Global Associations of Air Pollution and Conjunctivitis Diseases: A Systematic Review and Meta-Analysis

(1) Background: As the most common eye disease diagnosed in emergency departments, conjunctivitis has caused serious health and economic burdens worldwide. However, whether air pollution may be a risk factor for conjunctivitis is still inconsistent among current evidence. (2) Methods: We searched the literature on the relationship between air pollution and conjunctivitis in multiple English databases before 18 March 2019. Meta-analysis, meta-regression, and funnel plots were used to integrate the data, identify the sources of bias, and determine the publication bias, respectively. (3) Results: A total of 2450 papers were found, 12 of which were finally included. The pooled relative risk for each 10 μg/m3 increase of air pollution on conjunctivitis was 1.0006 (95%CI: 0.9993–1.0019) for CO, 1.0287 (1.0120–1.0457) for NO2, 1.0089 (1.0030–1.0149) for O3, 1.0004 (0.9976–1.0032) for PM2.5, 1.0033 (0.9982–1.0083) for PM10, and 1.0045 (0.9908–1.0185) for SO2. In the subgroup, PM2.5 and O3 had a greater impact on conjunctivitis risk in women than in men, and people <18 years old than those ≥18 years old. Relative humidity significantly modified the risk of O3 on conjunctivitis (p = 0.023), explaining 45% of the between-study heterogeneity. (4) Conclusion: Globally, air pollution has considerable health risks for conjunctivitis. Females and the youth were more vulnerable to PM2.5, NO2, and O3. Reductions of air pollution levels are still warranted to protect the vulnerable populations.


Introduction
Ambient air pollution is one of the most important risk factors that affects people worldwide [1][2][3]. Numerous epidemiological investigations have revealed the short-term or long-term associations between high concentrations of air pollutants and increased health outcomes, including stroke, heart disease, lung cancer, diabetes, and chronic lung disease. Dense innervations in the ocular surface are extremely sensitive to environmental chemical substances. In addition, human eyes are only protected by a thin layer of tear film, causing them to be very susceptible to the harmful effect of air pollution [4][5][6]. 5. The target was not conjunctivitis diseases (e.g., rhinoconjunctivitis or conjunctivitis of other organs).

Data Extraction
Data from all included studies that were extracted were as follows: the reference, study design, demographic data (e.g., GDP and population), average values of air pollutants, meteorological variables (e.g., temperature, relative humidity, and air pressure), and effect estimates (e.g., RR, regression coefficient, 95% confidence interval, and standard error). For articles with missing information, we contacted the corresponding authors by email to obtain the relevant data.

Quality Assessment
In order to distinguish between low-quality and high-quality studies, a quality assessment was performed. Due to the wide variety of study designs used in the literature, assessing the quality and their risk of bias can be difficult.
To the best of our knowledge, no validated scale has been developed to assess the quality of time-series and case-crossover studies. We selected and combined several items from the New Castle Ottawa Scale [15], the Cochrane risk of bias tool, and other tools [16], which were utilized in previous studies [17][18][19]. We created a five-point scoring system that included the following four aspects: a. Conjunctivitis disease occurrence verification (0-1 points) According to the International Classification of Diseases, studies on the causes of death encoded in revised version 9 (ICD-9), 10th revision, or ICPC-2 Code(s) (International Classification of Primary Care, Second Edition [20]) and official definitions of other countries are given a score of 1, but no score is given for studies that do not meet the criteria.
b. Quality of air pollutant measurements (0-1 points) The quality of the air pollutant measurement can be judged according to the measurement frequency and the existence of missing data. If the measurement is made at least once a day and the missing data is <25%, the research score is 1; otherwise, the quality is assessed with 0 points. c. Adjustment degree of confounders (0-3 points) Adjustment for temperature and humidity is given 1 point. Additional adjustments, for example, seasonality, wind speed, or rainfall, acquire 2 points. If the long-term trend and days of the week are considered, 3 points are given. Zero points are given if there is no adjustment for temperature and humidity.
If the study gets full marks for all three components, the study was considered to be of a good quality. If any of the three components were zero, the study quality was considered to be low. All other studies were considered to be of a medium quality.

