Observational studies investigating the association between particulate exposures in the workplace and cardiovascular outcomes, including both clinical (fatal and non-fatal) and intermediate outcomes were identified in PubMed (www.ncbi.nlm.nih.gov/pubmed) and EMBASE (www.embase.com) by searching free text and key words (Appendix). Search terms for occupational particulate exposures included “particulate matter” as well as specific types of PM exposures (silica, styrene, diesel exhaust, asphalt fumes, and metal or welding fumes), which contain particles in the fine or ultrafine size range [20–24]. The literature search was restricted to articles published in the English language from January 1, 1990 through April 2009. We also used secondary references cited by the articles identified in the primary search.

Because few mortality studies specifically investigated occupational PM exposures and CVD mortality, we chose to include general all-cause mortality studies where CVD was among the cause-specific associations presented. However, we excluded mortality studies that explicitly stated that the objective was to investigate an outcome other than CVD (e.g., cancer or respiratory disease) because not all such studies presented CVD mortality in findings. Hence, even if CVD mortality were presented in the findings, such studies were excluded. In addition, we excluded studies set in industries or among occupations for which PM was not the exposure of interest, or in which exposure to PM was not the predominant exposure. We also excluded publications that did not contain original research (e.g., reviews, editorials, and letters), studies not carried out in humans (e.g., animal and other experimental studies), and case reports and case series.

Additionally, for studies conducted in the same cohort with the same outcome and exposure, we chose the cohort with largest sample size. We included studies that re-analyzed data in the same cohort to identify specific components of PM responsible for health effects and also included studies that aimed to identify effect modifiers.

For each study that met the study criteria, we extracted information on the study characteristics (authors, year of publication, country, study design), population characteristics (inclusion criteria, age, industry, gender), exposure assessment, outcome ascertainment, and measures of association. Measures of association extracted or derived from the published data were standardized mortality ratios (SMR), incidence rate ratios (IRR), odds ratios (OR), proportional mortality ratios (PMR), regression coefficients (ß), and their standard errors. Where there were more than one analytical comparison group (i.e., external and internal), both measures of association were extracted. For mortality studies, data were extracted for the major categories of IHD (ICD–9 410–414), cerebrovascular disease (ICD–9 430–438), and overall CVD (ICD–9 390–459) when presented. For morbidity studies, we extracted data on any cardiovascular outcome presented, and for intermediate outcomes we investigated measures including systemic inflammation, blood pressure, cardiac abnormalities, heart rate and heart rate variability. Additional study characteristics, which may be used to assess the strengths and weaknesses of each of the mortality studies, were extracted and are included in the Appendix.

For mortality outcomes, pooled estimates of the SMR and IRR, “meta-SMR” and “meta-IRR”, respectively, with associated 95% confidence intervals were obtained when the study designs and contrasts were comparable. Formal tests for heterogeneity were conducted; however all meta-analyses were performed using random effects meta-analysis of the natural log transformed effect estimates regardless of the test result. For each of the major cardiovascular mortality outcomes, sub-group analyses were performed by type of particulate exposure where possible (SMR studies only). All analyses were performed with STATA Version 10.1 (College Station, TX, US). Meta-analyses were not performed for the morbidity studies or studies of intermediate outcomes because of the non-comparability in study designs.

Because of the small number of studies available for meta-analyses, no further analyses were performed to assess the influence of individual studies on the overall pooled estimates, from either the main analyses or sub-group analyses. Also because of the small number of studies, publication bias was not assessed formally.

Figure 1 presents the selection process for identifying original articles for the systematic review of occupational particulate exposures and CVD. The literature search identified 697 citations (Figure 1), of which a large majority was excluded based on review of the title and abstract. The full text of 58 articles were retrieved and reviewed, and 21 articles were subsequently excluded, including 6 studies of particulate exposures not considered to be germane to this review. A total of 37 articles were identified for review: 12 mortality studies; 5 studies considering morbidity due to CVD, two of which also included mortality [25,26]; and 20 studies concerning intermediate cardiovascular outcomes in relation to acute (17 studies) and/or chronic exposure (4 studies).

