Effect of exposure to ambient PM2.5 pollution on the risk of respiratory tract diseases: a meta-analysis of cohort studies.

The International Agency for Research on Cancer and the World Health Organization have designated airborne particulates, including particulates of median aerodynamic diameter ≤ 2.5 μm (PM2.5), as Group 1 carcinogens. It has not been determined, however, whether exposure to ambient PM2.5 is associated with an increase in respiratory related diseases. This meta-analysis assessed the association between exposure to ambient fine particulate matter (PM2.5) and the risk of respiratory tract diseases, using relevant articles extracted from PubMed, Web of Science, and Embase. In results, of the 1,126 articles originally identified, 35 (3.1%) were included in this meta-analysis. PM2.5 was found to be associated with respiratory tract diseases. After subdivision by age group, respiratory tract disease, and continent, PM2.5 was strongly associated with respiratory tract diseases in children, in persons with cough, lower respiratory illness, and wheezing, and in individuals from North America, Europe, and Asia. The risk of respiratory tract diseases was greater for exposure to traffic-related than non-traffic-related air pollution. In children, the pooled relative risk (RR) represented significant increases in wheezing (8.2%), cough (7.5%), and lower respiratory illness (15.3%). The pooled RRs in children were 1.091 (95%CI: 1.049, 1.135) for exposure to <25 μg/m3 PM2.5, and 1.126 (95%CI: 1.067, 1.190) for exposure to ≥ 25 μg/m3 PM2.5. In conclusion, exposure to ambient PM2.5 was significantly associated with the development of respiratory tract diseases, especially in children exposed to high concentrations of PM2.5.

Introduction

Air pollution is a complicated process involving the spread of distinct pollutants throughout the atmosphere. Air pollution has been found to induce diseases in humans and disorders in other living organisms, as well as destruction of the natural environment[ 1­ 2]. One type of pollutant, particulate matter (PM)[ 3], has been associated with serious public health problems[ 4], as has combinations of PM and other air pollutants[ 5]. PM is classified according to its aerodynamic diameter, and the gold standard used to evaluate its transport capacity in the atmosphere and inhalation capacity through the respiratory tract[ 6]. PM is primarily categorized as coarse (PM10), of median aerodynamic diameter≤10 μm, and fine (PM2.5), of median aerodynamic diameter≤2.5 μm[ 6].

PM originates from a wide range of sources, including road dust, agricultural dust, industrial emissions, construction sites, mining operations, river beds, crustal materials, and combustion, or as secondary aerosols from distant sources[ 7­ 9]. Due to the diversity of sources, human exposure is high. Entry of PM into the respiratory tract depends on the physical characteristics, breathing mode and rate, and size of an individual[ 10]. Moreover, PM size has been significantly related to the etiology of pertinent diseases. Frequently, smaller PM such as PM2.5 penetrates the respiratory tract more deeply at a higher rate, and is deposited in the respiratory bronchioles and alveoli or enters the bloodstream, influencing lung function and eventually causing other disorders[ 11]. Exposure to PM has been shown to be harmful to public health, increasing the incidence of respiratory symptoms, reducing lung function, and aggravating respiratory and cardiovascular diseases[ 12].

PM2.5 can also act as a carrier of other harmful constituents, such as heavy metal ions, which add to the deleterious effects of "inert" material[ 13]. Studies analyzing the induction of respiratory diseases by exposure to PM2.5 have yielded different outcomes. To quantitatively and accurately assess the effects of exposure to PM2.5 on respiratory tract diseases, we performed a meta-analysis that included all relevant cohort studies published to date. This meta-analysis showed that exposure to PM2.5 increased the risk of respiratory tract disease.

