The Long-Term Effect of Exposure to Particulate Matter Air Pollution on the Incidence of Myocardial Infarction: A Systematic Review and Meta-Analysis

This study systematically reviews the long-term impact of exposure to particulate matter (PM) air pollution with aerodynamic diameter ≤ 10 µm on the incidence of myocardial infarction (MI). The relevant databases were searched with appropriate keywords on February 29, 2020. A random-effects model through a generic inverse-variance method was used to calculate the pooled hazard ratio (HR) and 95% condence interval (CI) of MI. The number of 17 cohort studies with more than 18 million participants and 800,000 cases of MI were included. A signicantly higher risk of MI was observed per 1 µg/m 3 increment of PM with aerodynamic diameter ≤ 10 µm (HR= 1.02,95 % CI = 1.01, 1.03). Subgroup analysis according to the study population indicates subjects with cardiovascular diseases history had a signicantly greater risk of MI per 1 µg/m 3 increase in PM with aerodynamic diameter ≤ 10 µm level (HR= 1.05,95% CI= 1.01, 1.08). Subgroup analysis according to aerodynamic diameter of PM showed only a signicant stronger risk of MI per 1 µg/m 3 increase in PM with aerodynamic diameter < 2.5 µm (HR= 1.01,95% CI= 1.00, 1.02). The pooled result conrms a signicant association between the long-term exposure to PM air pollution and the developing of MI.


Introduction
Air pollution is a growing challenge in the world that seriously threatens health of people. According to the World Health Organization (WHO), more than 90% of the world's population, especially in low-income and middle-income countries lives in areas with inappropriate air pollution quality 1 . Although Air pollution is a complex cocktail of chemicals, more attentions are focused on particulate matter (PM) component mainly due to its high toxicity for lung and cardiovascular system 2 . According to recent data, from 1990 to 2017, global age-standardized summary exposure value for ambient PM air pollution raised by 41.2% averagely making it the rst environmental risk factor in 2017 3 .
A large number of adverse health effects can be associated with PM air pollution. Previous epidemiological studies indicted exposure to PM air pollution increased risk of lung disturbances involving mainly chronic obstructive pulmonary disease (COPD) 4 , asthma 5 , and lung cancer 6 and cardiovascular diseases, including heart failure 7 , hypertension 8 , arthrosclerosis 9 , arrhythmia 10 , carotid intima-media thickness 11 , and stroke [12][13] . Furthermore, metabolic dysfunctions such as diabetes [14][15] , neurological disorders 16 , including Parkinson's disease, dementia, Alzheimer's disease depression and Autism spectrum disorder (ASD) were reported. What is more, some studies indicated PM can be associated with the stronger risk of depression 17 and kidney disease 18 . The Other studies showed a signi cant association between PM air pollution and incidence of cancers 6, 19 . What is more, both short and long-term exposure to PM air pollution were showed increased risk of mortality and largely respiratory and cardiovascular mortality [20][21] . From 1990 to 2017, the number of deaths attributed to PM air pollution was increased nearly by 68 % and reached to about 5 % of all deaths 3 .
To current date, many studies have assessed the short-term effect of exposure to PM on MI. According to recent metaanalysis studies, short-term exposure to PM can be signi cantly associated with the stronger risk of MI [22][23][24][25] . However, the long-term effect of exposure to PM on MI is inconclusive or even contradictory. Although some studies showed the greater risk of MI per increase in PM levels 26-33 , others did not nd a signi cant association [34][35][36][37][38][39][40][41][42] . We designed this systematic and meta-analysis with the aim of synthesizing the current evidence to investigate the long-term association between PM air pollution and the incidence of MI.

Protocol and registration
We acted upon the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement 43 . Also, the elaboration and explanation of PRISMA statement were considered 44 . This study was approved by Research Deputy and Technology of Kermanshah University of Medical Sciences.

Eligibility criteria
Study population was de ned any subjects whose exposure to PM determined and then followed by time for MI occurrence. The hazard ratio (HR) of both primary and secondary MI de ned by ICD-9 code 410 or ICD-10 codes I21 to I23 was considered outcome. It should be noted we excluded studies with including ICD-10 codes I20, I24, and I25 under name of ischemic heart disease (IHD) or coronary heart disease (CHD) if did not provide an independent HR for MI. Given that the focus of this study was in clarifying the long-term effect of PM on MI, we included only prospective cohort studies. No publication date restriction was considered, however, we entered studies with English language.

