Association between PM_2.5 and risk of hospitalization for myocardial infarction: a systematic review and a meta-analysis

Air pollution (atmospheric pollution) is the release of harmful particles matter into air by one or more harmful gases. It is generally assumed that any exposure to outdoor particulate matter air pollution can pose a big challenge to both public health agencies and physicians in the world, especially in the developing countries [1]. It is also believed that outdoor air pollution is a threat factor contributing to universal mortality and disability-adjusted life-years (DALYs) which rank the fifth and sixth in the world, respectively [2, 3]. Based on the criteria released by the National Ambient Air Quality Standard (NAAQS), there are six major pollutants of ozone (O₃), carbon monoxide (CO), lead (PB), sulfur dioxide (SO₂), nitrogen dioxide (NO₂), particulate matter < 10 μm (PM₁₀), and particulate matter < 2.5 μm (PM_2.5). Recent findings suggest that any exposure to PM_2.5 can endanger lungs and blood stream more than other pollutants and can lead to adverse cardiovascular, respiratory, and neurological disorders (Stroke, Alzheimer and Parkinson) as well as premature birth [4‐8]. In most countries, it is thought that the level of PM_2.5 particles is higher than the defined standards, even higher than those set by the World Health Organization (WHO) (WHO) [9, 10]. It is assumed that automobiles and combustion activities are the main sources for the production of PM_2.5 [11]. Tehran Province in Iran, has been struggling with the highest air pollution in the last few decades, due to fast-growing industrial activities as well as the large number of automobiles on the road [12]. Cohen et al. (2017), in their study, indicated that 103.1 million years of life lost (YLL) and 4.2 million mortality occurred as a consequence of exposure to PM_2.5.

Myocardial infarction (MI) or cardiac infarction is generally defined as detection of an elevated cardiac troponin (cTn) value which is above the 99th percentile upper reference limit [13]. In recent years, the prevalence of MI has been increased in both developed and developing countries [14, 15]. Research evidences indicate that such risk factors as age, sex, and family history cannot be modulated, but some of the risk factors such as ambient air pollution and unhealthy life style are to a great extent preventable [16‐18]. Considering the fact that the age for onset of the first MI has been decreasing and that MI is multifactorial in nature, its fundamental function remains unknown [19].

Therefore, incidence of MI with simultaneous concentration of fine particulate matter has been extensively studied all over the world [20, 21] but only two systematic reviews and meta-analyses about the effect of particulate matter on MI were found Mustafic et al. (2012) revealed that all air pollutants, except for ozone, are significantly correlated with the increased risk of MI. In this meta-analysis, 13 research studies on feature of PM_2.5 were scrutinized to detect the risk of MI and it was found that the relative risk of overall PM_2.5 ranked the second after the relative risk of overall carbon monoxide [22]. Moreover, Luo et al. (2015) conducted a meta-analysis based on thirty-one time-series and case crossover studies in order to investigate the effect of particulate matter on the risk of MI. The results demonstrated that the exposure to PM_2.5 can increase the risk of MI much more than the exposure to PM₁₀. The findings also showed that there was a moderate heterogeneity in meta-analysis of the pooled estimates, but the subgroup analyses might not pinpoint the cause of this heterogeneity. Therefore, it is imperative to investigate the source of heterogeneity in a study with more details (Luo et al., 2015) [23]. In this meta-analysis, the rate of PM_2.5 and the risk of MI are focused. It should be noted that PM_2.5 was a subgroup in other meta-analysis studies. The time scope in this study is broader as the original studies conducted from January 2000 to June 2018 were attempted to be incorporated. This study has been performed in University of Medical Sciences, Iran in 2018.

