The built environment and cardiovascular disease: an umbrella review and meta-meta-analysis

Abstract

Aims

To provide a comprehensive overview of the current evidence on objectively measured neighbourhood built environment exposures in relation to cardiovascular disease (CVD) events in adults.

Methods and results

We searched seven databases for systematic reviews on associations between objectively measured long-term built environmental exposures, covering at least one domain (i.e. outdoor air pollution, food environment, physical activity environment like greenspace and walkability, urbanization, light pollution, residential noise, and ambient temperature), and CVD events in adults. Two authors extracted summary data and assessed the risk of bias independently. Robustness of evidence was rated based on statistical heterogeneity, small-study effect, and excess significance bias. Meta-meta-analyses were conducted to combine the meta-analysis results from reviews with comparable exposure and outcome within each domain. From the 3304 initial hits, 51 systematic reviews were included, covering 5 domains and including 179 pooled estimates. There was strong evidence of the associations between increased air pollutants (especially PM_2.5 exposure) and increased residential noise with greater risk of CVD. Highly suggestive evidence was found for an association between increased ambient temperature and greater risk of CVD. Systematic reviews on physical activity environment, food environment, light pollution, and urbanization in relation to CVD were scarce or lacking.

Conclusion

Air pollutants, increased noise levels, temperature, and greenspace were associated with CVD outcomes. Standardizing design and exposure assessments may foster the synthesis of evidence. Other crucial research gaps concern the lack of prospective study designs and lack of evidence from low-to-middle-income countries (LMICs).

Registration

PROSPERO: CRD42021246580

Graphical abstract

The built environment and cardiovascular disease. IHD, ischaemic heart disease; CeVD, cerebrovascular disease; CVD, cardiovascular disease.

Open in new tabDownload slide

Introduction

Cardiovascular diseases (CVD) are the leading cause of the global disease burden.¹ Worldwide CVD cases have doubled in the past 20 years. In 2019, 523 million cases led to 18.6 million deaths and 34.4 million years lived with disability.² The burden of CVD is not only an individual health issue but also a societal burden that strains healthcare and economic systems. The World Heart Federation estimates that the global cost of CVD will rise from roughly $863 billion in 2010 to $1044 billion in 2030.³ Therefore, it is important to deepen our understanding of the determinants of CVD in individuals and populations and develop sustainable strategies for reduction and prevention.

Lifestyle behaviours, like physical inactivity and unhealthy diet, are important risk factors of CVD.^1,2,4 Ecological models suggest that these behavioural risk factors are driven by contextual characteristics in the social, policy, and built environments, also known as ‘upstream determinants’.^5–7 The paradigm in CVD research has shifted, with the focus moving to these upstream determinants as promising entry points for population-level action to prevent CVD.⁸ The majority of the world’s population resides and spends most of its time in highly organized built environments. The built environment is a subset of the exposome, which is the sum of all environmental drivers of health and disease throughout life.⁹ The built environment is defined as all aspects of a person’s surroundings that are man-made or modified, such as buildings, parks, facilities, and infrastructure.¹⁰ Its direct effects, like air pollution, noise levels, and ambient temperature, are often included.¹¹

The mechanisms by which the built environment might affect CVD are not well established. Conceptually, there are two main pathways proposed.¹² The first pathway is between active built environmental exposure and behavioural risk factors.¹² For active exposure, one needs to actively use of the environment to be exposed. Attributes such as walkability, which is comprised of individual elements like sidewalks, connected streets, and proximity to key destinations, can facilitate a more active lifestyle.^13,14 Access to and availability of certain food resources may either improve or diminish diet quality, depending on whether these food resources are greengrocers or fast-food outlets, for example.¹⁵ The second pathway is between passive built environmental exposure and CVD. This includes exposures that occur when one is simply present in the environment, such as air pollution, residential noise, and ambient temperature.¹² Exposure to increased levels of air pollution can promote systemic inflammation and oxidative stress. As a result, a variety of pathological processes, such as increased thrombosis, hypercoagulability, and endothelial dysfunction, could eventually lead to CVD.^16,17 Noise and ambient temperature may cause typical physiological responses including hypertension, vasoconstriction, and tachycardia that may lead to CVD.^18,19

