Increased Incidence of Dysmenorrhea in Women Exposed to Higher Concentrations of NO, NO2, NOx, CO, and PM2.5: A Nationwide Population-Based Study

Background: Air pollution is speculated to affect the reproductive health of women. However, a longitudinal association between exposure to air pollution and dysmenorrhea has not been identified, which this study aimed to examine this point. Methods: Two nationwide databases, namely the Taiwan Air Quality Monitoring database and the Taiwan National Health Research Institutes database were linked. Women with a history of dysmenorrhea (International Classification of Disease, Ninth Revision, Clinical Modification code 625.3) before 2000 were excluded. All participants were followed from January 1, 2000 until the diagnosis of dysmenorrhea, withdrawal from National Health Insurance, or December 31, 2013. Furthermore, air pollutants were categorized into quartiles with three cut-off points (25th, 50th, and 75th percentiles). The Cox regression model was used to calculate the hazard ratios of dysmenorrhea. Results: This study enrolled 296,078 women. The mean concentrations of yearly air pollutants were 28.2 (±12.6) ppb for nitric oxides (NOx), 8.91 (±7.93) ppb for nitric oxide (NO), 19.3 (±5.49) ppb for nitrogen dioxide (NO2), 0.54 (±0.18) ppm for carbon monoxide (CO), and 31.8 (±6.80) μg/m3 for PM2.5. In total, 12,514 individuals developed dysmenorrhea during the 12-year follow-up. Relative to women exposed to Q1 concentrations of NOx, women exposed to Q4 concentrations exhibited a significantly higher dysmenorrhea risk [adjusted hazard ratio (aHR)= 27.9, 95% confidence interval (CI) = 21.6–31.3]; similarly higher risk was found for exposure to NO (aHR = 16.7, 95% CI = 15.4–18.4) and NO2 (aHR = 33.1, 95% CI = 30.9–37.4). For CO, the relative dysmenorrhea risk in women with Q4 level exposure was 28.7 (95% CI = 25.4–33.6). For PM2.5, women at the Q4 exposure level were 27.6 times (95% CI = 23.1–29.1) more likely to develop dysmenorrhea than those at the Q1 exposure level. Conclusion: Our results showed that women would have higher dysmenorrhea incidences while exposure to high concentrations of NO, NO2, NOx, CO, and PM2.5.

For PM 2.5 , women at the Q4 exposure level were 27.6 times (95% CI = 23.1-29.1) more likely to develop dysmenorrhea than those at the Q1 exposure level.

INTRODUCTION
Dysmenorrhea, characterized by painful cramps of the uterus during menstruation, affects 45-95% of menstruating women. Dysmenorrhea is the leading morbidity among gynecological disorders and the leading cause of pelvic pain (1), thus representing the greatest burden for menstruating women (2,3). Moreover, dysmenorrhea is related to negative performance in school and social and sports activities as well as shortterm school absenteeism among adolescents (4). Furthermore, dysmenorrhea affects the economy insofar as it impairs work productivity (5), and almost 600 million working hours are lost yearly due to dysmenorrhea (6). It is classified as either primary dysmenorrhea, which does not involve a related organic disease, or secondary dysmenorrhea, which is related to attributable causes such as endometriosis (7). Although the actual causes of primary dysmenorrhea are unclear, the most well-known pathogenesis is prostaglandin overproduction in the uterus, particularly overproduction of PGF 2a and PGF 2 (8). Therefore, the mechanism of primary dysmenorrhea involves arachidonic acid (AA) and inflammation (9), and thus, the effective treatment for primary dysmenorrhea is non-steroidal anti-inflammatory drugs and cyclooxygenase (COX) inhibitors (10,11).
Mahalingaiah et al. (12) and Carré et al. (13) observed that all size fractions of particulate matter (PM) exposure and traffic-related air pollution are associated with incidence of infertility and adverse reproductive health (14). Merklinger-Gruchala et al. (15) and Mahalingaiah et al. (16) showed that the mean concentrations of the air pollutants particles 10 microns and below(PM 10 ), Sulfur dioxide(SO 2 ), Nitrogen Oxides (NO x ), and Carbon monoxide (CO) were associated with shortening of the luteal phase, an irregular menstrual cycle, menstrual disorders and menstrual irregularity, respectively. The above evidence suggested that although infertility and irregular menstrual cycle had many different anatomic or hormonal causes, they would be affected by air pollutants. Therefore, air pollutants might have effect on menstruation. However, to the best of our knowledge, no epidemiological study has investigated the association between primary dysmenorrhea and air pollutants.
Therefore, we used nationwide medical records and nationwide air pollution monitoring databases to conduct a retrospective cohort study to determine whether exposure to air pollutants is associated with high dysmenorrhea risk.