Materials and methods

Systematic ascertainment of correlative studies

The online databases, including PubMed (National Library of Medicine, Bethesda, MD, USA), Web of Science (Thompson Scientific, Philadelphia, PA, USA), and Embase (Excerpta Medica Database, the Netherlands) were searched for cohort studies published and indexed through May 4, 2016 on the epidemiology of respiratory tract diseases associated with PM2.5 air pollution. Search strings, both free text and medical subject headings (MeSH), included (PM2.5 OR "particulate matter") AND (wheezing OR bronchitis OR cough OR asthma OR pneumonia OR COPD OR "lung cancer" OR "respiratory infections" OR "respiratory tract diseases"). Based on article titles, abstracts, and full texts, these cohort studies were screened for those fulfilling our inclusion criteria. Cohort-specific results reported in previous meta-analyses alone were also considered in the current analysis.

Inclusion criteria

All epidemiological studies involving the impact on human health of exposure to PM2.5 were screened. Health outcomes of interest were morbidities of respiratory tract diseases, according to ICD9 or ICD10, including pneumonia, asthma, bronchitis, upper respiratory tract infections, lower respiratory infections (LRI), wheezing, cough, chronic obstructive pulmonary disease, and lung cancer (). Studies were included if they had cohort designs for respiratory diseases and if they reported relative risk (RR) or odds ratio (OR) with 95% confidence intervals (CIs).

If multiple studies reported an association between respiratory tract disease and PM2.5 in the same study cohort at different times, the latest one was chosen. However, if these studies reported different outcomes in the same cohort, each was included. Only studies published in English and exclusively involving human subjects were included. Cohort-specific results reported in previous meta-analyses were considered. If that cohort was included in a meta-analysis that included several other cohorts, the latest publication by this cohort was included.

Exclusion criteria

Case-control studies, case series, and case reports were excluded, as were studies lacking appropriate data (e.g. useful RRs or ORs and 95% CIs in related cohort studies). These criteria were used to maximize sensitivity and ensure non-omission of any relevant study.

Data extraction

All selected publications were screened independently by two investigators (Q.L. and C.X.), and data extracted using a standardized form. Conflicts were resolved by discussions between the two investigators.

Meta-analysis

Effect estimates for all data collected from the selected studies were summarized using STATA software (version 11; Stata Corp, College Station, TX, USA). The unadjusted RRs and ORs with their 95% CIs were integrated to analyze the strength of the risk of respiratory tract diseases in participants exposed to PM2.5. Heterogeneity among studies was examined using a chi-square-based Q-statistic test and the standard I2 test. The Q-statistic test can only determine the presence or absence of heterogeneity, not the degree of heterogeneity. As the I2 test may quantify the extent of heterogeneity in a meta-analysis[ 14], two methods were chosen to simultaneously assess heterogeneity. Between-study heterogeneity reflected variations in study outcomes among different studies due to inherent differences in study design/populations/exposures, not to chance alone. All pooled RRs were calculated using a random effects model. I2 showed that heterogeneity accounts for a percentage of the overall variability in random error. I2<40%, 30%-60%, 50%-90%, and 75%-100% represented unimportant, medium, substantial, and high degree of heterogeneity, respectively[ 15].

Subjects were stratified into subgroups based on heterogeneity, age, region, PM2.5 exposure level and source, and differences between groups and risk factors were calculated[ 16]. Moreover, sensitivity analyses were performed to estimate the stability of the outcomes. That is, one study at a time was iteratively removed and the results of the remaining studies were determined[ 17]. Begg funnel plots and the Egger test of asymmetry were performed to statistically evaluate publication bias; a P value<0.05 was considered indicative of publication bias.

Results

Characteristics of the eligible studies assessed by meta-analysis

A review of the PubMed, Web of Science and Embase identified 1,126 eligible studies published in English. Based on the inclusion/exclusion criteria for the effects of exposure to PM2.5 pollution on respiratory tract diseases, 35 related studies were retrieved[ 18­ 52]. A flow diagram of the literature search and selection procedure is shown in . shows the details of each of the 35 studies; in these articles, there were 12 outcomes related to wheezing, 5 to bronchitis, 12 to cough, 14 to asthma, 1 to pneumonia, 4 to lower respiratory tract illness (LRI), 0 to upper respiratory tract illness (URI), 4 to lung cancer, 0 to COPD, and 2 to respiratory infections. These 35 articles included a total of 1,135,203 subjects with different respiratory tract diseases. Primary outcomes in our meta-analysis included the incidence of newly developed or exacerbated respiratory tract diseases. These participants resided in a variety of countries or regions and included subjects in different age groups. The mean, median and 50% interquartile range (IQR) for PM2.5 ranged from 3.60 to 100 μg/m3. There were seven articles on traffic-related air pollution and two meta-analyses that included unpublished data of some birth cohort studies ().

