Association of PM2.5 and PM10 with Acute Exacerbation of Chronic Obstructive Pulmonary Disease at lag0 to lag7: A Systematic Review and Meta-Analysis

Abstract

This study aimed to conduct a meta-analysis to investigate whether short-term exposure to fine (PM_2.5) and coarse (PM₁₀) particulate matter was associated with acute exacerbation of chronic obstructive pulmonary disease (AECOPD) hospitalization, emergency room visit, and outpatient visit at different lag values. PubMed, Embase, and the Cochrane Library were searched for relevant papers published up to March 2021. For studies reporting results per 1-µg/m³ increase in PM_2.5, the results were recalculated as per 10-µg/m³ increase. We manually calculated the RRs for these two studies and transferred the RRs to estimate 10 µg/m³ increases in PM_2.5. Automation tools were initially used to remove ineligible studies. Two reviewers independently screened the remaining records and retrieved reports. Twenty-six studies (28 datasets; 7,018,419 patients) were included. There was a significant association between PM_2.5 and AECOPD events on lag0 (ES = 1.01, 95%CI: 1.01-1.02, p < 0.001; I²=88.6%, P_{heterogeneity}<0.001), lag1 (ES = 1.00, 95%CI: 1.00-1.01, p < 0.001; I²=82.5%, P_{heterogeneity}<0.001), lag2 (ES = 1.01, 95%CI: 1.01-1.01, p < 0.001; I²=90.6%, P_{heterogeneity}<0.001), lag3 (ES = 1.01, 95%CI: 1.00-1.01, p < 0.001; I²=88.9%, P_{heterogeneity}<0.001), lag4 (ES = 1.00, 95%CI: 1.00-1.01, p < 0.001; I²=83.7%, P_{heterogeneity}<0.001), and lag7 (ES = 1.00, 95%CI: 1.00-1.00, p < 0.001; I²=0.0%, P_{heterogeneity}=0.743). The subgroup analyses showed that PM_2.5 influenced the rates of hospitalization, emergency room visits, and outpatient visits. Similar trends were observed with PM₁₀. The risk of AECOPD events (hospitalization, emergency room visit, and outpatient visit) was significantly increased with a 10-µg/m³ increment in PM_2.5 and PM₁₀ from lag0 to lag7.

List Of Abbreviations: particulate matter (PM_2.5 and PM₁₀); acute exacerbation of chronic obstructive pulmonary disease (AECOPD); Chronic obstructive pulmonary disease (COPD); Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA); Effect sizes [48]; confidence intervals (CIs)

Keywords:

• Fine particulate matter
• acute exacerbation of chronic obstructive pulmonary disease
• hospitalization
• emergency room
• outpatient

Introduction

Chronic obstructive pulmonary disease (COPD) is characterized by significant airflow limitation associated with chronic inflammatory responses in airways and lungs, resulting in the destruction of lung tissue [1]. COPD has several complications, including acute exacerbation, respiratory failure, and pulmonary hypertension. The 4-year mortality rate of COPD patients ranges from 28% (for mild-to-moderate COPD cases) to 62% (for moderate-to-severe COPD cases) [1].

The deterioration of symptoms beyond the daily variation range leading to the use of additional therapy is known as an acute exacerbation of COPD (AECOPD), a medical emergency associated with significant morbidity and mortality [1, 2]. COPD exacerbations are commonly caused by viral or bacterial infections, including pneumonia, and air pollution (e.g. tobacco smoke, ozone, and occupational exposure) [1, 2]. Other risk factors include a history of acute exacerbations, ambient temperature changes, emphysema, chronic bronchitis, severe COPD at baseline, deteriorating airflow limitation, and an increase in the ratio of pulmonary to the aorta [1–7]. The prognosis of AECOPD is poor. Indeed, the in-hospital mortality rates of patients with COPD exacerbations are generally around 2.5% and 10% in those with hypercarbia [1, 2, 8], and the all-cause mortality rate within 3 years of the index hospitalization can be as high as 49% [1, 9].