Data Source
To clarify dysmenorrhea risk in participants exposed to air pollutants, we used the Longitudinal Health Insurance Database 2000 (LHID 2000) from the Taiwan National Health Insurance Research Database (NHIRD), which includes ∼99% of the population in Taiwan. The LHID 2000 consists of 1 million patients who were randomly selected from NHIRD in 2000. LHID 2000 contains comprehensive deidentified health care information regarding patient demographics, outpatient visits, inpatient care, prescription drugs, and medical procedures from 1996 to 2013. The National Health Research Institutes confirmed that no significant differences were evident in the distribution of age, sex, or health care costs between the population in the LHID 2000 and that in the NHIRD. This study used the International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) to categorize disease diagnoses based on inpatient data.
Taiwan Environmental Protection Administration (EPA) have set up 76 air quality observation stations nationwide for the Air Quality Observation and Forecast Network at the end of 2006 (17) (Supplementary Figure 1) According to different pollution characteristics, weather conditions, and observational purposes of EPA, the air quality observation stations were distributed reasonably in cities and countries and divided into seven main island regions and three outlying islands regions (18) (Supplementary Figure 1). Geographic information system (GIS) (ArcGIS version 10; ESRI, Redlands, CA, USA) was used to identify and manage the locations of the monitoring stations and air pollution sources. The monitoring data were integrated into yearly point data and interpolated to pollutant surfaces using inverse distance weighting method (IDW) (19).
We obtained air pollution data [nitric oxides (NO x ), nitric oxide (NO), nitrogen dioxide (NO 2 ), carbon monoxide (CO), and PM 2.5 . from the Taiwan Air Quality Monitoring Database (TAQMD). The TAQMD provides daily air pollution data from community-based monitoring sites that are available in realtime on the organizations' website. We downloaded data for the pollutants NO x , NO, NO 2 , CO, and PM 2.5 and calculated their annual mean levels from January 2000 to December 2012. To define average exposure duration of women to these pollutants is really an issue. If exposure duration to air pollutants was defined too shorter (e.g., <1 year or only several months), the results that air pollutants is associated with risks of dysmenorrheal would not be convincing. In the other aspect, if duration was defined too long (e.g., longer than 10 or 5 years), the unknown and uncontrolled bias (i.e., body weight change, take hormone, change of residence, or pregnancy, etc) would be too much to make the results convincing. To avoid and minimize such bias, we defined duration of average exposure levels of women to these pollutants as 2 years. Then, we computed the average exposure levels of women to these pollutants for 2 years before the diagnosis of dysmenorrhea or end of the study period for each individual.

Outcome and Exposure Measurement
The study endpoint was the diagnosis of dysmenorrhea (ICD-9-CM code 625.3). We integrated daily concentrations of air pollutants corresponding to residential zip codes to calculate 2-year average exposure before the diagnosis of dysmenorrhea and the year of diagnosis of dysmenorrhea from 2000 to 2012.To estimate the effect of air pollutants on dysmenorrhea, we investigated the daily concentrations of NO x , NO, NO 2 , CO, and PM 2.5 by using the inverse distance weighting (IDW) method. IDW is a common and simple spatial interpolation method. It predicts values of unknown points on the basis of the similarity of two objects in terms of distance (22). Its principle is to use distance measurement to perform the weighted analysis of the interpolation. When the unknown point is estimated to be closer to the known point, the weighted value of the unknown point will be higher. Hence, when the distribution of the known measuring point is balanced, the value estimated using the IDW method will be more accurate (Supplementary Table 1; Supplementary Figure 1) (23,24). In this study, the generalized additive model (GAM) was used to quantitatively evaluate the acute effects of air pollution including NOx, NO, NO2, PM 2.5 , and CO concentrations and interquartile range on the incidence of dysmenorrhea in Taiwan. We found that the plots of the generalized additive model for interquartile range more linear than the plots of generalized additive model for air pollution concentrations. In this study, the air pollutant exposure levels were scaled to the interquartile range (IQR). Furthermore, Air pollutants were also categorized into quartiles with three cutoff point (25th, 50th and 75th percentiles) for NOx (Quartile 1,

