Current State of Research on the Risk of Morbidity and Mortality Associated with Air Pollution in Korea

Abstract

Purpose

The effects of air pollution on health can vary regionally. Our goal was to comprehensively review previous epidemiological studies on air pollution and health conducted in Korea to identify future areas of potential study.

Materials and Methods

We systematically searched all published epidemiologic studies examining the association between air pollution and occurrence of death, diseases, or symptoms in Korea. After classifying health outcomes into mortality, morbidity, and health impact, we summarized the relationship between individual air pollutants and health outcomes.

Results

We analyzed a total of 27 studies that provided 104 estimates of the quantitative association between risk of mortality and exposure to air pollutants, including particulate matter with aerodynamic diameter less than 10 µm, particulate matter with aerodynamic diameter less than 2.5 µm, sulfur dioxide, nitrogen dioxide, ozone, and carbon monoxide in Korea between January 1999 and July 2018. Regarding the association with morbidity, there were 38 studies, with 98 estimates, conducted during the same period. Most studies examined the short-term effects of air pollution using a time series or case-crossover study design; only three cohort studies that examined long-term effects were found. There were four health impact studies that calculated the attributable number of deaths or disability-adjusted life years due to air pollution.

Conclusion

There have been many epidemiologic studies in Korea regarding air pollution and health. However, the present review shows that additional studies, especially cohort and experimental studies, are needed to provide more robust and accurate evidence that can be used to promote evidence-based policymaking.

INTRODUCTION

The effect of air pollution on mortality and the burden of disease increases as air pollution increases, although estimates can vary from region to region. According to the Global Burden of Disease Study, ambient air pollution accounted for 7.5% of deaths globally in 2016 and was the sixth leading contributor to attributable disability-adjusted life years (DALYs) in that year.¹ Korea has experienced rapid economic growth in the last century, and the quality of the atmosphere has worsened. Air pollution reduction policies, such as the Special Law on Air Quality in the Seoul metropolitan area, have had limited effect on particulate matter (PM) pollution, and the overall air quality remains poor. Concentrations of PM with aerodynamic diameter less than 10 µm (PM₁₀) have improved over the past decade, reaching the lowest national average of 45 µg/m³ in 2012, and then rebounding to a level of 47 µg/m³ in 2016. However, the concentration of nitrogen dioxide (NO₂) has remained relatively constant, with no large changes. The average values of ozone (O₃) concentration are continuously increasing.²

Epidemiological studies on the health effects of air pollution have been actively conducted in many countries. In particular, time series studies to examine the short-term effects of air pollution have been conducted worldwide and have yielded relatively consistent results.³,⁴ However, cohort studies to assess the long-term effects of air pollution have been primarily conducted in Western countries that have relatively low concentrations of air pollutants. Due to a lack of direct evidence at higher global concentrations, the integrated expose–response (IER) model was developed. The IER combines information on PM-mortality associations from non-outdoor sources, including secondhand smoke, active smoking, and household air pollution,⁵ and has been used to estimate the disease burden attributable to PM with aerodynamic diameter less than 2.5 µm (PM_2.5).¹ As the use of IER requires a strict assumption of equal toxicity per unit dose across these non-outdoor sources, cohort studies are needed that reflect the different air pollution concentrations in different regions.⁶

The health effects of air pollution can vary regionally depending on the composition of pollutants or characteristics of the population at risk. The regional differences in PM_2.5 mortality risk estimates can likely be attributed to geographic variation in particle composition or the spatial heterogeneity of constituents,⁷ as well as differences in the total air pollution mixture.⁸ Regional differences of topography, which may lead to regional differences of exposure error, can contribute to regional differences in PM risk estimates.⁹

To accurately understand the impact of air pollution on health in Korea, the results of research performed specifically for Korea are needed. Since the publication of time series research starting in 1999 in Korea,¹⁰ many epidemiological studies have been conducted; however, the results of these studies have not been systematically summarized. To accurately assess the impact of air pollution in Korea and to clarify future research directions, systematic sorting of epidemiological studies on air pollution conducted in Korea is required. The aim of the present analysis was to comprehensively review previous epidemiological studies on air pollution and health conducted in Korea to identify future study needs.

LITERATURE SEARCH

We conducted a literature search in PubMed using the search terms (“air pollution”[MeSH Terms] OR (“air”[All Fields] AND “pollution”[All Fields]) OR “air pollution”[All Fields]) AND (“mortality”[Subheading] OR “mortality”[All Fields] OR “mortality”[MeSH Terms]) AND (“Korea”[MeSH Terms] OR “Korea”[All Fields]) and ((“air pollution”[MeSH Terms] OR (“air”[All Fields] AND “pollution”[All Fields]) OR “air pollution” [All Fields]) AND (“epidemiology”[Subheading] OR “epidemiology” [All Fields] OR “morbidity”[All Fields] OR “morbidity” [MeSH Terms]) AND (“Korea”[MeSH Terms] OR “Korea”[All Fields])) NOT (“mortality”[Subheading] OR “mortality” [MeSH Terms]) to find published studies on the associations of air pollution with mortality and morbidity respectively in Korea, between January 1990 and July 2018.

We also searched for health impact assessment studies using the same search engine and the search terms (“number” [All Fields] AND (“death”[MeSH Terms] OR “death”[All Fields] OR “deaths”[All Fields])) OR “burden of disease”[All Fields] OR “health impact assessment”[All Fields] AND “Korea”[All Fields] AND (“air pollution”[All Fields] OR “ambient”[All Fields]).

After reviewing the title and abstract of each article, we selected epidemiological studies that reported associations between exposure to air pollution and mortality or morbidity. We then summarized these articles according to their characteristics and results.

The initial search for mortality and morbidity returned 87 and 195 results, respectively. After excluding articles that did not meet the inclusion criteria (Fig. 1), there remained 27 (Table 1) and 37 studies (Table 2) on mortality and morbidity, respectively. One of the mortality study also reported morbidity results, so a total of 38 studies were included in the present review. The search for health impact analyses returned 22 studies; four articles remained after a review of titles and abstracts.