The effects of air pollution on respiratory health have been extensively studied [10–16]. Substantial epidemiological evidence has demonstrated that ground-level ambient pollutants such as nitrogen dioxide (NO₂), sulfur dioxide (SO₂), and ozone (O₃) are strongly associated with AECOPD [13, 17]. Among various air pollutants, fine particulate matter (PM_2.5) and coarse particulate matter (PM₁₀) are considered to be the most hazardous because they can easily reach the lower airways and carry many toxic components that trigger a variety of adverse responses [18–24]. According to a previous meta-analysis [13] published in 2016, short-term exposure to PM_2.5 significantly increased the burden of AECOPD, but this study only assessed the outcomes of AECOPD patients in terms of emergency hospitalization and mortality. Similarly, the meta-analysis published in 2013 showed that ambient PM₁₀ was associated with COPD hospitalization and mortality [25].

The effects of air pollution on COPD are not necessarily immediate, and a lag can be observed between exposure and the detrimental outcomes. Indeed, since the publication of the meta-analyses presented above, relevant time-series studies [26–38] and case-crossover studies [39–44] were published regarding the association between PM_2.5 exposure and AECOP hospitalization, emergency room visit, and outpatient visit.

Aims

Therefore, the present study aimed to conduct a meta-analysis to explore whether PM_2.5 and PM₁₀ were associated with AECOPD hospitalization, emergency room visit, and outpatient visit at different lag values.

Materials and methods

Literature search

This meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [45]. Based on the PICO principle [46], PubMed, Embase, and the Cochrane library were searched for available papers published up to March 2021 using the MeSH term ‘Pulmonary Disease’, and ‘Chronic Obstructive’, as well as relevant key words. The eligibility criteria were 1) patients with COPD; 2) patients exposed to PM_2.5 or PM₁₀; 3) the occurrence of AECOPD as shown by hospitalization, emergency room visit, or outpatient visit; 4) language was limited to English. The study selection process was performed by two investigators (** and **). Discrepancies in the final list of included studies were resolved by discussion with a third investigator (**) until a consensus was reached. The reference lists of the included reports were screened for additional potentially eligible studies.

Data extraction

The data on study characteristics (authors, year of publication, study design, country and city/district where the study was performed, sex percentage, and sample size), exposure parameters (average concentration of PM_2.5 and PM₁₀, analytical model, and lag time of the event), and hospitalization for AECOPD were independently extracted by two authors (Liniuniu and Wangliyun). Any disagreements were solved by discussion. Some articles could be separated into independent datasets because each article reported AECOPD hospitalization and emergency room visit separately.

Outcome

The primary outcome was all events or evidence of AECOPD, including hospitalization, outpatient visit, and emergency treatment.

Quality of the evidence

To our knowledge, there were no validated quality assessment scales recommended for time-series and case-crossover studies [47]. Thus, we evaluated the qualities of the evidence were based on the possible biases in the research team, first author, and institution where the study was performed, and the relevant data were analyzed. The authority of the journals was also a critical component of quality assessment.

Statistical analysis

All analyses were performed using STATA SE 14.0 (StataCorp, College Station, TX, USA). Not all studies reported their outcomes as risk ratio (RR). Some used odds ratio (OR) to evaluate the association between PM_2.5 or PM₁₀ and AECOPD. In our meta-analysis, RR and OR were treated as the same parameter. For studies that reported the RR for 1-µg/m³ increase in PM_2.5, we manually calculated the RRs for these studies and converted the RRs to estimate 10 µg/m³ increases in PM_2.5. Effect sizes [48] and the corresponding 95% confidence intervals (CIs) were used to compare the outcomes. Statistical heterogeneity among these studies was calculated using Cochran’s Q-test and I² index. An I² >50% and p < 0.10 for the Q-test indicated high heterogeneity, and the random-effects model was used; otherwise, the fixed-effects model was applied. A P-value ≤0.05 was considered statistically significant. We only assessed the potential publication bias on lag0, lag1, lag2, lag3, lag4, and lag5 by funnel plots and Egger’s test because more than 10 studies reported the RRs on these lag points. As for outcomes with less than 10 studies, the funnel plots and Egger’s test could yield misleading results and were not recommended [47]. Subgroup analyses were performed according to the individual event types (hospitalization, emergency room visit, and outpatient visit. Sensitivity analyses were also performed by sequentially excluding each study and observing whether the summary results changed significantly.