Literature search and article selection protocol used in the present meta-analysis.

Evidence synthesis

Pooled analysis of the 35 included articles showed that exposure to PM2.5 pollution significantly increased the RR for respiratory tract diseases for (RR= 1.076, 95% CI: 1.050, 1.103; Pheterogeneity<0.001, I2 = 93.6%). To assess differences among subgroups, studies were divided by the age of the participants (children or adults), geographic areas (North America, Europe, Oceania, or Asia), types of diseases, and sources of PM2.5 (traffic-related or non-traffic related air pollution) (). Studies in children were stratified by types of disease, and extent of exposure concentrations (, ). The pooled RRs in the random-effects model were 1.104 ( and ) for children, 1.099 for Europe, 1.090 for North America, 1.064 () for Asia, 1.073 for wheezing, 1.153 for LRI, and 1.048 () for cough. In addition, the pooled RRs exposed to traffic-related and non-traffic related air pollution were 1.085 and 1.076 (), respectively. The pooled RRs in children showed that the rates of wheezing (8.2%), cough (7.5%), and LRI (15.3%) were significantly increased. In adults, however, no positive association was found (). The pooled RRs in children exposed to PM2.5 concentrations<25 μg/m3 and≥25 μg/m3 were 1.091 and 1.126 (), respectively. However, other subgroups showed no association between RR and exposure to PM2.5.

Combined RRs with 95% CIs for the association between PM2.5 exposure and respiratory tract diseases in all subjects and in subpopulations of children and adults.

Combined RRs with 95% CIs for the association between PM2.5 exposure and respiratory tract diseases by geographic region (North America, Asia, Europe, and Oceania).

Combined RRs with 95% CIs for the association between PM2.5 exposure and respiratory tract diseases by type of disease (asthma, bronchitis, cough, LRI and wheezing).

Combined RRs with 95% CIs for the association between exposure to PM2.5 and respiratory tract diseases exposed to traffic-related and non-traffic-related air pollution.

Combined RRs with 95% CIs for the association between exposure to low and high concentrations of PM2.5 and respiratory tract diseases in children.

Sensitivity analysis

Sensitivity analysis was performed by re-analyzing RR after removing one study at a time. The correlation between exposure to PM2.5 pollution and the RR of respiratory tract disease was not driven by any individual study, with no alterations in the significance of the pooled RRs, suggesting that the combined RR remained stable and reliable. Sensitivity analysis indicated that the omission of any one study resulted in RRs between 1.057 (95% CI: 1.038, 1.075) and 1.080 (95% CI: 1.049, 1.112).

Publication bias

shows no funnel plot asymmetry, while P values of the Begg and Egger tests were greater than 0.05 in both the global and stratified analyses, respectively. These findings indicated a lack of publication bias.

Discussion

Previous epidemiological and experimental studies have been unable to definitively determine the specific mechanisms by which exposure to PM has adverse effects on human health. However, accumulated evidence suggests that the most deleterious effects of PM are dependent on particle size, with PM2.5 being especially harmful[ 53­ 54]. Exposure to PM2.5 has been found to increase health risks, particularly with regard to respiratory tract diseases[ 55]. Mortality has been regarded as the one important indicator of the effects of PM2.5 pollution on health outcomes[ 56]. Several case-control studies have also assessed daily hospital admissions or emergency department visits[ 57]. However, these variables are limited by the lack of persistent observation, restriction of end points, and recall bias. Therefore, this meta-analysis was performed to clarify the relationship between exposure to PM2.5 and the incidence or aggravation rate of respiratory tract diseases in cohort studies. In addition, studies were stratified by age, geographic location, and the source and concentration of PM2.5.