Statistical Analysis
We conducted a Cox proportional hazards regression for the multivariable analysis of dysmenorrhea diagnosis through increase in IQR (µg/m 3 or ppb) of long-term exposure to NO x , NO, NO 2 , CO, and PM 2.5 from 2000 to dysmenorrhea development or the end of the study period (25).
Dysmenorrhea incidence (per 10,000 person-years) was calculated at each level of air pollutant concentration. The relative risk of dysmenorrhea in participants who were exposed to Q2-Q4 levels of air pollutant concentrations relative to those exposed to the Q1 level of air pollutant concentrations was estimated using the Cox proportional hazards regression model. The potential confounders, namely age, monthly income, urbanization level, occupational class, and comorbidities, were defined as determinants of dysmenorrhea and associated with air pollution concentrations and further incorporated into models. Furthermore, we used the Cox regression model, which is stratified by age, urbanization level, monthly income, and occupational class, for analyzing dysmenorrhea risk between various yearly average concentrations of air pollutants. This study uses Generalized additive model (GAM) to analyze the impact of air pollutant on dysmenorrhea (26,27). GAM is a non-parametric extension of Generalized linear model (GLM) (28). It not only retains the basic framework of GLM, but also has many assumptions as other linear models (such as normal assumption or variance homogeneity), and its ability to handle non-linear models is better than other models (29). It is powerful and suitable for dealing with overly complex and non-linear relationships among many variables. In this study, air pollution data are usually non-linear, which limits the application of GLM, so it can be solved through GAM (30). The aim of this study was to use GAM to evaluate the relation between air pollution and dysmenorrheal by using data from the NHIR database and TAQM database (296,078 participants of whom 12,514 died during 11.7 years follow-up). Using GAM, we found that strong positive association between air pollution and dysmenorrhea, the association also appeared in low concentrations of air pollution (Supplementary Figure 2).