Data Synthesis and Statistical Analysis
The key objective of data synthesis was to unify the air pollutant concentration units, group the research population, and standardize the risk effect values. If studies used mg/m 3 , ppm, or ppb for the unit of measurement or unit of increment, all estimates were converted into µg/m 3 . Regarding population groupings, the included data were mainly divided into two groups: gender (male and female) and age group (>18 years old and <18 years old). In most studies, the risk estimates were expressed as ERs, ORs, or RRs with 95% CIs, and percent changes. The results presented as a regression coefficient and standard error were converted to RR. The summarized statistics are expressed as RRs with 95% CIs [21,22]. To pool the effect estimates, all estimates were standardized to an increment of 10 µg/m 3 of air pollutant (CO, O 3 , SO 2 , NO 2 , PM 2.5 , and PM 10 ) concentration.
The statistical analysis consisted of three steps: (1) computing the integrated estimates of each type of air pollutant using a fixed-or random-effect meta-analysis; (2) conducting a meta regression analysis based on the total population, GDP, and weather conditions; and (3) performing a sensitivity analysis. A meta-analysis was used to aggregate the risk estimates from all studies in detail. If the heterogeneity index (I 2 ) was greater than 25%, the aggregate estimates were calculated using a random effect model; otherwise, we selected the fixed effect model [23]. The second step was to judge and test the source of heterogeneity. Heterogeneity was classified as high (I 2 > 75%), medium (25 < I 2 < 75%), or low (I 2 < 25%) [24]. The sources of heterogeneity, such as the research design, regional GDP, geographic location (longitude and latitude, temperature, and humidity), and weather conditions, were further tested using a meta-regression analysis. Finally, we applied funnel charts and Begg's [25] and Egger's tests [26] to assess the potential impact of publishing bias. We conducted the sensitivity analysis by re-calculating the pooled effects by excluding each study to test whether our main findings were influenced by one study.
Statistical analysis and drawing were mainly conducted using R language software (R version 3.6.0; R Development Core Team, New Zealand, Australia).

Search Results and Study Characteristics
In this study, 2450 records were originally obtained from Scopus (n = 723), PubMed (n = 576), Embase (n = 440), and Web of Science (n = 710). Twelve articles from 10 regions met the inclusion criteria and were included in the meta-analysis (see Figure 1), covering 30,103,982 conjunctivitis patients. Among the 12 included studies, five were case-crossover studies [4,5,9,27,28], four were time-series studies [6,14,29,30], and three were other studies (e.g., spatial analysis and multi-level regression). Tables 1 and A2 summarize the basic characteristics of the included studies. The number of research papers including CO, NO 2 , O 3 , PM 2.5 , PM 10 , and SO 2 was two, seven, nine, four, seven, and seven, respectively.

Overall Analysis
Since significant heterogeneity (I 2 > 60%) was observed in the included studies, we used a random-effect meta-analysis to integrate the effect estimates of various air pollutants on conjunctivitis [31]. Figure 2 presents the pooled effect of six air pollutants on the risk of conjunctivitis among the included studies. The pooled relative risk for each 10 µg/m 3 increase of air pollutants on conjunctivitis was 1.0006 (95%CI: 0.9993-1.0019) for CO, 1.0287 (95%CI: 1.0120-1.0457) for NO 2

Subgroup Analysis
Given the limited number of articles, we could only combine the effect estimates by subgroup for PM 2.5 , NO 2 , and O 3 ( Table 2). The random-effect meta-analysis was used to pool the effect risk of air pollution on conjunctivitis among subgroups as the heterogeneity was significant. Generally, the impact of air pollution was higher among females and the youth than the other groups. However, only statistically significant effects of O 3 on males, with an RR value of 1.0321 (95%CI: 1.0000-1.0653), and NO 2     There was a much higher effect of nitrogen dioxide on visits for conjunctivitis when delayed effects were considered. Conjunctivitis was also significantly associated with PM 10 and ozone levels. There were higher risks of conjunctivitis in rural areas, but higher sensitization to air pollutants in urban cities. Children, females, and the older population were at higher risks for both types of conjunctivitis.

