Association between air pollution exposure and diabetic retinopathy among diabetics

Environmental Research xxx (xxxx) xxxx

Contents lists available at ScienceDirect

Environmental Research journal homepage: www.elsevier.com/locate/envres

Association between air pollution exposure and diabetic retinopathy among diabetics Shih-Chun Pana, Ching-Chun Huangb, Wei-Shan Chinc, Bing-Yu Chend, Chang-Chuan Chana, Yue Leon Guoa,b,e,∗ a

Institute of Environmental and Occupational Health Sciences, National Taiwan University, Taipei, Taiwan Environmental and Occupational Medicine, College of Medicine, National Taiwan University (NTU) and NTU Hospital, Taipei, Taiwan c School of Nursing, College of Nursing, Taipei Medical University, Taipei, Taiwan d Department of Medical Research and Development, Chang Gung Memorial Hospital, Keelung, Taiwan e National Institute of Environmental Health Sciences, National Health Research Institute, Miaoli, Taiwan b

ARTICLE INFO

ABSTRACT

Keywords: Diabetes mellitus Particulate matter Diabetic retinopathy Air pollution Ocular epidemiology

Background: Exposure to air pollution has been linked to adverse effects on vascular diseases. However, the effects of air pollution exposure on diabetic retinopathy (DR), a vascular disease, have not been studied. Objective: To determine the association of ambient air pollution exposure with DR risk. Methods: Patients newly diagnosed as having diabetes mellitus (DM) during 2003–2012 from Longitudinal Health Insurance Database 2005), a subset of National Health Insurance Research Database, were included as the study cohort. Newly diagnosed DR patients one year or later after DM diagnosis were identified as cases. Kriging was used to interpolate yearly concentrations of air pollutants at township levels and linked with every individual's residence in each year; average concentrations during the follow-up period were then calculated as personal exposure. Conditional logistic regressions with adjustments for age at DM diagnosis and comorbidities were applied. Results: Of newly diagnosed DM cases during 2003–2012, 579 were newly diagnosed as having DR over a mean follow-up period of 5.6 years. The Odds ratio (95% confidence interval) of DR occurrence for every 10-μg/m3 increase in particulate matter with ≤2.5 and 2.5–10-μm diameter was 1.29 (1.11–1.50) and 1.37 (1.17–1.61), respectively. Conclusion: In patients with DM, the higher particulate matter exposure, the higher is the DR risk.

1. Introduction

and macro-vascular disease, such as cardiovascular, renal, and retinal diseases. Diabetic retinopathy (DR) is a retinal vascular disease caused by abnormal blood flow in the retina. DR is a common DM-related complication and a leading cause of blindness among patients with DM. High blood sugar levels can obstruct the microvasculature that nourishes the retina and thus sever the blood supply to the eyes. This can result in the growth of newer underdeveloped, leaky blood vessels in eyes (Antonetti et al., 2012). As indicated by the Diabetes Prevention Program (DPP), DR is relatively common; DR is observed in 7.9% of patients with impaired glucose tolerance who did not progress to DM, whereas in 12.6% of the patients with impaired glucose tolerance who later progressed to DM (Diabetes Prevention Program Research, 2007).

Diabetes mellitus (DM) is no longer a disease predominant in developed countries alone; its prevalence is considerably increasing globally—particularly in low- or middle-income countries. According to the World Health Organization, DM prevalence in adults aged > 18 years increased from 4.7% in 1980 to 8.5% in 2014. In low- or middleincome countries, DM prevalence has risen faster than that in highincome countries (World Health Organization, 2016). In Taiwan, DM prevalence increased from 7.2% in 1990 to 10.6% in 2017 (Institute for Health Metrics and Evaluation, 2015), making it the fifth leading cause of death (Ministry of Health and Welfare, 2018). Medical expenditure associated with DM and its complications are high, including micro-

Corresponding author. Institute of Environmental and Occupational Health Sciences, National Taiwan University, Room 339, No. 17 Xu-Zhou Rd., Taipei, 100, Taiwan. E-mail addresses: [email protected] (S.-C. Pan), [email protected] (C.-C. Huang), [email protected] (W.-S. Chin), [email protected] (B.-Y. Chen), [email protected] (C.-C. Chan), [email protected] (Y.L. Guo). ∗

https://doi.org/10.1016/j.envres.2019.108960 Received 6 August 2019; Received in revised form 20 November 2019; Accepted 22 November 2019 0013-9351/ © 2019 Elsevier Inc. All rights reserved.

