healthy-worker selection, and loss of exposure contrast across follow-up — such that both 1.23 and the 3.4% population-attributable risk may themselves be underestimates [71,72]. The consortium replied defending its estimates [73]. This review does not adjudicate between the two critiques. It notes that the point estimate is bracketed by credible arguments in both directions, which is a further reason to treat 1.2–1.4 as a range rather than a measurement.

For stroke, the meta-analytic estimate was RR 1.22 (1.01–1.47) overall and 1.58 (1.12–2.23) for ischaemic stroke, while the more conservative individual-participant analysis gave HR 1.09 (0.94–1.26) for total stroke and 1.24 (1.05–1.47) for ischaemic stroke [4,5]. The ischaemic-specific signal is the more consistent of the two.

7.3 Perceived stress

Pooling six prospective cohorts and 118,696 participants, high perceived stress carried a risk ratio of 1.27 (95% CI 1.12–1.45) for incident coronary heart disease [6]. The exposure instruments varied across studies, which limits precision but also means the finding is not an artefact of one questionnaire.

7.4 Loneliness and social isolation

Across 16 longitudinal datasets and more than 181,000 adults, loneliness and social isolation carried RR 1.29 (1.04–1.59) for coronary heart disease and RR 1.32 (1.04–1.68) for stroke [7]. These estimates sit within the same band as job strain and perceived stress despite measuring a conceptually distinct exposure. That supports a broader psychosocial-risk association, but similarity between distinct constructs does not by itself constitute causal replication or imply a shared mechanism.

7.5 Caregiver strain

In the Caregiver Health Effects Study, 392 spousal caregivers aged 66–96 were followed alongside 420 non-caregiving controls. Caregivers reporting mental or physical strain had a 63% higher four-year all-cause mortality (RR 1.63, 95% CI 1.00–2.65) after adjustment for age, sex, health status, and subclinical disease [8]. The lower confidence bound touches 1.00, and the outcome is all-cause rather than cardiovascular mortality; the study is best read as corroborating rather than as independently establishing.

7.6 Acute triggering

这 INTERHEART case-crossover analysis of 12,461 first myocardial infarctions found that anger or emotional upset in the preceding hour was associated with OR 2.44（99% 置信区间 2.06–2.89）, with a population-attributable risk of 8.5% [9]. The within-person design controls for time-invariant 混杂因素 — 遗传学, socioeconomic position, chronic behaviour — which is a substantial methodological advantage over cohort designs. It does not control time-varying confounders, nor differential recall of what occurred during the hazard period.

But it answers a different question. The comparison is between one hour and another hour in the same person, whose coronary anatomy is identical in both. It establishes that a transient physiological state can precipitate an event in existing disease. It says nothing about 动脉粥样硬化发生. Related evidence — takotsubo 心肌病, population spikes in cardiac events after earthquakes and during emotionally charged sporting events, case-crossover data on anger and heavy exertion — belongs to this same category (Figure 4).

Figure 4. Two distinct timescales that are routinely conflated. Acute triggering is evidence that a transient physiological state can push pre-existing disease across the clinical threshold; it is not evidence that stress builds plaque.

7.7 Synthesis

Table 3. Human epidemiologic evidence linking stress exposures to ASCVD outcomes.

学习	Design / N	Exposure	Outcome / follow-up	Effect (95% CI unless noted)	Key limitation	等级
INTERHEART, 2004	Case-control; 24,767; 52 countries	Self-reported permanent stress, financial stress, life events	Acute MI	OR 2.17 (99% CI 1.84–2.55) permanent stress; combined psychosocial PAR ≈32.5%	Recall bias inherent to design	适度
INTERSTROKE, 2010	Case-control; 22 countries	Psychosocial stress; depression	Acute stroke	OR 1.30 (99% CI 1.06–1.60); depression 1.35 (1.10–1.66)	Same design limitation	适度
IPD-Work, 2012	IPD meta-analysis, 13 cohorts; 197,473; 2,358 events	Job strain (demand–control)	Incident CHD; mean 7.5 y	HR 1.23 (1.10–1.37); PAR ≈3.4%	Published cohorts 1.43 vs unpublished 1.16	适度
Huang et al., 2015	Meta-analysis, 6 cohorts; 138,782	Job strain	Incident stroke	RR 1.22 (1.01–1.47) total; 1.58 (1.12–2.23) ischaemic	Fewer cohorts; heterogeneous ascertainment	适度
Fransson et al., 2015	IPD meta-analysis	Job strain	Incident stroke	HR 1.09 (0.94–1.26) total; 1.24 (1.05–1.47) ischaemic	Null for total stroke	适度
Richardson et al., 2012	Meta-analysis, 6 cohorts; 118,696	Perceived stress scales	Incident CHD	RR 1.27 (1.12–1.45)	Instrument heterogeneity	适度
Valtorta et al., 2016	Meta-analysis, 16 datasets; >181,000	Loneliness / social isolation	CHD and stroke	RR 1.29 (1.04–1.59) CHD; 1.32 (1.04–1.68) stroke	Exposure definition varies	适度
Schulz & Beach, 1999	Prospective cohort; 812	Caregiver strain	4-y all-cause mortality	RR 1.63 (1.00–2.65)	All-cause outcome; CI touches null; no biomarkers	适度
INTERHEART trigger, 2016	Case-crossover; 12,461 MI cases	Anger / upset in prior hour	MI onset	OR 2.44 (99% CI 2.06–2.89); PAR 8.5%	Post-MI recall; triggering ≠ atherogenesis	Moderate–Strong for triggering

The synthesis is straightforward, with its scope stated. For several prospectively studied psychosocial constructs and CHD or ischaemic-stroke outcomes, adjusted estimates commonly fall in the approximate 1.2–1.4 range. That statement does not extend to every prospective entry in Table 3 — total stroke in the IPD analysis was 1.09, and the caregiver estimate of 1.63 is for all-cause mortality. They are directionally robust, replicated across exposures and populations, and consistent enough that sampling variation alone is unlikely to explain the pattern. They are also modest, vulnerable to residual confounding by socioeconomic position and behaviour, and, in the occupational literature, published cohorts produced materially larger estimates than unpublished cohorts.

A caution on calibration: it is tempting to rank this magnitude against smoking, established hypertension, or cumulative apoB exposure, but relative effects estimated on different exposure scales, over different induction periods, and by different designs are not directly commensurable. What can be said is that job strain’s population-attributable risk in IPD-Work was approximately 3.4%, which the investigators themselves described as substantially smaller than that of standard risk factors [3] — a comparison made within a single analysis rather than across literatures.

8. Cortisol and ASCVD: Cause, Mediator, or Marker?

Figure 5 arrays the principal estimates. What follows is the case for each interpretation and the resolution.

Figure 5. Cortisol measures and cardiovascular outcomes. Because exposure scales, cortisol matrices and outcomes differ, point estimates should not be compared quantitatively across rows. Estimates in grey cross the null. The largest point estimates come from the studies with the fewest events and the widest confidence intervals.

8.1 Diurnal rhythm

Whitehall II followed 4,047 civil servants for a mean of 6.1 years (139 deaths, 32 cardiovascular). A flatter diurnal cortisol slope predicted all-cause mortality (HR 1.30 per 1 SD reduction in slope steepness, 95% CI 1.09–1.55) and, more strongly, cardiovascular death (HR 1.87, 95% CI 1.32–2.64), independent of covariates. Morning cortisol and the CAR did 不 predict mortality [16]. This is an influential prospective observation: within a single well-characterized cohort, the rhythm predicted and the level did not. It rests on 32 cardiovascular deaths and requires replication.

KORA-F3 followed 1,090 participants for approximately 11 years (31 cardiovascular deaths) and found the same pattern from the other direction. A more pronounced CAR was associated with 较低 cardiovascular mortality (HR 0.59, 95% CI 0.36–0.96), as was a greater peak-to-bedtime ratio (HR 0.50, 0.34–0.73), while high late-night cortisol was associated with higher cardiovascular mortality (HR 1.49, 1.13–1.97). Preserved diurnal variability appeared protective [17].

在 coronary artery bypass patients, a steeper pre-surgical diurnal slope was protective for MACE and death (HR 0.73, 95% CI 0.56–0.96 per unit steeper) [18].

A broader synthesis provides an important qualification. A 2017 systematic review and meta-analysis of 179 associations from 80 studies found flatter diurnal cortisol slopes associated with poorer health overall (average r = 0.147), with significant associations in 10 of 12 outcome subtypes — including mortality, and largest for immune and inflammatory outcomes (r = 0.288). The cardiovascular-disease symptoms and diagnoses subgroup was not statistically significant in the random-effects model (r = 0.098, 95% CI −0.034 to 0.226) [78]. The cardiovascular-specific evidence is therefore suggestive rather than uniformly consistent, and this qualifies — rather than contradicts — the positive cardiovascular-mortality findings above.

Across the three cohorts, more preserved diurnal organization was generally associated with better outcomes, whereas amplitude at any single point was not. The measures and estimands nonetheless differ between studies, and the caveat is shared and serious: all three had few cardiovascular deaths — 32, 31, and a modest event count respectively — with correspondingly wide confidence intervals. Taken together, these data favour studying diurnal organization rather than relying on a single morning value, but they do not establish one prognostically validated cardiovascular HPA phenotype.

8.2 Integrated output

在 InCHIANTI, 861 adults aged 65 and over were followed for a mean 5.7 years. Twenty-four-hour urinary cortisol in the highest tertile carried HR 5.00 (95% CI 2.02–12.37) for cardiovascular mortality — but on only 41 cardiovascular deaths, with a confidence interval spanning a sixfold range, in an elderly cohort, from a single baseline collection [19]. Two features argue against the most obvious alternative explanations: the association was specific to cardiovascular rather than non-cardiovascular mortality, and it was consistent in participants with and without cardiovascular disease at baseline (p for interaction = 0.78), which constrains reverse 因果关系 [19]. The imprecision nonetheless remains severe, and the point estimate should never be quoted without its interval.

在 Lifelines cohort (6,341 hair samples; cortisone in 4,701), hair cortisone independently predicted incident cardiovascular disease, with the association concentrated in participants under 60 (OR 4.21, 95% CI 1.91–9.07 per log₁₀ unit). The scaling is difficult to translate clinically, and elevated hair glucocorticoids are not proof that psychological stress produced them [20].

Against this, a cross-sectional path analysis found that path coefficients through standard modifiable risk factors — hypertension, dyslipidaemia, 糖尿病 — accounted for much of the statistical association between hair cortisol and 冠状动脉疾病 [21]. Temporal mediation cannot be inferred from a cross-sectional design: exposure, putative mediator and outcome were measured at a single time point, so the ordering that the word mediation implies is assumed rather than observed. The result is consistent with cortisol acting through conventional risk factors, but it is equally consistent with reverse ordering or with shared upstream determinants.

8.3 Reactivity and subclinical atherosclerosis

Whitehall II also tested whether acute cortisol reactivity to standardized laboratory stressors predicted structural disease. Among 466 civil servants without known coronary disease, coronary artery 钙化 progression (Agatston increase >10) occurred in 38.2% over three years, and heightened cortisol stress reactivity independently predicted it (OR 1.27, 95% CI 1.02–1.60 per SD) after adjustment for age, sex, baseline CAC, employment grade, smoking, blood pressure, BMI, fibrinogen, and 他汀 use [22]. This is one of the few studies in the field connecting a stress-physiology measure to a structural endpoint with adequate adjustment, and it is graded accordingly.

这 MESA Stress Study produced a more mixed picture. Among 464 participants in the coronary calcium analysis and 610 in the ankle-brachial index analysis, salivary diurnal cortisol parameters showed weak or inconsistent associations with both, and the investigators described their own findings as providing only weak support for a link between cortisol and 亚临床动脉粥样硬化. In a later MESA analysis of 918 participants, 12-hour urinary cortisol was not associated with cardiovascular events overall. Higher cortisol was associated with greater coronary calcium progression in women but not men, and event associations varied by waist-to-hip ratio, with significant effect modification in women [23,24]. The inconsistency across matrices within a single cohort is itself informative and reinforces Section 6.

8.4 Genetic evidence

Two-sample Mendelian randomization using single-nucleotide polymorphisms at the SERPINA6/SERPINA1 locus on chromosome 14 — which govern corticosteroid-binding globulin and thereby morning plasma cortisol — provides an additional genetic causal-inference test.

The observational meta-analysis in the same investigation gave OR 1.18 (1.06–1.31) per SD of morning plasma cortisol for incident cardiovascular disease. The MR estimate across 122,737 coronary heart disease cases and 547,261 controls gave OR 1.06 (95% CI 0.98–1.15) per SD [25].

How to read this honestly. The genetic point estimate is directionally concordant with the observational estimate, but its confidence interval includes no effect and therefore does not distinguish a small causal effect from the null. Related work at the same locus, examining hepatic CBG expression and tissue gene expression, provides additional support for a causal contribution to ischaemic heart disease [26]. Two caveats are decisive. First, the instruments explain only about 0.5% of morning cortisol variance, which limits 统计功效 and the precision of the causal estimate; this is a matter of power rather than of instrument strength statistics, and the analysis should be described that way rather than as a “weak-instrument” result per se. Second, instrument validity is complicated by the biology of the SERPINA6/SERPINA1 locus itself: variants governing corticosteroid-binding globulin have been associated with other cardiometabolic traits, raising the possibility of horizontal pleiotropy [26]. The correct conclusion is that genetic data are compatible with a small causal contribution, and equally compatible with none.

8.5 The experiment of nature

Endogenous Cushing’s syndrome provides strong natural-experiment evidence for the cardiovascular toxicity of sustained pathological glucocorticoid excess: central 肥胖, diabetes, hypertension, dyslipidaemia, and premature death, with a standardized mortality ratio of 3.0 (95% CI 2.3–3.9; I² = 80.5%) and atherosclerotic disease with thromboembolism the leading cause of death (43.4%), ahead of infection (12.7%) and malignancy (10.6%). Notably, excess risk persists after biochemical remission — the standardized mortality ratio is 5.7 in active Cushing disease but remains elevated at 2.3 in remission — implicating cumulative prior exposure rather than concurrent hormone level — an idea structurally analogous to cumulative apoB exposure [28].