Meta-Regression
In order to assess the source of the between-study heterogeneity, a meta-regression was further conducted to test the influence of city-level characteristics (e.g., GDP, longitude and latitude, average temperature, relative humidity, and duration of sunshine) on the relationship between air pollution and conjunctivitis (see Table 3). Among these factors, only the relative humidity significantly modified the risk of O 3 for conjunctivitis (p = 0.023), explaining 45% of the between-study heterogeneity.

Publication Bias
Funnel plot, Begg's, and Egger's tests were applied to determine whether there was publication bias. Figure 3 shows the funnel plots of the meta-analysis for the association between air pollution and the risk of conjunctivitis. The results of PM 2.5 , SO 2 , and NO 2 presented a low probability of publication bias, reporting a p-value for both Begg's test and Egger's test of over 0.05. However, potential publication bias was detected for PM 10 (Egger's test: Z-value = 2.4238, p = 0.0154) and O 3 (Egger's test: Z-value = 5.4884, p < 0.001) (see Table 4). In addition, we performed the trim and fill method to validate the publication bias of PM 10 and O 3 (see Figure A1). The adjusted pooled relative risk of PM 10 for total conjunctivitis was 1.0026 (95%CI: 0.9975, 1.0077) and 1.0041 (95%CI: 0.9957, 1.0126) for O 3 .  Note: Egger's test was unavailable for the CO because of the limited number of studies on the association between CO and the risk of conjunctivitis. The trim-fill test was only performed for PM10 and O3, which showed significant publication bias. CO-carbon monoxide; NO2-nitrogen dioxide; SO2-sulfur dioxide; O3-ozone; PM2.5-particles smaller than 2.5 μm; PM10-particles smaller than 10 μm.

Sensitivity Analysis
Sensitivity analyses were performed to estimate the stability of the results by recalculating the pooled effect estimates after omitting one study each time [32][33][34]. We found that the effect estimate of each 10 µg/m 3 increase in the six air pollutants showed no significant change by removing one single study, suggesting that the combined results were relatively stable and reliable.

Discussion
To the best of our knowledge, this is the first systematic review and meta-analysis to assess the association between air pollution and conjunctivitis. Twelve studies, including 30,103,982 cases of conjunctivitis from 10 countries/regions around the world, were included. Positive associations between six common air pollutants and conjunctivitis were obtained, while statistical significance was only observed for NO 2 and O 3 . The female subgroup and those under 18 years old were most vulnerable to the risk of conjunctivitis caused by air pollution.