Fig. 1. Selection of papers.

Table 1. Epidemiological Studies on Air Pollution and Mortality in Korea between 1999 and 2018.

No.	Author (year)	Study design	Study period	Location	Outcome	Pollutant	Unit	Effect size
1	Lee, et al. (1999)10	Time series	1991–1995	Seoul	Non-accidental	SO2	50 ppb	RR 1.078 (1.057, 1.099)
				Ulsan	Non-accidental	SO2	50 ppb	RR 1.051 (0.991, 1.115)
				Seoul	Non-accidental	TSP	100 μg/m3	RR 1.051 (1.031, 1.072)
				Ulsan	Non-accidental	TSP	100 μg/m3	RR 0.999 (0.961, 1.039)
				Seoul	Non-accidental	1 hr max O3	50 bbp	RR 1.015 (1.005, 1.025)
				Ulsan	Non-accidental	1 hr max O3	50 bbp	RR 1.020 (0.889, 1.170)
2	Hong, et al. (1999)13	Time series	1995	Incheon	Total	TSP	10 μg/m3	1.2% (0.2, 2.2)
					Total	PM10	10 μg/m3	1.2% (0.2, 2.1)
3	Lee, et al. (1999)12	Case-crossover		Seoul	Non-accidental	SO2	50 ppb	RR 1.023 (1.016, 1.084)
					Non-accidental	Maximum O3	50 ppb	RR 1.023 (0.999, 1.048)
					Non-accidental	TSP	100 μg/m3	RR 1.010 (0.988, 1.032)
4	Hong, et al. (1999)11	Time series	1995–1996	Incheon	Total	PM10	10 μg/m3	RR 1.007 (1.001, 1.0013)
					Total	NO2		RR 1.0026 (1.0006, 1.0046)
					Total	SO2		RR 1.0023 (0.9996, 1.0051)
					Total	CO		RR 1.0019 (0.9990, 1.0049)
					Total	O3		RR 0.9951 (0.9908, 0.9994)
5	Lee, et al. (2000)59	Time series	1991–1997	7 cities	Total	TSP	100 μg/m3	0.5–4%
					Total	SO2	50 ppb	RR 1.03 (1.01, 1.05)
6	Kwon, et al. (2001)20	Time series	1994–1998	Seoul	Total	PM10	IQR (42.1 μg/m3)	OR 1.014 (1.006, 1.022)
					Total	CO	IQR (0.59 ppm)	OR 1.022 (1.017, 1.029)
					Total	NO2	IQR (14.6 ppb
					Total	SO2	IQR (9.9 ppb)	OR 1.020 (1.012, 1.028)
					Total	O3	IQR (20.5 ppb)	OR 1.010 (1.002, 1.017)
7	Hong, et al. (2002)60	Time series	1995–1998	Seoul	Stroke	PM10	IQR	1.5% (1.3, 1.8)
					Stroke	O3	IQR	2.9% (0.3, 5.5)
					Stroke	NO2	IQR	3.1% (1.1, 5.1)
					Stroke	SO2	IQR	2.9% (0.8, 5.0)
					Stroke	CO	IQR	4.1% (1.1, 7.2)
8	Hong, et al. (2002)21	Time series	1991–1997	Seoul	Ischemic stroke	TSP	IQR	RR 1.03 (1.00, 1.06)
					Ischemic stroke	SO2		RR 1.04 (1.01, 1.08)
					Ischemic stroke	NO2	IQR	RR 1.04 (1.01, 1.07)
					Ischemic stroke	CO	IQR	RR 1.06 (1.02, 1.09)
					Ischemic stroke	O3	IQR	RR 1.06 (1.02, 1.10)
9	Ha, et al. (2003)61	Cohort			Total (postneonates)	PM10	IQR (42.9 μg/m3)	RR 1.142 (1.096, 1.190)
					Respiratory (postneonates)	PM10	IQR (42.9 μg/m3)	RR 2.018 (1.784, 2.283)
10	Kim, et al. (2003)15	Time series	1995–1999	Seoul	Non-accidental	PM10	IQR (43.12 μg/m3)	3.7% (2.1, 5.4)
					Respiratory	PM10	IQR (43.12 μg/m3)	13.9% (6.8, 21.5)
					Cardiovascular	PM10	IQR (43.12 μg/m3)	4.4% (-1.0, 9.0)
					Cerebrovascular	PM10	IQR (43.12 μg/m3)	6.3% (2.3, 10.5)
11	Kim, et al. (2004)62	Time series	1997–2004	Seoul	Non-accidental	PM10 (mean)	IQR (42.11 μg/m3)	RR 1.021 (1.009, 1.035)
					Non-accidental	PM10 (SD)	IQR (11.93 μg/m3)	RR 1.025 (1.000, 1.028)
12	Lee, et al. (2007)63	Time series	2000–2004	Seoul	Non-accidental	Asian dust event		Larger effect sizes in the model without Asian dust event
13	Cho, et al. (2008)64	Time series	2001	Seoul	Respiratory	Fine particle count	IQR (10.221 number/cm3)	5.73% (5.03, 6.45)
					Respiratory	Respiratory particle count	IQR (10.38 number/cm3)	5.82% (5.13, 6.53)
14	Son, et al. (2008)22	Case-crossover	1999–2003	Seoul	Infant	PM10	1 μg/m3	OR 1.000 (0.998, 1.002)
					Infant	NO2	1 unit	OR 1.002 (0.994, 1.009)
					Infant	SO2	1 unit	OR 1.015 (0.973, 1.058)
					Infant	CO	1 unit	OR 1.029 (0.833, 1.271)
					Infant	O3	1 unit	OR 0.984 (0.977, 0.992)
15	Yi, et al. (2010)75	Case-crossover	2000–2006	Seoul	Non-accidental	PM10	10 μg/m3	0.28% (0.12, 0.44)
					Cardiovascular	PM10	10 μg/m3	0.51% (0.19, 0.83)
					Respiratory	PM10	10 μg/m3	0.59% (-0.08, 1.26)
16	Kim, et al. (2010)19	Case-crossover	2004	7 cities	Suicide	PM10	IQR	9.0% (2.4, 16.1)
					Suicide	PM2.5	IQR	10.1% (2.0, 19.0)
17	Park, et al. (2011)65	Time-series	1999–2007	Seoul	Non-accidental (high temp. ≥26.2℃)	SO2	0.5 ppb	0.83% (0.42, 1.25)