Study characteristics from each of the twelve mortality studies are presented in Table 1. All but two of the mortality studies [27,28] had a specific objective to study a cardiovascular endpoint. The studies were conducted in a large range of industries and occupations, such as gold mining, trucking, and synthetic rubber industry workers, and others. Mortality due to overall CVD was assessed in seven studies [27,29–34], IHD in ten studies [28,30–33,35–38], and cerebrovascular disease in five studies [29,30,31,35,38] (Table 1). The sample sizes were large, ranging from 3,431 to 176,309 workers, excluding external and internal control groups, and were primarily male. Almost all studies used an external reference group for statistical comparison (Table 2), though some used an internal reference only [32,33,38] (Table 3), and some used both [28,29,34,35,37] (Table 3). Half of the studies estimated exposure [30,32,33,35,37,38] however few actually presented estimates of exposure [32,35,37]. Exposure-response relationships using continuous exposure was presented in only one study [35]. Most studies obtained cause of death from death certificates [27,30,31,33,36–38] (Appendix) and most coding was performed with the International Classification of Disease (ICD) system, with the exception of a study of ceramics workers in China exposed to silica dust [29]. Additional study characteristics are presented in a table in the Appendix.

Studies that used external comparison groups standardized the effect estimates (SMRs) to typical population characteristics such as age and calendar year and in some cases race. Effect estimates varied widely across the cohorts and by major outcome (Tables 2). The SMR ranged from 0.90 [37] to 1.41 [31] for IHD; 0.69 [31] to 1.09 [35] for cerebrovascular disease; and 0.73 [31] to 1.06 [29] for overall CVD, with some statistically significant positive associations observed for IHD [31,35,36], none for cerebrovascular disease, and none for CVD. Significantly increased risk of IHD mortality was observed in miners and other workers exposed to silica and trucking industry workers exposed to diesel exhaust particles [31,35,36]. While studies that used an external comparison groups could not control for confounding by co-exposures or other risk factors for CVD such as smoking, one study of asphalt workers, in which a non-statistically significant increase in risk of IHD was found, compared smoking rates in a subset of the study population finding that they were similar to rates in the general male population, suggesting that confounding by smoking habits was minimal [30].

The SMR from seven studies that presented findings for IHD were combined [28,30,31,35–37], giving a slight but non-significant, excess in deaths due to IHD (meta-SMR = 1.09; 95% CI: 0.92, 1.30) (Figure 2). No excess in deaths due to cerebrovascular disease were found when the SMRs from four studies were combined [30,35–37] (Figure 2) or for overall CVD when the SMRs from four studies were combined [27,29–31] (Figure 2). There was statistically significant (p < 0.05) heterogeneity between studies in the three meta-analyses.

In sub-group meta-analyses, we were able to perform analyses for silica and styrene only (Figure 2). Excess deaths due to IHD mortality were observed in the silica-exposed cohorts [35,36] while a decreased risk of IHD mortality was found in the styrene exposed cohorts [28,37]. No associations were observed for silica and cerebrovascular disease or overall cardiovascular disease in the sub-group meta-analyses.

The study characteristics and major results of the five morbidity studies are presented in Table 4. All but one study, which evaluated IHD and overall CVD prevalence [40], specifically assessed incidence or prevalence of MI [25,26,41,42]. The study populations were also relatively large, ranging from 2,993 to 153,807 individuals and mostly male. All of the study populations were drawn from the general population. In general, hospital data were used to ascertain outcome. Job title [25,40,42] and exposure estimates [26,41] (lifetime average intensity and duration) were used to assign exposure. Several studies adjusted for smoking as a confounding factor of cardiovascular morbidity [40,41]. One study collected information on smoking but it was unclear if this was adjusted for in the analysis [26]. In the study of occupations exposed to traffic-related particles [25], the authors indirectly evaluated the impact of smoking and body mass index on the association between occupations exposed to PM and MI using varying distributions.

In a cross-sectional study, IHD and CVD prevalence rate ratios among transport workers compared to the general employed population were below unity and null [40]. While smoking and age were appropriately controlled for, the study included individuals as young as 16 years of age, who would be less likely to have any type of CVD. In another cross-sectional analysis, the prevalence of MI was found to be increased in individuals with long-term exposure (greater than five years) to welding and soldering fumes compared to those without long-term exposure. However smoking and other potential confounders did not appear to be controlled for in the analysis, although the authors did report collecting information on smoking [26]. In this same study, the incidence of MI was also studied prospectively finding an elevated risk of MI in individuals exposed to soldering fumes long-term, but not those exposed to welding fumes. Again it was unclear whether potential confounders were controlled for in the analysis.