The generalization must be resisted. Cushing’s syndrome involves cortisol concentrations far outside the range produced by chronic psychological stress. It establishes that the hormone 可以 cause cardiovascular disease at sufficient dose. It does not establish that ordinary chronic stress reaches that dose, and the evidence in Sections 5.7 and 8.1 suggests it usually does not.

8.6 The null and attenuated studies

Any honest account must weight the negative results equally.

LURIC followed 3,052 patients undergoing coronary angiography for a median 9.9 years. Baseline morning serum cortisol in the highest versus lowest 四分位数 initially predicted cardiovascular mortality (HR 1.32, 95% CI 1.04–1.67). After multivariable adjustment for age, sex, BMI, smoking, hypertension, diabetes, lipids, and coronary disease severity, the association was completely abolished (HR 0.97, 95% CI 0.76–1.25) [27].

This result admits two readings, and they are not equivalent. The mediation reading holds that cortisol acts through those risk factors, so adjusting for them is over-adjustment that removes real effect. The confounding reading holds that the crude association was never independent. LURIC alone cannot distinguish them. But the practical clinical implication is similar under either reading: morning serum cortisol was not independently associated with cardiovascular mortality after multivariable adjustment in established coronary disease. Note that the study did not perform formal incremental discrimination or 重新分类 analyses, so “adds no predictive value” is a stronger statement than the data support; what they support is the absence of an independent association.

Separately, a cross-sectional study found no association between circulating plasma cortisol and coronary or peripheral arterial disease [29], reinforcing that a single-timepoint measure is a weak marker.

8.7 Verdict

Table 4. Cortisol and cardiovascular outcomes.

学习	Design / N	Cortisol measure	结果	效果 (95% CI)	等级
Whitehall II, 2011	Prospective; 4,047; 32 CV deaths	Multi-sample salivary diurnal slope	CV mortality; 6.1 y	HR 1.87 (1.32–2.64) per SD flatter; morning cortisol and CAR null	适度
KORA-F3, 2022	Prospective; 1,090; 31 CV deaths	Salivary CAR, peak:bedtime, late-night	CV mortality; ≈11 y	CAR 0.59 (0.36–0.96); peak:bedtime 0.50 (0.34–0.73); late-night 1.49 (1.13–1.97)	适度
Ronaldson et al., 2015	Prospective, CABG patients	Pre-surgical diurnal slope	MACE and death	HR 0.73 (0.56–0.96) per unit steeper	适度
InCHIANTI, 2010	Prospective; 861; 41 CV deaths	24-h urinary cortisol	CV mortality; 5.7 y	Highest tertile HR 5.00 (2.02–12.37)	Moderate, imprecise
Lifelines, 2024	Prospective; 6,341 hair samples	Hair cortisol / cortisone	Incident CVD; 5–7 y	Age <60: OR 4.21 (1.91–9.07) per log₁₀ cortisone	适度
Stomby et al., 2022	Cross-sectional path model	Hair cortisol	冠状动脉疾病	Path coefficients through standard risk factors account for much of the association; temporal ordering not observed	有限 for mediation
Hamer et al. (Whitehall II), 2012	Prospective; 466	Salivary cortisol reactivity to lab stress	CAC progression; 3 y	OR 1.27 (1.02–1.60) per SD	Moderate–Strong
Hajat et al., 2013 (MESA)	Prospective; 464 (CAC), 610 (ABI)	Salivary diurnal cortisol	CAC, ABI	Weak and inconsistent associations	适度
Flynn et al., 2023 (MESA)	Prospective; 918	12-h overnight urinary cortisol	CVD events; CAC progression	No overall CVD-event association; greater CAC progression in women; waist-to-hip-ratio interaction for events	适度
Crawford et al., 2019	Nested cohorts + meta-analysis	Morning plasma cortisol	Incident CVD	Pooled OR 1.18 (1.06–1.31) per SD	适度
Crawford et al., MR	Two-sample MR; 122,737 cases	Genetically predicted morning cortisol	CHD	OR 1.06 (0.98–1.15) per SD	Limited–Moderate; low variance explained (≈0.5%), imprecise estimate, potential locus pleiotropy
LURIC, 2021	Prospective; 3,052; median 9.9 y	Morning serum cortisol, quartiles	CV mortality	Crude HR 1.32 (1.04–1.67); adjusted 0.97 (0.76–1.25)	适度
Limumpornpetch et al., 2022	Meta-analysis of cohorts	Endogenous Cushing’s syndrome	全因死亡率	SMR 3.0 (2.3–3.9), I² = 80.5%; atherosclerotic disease and thromboembolism 43.4% of deaths; SMR 2.3 in remission vs 5.7 active	Strong for extreme exposure; not generalizable
Reynolds et al., 2009	Cross-sectional	Circulating plasma cortisol	CAD, PVD	No association	Moderate (null)

The resolution. Cortisol is a biomarker and a plausible mediator; a small causal contribution remains possible but unproven. Its role differs by exposure, population, and measure. It is a biomarker of HPA and glucocorticoid physiology — albeit a nonspecific one, influenced by circadian phase, binding 蛋白质, illness, medication, sleep, assay and matrix, and therefore not specific to psychological stress. It is plausibly a mediator, though the supporting path analysis is cross-sectional and cannot establish temporal ordering [21]. A small causal contribution remains possible on genetic evidence that is directionally concordant but statistically inconclusive, and on the Cushing’s literature, which establishes causality only at an exposure range that ordinary chronic stress does not reach.

What cortisol is 不 is the unique causal transmitter of psychological stress to the artery. The simultaneous activation of sympathetic and immune systems, the decoupling introduced by glucocorticoid receptor resistance, the tissue-level amplification by 11β-HSD1, and prospective evidence suggesting that diurnal pattern may be more informative than isolated levels together make a cortisol-centric model untenable.

Clinically, this has one clear implication, developed in Section 25: do not order cortisol testing for cardiovascular risk stratification. It remains a research tool outside of suspected Cushing’s syndrome.

9. Mechanism I: Autonomic and Haemodynamic

Chronic sympathetic overactivity with reduced vagal tone raises resting heart rate, exaggerates blood-pressure reactivity, and promotes hypertension, vasoconstriction, and 动脉僵硬度. Each of these is an established atherogenic input, and hypertension in particular is a proven causal risk factor with an unambiguous randomized-trial evidence base of its own.

This is the pathway with the best combination of plausibility and human support, for a reason that is easy to overlook: the terminal step is already proven. One does not need to demonstrate that hypertension causes atherosclerosis; that is known. What remains to be demonstrated is that stress meaningfully contributes to blood-pressure burden. Prospective work linking cortisol responses to mental stress with incident hypertension supports the first half of that chain [74]. The argument is not complete on those terms alone: establishing a stress → hypertension → ASCVD causal pathway additionally requires evidence that stress alters sustained blood-pressure burden and that this pathway mediates a specifiable share of subsequent disease. Neither has been quantified.

Repeated pressor and flow changes also increase mechanical loading of the arterial wall and may alter local haemodynamic stresses at atherosclerosis-prone sites such as bifurcations, where local shear patterns are determined substantially by arterial geometry. Higher circulating norepinephrine and epinephrine concentrations are associated with higher cardiovascular risk in pooled observational data (norepinephrine RR 1.68, 1.37–2.06; epinephrine RR 1.58, 1.10–2.26), though circulating catecholamine concentration is an imperfect proxy for regional sympathetic nerve activity [10].

Grade: Moderate–Strong.

10. Mechanism II: Inflammation and Immunity

This is the mechanistically richest pathway in the human literature, and the one where translational evidence is strongest.

10.1 Acute stress produces measurable immune signalling

In a systematic review and meta-analysis of laboratory psychological stress, acute stressors increased IL-1β (d ≈ 0.66), 白介素-6 (d ≈ 0.35), IL-10 (d ≈ 0.69), and TNF-α (d ≈ 0.28), with IL-6 responses growing larger at later sampling points consistent with the known kinetics of cytokine production [13].

This matters more than its effect sizes suggest. Controlled laboratory paradigms use within-person pre–post contrasts, which removes many stable between-person confounders and shows that acute psychological stress can induce measurable inflammatory change in humans [13]. It therefore argues against explaining stress–inflammation associations entirely by lifestyle differences between people — though it does not abolish confounding, since baseline characteristics can still modify both the resting state and the magnitude of response, and the included studies are not uniformly randomized.

10.2 Glucocorticoid receptor resistance

As developed in Section 5.7, chronically stressed humans can show reduced glucocorticoid-responsive transcription and enhanced NF-κB-related transcription in monocytes without any obvious excess of daily cortisol secretion. This provides one plausible human mechanistic explanation for the anti-inflammatory-hormone paradox [14,15].

Grade: Moderate for the mechanism; Limited for its quantified contribution to ASCVD outcomes. No study has shown how much plaque is attributable to glucocorticoid resistance.

10.3 The amygdala–marrow–artery axis

A particularly influential integrative human study imaged 293 patients with ¹⁸F-FDG PET/CT and followed them for a median of 3.7 years, during which 22 had a cardiovascular event.

Resting amygdalar metabolic activity was associated with bone-marrow activity (r = 0.47, p < 0.0001), arterial inflammation (r = 0.49, p < 0.0001), and cardiovascular events (standardized HR 1.59, 95% CI 1.27–1.98, p < 0.0001), and remained significant after multivariable adjustment. Serial mediation analysis was consistent with an amygdala → bone marrow → arterial inflammation → event pathway: bone-marrow activity accounted for a reported 46% of the amygdala–arterial inflammation relationship and arterial inflammation for 39% of the amygdala–event relationship. These are observational mediation estimates derived from 22 events and should be read as hypothesis-generating rather than as a precise causal decomposition.

A separate cross-sectional psychometric substudy found perceived stress associated with amygdalar activity (r = 0.56, p = 0.049), arterial inflammation (r = 0.59, p = 0.035), and CRP (r = 0.83, p = 0.021) [11].

Three caveats are essential, and they are frequently omitted. First, the psychometric substudy had n = 13; correlations of 0.83 in a sample of thirteen are not stable estimates. Second, the longitudinal cohort had 22 events — enough for a hazard ratio, not enough for confident mediation decomposition. Third, the study is observational; stress was never manipulated. It is a single, well-conducted, blinded-adjudication study establishing an important mechanistic association, and it requires replication.

The translational complement comes from work showing monocytosis and neutrophilia in chronically stressed medical residents, coupled with murine experiments in which chronic stress activated β3-adrenergic signalling in the bone-marrow niche, reduced CXCL12, stimulated haematopoietic progenitor proliferation, and accelerated inflammatory plaque features. The detailed β3/CXCL12 causal chain is animal evidence. The human contribution is the upstream observation of stress-related leukocytosis. These should not be reported as a single seamless finding [12].

Grade: Moderate.

10.4 The NLRP3 qualification

NLRP3/IL-1β biology is clearly relevant to atherosclerosis, and higher NLRP3 expression in subcutaneous adipose tissue correlates with angiographic coronary atherosclerosis severity in humans, in a study that did not assess psychological stress [44]. However, the specific chain psychological stress → NLRP3 activation → human coronary events has not been demonstrated with anything like the strength of stress → IL-6/CRP or stress → 内皮功能障碍. Direct stress–NLRP3 experiments remain disproportionately animal or cellular.

Calling NLRP3 an established human mediator of stress-induced ASCVD would exceed the evidence. Grade: Limited for stress-specific mediation.

10.5 Is inflammation itself a modifiable causal node?

This question can be answered with randomized evidence, which is unusual in this review. A 2025 systematic review and meta-analysis pooling 37,056 individuals across 32 randomized trials found that anti-inflammatory therapies targeting the NLRP3/IL-1β/IL-6/CRP pathway reduced myocardial infarction (RR 0.85, 95% CI 0.78–0.93) and coronary 血运重建 (RR 0.80, 0.74–0.86), but did 不 significantly reduce overall 主要不良心血管事件, stroke, or mortality [43].

The interpretation is precise: these randomized data support selected inflammatory pathways as modifiable causal contributors to some coronary outcomes. Pharmacologic interruption of this pathway reduces some outcomes and not others, which supports the causal relevance of inflammation to selected coronary endpoints. It does 不, however, quantify how much stress-related risk is mediated by inflammation, and it should not be read as a ceiling on what an upstream behavioural intervention could achieve: such an intervention could act simultaneously through blood pressure, autonomic regulation, sleep, behaviour, metabolism and endothelial function, and differs from a targeted drug in specificity, 坚持 and off-target profile.

11. Mechanism III: Endothelial and Vascular

11.1 Acute mental stress impairs endothelial function

In landmark human experimental work, acute mental stress tasks — competitive mental arithmetic, public speaking — induced transient endothelial dysfunction in healthy individuals, with brachial artery flow-mediated dilation falling significantly from baseline at 30 and 90 minutes post-stress and requiring up to four hours to recover [31]. This has been demonstrated experimentally in more than one human laboratory, using different stressors and vascular beds [31,76], and constitutes strong experimental human physiology.

11.2 It has prognostic value

The more important finding is that this response predicts outcomes. In 569 patients with coronary artery disease, mean FMD fell from 4.8% before a public-speaking stress task to 3.9% afterward; 63.3% developed stress-induced endothelial dysfunction. Over a median 3.0 years, 74 major adverse cardiovascular events occurred, and stress-induced endothelial dysfunction was associated with subdistribution HR 1.78 (95% CI 1.15–2.76). Each one-percentage-point stress-related decrease in FMD carried sHR 1.15 (1.03–1.27) [32].

A separate ischaemic heart disease cohort found that each SD worse stress-induced endothelial response predicted MACE with HR 1.35 (1.07–1.71), with impaired microvascular stress responses appearing especially prognostic in women — suggesting the cardiovascular consequences of stress may differ by sex and vascular compartment [33].