Information sources
On July 14, 2019, we searched PubMed (https://www.ncbi.nlm.nih.gov/pubmed/), Scopus (https://www.scopus.com/), and Web of Science (https://www.webofknowledge.com/) to trace any relevant publication. The last updating for new articles was performed on February 29, 2020. We also screened references of included studies to nd any relevant publication that may not be detected by the searching.

Search
We searched the mentioned databases by appropriate keywords and terms such as "myocardial infarction" ,"heart attack", "acute coronary syndrome" as outcome, and "air pollution", "air pollutants", "particulate matter", and "PM" as exposure. Then, Endnote les of records were obtained for further assessment.

Study selection
As illustrated in Fig. 1, after removing duplicate records, we screened documents by title and abstract then excluded studies did not investigated the association. Next, we retrieved the full text of potentially relevant studies and examined them carefully. According to the eligibility criteria, at rst, studies did not investigated PM as an exposure or MI as an outcome were omitted. Second, studies assessed the short-term effect of air pollution on MI were removed. Third, we discarded studies assessed the effects of PM on MI mortality or studies mixed the incidence and death of MI. Forth, we removed studies with including ICD-10 codes I20, I24, and I25 under name of ischemic heart disease (IHD) or coronary heart disease (CHD) because did not provide an independent HR for MI. Finally, book chapters, conference papers, editorials, and review articles were eliminated.

Data collection process
We assessed carefully all included studies. The main characteristics of studies recorded in Microsoft excel version 2013.
During the process of data collection whenever there was a problem, we solved it with team-based discussion or consulted with other available researchers or the relevant books.

Data items
The following items were extracted from each study: study ID (name and publication year), country that study was conducted, characteristics of study population (age, sex, and sample size), PM measurement methods and types of Page 4/24 them, the way of diagnosing MI, and the estimated HR and 95% CI(con dence interval)of MI with adjusted variables (see Table 1).

Risk of bias in individual studies
We assessed the selected studies in terms of risk of bias. The Newcastle-Ottawa Scale (NOS) developed for investigating the quality of cohort studies was used 45 . The NOS is largely used due to recommendations from the Cochrane Collaboration 46 . This tool has three broad categories, including selection (four criteria), comparability of study groups (two criteria), and assessment of the outcome (three criteria). The scale ranged between 0 (the lowest study quality) and 9 (the highest study quality) points. The nal quality score according to the NOS was provided in the Table 1 for each included study.

Summary measures
We included only prospective cohort studies. All studies were reported HR and 95% CI for developing of MI per speci c unit increase in PM levels. For estimating the total effect of exposure to PM on the risk of MI, because some studies reported HR (95% CI) for PM 10 , PM 2.5, or/and PM 2.5−10 30, 35, 39 , we only included the estimated risk per PM 10 due to overlapping. Moreover, we found two studies reported HR of MI for different PM diameters on the same population [40][41] .
For total effect analysis, we only considered one of them reported HR of MI for PM 10 40 but for subgroup analyse according to PM diameters we included both studies [40][41] . Similarly, there were two studies on the same population and PM diameter [27][28] . We rst used xed effects meta-analysis model, then the estimated pooled HR (95% CI) was considered. In the present study, we assumed a linear relation between PM air pollution and MI. If applicable, we standardized the risk of MI per 1 µg/m 3 increment of PM air pollution by using the following formula:

Synthesis of results
The Random-effects model through a generic inverse-variance method was used to calculate the pooled HR (95%CI) for incidence of MI 47 . Heterogeneity presented with calculated I 2 index and I 2 values of 0%, 25%, 50%,and 75% represent no, low, moderate, and high heterogeneity, respectively 48-49 . We used the Jackknife approach to investigate the impact of each study on the pooled effect size and the heterogeneity across studies 52 . Besides, Subgroup analyses according to gender (female or both male and female subjects, study population (subjects without MI history and those with MI/CHD history), and diameter of PM (PM 2.5 , PM 2.5−10 , and PM 10 ) were performed. P-value of less than 0.05 was chosen to test null hypotheses in all analysis. Stata software version 14 was used to data analysis.

Risk of bias across studies
The Egger's and bagger's tests employed to investigate publication bias [50][51] . The p-value less than 0.05 chosen to test a signi cant publication bias across studies. In addition, Visual inspection of funnel plot was presented.