In the three case-control studies that investigated the incidence of MI, positive effect estimates were observed in almost all seven comparisons that were included in this review though not all associations were statistically significant (Table 4). In a study of urban males employed in occupations exposed to traffic-related particles, statistically significant effect estimates for risk of MI ranged from 1.31 to 1.53 for bus drivers, taxi drivers and long distance lorry drivers as compared to gainfully employed men in the general population while controlling for age and other potential confounders. Smoking and BMI were not adjusted for, but the authors were able to indirectly evaluate the impact of these factors, finding that smoking and obesity were unlikely to account for all of the increased risk in MI, especially in taxi and bus drivers.

In a subsequent study by the same authors [41], male and female workers exposed to high levels of respirable combustion-related PM as determined from a lifetime occupational history questionnaire experienced a more than two-fold increased risk of MI compared to unexposed individuals, while controlling for smoking and a number of other covariates, including age, diabetes and overweight. The authors did control for hypertension however, a factor which may be on the causal pathway between exposure and MI and which therefore may attenuate the effect estimate when included in the model. In the third case-control study by Bigert et al. [42] of subway drivers exposed to underground particulate matter, no increased risk of MI was found when workers were compared to gainfully employed men in the general population or other manual workers. The authors were only able to control for age and calendar year in this analysis.

Control for confounding factors and use of appropriate comparison groups (i.e., gainfully employed individuals as opposed to the general population) was in general better in these morbidity studies. However the lack of information on co-exposures such as CO and direct measurements of particulate exposures to estimate long-term exposure and evaluate exposure-response relationships remain as important limitations to these studies, which otherwise suggest an association between occupational PM exposure and MI.

The major study characteristics and results for the 20 studies concerning intermediate cardiovascular outcomes in relation to acute (17 studies) (Tables 5 and 6) and/or chronic exposure (4 studies) (Table 7) are presented in Tables 5–7. The studies were conducted in underground railroad workers [43], boilermaker construction workers [44–53], highway toll collectors [54], vehicle maintenance workers [55], highway patrol officers [56], traffic policemen [57], automobile mechanics [58], and a panel of healthy volunteers exposed to welding fumes [59]. Three studies focused on identifying effect modifiers [44,50,51] (Table 5) and four re-analyzed data to identify PM_2.5 sources and components responsible for observed associations [47,53,60,61]. In general, the studies of intermediate cardiovascular endpoints were conducted in small sample sizes (9 to 79 participants) using either short-term prospective study designs (Tables 5 and 6) with repeated measurements or cross-sectional (Table 7) designs, and were mostly male. A number of intermediate cardiovascular outcomes were studied—most commonly heart rate variability (HRV) (Table 5) and circulating markers of systemic inflammation (Table 6). Direct personal measurements of airborne PM concentrations were used to assign exposure in the boilermaker construction workers [44–53], vehicle maintenance workers [55], and highway patrol officers [23], while controlled exposures were studied in the panel of healthy volunteers exposed to welding fumes [59]. Job title was used in the other cohorts which were studied using a cross-sectional design [43,54,57,58] (Table 7). Because of the smaller number of participants, which generally makes it more feasible to collect more detailed information on a larger number of covariates, control or evaluation of important potential confounders such as smoking was conducted in all of the studies of intermediate cardiovascular outcomes, either through study design or in analysis. In addition, a number of other important potential confounders were considered in the vast majority of these studies, including co-pollutants such as CO [55,56].

Heart rate variability (HRV), a widely-used indicator of cardiac autonomic nervous system function measured by ambulatory electrocardiogram (ECG), was measured in four cohorts (boilermakers, healthy volunteers exposed to welding fumes, vehicle maintenance workers, and highway patrol officers). HRV can be characterized using a number of indices (e.g., short duration measures such as the 5-min SDNN and long duration measures such as the SDNNi), reflecting different components of the autonomic nervous system. Decreased HRV is associated with increased risk of mortality in the general population and the development of non-fatal cardiac events [63,64]. Both the HRV indices as well as the conditions for ECG ascertainment varied between these studies (Table 5). Findings also varied by cohort.

