背景

随着我国经济社会的发展，人口老龄化加剧，慢性非传染性疾病负担日益加重，尤其是心血管疾病（CVD）。近年来，我国健康政策逐步从疾病治疗转向健康促进与预防，并强调心理健康和环境因素在疾病防控中的作用。现有的心血管健康评估指标由美国心脏协会提出的Life’s Simple 7（LS7）发展为Life’s Essential 8（LE8），新增睡眠这一行为因素，体现了对多维度健康要素整合评估的趋势。然而，将抑郁状态作为心理健康的代表性指标纳入构建的Life’s Crucial 9（LC9），其在心血管疾病风险预测中的表现是否优于LE8，目前仍缺乏大规模人群研究的证据支持。此外，现行LE8评分采用等权重加总的方式，未能充分反映各组成因素在心血管结局预测中的差异性，如何通过差异化赋权优化评分系统仍有待进一步研究。与此同时，长期室外空气污染暴露和家庭空气污染暴露与心血管健康的交互作用仍不明确，其与个体健康状态之间的综合影响尚需深入探索。

目的

本研究基于中国前瞻性队列研究PURE-China数据，系统评估增加抑郁状态评分是否能够提升LE8模型对CVD及全因死亡风险的预测能力，同时探索采用多因素回归权重构建加权LE8评分后，其预测性能是否优于传统未加权LE8评分。在此基础上，进一步探究不同PM2.5暴露水平和家庭空气污染暴露与心血管健康指标对结局事件发生的关联是否存在交互作用与联合作用，从而为优化心血管健康评估工具和精准防控策略提供依据。

方法

心血管健康评分LE8评分包括4个行为因素：尼古丁暴露、饮食、体力活动、睡眠健康）；4个代谢因素：身体质量指数（Body Mass Index，BMI）、血脂、血糖、血压），每项评分0-100，最终得分为未加权平均值（0-100分）。抑郁症状信息通过使用简易版复合性国际诊断交谈量表（Composite International Diagnostic Interview-Short Form，CIDI-SF）来进行收集，量表总分为0-7分。按区间（0，1-2，3-4，5，6-7）分别赋值100、75、50、25、0，最终LC9评分为LE8评分与抑郁评分的均值。依据LE8写作组推荐，LE8/LC9评分分为三个等级：高（80-100分）、中（50-79分）、低（0-49分）。PM2.5暴露水平基于卫星遥感和地面监测数据估算，并按中位数分类。主要研究终点为死亡和主要心血管事件（包括心血管死亡、非致命性心肌梗死、卒中、心力衰竭）的复合终点，次要研究终点为主要研究终点的各个组分。使用加入随机效应项的Cox脆弱模型（Cox Frailty Model）评估LE8与LC9与结局事件的关联，估算风险比（Hazard Ratio，HR）及95%置信区间（Confidence Interval，CI），调整年龄、性别、城乡、社会经济因素等混杂变量。在构建加权LE8评分时，采用Cox脆弱模型中各组分在不同结局下的回归系数作为加权依据。使用限制性立方样条（Restricted Cubic Spline，RCS）评估LE8/LC9与结局发生的剂量-反应关系，并采用Harrell C统计量、赤池信息准则（Akaike Information Criterion，AIC）、净重分类改进指数（Net Reclassification Improvement，NRI）和综合判别改进指数（Integrated Discrimination Improvement，IDI）评估LE8和新增抑郁因素对结局发生风险预测能力的提升。通过相加交互作用指标交互作用导致的相对超额风险（Relative Excess Risk due to Interaction，RERI）和交互作用导致的疾病比例（Attributable Proportion due to Interaction，AP）以及相乘交互作用模型，探讨不同PM2.5暴露、家庭空气污染暴露水平与心血管健康评分的联合作用。

结果

本研究基于PURE-China前瞻性队列47931名参与者的数据开展，中位随访时间为11.97年，四分位数间距为9.59-12.61年。随访期间，共记录了4085例复合结局事件，其中包括1914例死亡事件及2881例主要心血管事件（包括638例心血管死亡、2082例卒中、738例心肌梗死及199例心力衰竭）。随着LE8评分的增加，每1000人年年龄性别标准化复合终点发生率从15.24（95%CI：13.27-17.50）降至6.20（5.78-6.64），趋势性检验P<0.001；每1000人年主要心血管事件发生率从11.82（10.05-13.90）降至3.94（3.61-4.30），趋势性检验P<0.001；每1000人年全因死亡发生率从6.26（5.16-7.59）降至2.76（2.49-3.06），趋势性检验P<0.001。