The two aforementioned pathways are not mutually exclusive, which increases complexity of research in this field.¹¹ Many environmental aspects may be interrelated. For example, the benefits of living in a dense, walkable environment might be diminished by increased exposure to traffic-related air pollution.²⁰ Attributes may also operate at multiple scales or contexts, from neighbourhoods to entire regions or from rural to urban areas. Consequently, the same built environmental aspects may have different effects from different perspectives or contexts. Urbanization, a context indicator of the urban development level, serves as a container of the abovementioned built environmental aspects. It is worthwhile to study urbanization as a proxy of a built environment and its confounding or interaction effect with other aspects. Furthermore, 55% of the world's population lives in urban areas, a proportion that is expected to increase to 68% by 2050.²¹ It is therefore relevant to understand what this means in terms of CVD risk.

Over the past two decades, a significant number of systematic reviews have examined the relationship between the built environment and CVD. However, these studies often address single (sub)domains of exposures, or the built environment was not the primary focus of the reviews. It is difficult for readers to assess and synthesize the piecemeal published evidence. As such, we aimed to provide a comprehensive overview of the current evidence, identify crucial research gaps, and determine the implications for public health, clinical medicine, policy, and regulation. To accomplish this, we conducted an umbrella review of systematic reviews and meta-analyses to investigate associations between the built environment and CVD events in adults. This review may also serve as a reference point for those who are new to the field.

Methods

The current umbrella review was conducted according to the protocol published in the International Prospective Register of Systematic Reviews (PROSPERO; ID CRD42021246580) and adheres to the guidelines of Transparent Reporting of Systematic Reviews and Meta-analyses (PRISMA) (see Supplementary material online, Appendix S1, for PRISMA checklist).

Literature search

We searched seven databases on 16 April 2021: Medical Literature Analysis and Retrieval System Online (MEDLINE), Excerpta Medica Database (EMBASE), Cumulative Index for Nursing and Allied Health Literature (CINAHL), Scopus, the Cochrane Database of Systematic Reviews (CDSR), the Joanna Briggs Institute (JBI) Database of Systematic Reviews and Implementation Reports, and PROSPERO. We built a search algorithm using search terms based on definitions and synonyms of the built environment, its attributes, and definitions of CVD events. A detailed search strategy for each database is presented in Supplementary material online, Appendix S2. We also screened the reference lists of the included reviews to identify additional eligible reviews.

Inclusion and exclusion criteria

We included systematic reviews of primary studies of the general population if they (i) reported on objectively measured long-term neighbourhood environmental exposures and covered at least one domain, including air pollution [e.g. particulate matter (PM), carbon monoxide (CO), nitrogen oxides (NO and NO₂), sulphur dioxide (SO₂), and ozone (O₃)], food environment (e.g. neighbourhood fast-food outlet density), physical activity environment (e.g. greenspace and walkability), urbanization, light pollution (light at night), residential noise from road, rail, and/or air traffic, and ambient temperature; (ii) reported associations between these factors and CVD events [i.e. prevalence, incidence, or mortality of coronary heart disease (CHD), stroke, transient ischaemic attack (TIA), peripheral arterial disease (PAD), atrial fibrillation (AF), aortic disease, and heart failure (HF) but not congenital heart disease) in adults (i.e. ≥ 18 years); (iii) used a systematic literature search, i.e. a reproducible search strategy with search strings corresponding to databases; and (iv) were published between 1 January 2000 and 16 April 16 2021 in English. In the domain of temperature, the studies of short-term and long-term were mixed and meta-analysed in systematic reviews. Therefore, the criteria of long-term exposure were not applied too strictly for temperature.

We excluded reviews if they (i) only focused on specific populations such as children, pregnant women, CVD patients, or patients whose CVD risk may be influenced by other conditions (e.g. cancer, diabetes, and chronic renal failure); (ii) were published as conference abstracts, case reports, editorials, and letters to editors; (iii) reviewed studies on the indoor built environment (e.g. home environment), occupational environment (e.g. workplace environment), or subjective assessments of environmental characteristics (e.g. perceptions of neighbourhood safety); or (iv) only examined acute (short-term) exposure.