This systematic review found that exposure to PM2.5 was positively correlated with risk of respiratory tract disease, especially in children; in subjects with wheezing, cough, and LRI; and in populations in Europe, North America, and Asia; The pooled RRs were greater for traffic-related than non-traffic-related air pollution. Ultimately, we found that PM2.5 in children was significantly associated with cough, wheezing, and LRI. Furthermore, in children, pooled RRs were greater for high (≥25 μg/m3) than low (<25 μg/m3) PM2.5 concentrations. These findings suggested that traffic-related PM2.5 did greater harm to the human body, and that exposure to PM2.5 pollution may pose an increased risk for respiratory tract disease, especially in children exposed to high concentrations of PM2.5. These results are consistent with those of previous reports[ 58­ 59], as well as being more accurate than the results of a previous meta-analysis of case-control studies[ 60].

Above all, this meta-analysis included many more articles with a larger population, especially birth cohorts, than previous meta-analyses. Second, longitudinal cohort studies with complete and reliable information are considered less likely to be influenced by confounding and better able to address the temporal sequence of events. Finally, this study not only revealed a strong association between PM2.5 and respiratory tract diseases in general, but also strong correlations between PM2.5 and specific types of respiratory tract diseases (wheezing, cough and LRI). Therefore, the results of this meta-analysis could amplify and complete those of earlier meta-analyses.

However, the pathological mechanisms underlying the effects of PM2.5 exposure on the respiratory tract are not fully understood. PM size may be affected by chemical, biologic, and physical properties, resulting in various pathological consequences[ 6]. There are several plausible biomedical explanations for associations between exposure to PM2.5 and respiratory tract diseases. Owing to its fine consistency, PM2.5 can be deposited more deeply into the lungs. Moreover, these particles may contain toxic components or contaminants, such as nitrates, sulfates, acids, and metals, originating from combustion processes or similar activities[ 61]. These particles can therefore lead to stress, inflammation, and allergy. Our results showed that PM2.5 exposure resulted in a greater incidence of cough, wheezing, and LRI in children than in adults, which may be related to differences in the structure of the respiratory tract in adults and children. Specifically, the immature respiratory system in children may be more sensitive to PM2.5. PMc of aerodynamic diameter 2.5-10 μm mainly derives from the soil and abrasive mechanical processes. These particles may transport biologic materials, such as bacteria, molds, and pollens, which may have harmful effects on the respiratory system[ 62­ 63]. Therefore, some infectious bacterial respiratory diseases or hay fever are more likely to manifest after lengthy exposure to PMc rather than PM2.5. Our results are consistent with these findings.

We also found that the correlation between PM2.5 exposure and respiratory diseases was stronger in European and North American populations, probably due to more studies of PM2.5 were conducted in these populations. The risk of respiratory diseases was also higher following exposure to high than low quantities of PM2.5, suggesting that the concentration of PM2.5 is also a risk factor for respiratory diseases.

This meta-analysis had several limitations. First, some of the outcomes demonstrated heterogeneity. We utilized a random-effects model, with stratified analysis performed to make up for this shortcoming. Second, non-English publications and unpublished results were excluded. Some of these studies may have included negative outcomes, which could have influenced our results. Third, we could not exclude residual confounders, which may have influenced our results. Fourth, the limitations of data collection from all studies prevented a comparison of results from different climate zones, with different temperatures and humidity, which may have influenced the correlation between PM2.5 exposure and respiratory diseases.

In conclusion, we found that PM2.5 may play an important role in respiratory tract diseases, especially in children exposed to high concentrations of PM2.5. Additional studies are needed to assess the quantification and identification of other, as yet undetermined, harmful compounds in ambient air particles, and to determine the underlying mechanisms that cause these particles to affect human health.