					Non-accidental (low temp. <26.2℃)	SO2	0.5 ppb	0.21% (0.07, 0.36)
18	Son, et al. (2011)66	Birth cohort	2004–2007	Seoul	All-cause infant	TSP	IQR	HR 1.44 (1.06, 1.97)
					All-cause infant	PM10	IQR	HR 1.65 (1.18, 2.31)
					All-cause infant	PM2.5	IQR	HR 1.53 (1.22, 1.90)
					All-cause infant	PM10-2.5	IQR	HR 1.19 (0.83, 1.70)
					Respiratory infant	TSP	IQR	HR 3.78 (1.18, 12.13)
					Respiratory infant	PM10	IQR	HR 6.20 (1.50, 25.66)
					Respiratory infant	PM2.5	IQR	HR 3.15 (1.26, 7.85)
					Respiratory infant	PM10-2.5	IQR	HR 2.86 (0.76, 10.85)
19	Son, et al. (2012)14	Case-crossover	2000–2007	Seoul	Total	PM10	IQR	0.94% (0.25, 1.62)
					Total	NO2	IQR	2.27% (1.03, 3.53)
					Total	SO2	IQR	1.94% (0.80, 3.09)
					Total	CO	IQR	2.21% (1.00, 3.43)
					Total	O3	IQR	Positive/NS
					Cardiovascular	PM10	IQR	1.95% (0.64, 3.27)
					Cardiovascular	NO2	IQR	4.82% (2.18, 7.54)
					Cardiovascular	SO2	IQR	3.64% (1.46, 5.87)
					Cardiovascular	CO	IQR	4.32% (1.77, 6.92)
					Cardiovascular	O3	IQR	Positive/NS
20	Heo, et al. (2014)67	Time-series	2003–2007	Seoul	Non-accidental, cardiovascular, respiratory	PM2.5 and components		Percentage of excess risk by PM3.5 and components
21	Lim, et al. (2014)68	GWR	2008–2010	Seoul	Cardiovascular	PM10		Mean β (SE) 0.956 (0.102)
22	Ha, et al. (2015)69	Case-crossover	2002–2008	7 citie	Unintentional injury	PM10	IQR (48.3 μg/m3)	NS
					Unintentional injury	SO2	IQR (0.005 ppm)	OR 1.119 (1.022, 1.226)
					Unintentional injury	NO2	IQR (0.02 ppm)	OR 1.208 (1.043, 1.400)
					Unintentional injury	O3	IQR (0.03 ppm)	NS
					Unintentional injury	CO	IQR (0.36 ppm)	OR 1.012 (1.000, 1.024)
23	Kim, et al. (2017)16	Time-series	1993–2009	7 cities	Non-accidental	PM10	10 μg/m3	0.51% (0.01, 1.01)
24	Kim, et al. (2018)70	Time-series	1993–2009	7 cities	Non-accidental	PM10	Daily concentrations of ≥75 μg/m3	0.48% (0.30, 0.60)
					Cardiovascular	PM10	Daily concentrations of ≥75 μg/m3	0.48% (0.14, 0.82)
					Respiratory	PM10	Daily concentrations of ≥75 μg/m3	1.13% (0.37, 1.89)
25	Kim, et al. (2017)18	Cohort	2007–2013	Seoul	Composite cardiovascular events	PM2.5	1 μg/m3	HR 1.41 (1.32, 1.50)
					All-cause	PM2.5	1 μg/m3	HR 1.32 (1.22, 1.43)
					Cardiovascular	PM2.5	1 μg/m3	HR 1.36 (1.11, 1.66)
					Composite cardiovascular events	CO	IQR (0.25 ppm)	HR 1.79 (1.61, 1.99)
					All-cause	CO	IQR (0.25 ppm)	HR 1.72 (1.52, 1.94)
					Cardiovascular	CO	IQR (0.25 ppm)	HR 2.96 (2.12, 4.14)
					Composite cardiovascular events	SO2	IQR (2.54 ppb)	HR 1.94 (1.78, 2.11)
					All-cause	SO2	IQR (2.54 ppb)	HR 1.73 (1.55, 1.92)
					Cardiovascular	SO2	IQR (2.54 ppb)	HR 1.50 (1.14, 1.96)
					Composite cardiovascular events	NO2	IQR (18.4 ppb)	HR 2.30 (2.08, 2.55)
					All-cause	NO2	IQR (18.4 ppb)	HR 1.79 (1.59, 2.03)
					Cardiovascular	NO2	IQR (18.4 ppb)	HR 2.67 (1.94, 3.69)
					Composite cardiovascular events	O3	IQR (15.9 ppb)	HR 0.63 (0.63, 0.73)
					All-cause	O3	IQR (15.9 ppb)	HR 0.68 (0.63, 0.73)
					Cardiovascular	O3	IQR (15.9 ppb)	HR 0.59 (0.49, 0.71)
26	Kim, et al. (2017)17	Cohort	2002–2014	Korea	Non-accidental	PM10	10 μg/m3	HR 1.05 (0.99, 1.11)
					Cardiovascular	PM10	10 μg/m3	HR 1.02 (0.90,1.16)
					Cerebrovascular	PM10	10 μg/m3	HR 1.14 (0.93, 1.39)
					Respiratory	PM10	10 μg/m3	HR 1.19 (0.91, 1.57)
					Cancer	PM10	10 μg/m3	HR 1.02 (0.95, 1.10)
					Lung cancer	PM10	10 μg/m3	HR 0.96 (0.82,1.13)
27	Lee, et al. (2018)71	Case-crossover	2002–2013	26 cities	Suicide	PM10	IQR	Increased OR 1.2% (0.2, 2.3)
						NO2	IQR	Increased OR 4.3% (1.9, 6.7)
						SO2	IQR	Increased OR 2.2% (0.7, 3.8)
						CO	IQR	Increased OR 2.4% (0.9, 3.8)
						O3	IQR	Increased OR 1.5% (-0.3, 3.2)

TSP, total suspended particles; IQR, interquartile range; OR, odds ratio; RR, relative risk; HR, hazard ratio; NS, not significant; GWR, geographically weighted regression; 7 cities, Seoul, Incheon, Daejeon, Gwangju, Daegu, Busan, Ulsan.