Study selection and data extraction

After removing duplicate records, two authors (M.L. and P.M.) screened all titles and abstracts independently. Then, these authors screened the full texts of potentially eligible articles separately and cross-checked a sample of each other’s work. Screening was done using Rayyan software, a non-commercial, web-based application.²² The two authors resolved any disagreements with discussion or, if no consensus could be reached, with discussion with other authors (T.M.L., E.J.T., I.V., and J.L.). Two authors (M.L. and P.M.) conducted the data extraction and verified each other’s work. For each eligible review, they extracted the following information: first author, year of publication, study design, study population, countries in which primary data were collected, exposure domain and type of environmental exposures, measures of the exposures, type and measure of the outcome studied, and summary of the (stratified) results. In the event that an included review was based on a meta-analysis, the following information was also extracted (where available): the pooled effect estimates and 95% confidence intervals (CIs), the effect size (ES) in the study with the largest study sample, the between-study heterogeneity using I²-statistic, the results of the Egger’s regression asymmetry test and excess statistical significance test, and the 95% prediction interval (PI).

Overlap of primary studies assessment

We assessed the overlap of primary studies across included reviews by a measure of corrected covered area (CCA).²³ The first occurrence of a primary study in included reviews was defined as the index study. We created a cross-table of index studies and reviews for each built environmental domain (see Supplementary material online, Appendix S4). The CCA-score was categorized into limited overlap (score: 0–5), moderate overlap (score: 6–10), high overlap (score: 11–15), and very high overlap (score: > 15).²³

Risk of bias assessment

We assessed the risk of bias in included systematic reviews using the validated Risk of Bias in Systematic Reviews (ROBIS) tool.²⁴ Any disagreements in the assessment were resolved with discussion.

Statistical analysis

For the syntheses of quantitative research, we used statistical methods in accordance with the most up-to-date recommendations.^25–28 Specifically, to rate the robustness of evidence for each review that reported pooled results from a meta-analysis, we considered the following statistics:

• Statistical heterogeneity. The extent of statistical heterogeneity was evaluated using the I²-statistic. When the I²-statistic exceeded 50%, heterogeneity was considered large. We also evaluated heterogeneity using the 95% PI. This measure assesses the uncertainty of expected outcomes in new studies of the same association.²⁹

• Small-study effect. The small-study effect refers to the observation that studies that include smaller sample sizes tend to yield larger ES than studies with larger study samples. The potential reasons for this include publication bias, reporting bias, and real heterogeneity. The Egger’s regression asymmetry test was used to assess whether or not a small-study effect was present. A P value <0.10 with a more conservative effect in larger studies was considered evidence of small-study effect.³⁰

• Excess significance bias. This measure was used to evaluate whether there is evidence of an excessive number of studies with statistically significant results in the meta-analysis. It may result from reporting bias and data dredging.³¹ The excess statistical significance test was used, with a P value <0.10 considered to be evidence of excess significance bias.

When these statistics were not available, a re-estimate based on the ES of primary studies in the included review was conducted. In line with up-to-date recommendations, the level of robustness for each pooled result was based on the following criteria²⁸:

• Strong evidence: P value <10⁻⁶ of the pooled estimate of meta-analysis, > 1000 individuals of the total number of participants in the primary studies that were included in the review, P value <0.05 of the largest study in the meta-analysis, I²-statistic <50%, no evidence of small-study effects, no evidence of excess significance bias, and the null value does not fall in the 95% PI.

• Highly suggestive evidence: P value <10⁻⁶ of the meta-analysis, > 1000 individuals in the review, and P value <0.05 of the largest study in the meta-analysis.

• Suggestive evidence: P value <10⁻³ of the meta-analysis and > 1000 individuals in the review.

• Weak evidence: P value <0.05 of the meta-analysis.

Within each category, all criteria had to be met.

We conducted meta-meta-analyses to combine the results from meta-analyses with comparable exposure and outcome.²⁷ We selected meta-analyses with comparable study populations, exposure and outcome assessment methods, and measure of association [relative risk (RR) or hazard ratio (HR)]. By matching meta-analyses, we extracted the unique primary ES included in the pooled analyses. After the removal of duplicates, random-effect meta-analyses were conducted with restricted maximum likelihood approach for the estimation of variance components. The Wald method was used for estimating 95% CIs. Heterogeneity was investigated with the I²-statistic. The Egger’s regression asymmetry test was conducted to examine the small-study effect, and excess statistical significance was tested for. To account for variability of covariate adjustment, we conducted sensitivity analyses by only including primary ES with similar covariate adjustment sets. All statistical analyses were conducted with the metaphor package in R software.^32,33

Results

Literature search results

The literature search identified 3304 unique publications (Figure 1). After screening all titles and abstracts, we excluded a total of 3164 studies. After reading the full texts, we included a total of 51 eligible. Additional details are presented in the PRISMA article selection process flow chart. A full list of included reviews is presented in supplementary material online, Appendix S3.