Tab.1

Characteristics of cohort studies with PM2.5 exposure

Author (published year)	Study year	Cohort / Study	Outcomes of included studies	Sample number	Age (year) (group)	Country (Continent)	PM2.5 (μg/m3)
Neas et al. 1994 [18]	1983-1988	-	wheezing, cough, bronchitis, asthma, LRI	1,237	7 to 11 (children)	United States (North America)	31.1
Romieu et al. 1996 [19]	1991-1992	-	wheezing, cough, LRI	71	5 to 13 (children)	Mexico (North America)	85.7
Tiitanen et al. 1999 [20]	1995	the PEACE study	cough	76	8-13 (children)	Finland (Europe)	15 a
Schwartz et al. 2000 [21]	1990-1991	the Harvard Six City Study	cough, LRI	1,844	school-aged (children)	United States (North America)	15
Gehring et al. 2002 [22]	1997-1999	the PIAMA birth cohort	respiratory infections	1,606	Infants (children)	German (Europe)	13.4 c
Gent et al. 2003 [23]	2001	-	wheezing, cough	271	<12 (children)	New England (Europe)	13.1
Mar et al. 2004 [24]	1997­1999	-	wheezing, cough	25	children and adults	United States (North America)	10
Millstein et al. 2004 [25]	1994-1995	-	wheezing	2,034	9.6 (0.4) (children)	United States (North America)	5.24
Pino et al. 2004 [26]	1995-1996	-	bronchitis	504	infants (children)	Chile (South America)	52
Johnston et al. 2006 [27]	2004	-	asthma	235	children and adults	Australian (Oceania)	11.1
Bennett et al. 2007 [28]	1998-2005	-	wheezing, cough, asthma	1,446	37.2 (7.2) (adult)	Australian (Oceania)	6.8
Brauer et al. 2007 [29]	1999-2003	the PIAMA birth cohort	bronchitis, cough	4,146	4 (children)	Netherlands (Europe)	16.9 c
Morgenstern et al. 2007 [30]	1999-2000	GINI and LISA birth cohort	wheezing, bronchitis, cough, respiratory infections	2,908	children (children)	Germany (Europe)	12.8 c
Picciotto et al. 2007 [31]	1994-2003	-	bronchitis	1,492	3 to 4.5 (children)	United States (North America)	>25
Rodriguez et al. 2007 [32]	1996-2003	-	wheezing, cough	263	5 (children)	Australian (Oceania)	8.534
Beelen et al. 2008 [52]	1986-1997	-	lung cancer	1,940	55 to 69 (adult)	Netherlands (Europe)	28.2 c
Nuñez et al. 2008 [33]	2003-2005	the EVA cohort	wheezing, cough	197	6 to 14 (children)	Mexico (North America)	27.8 c
Clark et al. 2010 [34]	1999-2000	-	asthma	3,484	3 to 4 (children)	United States (North America)	4.67
Gehring et al. 2010 [35]	1996-2006	the PIAMA birth cohort	wheezing	3,863	8 (children)	Netherlands (Europe)	16.9 c
Gurley et al. 2013 [36]	2008-2011	-	LRI	257	2 (children)	Bangladesh (Asia)	100
Li et al. 2013 [37]	2006-2009	-	asthma	412,832	>18 (adult)	United States (North America)	11.6
Nielsen et al. 2013 [51]	-	the ESCAPE study	lung cancer	2,095	43 to 73 (adult)	(Europe)	5 c
Evans et al. 2014 [38]	2002-2007	-	asthma	530	3 to10 (children)	United States (North America)	8.6
Loftus et al. 2014 [39]	2010-2012	-	wheezing	58	school-aged (children)	United States (North America)	6.9 b
MacIntyre et al. 2014 d[46]	-	10 European birth cohorts	pneumonia	14,009	36 month (children)	(Europe)	5
MÖlter et al. 2014 d [40]	-	6 birth cohort	asthma	10,377	8 / 10 (children)	(Europe)	5
Puett et al. 2014 [50]	1994-2010	-	lung cancer	2,155	67±8.3 (women)	United States (North America)	10
Wendt et al. 2014 [41]	2005-2007	-	asthma	18,289	1to17 (children)	United States (North America)	14.97
Young et al. 2014 [42]	2003-2009	-	wheezing, cough, asthma	50,884	55±9 (adult)	United States (North America)	3.6 b
Jacquemin et al. 2015 d [43]	-	the ESCAPE study	asthma	23,704	adult (adult)	(Europe)	5
Rice et al. 2015 [44]	1998-2011	-	wheezing, bronchitis, cough, asthma	4,444	50.4 (12.4) (adult)	United States (North America)	10.8 a
Teresa et al. 2015 [45]	1998-2006	the CNBSS study	asthma	29,549	40 to 59 (adult)	Canada (North America)	12.57
Gehring et al. 2016 [47]	1996-2010	4 European birth cohorts	asthma	6,864	14 to 16 (children)	(Europe)	10
Guo et al. 2016 [49]	1990-2009	-	lung cancer	368,762	>30 (adult)	China (Asia)	10
Tétreault et al. 2016 [48]	1996-2011	QICDSS	asthma	162,752	13 (children)	Canada (North America)	6.5 b