Table 2. Epidemiological Studies on Air Pollution and Morbidity in Korea between 1999 and 2018.

No.	Author (year)	Study design	Study period	Location	Outcome	Pollutant	Unit	Effect size
1	Lee, et al. (2002)23	Time series		Seoul	Asthma hospitalization	PM10	IQR (40.4 µg/m3)	RR 1.07 (1.04, 1.11)
					Asthma hospitalization	SO2	IQR (4.4 ppb)	RR 1.11 (1.06, 1.17)
					Asthma hospitalization	NO2	IQR (14.6 ppb)	RR 1.15 (1.10, 1.20)
					Asthma hospitalization	O3	IQR (21.7 ppb)	RR 1.12 (1.07, 1.16)
					Asthma hospitalization	CO	IQR (1.0 ppm)	RR 1.16 (1.10, 1.22)
2	Lee, et al. (2005)24	Panel study	2003	Seoul	Upper respiratory symptoms	NO2		OR 1.12 (1.01, 1.24)
					Lower respiratory symptoms	NO2		OR 1.18 (1.06, 1.31)
					Lower respiratory symptoms	SO2		OR 1.12 (1.01, 1.25)
					Lower respiratory symptoms	CO		OR 1.16 (1.02, 1.32)
3	Son, et al. (2006)25	Time series	2002	Seoul	Asthma hospitalization (highest SES)	O3		RR 1.12 (1.00, 1.25)
					Asthma hospitalization (moderate SES)	O3		RR 1.24 (1.08, 1.43)
					Asthma hospitalization (lowest SES)	O3		RR 1.32 (1.11, 1.58)
4	Lee, et al. (2007)72	Natural experiment	2002	Busan	Childhood asthma hospitalization		RR post Asian game period/RR baseline	0.73 (0.49, 1.11)
5	Lee, et al. (2006)73	Time series	2002	Seoul	Asthma hospitalization	PM10	IQR	31% (14, 51)
					Asthma hospitalization	SO2	IQR	29% (8, 53)
					Asthma hospitalization	NO2	IQR	29% (5, 58)
6	Seo, et al. (2007)33	Cohort	2002–2003	Seoul	Low birth weight	CO	IQR	RR 1.081 (1.002, 1.166)
					Low birth weight	SO2	IQR	RR 1.145 (1.036, 1.267)
					Low birth weight	PM10	IQR	RR 1.053 (1.002, 1.108)
					Low birth weight	NO2	IQR	RR 1.003 (0.954, 1.055)
7	Moon, et al. (2009)74			4 cities	Respiratory symptoms	5 criteria pollutants		Significant positive association with SO2 and NO2
8	Seo, et al. (2010)34	Cohort	2004	Seoul	Low birth weight	PM10		OR 1.08 (0.99, 1.18)
				Busan	Low birth weight	PM10		OR 1.24 (1.02, 1.52)
				Daegu	Low birth weight	PM10		OR 1.19 (1.04, 1.37)
				Incheon	Low birth weight	PM10		OR 1.12 (0.98, 1.28)
				Gwangju	Low birth weight	PM10		OR 1.22 (0.98, 1.52)
				Daejeon	Low birth weight	PM10		OR 1.06 (1.00, 1.11)
				Ulsan	Low birth weight	PM10		OR 1.19 (1.03, 1.38)
9	Yi, et al. (2010)75*	Case-crossover	2001–2006		Cardiovascular hospitalization	PM10	10 μg/m3	0.77% (0.53, 1.01)
					Respiratory hospitalization	PM10	10 μg/m3	1.19% (0.94, 1.44)
10	Kim, et al. (2011)26	Cohort			12-month prevalence of wheeze	O3	5 ppb	OR 1.372 (1.016, 1.852)
11	Lim, et al. (2012)31	Panel study		Seoul	Depression (SGDS-K)	PM10	IQR	17.0% (4.9, 30.5)
					Depression (SGDS-K)	NO2	IQR	32.8% (12.6, 65.6)
					Depression (SGDS-K)	O3	IQR	43.7% (11.5, 85.2)
12	Kim, et al. (2012)76	Panel study		Seoul	Insulin resistance	PM10, O3, NO2	IQR	Significantly increased
13	Kim, et al. (2013)77	Cross sectional			Allergic diseases	Traffic related pollutants	Polluted vs. non-polluted school	OR 2.12 (1.41, 3.19)
14	Kim, et al. (2013)27	Cohort			Airway hyperresponsiveness	O3		OR 1.60 (1.13, 2.27)
					New episodes of wheezing	O3		OR 1.92 (0.96, 3.83)
15	Han, et al. (2013)78				Hemorrhagic fever with renal syndrome	PM10	1 μg/m3	0.013 increase of monthly cases
16	Son, et al. (2013)79		2003–2008	8 cities	Allergic disease hospital admission	PM10	IQR (30.7 μg/m3)	2.2% (0.5, 3.9)
					Asthma hospital admission	PM10	IQR (30.7 μg/m3)	2.8% (1.3, 4.4)
					Respiratory hospital admission	PM10	IQR (30.7 μg/m3)	1.7% (0.9, 2.6)
					Cardiovascular hospital admission	PM10	IQR (30.7 μg/m3)
					Allergic disease hospital admission	NO2	IQR (12.2 ppb)	2.3% (0.6, 4.0)
					Asthma hospital admission	NO2	IQR (12.2 ppb)	2.2% (0.3, 4.1)
					Respiratory hospital admission	NO2	IQR (12.2 ppb)	2.2% (0.6, 3.7)
					Cardiovascular hospital admission	NO2	IQR (12.2 ppb)	2.2% (1.1, 3.4)
17	Park, et al. (2013)80	Time series		7 cities	Asthma admission	PM10, CO, O3, NO2	Children vs. adult	Lower risk in children for PM10 and CO
18	Kim, et al. (2014)81	Cohort			Neurodevelopment (MDI)	PM10		β=-2.83; p=0.003
19	Hwang, et al. (2014)82	Retrospective cohort		Seoul	Neurodevelopment (PDI)	PM10		β=-3.00; p=0.002