Characteristics of included reviews

The characteristics and a summary of the results of all 51 included systematic reviews are presented in Table 1. The number of relevant primary studies included in the reviews ranged from 2 to 67. Most reviews (n = 43) were not restricted to specific countries or regions in the search, but most evidence was obtained in Europe (n = 43), East and South Asia (n = 38), and North America (n = 37); relatively few studies were based in South America (n = 10), Middle East (n = 8), Oceania (n = 8), and Africa (n = 6). A variety of designs were used in primary studies, including ecological studies, cohort studies, cross-sectional studies, case-control studies, case-crossover studies, small-area studies, panel studies, and time-series studies. Over one-third of the reviews (n = 20) included only longitudinal studies.

Thirty-three reviews studied the domain of air pollution, 4 studied physical activity environment, 1 studied urbanization, 10 studied residential noise, and 8 studied ambient temperature. There were no systematic reviews found in the domains of food environment and light pollution. Most reviews (n = 46) conducted one or multiple meta-analyses. Seven reviews only conducted narrative syntheses of the primary studies. In total, the meta-analyses summarized 180 pooled estimates of the association between specific built environmental factors and CVD events. The CCA-score for each domain was below 1, indicating only limited overlap of primary studies (see supplementary material online, Appendix S4).

Domain-specific results

Air pollution

Overall, long-term increased air pollution was associated with a higher risk of CVD events, based on a total of 105 meta-analyses. Among them, three reviews contained pooled results classified as strong evidence.^34,36,41 Four air pollutants were associated with a higher risk of CVD events (Table 2). Chen et al.³⁶ found that both a 10 μg/m³ higher ambient SO₂ (HR, 1.005; 95% CI, 1.004–1.007) and CO concentration (HR, 1.017; 95% CI, 1.013–1.022) were associated with an increased incidence of AF. Furthermore, Yang et al.⁴¹ found that a 10 μg/m³ higher ambient NO₂ concentration was associated with an increased risk of ischaemic heart disease (IHD) events (RR, 1.05; 95% CI, 1.04–1.06). Alexeeff et al.³⁴ found that a 10 μg/m³ higher PM_2.5 concentration was associated with an increased risk of incident stroke (RR, 1.13; 95% CI, 1.11–1.15). Among these three reviews that provide strong evidence, Chen et al. and Yang et al. had a low risk of bias, while Alexeeff et al. had a high risk of bias for study eligibility criteria, identification, and selection (Table 2).

Thirteen specific pooled results were supported by highly suggestive evidence (Table 3). Associations were found between a higher ambient PM_2.5 concentration and higher risk of myocardial infarction (MI), IHD mortality, cerebrovascular disease (CeVD) mortality, CVD mortality, and stroke events. Associations were also found between a higher ambient PM₁₀ concentration and higher risk of AF and for NO₂ with CVD, ICH, and CHD mortalities. All evidence was limited by small-study effects, excess significance bias, or high heterogeneity. Only the studies by Alexeeff et al.³⁴ and Zou et al.⁴² had a high risk of bias. Furthermore, 32 pooled estimates were found to be non-significant (Table 3); most of these included a small number of studies (i.e. 2–4 studies).

