infertility: a cross-sectional study

5 a Department of Obstetrics and Gynecology, CHA Fertility Center Seoul Station, CHA 6 University School of Medicine, Seoul, South Korea 7 b Department of Preventive Medicine, Korea University Medical College, Seoul, South Korea 8 c Graduate School of Public Health, Seoul National University, Seoul, South Korea 9 d Department of Geography, Korea University, Seoul, South Korea 10 e Department of Cancer Control and Population Health, Graduate School of Cancer Science 11 and Policy, National Cancer Center, Gyunggi-do, South Korea 12 f Department of Environmental Health, Boston University School of Public Health, Boston, 13 MA, United States 14 g Department of Urology, CHA Fertility Center Seoul Station, CHA University School of 15 Medicine, Seoul, South Korea 16


50
Growing evidence suggests potential impacts of the outdoor environment on human 51 health. It has been suggested that components of built and the natural environment may 52 influence levels of psychological stress, physical activity, and social relationships; and 53 thereby, potentially improve or worsen human health and wellbeing [1][2][3]. For example, 54 neighborhood green space has been associated with many beneficial health effects, 55 including reduced all-cause and cardiovascular mortality and improved mental health, 56 possibly mediated by less air pollution, heat and stress, and increased physical activity and 57 social contacts [4]. 58 Male reproductive function is highly sensitive to various physical agents generated 59 by industrial activities [5,6]. In addition, semen quality itself reflects general health condition, 60 since it is affected during the early stage of medical disorders [7,8]. Therefore, assessing 61 the relationship between residential environment and semen quality would expand our 62 understanding of the potential role of environmental factors in human reproductive health. 63 Prior studies found that exposure to ubiquitous chemicals including endocrine disruptive 64 chemicals and air pollutants is associated with reduced semen quality [9][10][11]. Given the 65 association of physical environment with human fertility, male reproductive potential 66 represented by semen quality may be associated with features of the built environment. In 67 this study, we aimed to assess the association between the residential built environment and 68 six parameters indicative of semen quality among men with a history of infertility.  The Korean peninsula is mainly mountainous along its east coast, most of its river water 89 flows west, and highly populated towns are located mostly in the north-west region. Four 90 built environment components commonly used in prior studies were measured: distance to 91 fresh water, distance to the coast, distance to major roadways, and Normalized Difference 92 Vegetation Index (NDVI) [12][13][14][15]. We used distance to the nearest major roadway since it is 93 often used as a proxy for long-term residential levels of air and/or noise pollution due to  Fig. 1), as in previous studies [15,17]. The distance from the geocoded 97 address to the environmental variables was calculated using Arcmap's Spatial Join analytical 98 tool, which analyzes the spatial relationship between two geographical features. We defined 99 the distance between any two features as the shortest separation between them, such that

103
River and lake data were integrated into data on fresh water. Both data sets were 104 retrieved from the National Spatial Data Infrastructure (http://www.nsdi.go.kr). River data was 105 retrieved on January 21, 2016, and lake data on July 5, 2019. Integrated data is a  The original road data set was compiled on September 20, 2019, and was classified into 114 nine categories: national highways, metropolitan city highways, general national roads, 115 metropolitan city roads, government-financed provincial roads, provincial roads, district 116 roads, highway link lamps, and other roads. In this study, we defined major roads as national 117 highways, metropolitan city highways, metropolitan city roads, highway link lamps, and roads 118 more than six lanes wide in other classes.

119
For data on NDVI, we used Landsat 7 satellite data provided by the United States

167
The mean age of the study population was 39 years (Table 1)  The mean value of progressive sperm motility was different across quartiles of 185 distance to fresh water and a major roadway (Supplementary Table 1 were not evident except for NDVI within 500 m (Table 2). An IQR-increase in NDVI (0.1) was 192 associated with 0.05-increase in z-score of vitality (95% confidence interval (CI): 0.01, 0.09).

193
In the analyses using quartiles of exposures, living at the maximum distance to fresh water 194 (i.e., in the 4 th quartile) was generally associated with lower semen quality, but this did not   Coefficients are calculated using a multivariable linear regression model including four built environmental components, age, body mass index, occupation, smoking, season of semen test, and administrative district of home address.

206
We did not find a consistent association between built environment and semen quality among men with a history of infertility. In 207 single-and multi-exposure model, we observed that a higher value for NDVI within 500 m was positively associated with percentage of 208 sperm vitality. The observed associations between environmental components and semen quality indicators were generally non-linear. For 209 example, distance to fresh water was associated with lower percentage of progressive motility upon comparison of the first two quartiles. The 210 2 nd quartile of NDVI was associated with higher total motile sperm count compared to the 1 st quartile. To the best of our knowledge, this is the 211 first report to assess the association between land use and semen quality using large hospital-based data. with sperm motility and a positive association of remoteness to roadway and NDVI with vitality.

222
The results of this study need to be interpreted with caution. First, as a single fertility center study, our study population was mostly 223 restricted to white collar workers living in an urban area. Second, misclassification of exposure may have potentially occurred due to the use 224 of residential address for exposure assessment, or due to the distance between the home address and the workplace, where patients may 225 have spent a substantial amount of time. However, assuming that the misclassification was non-differential, it may have biased our results 226 towards the null [28]. We believe our study may still have important implications due to the use of hospital data belonging to a large infertile 227 population who is expected to be particularly vulnerable to environmental exposure.