Does air pollution exposure affect semen quality? Evidence from a systematic review and meta-analysis of 93,996 Chinese men

Background Air pollution may impair male fertility, but it remains controversial whether air pollution affects semen quality until now. Objectives We undertake a meta-analysis to explore potential impacts of six pollutants exposure during the entire window (0–90 days prior to ejaculation) and critical windows (0–9, 10–14, and 70–90 days prior to ejaculation) on semen quality. Methods Seven databases were retrieved for original studies on the effects of six pollutants exposure for 90 days prior to ejaculation on semen quality. The search process does not limit the language and search date. We only included original studies that reported regression coefficients (β) with 95% confidence intervals (CIs). The β and 95% CIs were pooled using the DerSimonian-Laird random effect models. Results PM2.5 exposure was related with decreased total sperm number (10–14 lag days) and total motility (10–14, 70–90, and 0–90 lag days). PM10 exposure was related with reduced total sperm number (70–90 and 0–90 lag days) and total motility (0–90 lag days). NO2 exposure was related with reduced total sperm number (70–90 and 0–90 lag days). SO2 exposure was related with declined total motility (0–9, 10–14, 0–90 lag days) and total sperm number (0–90 lag days). Conclusion Air pollution affects semen quality making it necessary to limit exposure to air pollution for Chinese men. When implementing protective measures, it is necessary to consider the key period of sperm development.

Introduction 8-12% of reproductive-age couples are infertile in the world and its prevalence may be increasing (1). Male factors cause 40-50% of infertile couples (2). Total sperm number, sperm concentration, progressive and total motility are commonly adopted to evaluate male reproductive potential. Sperm quality of sperm donors in China's Henan Province showed a decreasing trend from 2009 to b2019 (3). Although the exact cause remains unclear, air pollution might be a hazard factor for declining semen quality (4).
The growth period of mature sperm is approximately 90 days, including three critical windows: 0-9 days prior to ejaculation (epididymal storage), 10-14 days prior to ejaculation (development of sperm motility), and 70-90 days prior to ejaculation (spermatogenesis) (48). There are fewer studies on which stage of sperm development is most vulnerable to air pollution, but the findings remain controversial (27, 29, 33, 34, 36-40, 42-44, 47). A meta-analysis of relevant research data is needed.
Although there are five systematic review and meta-analyses on whether semen quality is affected by air contaminants (49)(50)(51)(52)(53), the measured indicators of the four systematic review and meta-analyses were the mean differences and the exposure periods were not 90 days (49)(50)(51)(52). The four systematic review and meta-analyses compare semen quality between men exposed to high levels of air pollution and men exposed to low levels of air pollution and were not standardized when merging the effects of air pollution from different studies (49)(50)(51)(52). The main distinction between the reported four meta-analyses and the present work is that we have studied the association air pollution exposure during the whole 90 day period as well as the three critical windows of sperm development. A systematic review and meta-analysis by Xu et al. reported the effect of air pollution exposure during lag 0-90 days or 0-12 weeks on semen quality based on exposure-response relationships but did not report the effect of air pollution exposure during the three critical windows of sperm development (53). The included articles did not include those published in Chinese and those published recently in 2023, and subgroup or sensitivity analyses were also not performed (53). There is still no systematic review on whether semen quality is affected by air pollution exposure during the three critical windows of sperm development.
Therefore, the first meta-analysis was done for analyzing the relation of air pollution exposure during the whole and three critical windows of sperm development and sperm quality in China.

Methods
The present meta-analysis was performed in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (54) as well as PRISMA 2020 checklist had been provided in Supplementary Materials A. This meta-analysis was registered on the PROSPERO website (No. CRD42022374712). Literature search.
We retrieved the Cochrane Library, EMBASE, Web of Science, PubMed, VIP, China National Knowledge Infrastructure (CNKI) as well as Wanfang databases for articles. The search process does not limit the language and search date. Only epidemiological observational studies published in Chinese or English would be included. The applied search words and detailed search strategies are shown in Supplementary Table S1; Supplementary Materials B, respectively. Searches were performed independently by RL and JY Disagreement was resolved by a third author (JL)

Outcomes
Outcomes included total sperm number, sperm concentration, total and progressive motility.