We identified three specific exposure and outcome analyses suitable for meta-meta-analyses (studies presented in bold in Table 3). Firstly, a meta-meta-analysis of 29 unique primary studies from Alexeeff et al. and Chen et al.^34,51 showed that a 10 µg/m³ increase in PM_2.5 was associated with increased IHD mortality (RR, 1.21; 95% CI, 1.10–1.33; I², 98.16%) (Figure 2) with a large degree of heterogeneity across studies (high I²-statistic and larger 95% PI than the 95% CI) (Table 4). As a sensitivity analysis, we conducted the meta-meta-analysis excluding studies that did not adjust for smoking, which decreased the degree of heterogeneity, but did not affect the results (Table 4 and see Supplementary material online, Figure S1). Secondly, a meta-meta-analysis of 26 unique primary studies from Yang et al. and Chen et al.^41,51 showed that each 10 µg/m³ increase in PM_2.5 was associated with increased CVD mortality (RR, 1.11; 95% CI, 1.09–1.14; I², 74.23%) (Figure 3) with a large degree of heterogeneity across these 26 studies (Table 4). Excluding studies that did not adjust for any lifestyle behaviour did not result in any substantial change in effect estimates (Table 4 and see Supplementary material online, Figure S2). Lastly, a meta-meta-analysis of 14 unique primary studies from Stieb et al. and Atkinson et al.^35,40 showed a non-significant association of NO₂ with CeVD mortality (HR, 1.03; 95% CI, 1.00–1.06; I², 45.64%) (Figure 4). There was no evident small-study effect or excess significance bias in these three meta-analyses.

Five narrative reviews were found on air pollution with consistent results to other reviews (Table 1).^{37–39,52,53}

Physical activity environment

Three reviews studied associations in the physical activity environment domain including eight meta-analyses. All reviews in this domain studied greenspace exposure. Two associations related to greenspace were supported by suggestive evidence (Table 3). Gascon et al. found an inverse association for 10% higher greenspace exposure and CVD mortality; however, their review was at high risk of bias for study identification.⁵⁴ Six pooled estimates were not statistically significant. Among these, Yuan et al.⁵⁵ reviewed the association between greenspace and IHD mortality and Twohig-Bennett et al.⁴³ reviewed the association between greenspace and IHD incidence with no obvious risk of bias, though they comprised a small number of studies (i.e. 2–3 studies).

Urbanization

One review was identified in the urbanization domain.⁴⁴ Angkurawaranon et al. reviewed studies focusing on Southeast Asian populations and included four meta-analyses. Two associations were supported by weak evidence. One pooled estimate showed that urban exposure was associated with higher odds of CHD (OR, 2.48; 95% CI, 1.20–5.11), while the other showed lower odds of rheumatic heart disease (RHD) (OR, 0.31; 95% CI, 0.13–0.76). The other two pooled estimates for stroke and non-specific heart disease were non-significant. The eligibility criteria were not clearly reported in this review, resulting in an unclear risk of bias (Table 3).

Residential noise

In the residential noise domain, 9 reviews were identified, including 30 meta-analyses (Table 3). Cai et al. found strong evidence of an association between 10 dB higher aircraft traffic noise and increased CVD mortality (HR, 1.17; 95% CI, 1.10–1.25), though this review had a high risk of bias in study eligibility criteria, identification, and selection.⁵⁶ Two associations were supported by suggestive evidence. Babisch et al.⁴⁵ found an association between 10 dB higher road traffic noise exposure and higher risk of CHD. Vienneau et al.⁴⁷ found a similar association between traffic noise exposure and higher risk of IHD. Seventeen pooled estimates were non-significant. Among these, Weihofen et al.⁴⁸ reviewed seven studies and found a marginal association between 10 dB higher aircraft traffic noise exposure and higher risk of stroke (RR, 1.013; 95% CI, 0.998–1.028). This review did not suffer from obvious risks of bias, heterogeneity, small-study effect, or excess significance bias.

One systematic review covered multiple domains: air pollution, greenspace, walkability, and noise. Using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) system, Rugel et al. found sufficient evidence of an association between increased noise and higher risk of CVD.⁵⁷

Ambient temperature

In the ambient temperature domain, eight reviews were included. These reviews included 37 meta-analyses (Table 3). Associations between both higher and lower temperature in relation to higher risk of CVD were supported by suggestive evidence (3 reviews), 6 associations were supported by weak evidence, and 22 pooled estimates were non-significant. All results suffered from a high risk of bias. Two narrative reviews were found in this domain with inconsistent results.