LE8、LC9及加权LE8评分均与复合终点、主要心血管事件和全因死亡风险呈显著负相关（P < 0.001），随着评分水平的增加，各类结局事件的发生风险显著下降（趋势性检验 P < 0.001）。剂量-反应关系分析显示，LE8和LC9评分与复合终点及主要心血管事件之间未观察到显著的非线性关联，而加权LE8评分则在复合终点、主要心血管事件和全因死亡方面均呈现出显著的非线性负相关（非线性 P < 0.001），其中在中低分段的风险下降更为显著，而在高分段趋于平台。

在进一步比较模型预测性能方面，与LE8模型相比，加入CIDI-SF评分后，连续 为-0.043（95%CI：-0.045--0.041），P<0.001； 为0（95%CI：0-0），P<0.001。而加入抑郁症状评分后，连续 为-0.043（-0.050-0.300），P=0.192； 为0（0-0），P=0.188。表明无论采用哪种抑郁状态指标，原模型的预测能力并没有显著改善。相比之下，基于各结局的回归系数对LE8评分进行加权处理，可在一定程度上提升模型的预测性能，尽管整体提升幅度较为有限。在六项结局指标中，加权LE8模型的AIC值均略低于对应的未加权模型，提示其拟合优度有所改善。例如，复合终点的C统计量由未加权模型的0.7307（95% CI：0.7230–0.7384）提高至加权模型的0.7352（95% CI：0.7276–0.7428）；主要心血管事件的C统计量从0.7282上升至0.7380；全因死亡的C统计量亦从0.7695提升至0.7715，表明加权策略在模型区分能力上具有一定优化效果。

PM2.5暴露水平与LE8评分对复合终点、主要心血管事件和全因死亡无显著的相乘交互作用（P=0.546，0.247，0.479）；也没有显著的协同相加交互作用，复合终点 ：0.237（95%CI：-0.357-0.831）， ：0.101（95%CI：-0.152-0.354）；主要心血管事件 ：0.224（-0.560-1.009）， ：0.083（-0.208-0.374）；全因死亡 ：0.395（-0.322-1. 112）， ：0.083（-0.162-0.559）。联合作用分析显示，随着PM2.5暴露水平的提高，所有LE8评分水平的结局事件发生风险相应地增加；在高PM2.5暴露时，低LE8评分者的复合终点和主要心血管事件风险最高：复合终点 =1.72（95%CI：1.16-2.56），主要心血管事件 =1.88（1.17-3.01），表明高PM2.5暴露对心血管健康水平较差者影响更大。

是否暴露于家庭空气污染与LE8评分对复合终点、主要心血管事件和全因死亡无显著的相乘交互作用（P=0.688，0.393，0.597）；也没有显著的协同相加交互作用，复合终点 ：0.200（95%CI：-0.308-0.707）， ：0.108（95%CI：-0.166-0.382）；主要心血管事件 ：0.243（-0.364, 0.850）， ：0.130（-0.195, 0.455）；全因死亡 ：0.255（-0.459, 0.969）， ：0.129（-0.233, 0.491）。联合作用分析显示，与暴露组相比，非暴露组所有LE8评分水平的结局事件发生风险相应地降低；高LE8评分者在无家庭空气污染暴露时的结局事件风险最低：复合终点 =0.38（95%CI：0.29-0.49），主要心血管事件 =0.32（0.24-0.43），全因死亡 =0.43（0.31-0.60），表明无家庭空气污染暴露对心血管健康水平较高者影响更大。

结论

LE8和LC9评分均可有效预测心血管疾病（CVD）及全因死亡风险，但将抑郁状态评分纳入LE8模型后，并未显著提升其预测能力。进一步地，基于Cox脆弱模型回归系数构建的加权LE8评分在预测CVD及死亡风险方面表现出略优于未加权评分的性能，提示差异化赋权在优化心血管健康评估工具方面具有一定可行性，尽管整体提升幅度有限。同时，PM2.5暴露和家庭空气污染暴露与心血管健康水平的联合作用显著影响结局发生风险，心血管健康水平较低的个体在高污染环境中风险尤为突出。本研究为中国人群的心血管健康评估、空气污染治理及综合干预策略提供了实证支持，也为优化公共卫生政策和精准健康管理措施提供了科学依据。

Background

With the rapid development of China's economy and the acceleration of population aging, the burden of non-communicable diseases (NCDs)—particularly cardiovascular disease (CVD)—has continued to rise. In recent years, national health policy has gradually shifted from a focus on treatment to a greater emphasis on health promotion and disease prevention, with increasing attention to the roles of mental health and environmental factors in disease control. The existing cardiovascular health assessment metric, originally proposed by the American Heart Association as Life's Simple 7 (LS7), has evolved into Life's Essential 8 (LE8) through the addition of sleep health as a behavioral component, reflecting a trend toward multidimensional health evaluation. However, the integration of depression—a representative indicator of mental health—into the extended Life's Crucial 9 (LC9) metric has not yet been validated by large-scale population-based studies to determine whether it offers improved predictive value for CVD risk compared to LE8.

