Association of NO₂ and Other Air Pollution Exposures With the Risk of Parkinson Disease

Key Points

Question  Which air pollutant is associated with the development of Parkinson disease (PD)?

Findings  In this cohort study including a nationally representative cohort from a metropolitan city in South Korea (n = 78 830), a statistically significant association was found between exposure to NO₂, especially at high levels, and incidence of PD.

Meaning  These findings suggest that regulation of air pollutants might reduce the incidence of PD.

Importance  The development of Parkinson disease (PD) may be promoted by exposure to air pollution.

Objective  To investigate the potential association between exposure to particulate matters (PM_2.5 and PM₁₀), nitrogen dioxide (NO₂), ozone (O₃), sulfur dioxide (SO₂), and carbon monoxide (CO) and the risk of incident PD.

Design, Setting, and Participants  This retrospective cohort study used data from the Korean National Health Insurance Service. Among the 1 021 208 Korean individuals in the database, those who had lived in Seoul from January 2002 to December 2006 (n = 176 875) were screened for eligibility. A total of 78 830 adults older than 40 years without PD and who lived in Seoul between January 2002 and December 2006 were included in this study. Individuals diagnosed with PD before 2006 (n = 159) and individuals 40 years or younger (n = 97 886) were excluded. Each participant was followed up with annually from January 2007 to December 2015, thereby adding up to 757 704 total person-years of follow-up. Data were analyzed from January to September 2020.

Exposures  Individual exposure levels to PM_2.5, PM₁₀, NO₂, O₃, SO₂, and CO were estimated based on the participants’ residential address at the district level. To evaluate long-term exposure to air pollution, time-varying 5-year mean air pollutant exposure was calculated for each participant.

Main Outcomes and Measures  The outcome measure was the association between air pollution and the risk of incident PD measured as hazard ratios after adjusting for demographic factors, socioeconomic factors, and medical comorbidities.

Results  At baseline, the mean (SD) age of the 78 830 participants was 54.4 (10.7) years, and 41 070 (52.1%) were female. A total of 338 individuals with newly diagnosed PD were identified during the study period. Exposure to NO₂ was associated with an increase in risk of PD (hazard ratio for highest vs lowest quartile, 1.41; 95% CI, 1.02-1.95; P for trend = .045). No statistically significant associations between exposure to PM_2.5, PM₁₀, O₃, SO₂, or CO and PD incidence were found.

Conclusions and Relevance  In this large cohort study, a statistically significant association between NO₂ exposure and PD risk was identified. This finding suggests the role of air pollutants in PD development, advocating for the need to implement a targeted public health policy.

Air pollution is a significant public health hazard, and more than 80% of urban area residents are exposed to levels of air pollutants that exceed the limits set by the World Health Organization.¹ Recently, long-term exposure to air pollution has been identified to be associated with neurodegenerative diseases through systemic inflammation, oxidative stress, and direct invasion into the brain.^2-4

Parkinson disease (PD) is the second most prevalent neurodegenerative disease next to Alzheimer disease and affects more than 6 million people worldwide.⁵ The Braak stage of PD proposed that the pathologic aggregation of α-synuclein begins in the olfactory bulb and gut before spreading through the central nervous system.⁶ Therefore, exposures to environmental pollutants, such as pesticides, metals, and the microbiome as well as air pollution, have been suggested as risk factors for PD. In an animal study, exposure to ambient particles resulted in neuropathological damages in the dopaminergic neurons, which is a hallmark of PD.⁷ In an autopsy study, healthy individuals exposed to high levels of air pollutants showed ultrafine particles in the olfactory bulb and some α-synuclein and Aβ42 in neurons and glial cells.^8,9

However, the results of epidemiologic studies were inconsistent, and meta-analysis showed a statistically significant association only between PD and ozone (O₃), while particulate matters (PM_2.5) and nitrogen dioxide (NO₂) suggested some associations.¹⁰ Since most studies were conducted in North American and European countries, there is a need for investigation in Asian countries, where air pollutant levels are generally higher than those in other countries. Indeed, the mortality risk of stroke related to air pollution in Asia was 2-fold to 9-fold higher than those in North America and Europe.¹¹ In this study, we evaluated the associations between the incidence of PD and 6 types of ambient air pollutants (PM_2.5, PM₁₀, NO₂, O₃, sulfur dioxide [SO₂], and carbon monoxide [CO]) in a nationally representative cohort of South Korea.