Please cite this article as: Shih-Chun Pan, et al., Environmental Research, https://doi.org/10.1016/j.envres.2019.108960

Environmental Research xxx (xxxx) xxxx

S.-C. Pan, et al.

Exposure to particulate matter (PM) and gaseous air pollutants has been linked to macrovascular diseases (e.g., cerebrovascular and cardiovascular diseases) (Andersen et al., 2010; Brook et al., 2010; Collart et al., 2018; Dominici et al., 2006; Gurung et al., 2017; Milojevic et al., 2014; O'Donnell M et al., 2011; Phung et al., 2016; Shah et al., 2015; Wellenius et al., 2005), progression in coronary artery calcification, and atherosclerosis acceleration (Kaufman et al., 2016). In some studies, patients with DM were particularly susceptible to cardiovascular damage by PM and other air pollutants (Hart et al., 2015; Pinault et al., 2018; Pope et al., 2015; Tibuakuu et al., 2018; Zanobetti and Schwartz, 2002), even myocardial infarction and stroke (Akbarzadeh et al., 2018; Ashkenazi et al., 1992). Among the general population as well as patients with DM, renal function deterioration and end-stage renal disease are related to exposure to ambient PM (Bowe et al., 2018; Chen et al., 2018; Chin et al., 2018; Kim et al., 2018). Etiologically, the aforementioned processes are potentially microvascular (Kahn et al., 2004; King et al., 2005). However, no study has investigated the effect of air pollution exposure on DR. The present study determined whether air pollution exposure accelerates DR development in patients with DM. A nationwide representative insurance data set was combined with Taiwan Environmental Protection Administration's air monitoring data to examine the association between air pollution and DR risk in patients with DM.

2.4. Exposure assessment The 2003–2013 air pollution data were extracted from 77 fixed-site air monitoring stations in Taiwan. At each station, the hourly concentrations of PM with ≤2.5-μm diameter (PM2.5; measured since 2006) and 2.5–10-μm diameter (PM2.5–10), carbon monoxide (CO), nitrogen dioxide (NO2), sulfur dioxide (SO2), and ozone (O3) were automatically monitored; these concentrations were monitored using β-ray attenuation (in μg/m3), nondispersive infrared (in parts per million [ppm]), chemiluminescence (in parts per billion [ppb]), ultraviolet fluorescence (in ppb), and ultraviolet absorption (in ppb) methods, respectively. The yearly average for each pollutant concentration at each station was calculated on the basis of at least 6000 hourly measurements; fewer data points than 6000 hourly measurements would result in a missing value for that year. The calculated yearly average of each pollutant concentration at each monitoring station was used for further spatial interpolation. ArcGIS Desktop (version 10; ESRI Inc., Redlands, CA, USA) with the ordinary kriging method was used to interpolate exposure concentration to a regular grid (250 × 250 m) across Taiwan. The interpolated concentrations were then averaged into township levels as yearly township exposure levels of air pollutants. People were assumed to have sought medical care at clinics nearest to their residence. The corresponding townships where medical attention were most frequently sought were identified as individual's townships of residence in each year (Lin et al., 2011). Finally, yearly township exposure levels were linked with the township of residence for each year and the average concentrations during the follow-up period were calculated as personal exposure levels.

2. Material and methods 2.1. Database The National Health Insurance (NHI) program, implemented since 1995, covers approximately 99% of the Taiwan's population. The NHI Research Database (NHIRD) contains comprehensive medical claims records of the beneficiaries, including registry for beneficiaries, inpatient expenditures by admissions, ambulatory care expenditures by visits and details of ambulatory care orders. The detail diagnostic codes or therapeutic treatment at each medical visit can be obtained from these claims data. We used Longitudinal Health Insurance Database 2005 (LHID2005), which contains all the original claims data of randomly sampling 1,000,000 individuals from the NHI beneficiary registry of 2005. The study was approved by institutional review boards at National Taiwan University Hospital (Number: 201302039RINC).

2.5. Covariates We extracted each individual's chronic comorbidity status, including hypertension (ICD-9-CM: 401.XX), hyperlipidemia (ICD-9-CM: 272.XX), coronary artery disease (CAD; ICD-9-CM: 414.XX), heart failure (ICD-9-CM: 438.XX), arrhythmia (ICD-9-CM: 427.9), and glaucoma (ICD-9-CM: 365.XX). Patient characteristics, such as age at DM diagnosis and comorbidities, were used as covariates. Age at DM diagnosis was categorized into three groups, 20–44, 45–64, and ≥65. Whereas conditions of comorbidities were categorized into two groups (with and without).