Grade: Moderate–Strong. This is a reproducible human experimental effect with demonstrated prognostic value in the population of clinical interest.

11.3 Molecular mechanism: what is human and what is not

A largely translational molecular model proposes the following: sympathetic activation stimulates the renin–angiotensin–aldosterone system and upregulates angiotensin II type 1 receptor (AT₁R) signalling in 内皮细胞; AT₁R activation stimulates NADPH oxidase, generating superoxide; superoxide reacts with 一氧化氮 形成 过氧亚硝酸盐; peroxynitrite oxidizes the eNOS cofactor tetrahydrobiopterin, causing eNOS uncoupling, whereupon the enzyme produces superoxide rather than nitric oxide in a self-sustaining cycle. Reduced NO bioavailability then permits NF-κB-driven expression of 血管细胞黏附分子-1, 细胞间黏附分子-1, and E-selectin, which capture circulating monocytes for transmigration into the 内皮下间隙.

Downstream of this, a further set of mechanisms — disruption of endothelial tight-junction proteins (ZO-1, claudin-5, occludin), MMP-9 upregulation with TIMP-1 suppression, VEGF- and angiogenin-driven intra-plaque neovascularization, and TNF-α-mediated impairment of 内皮祖细胞 — has been proposed. Each of these steps derives from animal or cell-system work rather than from human plaque, and none is individually sourced to a human study here. They are listed as candidate mechanisms, not as findings.

A necessary demarcation, and it matters more than it may appear. Two components of this account do have direct human experimental support. Selective endothelin-A receptor antagonism with intra-arterial BQ-123 prevented the prolonged endothelial dysfunction that followed a mental stress task in healthy subjects, implicating endothelin-A signalling causally in the human response [76]. And in a three-way randomized crossover study of 15 overweight or obese men, AT₁R blockade with olmesartan and ascorbic acid infusion each altered the stress-induced fall in flow-mediated dilation and the accompanying redox profile, implicating AT₁R-mediated redox imbalance [77]. Both are small studies, and the second is confined to one sex and body-composition stratum.

The remainder of the cascade — the NADPH-oxidase/peroxynitrite/tetrahydrobiopterin/eNOS-uncoupling sequence and the downstream adhesion-molecule and plaque steps — remains substantially translational. It should not be cited to the demonstration that endothelial glucocorticoid receptor signalling suppresses atherogenesis, which is an ApoE-deficient mouse experiment [45] and speaks to endothelial GR biology rather than to this pathway. The tight-junction, MMP-9/TIMP-1, VEGF/angiogenin, and endothelial progenitor components derive predominantly from animal models, cell systems, and mechanistic reviews. They are biologically coherent and worth stating — but they are hypotheses about human plaque, not observations of it, and this review declines to present them at the same evidentiary level as the FMD data in Section 11.2. Related work showing that acute mental stress rapidly redistributes leukocytes in humans, with locally derived norepinephrine increasing endothelial adhesion signalling and plaque leukocyte recruitment in mice, illustrates the same split precisely [34].

Grade: Moderate for stress-induced endothelial dysfunction in humans; Limited for the detailed downstream molecular cascade in human plaque.

12. Mechanism IV: Metabolic

Glucocorticoids increase hepatic 葡萄糖 production, antagonize peripheral insulin action, and contribute to visceral adiposity through tissue-specific glucocorticoid signalling [30]. These are established features of glucocorticoid physiology; none has been demonstrated to mediate a quantified share of stress-related ASCVD.

In community populations, altered diurnal cortisol — flatter slopes, higher bedtime cortisol — is associated with type 2 diabetes, and long-term hair glucocorticoid measures correlate with cardiometabolic traits [38].

12.1 The lipid pathway requires particular caution

It is sometimes asserted that stress raises 低密度脂蛋白胆固醇 and thereby causes atherosclerosis. This is too simple, and the evidence does not support it.

ApoB-containing 脂蛋白 remain the substrate that atherosclerotic plaque formation requires, a position supported by convergent genetic, epidemiologic and interventional evidence [75]. Psychological stress may alter diet, adiposity, 胰岛素敏感性, and hepatic metabolism, and hypercortisolaemia with elevated free fatty acid flux can plausibly increase hepatic 极低密度脂蛋白 and apoB secretion while lowering 高密度脂蛋白 cholesterol. But studies of job strain have not consistently found large adverse changes in 总胆固醇, 低密度脂蛋白 cholesterol, or fibrinogen [39].

The defensible model is that stress interacts with pre-existing lipoprotein exposure rather than creating it. A person with low lifetime apoB exposure and high chronic stress is in a materially different position from a person with high lifetime apoB exposure and high chronic stress, and the literature gives no reason to think stress substitutes for the lipoprotein term.

12.2 Separating hormone from behaviour

The deeper problem with the metabolic literature is that in everyday chronic stress, behavioural pathways — diet, inactivity, 酒精, and sleep loss — may contribute importantly to the metabolic effects observed. Distinguishing direct glucocorticoid effects from behavioural ones in observational human data is extremely difficult, and few studies have attempted it rigorously.

Grade: Moderate overall; Limited for a stress → cortisol → dyslipidaemia → atherosclerosis pathway specifically.

13. Mechanism V: Thrombosis and Acute Triggering

These mechanisms are distinct from long-term atherogenesis and should be reported separately (Figure 4).

Acute psychological stress produces measurable haemostatic change. A systematic review found relatively consistent stress-related alterations in coagulation, 纤维蛋白溶解, haemorheology, and platelet measures, though with considerable methodological weakness. Controlled norepinephrine infusion increases factor VIII activity, fibrinogen, and D-dimer, providing direct evidence that adrenergic signalling alters coagulation [35].

The proposed components are: epinephrine binding platelet α₂-adrenergic receptors, lowering cyclic AMP and the aggregation threshold, with P-selectin and activated glycoprotein IIb/IIIa rising within minutes; elevated factor VIII, von Willebrand factor, and fibrinogen; increased tissue factor expression; and elevated 纤溶酶原 activator inhibitor-1 impairing endogenous fibrinolysis [35].

An important nuance is usually lost. In healthy subjects, coagulation and fibrinolysis often rise together, preserving haemostatic balance. It is in people with vascular disease or other vulnerabilities that a relative prothrombotic imbalance becomes consequential. The acute-stress prothrombotic state is therefore best understood as a modifier of existing risk, not a universal hazard.

Mental stress-induced myocardial ischaemia (MSIMI) roughly doubles the risk of adverse cardiac events and mortality in patients with coronary artery disease. The supporting meta-analysis rested on only five small studies with fewer than 50 events each — a thin base for a widely cited conclusion [36]. Mechanistically, in diseased 冠状动脉 with endothelial dysfunction, catecholamine release fails to produce normal flow-mediated vasodilation; direct α-adrenergic stimulation of vascular smooth muscle instead produces paradoxical vasoconstriction.

Grade: Moderate–Strong for acute triggering; Moderate for the specific haemostatic mechanisms.

14. Mechanism VI: Behavioural Mediation

Multiple behavioural pathways plausibly contribute to the stress–ASCVD relationship: smoking, alcohol, diet quality, physical inactivity, sleep disruption, medication non-adherence, obesity, and co-occurring depression or anxiety. The share attributable to them is not established.

The standard analytic response is to adjust for these. Many prospective studies do so and find attenuated but still significant associations. This is often described as showing a “partly independent” effect, but the accurate statement is narrower: associations persist after adjustment for measured covariates. Statistical adjustment does not establish biological independence.

But adjustment cuts both ways, and this is under-appreciated. Chronic distress can contribute to smoking, inactivity, sleep disturbance and non-adherence, while these factors can also share upstream determinants with distress and can feed back on it. Depending on the causal structure, a given behaviour may act as a mediator, a confounder, a consequence of a common cause, or some combination. So:

• A minimally adjusted estimate retains pathways operating through behaviour, but also retains more confounding.
• A covariate-adjusted estimate conditions on measured variables and attenuates behaviourally mediated pathways — but conditioning on post-exposure variables can also induce collider bias, and it does 不 by itself identify a controlled or natural direct effect.
• Neither is “the” answer, and reporting only one is misleading.

This is a point of formal causal inference, not a stylistic preference. Estimating how much of the stress–ASCVD relationship is transmitted biologically rather than behaviourally requires a specified causal estimand and mediation analysis under explicit identification assumptions — no-unmeasured-confounding of exposure–outcome, mediator–outcome, and exposure–mediator relationships, and correct measurement and temporal ordering. Ordinary multivariable covariate adjustment alone does not establish these identification assumptions and does not by itself identify a controlled or natural direct effect.

This resolves an apparent conflict in the literature. The cross-sectional path analysis showing that standard risk factors mediate most of the hair-cortisol–CAD association and the INTERHEART finding that the psychosocial–MI association persists after adjustment for income and education are not contradictory. They are estimating different quantities.

There is no defensible single percentage for how much of stress-related ASCVD is behavioural versus direct. The answer varies by stressor, population, life stage, baseline disease, and mediator definition. Any review that supplies a specific figure is over-claiming.

The 2025 AHA scientific statement on post-myocardial-infarction psychological distress explicitly identifies these behavioural routes as credible pathways by which distress worsens prognosis [48] — which is also, usefully, where the clinical leverage lies (Section 21).

Grade: Moderate–Strong.

15. Mechanistic Evidence Matrix

Table 5. Mechanisms linking psychological stress to ASCVD, graded by human evidence.

机制	Proposed pathway	Human evidence	Consistency	等级
HPA-axis dysregulation	Threat circuitry → CRH/ACTH → altered cortisol level, timing, recovery	Prospective links to diurnal slope, late-night cortisol, urinary and hair glucocorticoids	Direction depends on metric; several cohorts associate flatter or less differentiated rhythms with adverse outcomes, but cardiovascular findings are not uniformly consistent in meta-analysis	适度
Glucocorticoid receptor resistance	Repeated stress → reduced immune-cell GR signalling → failure to suppress NF-κB	Human monocyte transcriptional data in chronically stressed caregivers	Biologically coherent; few outcome-linked studies	适度 mechanism; 有限 outcome mediation
SAM / catecholamine activation	Stress → sympathetic output → heart rate, blood pressure, vasoconstriction, immune effects	Strong acute physiology; hormone–CVD meta-analysis; controlled norepinephrine infusion	Generally consistent	Moderate–Strong
Hypertension / haemodynamic load	Recurrent pressor and flow changes → mechanical wall stress and sustained BP burden; possible alteration of local haemodynamic forces	Cortisol responses to mental stress predict incident hypertension [74]; terminal step independently proven causal	Consistent in direction	Moderate–Strong
Cytokine inflammation	Stress → adrenergic/HPA/immune signalling → IL-6, IL-1β, TNF-α, CRP	Experimental acute-stress meta-analysis; chronic-stress genomic studies	Strongest for IL-6; magnitude varies	强大 for acute activation; 适度 for ASCVD mediation
Amygdala–marrow–artery axis	Central threat response → sympathetic marrow signalling → leukopoiesis → arterial inflammation	One PET/CT cohort (n = 293, 22 events) with serial mediation; animal β3/CXCL12 chain	Highly coherent translationally; single human study	适度
NLRP3 / IL-1β	Stress → inflammasome → IL-1β, IL-18	NLRP3 linked to human atherosclerosis; stress-specific chain mostly animal/cellular	Stress-specific human evidence sparse	有限 for stress-specific mediation
内皮功能障碍	Stress → neurohumoral and redox signalling → impaired endothelial NO-dependent function / FMD	Reproducible experimental FMD impairment; prognostic for MACE in CAD	Reproduced clinically	Moderate–Strong
Downstream endothelial molecular cascade	Tight-junction loss, MMP-9/TIMP-1 imbalance, VEGF/angiogenin neovascularization, EPC suppression	Predominantly animal and cell-system; mechanistic reviews	Coherent but unverified in human plaque	有限
Insulin resistance / visceral adiposity	Cortisol + behaviour + sleep loss → insulin resistance, central adiposity, diabetes	Strong endocrine plausibility; diurnal cortisol–type 2 diabetes association	Relation plausible; mediator-specific estimates lacking	适度
ApoB / lipoprotein pathway	Stress-related metabolic change → altered atherogenic lipoproteins	Inconsistent direct stress–lipid association in occupational cohorts	Not consistently adverse	有限
Coagulation / 血小板	Acute stress → factor VIII, fibrinogen, vWF, platelet activity, impaired fibrinolysis	Laboratory studies and systematic review; NE infusion experiments	Broadly procoagulant but heterogeneous	适度
Acute event triggering	Stress surge → demand, vasoconstriction, endothelial and thrombotic shift → ischaemia	Large international case-crossover data; MSIMI cohorts	Strong temporal signal	Moderate–Strong
Behavioural mediation	Stress → smoking, diet, inactivity, sleep, alcohol, non-adherence → CVD	Large observational base; consensus statements	Repeatedly observed	Moderate–Strong

Figure 6 renders this matrix as a pathway diagram with per-link weighting. The shape of the evidence is worth noting explicitly: the chain is strongest at its two ends and weakest in the middle. That stress acutely activates these systems is well established in human laboratory physiology; that stress predicts events is well established epidemiologically. The quantitative contribution of each intermediate step to 斑块负荷 in humans is not.

Figure 6. Integrated pathway from psychological stress to atherosclerotic events, with per-link human evidence weight. Labels grade the strength of human evidence for each arrow, not the size of the effect.

16. The Intervention Question: Taxonomy and the Claim Ladder

16.1 Meditation is not one treatment

Analysing “meditation” as a single intervention treats a clinically and methodologically heterogeneous class as though it were one treatment.