We used the nationwide population-based cohort data from the National Health Insurance Service (NHIS) of South Korea. Approximately 97% of Korean residents are currently covered by health insurance based on employment or residential area and the rest are covered by the medical aid program (Medicaid). NHIS built the NHIS–National Sample Cohort version 2.0 (NHIS-NSC) by selecting a total of 1 million Korean individuals using a systematic stratified sampling method, maintaining approximately 2% of the Korean population that was covered by the NHIS, with 2006 as the reference year.¹² Information on patient demographic characteristics, medical treatment record, and detailed diagnoses coded with the Korean Standard Classification of Disease Version 5 (modification of the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD-10]) were collected from all individuals for 14 years from January 2002 to December 2015. As the reference year of sampling was 2006, information from January 2002 to December 2006 was collected retrospectively, while information from January 2007 to December 2015 was collected prospectively. This study was approved by the Institutional Review Board of Asan Medical Center, and all methods were performed in accordance with the relevant guidelines and regulations. As the NHIS provided the data after encryption to protect private information, the need for informed consent was waived. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Among 1 021 208 Korean individuals in the NHIS-NSC, we selected those who had lived in Seoul from January 2002 to December 2006, since the reference year was 2006 (n = 176 875) (Figure 1). We chose Seoul because it is the largest metropolitan city in South Korea, has a high population density, and retains the measurement data for each air pollutant in all its districts. We excluded (1) individuals diagnosed with PD before 2006 (n = 159) and (2) individuals 40 years or younger (n = 97 886) to exclude young-onset PD.¹³ As a result, a total of 78 830 adults older than 40 years without PD and who lived in Seoul between January 2002 and December 2006 were included in this study.

The Seoul Research Institute of Public Health and Environment provides hourly monitored data on PM_2.5, PM₁₀, NO₂, O₃, SO₂, and CO from 25 monitoring sites (1 per district). The eMethods in the Supplement provide measurement methods.^14,15 By linking the annually updated individual address with the annual mean pollution level from the monitoring sites, we estimated the annual air pollutants for all participants. This enabled us to reflect mobility within Seoul. For long-term air pollution exposure, we calculated the average of the annual air pollutants during the previous 5 years from the index date (time-varying 5-year mean air pollutant exposure), as used in previous studies.^16-19 In cases of missing data, air pollution data from the adjacent district were used.

National Health Insurance (NHI) uses a differential copayment system for rare intractable diseases, including PD. To register patients with PD for the rare intractable disease registration program, diagnosis based on NHI criteria by a neurologist is necessary. The NHI criteria for PD is similar to the UK Parkinson’s Disease Society Brain Bank clinical diagnostic criteria (eMethods in the Supplement).²⁰ After registration, the PD registration code (V124) is given to PD-related claims for reduction of the copayment.²¹ As a result of this registration system, all patients with a PD registration code are assured of being diagnosed by a neurologist with the definite criteria. Hence, we defined newly diagnosed cases of PD as individuals who were newly diagnosed with a primary or subsidiary diagnosis with an ICD-10 code for PD (G20) together with the rare intractable disease registration code for PD (V124), as used in previous studies.^21-23 If there were several claims with PD codes (G20), the first instance it occurred was considered the time of PD diagnosis. Each participant was followed up with annually from January 2007 to the index date (first instance of PD, death, relocation from Seoul, or December 31, 2015, whichever came first). Individuals were censored at the date of death or relocation from Seoul.

We used individual-level covariates recorded in the database, including age, sex, type of health insurance, and preexisting comorbidities. The health insurance type was composed of 2 parts: health insurance based on employment or residential area (97% of the whole population) and lower-income Medicaid group (3% of the whole population).²⁴ The diagnosis of medical comorbidities, including hypertension, diabetes, dyslipidemia, chronic kidney disease, congestive heart failure, ischemic heart disease, and traumatic brain injury, was based on ICD-10 codes (eMethods in the Supplement)

In a subgroup of the study population who underwent a biennial national health screening program, we were able to obtain data on body mass index, physical activity (0, 1 to 2, 3 to 4, or 5 or more times per week), smoking habits (never, former, or current), and drinking habits (none, mild to moderate, or heavy).²⁵ Physical activity was defined as the weekly frequency of strenuous exercise for at least 20 minutes.