Results

Retrieval of the studies

Figure 1 presents the study selection process. The initial search identified 386 records, and two were found from other sources. A total of 330 records were left after removing the duplicates and screened, and 268 were excluded. The remaining 62 papers were assessed for eligibility, and 36 were excluded because of inappropriate outcomes (n = 11) and no usable data (n = 25). Finally, 26 studies were included, and 28 datasets were analyzable [26–30, 39–42, 49–53] because the studies by Pothirat et al. [28] and Jo et al. [27], respectively, contained two datasets that could be assessed separately.

Figure 1. PRISMA 2009 Flow Diagram.

Table 1 presents the characteristics of selected studies. There were six case-crossover studies [39–44], 14 time-series studies (16 datasets) [26–38], three cross-sectional studies [49, 50, 53], two ecological studies [51, 52], and one cohort study [54], involving a total of 7,018,419 patients. The assessed lag varied from 0-2 to 0-14 days, except for one study evaluating 0-5 years [54]. A summarized table (Table S1) shows the overall outcomes on each lag day and subgroup analyses stratified by the type of admission.

Table 1. Literature search and study characteristic.

Author name	Study period	Design	Country	Study location	Sample size	Sex, female (%)	PM2.5 or PM10 concentration(μg/m3)	Lag	Analytical method
Alman, 2016	2012.6-2012.7	Case-crossover	USA	Colorado	1136	/	/	0-2	Conditional logistic regression
Belleudi, 2010	2001.4-2005.12	case-crossover	Italy	Rome	15087	/	22.8 ± 11.8/39.3 ± 17.2	0-6	Conditional logistic regression
Chen, 2020	2014-2017	Cross-sectional	China	Shenyang	17,655	47.70%	60 ± 50	0-5	Generalized additive model
Chen, 2019	2013-2015	Cross-sectional	China	Jinan	6981	29.50%	94 ± 56	0-7	Generalized additive model
Hwang, 2017	2006–2012	Time series analysis	China	Taiwan	38715	23.50%	38.8(mean)	0-5	Generalized additive model
Jo, 2018	2006-2012	Time series analysis	Korea	Chuncheon,	/	/	35.0 ± 25.2	0-5	Zero-inflated Poisson model
Kloog, 2014	2000–2006	case-crossover	USA	Mid-Atlantic	416778	57.70%	/	0-2	Conditional logistic regression
Ko, 2007	2000-2005	Ecological study	China	region	119225	/	35.7 ± 20.6	0-5	Generalized additive model
Li, 2017	2013/12/2-2014/12/1	Case-crossover	China	Beijing	1015	/	66(Median)	0-5	Conditional logistic regression
Liang, 2019	2013/1/18-2017/12/31	Ecological analysis	China	Beijing	161613	/	/	0-4	Generalized additive model
Pothirat, 2019	2016/3/1-2017/3/31	Time series study	Thailand	Chiang Dao	335	/	20.42(12.62-42.96)	0-7	General linear model
Pothirat, 2016	2006-2009	Cross-sectional	Thailand	Chiang Mai	740	47.43%	/	0-7	General linear model
Qiu, 2012	2000-2005	Time series analysis	China	Hong Kong	518,864	/	39.4	0-5	Generalized additive modeling
Raji, 2020	2008/3-2018/3	Time series study	Iran	Ahvaz	4534	30.17%	85.9 ± 150.4	0-7	General linear model
Santus, 2012	2007/1/1-2008/12/31	Case-crossover	Italy	Milan	1825	43.40%	32.8 ± 26.92	0-5	Conditional logistic regression
Sun, 2019	2015/1/1-2017/12/31	Time-series study,	China	Yancheng	4761	31.30%	45.2 ± 34.5	0-2	Over-dispersed generalized additive models
Sun, 2018	2010/10-2011/8	Time-series study	China	Shanghai	101	48.50%	55.4 ± 12.6	0-7	Poisson regression
Tao, 2014	2001-2005	Time-series study	China	Lanzhou	28057	33.90%	139	0-6	Poisson regression models
Tian, 2014	2001-2007	Time-series study	China	Hong Kong	117,329	/	/	0-2	Poisson regression models
Tian, 2018	2010-2012	Time-series study	China	Beijing	3630295	46.15	99.5	0-3	Generalized additive Poisson model
Wang, 2021	2013-2017	Time-series study	China	Guangdong	483861	/	37.0 ± 29.0	0-14	Quasi-Poisson generalized additive model
Weichenthal, 2017	1996-2012	cohort	Canada	Toronto	1,105,258	53%	/	0-5year	Random-effect Cox Proportional hazard models
Weichenthal, 2016	2004-2011	Case-crossover	Canada	Ontario	298,751	52.40%	/	0-3	Conditional logistic regression
Wong, 1999	1994-1995	Time-series study	China	Hong Kong	/	/	/	0-3	Poisson regression models
Xu, 2016	2013/1/1-2013/12/31	Time-series study	China	Beijing	712	40.30%	102.1 ± 73.6	0-5	Generalized-additive model
Zhao, 2017	2013-2015	Time-series study	China	Dongguan	44,801	/	42.57	0-3	Quasi-Poisson generalized additive model