Inclusion and exclusion criteria
Inclusion criteria were: (a) reporting the effect of at least one air pollutant exposure during the whole window and/or critical stages of sperm development on sperm quality; (b) cross-sectional or cohort studies; (c) reporting regression coefficients (β) and 95% confidence intervals (CIs); (d) Chinese males; and (e) English and Chinese articles. The measured indicators of case-control studies were the means and standard deviations (SDs) rather than β and 95% CIs.
The following exclusion criteria were adopted: (a) animal studies, case reports, commentaries, reviews, protocols, editorials, conference abstracts, letters, or book chapters; (b) case-control studies; (c) studies in countries other than China; (d) reported shorter or longer exposure period; (e) focused on indoor air pollution; and (f) multivariate logistic regression.

Study selection
Two authors (RL and JY) conducted the literature selection independently. If any disagreement arose during the selection process, it would be resolved by discussing with the third author (JL).

Data extraction
Using a standardized form, the following information was extracted independently from eligible publications by two authors (RL and JY): publication year, first author, design of study, region, setting, research period, study subjects, size of the sample, pollutants exposure measurement, outcome, exposure period, statistical model, adjusted confounding factors, adjusted β with their corresponding 95% CIs. Through discussion with the third author (JL), any disagreement in the data extraction was resolved. The missing information of the original study was requested by contacting the corresponding author.

Quality assessment
Quality assessments of eligible publications were executed independently by two researchers (QW and LW). If there was any Frontiers in Public Health 03 frontiersin.org inconsistent opinion, it would be resolved by discussing with the third researcher (YD). The Newcastle-Ottawa Scale (NOS) checklist was adopted for evaluating the quality of retrospective as well as prospective cohort studies (55). The Joanna Briggs Institute (JBI) critical appraisal checklist was adopted for evaluating the quality of cross-sectional studies (56). Based on the Grading of Recommendations Assessment, Development and Evaluation (GRADE) guidelines (57), the certainty of evidence was started with moderate and further downgraded based on the following items: publication bias, directness, study limitations, consistency, and precision (58, 59), and upgraded for dose-response gradient, strong effect size as well as plausible confounding effect (60).

Data analyses
If the articles did not give interquartile range (IQR) values or original incremental units of pollutant exposure, we would contact the authors by email. For parts per billion (ppb) units, the following equations were used to convert to μg/ Statistical analyses were conducted with Stata v12.1 (Stata Corp., United States). The β and 95% CIs were combined using the DerSimonian-Laird random effect models. Chi-squared test and I 2 statistics were used to quantify the heterogeneity. Heterogeneity existed when p < 0.05 or I 2 > 50% (66). In order to find sources of heterogeneity, we conducted sub-group analyses based on design of the study (crosssectional and cohort), location (northern and southern China), and exposure assessment approaches (estimating models or monitoring station). Egger's test as well as funnel plots were adopted for assessing publication bias. Stability of the findings was judged with the help of sensitivity analysis. p < 0.05 was statistical significance.

Study characteristics
As depicted in Figure 1, 3,952 publications were retrieved from the seven databases, and 34 articles remained after duplicate literature, abstracts and titles exclusion. After reading the full article, 14 articles were further excluded and detailed exclusion reasons were given in Supplementary Table S2. The remaining 21 eligible publications were eventually included in this meta-analysis. Missing data of original articles were requested by contacting the authors via email or WeChat. Studies with missing information were excluded if multiple contacts with the corresponding author remained unanswered. Table 1 illustrates the primary characteristics of the eligible publications. Table 2 demonstrates the original incremental units, outcomes, statistical models used and adjusted confounding factors of all the eligible studies. If the increment unit of the original study was not