2.2. Study population

2.6. Statistical analysis

This retrospective population-based nested case-control study was designed to examine the association between air pollution and DR risk in patients with DM. The International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) and A-codes were applied for classified diagnosis. DM was defined by three or more outpatient visits or one inpatient discharge with the relevant diagnostic codes (i.e., ICD-9-CM: 250.XX or A code: A181) within 1 year and oral hypoglycemic agents administered or insulin used after diagnosed. The recruited patients were aged > 20 years and were newly diagnosed as having DM cases on or after January 1, 2003.

Demographic characteristics and potential covariates were compared between DM patients with and without DR by using the chisquare test. The separate conditional logistic regressions were performed for examining the association between long-term air pollution exposure and DR occurrence. For air pollutants significantly associated with DR, other pollutants were added one at a time to form a twopollutant model, thus excluding the effects of the second pollutant. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated for each interquartile range (IQR) and 10-μg/m3 increment in PM2.5, PM2.5–10 and PM10 concentrations, 0.1-ppm increment in CO concentrations, and 1-ppb increment in NO2, SO2, and O3 concentrations. All analyses were performed using SAS (version 9.4; SAS Institute Inc., Cary, NC, USA).

2.3. Cases and controls DR, as case group, was defined by two or more outpatient visits with the main diagnosis (ICD-9-CM: 362.0 and 362.0X) by an ophthalmologist in 1 year. We excluded individuals with retinal detachment and defects (ICD-9-CM: 361.XX), other retinal disorders (ICD-9-CM: 362.XX), blindness or low vision (ICD-9-CM: 369.XX) before DM diagnosis, previous DR diagnosis, and follow-up period < 1 year. The control group was selected from diabetics without DR diagnosis in study population and was 10-fold frequency matched according to sex, index year of DM diagnosis and years of follow-up.

3. Results A total of 579 DR cases were identified, and the control group was consisted of 5790 patients. There were greater proportion of these DR cases being at age of 45–64 years old and without hyperlipidemia (Table 1). Townships were categorized according to level of urbanization, namely, highly urbanized, intermediately urbanized, emerging towns, and rural towns. We did not find significant rural-urban 2

Environmental Research xxx (xxxx) xxxx

S.-C. Pan, et al.

Table 1 Characteristics of diabetics during 2003–2012 in Taiwan. Non-Retinopathy

Retinopathy

n 5790

(%) 90.9

n 579

(%) 9.1

48.2 51.8

279 300

48.2 51.8

30.0 57.8 12.2

137 377 65

23.7 65.1 11.2

41.6 45.4 13.0

251 265 63

43.3 45.8 10.9

49.1

281

48.5

66.2

321

55.4

12.4

58

10.0

4.0

27

4.7

3.0

11

1.9

4.0

30

5.2

Sex Female 2790 Male 3000 Age 20-44 1736 45-64 3349 ≥65 705 Income, TWD < 19,200 2409 19,200–40,100 2629 ≥40,100 752 Hypertension With 2846 Hyperlipidemia With 3834 Coronary artery disease With 719 Heart failure With 232 Arrhythmia With 174 Glaucoma With 231

Table 4 Hazard ratios (95% confidence intervals) for diabetic retinopathy: single-pollutant models.

P-value

1.000 0.002

PM2.5 PM2.5-10 PM10 CO NO2 SO2 O3 TEMP

3

< 0.001 0.096 0.440 0.153 0.185

μg/m μg/m3 μg/m3 ppm ppb ppb ppb °C

SD

25th

Median

75th

IQR

33.4 25.8 59.4 0.5 16.6 3.8 28.8 23.7

6.0 5.5 10.6 0.1 3.5 0.9 1.6 0.7

28.9 21.7 51.3 0.4 14.4 3.3 28.0 23.2

34.3 24.9 59.2 0.5 16.1 3.7 29.0 23.7

38.0 29.4 67.3 0.5 18.3 4.2 29.8 24.1

9.1 7.7 15.9 0.1 3.9 0.9 1.8 0.9

PM2.5 PM2.5-10 PM10 CO NO2 SO2 O3

CO

NO2

SO2

O3

TEMP

0.721 1

0.934 0.912 1

−0.036 −0.219 −0.129 1

0.090 −0.045 0.024 0.839 1

0.361 0.528 0.456 0.166 0.497 1

0.224 0.328 0.278 −0.736 −0.725 −0.107 1

0.745 0.703 0.777 −0.107 −0.102 0.286 0.189

Per IQR

HRa

95% CI

PM2.5 PM2.5–10 PM10 CO NO2 SO2 O3

10 μg/m3 10 μg/m3 10 μg/m3 0.1 ppm 1 ppb 1 ppb 1 ppb

1.29 1.37 1.18 0.96 1.00 1.05 1.03

(1.11–1.50) (1.17–1.61) (1.08–1.28) (0.87–1.05) (0.97–1.02) (0.96–1.15) (0.97–1.09)