Study selection
Flow diagram of systematic review process was presented in the Fig. 1

Study characteristics
As provided in Supplemental Table 1, we included the number of 17 prospective cohort studies with more than 18 million participants and 800,000 cases of MI. The large number of the included studies (10 out of 17) was performed in North America region (the eight studies in the United States of America and the two studies in Canada). Besides, there were four studies came from Europa region (two studies from the United Kingdom, another study from Netherlands, and the other study from Italy), two studies from East Asia (one study from Korea and another study from Japan), and one study from Middle East (Israel). All studies carried out on adult subjects (the youngest age was ≥ 18 years). Most of the studies included both female and male subjects but in the four studies participants had female gender. In the 14 studies only subjects without CVDs included but in the three studies participants had previous history of MI or CVDs. Most studies used the different models to determine the rate of exposure to PM for the subjects (12 studies). By considering the aerodynamic diameter of particulates, we observed in the nine studies only PM < 2.5 µm investigated and in the three studies only PM ≤ 10 µm considered. Furthermore, in the other three studies assessed both PM < 2.5 and PM ≤ 10 µm. Moreover, one study investigated PM < 2.5 and PM 2.5-10 (Coarse) and in the other in addition to PM < 2.5 and PM 2.5-10 µm included PM ≤ 10 µm. the majority of included studies diagnosed MI based on medical records registers by ICD-9 codes 140 (three studies) and/or ICD-10 codes I21(three studies) or I21-23(four studies). All studies provided the adjusted hazard ratio (HR) and 95 % con dence interval (CI) of per speci c unit increment of PM.

Risk of bias within studies
The results of risk of according to the NOS tool indicates all included studies have good quality (all studies met 3/4 items of Selection part, 1/2 items of Comparability par, and 2/3 items of Outcome part).

Results of individual studies
The results of studies were differed according to type of PM. OF the 13 studies investigated the risk of MI per speci c unit increment of PM ≤ 2.5 µm, seven studies reported a higher risk of MI that in the most of them ( ve studies) the risk was signi cant. Besides, in the other six studies did not show a higher risk of MI that even in a study the risk of MI decreased per speci c unit increment of PM ≤ 2.5 µm. for particulates with aerodynamic range of 2.5 to 10 µm, of 3 six studies investigated association, the number of two studies found a signi cantly higher risk of MI and another study did not observe a higher risk of MI per increasing in the PM. In related to PM ≤ 10 µm, we found 6 studies examined the relationship. Of three studies reported a higher risk of MI, the risk of MI was signi cant in two studies. Moreover, in the other three studies did not nd a higher risk of MI per increment of PM.
Furthermore, we observed a signi cantly high heterogeneity across studies (I 2 = 96.8% [p-value < 0.001]). By considering the study population (subjects without MI history and those with MI/CHD history), As showed in Fig. 3 respectively. We also performed subgroup analysis according to gender, including female and both genders. As plotted in Supplemental Fig. 2, there was no signi cant higher risk of MI in female subgroup, but we found a signi cantly higher risk of MI in the other subgroup. Apart from, we used the Jackknife approach to investigate the impact of each study on the pooled effect size and the heterogeneity across studies (see Fig. 4