The time-dependent Cox proportional hazards regression model was used to estimate the hazard ratios (HRs) for PD associated with the 6 types of air pollutants (PM_2.5, PM₁₀, NO₂, O₃, SO₂, and CO). The time-varying 5-year mean air pollutant exposure in each participant was used as the main time-related variable of interest and was expressed as interquartile range. For tests of trend, we assigned the order of quartile as a continuous variable and fitted them in a Cox regression model.^26-28 Deviations from linearity of continuous variables were assessed using restricted cubic regression splines. The time-dependent Cox proportional hazards regression model (unadjusted: model 1) was adjusted for age, sex, insurance type (model 2); model 2 factors plus comorbidities (model 3); and model 3 factors plus body mass index, smoking, alcohol, and physical activities (model 4), which could affect PD risk.^29-33 We applied a multiple-imputation procedure using the fully conditional specification method to impute the missing values in health examination data for model 4. The 5 imputed data sets were created with 20 burn-in iterations that were analyzed by the same analytical procedures, and the results were combined to obtain an overall estimate.

P values were calculated using t tests and χ² tests, as appropriate. Significance was set at a P value less than .05, and all P values were 2-tailed. All data analyses were performed using the SAS Enterprise Guide software version 7.1 (SAS Institute). Restricted cubic splines in the Cox regression models were presented using the rms and spline packages in R software version 3.3.3 (The R Foundation).

We additionally considered 1-year and 2-year lags in exposure in the Cox proportional hazards regression model 3. For 1-year lags in exposure, we used the time-varying 4-year mean air pollutant exposure for each air pollutant; for 2-year lags in exposure, we used the time-varying 3-year mean air pollutant exposure.

At baseline, the mean (SD) age of the 78 830 participants was 54.4 (10.7) years, and 41 070 (52.1%) were female (Table 1). During the maximum 9 years of follow-up (757 704.2 person-years; median, 8.6 years), a total of 338 participants (0.4%) developed incident PD, and the incidence rate of PD was 44.6 cases per 100 000 person-years. Compared with the entire study population, those who developed incident PD were significantly older (mean [SD] age, 54.4 [10.7] years vs 66.5 [9.7] years; P < .001) and were more likely to be receiving Medicaid (2.8% [2208 of 78 830] vs 7.1% [24 of 338]; P < .001). Among 338 participants with incident PD, the most common preexisting comorbidities were hypertension (158 [46.7%]), diabetes (102 [30.2%]), dyslipidemia (93 [27.5%]), and ischemic heart disease (51 [15.1%]) (Table 1). Among the entire study population, approximately 40% underwent a biennial national health screening program (eTable 1 in the Supplement).

In the 5 years preceding the index date, the median (range) level of PM_2.5 exposure was 26.5 (18.0-44.4) μg/m³; PM₁₀, 55.5 (41.0-79.0) μg/m³; NO₂, 0.033 (0.026-0.045) ppm; O₃, 0.019 (0.013-0.025) ppm; SO₂, 0.0053 (0.0036-0.0074) ppm; and CO, 0.59 (0.40-0.82) ppm. The annual mean concentrations of PM_2.5 and PM₁₀ were the highest in 2002, gradually decreased until 2009, and plateaued. The annual mean concentration of O₃ continuously showed an increasing trend during the study period, while those of NO₂, SO₂, and CO remained relatively stable (eFigure 1 in the Supplement).

Exposure to NO₂ was associated with an increase in risk of PD (HR for highest vs lowest quartile, 1.41; 95% CI, 1.02-1.95; P for trend = .045). The association was robust even after adjusting for age, sex, insurance type (model 2); model 2 factors and comorbidities (model 3); and model 3 factors and lifestyle (model 4). In contrast, we did not find statistically significant associations between exposure to other air pollutants (PM_2.5, PM₁₀, O₃, SO₂, and CO) and the risk of incident PD in both the unadjusted and adjusted analyses (Table 2).