9.1 μg/m3 7.7 μg/m3 15.9 μg/m3 0.1 ppm 3.9 ppb 0.9 ppb 1.8 ppb

1.26 1.28 1.30 0.96 0.98 1.04 1.05

(1.10–1.45) (1.13–1.44) (1.14–1.49) (0.87–1.05) (0.89–1.09) (0.96–1.14) (0.95–1.16)

To our knowledge, this is the first study to note a positive association between particulate matter exposure and DR risk. Based on the IQRs of the concentrations of PM2.5 (28.9–38.0 μg/m3) and PM2.5–10 (21.7–29.4 μg/m3), patients with DM in the high exposure groups had 26% and 28% higher DR risk than did those in the low exposure groups, respectively. We did not find any significant association between traffic-related air pollutant (e.g. CO and NO2) exposure and DR. Air pollutant exposure can increase the risk of vascular conditions, including both macrovascular and microvascular conditions. Patients with DM have a similar (Ho et al., 2018; Kim et al., 2017) or even higher (Akbarzadeh et al., 2018; Hart et al., 2015; Pinault et al., 2018; Pope et al., 2015; Tibuakuu et al., 2018; Zanobetti and Schwartz, 2002) susceptibility to the cardiovascular effects of PM and other air pollutants. Furthermore, individuals with and without DM did not have differences in susceptibilities to renal damage due to PM and NOx exposure (Bragg-Gresham et al., 2018; Chan et al., 2018; Lin et al., 2018; Xu et al., 2016). DR, noted only in patients with DM, involves vascular

Table 3 The spearman correlation coefficients between air pollutants. PM10

95% CI

4. Discussion

Note: TEMP, ambient temperature; ppm, parts per million; ppb, parts per billion.

PM2.5-10

HRa

hyperlipidemia, each IQR increment in PM2.5, PM2.5–10, and PM10 concentrations was significantly associated with increasing DR risk (OR [95% CI] = 1.26 [1.10–1.45], 1.28 [1.13–1.44], and 1.30 [1.14–1.49], respectively); this significant association was also noted for each 10-μg/ m3 increment in PM2.5, PM2.5–10, and PM10 concentrations (OR [95% CI] = 1.29 [1.11–1.50], 1.37 [1.17–1.61], and 1.18 [1.08–1.28], respectively). In the two-pollutant model, each 10-μg/m3 increment in PM2.5 exposure remained robust when CO (1.17 [1.02–1.35]), NO2 (1.20 [1.04–1.39]), or SO2 (1.25 [1.07–1.46]) was added, respectively (Table 5). Also, the effects of PM2.5-10 was remaining while NO2 (1.20 [1.03–1.40]) or SO2 (1.31 [1.10–1.56]) was added, respectively. Sensitivity analysis was done to examine whether particulates’ effects on occurrence of DR were different by gender or age, and statistical interactions were not found.

0.794

Mean

Unit

Note: HR, hazard ratio; CI, confidence interval. HRa, single pollutant model adjusted for age at DM diagnosis, and hyperlipidemia.

0.328

Table 2 The distribution of air pollution exposure levels of diabetes mellitus patients. Units

Pollutants

Table 5 Hazard ratios (95% confidence intervals) for diabetic retinopathy: two-pollutant models.

Note: TEMP, ambient temperature. *by Spearman's rank correlation coefficient, all correlation coefficients are statistically significant at p < 0.05.

PM2.5

differences in DR prevalence in our study population. Thus, urbanization was not included into the final models. Table 2 presents the distribution of the air pollutant exposure level in participants during the follow-up period. Fine and coarse PM concentrations were strongly correlated (Table 3). CO concentrations were highly related to NO2 concentrations (r = 0.839), and both were negatively correlated with O3 concentrations. Table 4 demonstrates the effect of air pollutant exposure on DR development, determined using the conditional logistic regression. In single-pollutant models, after adjustment for age at DM diagnosis and

Pollutants Adj. PM2.5 Adj. PM2.5-10 Adj. CO Adj. NO2 Adj. SO2 Adj. O3

HRa – 1.10 1.17 1.20 1.25 1.16

PM2.5-10 95% CI (0.90–1.33) (1.02–1.35) (1.04–1.39) (1.07–1.46) (1.00–1.33)