• Mindfulness-based interventions (MBIs), including MBSR, combine attentional training, acceptance and non-reactivity, body awareness, and frequently movement, yoga, and explicit behavioural instruction. MBSR is conventionally an eight-week instructor-led programme with substantial home practice.
• Mindfulness-based cognitive therapy (MBCT) adds cognitive-behavioural technique.
• Transcendental Meditation (TM) is a standardized mantra-based practice, typically 15–20 minutes twice daily, taught through an organized certification structure.
• Focused-attention and paced-breathing practices concentrate on a physiological anchor, such as breathing at approximately six breaths per minute.

The heterogeneity is not cosmetic. The Mindfulness-Based Blood Pressure Reduction (MB-BP) programme explicitly teaches DASH-pattern diet, physical activity, medication adherence, and alcohol reduction alongside meditation. A trial of MB-BP is a trial of a multicomponent behavioural intervention, and attributing its results to meditation specifically is unwarranted [60].

Focused-attention meditation deserves separate mention: cardiovascular trials generally bundle attentional practices into broader mindfulness programmes rather than testing a standardized focused-attention intervention in isolation. There is therefore insufficient evidence to rank focused-attention meditation against MBSR or TM for ASCVD endpoints.

16.2 The comparator problem

An inactive control — waitlist, no treatment, usual care — does not isolate meditation. It bundles instructor attention, expectancy, social contact, time spent sitting quietly, repeated measurement, and the structure of a programme. A waitlist-controlled trial therefore does not isolate the practice-specific effect; it estimates the effect of the whole package and its expectancy context, and inactive comparators tend to yield larger estimates in this literature.

Figure 7 shows what this does to the numbers. Comparator choice materially influences the estimated effect — though intervention type, trial size, adherence, population, background therapy, measurement method and heterogeneity all contribute as well.

Figure 7. In meditation trials, the choice of comparator materially influences the answer. The only moderate-certainty systolic blood-pressure estimate in the Cochrane review — Transcendental Meditation against an active comparator — is also the smallest.

16.3 Five separable claims

Because these are routinely conflated, they are separated throughout (Figure 8):

• Claim A: meditation reduces perceived stress.
• Claim B: meditation changes stress physiology (cortisol, HPA function).
• Claim C: meditation improves established risk factors (blood pressure, HRV, inflammation, metabolic markers).
• 主张D： meditation slows measurable atherosclerosis.
• Claim E: meditation reduces hard events — myocardial infarction, stroke, cardiovascular death.

Evidence weakens sharply from A to E, and E cannot be inferred from A through D.

Figure 8. Five separable claims about meditation, and the evidence for each. An intervention can reliably reduce how stressed a person feels, modestly lower their blood pressure, and still have no demonstrated effect on whether they have a heart attack.

17. The Benchmark Synthesis

这 2024 Cochrane review is the reference standard for this question. It included 81 randomized trials and 6,971 participants, requiring interventions of at least 12 weeks in adults at high cardiovascular risk or with established cardiovascular disease, and examined four prespecified comparisons. Most trials were at unclear risk of bias, many were small, and heterogeneity was substantial across most outcomes.

Table 6. Cochrane 2024 — findings by comparison.

Comparison	血压	Psychological outcomes	Clinical events	确定性
MBIs vs active comparators (29 RCTs, 2,883 participants)	SBP MD −6.08 mmHg (−12.79 to 0.63), I² = 88%; DBP −5.18 (−10.65 to 0.29), I² = 91%; 6 trials, 388 participants	Perceived stress SMD −0.24 (−0.45 to −0.03), I² = 0%, 6 trials, 357 participants; anxiety SMD −0.06 (−0.25 to 0.13), I² = 0%; depression ≈ null; wellbeing little or no effect	None reported. Smoking cessation RR 1.45 (0.78–2.68), I² = 79%	Low for BP; moderate for perceived stress, anxiety, depression
MBIs vs non-active comparators (38 RCTs, 2,905 participants)	SBP MD −6.62 mmHg (−13.15 to −0.10), I² = 87%; DBP −3.35 (−5.86 to −0.85), I² = 61%; 9 trials, 379 participants	Larger effects than against active controls	One trial (110 participants): RR 0.94 (0.37–2.42)	Low for BP; very low for events
TM vs active comparators (8 RCTs, 830 participants; SBP analysis 8 RCTs, 774 participants)	SBP MD −2.33 mmHg (−3.99 to −0.68), I² = 2% — TM probably reduces SBP; DBP less certain	—	One trial (201 participants): RR 0.91 (0.56–1.49)	Moderate for SBP; 低 for events
TM vs non-active comparators (2 RCTs, 186 participants)	SBP MD −6.34 mmHg (−9.86 to −2.81), I² = 0%; DBP −5.13 (−9.07 to −1.19), I² = 18%; 2 trials, 139 participants	One trial (112 participants): anxiety SMD −0.71 (−1.09 to −0.32); depression −0.48 (−0.86 to −0.11)	None reported. No adverse events or smoking data	Low for SBP; very low for DBP

Two observations dominate. First, of 81 randomized trials, exactly two contributed cardiovascular clinical-event data — one at very low certainty (MBIs vs inactive, RR 0.94, 0.37–2.42) and one at low certainty (TM vs active, RR 0.91, 0.56–1.49) — and neither showed a detectable effect. Because the Cochrane search closed in November 2021, this characterises the trials it included rather than the entire literature to the present day. Second, the only moderate-certainty blood-pressure estimate in the entire review is also the smallest: −2.33 mmHg for TM against an active comparator, with essentially no heterogeneity (I² = 2%) and the largest participant total — 774 of the 830 participants in that comparison contributed systolic data. Every estimate near −6 mmHg carries either I² ≥ 87% or fewer than 150 participants.

The Cochrane authors’ own summary is that they found very little information on clinical endpoints, limited information on blood pressure and psychological outcomes, substantial between-study heterogeneity, and a body of evidence generally of low certainty [50].

A note on a common citation error. The −2.33 mmHg figure is frequently attributed to mindfulness-based interventions, or to a non-active comparison. It belongs to TM versus active comparators. The distinction matters: the figure is often deployed to argue that mindfulness produces meaningful blood-pressure reduction, when it in fact represents the most rigorously controlled — and smallest — estimate in the review.

18. Claim A: Perceived Stress

Mindfulness-based interventions probably produce a small reduction in perceived stress against active comparators, in the cardiovascular-risk populations Cochrane included. Against active comparators, the Cochrane pooled estimate was SMD −0.24 (−0.45 to −0.03; I² = 0%; 6 trials, 357 participants; moderate certainty) — a small effect with no detected statistical heterogeneity (I² = 0%), across six trials. Against inactive comparators, effects are larger.

Two qualifications. First, pooled effects on anxiety and depression against active comparators were approximately null (anxiety SMD −0.06, I² = 0%, moderate certainty). Reducing perceived stress is not the same as treating an anxiety or depressive disorder, and this literature does not support the latter claim in cardiovascular populations [50]. Second, the difference between active-controlled and waitlist-controlled effects is itself the finding: the smaller effect against active comparators suggests that non-specific intervention and context effects contribute meaningfully to estimates obtained in uncontrolled or waitlist designs.

Grade: Moderate.

19. Claim B: Cortisol and HPA Physiology

The strongest aggregate evidence comes from a systematic review and meta-analysis of 58 randomized trials with 3,508 participants (Figure 9A). Across blood, salivary, and hair matrices, psychological stress-management interventions changed cortisol with a pooled Hedges’ g of 0.282. That pooled figure covers stress-management interventions generally, not meditation specifically; the mindfulness/meditation subgroup estimate was g = 0.345. Effects were larger for CAR measures (g = 0.644) than diurnal cortisol measures (g = 0.255). By modality, mindfulness and meditation (g = 0.345) and relaxation (g = 0.347) were most effective, while mind–body (g = 0.129) and talking therapies (g = 0.107) were non-significant [51].

Figure 9. Meditation alters stress biology modestly and inconsistently. ‘Modulation’ is a more accurate description than ‘cortisol lowering’; no study has shown that a meditation-induced cortisol change mediates a reduction in cardiovascular events.

Complementary evidence: across 10 randomized trials using blood sampling, meditation had a medium effect on cortisol (g = 0.62, 95% CI 0.22–1.02, p = 0.003) — but only in at-risk or somatically ill samples, not in healthy participants without risk factors [52]. A meta-analysis of MBIs on salivary cortisol in healthy adults (5 RCTs, n = 190) found smaller and less consistent effects [53].

Three interpretive points.

First, “modulation” is safer than either “lowering” or “normalisation.” Depending on baseline phenotype, a favourable change could mean a stronger CAR, a steeper diurnal decline, lower evening cortisol, or altered reactivity — potentially in opposite directions in different people. Reporting a pooled standardized effect as “meditation lowers cortisol” misdescribes what was measured; but neither do these pooled effects demonstrate movement toward a validated healthy physiological reference, which is what “normalisation” would assert.

Second, the observation that active-controlled trials in that meta-analysis did 不 show weaker cortisol effects than passive-controlled trials argues against explaining the entire effect by no-treatment expectancy. It does not identify a meditation-specific mechanism, since active controls differ widely in credibility and intensity.

Third, and decisively: no study has demonstrated that a meditation-induced cortisol change mediates any cardiovascular benefit. Observing a cortisol decrease and a blood-pressure decrease in the same trial does not establish that the former caused the latter, let alone that either prevented atherosclerosis.

Grade: Limited–Moderate. Some evidence suggests larger effects in at-risk or somatically ill samples, with the clearest signal in pooled blood-cortisol studies and weaker salivary findings.

20. Claim C: Established Risk Factors

20.1 Blood pressure — the best-supported surrogate

The Cochrane estimates are in Table 6. Older TM-specific meta-analyses estimated larger reductions of approximately −4.7/−3.2 mmHg, and a separate TM meta-analysis of 12 studies and 996 participants reported −4.26 mmHg systolic and −2.33 mmHg diastolic [62,63]. The AHA statement on alternative approaches to blood-pressure lowering rated TM only Class IIb, Level of Evidence B, explicitly citing study-quality concerns [49], and the same statement noted that effect estimates in this literature vary with study quality.

Individual trials illustrate why pooling is contentious:

• MBSR versus progressive muscle relaxation in 56 adults with prehypertension: clinic SBP fell 4.8 mmHg versus 0.7 mmHg (p = 0.016) and DBP fell 1.9 mmHg versus a 1.2 mmHg rise (p = 0.008) — but ambulatory blood pressure did not differ. [58]
• HARMONY: 101 adults with untreated stage-1 hypertension showed essentially no between-group difference in 24-hour ambulatory blood pressure at 12 weeks (+0.4/0.0 versus +0.4/−0.4 mmHg) [59].
• MB-BP, a mindfulness-based multicomponent blood-pressure programme rather than a meditation-only trial: 201 participants with elevated office blood pressure showed a prespecified between-group difference in office systolic pressure of −4.5 mmHg (95% CI −9.0 to −0.1) at six months versus enhanced usual care. But the intervention also targeted diet, physical activity, medication adherence, alcohol, and stress; sedentary time fell by 350.8 minutes per week. Improved health behaviour is one plausible contributor, but the multicomponent design prevents attributing the blood-pressure effect to meditation alone, and co-occurrence of behavioural and blood-pressure change does not by itself establish mediation. A 2026 secondary analysis found a small improvement in composite cardiovascular health (SMD 0.144, 0.023–0.266) while individual component confidence intervals generally crossed the null. The lead investigator disclosed ownership of a company providing mindfulness training, with preregistration, restricted data access, and independent statistical analysis as declared safeguards [60,61].

In several of these trials, clinic blood pressure moved while ambulatory blood pressure did not. This discrepancy warrants caution, though it is drawn from a small number of trials and should not be generalized to the whole literature. Ambulatory measurement is less susceptible to office and white-coat effects and is generally more prognostically informative.

Grade: Moderate. A realistic expectation differs by modality and comparator rather than resolving to one number: approximately −2.3 mmHg systolic for TM against an active comparator (moderate certainty); larger but imprecise and highly heterogeneous estimates for MBIs (low certainty, confidence intervals crossing or near the null against active comparators); and approximately −4.5 mmHg office systolic for the multicomponent MB-BP programme, which cannot be attributed to meditation alone.

20.2 Heart-rate variability

Popular accounts assert that meditation “restores vagal tone.” The randomized evidence does not support this as a consistent chronic effect. A meta-analysis of 19 RCTs found Hedges’ g 0.38 (95% CI −0.014 to 0.77; I² = 89.1%) for resting vagally mediated HRV — not statistically convincing, and extraordinarily heterogeneous. Removing a single extreme outlier reduced the estimate to 0.19 (−0.02 to 0.39) [57].

Grade: Inconclusive.

20.3 Inflammation

The picture is mixed and, importantly, level-dependent (Figure 9B).

Pooled across 48 randomized trials with 4,683 participants, MBIs produced small reductions in CRP at post-treatment (standardized mean change difference −0.14, −0.26 to −0.01) and IL-6 (−0.35, −0.67 to −0.03). At follow-up, CRP remained reduced (−0.39, −0.68 to −0.10) while the IL-6 interval included no effect (−0.13, −0.29 to 0.03) [54].

But a relatively large, matched-active-control randomized trial complicates this. In 190 lonely older adults, MBSR versus an active health-enhancement programme reduced pro-inflammatory NF-κB-related gene expression (d = 0.17, p = 0.028) but did 不 reduce circulating IL-6 or CRP, nor alter CREB, IRF, or glucocorticoid-receptor transcriptional activity. An earlier positive gene-expression trial had only n = 40 [55,56].

The lesson is specific and generalizable: transcriptional signalling and systemic protein biomarkers do not necessarily move together. Demonstrating a change at one level of the inflammatory cascade is not evidence of change at another, and neither is evidence of change in plaque.

Grade: Limited–Moderate.

20.4 Metabolic outcomes

Small improvements in glucose, lipids, and metabolic-syndrome components appear in some trials but are underpowered and inconsistent. A meta-analysis in diabetes found psychological benefits from MBSR but no detectable pooled HbA1c improvement at post-intervention or follow-up [64]. Cochrane found no significant lipid or glycaemic changes for MBIs against active controls.