20	Han, et al. (2015)28	Time series	2004–2013		Tuberculosis	SO2	IQR	RR 1.07 (1.03, 1.12)
21	Kim, et al. (2015)83	Case-crossover		Korea	Hourly asthma ED visit	PM10-2.5	IQR	OR 1.05 (1.00, 1.11)
					Hourly asthma ED visit	O3	IQR	OR 1.10 (1.04, 1.16)
22	Jang, et al. (2015)84	Ecological		Korea	Monthly malaria incidence	NO2		β=-0.884, p<0.01
23	Shim, et al. (2016)85	Cross sectional	2010–2013	Korea	Benign prostate hyperplasia	NO2		OR 2.23 (1.55, 2.39)
					Benign prostate hyperplasia	SO2		OR 2.02 (1.42, 2.88)
24	Kang, et al. (2016)86	Time series	2006–2013	Seoul	Cardiac arrest	PM2.5	10 μg/m3	1.30% (0.20, 2.41)
25	Kim, et al. (2016)87	Cross sectional			Allergic rhinitis	CO (during the first year of life)	100 ppb	OR 1.10 (1.03, 1.19)
					Atopic dermatitis	CO (past 12 months)	1 ppm	OR 8.11 (1.06, 62.12)
26	Kim, et al. (2016)88	Cross sectional			Asthma	NO2		OR 1.67 (1.03, 2.71)
					Allergic rhinitis	Black carbon		OR 1.60 (1.36, 1.90)
					Allergic rhinitis	SO2		OR 1.09 (1.01, 1.17)
					Allergic rhinitis	NO2		OR 1.18 (1.07, 1.30)
27	Han, et al. (2016)89	Time series	2004–2014	Seongdong-gu, Seoul	Intracerebral hemorrhage	PM10	RR	1.09 (1.02, 1.15)
					Subarachnoid hemorrhage	O3	RR	1.32 (1.10, 1.58)
28	Kim, et al. (2016)32	Cohort	2002–2010	Korea	Major depressive disorder	PM2.5	10 μg/m3	HR 1.44 (1.17-1.78)
29	Lee, et al. (2016)90	Cross sectional	2008–2010	Korea	Hypertension	PM10	10 μg/m3	OR 1.042 (1.009, 1.077)
					Hypertension in >30 years old	PM10	10 μg/m3	OR 1.044 (1.009, 1.079)
					Stroke	PM10	10 μg/m3	OR 1.044 (0.979, 1.114)
					Angina	PM10	10 μg/m3	OR 0.977 (0.901, 1.059)
					Hypertension	NO2	10 ppb	OR 1.077 (1.044, 1.112)
					Hypertension	in >30 years old NO2	10 ppb	OR 1.080 (1.043, 1.118)
					Stroke	NO2	10 ppb	OR 1.073 (0.994, 1.157)
					Angina	NO2	10 ppb	OR 1.047 (0.968, 1.134)
					Hypertension	CO	10 ppb	OR 1.123 (0.963, 1.310)
					Hypertension in >30 years old	CO	10 ppb	OR 1.129 (0.963, 1.387)
					Stroke	CO	10 ppb	OR 1.336 (0.987, 2.011)
30	Lee, et al. (2017)91	Cross sectional	2008–2011	Korea	Pterygium	PM10	5 μg/m3	OR 1.23 p=0.023
31	Chung, et al. (2017)92				Cardioembolic stroke	PM10		Significantly increased
					Cardioembolic stroke	SO2		Significantly increased
32	Kim, et al. (2016)93	Randomized intervention trial			Atopic dermatitis	Indoor VOC	Environmentally friendly vs. PVC wallpaper	More improvement in environmentally friendly wallpaper group
33	Ha, et al. (2017)29	Ecological	1999–2008		Male lung cancer	Indoor radon	10 Bq/m3	0.01
					Female children non-Hodgkin's lymphoma	Indoor radon	10 Bq/m3	0.07
34	Hwang, et al. (2017)94	Time series			Cardiovascular ED visit	NH4+ (PM2.5 component)		RR 1.05 (1.01, 1.09)
35	Kim, et al. (2017)18*	Cohort	2007–2013	Seoul	Acute myocardial infarction	PM2.5	1 μg/m3	1.36 (1.19, 1.56)
					Congestive heart failure	PM2.5	1 μg/m3	1.44 (1.29, 1.61)
					Stroke	PM2.5	1 μg/m3	1.39 (1.27, 1.52)
					Acute myocardial infarction	CO	IQR (0.25 ppm)	2.12 (1.72, 2.61)
					Congestive heart failure	CO	IQR (0.25 ppm)	1.86 (1.56, 2.21)
					Stroke	CO	IQR (0.25 ppm)	2.00 (1.73, 2.30)
					Acute myocardial infarction	SO2	IQR (2.54 ppb)	1.82 (1.52, 2.19)
					Congestive heart failure	SO2	IQR (2.54 ppb)	2.00 (1.73, 2.32)
					Stroke	SO2	IQR (2.54 ppb)	2.25 (2.00, 2.54)
					Acute myocardial infarction	NO2	IQR (18.4 ppb)	1.81 (1.46, 2.25)
					Congestive heart failure	NO2	IQR (18.4 ppb)	2.40 (2.02, 2.85)
					Stroke	NO2	IQR (18.4 ppb)	2.65 (2.29, 3.06)
					Acute myocardial infarction	O3	IQR (15.9 ppb)	0.71 (0.63, 0.82)
					Congestive heart failure	O3	IQR (15.9 ppb)	0.64 (0.58, 0.71)
					Stroke	O3	IQR (15.9 ppb)	0.60 (0.55, 0.65)
36	Lamichhane, et al. (2017)30	Case-control		Korea	Lung cancer	PM10	10 μg/m3	OR 1.09 (0.96, 1.23)
						NO2	10 ppb	OR 1.10 (1.00, 1.22)
37	Yi, et al. (2017)95	Cross sectional	2010	Seoul	Children's atopic eczema	Road density		OR 1.08 (1.01, 1.15)
						Road proximity		OR 1.15 (1.01, 1.31)
38	Lamichhane, et al. (2018)96	Birth cohort			Fetal growth (BPD)	PM10	10 μg/m3	-0.26 mm (-0.41, -0.11)
						NO2	10 μg/m3	-0.30 mm (-0.59, -0.03)