Association between PM_2.5 and AECOPD events on lag0

Seventeen datasets could be analyzed for the relationship between particulate matters and AECOPD events on lag0. There was a significant association between PM_2.5 and AECOPD events on lag0 (ES = 1.01, 95%CI: 1.01-1.02, p < 0.001; I²=88.6%, P_{heterogeneity}<0.001) (Figure 2A and Supplementary Table S1). Subgroup analyses showed significant associations between PM_2.5 and hospitalization (ES = 1.01, 95%CI: 1.00-1.02, p < 0.001; I²=91.7%, P_{heterogeneity}<0.001), emergency room visit (ES = 1.01, 95%CI: 1.01-1.01, p < 0.001; I²=0.0%, P_{heterogeneity}=0.533), and outpatient visit (ES = 1.01, 95%CI: 1.00-1.03, p < 0.001) (Figure 3A and Supplementary Table S1). Significant associations of PM_2.5 (ES = 1.01, 95%CI: 1.01-1.02, p < 0.001; I²=91.7%, p < 0.001) and PM₁₀ (ES = 1.00, 95%CI: 0.99-1.02, p < 0.001) with hospitalization were also found (Figure 4A).

Figure 2. Forest plots for the association between PM_2.5 and AECOPD hospitalization, outpatient, and emergency room visit on lag0 (A), lag1 (B), lag2 (C), lag3 (D), lag4 (E), lag5 (F), lag7 (G), lag0-1 (H), lag0-2 (I), and lag0-3 (J).

Figure 3. Forest plots for the subgroup analyses (hospitalization, emergency room visit, and outpatient visit) on the association between PM_2.5 and AECOPD on lag0 (A), lag1 (B), lag2 (C), lag3 (D), lag4 (E), lag5 (F), lag7 (G), lag0-1 (H), lag0-2 (I), and lag0-3 (J).

Figure 4. Forest plots for the subgroup analyses on the association of PM_2.5 and PM₁₀ with hospitalization on lag0 (A), lag3 (B), lag4 (C), lag7 (D), and lag0-3 (E).

Association between PM_2.5 and AECOPD events on lag1

Fourteen datasets analyzed the correlation of PM_2.5 with AECOPD events on lag1. There was a significant association between PM_2.5 and AECOPD events on lag1 (ES = 1.00, 95%CI: 1.00-1.01, p < 0.001; I²=82.5%, P_{heterogeneity}<0.001) (Figure 2B and Supplementary Table S1). Subgroup analyses showed a significant association between PM_2.5 and hospitalization (ES = 1.00, 95%CI: 1.00-1.00, p < 0.001; I²=0.0%, P_{heterogeneity}=0.482), outpatient visit (ES = 1.00, 95%CI: 1.00-1.00, p < 0.001; I²=0.0%, P_{heterogeneity}=0.934), and emergency room visit (ES = 1.02, 95%CI: 1.01-1.03, p < 0.001; I²=71.7%, P_{heterogeneity}=0.007) (Figure 3B and Supplementary Table S1).

Association between PM_2.5 and AECOPD events on lag2

Pooled data of fourteen datasets showed a significant association between PM_2.5 and AECOPD events on lag2 (ES = 1.01, 95%CI: 1.01-1.01, p < 0.001; I²=90.6%, P_{heterogeneity}<0.001) (Figure 2C and Supplementary Table S1). Subgroup analyses showed a significant association between PM_2.5 and hospitalization (ES = 1.01, 95%CI: 1.00-1.01, p < 0.001; I²=84.4%, P_{heterogeneity}<0.001), outpatient visit (ES = 1.00, 95%CI: 1.00-1.00, p < 0.001; I²=1.7%, P_{heterogeneity}=0.313), and emergency room visit (ES = 1.03, 95%CI: 1.01-1.04, p < 0.001; I²=89.0%, P_{heterogeneity}<0.001) (Figure 3C and Supplementary Table S1).