In this domain, we identified one specific exposure and outcome analysis that was suitable for a meta-meta-analysis. Bunker et al.⁴⁹ and Moghadamnia et al.⁵⁰ meta-analysed the RR of CVD mortality per 1°C change in temperature. The meta-meta-analysis of 62 unique primary studies showed a significant association between increasing temperature and CVD mortality (RR, 1.04; 95% CI, 1.03–1.04; I², 98.63%) (Figure 5). Both a small-study effect and excess significance bias were evident in this analysis. As sensitivity analyses, we conducted the meta-meta-analysis excluding studies that did not adjust for any confounders (see Supplementary material online, Appendix S7, and Figure S3), did not adjust for air pollution (see Supplementary material online, Appendix S7, and Figure S4), and adjusted only for air pollution (see Supplementary material online, Appendix S7, and Figure S5). These analyses did not substantially change the results of the main analyses.

Discussion

Principal findings

Our umbrella review found strong evidence of an association between increased air pollution, including increased ambient PM_2.5, SO₂, CO, and NO₂ exposure, and CVD outcomes. There was also consistent evidence of associations between increased noise levels and CVD outcomes, including strong evidence for air traffic noise and CVD mortality. An association between increased ambient temperature and CVD mortality was supported by highly suggestive evidence. Review evidence in other domains including physical activity environment, food environment, light pollution, and urbanization in relation to CVD are scarce or lacking. However, the limited available evidence is suggestive of a protective effect of greenspace exposure in terms of CVD mortality.

Strengths and limitations

To the best of our knowledge, this is the first umbrella review to comprehensively summarize the relationship between aspects of the built environment and CVD. Following a pre-specified protocol, we conducted a wide-ranging search across seven databases. The included reviews underwent extensive critical appraisal using a validated risk of bias assessment method. Both methodological quality and statistical evidence were evaluated to determine the robustness of evidence. Finally, meta-meta-analyses were performed where possible to combine all available quantitative evidence.

To contextualize the findings, certain limitations of our umbrella review need to be addressed. Firstly, the search strategy was limited to reviews published in the English language. Therefore, we may have omitted key studies published in other languages. Secondly, we included systematic reviews published between 1 January 2000 and 16 April 2021. The most recently published evidence has not yet been taken up in systematic reviews and is, therefore, not included here. Lastly, many included reviews (and studies to date) report on cross-sectional analyses, which inhibits identification of causality.

Evidence in relation to other studies

Our results were generally consistent with the World Health Organization (WHO) 2021 global air quality guidelines, as well as earlier umbrella reviews investigating air pollution and CVD outcomes.^58–62 Positive associations are consistently found for a variety of pollutants, as well as CVD outcomes. However, the association between PM_2.5 and CVD was the most widely investigated. NO₂ exposure was, thus far, not encompassed in the scope of the WHO guidelines because of the absence of clear quantitative evidence. Our results contribute strong evidence of an association between higher NO₂ exposure and a higher risk of IHD and highly suggestive evidence of an association between higher NO₂ exposure and a higher risk of CVD mortality. While earlier work demonstrated strong evidence of a positive association between PM_2.5 and stroke only in Europe,⁶² our results show that this holds true when non-European countries are included.

This umbrella review indicates that studies of greenspace were heterogeneous in design and exposure assessment, which partly explains the limited combined evidence of an association between greenspace exposure and CVD. For example, greenspace assessment methods included the Normalized Different Vegetation Index (NDVI), percentage of greenspace coverage, and distance to the nearest greenspace. While the first two may be good indicators of the amount of greenspace in certain areas, they do not reflect access to those greenspaces. For distance to the nearest greenspace, the opposite may be true. Therefore, studies of greenspace using different assessment methods are not directly comparable.^63,64 This same issue of limited combined evidence may also hold for other domains like physical activity, food environment, light pollution, and urbanization, although our results did not provide information to support this.

Our findings about traffic noise and CVD events are in line with those from two WHO reports.^46,65 The WHO Environmental Noise Guidelines for the European Region (2015) included a systematic review (included in the current review).⁴⁶ According to an adapted GRADE assessment, high-quality evidence was found on road traffic noise and IHD, and moderate-quality evidence was found for aircraft noise and stroke mortality, and for road traffic noise, IHD mortality, and stroke events. Our study updated the evidence base with strong evidence of an association between air traffic noise and CVD mortality. Furthermore, we observed that the existing evidence on rail traffic noise is still limited.