Discussion
According to the best of our knowledge, this systematic review and meta-analysis study was the rst attempt to investigate the effect of long-term exposure to PM air pollution on the incidence of MI. Overall, The number of 17 prospective cohort studies with more than 18 million participants and 800,000 cases of MI was included. Obtained results indicated per 1 µg/m 3 increment of PM 10 , the risk of MI raised signi cantly by 2% averagely. This nding was in line with some studies found a signi cantly greater risk of CVDs results from long-term exposure to PM 8, 12,53 . Besides, in comparison to previous meta-analysis investigated the short-term effect of exposure to PM 10 on the risk of MI, we observed stronger association between PM and MI [22][23][24][25] . Furthermore, subgroup analysis based on types of study population reveals subjects with MI/CHD history in comparison to who without MI history had a signi cantly 3 % greater risk of MI per 1 µg/m 3 increase in the levels of PM averagely. Apart from this, by considering aerodynamic diameter of PM, We found only a signi cant stronger risk of MI per 1 µg/m 3 increase in PM with aerodynamic diameter less than 2.5 µm, but not for PM 2.5−10 and PM 10 . This nding was in line of several previous studies reported a signi cantly association between PM 2.5 and CVD in uence and mortality 12,13,54 . This nding also con rms the greater effect of the smaller size fractions of PM on the cardiovascular system 2, 55 . In sum, this study indicates a signi cant association between long-term exposure to PM ≤ 10 µg/m 3 and the developing of MI. Although biological mechanisms underlining the association not fully understood, the possible biological pathways were discussed with details in several invaluable studies 2, 56-59 . Brie y, PM air pollution can result in impairments in the cardiovascular system, including heart, vascular, and blood through several pathways, including lung in ammation, neuroendocrine activation, particle translocation, and unde ned bloodborne mediator. It is supposed that in long-term exposure to PM, these impairments can raise the developing of cardiovascular morbidity and mortality 2 .
This study accompanied with several advantages. This study was the rst attempt to quantify the long-term effect of exposure to PM on the incidence of MI. We included the prospective cohort studies with appropriate sample size (more than 18 million participants), cases of MI (more than 800,000 cases). However, there are some shortcomings in the both study and outcome levels that should be taken into account when interpreting the results of this review. First of all, for measuring exposure to PM, the included studies applied various methods having different precision. Moreover, most studies did not determine exposure to PM in other situations such as indoor and occupational sources. Apart from, PM health effects can be vary depending on its nature. Studies did not account this and only considered the diameter of PM.
The studies provided adjusted HR (95% CI) of MI per different increases in PM levels. Thus, we assumed a linear relation between PM and MI and standardized the risk of MI per 1 µg/m3 increase in PM levels. This also can be associated with error in measuring the effects size of association between. Second, the de nition of MI across studies is problematic.
Some studies used ICD codes but the others did not. Among studies provided ICD codes, some studies de ned MI as ICD-9: 410 or ICD-10:I21 (primary MI) but the others considered both primary and secondary types of MI and applied ICD-10: I21-23. Last but not the least, we found a high heterogeneity across studies. What is more, in the large number of included studies, MI incidence data obtained from registry information. This can potentially increase the risk of nondifferential bias because of coding errors. Third, confounding effect is another important limitation across studies.
Recent meta-analysis study has indicated that exposure to environmental noise can be associated with higher risk of MI 60 . The most selected studies did not adjust this effect. None of the included studies did not adjust exposure to occupational risk factor such as noise, shift work and life style risk factors such as sleep status, diets patterns, and Page 16/24 physical activity that according to previous studies can be associated with the higher odds of MI. Forth, combining HR (95% CI) adjusted for different covariates may have reduced the consistency across studies and reduced the precision of our summary estimates. Finally, although subgroup analyses based on type of population and sensitivity analyses according to Jackknife approach were performed, we found a high heterogeneity, especially in studies conducted among subjects without history of MI. More studies are needed to explore the sources of heterogeneity.

Conclusion
This study was the rst attempt, according to the best of our knowledge, to investigate the effect of long-term exposure to PM air pollution on the risk of MI. Obtained results indicated a signi cantly higher risk of MI per 1 µg/m3 increase in the level of PM 10 . Moreover, Subgroup analysis based on types of study population reveals subjects with MI/CHD history in comparison to who without MI history had a signi cantly greater risk of MI per 1 µg/m3 increase in the levels of PM 10 .
Considering aerodynamic diameter of PM shows only a signi cant stronger risk of MI per 1 µg/m3 increase in PM 2.5 .
More studies are warranted to overcome the above-mentioned limitations and con rm our ndings. Figure 1 Flow diagram of systematic review process.

Figure 2
Forest plot for the association between Particulate matter (PM ) air pollution with aerodynamic ≤ 10 µm and the risk of myocardial infarction (MI). HR, hazard ratio; CI, con dence interval.

Figure 3
Forest plot for the association between Particulate matter (PM ) air pollution with aerodynamic ≤ 10 µm and the risk of myocardial infarction (MI) according to the study population (with or without CVDs). CVDs, cardiovascular diseases; HR, hazard ratio; CI, con dence interval.

Figure 4
Sensitivity analysis using the jackknife approach. Forest plot for the association between Particulate matter (PM ) air pollution with aerodynamic ≤ 10 µm and the risk of myocardial infarction (MI) after removing Kim study according to Jackknife approach. HR, hazard ratio; CI, con dence interval.

Figure 6
Forest plot for the association between Particulate matter (PM) air pollution and the risk of myocardial infarction (MI) according to PM diameters. HR, hazard ratio; CI, con dence interval.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.