SES, socioeconomic status; SGDS-K, Short Geriatric Depression Scale-Korean; MDI, mental developmental index; PDI, psychomotor developmental index; BPD, biparietal diameter; VOC, volatile organic carbon; PVC: polyvinyl chloride; IQR, interquartile range; OR, odds ratio; RR, relative risk; HR, hazard ratio; 7 cities, Seoul, Incheon, Daejeon, Gwangju, Daegu, Busan, Ulsan.

^*From the search results of mortality studies.

Air pollution and mortality

Among the included studies, the earliest reports regarding an association between air pollution and mortality in Korea were published in 1999.¹⁰,¹¹,¹²,¹³ Three of these were time series studies and one was a case-crossover study. Both time series and case-crossover designs are suitable for analysis of acute effects (in days) of short-term exposure to air pollution. One time series analysis was conducted in Seoul and Ulsan. That study reported that the daily variation of ambient concentrations of sulfur dioxide (SO₂), total suspended particles (TSP), and O₃ in Seoul were significantly associated with increased non-accidental mortality.¹⁰ In the same year, the results of reanalysis of Seoul data from the previous time series using a case-crossover approach, in which each participant became its own control, were reported, showing that only SO₂ was significantly associated with non-accidental mortality.¹² Another time series study conducted in Incheon showed that, in addition to TSP, a 10-µg/m³ increase in the daily mean concentration of PM₁₀ was also associated with a 1.2% increase in total mortality.¹³ The remaining study was the first to examine the effects of all five criteria pollutants [PM₁₀, SO₂, NO₂, carbon monoxide (CO), and O₃] on mortality in Seoul. That study reported that the previous day's concentrations of PM₁₀ and NO₂ were significantly associated with increased daily mortality [relative risks (RRs) of 1.0007 and 1.0026 for PM₁₀ and NO₂, respectively].¹¹

After 1999, most subsequent studies examined the associations between air pollutants and total or non-accidental mortality using time series analysis and a case-crossover design. However, the effect sizes varied according to different studies. For instance, the percent increase in mortality for an interquartile range (IQR) increment in PM₁₀ ranged between 0.9%¹⁴ and 3.7%.¹⁵ This may be due to different factors of these studies, including the study period and area, and a multi-city study may provide more robust effect size. There were few multi-city studies and even fewer reported associations with total mortality. The most recent such study stated that a 10-µg/m³ increase in daily ambient PM₁₀ was associated with a 0.51% increase in mortality.¹⁶

The effects of air pollution are not only acute but also chronic, and long-term exposure is generally expected to have a much higher effect size than short-term exposure. However, the chronic effect of air pollution has rarely been examined in Korea. In fact, there were only two studies reporting long-term effects of PM exposure on mortality among our search results, one each for PM₁₀ and PM_2.5. Kim, et al.¹⁷ analyzed a sample cohort of the National Health Insurance Service and reported a marginally significant 5% increase in mortality per a 10-µg/m³ increase in annual PM₁₀ concentration. Another study reported a hazard ratio (HR) of 1.32 for all-cause mortality with an increment of 1 µg/m³ in PM_2.5.¹⁸ Long-term exposure to other gaseous pollutants was also found to be associated with increased risk of mortality, and CO, SO₂, and NO₂ showed HRs of 1.72, 1.73, and 1.79 for each IQR increase, respectively.

The effect of air pollution exposure on mortality is cause-specific, and the related cardiovascular and respiratory effects are well known. There have been several reports on cardiovascular and respiratory mortality owing to air pollution in Korea. An interesting cause of death that shows an association with air pollution is suicide. In a case-crossover study conducted using data from seven metropolitan cities in Korea (Seoul, Incheon, Daejeon, Gwangju, Daegu, Busan, and Ulsan), the authors reported that an IQR increase of PM_2.5 was associated with a 10.1% increase in the number of suicides.¹⁹

Most gaseous air pollutants (SO₂, NO₂, and CO) showed consistently significant associations with increased mortality. For acute exposure, an IQR increase of SO₂, NO₂, and CO increased daily mortality about 2%, and an IQR increase in chronic exposure to those three pollutants showed consistent RRs of around 1.7 (Table 1). However, the association between ambient O₃ concentration and mortality seems inconclusive. Two studies reported significant positive associations of O₃ concentration with total mortality²⁰ and ischemic stroke mortality.²¹ However, we also found reports of significant negative associations with all-cause,¹³,¹⁸ cardiovascular,¹⁸ and infant mortality.²²

Air pollution and morbidity

Asthma and respiratory diseases were among the first specific disorders analyzed in Korea. A time series analysis conducted in Seoul reported that an IQR increase in PM₁₀, SO₂, NO₂, CO, and O₃ showed significant RRs for children's asthma hospitalization of 1.07, 1.11, 1.15, 1.16 and 1.12, respectively.²³ Another study reported the results of a children's panel for NO₂ exposure showing an OR of 1.12 for upper respiratory symptoms and ORs for lower respiratory symptoms of 1.18, 1.12, and 1.16 for increased exposures to NO₂, SO₂, and CO, respectively.²⁴ O₃ was also associated with children's asthma hospitalization, especially in groups with lower socioeconomic status (RR: 1.32, 95% CI: 1.11, 1.58).²⁵ In a cohort study, O₃ concentration was associated with a 12-month prevalence of wheeze²⁶ and airway hyperresponsiveness²⁷ in children. Other allergic disorders, such as allergic rhinitis and atopic dermatitis, were also associated with air pollution (Table 2).