Behavioural change driven by these programmes may still be cardiovascularly useful — but that is a behavioural mechanism, not a demonstration of a cortisol → insulin-sensitivity pathway.

Grade: Limited.

21. Claim D: Atherosclerosis

Direct randomized evidence is sparse and inconsistent.

The most-cited trial randomized 138 hypertensive African American adults to TM versus health education. Only about 60 of the 138 randomized participants contributed paired carotid 内膜中膜厚度 data (roughly 43.5% retention), and among those completers cIMT regressed in the TM group (−0.098 mm) while progressing in controls (+0.054 mm; p = 0.038) [65]. Attrition of this magnitude creates a high risk of attrition bias, and the finding is best regarded as preliminary.

A randomized trial conducted from 2000 to 2014 and published in 2025, in a different population and with a different follow-up structure, did not demonstrate a between-group cIMT benefit. Of 197 randomized participants, 136 completed post-test cIMT. After one year, cIMT changed by −0.0004 mm in the TM group and −0.0003 mm in health education — no significant difference. There were also no differences in lipids or blood pressure at one year. The authors noted that both groups progressed less than historical non-randomized controls, but that comparison is not randomized and cannot support a causal inference about either arm [67].

Beyond these, small mind-body studies have evaluated flow-mediated dilation or endothelial biomarkers, but none establishes that meditation changes coronary 斑块体积, carotid plaque, coronary artery calcium progression, or plaque composition.

The hierarchy matters here more than anywhere. Improving FMD is not slowing plaque. Slowing a surrogate is not preventing myocardial infarction.

Grade: Limited / Inconclusive — with the later trial’s failure to demonstrate a between-group cIMT benefit carrying substantial weight. Because populations, designs and follow-up differ, this is a failure to demonstrate benefit in a related trial rather than a strict replication attempt.

22. Claim E: Hard Clinical Events

Two small randomized trials, both from the same TM-affiliated research group at institutions linked to the organization that teaches and licenses the technique, report hard-endpoint benefit.

• 201 Black patients with coronary heart disease randomized to TM versus health education. Over a mean 5.4 years, TM was associated with a 48% reduction in the composite primary endpoint of all-cause mortality, myocardial infarction, and stroke: HR 0.52 (95% CI 0.29–0.92; p = 0.025), alongside a systolic blood-pressure reduction of approximately 4.9 mmHg [66].
• The trial published in 2025 had a null primary endpoint. This trial was conducted from 2000 to 2014 and published in 2025 after delayed analysis; it is not a newly recruited contemporary trial, and its registration (NCT05642936) post-dates data collection by many years. Its primary endpoint, carotid intima-media thickness, showed no between-group difference. The secondary event analysis reported a 65% relative MACE reduction at five years (HR 0.346, 95% CI 0.134–0.893; p = 0.017), with the 14-year analysis null (HR 0.68, 0.35–1.31) [67].
• Three features require explicit flagging. First, events were analysed across multiple maximum-follow-up windows (1, 5, 10 and 14 years) without a stated multiplicity adjustment; the five-year window is best treated as the main secondary analysis and the others as exploratory. Second, event counts are small throughout. Third, the reported ten-year result is internally inconsistent as published: the paper gives HR 0.485 with a 95% confidence interval of 0.226–1.044 alongside p = 0.0435. An interval that includes 1.0 cannot accompany a p-value below 0.05. Until that discrepancy is resolved by the authors, the ten-year result should not be described as statistically significant, and this review does not do so.

These findings must be weighed carefully rather than dismissed or accepted.

In their favour: both are randomized, both used an active comparator (health education rather than waitlist), and the direction is consistent.

Against them: both are small; both come from a single research group; the confidence intervals are wide (the 2025 upper bound of 0.893 sits close to unity); the 2025 trial’s own primary endpoint — cIMT — was null, so the event finding is a secondary outcome in a trial that failed its primary; multiple maximum-follow-up windows were analysed and significance was lost at 14 years; trial registration post-dated the original data collection; and both carry researcher-allegiance and institutional-affiliation concerns of a degree that would attract close scrutiny in any pharmaceutical context. The distinction matters and this review observes it: the 2025 trial explicitly declared no commercial or financial conflict of interest, while several authors hold institutional affiliations with organizations involved in teaching and researching the intervention. That is an allegiance concern, not a documented undeclared financial conflict, and should be described as such. A 2025 commentary in a major cardiology review journal advocating TM within cardiovascular prevention frameworks originates from the same investigator group and should be read with the same awareness [69].

A separate 2025 multicentre trial of meditation and health education for cardiometabolic disease prevention in 201 Black women reported no between-group difference in carotid intima-media thickness, with some metabolic signals on secondary outcomes [68]. It is a separate trial but arises from the same investigator network, and therefore does not constitute independent replication.

The Cochrane review, applying prespecified methods to trials published to November 2021, found very sparse clinical-event evidence — two comparisons, one low and one very low certainty, neither showing a detectable effect [50]. Because its search closed in 2021, Cochrane cannot by itself adjudicate the 2025 and 2026 publications discussed above; our own structured search to 20 August 2026 identified no independent replication among them.

The correct statement is therefore precise, and scoped to what we searched: in the literature identified through our search to 20 August 2026, we found no independent, replicated, low-bias randomized evidence that meditation prevents cardiovascular events or death. This is not the same as evidence of no effect. It is the absence of the evidence that would be required to make the claim — and, given a structured rather than systematic search, it is a review finding rather than a proof of non-existence.

Grade: Inconclusive.

Table 7. Meditation and stress-reduction trials and syntheses.

学习	Intervention vs comparator	N / duration	Principal outcomes	Effect	Principal methodological concerns
Cochrane review, 2024	MBI or TM vs active or non-active	81 RCTs; 6,971	BP, psychological outcomes, lipids; CVD events sought	See Table 6; only 2 trials reported events	Mostly unclear; low to moderate certainty
Rogerson et al., 2024	Stress-management vs pooled controls	58 RCTs; 3,508	Cortisol (CAR, diurnal, single-sample)	g = 0.282 overall; CAR 0.644; meditation 0.345	Moderate; heterogeneous comparators
Koncz et al., 2021	Meditation vs control	10 RCTs (blood samples)	Blood cortisol	g = 0.62 (0.22–1.02); at-risk samples only	Moderate; small trials
Sanada et al., 2016	MBI vs control, healthy adults	5 RCTs; 190	Salivary cortisol	Smaller, inconsistent	适度
Dunn & Dimolareva, 2022	MBI vs randomized controls	48 RCTs; 4,683	CRP, IL-6	CRP −0.14 (−0.26 to −0.01); IL-6 −0.35 (−0.67 to −0.03)	Population and dose heterogeneity
MBSR gene-expression RCT, 2023	MBSR vs active health-enhancement, older adults	190; 8 weeks	NF-κB gene expression; IL-6; CRP	Gene expression d = 0.17 (p = 0.028); IL-6 and CRP unchanged	Low–moderate; best-designed inflammation trial
Brown et al., 2021	Mindfulness / meditation vs control	19 RCTs	Vagally mediated HRV	g = 0.38 (−0.014 to 0.77), I² = 89%; 0.19 without outlier	Extreme heterogeneity
Hughes et al., 2013	MBSR vs progressive muscle relaxation	56; 8 weeks	Clinic and ambulatory BP	Clinic SBP −4.8 vs −0.7 mmHg (p = 0.016); ambulatory null	Small, short
Blom et al. (HARMONY), 2014	MBSR vs waitlist	101; 12 weeks	24-h ambulatory BP	+0.4/0.0 vs +0.4/−0.4 mmHg — no effect	Pilot; waitlist
Loucks et al. (MB-BP), 2023	8-wk adapted mindfulness + CV behaviour education vs enhanced usual care	201; 6-mo follow-up	Office SBP; behaviour	SBP −4.5 mmHg (−9.0 to −0.1); sedentary time −350.8 min/wk	17.4% loss to follow-up; multicomponent; investigator commercial interest
MB-BP secondary analysis, 2026	Same programme	201; 6 mo	Composite CV health	SMD 0.144 (0.023–0.266); component CIs cross null	Secondary analysis
Anderson et al., 2008	TM vs control	Meta分析	SBP, DBP	−4.7 mmHg SBP; −3.2 DBP	Older meta-analysis; study-quality and investigator-allegiance concerns; AHA rated TM Class IIb, LOE B
Castillo-Richmond et al., 2000	TM vs health education	138 randomized; ≈60 paired cIMT (43.5% retention)	Carotid IMT	−0.098 vs +0.054 mm (p = 0.038)	Severe attrition; small analysed sample; preliminary
Schneider et al., 2012	TM vs health education, CHD patients	201; mean 5.4 y	Composite mortality, MI, stroke	HR 0.52 (0.29–0.92), p = 0.025	Small trial; single investigator network; allegiance concern
Norris et al., 2025	TM vs health education	197 randomized; 136 cIMT	cIMT at 12 months (primary); MACE to 14 y	颈动脉内中膜厚度 null (−0.0004 vs −0.0003 mm); MACE HR 0.346 (0.134–0.893) at 5 y, null at 14 y	Primary endpoint null; secondary events across multiple windows; retrospective registration; allegiance concern

23. Evidence Quality, Bias, and Adverse Effects

Table 8. Author-assessed narrative confidence by claim, with explicit basis for downgrade.

Downgrade domains: RoB = risk of bias · IND = indirectness · IMP = imprecision · INC = inconsistency · CONF = confounding · PB = publication bias.

Claim	Confidence	Downgrade domains	Basis
Stress → ASCVD association (several psychosocial constructs, CHD/CVD)	强大	CONF	Consistent associations across case-control and prospective cohort studies for several psychosocial constructs; imaging and mechanistic studies provide supporting evidence for selected intermediate pathways; residual socioeconomic and behavioural confounding remains
Stress → ASCVD causation (clinical events)	适度	CONF, IND	Temporality and experimental physiology support it; no randomized test; residual confounding cannot be excluded
Stress → plaque initiation / progression	Limited–Moderate	IND, IMP	Direct human serial-imaging evidence is sparse; most support inferred from event and surrogate endpoints
Amygdala–marrow–artery pathway	适度	IMP, RoB	One longitudinal imaging study; 22 events; observational mediation; psychometric substudy n = 13
Cortisol diurnal dysregulation → CV mortality	适度	IMP, INC	Several prospective cohorts report adverse associations with flatter or less differentiated rhythms, but cardiovascular event counts are small (31–32 deaths), measures differ, and a broader meta-analysis found no significant random-effects association for the cardiovascular-disease subgroup [78]
Cortisol Mendelian randomization → CHD	Limited–Moderate	IMP, IND	CI includes null; instruments explain ≈0.5% of variance, limiting power; SERPINA6/SERPINA1 pleiotropy unresolved; instruments cortisol, not stress
Spot cortisol → CV outcomes	有限	CONF, INC	Attenuated to null on adjustment (LURIC); null cross-sectional study; matrix mismatch with chronic exposure
Hypertension / autonomic pathway	Moderate–Strong	IND	Terminal step independently proven causal; exposure–mediator link and mediated fraction unquantified
Inflammation — acute stress-induced activation	强大	—	Experimental human meta-analysis with consistent direction
Inflammation — mediation of stress-related ASCVD risk	适度	IND	Anti-inflammatory RCTs establish causal relevance of inflammation to selected coronary outcomes, not the mediated fraction of stress risk
Endothelial dysfunction — acute human response	Moderate–Strong	IMP	Reproducible experimental effect; prognostic in CAD; small experimental samples
Endothelial molecular cascade (downstream steps)	有限	IND	Predominantly animal and cell-system evidence; endothelin-A and AT₁R components have small human support
Metabolic pathway	适度	IND	Endocrine plausibility strong; stress-specific mediated fraction unquantified
Lipid / apoB pathway	有限	INC	Occupational-cohort stress–lipid associations inconsistent
Thrombotic and acute-triggering pathways	Moderate–Strong	RoB	Large case-crossover data; mechanism studies methodologically heterogeneous
Behavioural mediation	Moderate–Strong	IND	Repeatedly observed; no defensible single quantification of mediated fraction
A. Meditation → reduced perceived stress	适度	IMP	SMD −0.24 against active comparators, I² = 0%; small effect, modest sample
B. Meditation → cortisol / HPA change	Limited–Moderate	INC, IND	Effects mainly in at-risk samples; heterogeneous matrices; pooled estimate not meditation-specific; mediation unproven
C. Meditation → blood pressure	适度	INC, IMP	Only moderate-certainty systolic estimate is −2.33 mmHg (TM vs active); MBI estimates heterogeneous (I² ≥ 87%)
C. Meditation → HRV	无定论	INC, IMP	Pooled effect crosses null; I² = 89%; outlier-sensitive
C. Meditation → inflammation	Limited–Moderate	INC, IND	Small pooled biomarker effects; best-controlled trial found gene-expression change without CRP or IL-6 change
D. Meditation → reduced atherosclerosis	Limited / Inconclusive	RoB, IMP	Original cIMT finding had 43.5% retention; later trial showed no between-group difference
E. Meditation → reduced hard events	无定论	RoB, IMP, PB	Two small allegiance-linked trials; sparse low-/very-low-certainty Cochrane event data; no independent replication identified

23.1 Systematic problems

Comparator selection. Waitlist and no-treatment controls tend to yield larger estimated effects than active comparators in this literature. Cochrane found consistently larger effects against inactive comparators.

Publication bias and related reporting concerns. Publication status was directly associated with effect size in the occupational-stress literature (1.43 published versus 1.16 unpublished). In the TM literature, small samples, selective-reporting possibilities, retrospective registration in some studies, and researcher allegiance create related but separately defined concerns, without an equivalent empirical quantification of publication bias.