Furthermore, the current LE8 scoring system applies equal weights to each component, which may not adequately reflect the relative contributions of individual factors to cardiovascular outcomes. The potential for improving predictive performance through differential weighting remains underexplored. At the same time, the interactions between long-term ambient air pollution and household air pollution exposures with cardiovascular health status are still unclear, and the combined effects of these exposures on individual health require further investigation.

Objectives

Based on data from the prospective PURE-China cohort study, this research aims to systematically evaluate whether incorporating a depression score enhances the predictive ability of the LE8 model for CVD and all-cause mortality. In addition, the study explores whether constructing a weighted LE8 score—based on regression-derived weights—can improve prediction performance compared to the conventional unweighted LE8. Building on these analyses, the study further investigates whether there are interaction or joint effects between cardiovascular health scores and different levels of PM2.5 exposure or household air pollution, with the goal of informing the optimization of cardiovascular health assessment tools and the development of targeted prevention strategies.

Methods

The LE8 cardiovascular health score comprises four behavioral factors—nicotine exposure, diet, physical activity, and sleep health—and four metabolic factors—body mass index (BMI), blood lipids, blood glucose, and blood pressure. Each component is scored from 0 to 100, and the final LE8 score is calculated as an unweighted average (ranging from 0 to 100). Depression symptom information is collected using the Composite International Diagnostic Interview-Short Form (CIDI-SF), with a total score ranging from 0 to 7. The depression score is categorized into intervals (0, 1–2, 3–4, 5, 6–7), assigned values of 100, 75, 50, 25, and 0, respectively. The final LC9 score is derived as the average of the LE8 score and the depression score. Based on the LE8 writing group's recommendations, LE8/LC9 scores are classified into three categories: high (80–100), moderate (50–79), and low (0–49). PM2.5 exposure levels are estimated using satellite remote sensing and ground monitoring data and are classified according to the median. The primary outcome is a composite outcome of mortality and major cardiovascular events, including cardiovascular death, nonfatal myocardial infarction, stroke, and heart failure. Secondary outcomes include the individual components of the primary outcome. A Cox frailty model incorporating random effects is used to assess the association between LE8, LC9, and outcome events, estimating hazard ratios (HRs) and 95% confidence intervals (CIs) while adjusting for potential confounders such as age, sex, urban or rural residence, and socioeconomic factors. When constructing the weighted LE8 score, the regression coefficients of each component derived from the Cox frailty models for different outcomes were used as the weighting factors. Restricted cubic spline (RCS) regression is employed to evaluate the dose-response relationship between LE8/LC9 scores and health outcomes. Improvements in risk prediction by incorporating depression into the LE8 score are assessed using Harrell's C-statistic, Akaike Information Criterion (AIC), Net Reclassification Improvement (NRI), and Integrated Discrimination Improvement (IDI). The joint associations of different PM2.5 exposure levels, household air pollution exposure, and cardiovascular health scores are examined using additive interaction indices, including relative excess risk due to interaction (RERI) and attributable proportion due to interaction (AP), as well as multiplicative interaction models.

Results

This study was based on data from the PURE-China prospective cohort, including 47,931 participants, with a median follow-up time of 11.97 years (interquartile range: 9.59–12.61 years). During the follow-up period, a total of 4,085 composite outcome events were recorded, comprising 1,914 deaths and 2,881 major cardiovascular events (including 638 cardiovascular deaths, 2,082 strokes, 738 myocardial infarctions, and 199 heart failure cases). With increasing LE8 scores, the age- and sex-standardized incidence rates of composite outcomes per 1,000 person-years decreased from 15.24 (95% CI: 13.27–17.50) to 6.20 (95% CI: 5.78–6.64) (P for trend <0.001). Similarly, the incidence of major cardiovascular events per 1,000 person-years declined from 11.82 (95% CI: 10.05–13.90) to 3.94 (95% CI: 3.61–4.30) (P for trend <0.001), and all-cause mortality rates declined from 6.26 (95% CI: 5.16–7.59) to 2.76 (95% CI: 2.49–3.06) (P for trend <0.001).