Air pollutants and sperm quality
Six air pollutants exposure during the whole window did not affect sperm concentration (Supplementary Table S4; Figure 2). PM 10 , SO 2 , and NO 2 exposure during the whole window were related with decreased total sperm number, while such association was not found for PM 2.5 , CO, and O 3 exposure (Supplementary Table S4; Figure 2). PM 2.5 , PM 10 as well as SO 2 exposure during the entire window were negatively related with total motility, while such association was not found for other pollutants.
In order to find sources of heterogeneity, we conducted sub-group analyses based on design of the study (cohort and cross-sectional), location (northern China and southern China), and exposure assessment approaches (monitoring station or estimating models). The majority of sub-group results were consistent with the pooled results (Supplementary Table S5; Figure 3).
During 0-9 lag days, only SO 2 exposure was related with declined total motility (Supplementary Table S5; Figure 3). During 10-14 lag days, PM .2.5 exposure was adversely related with total sperm number and total motility, SO 2 with total motility (Supplementary Table S5; Figure 3). During 70-90 lag days, PM 10 and NO 2 exposure were adversely related with total sperm number, PM 2.5 with total motility (Supplementary Table S5; Figure 3).

Sensitivity analysis
In the sensitivity analyses for six pollutants exposure during the whole window and sperm quality, pooled effect sizes did not change significantly by omitting one study from each analysis, thus indicating that our findings were stable (Supplementary Table S4; Supplementary Figure S1). However, when the study of Wu et al. (40) was omitted from sensitivity analyses of PM 2.5 exposure and progressive motility, a significant association disappeared (p = 0.081; Supplementary Table S4; Supplementary Figure S1A). When the study of (34) was omitted from sensitivity analyses of PM 10 exposure and sperm concentration, a significant association disappeared (p = 0.119; Supplementary Table S4; Supplementary Figure S1B). When the study by Ma et al. (33) was omitted from the sensitivity analysis of O 3 exposure and total motility, a significant association disappeared (p = 0.104; Supplementary Table S4; Supplementary Figure S1F).
In the sensitivity analyses of six pollutants exposure during critical windows and sperm quality, the pooled effect sizes did not change significantly by omitting one study from each analysis, thus indicating that our findings were stable. However, when the study of Ma et al. (33) was omitted from the sensitivity analyses of O 3 (70-90 lag days) exposure and total motility, a significant association disappeared (p = 0.197) with heterogeneity decreasing from 51 to 0% (Supplementary Table S5).

Summary of study results
China has a population of more than 1.4 billion and covers a land area of approximately 9.6 million km 2 . Due to the vast territory of China, it varies greatly in climate conditions, landforms, geography, population density, and economic development level in different regions. Based on economic development levels and climatic conditions, China is generally grouped into seven geographic regions (67-69). Detailed geographic location is presented in Supplementary Figure S2. China is roughly classified as southern and northern China (70)(71)(72). Distribution of southern and northern China is shown in Figure 4. As a result of the limited sample size, we performed sub-group analysis by location (northern China and southern China). Air quality is closely related with climatic conditions and economic development levels. Air quality is better in western China than in eastern China (67). Economic development levels in western and eastern regions result in different chemical compositions of pollutants (73,74). In the eastern and central regions, industry and traffic are the primary causes of air pollution (75). Biomass burning and soil dust are the primary reasons of air pollution in the western region. Different sources of air pollution in different regions result in different toxicity, concentrations, and chemical compositions. This may explain, to some extent, the inconsistent results.
Different individual exposure assessment approaches can partially explain the controversial results. Lao et al. estimated individual exposure levels of PM 2.5 using a high-resolution (1 × 1 km) spatiotemporal model (31). Zhou et al. (44) adopted the ordinary Kriging model to measure individual exposure concentrations. Some studies used the land-use random forest model (41) or inverse distance weighting model (28,29,32,34,39,40) to assess the actual individual exposure levels. Some other studies used the averaged levels of the city-wide or the nearest monitoring station to assess actual individual pollutant exposure concentrations (27, 30, 35-38, 42, 43, 45, 46). This is the first meta-analysis to analyze potential impacts of ambient air pollution exposure during the whole window and three critical windows on semen quality in China. Sperm motility, a conventional semen parameter, is one of the common indicators of fertility assessment. Sperm motility is commonly used as one of the most important sperm functions to determine whether female partners can successfully conceive without any assisted reproductive FIGURE 2 Regression coefficients and 95% confidence intervals for the relation between six pollutants exposure during the whole window and sperm quality. Regression coefficients and 95% confidence intervals for the relation between six pollutants exposure during three critical windows and sperm quality.
Frontiers in Public Health 09 frontiersin.org technology (ART). Sperm motility parameters are also sensitive indicators of male reproductive toxicity (76). PM 10 , PM 2.5 as well as SO 2 exposure were adversely related with total motility during 0-90 days prior to ejaculation. PM 2.5 , CO as well as O 3 exposure were adversely related with total sperm number during 0-90 lag days. In order to find sources of heterogeneity, we conducted sub-group analyses based on design of the study (cohort and cross-sectional), location (northern China and southern China), and exposure assessment approaches (monitoring station or estimating models). Although subgroup analysis reduced heterogeneity to some extent, heterogeneity remained high level in some subgroups, and it was necessary to continue to explore potential sources of betweenstudies heterogeneity.
In addition, the possible exposure susceptibility window was also investigated. PM 2.5 exposure affected total motility (10-14 and 70-90 lag days) and total sperm number (10-14 lag days). PM 10 affected total sperm number (70-90 lag days). SO 2 influenced total sperm number (0-9 and 10-14 lag days). NO 2 affected total sperm number (70-90 lag days). The findings suggested that pollutants exposure might affect total motility and total sperm number.