This protocol was registered in PROSPERO, the International Prospective Register of Systematic Reviews, on 2 January, 2019 (registration number CRD42019118998). The findings of this study were based on the accommodation guidelines: “Preferred reporting items for a protocol for a meta-analysis (PRISMA-P) 2015” [24]. In this study, two reviewers (ZF and MAD) conducted a research on such electronic bibliographic databases including Scopus, Web of Science, PubMed, and EMBASE. They also searched for components such as “air pollution” and “myocardial infarction” and found synonyms using the Medical Subject Headings (MeSH). In addition, the results were combined using the Emtree term and incorporated all other synonyms, except for those found in PubMed. In this way, it was possible to narrow the syntax down to a specific period from Jan. 1, 2000 to Jun. 1, 2018. It is worth mentioning that any study dealing with the short-term relationship between the pollutants and myocardial infarction was thoroughly reviewed. A thorough search on Google Scholar was also performed using dual combinations of the two main components. In an attempt not to miss any study, the grey literature and conference proceedings were explored, and a list of references was ultimately reviewed [25]. There was no language restriction on the search engines. Having completed the search, one of the reviewers (ST) did the duplications using Endnote software version 8 and started to conduct the initial screening through titles and abstracts. Then, the two reviewers reviewed the full-text of the articles carefully for any potentially-relevant studies according to inclusion and exclusion criteria [26]. The inclusion criteria for this meta-analysis allowed for utilization of original studies with time-series or case-crossover designs dealing with any exposure to particulate matter (PM_2.5), including even a short-term exposure such as the same day or 7 days before the occurrence of MI. The excluded studies had the following traits: 1) not being case-crossover or time-series designs, 2) non-original studies, 3) patients with MI, (4) long-term exposure to particulates matter PM₁₀, and 5) no reported relative risk (RR)/odds ratio (OR) and 95% Confidence Interval (CI 95%). Any disagreement in arbitration for the eligibility of the paper was discussed until a consensus was reached by the reviewers.

Study characteristics, the risk of bias, and study selection for the included research studies. The selection procedure for the meta-analysis is shown in Fig. 1.

Totally, 13,468 papers were identified. Having excluded 2397 duplicated studies, 11,071 papers were obtained, out of which 10,918 were excluded by title and/or abstract. One hundred fifty-seven full-text papers were opted out and then thoroughly assessed. Finally, 131 unrelated studies were left out and 26 papers were identified eligible for the study. The total number of participants with hospitalization for myocardial infarction was 2,250,473. The largest number of participants was 452,343, which belonged to the study by Milojevic et al., (2014), and the smallest number of participants was 208, which was reported by Akbarzadeh et al., (2018). Considering Berglind et al. (2010), the research was conducted in three cities (Boston, Seattle, and Augsburg), the lag averaging time was 2 h, and the adjusted odds ratio for the PM_2.5 pollutant was applied in the analysis [39].

Figure 2 is based on 28 comparisons illustrating the association between a 10 μg/m3 increase in the risk of MI and PM_2.5. The heterogeneity of the research studies was evaluated through random-effects with P < 0.0001, and I2 was 69.52%, showing a moderate heterogeneity. The meta-analysis showed a significant positive association between per 10 μg/m3 elevation in PM_2.5 and MI risk (RR = 1.02; 95% CI 1.01–1.03; P ≤ 0.0001) at lag exposure of 0 and 0–1 days.

Considering the quality of study subgroup (17 high-quality studies and 11 low-quality studies), a significantly higher rate of MI risk was seen in high quality studies (RR = 1.02, 95% CI: 1.01–1.03, P ≤ 0.0001) with a moderate degree of heterogeneity (I² = 62.37, P ≤ 0.0001) and in low quality studies (RR = 1.02, 95% CI: 1.01–1.03, p = 0.002) with a moderate to high degree of heterogeneity (I² = 71.96, P ≤ 0.0001), which was consistent with the results of the overall analyses (Fig. 3).

With respect to the study design subgroup, there were a positive association among MI risks in 21 case-crossover studies (RR = 1.03, 95% CI: 1.02–1.04, p ≤ 0.0001; I² = 75.78, P ≤ 0.0001), which was basically consistent with the overall analyses. There was also statistical significance for 7 time series study subgroup (RR = 1.01, 95% CI: 1.01–1.02, P ≤ 0.0001; I² = 41.61, P ≤ 0.0001) (Fig. 4).

The subgroup analysis of the study period, Accordingly, the original studies were divided into two subgroups based on the follow-up times: the follow-up of less than 4 years as short study period and the follow-up time of more than 4 years as long study period (10 long study period and 19 short study period) revealed a significantly increased MI risk in the long study period (RR = 1.02, 95% CI: 1.01–1.02, P = 0.014) with a moderate degree of heterogeneity (I² = 41.78 P ≤ 0.0001) and the short study period (RR = 1.03, 95% CI: 1.01–1.04, P ≤ 0.0001) with a moderate degree of heterogeneity (I² = 76.37, P ≤ 0.0001), which was consistent with the overall analyses (Fig. 5).