HRa 1.13 – 1.16 1.20 1.31 1.17

95% CI (0.91–1.40) (0.99–1.36) (1.03–1.40) (1.10–1.56) (0.99–1.37)

Note: HR, hazard ratio; CI, confidence interval. HRa, two-pollutant models adjusted for age at DM diagnosis, and hyperlipidemia. Adj.: adjusted by adding following air pollutant. 3

Environmental Research xxx (xxxx) xxxx

S.-C. Pan, et al.

permeability increase, pericyte loss, hypoxia, and ischemia in the early stage, followed by hemorrhage and retinal neovascularization later. However, the Multi-Ethnic Study of Atherosclerosis (MESA) reported chronic and short-term PM2.5 exposure to be associated with a narrow retinal arteriolar diameter (Adar et al., 2010). In this study, patients without DM were not included; nevertheless, our study findings suggest that potential damages of air pollution to micro-vasculatures occur especially among susceptible individuals like diabetics, and such damages can involve mechanisms to induce retinopathy. It is possible that those with DM in lower SES would be more likely to be live in highly polluted areas and having higher risk of DR. We thus used the insurance premium, which was determined by personal monthly income, as an indicator of SES status. The insurance premium was categorized into three groups (< 19,200, 19,200–40,100, ≥40,100). We did not find significant difference of time to DR onset among income groups. Thus, insurance premium was not included into the final models. Due to the delayed diagnosis of DM, it is possible that the diagnosis is made when retinopathy is diagnosed. Up to 21% of patients with type 2 diabetes (T2DM) have retinopathy at the time of first diagnosis of diabetes (Fong et al., 2003). This issue was not included in this current investigation. Since the research interest was on whether air pollution's effects on DR occurrence among diabetics, this study only included those without DR when DM diagnosis was made. Our results showed that DR incidence was 3.4% within a median follow-up duration of 5.5 years. This figure was similar to a study conducted in UK, in which a 5-year cumulative DR incidence was 4% among patients with T2DM (Jones et al., 2012). In addition, the observed period shows a decrease in average yearly concentrations of air pollutants. To avoid bias caused by differential follow-up periods between those with and without DR, we matched DR cases and non-DR controls by index year of DM diagnosis and follow-up years. Hyperglycemia in retinal vasculature leads to excessive reactive oxygen species (ROS) production through the mitochondrial transport chain and thus causes oxidative stress (Kowluru and Chan, 2007). Oxidative stress and hyperglycemia contribute to inflammation through cytokines, adhesion molecules, and vascular endothelial growth factor (VEGF) signaling. VEGF has a pro-inflammatory role in the diabetic retina and promotes retinal neovascularization and damage (Yang et al., 2013). Formation and increasing permeability of new and weak vessels, breakdown of the blood-retinal barrier due to VEGF in turn induces hemorrhages in the retina (Shin et al., 2014; Spijkerman et al., 2007; van Hecke et al., 2005). VEGF has also been reported to be associated with leukocyte adhesion to endothelial cells through the induction of nitric oxide synthase and intracellular adhesion molecule-1 (ICAM-1) expression (Joussen et al., 2002). Endothelial dysfunction and increased levels of inflammatory mediators, such as ICAM-1 and tumor necrosis factor-alpha (TNF-α), contribute to leukocyte dysfunction and leukostasis (Larson and Springer, 1990; Takami et al., 1998). Leukostasis also plays a crucial role in DR through capillary occlusion and ROS-mediated cell death (Lutty et al., 1997; Schroder et al., 1991). The mechanism underlying the association of air pollution exposure with DR risk remains unclear. Particulate air pollutants induce oxidative stress, systemic inflammation, and elevated serum cytokines (e.g., TNF-α, VEGF, ICAM-1, and interleukin 6). Previous study has been reported that high concentrations of some metals such as nickel, copper, and arsenic in fine particles were related to higher levels of markers of inflammation (IL-6 and VEGF) (Niu et al., 2013). This might partially explain the contribution of PM to DR occurrence. These effects might add to the effects of oxidative stress and inflammation induced by DM. Endothelial dysfunction and atherosclerosis are pathological features of DR, and both have been observed in animals and humans exposed to particulate air pollutants (Jomova and Valko, 2011; Miller et al., 2012; Niu et al., 2013). This study has several strengths. First, it used data from a nationwide and representative longitudinal database, NHIRD, with a > 99% coverage rate of the total population. Follow-up to occurrence of event