The 2019 Global Burden of Disease study estimated that global age-standardized CVD disability-adjusted life years (DALYs) attributable to high temperature was 25.59 per 100 000 people.⁶⁶ The heat-related disease burden was the highest for stroke and was greater in regions with a lower socio-demographic index (SDI). However, the observational evidence on specific outcomes and regional differences, as reported in the current review, was insufficient. The effect estimate found in the current meta-meta-analysis might be small (Figure 5), but increasing temperature has a potentially large and increasing population reach.

Mutual confounding and interaction

Many aspects of the built environment may be interconnected, potentially leading to confounding and/or interaction effects. However, most existing reviews and meta-analyses focus on single (sub)domains of exposure. We only found one systematic review that covered multiple domains. This study observed sufficient evidence of an association between higher noise exposure and increased CVD morbidity, after adjusting for traffic-related air pollution.⁵⁷ A meta-analysis by Vienneau et al.⁴⁷ observed an association between higher noise levels and increased IHD which remained robust after including studies that adjusted for air pollution exposure. Despite the substantial amount of studies on the effects of air pollution and temperature on CVD, there is limited research available on their mutual confounding or interaction effects. One systematic review found that long-term air pollution exposure and colder temperatures are independently associated with an increased risk of CVD.⁶⁷ Additionally, another US cohort study found that the association between higher PM_2.5 and CVD was stronger in areas with higher green space, lower O₃ levels, and lower temperatures.⁶⁸ Furthermore, as a proxy for context, urbanization may also have significant confounding or interaction effects with various aspects of the built environment. For example, a Dutch cohort study indicated associations between fast-food restaurant density and CVD in urban areas but not in rural areas.⁶⁹

Possible mechanisms and explanations

Several potential mechanisms have been theorized that may explain the associations found in this study. Exposure to air pollutants may promote systemic inflammation and oxidative stress in the lungs, which eventually increases systemic inflammation and oxidative stress.¹⁷ As a result, a variety of pathological processes could take place that eventually lead to CVD, such as increased thrombosis, hypercoagulability, endothelial dysfunction, atherosclerosis progression, insulin resistance, and dyslipidaemia.^16,17 Pollutants can also induce oxidative stress and inflammation in the central nervous system, especially in the hypothalamus, leading to an imbalance in the cardiac autonomic nervous system.¹⁷ Moreover, smaller size pollutants can directly enter the circulation and damage the cardiovascular system and organs.¹⁷ These mechanisms are better understood for PM exposure than for other pollutants. Carbon monoxide typically hampers oxygenation of tissues via carboxyhaemoglobin production, which has a more significant influence on the myocardium than peripheral tissues.⁷⁰ These mechanisms may explain the more evident associations found for AF. Atrial fibrillation is an early CVD endpoint that can lead to other CVD outcomes in the long term. For residential noise exposure, both direct and indirect pathways have been theorized.¹⁸ The direct pathway is induced by the instant interaction between the acoustic nerve and the central nervous system. The indirect pathway relates to cognitive perception and subsequent cortical activation and emotional response. Both pathways can cause physiological stress responses, and, in the long term, pathophysiologic alterations and CVD. The adverse effect of residential noise has been reported independent of sleep quality and self-reported noise sensitivity. But noise can indeed disturb sleep and subsequently cause reduced immune function and a pro-inflammatory state.²¹ Changes in temperature can cause typical physiological responses that might lead to MI, including tachycardia and increased blood viscosity.¹⁹ Both cold and heat exposure may increase sympathetic activation, which could lead to high blood pressure.¹⁹ Lastly, it is hypothesized that greenspace improves air quality. By reducing the concentration of air pollutants, greenspace may reduce the impact of air pollution on CVD.⁶³ Additionally, greenspaces may facilitate physical activity, thereby promoting cardiovascular health.⁶³

Gaps in current knowledge and suggestions for future research

Future primary studies should investigate the effects of PM₁₀, SO₂, CO, NO, and PM_0.1, especially long-term exposure. As we observed evidence of early-onset CVD (e.g. AF), more studies of other CVD outcomes that occur later (e.g. IHD) are warranted. Research is needed on light pollution. Despite there being multiple primary studies on the food environment, they differ in definition of food environment and vary with regard to buffer size. As such, these studies cannot be meta-analysed. Standard design and exposure assessments should be developed and implemented to foster synthesis of evidence. More studies are also needed for residential noise and ambient temperature, especially from low-to-middle-income countries (LMICs). Future primary studies should preferably use prospective study designs in the domains of physical activity environment, urbanization, and residential noise to increase causal inference power. Furthermore, primary studies on interactions or interrelationships of built environmental aspects in relation to CVD are also warranted.