RESULTS
There were total 296,078 participants in this study, and there were 12,514 participants had diagnosis of dysmenorrhea. The distribution of urbanization level among different quartile of air pollutant levels was displayed in Supplementary Table 2. Participants exposed to the Q4 level air pollutant including NOx, NO, NO 2 , PM 2.5 , and CO mostly lived in the level 1 urbanization area. Table 2 shows the incidence rate of dysmenorrhea among levels of all air pollutant concentrations. After controlling for potential confounding factors, Q4 air pollutant level exposure had higher risk of dysmenorrhea significantly compared with Q1 level of air pollutants. For NOx, relative to Q1 concentrations, the Q4 (adjusted = 27.9, 95%CI = 21.6-31.3) concentrations were had a significant higher risk of dysmenorrhea. Relative to Q1 NO concentrations, the Q4 (adjusted = 16.7, 95% CI = 15.4-18.4) also had a significant higher risk of dysmenorrhea. Relative to Q1 NO 2 concentrations, the Q4 (adjusted = 33.1, 95% CI = 30.9-37.4) also had a significant higher risk of dysmenorrhea. The relative risks of dysmenorrhea in Q4 level exposure were 27.6 (95% CI = 23.1-29.1) for PM 2.5 . People under Q4 CO level exposure had 28.7 (95% CI = 25.4-33.6) fold increased risk to develop dysmenorrhea than those under Q1 level. Table 3 showed the result of stratified Cox regression analysis by considering subject exposed to Q1 level of air pollutants as reference group. Relative to Q1 concentrations of NOx, NO, NO 2, PM 2.5 , and CO, the participants who exposed in Q4 level of air pollutants had a significant higher risk of dysmenorrhea in each age group, monthly income, occupational class, and urbanization level 1, 2, and 3. Participants who aged 30-44 exposed to Q4 quartile of NOx and NO had 33.7 (95% CI = 28.4-36.5) fold   Participants whose occupational class as others exposed to Q4 quartile of NOx, NO, and NO 2 had increased risk of dysmenorrheal up to 34.5 (95% CI = 29.2-84.5), 48.6 (95% CI = 43.5-69.8), 64.5 (95% CI = 49.4-87.8) fold accordingly compared with those under Q1 level. Supplementary Figures 3-7 provides illustrations of these results. Table 4 showed associated HRs in participants exposed to every increment of one unit of yearly average concentrations of air pollutants. For every increment of 1 unit yearly average of NOX, the adjusted HR of dysmenorrhea would be increased to 1.51. For every increment of 1 unit yearly average of NO2, the adjusted HR of dysmenorrhea would be increased to 1.72. For every increment of 1 unit yearly average of NO, the adjusted HR of dysmenorrhea would be increased to 1.53. For every increment of 1 unit yearly average of CO, the adjusted HR of dysmenorrhea would be increased to 1.009. For every increment of 1 unit yearly average of PM 2.5 , the adjusted HR of dysmenorrhea would be increased to 1.81. Table 5 showed the synergy effects of PM 2.3 and NOx, NO, NO 2 , and CO on the risks of dysmenorrhea. The cut value was median value of air pollutant which exposure concentration being less than median value was defined as low and being more than median value was defined as high. Women exposed to low PM 2.5 and high NOx had 1.30-fold adjusted risk of dysmenorrhea (95% CI = 1.21-1.39), compared with those exposure to low PM 2.5 and low NOx. Women exposed to low PM 2.5 and high NO had 2.64-fold adjusted risk of dysmenorrhea (95% CI = 2.46-2.84), compared with those exposure to low PM 2.5 and low NO. Women exposed to low PM 2.5 and high NO 2 had 3.75-fold adjusted risk of dysmenorrhea (95% CI = 3.84-4.05), compared with those exposure to low PM 2.5 and low NO 2 . Women exposed to low PM 2.5 and high CO had 5.03-fold adjusted risk of dysmenorrhea (95% CI = 4.67-5.42), compared with those exposure to low PM 2.5 and low CO. Women exposed to high PM 2.5 and high concentrations of NOx, NO, NO 2 , and CO had highest risk of dysmenorrhea in each category, which 10.8-fold adjusted risk of dysmenorrhea (95% CI = 10.0-11.7) for NO X , 6.46-fold adjusted risk of dysmenorrhea (95% CI = 6.01-6.94) for NO, 9.77-fold adjusted risk of dysmenorrhea (95% CI =