Researcher allegiance and institutional affiliation. Particularly pronounced in TM research, where the trials reporting hard-endpoint benefit originate from investigators institutionally affiliated with organizations that teach and research the technique. Where financial interests are separately disclosed — as in the MB-BP programme and the 2026 commentary — this review says so explicitly; where trials declare no financial conflict, the concern is allegiance and affiliation rather than undeclared financial interest. The AHA statement explicitly called for adequately powered randomized trials with under 20% dropout conducted by investigators without inherent bias in outcome. That call has not been answered [46].

Underpowering and attrition. Studies are frequently underpowered per arm; attrition ranges from 17% to more than 50% among the key trials reviewed here.

Surrogate reliance and short follow-up. Trial durations in the Cochrane corpus were required to be at least 12 weeks, and most were short relative to what would be needed to assess structural plaque remodelling or event rates.

Reverse causality in cortisol cohorts. In cohorts with existing cardiovascular disease, hormone measurements may reflect disease severity, medications, acute illness, sleep, or renal function rather than psychological stress.

23.2 Adverse effects

Mind-body interventions are generally safe but not free of adverse effects. A 2020 systematic review pooling heterogeneous designs estimated an overall adverse-event prevalence of approximately 8% among people practising mindfulness or meditation — chiefly heightened anxiety, panic, depressive symptoms, or depersonalization — broadly comparable to rates reported for standard psychological therapies, though estimates varied widely by study design and reporting was substantially incomplete [70]. Notably, the Cochrane review found that most included trials did not report adverse events at all, which is itself a reporting failure [50].

24. Integrated Model with Per-Link Evidence Weight

Disease pathway (Figure 6). Grades below apply to each individual link and distinguish acute physiological certainty from chronic disease mediation. Chronic stress → cortico-limbic threat circuitry [Moderate–Strong] → acute HPA and sympathetic activation [Strong for acute physiology; chronic phenotype variable] → catecholamine response [Strong for acute physiology; Moderate for chronic CVD association], cortisol dysregulation [Moderate], reduced chronic vagal tone [Limited–Moderate], glucocorticoid receptor resistance [Moderate as mechanism; Limited as ASCVD mediator] → inflammation via marrow haematopoiesis, IL-6/CRP, NF-κB [Moderate–Strong in humans], endothelial dysfunction [Moderate–Strong], stress → sustained blood-pressure burden [Moderate]; blood pressure → ASCVD [Strong], insulin resistance and visceral adiposity [Moderate, partly behavioural], platelet and coagulation activation [Moderate, mainly acute] → atherosclerosis initiation and progression [Limited–Moderate: direct human serial-imaging evidence is sparse] → plaque instability and thrombosis [Moderate] → myocardial infarction, stroke, cardiovascular death [association Strong; mechanism-specific causal weight Moderate].

Intervention pathway. Meditation → reduced perceived stress [Moderate] → possible improvement in HPA and autonomic regulation [Limited–Moderate] → reduced blood pressure [Moderate], some reduction in inflammatory gene expression [Limited], small metabolic change [Limited] → possible reduction in atherosclerosis [Inconclusive; early positive cIMT trial followed by a later related null trial] → possible reduction in events [Inconclusive].

24.1 The two competing narratives, stated precisely

The strongest available interpretation of the mechanistic evidence is 不:

stress → continuously elevated cortisol → plaque.

A model more consistent with the available evidence — offered as a conceptual scheme, not a demonstrated serial mediation chain — is:

repeated psychological threat → altered central threat processing → repeated and dysregulated HPA and sympathetic activity → abnormal cortisol timing and reactivity, sometimes with glucocorticoid resistance and tissue-level amplification, together with catecholamine and haemodynamic activation → inflammatory, vascular, metabolic and behavioural change → increased susceptibility to atherosclerosis, and, in people who already have disease, increased vulnerability to acute triggering.

The second model accommodates what the first cannot: that some chronically stressed populations show high cortisol, others blunted reactivity, and others relatively normal circulating profiles with impaired glucocorticoid signalling.

24.2 Where stress sits relative to apoB

A point of framing that this review considers important. Atherosclerosis requires the retention of apoB-containing lipoproteins in the arterial 内膜 and a chronic inflammatory response to them. Psychological stress does not replace that process; current atherosclerosis biology provides no basis for treating psychological stress as an alternative to apoB-containing lipoprotein retention in plaque initiation [75].

Stress is more plausibly a modifier of atherogenic lipoprotein exposure and its vascular consequences than an alternative to apoB-driven plaque biology. What it can plausibly do is modify the rate at which that process runs, and modify the probability that established plaque converts to a clinical event. The magnitude of that modification will depend heavily on baseline apoB burden, smoking, blood pressure, diabetes, sleep, age, genetics, existing plaque burden, and behaviour. This is why the same 相对危险度 means very different absolute things in different people — and why stress management is properly framed as a modifier of risk, not as a foundation of prevention.

25. Clinical and Prevention Implications

Stress management, including meditation, is a reasonable, low-risk adjunct for interested patients — particularly for blood pressure, psychological wellbeing, and behavioural self-regulation. It should 不 replace statins or other apoB-lowering therapy, 抗高血压药, smoking cessation, diabetes treatment, physical activity, 心脏康复, or any other guideline-directed intervention. This is also the position of the American Heart Association, which concluded that meditation “may be considered as an adjunct” while noting that benefits remain to be better established [46,47].

Staged, actionable recommendations:

1. Assess psychosocial stressors, depression, anxiety, and social isolation as risk markers — particularly in patients with unexplained risk, poor control, or after myocardial infarction. The value of screening lies in identifying people who need behavioural support, not in generating a risk score.
2. Address established behavioural risk factors and psychosocial barriers alongside guideline-directed prevention. Smoking, physical activity, sleep, alcohol use and medication adherence are clinically important behavioural factors that can accompany psychological distress and should be addressed according to established prevention guidance. This is where the practical leverage is likely to lie, though the ordering is a clinical judgment rather than a directly tested comparison.
3. Optimize guideline-directed prevention regardless of stress status — blood pressure, LDL cholesterol and apoB, glucose, smoking. Stress status does not change these targets.
4. Offer meditation or structured stress reduction to patients who want it, framing the expected benefit honestly: modest blood-pressure and perceived-stress reduction, unproven event reduction. Where stress contributes to hypertension, poor sleep, inactivity, unhealthy eating, or difficulty adhering to treatment, the case for intervention is stronger — although the relative contribution of behavioural versus direct physiological mediation is not established.
5. Do not order cortisol testing for cardiovascular risk stratification. It remains a research tool except where Cushing’s syndrome is clinically suspected. In LURIC, morning serum cortisol showed no independent association with cardiovascular mortality after adjustment for conventional risk factors [27], and no cortisol measure has been shown in a formal incremental-prediction analysis to improve risk estimation or to identify treatment responders.
6. Do not present meditation as an alternative to pharmacotherapy. The most consequential harm in this area is not the intervention — it is the substitution.
7. Counsel patients with established coronary disease about acute triggering where clinically appropriate. This is a distinct and better-supported phenomenon than chronic atherogenesis. Note, however, that evidence establishing acute triggering does not establish that counselling about it prevents events; no trial has tested that.

26. Thresholds That Would Change These Conclusions

Stating in advance what evidence would change one’s mind is a discipline this literature would benefit from.

• 一个 independently conducted, actively controlled, event-driven randomized trial, powered a priori for a clinically plausible MACE reduction given the expected baseline event rate, follow-up duration, non-adherence and competing risk, showing that meditation reduces MACE would justify elevating it from adjunct to recommended therapy. Independent replication by investigators without intervention-specific allegiance would materially increase confidence.
• Replication of the amygdala–marrow–artery finding in a larger cohort with adequate event numbers would move that pathway from Moderate to Strong.
• Demonstration that experimentally modifying an adverse diurnal cortisol pattern in a prespecified direction reduces events — not merely correlates with them — would justify cortisol-rhythm monitoring and would establish cortisol as a modifiable target rather than a readout.
• 系列 冠状动脉成像 in a randomized stress-reduction trial showing changed plaque burden or phenotype would fill the Claim D gap directly.
• Stronger genetic instruments for cortisol, explaining substantially more than 0.5% of variance, would sharpen the causal estimate in either direction.

We identified none of these in our search to 20 August 2026.

27. Major Unanswered Research Questions

1. Does reducing stress, by any means, reduce hard cardiovascular events in an adequately powered, low-bias, independent randomized trial?
2. Does stress reduction alter serial coronary plaque burden or inflammatory 斑块表型?
3. Is modification of an adverse diurnal cortisol pattern itself cardioprotective, or is the pattern merely a marker of underlying health?
4. Which meditation type, dose, frequency, and duration — if any — is most effective, via which mechanism, against an active control?
5. How much of the stress–ASCVD link is genuinely causal versus confounded by socioeconomic position and behaviour?
6. Can hair cortisol or cortisone add incremental predictive value over standard risk factors?
7. Do baseline neuroimaging or cortisol phenotypes identify subgroups deriving disproportionate benefit from mind-body intervention?
8. Do effects differ by sex, race, socioeconomic stress burden, PTSD or depression status, or baseline HPA phenotype?
9. Do stress-reduction practices reproducibly modify immune gene expression, and if so, through what epigenetic mechanisms — including FKBP5-related regulation and GR signalling?

The field needs better mediation designs: an ideal study would randomize a standardized intervention, repeatedly measure perceived stress, sleep, behaviour, ambulatory blood pressure, catecholamines, multi-timepoint salivary cortisol, hair glucocorticoids, inflammatory signalling, endothelial function, and validated arterial imaging, then follow adjudicated cardiovascular events. Without that temporal mediator structure, an observed cortisol decrease and an observed blood-pressure decrease cannot establish that the former caused the latter.

28. Conclusions: Direct Answers to the Core Questions

Table 9. Summary answers.

Question	Answer	等级
Do chronic psychosocial stressors increase ASCVD risk?	Several distinct constructs — job strain, perceived stress, loneliness and social isolation — are each associated with modestly higher risk. Adjusted prospective estimates commonly ~1.2–1.3 for several psychosocial constructs; case-control and trigger estimates larger and answering different questions	协会 强大
How strong is the causal evidence?	Moderate. Supported by temporality, consistency and human experimental physiology, but materially weaker than the causal evidence for apoB exposure, smoking and blood pressure; residual confounding remains possible	适度
Which mechanisms have the strongest human evidence?	Autonomic and haemodynamic activation; inflammation; endothelial dysfunction; behavioural mediation; acute triggering. Brain–marrow–artery signalling is compelling but rests on one study	Moderate–Strong
What role does cortisol play?	A nonspecific biomarker of HPA/glucocorticoid physiology and a plausible mediator; a small causal contribution is possible but unproven. Not the unique transmitter of stress	Limited–Moderate for causation
Is elevated cortisol pathogenic, or is dysregulation the better model?	Dysregulation is the better framework, without one mandatory phenotype. Frank hypercortisolism is unequivocally pathogenic but is not the usual chronic-stress state	强大 evidence against the simple model; 适度 for any specific dysregulated phenotype
How much is mediated by BP, inflammation, insulin resistance, adiposity, sleep, and behaviour?	Unknown, and probably heterogeneous across populations and stressors. No single percentage is defensible; minimally adjusted and covariate-adjusted estimates answer different questions, and neither identifies a direct biological effect	—
Does meditation reduce perceived stress?	Yes, modestly; effect shrinks against credible active controls	适度
Does meditation alter cortisol or HPA physiology?	Small-to-moderate pooled effects, mainly in at-risk samples. “Modulation” is more accurate than either blanket “lowering” or “normalisation”; favourable direction depends on baseline HPA phenotype and cortisol metric. Mediation of benefit unproven	Limited–Moderate
Does meditation improve established risk factors?	Blood pressure modestly — approximately −2.3 mmHg systolic for TM versus active comparators in the moderate-certainty Cochrane estimate, with larger but less certain estimates in other comparisons; HRV not convincingly; inflammation inconsistently; metabolic markers not reliably	适度 for BP
Does meditation slow atherosclerosis?	Not demonstrated consistently. An early small trial reported cIMT benefit with substantial attrition; a later randomized trial in a different population did not demonstrate a between-group benefit	Limited / Inconclusive
Does meditation prevent MI, stroke, or CV death?	Not established. Two small allegiance-linked trials suggest benefit; no independent replication; Cochrane found sparse low-/very-low-certainty event evidence	无定论

28.1 What is justified, and what is overreach

Justified. Several forms of chronic psychosocial stress and distress are prospectively associated with modestly higher ASCVD risk after adjustment for measured covariates, and probably contribute causally to some degree. HPA-axis dysregulation — not simple hypercortisolism — is a more defensible physiological framework, though no single dysregulated phenotype is established. Acute emotional stress can trigger events in people with existing disease. Meditation lowers perceived stress, modestly lowers blood pressure, is generally low risk, and is a reasonable adjunct within a complete prevention strategy — though adverse psychological experiences do occur and cardiovascular trials have reported harms incompletely.

Overreach. That stress is a major independent driver of atherosclerosis comparable to lipids or smoking. That cortisol testing is clinically useful for cardiovascular risk. That meditation “prevents heart attacks or strokes.” That meditation “reverses atherosclerosis.” That meditation “normalizes cortisol to prevent cardiovascular disease.” That any of this justifies deferring or substituting for guideline-directed therapy.

The gap between these two lists is where most public communication on this topic goes wrong — and, because cumulative apoB exposure drives atherosclerotic risk over time [75], the substitution error is not a harmless one.

29. Limitations of This Review

This is a narrative evidence review, not a systematic review with a registered protocol, prespecified search strategy, or formal risk-of-bias instrument applied to each included study. Study selection reflects the judgement of the author, informed by four independently prepared source syntheses, and may not be exhaustive.

The grading scheme in Section 2.3 is explicit but is not GRADE, and the grades are the author’s assessments rather than formal consensus ratings. Where full texts were paywalled, quantitative values were taken from published abstracts. Effect estimates drawn from meta-analyses inherit the limitations of those analyses, including heterogeneity and the publication bias documented in Section 7.2.