Risk Analysis of Air Pollution and Conjunctivitis in the Whole Population
In the past decade, the effect of air pollution on conjunctivitis has attracted increasing interest [4,35,36]. However, the evidence so far is inconsistent (Figure 2). For instance, Fu et al. [5] revealed that the risk of NO 2 and conjunctivitis in the population was significant, with an RR value of 1.0403 (95%CI: 1.0228, 1.0581), while Jamaludin et al. [30] did not find any significant effects on the risk of conjunctivitis in the population, with an RR value of 0.9989 (95%CI: 0.9205, 1.0840). For PM 10 , Chang et al.'s [4] study revealed that PM 10 was significantly associated with the conjunctivitis risk among people, with an RR value of 1.0020 (95%CI: 1.0005, 1.0036). However, in the study of Chiang et al. [29], NO 2 had no significant effect on the risk of conjunctivitis in people, with an RR value of 0.9933 (95%CI: 0.9867, 1.0000). For SO 2 , Fu et al.'s study [5] revealed that the risk of conjunctivitis between SO 2 and the population was significant, with an RR value of 1.0480 (95%CI: 1.0040, 1.0939). In the study of Jamaludin et al. [30], SO 2 had a protective effect on the conjunctivitis risk among people, with an RR value of 0.8468 (95%CI: 0.7371, 0.9730). Air pollution is gradually occupying an important position in the risk factors of conjunctivitis. Our study shows that all six air pollutants have a positive correlation with conjunctivitis. Among them, NO 2 had the most significant effect, followed by O 3 . This may be due to differences in the physical and chemical properties between pollutants, resulting in different risk outcomes. Both NO 2 and O 3 are highly oxidative and irritating to the eyes [37][38][39][40]. According to the chemical properties of O 3 and NO 2 , O 3 is easily removed by a reaction, so the lifetime of NO 2 is longer than that of O 3 [41,42]. In addition, in terms of toxicity, the toxicity of O 3 may be more complex than that of NO 2 [43,44], which may have a significant potential impact on eye tissue cells. From the comprehensive analysis of the toxicity degree and lifetime of pollutants, NO 2 and O 3 have obvious risks for conjunctivitis in the population, among which, NO 2 has the highest risk value, followed by O 3 .

Risk Analysis of Air Pollution and Conjunctivitis in Subgroups
According to the research analysis, PM 2.5 , NO 2 , and O 3 present a higher risk for conjunctivitis in women than in men; meanwhile, PM 2.5 and O 3 exhibit a higher risk for conjunctivitis for people under 18 years of age than people over 18 years of age, whereas NO 2 had the opposite effect. Between genders, there are three possible reasons for the greater risk of conjunctivitis in women. First, women's physical function is generally not as good as men's [45], so their ability to resist air pollution is relatively weak. Second, women spend more time indoors than men [46,47], and indoor air circulation is not strong, so more toxic and harmful air pollutants may more easily accumulate and then be absorbed. Third, compared with men, women prefer makeup [48], especially eye shadows, eyelashes, and contact lenses. Studies have shown that these types of eye makeup can cause discomfort to the eyes, such as dryness, pain, etc. [49][50][51][52], which may increase the risk of conjunctivitis. Therefore, in combination with the above points, the risk for females of conjunctivitis is greater than that for males. In terms of the age group, for people younger than 18 years old, the development of physical function and the defense ability is still immature and they are thus vulnerable to air pollutants. The effects of NO 2 on people over 18 years of age was significantly greater than that on people under 18 years of age, which may be related to people's living and working habits. People over the age of 18 go to work, which often involves the need to travel between cities, so there is a relatively high chance of exposure to severe air pollution scenarios [53]. Exposure to more mobile sources of pollution, such as NO 2 emitted by automobiles [54], increases the risk of conjunctivitis in adults.

Source of Heterogeneity and Possible Bias
For GDP, latitude, longitude, temperature, and humidity, we observed substantial heterogeneity in the pooled effect sizes of air pollutants (NO 2 , O 3 , PM 2.5 , PM 10 , and SO 2 ) for conjunctivitis. We found that there was a negative correlation between relative humidity and the risk of conjunctivitis for five kinds of air pollutants. There may be several explanations for this.
First, the higher the humidity in the air, the easier it is to condense and settle the solid particles in the air [55], and the easier it is to dilute the liquid or gaseous pollutants. These processes can reduce the concentration of pollutants in the air, thereby reducing the risk of conjunctivitis. Second, a high humidity will affect visibility [56], which will affect people's travel habits; therefore, to a certain extent, it can reduce the risk of exposure to conjunctivitis. Finally, from a physiological point of view, in greater humidity, the eyes will be relatively comfortable (so it is not easy to itch the eyes, not easy to rub the eyes, etc.) and thus dry eye will not be easily caused [57]. Furthermore, it reduces the risk of conjunctivitis.