The LE8, LC9, and weighted LE8 scores were all significantly and inversely associated with the risks of composite outcomes, major cardiovascular events, and all-cause mortality (P < 0.001). As the score levels increased, the risks of these adverse outcomes decreased significantly (P for trend < 0.001). Dose–response analyses revealed no significant non-linear associations between LE8 or LC9 scores and the risks of composite outcomes or major cardiovascular events. In contrast, the weighted LE8 score showed significant non-linear inverse associations with composite outcomes, major cardiovascular events, and all-cause mortality (P for non-linearity < 0.001), with more pronounced risk reductions observed in the lower-to-middle score ranges, and a plateauing effect at higher score levels.

Further comparison of model predictive performance showed that adding the CIDI-SF depression score to the LE8 model resulted in a continuous  of -0.043 (95% CI: -0.045 to -0.041, P < 0.001) and an  of 0 (95% CI: 0–0, P < 0.001). Similarly, incorporating a general depressive symptom score yielded a continuous  of -0.043 (95% CI: -0.050 to 0.300, P = 0.192) and an  of 0 (P = 0.188), indicating that neither depression indicator meaningfully improved the model's predictive performance.

In contrast, applying regression-based weights for each LE8 component according to different outcome associations led to a modest improvement in predictive performance. Across six clinical outcomes, the weighted LE8 models consistently exhibited slightly lower AIC values compared to their unweighted counterparts, suggesting better model fit. For instance, the C-statistic for the composite outcome increased from 0.7307 (95% CI: 0.7230–0.7384) in the unweighted model to 0.7352 (95% CI: 0.7276–0.7428) in the weighted model; for major cardiovascular events, the C-statistic improved from 0.7282 to 0.7380; and for all-cause mortality, from 0.7695 to 0.7715. These findings suggest that weighing the LE8 components offers a limited yet consistent enhancement in model discrimination.

There were no significant multiplicative interactions between PM2.5 exposure levels and LE8 scores for composite outcomes, major cardiovascular events, and all-cause mortality (P=0.546, 0.247, 0.479, respectively). Similarly, no significant additive synergistic interactions were observed: composite outcomes ( : 0.237, 95% CI: -0.357 to 0.831; : 0.101, 95% CI: -0.152 to 0.354), major cardiovascular events ( : 0.224, 95% CI: -0.560 to 1.009; : 0.083, 95% CI: -0.208 to 0.374), and all-cause mortality ( : 0.395, 95% CI: -0.322 to 1.112; : 0.083, 95% CI: -0.162 to 0.559). Joint-effect analyses revealed that higher PM2.5 exposure corresponded to increased risks of outcome events across all LE8 score levels, with the highest risks among individuals with low LE8 scores under high PM2.5 exposure conditions: composite outcome =1.72 (95% CI: 1.16–2.56), major cardiovascular events ==1.88 (95% CI: 1.17–3.01), indicating a greater impact of high PM2.5 exposure among individuals with poorer cardiovascular health.

No significant multiplicative interactions were identified between household air pollution exposure and LE8 scores for composite outcomes, major cardiovascular events, and all-cause mortality (P=0.688, 0.393, 0.597, respectively). Similarly, additive interactions were not significant: composite outcomes ( : 0.200, 95% CI: -0.308 to 0.707; : 0.108, 95% CI: -0.166 to 0.382), major cardiovascular events ( : 0.243, 95% CI: -0.364 to 0.850; : 0.130, 95% CI: -0.195 to 0.455), and all-cause mortality ( : 0.255, 95% CI: -0.459 to 0.969; : 0.129, 95% CI: -0.233 to 0.491). Joint-effect analyses demonstrated lower risks of outcome events across all LE8 score levels in the non-exposed group compared to the exposed group. Individuals with high LE8 scores and no household air pollution exposure exhibited the lowest risk of outcome events: composite outcomes =0.38 (95% CI: 0.29–0.49), major cardiovascular events =0.32 (95% CI: 0.24–0.43), and all-cause mortality =0.43 (95% CI: 0.31–0.60), suggesting a greater protective effect of non-exposure among individuals with better cardiovascular health.

Conclusion

Both the LE8 and LC9 scores were effective in predicting the risk of cardiovascular disease (CVD) and all-cause mortality. However, incorporating depressive symptoms into the LE8 model did not significantly improve its predictive performance. Additionally, a weighted LE8 score—constructed using regression coefficients from Cox frailty models—demonstrated slightly better predictive ability compared to the unweighted version, suggesting the potential feasibility of differential weighting to optimize cardiovascular health assessment, although the overall improvement was modest. Moreover, the joint effects of PM2.5 exposure and household air pollution with cardiovascular health status significantly influenced outcome risks, with individuals exhibiting poor cardiovascular health facing particularly elevated risks under high pollution conditions. This study provides empirical evidence supporting cardiovascular health assessment, air pollution mitigation, and integrated intervention strategies in the Chinese population, offering a scientific basis for improving public health policy and precision health management.