The review is weighted toward literature published in English and toward cohorts in high-income countries, with the notable exceptions of INTERHEART and INTERSTROKE. Finally, this is a rapidly moving area — the Cochrane authors themselves noted a large number of ongoing eligible studies — and conclusions about Claims D and E in particular should be expected to change.

30. References

1. Rosengren A, Hawken S, Ôunpuu S, Sliwa K, Zubaid M, Almahmeed WA, et al. Association of psychosocial risk factors with risk of acute myocardial infarction in 11 119 cases and 13 648 controls from 52 countries (the INTERHEART study): case-control study. Lancet. 2004;364(9438):953–62. doi:10.1016/S0140-6736(04)17019-0. PMID: 15364186.
2. O’Donnell MJ, Xavier D, Liu L, Zhang H, Chin SL, Rao-Melacini P, et al. Risk factors for ischaemic and intracerebral haemorrhagic stroke in 22 countries (the INTERSTROKE study): a case-control study. Lancet. 2010;376(9735):112–23. doi:10.1016/S0140-6736(10)60834-3. PMID: 20561675.
3. Kivimäki M, Nyberg ST, Batty GD, Fransson EI, Heikkilä K, Alfredsson L, et al. Job strain as a risk factor for coronary heart disease: a collaborative meta-analysis of individual participant data. Lancet. 2012;380(9852):1491–7. doi:10.1016/S0140-6736(12)60994-5. PMID: 22981903.
4. Huang Y, Xu S, Hua J, Zhu D, Liu C, Hu Y, et al. Association between job strain and risk of incident stroke: a meta-analysis. Neurology. 2015;85(19):1648–54. doi:10.1212/WNL.0000000000002098. PMID: 26468409.
5. Fransson EI, Nyberg ST, Heikkilä K, Alfredsson L, Bjorner JB, Borritz M, et al. Job strain and the risk of stroke: an individual-participant data meta-analysis. Stroke. 2015;46(2):557–9. doi:10.1161/STROKEAHA.114.008019. PMID: 25563644.
6. Richardson S, Shaffer JA, Falzon L, Krupka D, Davidson KW, Edmondson D. Meta-analysis of perceived stress and its association with incident coronary heart disease. Am J Cardiol. 2012;110(12):1711–6. doi:10.1016/j.amjcard.2012.08.004. PMID: 22975465.
7. Valtorta NK, Kanaan M, Gilbody S, Ronzi S, Hanratty B. Loneliness and social isolation as risk factors for coronary heart disease and stroke: systematic review and meta-analysis of longitudinal observational studies. Heart. 2016;102(13):1009–16. doi:10.1136/heartjnl-2015-308790. PMID: 27091846. PMCID: PMC4941172.
8. Schulz R, Beach SR. Caregiving as a risk factor for mortality: the Caregiver Health Effects Study. JAMA. 1999;282(23):2215–9. doi:10.1001/jama.282.23.2215. PMID: 10605972.
9. Smyth A, O’Donnell M, Lamelas P, Teo K, Rangarajan S, Yusuf S; INTERHEART Investigators. Physical activity and anger or emotional upset as triggers of acute myocardial infarction: the INTERHEART study. Circulation. 2016;134(15):1059–67. doi:10.1161/CIRCULATIONAHA.116.023142. PMID: 27753614.
10. Tsai S-Y, Chi N-F, Wang Y-C, et al. Association of stress hormones and the risk of cardiovascular diseases: a systematic review and meta-analysis. Int J Cardiol Cardiovasc Risk Prev. 2024;23:200305. doi:10.1016/j.ijcrp.2024.200305. PMID: 39319239. PMCID: PMC11420448. Outcome is broad cardiovascular disease, not ASCVD specifically.
11. Tawakol A, Ishai A, Takx RAP, Figueroa AL, Ali A, Kaiser Y, et al. Relation between resting amygdalar activity and cardiovascular events: a longitudinal and cohort study. Lancet. 2017;389(10071):834–45. doi:10.1016/S0140-6736(16)31714-7. PMID: 28088338.
12. Heidt T, Sager HB, Courties G, Dutta P, Iwamoto Y, Zaltsman A, et al. Chronic variable stress activates hematopoietic stem cells. Nat Med. 2014;20(7):754–8. doi:10.1038/nm.3589. PMID: 24952646.
13. Marsland AL, Walsh C, Lockwood K, John-Henderson NA. The effects of acute psychological stress on circulating and stimulated inflammatory markers: a systematic review and meta-analysis. Brain Behav Immun. 2017;64:208–19. doi:10.1016/j.bbi.2017.01.011. PMID: 28089638.
14. Miller GE, Chen E, Sze J, Marin T, Arevalo JMG, Doll R, et al. A functional genomic fingerprint of chronic stress in humans: blunted glucocorticoid and increased NF-κB signaling. Biol Psychiatry. 2008;64(4):266–72. doi:10.1016/j.biopsych.2008.03.017. PMID: 18440494.
15. Cohen S, Janicki-Deverts D, Doyle WJ, Miller GE, Frank E, Rabin BS, et al. Chronic stress, glucocorticoid receptor resistance, inflammation, and disease risk. Proc Natl Acad Sci U S A. 2012;109(16):5995–9. doi:10.1073/pnas.1118355109. PMID: 22474371.
16. Kumari M, Shipley M, Stafford M, Kivimäki M. Association of diurnal patterns in salivary cortisol with all-cause and cardiovascular mortality: findings from the Whitehall II study. J Clin Endocrinol Metab. 2011;96(5):1478–85. doi:10.1210/jc.2010-2137. PMID: 21346074.
17. Karl S, Johar H, Ladwig KH, Peters A, Lederbogen F. Dysregulated diurnal cortisol patterns are associated with cardiovascular mortality: findings from the KORA-F3 study. Psychoneuroendocrinology. 2022;141:105753. doi:10.1016/j.psyneuen.2022.105753. PMID: 35395560.
18. Ronaldson A, Kidd T, Poole L, Leigh E, Jahangiri M, Steptoe A. Diurnal cortisol rhythm is associated with adverse cardiac events and mortality in coronary artery bypass patients. J Clin Endocrinol Metab. 2015;100(10):3676–82. doi:10.1210/jc.2015-2617. PMID: 26305622. PMCID: PMC4889728.
19. Vogelzangs N, Beekman ATF, Milaneschi Y, Bandinelli S, Ferrucci L, Penninx BWJH. Urinary cortisol and six-year risk of all-cause and cardiovascular mortality. J Clin Endocrinol Metab. 2010;95(11):4959–64. doi:10.1210/jc.2010-0192. PMID: 20739384.
20. van der Valk ES, Mohseni M, Iyer AM, et al. Long-term glucocorticoid exposure and incident cardiovascular diseases — the Lifelines cohort. J Clin Endocrinol Metab. 2024;109(10):2520–9. doi:10.1210/clinem/dgae081. PMID: 38500477. PMCID: PMC11403319. Outcome is broad CVD.
21. Stomby A, Strömberg S, Theodorsson E, Faresjö ÅO, Jones M, Faresjö T. Standard modifiable cardiovascular risk factors mediate the association between elevated hair cortisol concentrations and coronary artery disease. Front Cardiovasc Med. 2022;8:765000. doi:10.3389/fcvm.2021.765000. PMID: 35146006. PMCID: PMC8821522. Cross-sectional mediation analysis; not a SCAPIS analysis.
22. Hamer M, Endrighi R, Venuraju SM, Lahiri A, Steptoe A. Cortisol responses to mental stress and the progression of coronary artery calcification in healthy men and women. PLoS One. 2012;7(2):e31356. doi:10.1371/journal.pone.0031356. PMID: 22328931. PMCID: PMC3273460.
23. Hajat A, Diez-Roux AV, Sanchez BN, Holvoet P, Bild D, Seeman T, et al. Examining the association between salivary cortisol levels and subclinical measures of atherosclerosis: the Multi-Ethnic Study of Atherosclerosis. Psychoneuroendocrinology. 2013;38(7):1036–46. doi:10.1016/j.psyneuen.2012.10.007. PMID: 23146655. PMCID: PMC4020284.
24. Flynn S, Srikanthan P, Ravellette K, Inoue K, Watson K, Horwich T. Urinary cortisol and cardiovascular events in women vs. men: the Multi-Ethnic Study of Atherosclerosis. Am Heart J Plus. 2023;36:100344. doi:10.1016/j.ahjo.2023.100344. PMID: 37982128. PMCID: PMC10655947.
25. Crawford AA, Soderberg S, Kirschbaum C, Murphy L, Eliasson M, Ebrahim S, et al. Morning plasma cortisol as a cardiovascular risk factor: findings from prospective cohort and Mendelian randomization studies. Eur J Endocrinol. 2019;181(4):429–38. doi:10.1530/EJE-19-0161. PMID: 31325907. PMCID: PMC6733337.
26. Crawford AA, Bankier S, Altmaier E, Barnes CLK, Clark DW, Ermel R, et al. Variation in the SERPINA6/SERPINA1 locus alters morning plasma cortisol, hepatic corticosteroid binding globulin expression, gene expression in peripheral tissues, and risk of cardiovascular disease. J Hum Genet. 2021;66(6):625–36. doi:10.1038/s10038-020-00895-6. PMID: 33469137.
27. Pilz S, Theiler-Schwetz V, Trummer C, et al. Associations of serum cortisol with cardiovascular risk and mortality in patients referred to coronary angiography (LURIC). J Endocr Soc. 2021;5(5):bvab017. doi:10.1210/jendso/bvab017. PMID: 33869979. PMCID: PMC8041336.
28. Limumpornpetch P, Morgan AW, Tiganescu A, Baxter PD, Nyawira Nyaga V, Pujades-Rodriguez M, et al. The effect of endogenous Cushing syndrome on all-cause and cause-specific mortality. J Clin Endocrinol Metab. 2022;107(8):2377–88. doi:10.1210/clinem/dgac265. PMID: 35486378. PMCID: PMC9282270.
29. Reynolds RM, Ilyas B, Price JF, Fowkes FGR, Newby DE, Webb DJ, Walker BR. Circulating plasma cortisol concentrations are not associated with coronary artery disease or peripheral vascular disease. QJM. 2009;102(7):469–75. doi:10.1093/qjmed/hcp057. PMID: 19458201.
30. Ortiz R, Kluwe B, Lazarus S, Teruel MN, Joseph JJ. Cortisol and cardiometabolic disease: a target for advancing health equity. Trends Endocrinol Metab. 2022;33(11):786–97. doi:10.1016/j.tem.2022.08.002. PMID: 36266164. PMCID: PMC9676046. Narrative review; cited as a secondary synthesis for glucocorticoid receptor, pre-receptor metabolism and cardiometabolic framework, not as primary evidence for any individual molecular step.
31. Ghiadoni L, Donald AE, Cropley M, et al. Mental stress induces transient endothelial dysfunction in humans. Circulation. 2000;102(20):2473–8. doi:10.1161/01.CIR.102.20.2473. PMID: 11076819. Small experimental sample.
32. Lima BB, Hammadah M, Kim JH, et al. Association of transient endothelial dysfunction induced by mental stress with major adverse cardiovascular events in men and women with coronary artery disease. JAMA Cardiol. 2019;4(10):988–96. doi:10.1001/jamacardio.2019.3252. PMID: 31509180. PMCID: PMC6739728.
33. Sullivan S, Young A, Garcia M, et al. Sex differences in vascular response to mental stress and adverse cardiovascular events among patients with ischemic heart disease. Arterioscler Thromb Vasc Biol. 2023;43(5):e112–e120. doi:10.1161/ATVBAHA.122.318576. PMID: 36857628. PMCID: PMC10164352.
34. Hinterdobler J, Schott S, Jin H, et al. Acute mental stress drives vascular inflammation and promotes plaque destabilization in mouse atherosclerosis. Eur Heart J. 2021;42(39):4077–88. doi:10.1093/eurheartj/ehab371. PMID: 34279021. PMCID: PMC8516477. Human data concern leukocyte redistribution; plaque destabilization is mouse evidence.
35. Thrall G, Lane D, Carroll D, Lip GYH. A systematic review of the effects of acute psychological stress and physical activity on haemorheology, coagulation, fibrinolysis and platelet reactivity: implications for the pathogenesis of acute coronary syndromes. Thromb Res. 2007;120(6):819–47. doi:10.1016/j.thromres.2007.01.004. PMID: 17321571.
36. Wei J, Rooks C, Ramadan R, Shah AJ, Bremner JD, Quyyumi AA, et al. Meta-analysis of mental stress-induced myocardial ischemia and subsequent cardiac events in patients with coronary artery disease. Am J Cardiol. 2014;114(2):187–92. doi:10.1016/j.amjcard.2014.04.022. PMID: 24856319.
37. Kivimäki M, Steptoe A. Effects of stress on the development and progression of cardiovascular disease. Nat Rev Cardiol. 2018;15(4):215–29. doi:10.1038/nrcardio.2017.189. PMID: 29213140.
38. Hackett RA, Steptoe A, Kumari M. Association of diurnal patterns in salivary cortisol with type 2 diabetes in the Whitehall II study. J Clin Endocrinol Metab. 2014;99(12):4625–31. doi:10.1210/jc.2014-2459. PMID: 25215558.
39. Alfredsson L, Hammar N, Fransson E, et al. Job strain and major risk factors for coronary heart disease among employed males and females in a Swedish study on work, lipids and fibrinogen. Scand J Work Environ Health. 2002;28(4):238–48. doi:10.5271/sjweh.671. PMID: 12199425.
40. Brindle RC, Pearson A, Ginty AT. Adverse childhood experiences (ACEs) relate to blunted cardiovascular and cortisol reactivity to acute laboratory stress: a systematic review and meta-analysis. Neurosci Biobehav Rev. 2022;134:104530. doi:10.1016/j.neubiorev.2022.104530. PMID: 35031343.
41. Salomon RE, Tan KR, Vaughan A, Adynski H, Muscatell KA. Minimally-invasive methods for examining biological changes in response to chronic stress: a scoping review. Int J Nurs Stud. 2020;103:103419. doi:10.1016/j.ijnurstu.2019.103419. PMID: 31945603. PMCID: PMC7067628.