Possible Mechanisms Explaining the Relation between Conjunctivitis and Air Pollution
To date, the underlying pathophysiological mechanism of conjunctivitis caused by air pollutants is still unclear. As a human's eyes are directly exposed to air pollution, some studies have speculated that PM 2.5 [35,58,59] and PM 10 [60] particles may easily cause the inadaptability of intraocular epidermal cells, leading to cell death and the inflammation of tissue cells. Second, NO 2 and O 3 have strong oxidative stress effects [61], which may stimulate conjunctival cell inflammation. Finally, NO 2 is an acidic gas. When it enters the eyes, it easily changes the acidic and alkaline environment of the inner epidermis cells of the eyes [62], breaking the function of the eye cells and causing inflammation [63,64]. It is plausible that the association between air pollution and the risk of conjunctivitis events is a result of these important mechanistic pathways.

Limitations and Implications
Several limitations of our study should be considered. First, almost all the included references used the air pollutant data from fixed environmental monitoring stations instead of individual-level air pollutant exposures, which may have led to measurement error. Second, we included studies in the same place at different times (for example, Taiwan), which may have also had an impact on the combined value of conjunctivitis risk. Finally, few studies were available on the association between some types of air pollutants (e.g., carbon monoxide) and the risk of conjunctivitis, which led to a relatively low statistical power and limited the further stratified assessment for subgroups. Therefore, future epidemiological evidence from more countries and/or cities with a well-designed strategy is required to be able to develop more comprehensive knowledge on the effect of air pollution on the risk of conjunctivitis. Further investigations are also needed to identify the subgroups that are most vulnerable to air pollution, and the socioeconomic status should be considered. It would also be useful to explore the use of alternative exposure metrics that are more representative of individual exposure, and it would be beneficial to examine the mechanism underlying the harmful effect of air pollution on patients with conjunctivitis. Additionally, a cost-effectiveness of preventive measures for improving the air quality to reduce the incidence of conjunctivitis is also needed in future research.