As for systematic reviews, future work should provide evidence on O₃ that is of higher quality. Review evidence on the physical activity environment, food environment, and urbanization was limited. There are primary studies on these domains, but future reviews to harmonize research data (e.g. in terms of the heterogeneity in the definition and measurement of exposures) and synthesize evidence are required. For example, in terms of greenspace, accessibility and composition may be crucial to determining its effects.⁷¹ Furthermore, more high-quality review evidence is needed on residential noise and ambient temperature, especially for LMICs. It is highly recommended that future systematic reviews on the built environment adhere to the PRISMA reporting guidelines and pre-register a protocol in a dedicated registry such as PROSPERO to guarantee transparency, prevent overlap in topics, and promote methodological quality. Furthermore, the use of quality assessment should become common practice, using standardized, validated tools, as recommended by sources such as PRISMA and Cochrane.^72,73 We also observed that most primary studies and reviews focus on one single built environment aspect. However, these aspects co-exist and act together. Accounting for multiple elements of the built environment may provide a better understanding of the relative contribution of each aspect to CVD risk. Systematic reviews should also consider comparing single-exposure to multi-exposure models to examine potential confounding effects between built environment exposures.

Implications for public health

There was strong evidence of a detrimental effect of air pollution on CVD. Therefore, it is important that governments and individuals take measures to reduce outdoor air pollution. The WHO has developed the 2021 long-term air quality guideline levels with interim targets.⁶¹ Each country should establish air quality standards with adapted interim targets, specified measurement, and statistical procedures to reduce CVD events attributable to air pollution. Another strategy could be the development of a framework for clinical recommendations and individual exposure mitigation strategies. The American Heart Association has stated such individual strategies to prevent air pollution, like using portable air cleaners, high-efficiency heating ventilation and air conditioning systems, and N95 respirator.⁷⁴

We found suggestive evidence of a protective effect of greenspace on CVD. This supports the call for urban planners, city designers, project developers, and policymakers to take greenspace into account.⁷⁵ There was also clear evidence of increased residential noise being associated with a higher CVD risk. The WHO developed guidelines pertaining to noise levels for the European regions in 2015.⁴⁶ However, these recommendations may not be directly applicable to other regions. One general public health approach is to address the major sources of pollution (i.e. combustion, traffic, industry, and power generation). Another approach is to develop healthy building guidelines (e.g. including advanced insulation) to reduce the exposure to outdoor air pollution, light pollution, residential noise, and extreme temperatures.

Conclusion

This umbrella review provides a comprehensive overview of the observational evidence on the built environment and CVD. We found strong evidence of associations between increased air pollution, especially PM_2.5, and increased residential noise with higher risk of CVD. The present meta-meta-analysis produced highly suggestive evidence of an association between temperature and CVD. Evidence on physical activity environment, food environment, light pollution, and urbanization is scarce or lacking. Our identification of several knowledge gaps and methodological limitations in the current literature may improve and inform future research and contribute to a better understanding of the effect of the built environment on CVD.

Author contributions

M.L., P.M., and J.L. designed the study and developed the review question. M.L. and P.M. performed the literature search and were the primary reviewers. T.M.L., E.J.T., I.V., and J.L. acted as additional reviewers where required. E.J.T., D.E.G., J.W.J.B., I.V., and J.L. supervised the study. M.L. and P.M. drafted the manuscript. All authors reviewed and revised the protocol and manuscript. J.L. acts as guarantor. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

Supplementary material

Supplementary material is available at European Journal of Preventive Cardiology.

Funding

M.L. had financial support from China Scholarships Council; all other authors had financial support from NWO Gravitation grant Exposome-NL (Grant/Award Number: 024.004.017). The funders didn’t have any role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the article for publication. We confirm the independence of researchers from funders and that all authors had full access to all of the data (including statistical reports and tables) in the study and can take responsibility for the integrity of the data and the accuracy of the data analysis.

Data availability

The data underlying this article are available in the article and in its online supplementary material.

References

Author notes