42. Chauntry AJ, Diaz KM, Sloan RP, et al. Psychobiological reactivity to acute psychological stress as a predictor of cardiovascular disease and mortality: the Multi-Ethnic Study of Atherosclerosis. Psychoneuroendocrinology. 2026;183:107657. doi:10.1016/j.psyneuen.2025.107657. PMID: 41135261.
43. Pan Y, Fan F, Jiang J, Zhang Y. Clinical outcomes of anti-inflammatory therapies inhibiting the NLRP3/IL-1β/IL-6/CRP pathway in coronary artery disease patients: a systematic review and meta-analysis. Inflamm Res. 2025;74:99. doi:10.1007/s00011-025-02058-9. PMID: 40583093. PMCID: PMC12206679.
44. Bando S, Fukuda D, Soeki T, et al. Expression of NLRP3 in subcutaneous adipose tissue is associated with coronary atherosclerosis. Atherosclerosis. 2015;242(2):407–14. doi:10.1016/j.atherosclerosis.2015.07.043. PMID: 26282945. NLRP3 measured in subcutaneous adipose tissue; psychological stress not assessed.
45. Goodwin JE, Feng Y, Velazquez H, Zhou H, Sessa WC. Endothelial glucocorticoid receptor suppresses atherogenesis — brief report. Arterioscler Thromb Vasc Biol. 2015;35(4):779–82. doi:10.1161/ATVBAHA.114.304525. PMID: 25810297. PMCID: PMC4375730. ApoE-deficient mouse model; not human evidence.
46. Levine GN, Lange RA, Bairey-Merz CN, Davidson RJ, Jamerson K, Mehta PK, et al. Meditation and cardiovascular risk reduction: a scientific statement from the American Heart Association. J Am Heart Assoc. 2017;6(10):e002218. doi:10.1161/JAHA.117.002218. PMID: 28963100.
47. Levine GN, Cohen BE, Commodore-Mensah Y, Fleury J, Huffman JC, Khalid U, et al. Psychological health, well-being, and the mind-heart-body connection: a scientific statement from the American Heart Association. Circulation. 2021;143(10):e763–83. doi:10.1161/CIR.0000000000000947. PMID: 33486973.
48. Levine GN, Carney RM, Cohen BE, et al. Post-myocardial infarction psychological distress: a scientific statement from the American Heart Association. Circulation. 2025;152:e298–e310. doi:10.1161/CIR.0000000000001381. PMID: 40977387. PMCID: PMC12919889.
49. Brook RD, Appel LJ, Rubenfire M, Ogedegbe G, Bisognano JD, Elliott WJ, et al. Beyond medications and diet: alternative approaches to lowering blood pressure — a scientific statement from the American Heart Association. Hypertension. 2013;61(6):1360–83. doi:10.1161/HYP.0b013e318293645f. PMID: 23608661.
50. Rees K, Takeda A, Court R, Kudrna L, Hartley L, Ernst E. Meditation for the primary and secondary prevention of cardiovascular disease. Cochrane Database Syst Rev. 2024;2(2):CD013358. doi:10.1002/14651858.CD013358.pub2. PMID: 38358047. PMCID: PMC10867897.
51. Rogerson O, Wilding S, Prudenzi A, O’Connor DB. Effectiveness of stress management interventions to change cortisol levels: a systematic review and meta-analysis. Psychoneuroendocrinology. 2024;159:106415. doi:10.1016/j.psyneuen.2023.106415. PMID: 37879237.
52. Koncz A, Demetrovics Z, Takacs ZK. Meditation interventions efficiently reduce cortisol levels of at-risk samples: a meta-analysis. Health Psychol Rev. 2021;15(1):56–84. doi:10.1080/17437199.2020.1760727. PMID: 32635830.
53. Sanada K, Montero-Marin J, Alda Díez M, Salas-Valero M, Pérez-Yus MC, Morillo H, et al. Effects of mindfulness-based interventions on salivary cortisol in healthy adults: a meta-analytical review. Front Physiol. 2016;7:471. doi:10.3389/fphys.2016.00471. PMID: 27807420. PMCID: PMC5069287.
54. Dunn TJ, Dimolareva M. The effect of mindfulness-based interventions on immunity-related biomarkers: a comprehensive meta-analysis of randomised controlled trials. Clin Psychol Rev. 2022;92:102124. doi:10.1016/j.cpr.2022.102124. PMID: 35078038.
55. Lindsay EK, Marsland AL, Cole SW, et al. Mindfulness-based stress reduction reduces proinflammatory gene regulation but not systemic inflammation among older adults: a randomized controlled trial. Psychosom Med. 2024;86(5):463–72. doi:10.1097/PSY.0000000000001264. PMID: 37982547. PMCID: PMC11098967. Published online 2023.
56. Creswell JD, Irwin MR, Burklund LJ, Lieberman MD, Arevalo JMG, Ma J, Breen EC, Cole SW. Mindfulness-Based Stress Reduction training reduces loneliness and pro-inflammatory gene expression in older adults: a small randomized controlled trial. Brain Behav Immun. 2012;26(7):1095–101. doi:10.1016/j.bbi.2012.07.006. PMID: 22820409. PMCID: PMC3635809.
57. Brown L, Rando AA, Eichel K, Van Dam NT, Celano CM, Huffman JC, et al. The effects of mindfulness and meditation on vagally mediated heart rate variability: a meta-analysis. Psychosom Med. 2021;83(6):631–40. doi:10.1097/PSY.0000000000000900. PMID: 33395216.
58. Hughes JW, Fresco DM, Myerscough R, van Dulmen MHM, Carlson LE, Josephson R. Randomized controlled trial of mindfulness-based stress reduction for prehypertension. Psychosom Med. 2013;75(8):721–8. doi:10.1097/PSY.0b013e3182a3e4e5. PMID: 24127622.
59. Blom K, Baker B, How M, Dai M, Irvine J, Abbey S, et al. Hypertension analysis of stress reduction using mindfulness meditation and yoga: results from the HARMONY randomized controlled trial. Am J Hypertens. 2014;27(1):122–9. doi:10.1093/ajh/hpt134. PMID: 24038797.
60. Loucks EB, Schuman-Olivier Z, Saadeh FB, et al. Effect of adapted mindfulness training in participants with elevated office blood pressure: the MB-BP study — a randomized clinical trial. J Am Heart Assoc. 2023;12(11):e028712. doi:10.1161/JAHA.122.028712. PMID: 37218591. PMCID: PMC10381983. Multicomponent intervention; investigator commercial interest disclosed.
61. Wu F, Brewer LC, Neves VV, Scarpaci MM, Saadeh FB, Proulx JA, Loucks EB. Impact of the mindfulness-based blood pressure reduction (MB-BP) program on cardiovascular health: a randomized clinical trial. Am J Prev Cardiol. 2026;27:101584. doi:10.1016/j.ajpc.2026.101584. PMID: 42291030. PMCID: PMC13261222.
62. Anderson JW, Liu C, Kryscio RJ. Blood pressure response to Transcendental Meditation: a meta-analysis. Am J Hypertens. 2008;21(3):310–6. doi:10.1038/ajh.2007.65. PMID: 18311126.
63. Bai Z, Chang J, Chen C, Li P, Yang K, Chi I. Investigating the effect of transcendental meditation on blood pressure: a systematic review and meta-analysis. J Hum Hypertens. 2015;29(11):653–62. doi:10.1038/jhh.2015.6. PMID: 25673114.
64. Fisher V, Li WW, Malabu U. The effectiveness of mindfulness-based stress reduction (MBSR) on the mental health, HbA1c, and mindfulness of diabetes patients: a systematic review and meta-analysis of randomised controlled trials. Appl Psychol Health Well Being. 2023;15(4):1733–49. doi:10.1111/aphw.12441. PMID: 36855248.
65. Castillo-Richmond A, Schneider RH, Alexander CN, Cook R, Myers H, Nidich S, et al. Effects of stress reduction on carotid atherosclerosis in hypertensive African Americans. Stroke. 2000;31(3):568–73. doi:10.1161/01.STR.31.3.568. PMID: 10700487.
66. Schneider RH, Grim CE, Rainforth MV, Kotchen T, Nidich SI, Gaylord-King C, et al. Stress reduction in the secondary prevention of cardiovascular disease: randomized, controlled trial of Transcendental Meditation and health education in Blacks. Circ Cardiovasc Qual Outcomes. 2012;5(6):750–8. doi:10.1161/CIRCOUTCOMES.112.967406. PMID: 23149426.
67. Norris KC, Salerno J, Bairey Merz CN, Kaushik V, Gelleta S, Castillo A, Nidich S, Gaylord-King C, Schneider RH. A randomized controlled trial of meditation and health education on carotid intima-media thickness and major adverse cardiovascular events in Black men and women. Front Med (Lausanne). 2025;12:1513699. doi:10.3389/fmed.2025.1513699. PMID: 40177279. PMCID: PMC11962031. Trial conducted 2000–2014; registered retrospectively (NCT05642936). Declares no commercial or financial conflict of interest; author institutional affiliations constitute an allegiance concern.
68. Gaylord-King C, Harris CL, Xu S, Salerno JW, Walton KG, Bairey-Merz CN, Nidich S, Schneider RH. A multicenter randomized controlled trial of meditation and health education in the prevention of cardiometabolic disease in Black women. J Womens Health. 2025;34(10):1247–56. doi:10.1177/15409996251364663. PMID: 40759176.
69. Schneider RH, Norris KC, Brook RD. Transcendental Meditation to combat psychosocial stress, hypertension and cardiovascular disease. Nat Rev Cardiol. 2026;23(1):4–5. doi:10.1038/s41569-025-01235-x. PMID: 41266880. Commentary, not primary outcome evidence; cited to document investigator allegiance. Disclosed conflicts include an author commercial interest.
70. Farias M, Maraldi E, Wallenkampf KC, Lucchetti G. Adverse events in meditation practices and meditation-based therapies: a systematic review. Acta Psychiatr Scand. 2020;142(5):374–93. doi:10.1111/acps.13225. PMID: 32820538.
71. Landsbergis P, Schnall P. Job strain and coronary heart disease [correspondence]. Lancet. 2013;381(9865):448. doi:10.1016/S0140-6736(13)60242-1. PMID: 23399068.
72. Choi B, Schnall P, Ko S, Dobson M, Baker D. Job strain and coronary heart disease [correspondence]. Lancet. 2013;381(9865):448. doi:10.1016/S0140-6736(13)60243-3. PMID: 23399069.
73. Kivimäki M, Singh-Manoux A, Nyberg S, Batty GD. Job strain and coronary heart disease — authors’ reply. Lancet. 2013;381(9865):448–9. doi:10.1016/S0140-6736(13)60244-5. PMID: 23399067.
74. Hamer M, Steptoe A. Cortisol responses to mental stress and incident hypertension in healthy men and women. J Clin Endocrinol Metab. 2012;97(1):E29–34. doi:10.1210/jc.2011-2132. PMID: 22031509.
75. Borén J, Chapman MJ, Krauss RM, et al. Low-density lipoproteins cause atherosclerotic cardiovascular disease: pathophysiological, genetic, and therapeutic insights — a consensus statement from the European Atherosclerosis Society Consensus Panel. Eur Heart J. 2020;41(24):2313–30. doi:10.1093/eurheartj/ehz962. PMID: 32052833. PMCID: PMC7308544.
76. Spieker LE, Hürlimann D, Ruschitzka F, Corti R, Enseleit F, Shaw S, Hayoz D, Deanfield JE, Lüscher TF, Noll G. Mental stress induces prolonged endothelial dysfunction via endothelin-A receptors. Circulation. 2002;105(24):2817–20. doi:10.1161/01.CIR.0000021598.15895.34. PMID: 12070106.
77. Rocha HNM, Batista GMS, Storch AS, Garcia VP, Teixeira GF, Mentzinger J, Gomes EAC, Campos MO, Nóbrega ACL, Rocha NG. Mental stress induces endothelial dysfunction by AT1R-mediated redox imbalance in overweight/obese men. Braz J Med Biol Res. 2023;56:e12547. doi:10.1590/1414-431X2023e12547. PMID: 36995873. PMCID: PMC10041671. Three-way randomized crossover, n = 15; single sex and body-composition stratum.
78. Adam EK, Quinn ME, Tavernier R, McQuillan MT, Dahlke KA, Gilbert KE. Diurnal cortisol slopes and mental and physical health outcomes: a systematic review and meta-analysis. Psychoneuroendocrinology. 2017;83:25–41. doi:10.1016/j.psyneuen.2017.05.018. PMID: 28578301. PMCID: PMC5568897. 179 associations from 80 studies to January 2015; overall r = 0.147; cardiovascular-disease subgroup not significant in random-effects analysis.

Verification Note

Quantitative estimates were cross-checked against the primary publication or the cited systematic review wherever available. INTERHEART and INTERSTROKE figures are reported with 99% confidence intervals per the original study designs; all other intervals are 95% unless stated.

Three points of interpretation are flagged for the reader. First, the Tawakol psychometric substudy comprised n = 13 participants, and correlations from that substudy should be treated as exploratory. Second, the trial published in 2025 reports a ten-year hazard ratio of 0.485 with a 95% confidence interval of 0.226–1.044 alongside p = 0.0435; these are internally inconsistent as published, and this review therefore does not describe that result as statistically significant. Third, the Transcendental Meditation hard-endpoint literature (references 65–69) should be read with explicit awareness of researcher allegiance and institutional affiliation; where a financial interest is separately disclosed — as for the MB-BP programme (reference 60) and the 2026 commentary (reference 69) — this is stated, and where trials declare no financial conflict the concern is allegiance rather than undeclared financial interest.

Figures 1A and 6 are schematic representations, not plots of measured data. All other figures plot values reported in the cited primary sources.