This study aimed to assess the association between exposure to PM_2.5 and MI hospitalization. The subgroup analyses and the overall analysis were performed based on pooled estimates and relationship between 10 μg/m3 increase in the short-term exposure to PM_2.5 and the risk of incident MI was pinpointed. Notably, two previous reviews had also shown this association, and in both of them the heterogeneity was supposedly moderate. In an attempt to identify the sources of heterogeneity, Mustafic et al. (2012), formed two subgroups based on study quality and lag exposure, and Luo et al. (2015), formed four subgroups based on study design, study quality, lag exposure, and geographic locations subgroup analyses. However, neither of them could successfully describe the sources of heterogeneity. In the same vein, three subgroup analyses (study quality, study design, and study period) were performed in the present study. According to the results, there was a relatively little difference among the high quality subgroup of the study, the case-crossover study design, and the short study period, all of which contributed to the overall analyses in terms of statistical significance and evidence of heterogeneity. Except for the low quality subgroup of the study which was not statistically significance, the rest were statistically significant. It is assumed that subgroup time series study design and long study period could substantially decrease heterogeneity (I² = 41.61, 41.78). Time-series analysis examined both pre-adjustment and co-adjustment. The pre-adjustment method picks up temporal trends from both the health and air pollution, whereas the co-adjustment approach considers air pollution forecasters and temporal trends [54]. It is presumed that the case-crossover design can create bias as a unidirectional control sampling devoid of time trends [55]. Time-series method is more likely to result in more accurate estimates of risk than the case-crossover method [54]. It is reckoned that the reasons behind the observed heterogeneity in the present study could be the varied design of the included original studies, and also the use of case-crossover and time series studies. Future studies are, nonetheless, expected to use time-series studies, which may help clarify the source of heterogeneity. The follow-up accuracy is also a prerequisite for estimating valid consequences and should be acclaimed systematically. The follow-up index is easy to achieve and could be applied as a reporting criterion for indicators [56]. It is thought that the priority put on the long-term follow-up could enhance the capability to prepare more precise estimates [57]. We also found derivation errors in the second meta-analysis which was performed in 2015. The errors were found in three papers authored by Linn et al. (2000) [58], Xie et al. (2014) [59], and Wichmann et al. (2014), [46]. All these studies might be unintentionally entered the forest plot (PM_2.5) and did not measure the effect of PM_2.5 on the risk of incident MI, which could affect the pooled estimates of the study. All these studies might be unintentionally entered the forest plot (PM_2.5) and did not measure the effect of PM_2.5 on the risk of Incident MI, which could affect the pooled estimates of the study. In addition, the number of studies identified in the work of Chowdhary et al. (2018), was higher than that of the two previous studies conducted by Mustafic et al. (2012) (13) and Luo et al. (2015) (16). Also, the two reviewers in this study conducted the data extraction phase independently, appraised the papers, investigated all the data, and removed the difference through a third person. Moreover, this study was extensive enough to lower the possibility of publication bias. Even the gray studies were enveloped without any language limitation. This study had some limitations. First, the included original research papers had a great variety and substantially differed from one another in case population, number of people from below 1000 to more than 400,000 people, the city examined, and the study period from under 6 months to over 300 months. Secondly, the assessment of the effect of air pollution on MI could not be well-established as MI is a multifactorial disease associated with diabetes, hypertension, smoking, alcohol, and obesity [60]. Thirdly, the population is chosen in different age group, while the cause of MI in young adults differs from the elderly people. Most of the people with MI are elderly ones with heart problems beforehand. It is estimated that airborne contamination could trigger the undesirable effects to be over-represented for this age group [33, 61].

Finally, as the people who are exposed to a mix combination of air pollutants for longest periods are those at elevated risk of adverse health, outcomes related to individual exposure to a single pollutant cannot be obtained with a high degree of certainty [61].

The results of this Meta-analysis demonstrated the severity of the relationship between PM_2.5 and MI with more accurate estimates than analysis presented by 26 studies alone, substantiating the notion that PM_2.5 levels are key factors in the development of MI hospitalizations. It is highly imperative to conduct further investigations to determine all possible causal relationships and explore potential mechanisms affecting MI. The economic burden of air pollution health-related outcomes is very significant, especially for healthcare providers. Fiscal implications attributed to air pollution are calculated as 253 million to 2.9 billion USD in Asia. Policy makers adopt more effective strategies to help improve air pollution, especially in developing countries, or prevent exposure to PM_2.5 so as to protect public health.