Association between PM_2.5 and AECOPD events on lag3

Pooled analyses of eleven datasets suggested a significant association between PM_2.5 and AECOPD events on lag3 (ES = 1.01, 95%CI: 1.00-1.01, p < 0.001; I²=88.9%, P_{heterogeneity}<0.001) (Figure 2D and Supplementary Table S1). Subgroup analyses showed significant associations between PM_2.5 and hospitalization (ES = 1.01, 95%CI: 1.00-1.02, p < 0.001; I²=94.0%, P_{heterogeneity}<0.001), outpatient visit (ES = 1.00, 95%CI: 1.00-1.00, p < 0.001; I²=74.4%, P_{heterogeneity}=0.048), and emergency room visit (ES = 1.01, 95%CI: 1.00-1.02, p < 0.001; I²=0.0%, P_{heterogeneity}=0.889) (Figure 3D and Supplementary Table S1). Significant correlations of hospitalization with PM_2.5 (ES = 1.01, 95%CI: 1.01-1.02, p < 0.001; I²=94.0%, p < 0.001) and PM₁₀ (ES = 1.03, 95%CI: 1.00-1.06, p < 0.001) were also demonstrated (Figure 4B).

Association between PM_2.5 and AECOPD events on lag4

There was a significant association between PM_2.5 and AECOPD events on lag4 (ES = 1.00, 95%CI: 1.00-1.01, p < 0.001; I²=83.7%, P_{heterogeneity}<0.001) based on the pooled analysis of ten datasets (Figure 2E and Supplementary Table S1). Subgroup analyses showed significant associations between PM_2.5 and hospitalization (ES = 1.01, 95%CI: 1.00-1.01, p < 0.001; I²=90.1%, P_{heterogeneity}<0.001), outpatient visit (ES = 1.00, 95%CI: 1.00-1.00, p < 0.001; I²=86.7%, P_{heterogeneity}=0.006), and emergency room visit (ES = 1.00, 95%CI: 0.98-1.03, p < 0.001; I²=66.2%, P_{heterogeneity}=0.052) (Figure 3E and Supplementary Table S1). Subgroup analyses showed significant associations of hospitalization with PM_2.5 (ES = 1.01, 95%CI: 1.00-1.01, p < 0.001; I²=90.1%, p < 0.001) and PM₁₀ (ES = 1.03, 95%CI: 1.00-1.06, p < 0.001) (Figure 4C).

Association between PM_2.5 and AECOPD events on lag5

According to pooled analysis of ten datasets, there was a significant association between PM_2.5 and AECOPD events on lag5 (ES = 1.00, 95%CI: 1.00-1.00, p < 0.001; I²=75.7%, P_{heterogeneity}<0.001) (Figure 2F and Supplementary Table S1). Subgroup analyses showed significant associations between PM_2.5 and hospitalization (ES = 1.00, 95%CI: 1.00-1.01, p < 0.001; I²=84.3%, P_{heterogeneity}<0.001), emergency room visit (ES = 0.99, 95%CI: 0.96-1.03, p < 0.001; I²=77.8%, P_{heterogeneity}=0.011), and outpatient visit (ES = 1.00, 95%CI: 1.00-1.00, p < 0.001) (Figure 3F and Supplementary Table S1).

Association between PM_2.5 and AECOPD event on lag7

Pooled estimate from three datasets showed a significant association between PM_2.5 and AECOPD events on lag7 (ES = 1.00, 95%CI: 1.00-1.00, p < 0.001; I²=0.0%, P_{heterogeneity}=0.743) (Figure 2G). Subgroup analyses showed significant associations between PM_2.5 and hospitalization (ES = 1.00, 95%CI: 1.00-1.00, p < 0.001; I²=0.0%, P_{heterogeneity}=0.510), and emergency room visit (ES = 0.98, 95%CI: 0.88-1.08, p < 0.001) (Figure 3G). Subgroup analyses showed significant correlations of hospitalization with PM_2.5 (ES = 0.98, 95%CI: 1.00-1.08, p < 0.001) and PM₁₀ (ES = 1.03, 95%CI: 1.01-1.05, p < 0.001) (Figure 4D).