In sensitivity analyses for NO₂, we found a positive association between high exposure to NO₂ and incident PD in the 1-year lag analysis (HR for highest vs lowest quartile, 1.34; 95% CI, 0.97-1.84; P for trend = .048) but not in the 2-year lag analysis (HR for highest vs lowest quartile, 1.17; 95% CI, 0.86-1.58; P for trend = .16). In both the 1-year and 2-year lag analyses, there was no evidence of a statistically significant association between the incidence of PD and exposure to PM_2.5, PM₁₀, O₃, SO₂, or CO (Table 3).

In multivariable Cox regression models with restricted cubic splines, the HRs for incident PD according to NO₂ exposure showed a statistically significant increase when the NO₂ levels were in the fourth quartile (greater than 0.038 ppm) (Figure 2C). No statistically significant association was found in restricted cubic splines between the PD incidence and exposure to other pollutants (Figure 2).

By using the nationally representative cohort, we examined the association between time-varying 5-year mean air pollutant exposure and incident PD in Seoul, a large-sized metropolitan city in South Korea. We found a statistically significant association between the incidence of PD and exposure to NO₂, especially at high exposure levels (greater than 0.038 ppm). The association between PD and NO₂ was robust even after adjusting for covariates as well as in sensitivity analysis using 1-year lags. Other air pollutants (PM_2.5, PM₁₀, O₃, SO₂, and CO) did not show statistically significant associations with the incidence of PD.

In this study, we found an association between the risk of incident PD and NO₂ concentration, and the association was evident when the exposure level of NO₂ was in the highest quartile (greater than 0.038 ppm). Seven studies evaluated the risk of PD according to NO₂ concentration (eTable 2 in the Supplement); of those, 3 showed positive correlations,^17,34,35 1 showed a negative correlation,³⁶ and 3 showed no correlations.^37-39 In the studies that did not show significant correlations between NO₂ and the risk of PD,^37-39 the maximum NO₂ concentrations were less than 0.038 ppm, which was the lower limit of the fourth quartile in this study. On the other hand, 2 of 3 studies that showed positive correlations, which were conducted in Taiwan³⁴ and Canada,¹⁷ reported wide ranges of NO₂ concentrations. Therefore, the heterogeneity between studies was likely due to the lack of contrast in NO₂ exposure, as well as other methodologic differences.

The association of NO₂ exposure with public health is often attributed to the exposure to traffic-related emissions, such as ultrafine particles.³⁹ However, some studies have suggested that NO₂ may directly exert toxic effects on the brain. NO₂ inhalation was shown to aggravate the accumulation of amyloid-β42 and decline in cognition through prostaglandin E₂ metabolism,⁴⁰ cause synaptic dysfunction coupled with tauopathy,⁴¹ and increase the levels of proinflammatory markers in the brain.⁴² Moreover, oxidative stress and systemic inflammation after NO₂ inhalation have been reported.⁴³ The direct toxic effect of NO₂ is in line with the results of a number of epidemiologic studies showing correlations between NO₂ and neurological disorders, such as PD, cognitive decline, stroke, and amyotrophic lateral sclerosis.^44-46

In animal studies, concentrated ambient PM induced central nervous system inflammation and neurodegeneration.^2,4,47 Despite evidence gathered from preclinical studies, we did not find significant associations between exposure to PM_2.5 or PM₁₀ and the risk of incident PD. To our knowledge, to date, 3 studies reported positive associations between PM_2.5 and the risk of PD,^16,19,48 while 5 studies showed no associations^36-38,49,50 (eTable 2 in the Supplement). For PM₁₀, all 6 studies failed to show significant associations with PD,^{34,36-38,49,50} while 1 study showed a significant correlation between PM₁₀ and PD in a subgroup analysis of women.³⁷ The studies with high exposure level of PM_2.5 or a wide range of exposure contrast, including ours, did not show an association between PM_2.5 and the risk of PD. Because PM is the mixture of organic and inorganic particles, the composition of PM might be more important than total amounts. For example, airborne metals, such as manganese, copper, and mercury, are strong risk factors for PD.^51,52