In the boilermaker cohort, an inverse association between increasing PM_2.5 exposure from welding fume and residual oil fly ash (ROFA) and decreasing HRV was first reported in 2001 using a repeated measures study design [52]. This study found a 2.66% (95%CI: −3.75% to −1.58%) decrease in 5-min SDNN for every 1 mg/m³ increase in the personal 4-hour moving average PM_2.5 concentration, while controlling for smoking, age, and diurnal variation, suggesting an acute adverse effect of PM_2.5 on the autonomic nervous system, a potential mechanism by which PM may be associated with adverse clinical cardiovascular events. Subsequently, consistent inverse associations between personal PM_2.5 measurements and other measures of HRV were observed in additional boilermaker studies while controlling for smoking, age, and diurnal variation [46,48,51]. These studies strongly suggest acute adverse cardiovascular health effects of PM_2.5 exposure from welding fumes and residual oil fly ash. Presence of co-pollutants was reported to be minimal and uncorrelated with PM_2.5 in these cohorts, providing further support that the observed effects were due to PM_2.5. Re-analyses of some of these studies using PM_2.5 components showed that nighttime root mean square successive differences (rMSSD) in heart period series, a high-frequency component of HRV that reflects the relative influence of the parasympathetic nervous system, was inversely associated with manganese [47]. In an earlier study, the mean of the standard deviations of all normal to normal intervals (SDNNi) for all five minute segments of the ECG recording, an HRV measure that reflects overall contributions of the parasympathetic and sympathetic nervous system, was positively associated with lead and vanadium [53], a finding that suggests that these PM_2.5 metal components are not the responsible constituents for observed deceases in HRV with increasing PM_2.5 [53]. Additional studies in this cohort showed that obesity, poorer cardiovascular disease risk profile as measured by the Framingham heart score, as well poorer vascular function and greater systemic inflammation aggravate the PM_2.5-HRV association up to four-fold [44,50,51].

In contrast to the boilermaker studies, using a repeated measures study design, associations between PM_2.5 and a number of HRV indices immediately after a work-shift were generally null in non-smoking highway patrol officers [56], though positive associations with a number of HRV parameters were observed the morning after work (e.g., 11.7% change in next-morning SDNN per 10 μg/m³ increase in PM_2.5, p = 0.006) (Table 3). Because the study participants were young and in good physical health, the authors suggested that the positive associations may reflect a healthy physiologic response to PM exposure in comparison with older and less healthy individuals and workers. The effects of co-exposures from CO, NO₂, and relative humidity were assessed, giving minimal changes in the results, but because of the correlation with PM_2.5 potentially confounding effects from these co-exposures could not be entirely ruled out. In subsequent analyses, PM_2.5 associated with vehicular “speed change” (e.g., engine emissions and break wear) which include the components of copper and sulfur were associated with the increased HRV indices in the patrol officers [61]. In studies of healthy volunteers exposed to welding fumes and in vehicle maintenance workers [55,59] null associations between exposure and HRV were also observed..

Circulating markers of inflammation, coagulation, and vascular function were investigated in the cohorts of underground railways workers [43], boilermaker construction workers[49,62], healthy volunteers exposed to welding fumes [59], and highway patrol troopers [56] in relation to short-term PM exposure. The types of markers measured and their associations varied from study to study, however positive associations with inflammatory markers were more consistently observed than HRV across cohorts. For example, C-reactive protein (CRP), a commonly measured biomarker of systemic inflammation, was measured in four studies [43,49,56,59]. A positive baseline to next morning change (24 hours) in CRP (mg/L) was observed in boilermaker construction workers exposed to welding fume PM_2.5 for approximately six hours, after adjusting for smoking and time of day (β = 0.95; 95% CI: 0.23–1.67) [49]. Similarly, in non-smoking highway patrol officers exposed to in-vehicle PM_2.5 during a full work shift, CRP (mg/L) measured 14 hours after a shift increased 32% (p = 0.02) per 10 μg/m³ increase in PM_2.5 [56]. No associations however were found in healthy volunteers exposed to welding fume using a crossover study design [59] or in underground railroad system workers with measurements taken two days apart [43]. The short duration of exposure in the healthy volunteers and/or the timing of the outcome measurements may explain the lack of effects observed in this study. An increase in absolute neutrophil counts from baseline to post-shift (× 10³/μl), another indicator of systemic inflammation, was also associated with increasing PM_2.5 exposure in non-smoking boilermakers (β = 0.3; 95% CI: 0.02–0.6) [49]. In the non-smoking highway patrol officers, a 6.2% increase in percent neutrophils per 10 μg/m³ increase in PM_2.5 was observed 14 hours post-shift [56]. The increase in percent neutrophils in the highway patrol officers was also linked with copper and sulfur in PM_2.5 from vehicular “speed change”. In contrast, to these findings, no changes were observed in circulating levels of inflammatory cytokines (tumor necrosis factor [TNF]-α, interleukin-6 [IL-6], or interleukin-8 [IL-8]) in healthy volunteers exposed to welding fumes for two hours [59], which may be due to the short duration of exposure and/or timing of outcome measurements.