Association between PM_2.5 and AECOPD events on lag0-1

There was a significant association between PM_2.5 and AECOPD events on lag0-1 (ES = 1.01, 95%CI: 1.00-1.03, p < 0.001; I²=82.0%, P_{heterogeneity}<0.001) based on five datasets (Figure 2H). Subgroup analyses showed significant associations between PM_2.5 and hospitalization (ES = 1.00, 95%CI: 0.99-1.03, p < 0.001; I²=85.0%, P_{heterogeneity}<0.001), and emergency room visit (ES = 1.03, 95%CI: 1.01-1.04, p < 0.001) (Figure 3H).

Association between PM_2.5 and AECOPD events on lag0-2

There was a significant association between PM_2.5 and AECOPD events on lag0-2 (ES = 1.02, 95%CI: 1.02-1.03, p < 0.001; I²=75.0%, P_{heterogeneity}<0.001) based on eight datasets (Figure 2I). Subgroup analyses showed significant associations between PM_2.5 and hospitalization (ES = 1.02, 95%CI: 1.01-1.03, p < 0.001; I²=54.8%, P_{heterogeneity}=0.065), and emergency room visit (ES = 1.04, 95%CI: 1.01-1.06, p < 0.001) (Figure 3I).

Association between PM_2.5 and AECOPD events on lag0-3

Pooled analysis of four datasets showed a significant association between PM_2.5 and AECOPD events on lag0-3 (ES = 1.03, 95%CI: 1.02-1.04, p < 0.001; I²=0.0%, P_{heterogeneity}=0.600) (Figure 2J). Subgroup analyses showed significant associations between PM_2.5 and hospitalization (ES = 1.03, 95%CI: 1.02-1.04, p < 0.001; I²=0.0%, P_{heterogeneity}=0.916), emergency room visit (ES = 1.03, 95%CI: 1.01-1.05, p < 0.001), and outpatient visit (ES = 1.08, 95%CI: 1.01-1.16, p < 0.001) (Figure 3J). Significant associations of hospitalization with PM_2.5 (ES = 1.03, 95%CI: 1.01-1.05, p < 0.001) and PM₁₀ (ES = 1.01, 95%CI: 1.01-1.01, p < 0.001) were also revealed (Figure 4E).

Sensitivity analysis

Sensitivity analysis results showed that no individual study could influence the overall association relationship between PM_2.5 and AECOPD events on lag0, lag1, lag2, lag3, lag4, and lag5 (Supplementary Figure S1).

Publication bias

The funnel plot indicated that one study [32] had a significant publication bias (Supplementary Figure S2).

Discussion

The results showed that AECOPD events were significantly increased with an increment of 10 µg/m³ in PM_2.5 from lag0 to lag7. Similar results were observed with PM₁₀, as well as with the individual event types.

These results were supported by a previous meta-analysis by Li et al. published in 2016 [13]. Their review [13] included 59 studies that examined the association between various air pollutants (O₃, CO, NO₂, SO₂, PM₁₀, and PM_2.5) and AECOPD and reported that short-term exposure to those pollutants increased the burden of AECOPD. They also reported that those associations were stronger on lag0 and lag3, and the present study observed associations at all time points from lag0 to lag7. Wang et al. [55] performed a meta-analysis of studies from mainland China, Hong Kong, Macao, and Taiwan and showed an association of PM₁₀ and PM_2.5 with hospitalization burden of COPD, but they did not examine the outcomes based on days since exposure. Nevertheless, why the associations were stronger on lag0 and lag3 but not on lag1 or lag5 in previous studies was difficult to explain. Based on the sub-analysis of PM_2.5, Li et al. [13] observed that the main source of heterogeneity was the differences in lag periods and confounders among the studies. These two sources of bias were identified as early as 2001 [56]. In addition, the weaker or lack of association on lag5 could be due to a decreased inflammatory response with elapsing time from exposure. Still, despite the early recognition of the lag effect, these previous meta-analyses did not consider it, mainly because of a lack of data. In the present meta-analysis, the lag periods varied from 0-2 to 0-7 among the included studies. Moreover, our meta-analysis included over 7 million patients and showed that the associations were significant at all lag points, suggesting that the risk of AECOPD remains significant for 7 consecutive days following PM_2.5 and PM₁₀ exposure.