or to censoring has been complete due to rather complete records on the date of withdrawing from insurance or death in each individual. Second, the diagnosis of DR and the related comorbidities was based on the aforementioned NHIRD medical records and thus was free of recall bias. However, some limitations of this study are as follows. First, only diagnosis and treatment codes were obtained from the longitudinal medical data set, LHID2005. Detailed information related to DM severity was lacking (e.g., body mass index and hemoglobin A1C levels). However, the recruitment of patients with DM was started after their diagnosis, and there is no reason to believe severity was different to start with in regions with higher or lower air pollutants. The progression was faster in patients with DM in highly polluted regions; this was potentially an effect of air pollution exposure. Second, delay in the diagnosis of T2DM has been well documented (Harris et al., 1992). It is possible that those used medical services more frequently, especially among older people, might have had a higher chance of being diagnosed as having DM. Due to the high availability and accessibility of healthcare in Taiwan, the diagnosis of T2DM should be more ready. Previous study using the definition of three or more outpatient visits or one inpatient within one year found accuracy of DM diagnosis of 74.6% (Lin et al., 2005). The study using two clinic visits, prescription, and/or hospitalization during a two-year period found reasonably high validity, i.e., sensitivity of 90.9%, specificity of 94.9%, positive predictive value of 92.0%, and negative predictive value of 94.2% (Sung et al., 2016). This current study combined both approaches to ensure better validity of DM diagnosis. Thus, the diabetics we recruited should be more accurate. Third, given the severity of DR could be different in DM patients at first visit for DR problem, time of DR diagnosis might not be the time of disease development (Looker et al., 2012; Porta et al., 2014). In NHI program of Taiwan, diabetic patients are recommended to receive examination by an ophthalmologist yearly after the diagnosis of DM. Therefore, our case definition of DR involved a diagnosis by an ophthalmologist. This might have reduced the difference between the date of DR diagnosis and the time of disease development. Nevertheless, the diagnosis of DR can also be delayed (Looker et al., 2012; Porta et al., 2014). However, we considered such delay in diagnosis likely to be non-differential, i.e., the degree of exposure misclassification was independent of the risk of air pollution exposure. Such nondifferential misclassification of outcome would tend to underestimate relationship between DR and air pollution exposure. Therefore, the observed relation between PM2.5 and the development of DR is likely true. In addition, exposure assessment was performed at township levels, estimated using ordinary kriging, and the analysis on the individual level was lacking. This might have caused exposure misclassification and reduced capability in finding relationships between air pollutants and DR. In other words, our findings are likely biased toward the null hypothesis. Thus, our finding that air pollution exposure increases DR risk is likely true, but the effects may have been underestimated. People actually may sometimes seek medical care nearer their places of work. However, only 49.2% of the population was working. Among them, 42.9% went to a different township for work, and only 13.8% of people go to a different county for work. Since the air pollution were not drastically different in neighboring townships, the misclassification of exposure was assumed to be small. Furthermore, misclassification of this kind might have driven the findings towards the null hypothesis. Therefore, we believe our findings of significant effects of air pollution was unbiased. In conclusion, in patients with DM, ambient PM exposure is associated with DR risk. Future studies for confirming these findings and identifying the underlying mechanisms are warranted. Financial disclosure The authors have no financial relationships relevant to this article to disclose. 4

Environmental Research xxx (xxxx) xxxx

S.-C. Pan, et al.