Soluble adhesion molecules, which relate to both inflammation and endothelial function, were also measured in relation to PM_2.5 exposure in some studies [59,62]. Soluble inter-cellular adhesion molecule-1 (sICAM-1) was measured in boilermakers [62] and healthy volunteers exposed to welding fumes [59], but no associations with exposure were observed in either cohorts. An acute decrease in soluble vascular cell adhesion molecule-1 (sVCAM-1) however was associated with increasing PM_2.5 exposure in boilermakers [62]. This may reflect an initial down-regulation followed by an up-regulation in VCAM-1 as explained by the authors. Von Willebrand factor (vWF), a coagulatory molecule, also related to endothelial function and inflammation, was found to increase in the highway patrol officers in relation to increasing personal PM_2.5 exposures [56]. An acute increase in vWF across a work-shift was also observed in the cohort of boilermaker construction workers in association with personal PM_2.5 levels [62]. Higher levels of vWF were also observed in healthy volunteers after exposure to welding fumes as compared to a control day, though this difference was only marginally statistically significant [59].

Plasminogen activator inhibitor-1 (PAI-1) and fibrinogen, which characterize a pro-thrombotic state, were measured in the underground railroad system workers, and acute increases were observed in the low-exposed workers but not the higher exposed workers [43], which could indicate a greater reactivity to occupational PM among lower exposed individuals who may have a lower baseline inflammatory state. In the healthy volunteers exposed to welding fumes, however, no differences in fibrinogen levels were observed under exposed conditions versus unexposed conditions [59].

In the boilermakers cohort, acute vascular effects of PM exposure were also assessed non-invasively with aortic pulse wave analysis derived from radial waveforms [45]. An acute alteration in vascular function was observed in relation to PM_2.5 exposure as measured by the augmentation index (AIx), a correlate of arterial stiffness, while adjusting for smoking, age, and diurnal variation. Increasing AIx is an independent marker of coronary artery disease and correlates with risk of developing coronary artery disease [65,66].

To assess the chronic effects of exposure, four cross-sectional studies investigated the association between exposure as measured by job title and levels of circulating inflammatory markers [43], cardiovascular abnormalities [57], and measures of pre-clinical atherosclerosis [54,58] while controlling for smoking and a number of potentially important confounders (Table 7). In underground railway workers with relatively high exposure to PM_2.5, increased baseline levels of PAI-1 and CRP were found as compared to low-exposed workers, suggesting a chronic systemic inflammatory effect of PM exposure [43]. In traffic policemen exposed to traffic-related PM, increased resting diastolic blood pressure, alterations in the ST segment of the electrical activity of the heart and arrhythmias with exercise were observed in comparison with office workers [57]. Increased homocystiene levels and decreased ocular blood flow velocity, indicators of pre-clinical atherosclerosis were observed in toll collectors versus non-exposed controls in Turkey, suggesting an effect of chronic PM exposure on the development of atherosclerosis [54]. Increased plasma susceptibility to oxidation in automobile mechanics as compared to white collar further supports a potential effect of chronic occupational PM exposure on pre-clinical atherosclerosis [58]. While these studies controlled for a number of potential confounders, a limitation was the lack of data on co-pollutants such as CO.