DISCUSSION
Our results demonstrated that women exposed to relatively high concentrations of NO, NO 2 , NO x , CO, and PM 2.5 had relatively high incidences of dysmenorrhea. This was the first cohort study to identify the association between dysmenorrhea and exposure to certain air pollutants.
Since no previous studies have reported association between air pollution and dysmenorrhea, there was no available direct information to be cited scientifically for our findings. However, in recent years, air pollution has been reported to be associated with many diseases, which resulting from direct injures to lung as well as provoking systematic inflammation (31,32). Pope et al. reported that exposure to PM 2.5 was associated with endothelial injury and systemic inflammation, which involves increased thrombotic propensity, loss of apoptosis of endothelial cells, and changes in cytokines (33). Chen et al. showed that exposure to fine-particle air pollutants was associated with increased urinary metabolites, namely AA, which is indicative of the inflammatory property of air pollution (34). Therefore, systematic inflammation and associated increasing oxidative stress would be plausible pathway for association between air pollution and dysmenorrhea.
First, increased levels of 8-epi-PGF2α have been detected in exhaled breath and urine after exposure to air pollution PM 2.5 (44)(45)(46)(47)(48). Furthermore, Schneider et al. and Samet et al. found that air pollution could activate macrophages, leading to the production of prostaglandin E2 and induce induce prostaglandin H synthase 2 (35,49). Yan et al. found that NO 2 inhalation could promote COX-2 elevation-mediated prostaglandin E2 production (36). Sang et al. showed that sulfur dioxide (SO 2 ) could induce cyclooxygenases-2-derived prostaglandin E2 production and its downstream signaling pathway in neurons (50). Therefore, exposing to air pollution NOx, NO2, and PM 2.5 could cause increasing levels of prostaglandin E2 production within body. Since increasing levels of prostaglandin E2 had been related with dysmenorrhea (51, 52), we supposed that it would be one possible explanation pathway for our findings.
Second, myeloperoxidase activation plays a role in the association between air pollution and increased incidence of dysmenorrhea. Kaplan et al. summarized that myeloperoxidase as well as oxidative stress and calcium ion levels are involved in the pathogenesis of primary dysmenorrhea (53). Also, oxidative stress and increased serum levels of the cytokines malondialdehyde and interleukin-6 are associated with dysmenorrhea (54). Constantin showed that SO 2 and NO 2 induce oxidative burst and myeloperoxidase activation (37). Moreover, several studies have shown an association between air pollution and oxidative stress in humans (38)(39)(40)(41)(42). Szmidt et al. have systematic reviewed that oxidative stress was associated with development of dysmenorrhea (55). Results of Szmidt et al. indicated an elevated level of oxidative stress, especially of lipid peroxidation existed among dysmenorrhea women (55). Yang et al. have found that exposure to PM 2.5 will cause dysregulation of lipid metabolism and oxidation (56). Therefore, women exposure to NO 2, NO X, as well as PM 2.5 might lead to increasing oxidative stress, therefore had higher risks of dysmenorrhea.
Third, an irregular menstrual cycle is also associated with dysmenorrhea in women exposed to high concentrations of the air pollutants SO 2 and NO x . Merklinger-Gruchala et al. (15) and Mahalingaiah et al. (16) found that the concentrations of the air pollutants were associated with shortening of the luteal phase, an irregular menstrual cycle, and menstrual irregularity. An irregular menstrual cycle is also a risk factor for dysmenorrhea (57), and therefore, air pollution might lead to dysmenorrhea through irregular menstruation. Additionally, Beck revealed that CO intoxication is related to dysmenorrhea, menorrhagia, and amenorrhea in women (58). Although we have adjusted for irregular menses, irregular menses might also be speculated as one plausible pathway for air pollution related dysmenorrhea since not all patients with irregular menses would seek for medical therapy.
Finally, emotional stress resulting from exposure to air pollution may contribute to increased incidence of dysmenorrhea. Power et al. reported that exposure to fine particulate air pollution was associated with anxiety (43). Further, extensive long-term exposure to air pollution may increase the odds of depression, antidepressant use, and mental health problems in children (59,60). Since mental stress has been reported to be a risk factor for dysmenorrhea (61)(62)(63)(64), therefore exposure to air pollution might increase mental stress and then dysmenorrhea occurrs.
Another novel interesting finding is that our results showed synergy effects of PM 2.5 and NO, NO2, NOx, CO on the risks of dysmenorrheal. The risks of dysmenorrhea appeared to be highest in women exposed to high concentrations of NO, NO2, NOx, CO and high PM 2.5 . We supposed that it would be that high concentrations of PM 2.5 and high concentrations of acid gases could provoke the highest oxidative stress as well as systemic inflammation. Further, the risk of dysmenorrhea would be generally higher if women exposed to high concentrations of NO, NO2, NOx, or CO, despite women exposed to low or high concentrations of PM 2.5 . This observational finding warrants attention since it implied that NO, NO2, NOx, or CO would have more harms for risks of dysmenorrheal. We suppose that high concentrations of NO, NO2, NOx, or CO would also have effects on epigenetic modifications (65). Further studies are required to explore the underlying mechanisms.
This study has several limitations. First, the NHIRD does not provide information regarding risk factors for dysmenorrhea, including family history of dysmenorrhea, an early age of menarche (younger than 12 years), nulliparity, heavy menses or irregular menses, smoking, obesity, dietary habits, life stress, use of oral contraceptives, and social network support (20,66). Second, although National Health Insurance covers up to 99% of the population, patients with dysmenorrhea might take overthe-counter medications instead of visiting a doctor. Therefore, the incidences of dysmenorrhea would be underdiagnosed in all quartiles of air pollutants. Therefore, it should be cautious in interpreting results of this study. Third, although we have considered pelvic inflammatory disease, cervical stenosis and polyps, functional ovarian cysts, malignant ovarian tumors, and inflammatory bowel disease, we did not consider adenomyosis, inflammation and scarring of uterus, intracavity or intramural fibroids, functional ovarian cysts, and intrauterine devices, which are also associated with dysmenorrhea (67). Forth, although there is considerable difference between primary and secondary dysmenorrhea, information about hormone measurement, regularity of menses, distribution of body fat, and signs of hirsutism were unavailable in NHIRD. Fifth, although we have used zip codes to define individual exposure to air pollutants, misclassification of exposure measurement should occur in this study since each participant was free mobile. Sixth, considerations for analyzing designs inclusive 1-year exposure average window, IDW methods, as well as GAM for categorical variables should be announced here. Since this is a retrospective study, it would be many uncontrolled variables generating during this exposure window such as those causes of secondary dysmenorrhea. If exposure window was too long, it would be more hardly to convince the association between air pollution and dysmenorrhea even if results are of statistical significance. Thus, although a 1-year exposure average window seems too short for dysmenorrhea investigation, we decided 1-year as exposure average window. The association between air pollution and primary dysmenorrhea would be more convincing on the basis on current design of 1-year exposure average window. The advantage of using IDW method is that IDW is assumed substantially that the rate of correlations and similarities between neighbors is proportional to the distance between them that can be defined as a distance reverse function of every point from neighboring points. However, on an area of 36,193 square kilometers of Taiwan, the density of 78 air quality monitoring stations is not very high. Thus, another limitation of this study would be limited numbers of air pollution monitoring stations. The principal advantage of GAM is its ability to model highly complex non-linear relationships when the number of potential predictors is large. The disadvantage of GAM is that like other non-parametric methods, GAM has a high propensity for overfitting. Finally, this retrospective study was based on analysis of existing database, therefore the level of internal bias of this study would be more compared with those prospective randomized control trial studies. However, our study demonstrated real-world observational results.
In conclusion, our study showed that exposure to high concentrations of the air pollutants SO 2 , NO x , NO, and NO 2 was associated with increased incidences of dysmenorrhea. Whether personal protection against air pollution could help lessen the risk of dysmenorrhea will require further investigations.