Declaration of competing interest

initiates early diabetic retinal leukocyte adhesion in vivo. Am. J. Pathol. 160, 501–509. Kahn, C.R., et al., 2004. Joslin's Diabetes Mellitus. Lippincott Williams and Wilkins. Kaufman, J.D., et al., 2016. Association between air pollution and coronary artery calcification within six metropolitan areas in the USA (the Multi-Ethnic Study of Atherosclerosis and Air Pollution): a longitudinal cohort study. Lancet 388, 696–704. Kim, H., et al., 2017. Cardiovascular effects of long-term exposure to air pollution: a population-based study with 900 845 person-years of follow-up. J. Am. Heart Assoc. 6. Kim, H.J., et al., 2018. Association between exposure to ambient air pollution and renal function in Korean adults. Ann. Occup. Environ. Med. 30, 14. King, K.D., et al., 2005. Microvascular and macrovascular complications of diabetes mellitus. Am. J. Pharmaceut. Educ. 69, 1–10. Kowluru, R.A., Chan, P.S., 2007. Oxidative stress and diabetic retinopathy. Exp. Diabetes Res. 2007, 43603. Larson, R.S., Springer, T.A., 1990. Structure and function of leukocyte integrins. Immunol. Rev. 114, 181–217. Lin, C.C., et al., 2005. Accuracy of diabetes diagnosis in health insurance claims data in Taiwan. J. Formos. Med. Assoc. 104, 157–163. Lin, M.H., et al., 2011. Using regional differences and demographic characteristics to evaluate the principles of estimation of the residence of the population in national health insurance research databases (NHIRD). Chin. J. Publ. Health 30, 347–360. Lin, S.Y., et al., 2018. Association of exposure to fine-particulate air pollution and Acidic gases with incidence of nephrotic syndrome. Int. J. Environ. Res. Public Health 15. Looker, H.C., et al., 2012. Diabetic retinopathy at diagnosis of type 2 diabetes in Scotland. Diabetologia 55, 2335–2342. Lutty, G.A., et al., 1997. Relationship of polymorphonuclear leukocytes to capillary dropout in the human diabetic choroid. Am. J. Pathol. 151, 707–714. Miller, M.R., et al., 2012. From particles to patients: oxidative stress and the cardiovascular effects of air pollution. Future Cardiol. 8, 577–602. Milojevic, A., et al., 2014. Short-term effects of air pollution on a range of cardiovascular events in England and Wales: case-crossover analysis of the MINAP database, hospital admissions and mortality. Heart 100, 1093–1098. Ministry of Health and Welfare, 2018. Taiwan Death Report 2017. Taiwan. Niu, J., et al., 2013. The role of metal components in the cardiovascular effects of PM2.5. PLoS One 8, e83782. O'Donnell, J.M., et al., 2011. Fine particulate air pollution (PM2.5) and the risk of acute ischemic stroke. Epidemiology 22, 422–431. Phung, D., et al., 2016. Air pollution and risk of respiratory and cardiovascular hospitalizations in the most populous city in Vietnam. Sci. Total Environ. 557–558, 322–330. Pinault, L., et al., 2018. Diabetes status and susceptibility to the effects of PM2.5 exposure on cardiovascular mortality in a national Canadian cohort. Epidemiology 29, 784–794. Pope 3rd, C.A., et al., 2015. Relationships between fine particulate air pollution, cardiometabolic disorders, and cardiovascular mortality. Circ. Res. 116, 108–115. Porta, M., et al., 2014. Estimating the delay between onset and diagnosis of type 2 diabetes from the time course of retinopathy prevalence. Diabetes Care 37, 1668–1674. Schroder, S., et al., 1991. Activated monocytes and granulocytes, capillary nonperfusion, and neovascularization in diabetic retinopathy. Am. J. Pathol. 139, 81–100. Shah, A.S., et al., 2015. Short term exposure to air pollution and stroke: systematic review and meta-analysis. BMJ 350, h1295. Shin, E.S., et al., 2014. Diabetes and retinal vascular dysfunction. J. Ophthalmic Vis. Res. 9, 362–373. Spijkerman, A.M., et al., 2007. Endothelial dysfunction and low-grade inflammation and the progression of retinopathy in Type 2 diabetes. Diabet. Med. 24, 969–976. Sung, S.F., et al., 2016. Validation of algorithms to identify stroke risk factors in patients with acute ischemic stroke, transient ischemic attack, or intracerebral hemorrhage in an administrative claims database. Int. J. Cardiol. 215, 277–282. Takami, S., et al., 1998. High concentration of glucose induces the expression of intercellular adhesion molecule-1 in human umbilical vein endothelial cells. Atherosclerosis 138, 35–41. Tibuakuu, M., et al., 2018. Air pollution and cardiovascular disease: a focus on vulnerable populations worldwide. Curr. Epidemiol. Rep. 5, 370–378. van Hecke, M.V., et al., 2005. Inflammation and endothelial dysfunction are associated with retinopathy: the Hoorn study. Diabetologia 48, 1300–1306. Wellenius, G.A., et al., 2005. Air pollution and hospital admissions for ischemic and hemorrhagic stroke among medicare beneficiaries. Stroke 36, 2549–2553. World Health Organization, 2016. Global Report on Diabetes. Xu, X., et al., 2016. Long-term exposure to air pollution and increased risk of membranous nephropathy in China. J. Am. Soc. Nephrol. 27, 3739–3746. Yang, W., et al., 2013. Attenuation of streptozotocin-induced diabetic retinopathy with low molecular weight fucoidan via inhibition of vascular endothelial growth factor. Exp. Eye Res. 115, 96–105. Zanobetti, A., Schwartz, J., 2002. Cardiovascular damage by airborne particles: are diabetics more susceptible? Epidemiology 13, 588–592.