Biological mechanisms
The biological mechanisms that environmental pollutant exposure may damage the development of total motility have not been elucidated. PM 10 , PM 2.5 , and O 3 exposure can lead to elevated concentrations of reactive oxygen species (77,78), which may disrupt the blood-testis barrier, detriment spermatogenesis and result in declined sperm motility (79)(80)(81)(82). PM exposure can also cause systemic inflammatory reactions by elevating tumor necrosis factor (TNF) as well as interleukin-1β (IL-1β) levels (83)(84)(85)(86). Higher concentrations of IL-1β and TNF are related with impaired total sperm motility (87-89). Significant reduction in air pollutants emissions was accompanied by improvements in people's markers of inflammatory conditions, thrombosis as well as oxidation stress (90). We hypothesized that environmental pollutant exposure would elevate oxidative stress levels and inflammatory reactions, which could lead to decreased total sperm motility. This hypothesis requires further toxicological studies to elucidate the detailed mechanism of reduced sperm motility caused by environmental pollutant exposure.

Strengths and limitations
This present meta-analysis has three advantages. First, it is the first meta-analysis to analyze whether semen quality is affected by air pollution exposure during the whole and critical windows. Second, the findings are relatively new as a result of most eligible studies being published within recent 4 years. Third, results of different original studies were difficult to compare since the exposure increment units Distribution of northern and southern China.
Frontiers in Public Health 10 frontiersin.org were different in most cases. Therefore, the comparability of the results was improved by standardizing the data through transformation. However, the present meta-analysis still has four limitations. First, a high degree of heterogeneity for some pollutants was found, which may be explained by differences in pollutant concentrations, types of air pollutants, chemical components of particulate matter, individual exposure assessment approaches, design of the study, study setting, sample size, study regions, selection bias, and adjustment confounding factors. Due to the high degree of heterogeneity, caution should be given when interpreting some pooled effects. A high degree of heterogeneity may also hinder the detection of publication bias. Second, selective bias may occur due to some of the included studies selecting patients from infertility clinics. Third, subgroup analysis by exposure assessment approaches was not performed as a result of the insufficient sample size. Fourth, the sample size is still inadequate, with only 2 articles from northern China being included. Insufficient data might lead to inescapable errors, and the original researches need to be further supplemented. Fifth, many of the included studies obtained estimates of air pollution exposure from ecological data or modeling and did not examine individual exposure to air pollution.

Conclusion
This evidence suggested that ambient air pollution could reduce semen quality in Chinese men and may even lead to infertility. For Chinese men, there is a need to reduce the duration of exposure. Further studies should be conducted to explore the possible biological mechanisms behind the findings observed in this study.

Author contributions
JL and YD proposed the idea and designed the present study, interpreted the findings, and were responsible for statistical analysis and manuscript writing. RL and JY performed literature retrieval, study selection, and data extraction. QW and LW performed the quality assessment. All authors contributed to the article and approved the submitted version.

Funding
The present study was funded by the Henan Provincial Science and Technology Research Project (No. LHGJ20190389).