LRI : Lower respiratory illness ; PM2.5 : mean, a median or b 50% IQR( Interquartile Range ) ; Traffic-related air pollution : c ; meta : d 6 birth cohorts: MAAS, BAMSE, PIAMA, GINI and LISA birth cohort; 10 European birth cohorts: BAMSE, GASPII, GINI and LISA, MAAS, PIAMA and four INMA cohorts; 4 European birth cohorts: BAMSE, PIAMA, GINI and LISA birth cohort

Tab.2

Association of PM exposure with episodes of respiratory tract diseases in children and adult group.

	N	RR (95%CI) Random-effects	P (I2)	P for Egger / Begg bias test
Children Symptom
Wheezing	10 [18,19,23-25,30,32,33,35,39]	1.082 (1.011, 1.158)	0.000 (77.2%)	0.079 / 0.107
Bronchitis	5 [18,26,29-31]	1.145 (0.957, 1.370)	0.000 (90.2%)	0.593 / 1.000
Cough	10 [18-21,23,24,29,30,32,33]	1.075 (1.019, 1.134)	0.002 (65.4%)	0.007 / 0.721
Asthma	8 [18,27,34,38,40,41,47,48]	1.119 (0.989, 1.266)	0.000 (98.5%)	0.910 / 0.536
Lower respiratory illness	4 [18,19,21,36]	1.153 (1.033, 1.287)	0.006 (76.2%)	0.064 / 0.308
Respiratory infections	2 [22,30]	1.050 (0.930,1.184)	0.409 (0.00%)	N/A / 1.000
Pneumonia	1 [46]	2.580 (0.910,7.270)	-	-
Total	40	1.104 (1.069, 1.139)	0.000 (95.4%)	0.025 / 0.408
Adult Symptom
Wheezing	4 [24,28,42,44]	1.053 (0.964, 1.150)	0.229 (30.5%)	0.593 / 1.000
Bronchitis	1 [44]	0.980 (0.860, 1.110)	-	-
Cough	4 [24,28,42,44]	0.951 (0.885, 1.021)	0.493 (0.00%)	0.713 / 0.308
Asthma	7 [27,28,37,42-45]	0.999 (0.978, 1.021)	0.553 (0.00%)	0.581 / 1.000
Lung cancer	4 [49,50,51,52]	1.064 (0.991, 1.142)	0.216 (30.8%)	0.691 / 0.806
Total	20	1.022 (0.986, 1.058)	0.000 (64.6%)	0.132 / 0.546

P (I2): P-value for test of heterogeneity, I2 of Higgins and Thompson reflecting the proportion of total variation in the estimate that is due to heterogeneity between studies. LRI: lower respiratory illness. Bold: significant results.