DATA AVAILABILITY STATEMENT
The datasets presented in this article are not readily available because The dataset used in this study is held by the Taiwan Ministry of Health and Welfare (MOHW). The Ministry of Health and Welfare must approve our application to access this data. Any researcher interested in accessing this dataset can submit an application form to the Ministry of Health and Welfare requesting access. Please contact the staff of MOHW (Email: stcarolwu@mohw.gov.tw) for further assistance. Taiwan Ministry of Health and Welfare Address: No. 488, Sec. 6, Zhongxiao E. Rd., Nangang Dist., Taipei City 115, Taiwan (R.O.C.). Phone: +886-2-8590-6848. All relevant data are within the paper.
Requests to access the datasets should be directed to Please contact the staff of MOHW (Email: stcarolwu@mohw.gov.tw) for further assistance.

ETHICS STATEMENT
This study was approved to fulfill the condition for exemption by the Institutional Review Board (IRB) of China Medical University (CMUH104-REC2-115-CR4). The IRB also specifically waived the consent requirement. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

AUTHOR CONTRIBUTIONS
S-YL and C-HK: conception/design and collection and/or assembly of data. C-HK: provision of study materials. S-YL, W-HH, C-CL, I-KW, and C-HK: data analysis and interpretation. S-YL, Y-CY, C-CL, CC, W-HH, I-KW, C-DL, C-YH, and C-HK: manuscript writing and final approval of manuscript. All authors contributed to the article and approved the submitted version.  Supplementary Table 1 | Description correlation matrix for air pollutants. * * * correlation is significant at the <0.0001 level (two tails).
Supplementary Table 2 | Baseline urbanization level among quartiles of daily average concentration of air pollutants in Taiwan. P-value using chi-square for the comparisons between urbanization level among quartiles of daily average concentration of air pollutants. The urbanization level was categorized by the population density of the residential area into 4 levels, with level 1 as the most urbanized and level 4 as the last urbanized. The daily average air pollutant concentrations were categorized based on quartiles for each air pollutants.