Similar to the association between air pollution and cardiovascular mortality, the morbidity of cardiovascular and cerebrovascular diseases, such as stroke, myocardial infarction, and hypertension, were also significantly associated with increased exposure to air pollution. A time series analysis reported that NO₂ increased stroke (RR=1.2, p-value=0.001),²⁸ and a cohort study reported that long-term exposure to PM_2.5, CO, SO₂, and NO₂ increased the risk of acute myocardial infarction, congestive heart failure, and stroke (Table 2).¹⁸

We found two studies examining the association of air pollution with cancer. In these recent studies, indoor radon concentrations were associated with an increased risk of male lung cancer and non-Hodgkin's lymphoma in girls,²⁹ and conventional air pollutants (PM₁₀ and NO₂) were associated with lung cancer with marginal significance.³⁰

Similar to the association of suicide with air pollution, depressive symptoms were also found to be associated with air pollution in Korea. A panel study examining air pollution and depressive symptoms was one of the first to report such an association.³¹ An association between PM_2.5 and major depressive disorder was also found in a community-based urban cohort.³²

Birth outcome has been another subject of analysis. PM₁₀, SO₂, and CO exposures were reported to have significant associations with low birth weight in a cohort study.³³,³⁴

Health impact assessment

Among four studies (Table 3), two calculated the attributable number of deaths,³⁵,³⁶ one calculated the attributable number of deaths and morbidity,³⁷ and a fourth calculated DALYs.³⁸

Table 3. Studies Estimating Health Impact of Air Pollution Conducted in Korea.

No.	Autor (year)	Study period	Location	Outcome	Pollutant	Health impact (person)
1	Leem, et al. (2015)37	2010	Seoul metropolitan area	Attributable number of deaths	PM2.5	15346
				Attributable number of respiratory hospital admission	PM10	12511
				Attributable number of cardiovascular hospital admission	PM10	12351
				Attributable number of lung cancer incidence	PM10	1403
				Attributable number of asthma attack (children)	PM10	11389
				Attributable number of asthma attack (adults)	PM10	44006
				Attributable number of chronic bronchitis	PM10	20490
				Attributable number of acute bronchitis	PM10	278346
2	Yorifuji, et al. (2015)35	2009	Seoul	Attributable number of deaths	PM10 over 20 μg/m3	5840
			Busan	Attributable number of deaths	PM10 over 20 μg/m3	2465
			Daegu	Attributable number of deaths	PM10 over 20 μg/m3	1466
			Incheon	Attributable number of deaths	PM10 over 20 μg/m3	1931
			Daejeon	Attributable number of deaths	PM10 over 20 μg/m3	599
			Gwangju	Attributable number of deaths	PM10 over 20 μg/m3	698
			Ulsan	Attributable number of deaths	PM10 over 20 μg/m3	539
3	Yoon, et al. (2015)38	2007	Korea	Disability-adjusted life years	Out door air pollution	6.89/1000 person
4	Han, et al. (2018)36		Korea	Attributable number of deaths	PM2.5	11924
			Seoul	Attributable number of deaths	PM2.5	1763
			Busan	Attributable number of deaths	PM2.5	947
			Daegu	Attributable number of deaths	PM2.5	672
			Incheon	Attributable number of deaths	PM2.5	309
			Gwangju	Attributable number of deaths	PM2.5	657
			Daejeon	Attributable number of deaths	PM2.5	342
			Ulsan	Attributable number of deaths	PM2.5	222
			Sejong	Attributable number of deaths	PM2.5	49

There were substantial differences in the attributable number of deaths among the study results. For instance, Leem, et al.³⁷ estimated the number of deaths attributable to PM_2.5 to be 15346 in the Seoul metropolitan area, whereas Han, et al.³⁶ estimated this number to be 1763. Yorifuji, et al.³⁵ estimated the number of deaths attributable to PM₁₀ over 20 µg/m³ at 5840 in Seoul. These numbers are substantially different, even when considering the differences in study area, study period, and pollutants investigated.

DISCUSSION

Beginning in 1999, many studies have been conducted to elucidate the health effects of air pollution in Korea. These studies have reported associations with mortality (all-cause, respiratory, cerebrovascular, cardiovascular, infant, injury, and suicide) and morbidity (allergic, respiratory, cardiovascular, cerebrovascular, adverse birth outcomes, depression, and cancer). Most studies examined the short-term effects of air pollution using a time series or case-crossover study design; we found only three cohort studies that examined long-term effects. There were four studies that estimated the health impacts of air pollution, and except for one study that reported DALYs, three studies had inconsistent estimations of the attributable number of deaths.

Estimating health impacts is usually conducted later than other research as previously estimated associations between exposure and outcome, or concentration-response function (C-R function) are required.³⁹ Naturally, the estimated health impact depends on the C-R function used. We suspect that differences in the attributable number of deaths estimated in the three studies reviewed here is partly due to the different C-R functions applied by the authors. Specifically, Yorifuji, et al.³⁵ and Leem, et al.³⁷ used C-R functions for mortality derived from epidemiological studies conducted in the United States (U.S.), whereas Han, et al.³⁶ used an IER function developed for the Global Burden of Disease 2010 and 2013. The C-R function derived from U.S. studies only accounted for a relatively low level of PM; thus, it may be inadequate for estimation of health impacts in Korea where exposure to higher concentrations of PM is observed. The IER function was developed by integrating various C-R functions of other exposures, such as tobacco smoke and burning of indoor solid fuel, to fill the gap in exposure range.³⁶ However, it remains uncertain whether the C-R function is comparable to the higher exposure range observed in Korea. Considering this, it is important to produce C-R functions using Korean data to accurately estimate the health impacts of exposure to air pollution.