Conclusions
This meta-analysis found that air pollution is an important factor for the risk of conjunctivitis. NO 2 presented the highest impact on patients with conjunctivitis, followed by O 3 . For different sub-groups of patients with conjunctivitis, females and the age group under 18 years old were more sensitive to the air pollution. Notable inconsistencies in the various studies have been found for the association between air pollution and conjunctivitis, while only relative humidity significantly modified the risk of O 3 for conjunctivitis, which explained 45% of the between-study heterogeneity. Our findings highlight the necessity for the reduction of air pollution levels and protection of vulnerable populations. Further research is needed to better understand the mechanisms underlying the harmful effect of air pollutants on the risk of conjunctivitis. Future well-designed epidemiological studies from more countries and/or cities are still warranted to be able to get more comprehensive knowledge and powerful evidence about the effect of air pollution on the risk of conjunctivitis and identification of the subpopulations sensitive to air pollution.  (TS=("conjunctivitis" OR "endophthalmitis" OR "ophthalmia" OR "pinkeye" OR "conjunctivitis" OR "Pink eye") OR TI=("conjunctivitis" OR "endophthalmitis" OR "ophthalmia" OR "pinkeye" OR "conjunctivitis" OR "Pink eye")) TITLE-ABS-KEY("conjunctivitis" OR "endophthalmitis" OR "ophthalmia" OR "pinkeye" OR "conjunctivitis" OR "Pink eye") AND TITLE-ABS-KEY("air pollution" OR "ambient air pollution" OR "outdoor air pollution" OR "atmospheric pollution") "conjunctivitis":ti,ab,kw OR "endophthalmitis":ti,ab,kw OR "ophthalmia":ti,ab,kw OR "pinkeye":ti,ab,kw OR "pink eye":ti,ab,kw [2] ("air pollution" (TS=("air pollution" OR "ambient air pollution" OR "outdoor air pollution" OR "atmospheric pollution") OR TI=("air pollution" OR "ambient air pollution" OR "outdoor air pollution" OR "atmospheric pollution")) TITLE-ABS-KEY("conjunctivitis" OR "endophthalmitis" OR "ophthalmia" OR "pinkeye" OR "conjunctivitis" OR "Pink eye") AND TITLE-ABS-KEY( "PM 2.5 " OR "Particulate Matter2.5" OR "particulate matter" OR "PM 10 " OR "Particulate Matter 10" OR "SO 2 " OR "Sulfur dioxide" OR "NO 2 " OR "Nitrogen dioxide" OR "NO x " OR "Nitrogen oxides" OR "O 3 " OR "ozone" OR "CO" OR "Carbon monoxide" OR "Smog" OR "black carbon") "air pollution":ti,ab,kw OR "ambient air pollution":ti,ab,kw OR "outdoor air pollution":ti,ab,kw OR "atmospheric pollution":ti,ab,kw (TS=( "PM 2.5 " OR "Particulate Matter2.5" OR "particulate matter" OR "PM 10 " OR "Particulate Matter 10" OR "SO 2 " OR "Sulfur dioxide" OR "NO 2 " OR "Nitrogen dioxide" OR "NO x " OR "Nitrogen oxides" OR "O 3 " OR "ozone" OR "CO" OR "Carbon monoxide" OR "Smog" OR "black carbon") OR TI=( "PM 2.5 " OR "Particulate Matter2.5" OR "particulate matter" OR "PM 10 " OR "Particulate Matter 10" OR "SO 2 " OR "Sulfur dioxide" OR "NO 2 " OR "Nitrogen dioxide" OR "NO x " OR "Nitrogen oxides" OR "O 3 " OR "ozone" OR "CO" OR "Carbon monoxide" OR "Smog" OR "black carbon")) TITLE-ABS-KEY("conjunctivitis" OR "endophthalmitis" OR "ophthalmia" OR "pinkeye" OR "conjunctivitis" OR "Pink eye") AND TITLE-ABS-KEY("air pollution" OR "ambient air pollution" OR "outdoor air pollution" OR "atmospheric pollution") AND TITLE-ABS-KEY( "PM 2.5 " OR "Particulate Matter2.5" OR "particulate matter" OR "PM 10 " OR "Particulate Matter 10" OR "SO 2 " OR "Sulfur dioxide" OR "NO 2 " OR "Nitrogen dioxide" OR "NO x " OR "Nitrogen oxides" OR "O 3 " OR "ozone" OR "CO" OR "Carbon monoxide" OR "Smog" OR "black carbon") "PM 2.5 ":ti,ab,kw OR "particulate matter2.5":ti,ab,kw OR "particulate matter":ti,ab,kw OR "PM 10 ":ti,ab,kw OR "particulate matter 10":ti,ab,kw OR "SO 2 ":ti,ab,kw OR "sulfur dioxide":ti,ab,kw OR "NO 2 ":ti,ab,kw OR "nitrogen dioxide":ti,ab,kw OR "NO x ":ti,ab,kw OR "nitrogen oxides":ti,ab,kw OR "O 3 ":ti,ab,kw OR "ozone":ti,ab,kw OR "CO":ti,ab,kw OR "carbon monoxide":ti,ab,kw OR "smog":ti,ab,kw OR "black carbon":ti,ab,kw    Note. The number in the column of "literature" denotes the number of the literature from Table A3 that was excluded, and effect estimates from the rest of the literature were then pooled using a meta-analysis. Q-p denotes the p-value for the Q test. Figure A1. Funnel plot of PM10 and O3 on conjunctivitis using Trim-fill method. The solid circles denote effect estimates from included studies and open circles denotes estimates provided by Trim-fill method.