In this study, exposure to NO₂ in the preceding 5 years but not exposure to PM was associated with the development of PD. Unlike PM, which showed high variation in the past, the level of NO₂ remained stable in Seoul since 1980, which suggests that the 5-year exposure to NO₂ might be similar to the prior exposure (eFigure 2 in the Supplement). On the other hand, even the 5-year exposure to NO₂ could have attributed to the development of PD. Aggregation of pathogenic α-synuclein in the gut and the olfactory system are thought to take place decades before clinical symptoms appear.⁶ However, the aggregation of α-synuclein in the gut and olfactory bulb have also been shown in healthy older people.^53-55 A 2019 study suggested that what differentiates those who develop PD from those who do not might be the presence of facilitators that amplify and spread the α-synuclein aggregates along the vagus nerve and into the brainstem.⁵⁶ Inflammation and lysosomal dysfunction have been suggested to contribute to the spread of α-synuclein pathology.^57-60 Hence, we speculate that the inflammation provoked by NO₂ acts to amplify and propagate α-synuclein.

In the sensitivity analysis using the 1-year lags in exposure, the association between NO₂ and PD was significant; in contrast, the association was not evident in the analysis using 2-year lags in exposure. This might be explained by the short exposure duration in the 2-year lag analysis in this study. A large population-based cohort study also reported the attenuation of such positive association when 5-year and 10-year lags in exposure were used.¹⁷ There is a lack of evidence regarding the duration or lags in exposure to air pollutants required for developing PD, and further study is required to evaluate the optimal time lags between air pollutant exposure and the development of PD.

The strength of this study is that we used data from a large and well-characterized nationally representative cohort, with up to 9 years of follow-up duration. Using this cohort, we were able to incorporate the annually updated information on demographic characteristics, residence, lifestyle factors, and medical history in all individuals. Second, we comprehensively evaluated 6 different types of air pollutants using the annually updated air pollutant concentrations that allowed the evaluation of an association between various air pollutants and the development of PD. Lastly, this study was one of the few studies that evaluated the association between air pollutants and PD risk in Asian countries, where the high levels of air pollution are a major health threat.

Our study has several limitations. First, ambient air pollution exposure was only measured outdoors and may not reflect their indoor concentrations. Nevertheless, most epidemiologic studies also used outdoor measurements of air pollutants to obtain large-scale data. Second, as the reference exposure was based on district-level measurements, the relatively coarse spatial scale of exposure assessment and the narrow exposure range might have affected the result. The measurement error in exposure estimates tends to cause bias toward a null,⁶¹ which underestimates the association. However, the results are not likely to be exaggerated. Third, we did not account for occupational exposure to air pollutions, such as heavy metals, pesticides, and trichlorethylene. Considering that we used a population in a metropolitan city, where the composition of agricultural and forestry workers was 0.28%, exposure to pesticide would be low.⁶² Other occupational risk factors for PD should be considered in future studies. Fourth, we could not investigate the long lag period in exposure. To account for the long prodromal stage of PD, the effect of air pollutant according to various exposure duration and lag analysis should be investigated.

In conclusion, we identified a statistically significant association between the risk of PD and exposure to NO₂ for the previous 5 years, especially at high exposure levels. We found no evidence for the association between the risk of PD and exposure to PM_2.5, PM₁₀, O₃, SO₂, or CO.

Accepted for Publication: March 5, 2021.

Published Online: May 17, 2021. doi:10.1001/jamaneurol.2021.1335

Corresponding Author: Sun Ju Chung, MD, PhD, Department of Neurology, Asan Medical Center, University of Ulsan College of Medicine, 88, Olympic-ro 43-gil, Songpa-gu, Seoul 05505, South Korea (sjchung@amc.seoul.kr).

Author Contributions: Dr Chung had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Jo, Y. Kim, B. Kim, Chung.

Acquisition, analysis, or interpretation of data: Jo, Y. Kim, Park, Hwang, Lee, Chung.

Drafting of the manuscript: Jo, Chung.

Critical revision of the manuscript for important intellectual content: Y. Kim, Park, Hwang, Lee, B. Kim, Chung.

Statistical analysis: Jo, Y. Kim, Park, Hwang.

Obtained funding: Chung.

Administrative, technical, or material support: Jo.

Study supervision: B. Kim, Chung.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by the Korea Healthcare Technology R&D Project, Ministry of Health & Welfare, Republic of Korea (grant HI19C0256).

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.