There is a wealth of literature about the relationship between air pollutants and health [12], COPD [11, 14–16], and AECOPD [10, 56, 57]. PM_2.5 can absorb chemicals from the environment, inducing oxidative stress in the airways [58–60], impairing the protective function of the immune system [60, 61], and triggering inflammatory responses that further damage the respiratory system in COPD patients [60, 62]. Nevertheless, the associations of AECOPD with smoking and pneumonia have been well documented [48, 63, 64], but data are still lacking about the link with air pollutants. It is, of course, complicated by the presence of a wide variety of pollutants both indoors and outdoors. The present meta-analysis and previous ones [13, 55] supported the effect of short-term exposure to air pollutants on AECOPD, but the impact of long-term exposure was not studied. Indeed, long-term exposure to air pollutants is a well-known risk factor for cardiovascular events and mortality [65–67]. Nevertheless, one meta-analysis revealed the association between PM_2.5 and COPD risk [68]. The associations were observed at all lag values, suggesting that the effects of PM exposure activate systemic immunity and that this activation remains. Still, the exact mechanisms involved will have to be examined. The subgroup analyses showed that the individual AECOPD events (i.e. hospitalization, emergency room visits, and outpatient visits) were associated with PM_2.5 exposure. Although these events indicate different severity levels of the patients’ condition, they nevertheless indicate that the patients feel ill, and they seek consultation according to their perceived degree of illness severity. Future studies should refine the relationship between PM_2.5 and AECOPD.

The results of the present meta-analysis must be explained with its limitations. First, the included studies were controlled for crucial covariates such as weather, temperature, humidity, and day of the week, but the adjusted covariates were different among the included studies. Second, 16 of 26 studies were from China, which might cause bias in the results. Indeed, many households in China still used coal for heating and cooking [69], which could predispose COPD patients to disease exacerbation when they were exposed to outdoor PM_2.5. Third, only the parameters in single-pollutant models were extracted for analysis purposes. The underlying interactive effects might exist among pollutants. It needs further analysis. Finally, despite being statistically significant, the ESs were small and must be interpreted with caution.

Conclusions

In conclusion, this meta-analysis showed that the risk of AECOPD events (hospitalization, emergency room, and outpatient visits) was significantly increased with an increment of 10 µg/m³ in PM_2.5 and PM₁₀ from lag0 to lag7. The present study enabled the investigation of the association between short-term exposure to PM_2.5 and PM₁₀ and AECOPD risk. Nevertheless, result interpretation must be cautious due to the study limitations. Future studies should focus on the impact of relatively longer lag-time points on AECOPD events.

Relevance to clinical practice

Short-term exposure to PM_2.5 and PM₁₀ were reported to increase the burden of AECOPD significantly. Therefore, this study aimed to conduct a meta-analysis to explore whether PM_2.5 and PM₁₀ were associated with AECOPD hospitalization, emergency room visit, and outpatient visit. The results showed that the risk of AECOPD events (hospitalization, emergency room visit, and outpatient visit) was significantly increased with an increment of 10 µg/m³ in PM_2.5 and PM₁₀ from lag0 to lag7.

Impact statement

What does this paper contribute to the wider global clinical community?

The results showed that an increment of 10 µg/m³ in particulate matter PM_2.5 and PM₁₀ from lag0 to lag7 might significantly increase the AECOPD events, including hospitalization, emergency room, and outpatient visits.

Author contributions

Niuniu Li and Liyun Wang carried out the studies, participated in collecting data, and drafted the manuscript. Niuniu Li and Liyun Wang performed the statistical analysis and participated in its design. Kun Ji and Jianling Ma participated in the acquisition, analysis, or interpretation of data and drafted the manuscript. All authors read and approved the final manuscript.

Statement

Manuscripts have been read and approved by all authors to meet the authorship requirements of this journal, and each author believes that the manuscript represents honest work if that information is not provided in another form.

Health and safety

This study does not involve any experimental operations, so there are no risks that might be involved in experiments or procedures, or that might involve instructions, materials, or formulas. This article complies with the relevant rules.

Declaration of interest

The authors report no conflict of interest.

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article [and/or] its supplementary materials.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China (No.81373588) and Beijing University of Chinese Medicine (2018-JYbZZ-JS144).