The authors have no conflicts of interest relevant to this article to disclose. Acknowledgments This study was partially supported by the Ministry of Science and Technology grant MOST 108-2621-M-002-018 and MOST 107-2621-M002-007, and by the National Health Research Institute, Taiwan grant NHRI-EM-105-SP08. The views expressed herein are the authors' own. Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.envres.2019.108960. References Adar, S.D., et al., 2010. Air Pollution and the microvasculature: a cross-sectional assessment of in vivo retinal images in the population-based multi-ethnic study of atherosclerosis (MESA). PLoS Med. 7, e1000372. Akbarzadeh, M.A., et al., 2018. The association between exposure to air pollutants including PM10, PM2.5, ozone, carbon monoxide, sulfur dioxide, and nitrogen dioxide concentration and the relative risk of developing STEMI: a case-crossover design. Environ. Res. 161, 299–303. Andersen, Z.J., et al., 2010. Association between short-term exposure to ultrafine particles and hospital admissions for stroke in Copenhagen, Denmark. Eur. Heart J. 31, 2034–2040. Antonetti, D.A., et al., 2012. Diabetic retinopathy. N. Engl. J. Med. 366, 1227–1239. Ashkenazi, A., et al., 1992. The quality assurance committee in a general hospital: its use in improvement of the medical record. Isr. J. Med. Sci. 28, 714–717. Bowe, B., et al., 2018. Particulate matter air pollution and the risk of incident CKD and progression to ESRD. J. Am. Soc. Nephrol. 29, 218–230. Bragg-Gresham, J., et al., 2018. County-level air quality and the prevalence of diagnosed chronic kidney disease in the US Medicare population. PLoS One 13, e0200612. Brook, R.D., et al., 2010. Particulate matter air pollution and cardiovascular disease: an update to the scientific statement from the American Heart association. Circulation 121, 2331–2378. Chan, T.C., et al., 2018. Long-term exposure to ambient fine particulate matter and chronic kidney disease: a cohort study. Environ. Health Perspect. 126, 107002. Chen, S.Y., et al., 2018. Traffic-related air pollution associated with chronic kidney disease among elderly residents in Taipei city. Environ. Pollut. 234, 838–845. Chin, W.S., et al., 2018. Effects of long-term exposure to CO and PM2.5 on microalbuminuria in type 2 diabetes. Int. J. Hyg Environ. Health 221, 602–608. Collart, P., et al., 2018. Short-term effects of nitrogen dioxide on hospital admissions for cardiovascular disease in Wallonia, Belgium. Int. J. Cardiol. 255, 231–236. Diabetes Prevention Program Research, G., 2007. The prevalence of retinopathy in impaired glucose tolerance and recent-onset diabetes in the Diabetes Prevention program. Diabet. Med. 24, 137–144. Dominici, F., et al., 2006. Fine particulate air pollution and hospital admission for cardiovascular and respiratory diseases. J. Am. Med. Assoc. 295, 1127–1134. Fong, D.S., et al., 2003. Diabetic retinopathy. Diabetes Care 26 (Suppl. 1), S99–S102. Gurung, A., et al., 2017. Particulate matter and risk of hospital admission in the Kathmandu valley, Nepal: a case-crossover study. Am. J. Epidemiol. 186, 573–580. Harris, M.I., et al., 1992. Onset of NIDDM occurs at least 4-7 yr before clinical diagnosis. Diabetes Care 15, 815–819. Hart, J.E., et al., 2015. Effect modification of long-term air pollution exposures and the risk of incident cardiovascular disease in US women. J. Am. Heart Assoc. 4. Ho, A.F.W., et al., 2018. The relationship between ambient air pollution and acute ischemic stroke: a time-stratified case-crossover study in a city-state with seasonal exposure to the Southeast asian haze problem. Ann. Emerg. Med. 72, 591–601. Institute for Health Metrics and Evaluation, GBD Compare, 2015. IHME. University of Washington, Seattle, WA. Jomova, K., Valko, M., 2011. Advances in metal-induced oxidative stress and human disease. Toxicology 283, 65–87. Jones, C.D., et al., 2012. Incidence and progression of diabetic retinopathy during 17 years of a population-based screening program in England. Diabetes Care 35, 592–596. Joussen, A.M., et al., 2002. Retinal vascular endothelial growth factor induces intercellular adhesion molecule-1 and endothelial nitric oxide synthase expression and

5