As mentioned above, the effect of air pollution exposure can be divided into short-term and long-term effects. Typically, short-term effects are examined using time series and case-crossover studies, and long-term effects are investigated in cohort studies. The most recent time series study in Korea reported a 0.51% increase in mortality for each 10-µg/m³ increase in PM₁₀.¹⁶ This is comparable to the results of a recent meta-analysis of studies from East Asian cities, including Seoul and Incheon, which reported a 0.47% increase in total mortality for the same amount of increase in PM₁₀.⁴⁰ Similarly, although we could not find health impact assessment studies regarding air pollutants other than PM, we believe that previous epidemiological studies can provide relatively robust C-R functions for NO₂ and SO₂ to estimate health impacts.

Previous studies have reported inconsistent associations between O₃ exposure and mortality. Some published studies have reported a negative association, and the cause of this negative association has been an intriguing subject for additional analysis. One hypothesis is that the C-R function between O₃ concentration and mortality is not linear.⁴¹ Time series analyses conducted in Korea and Japan support this hypothesis in short-term associations.⁴²,⁴³ However, such non-linearity has not been observed in other studies,⁴⁴,⁴⁵ and the shape of the C-R function between O₃ concentration and acute mortality is still controversial. Nevertheless, studies analyzing the C-R function for long-term exposure of O₃ and mortality consistently report no evidence of a threshold.⁴⁶,⁴⁷ However, these studies may not have accounted for lower concentrations of O₃; this may be the reason for not observing a non-linear association, as the reported threshold of non-linear associations tends to be at lower concentrations. The negative association reported in a cohort study conducted by Kim, et al.¹⁸ may suggest the existence of a non-linear C-R function between long-term exposure to O₃ and mortality because Korea has lower concentrations of O₃ than the U.S.;⁴³ however, no analysis has been conducted using Korean data, as far as we know.

Among the two cohort studies on air pollution and mortality, one study examined the long-term health effects of PM_2.5 exposure. Although it is a valuable addition to the current knowledge, the results of that study seem inconsistent with previous reports. For instance, Kim, et al.¹⁸ reported an HR of 1.32 for all-cause mortality for a 1-µg/m³ increment of PM_2.5 in a cohort constructed using the National Health Insurance Service database, and a recent U.S. study analyzing a cohort constructed from a Medicare database reported an HR of 1.073 for a 10-µg/m³ increment of PM_2.5.⁴⁷ Kim, et al.¹⁸ suggested possible differences in the effect and composition of PM_2.5, genetic characteristics, and range of exposure between these studies, although we find a more than 30-fold greater HR difficult to explain. The largest difference between these two studies was in exposure assessment. Kim, et al.¹⁸ linked the concentration measured at a fixed monitoring station to the addresses of participants, whereas Di, et al.⁴⁷ used a model-based estimation of individual exposure. Another cohort study examined the long-term effect of PM₁₀ exposure.¹⁷ Those authors reported similar effects for PM₁₀ exposure, although the association was not statistically significant. However, this latter study applied an exposure assessment strategy, which could alleviate the effect of misclassification caused by participant mobility and exposure measurement at fixed monitoring stations.

Conventionally, air pollution studies use concentrations measured at fixed monitoring stations for exposure, which is an advantage for providing a large amount of data for a wide range of pollutants. However, data linked to study participants' addresses may not reflect individual exposure, especially when the mobility pattern of individuals is not accounted for.⁴⁸ This limitation may lead to misclassification, which may have substantial implications for the interpretation of results.⁴⁹ In recent years, advanced sensor and modeling technologies have facilitated individual exposure measurement in air pollution studies with the use of personal sensors and various exposure models based on dispersion models, geographical information, and satellite images.⁴⁸,⁵⁰,⁵¹ Estimation of exposure using these methods in Korea has been reported recently,⁵² and these individual exposure estimation methods should be applied in future studies to reduce uncertainty.

In addition to observational studies, there have been many intervention studies on air pollution and its health effects. Recent intervention studies have explored the benefits of exposure reduction using devices, such as an air purifier⁵³,⁵⁴ and facemasks,⁵⁵ in randomized controlled trials. The strength of intervention studies is two-fold: First, intervention studies may provide more robust evidence regarding the health effects of exposure to air pollution. Second, these trials may provide evidence regarding the effectiveness of personal measures that can be used to reduce the effects of air pollution. However, due to ethical and practical limitations, randomized controlled trials can only be applied to evaluate acute effects of exposure to air pollution. For instance, it may be unfeasible and unethical to design a study in which a portion of study participants are asked to wear facemasks for a long period (e.g., years). Causal modeling is a method that has been proposed to mitigate the shortcomings of observational studies without the need to conduct a randomized trial. This approach includes marginal structure modeling, instrumental variable analysis, and negative exposure control.⁵⁶ The causal modeling approach provides associations that are free of confounding under certain assumptions, which can be interpreted as causal, similar to the results of a trial. To date, there had been reports on the causal associations of PM_2.5, black carbon, and NO₂ in various circumstances.⁵⁷,⁵⁸ Such experimental studies are necessary so as to correctly assess the effects of air pollution on health and to facilitate more effective interventions through which to reduce exposure and to mitigate the health effects of air pollution.

Finally, despite our best efforts to comprehensively summarize the study results regarding the health effects of air pollution exposure in Korea, it is possible that we did not compile a complete list of all relevant research, which should be considered a limitation of the present review.

CONCLUSION

In the present review, we presented epidemiological studies conducted in Korea examining the health effects of exposure to air pollution. For the past 2 decades, there has been a considerable accumulation of knowledge regarding air pollution and health in Korea. However, the present review highlights that additional studies, especially cohort and experimental studies, are needed to provide more robust and accurate evidence that can be used to promote evidence-based policymaking.