While the precise mechanisms behind air pollution-related female infertility are not yet fully understood, we propose a potential mechanism: maternal exposure to contaminants during the pre- and peri-conceptional periods could lead to abnormalities in oocytes, embryos, and fetuses [22]. This could potentially hinder safe childbirth, impact overall health and mental well-being, and thereby increase vulnerability to and susceptibility to infertility. Air pollutants disrupt both animal and human gametogenesis, reducing reproductive efficacy [8]. These endocrine-disrupting properties could activate the Ras/Erk pathway via interactions with nuclear receptors like estrogen or androgen receptors and specific cytosolic targets, while diesel exhaust particles and PM_2.5 impact ovarian function by disrupting the endocrine system, increasing oxidative stress and inflammation, and activating specific targets that trigger abnormal MAPK signaling [23]. Indeed, in a recent study of 777 men, increased air pollutant concentrations were significantly associated with both the hypomethylation of F3, ICAM-1, and TLR-2, and the hypermethylation of IFN-γ and IL-6 [24]. What's most notable is that research has demonstrated that adding antioxidant factors to ovarian stimulation can enhance reproductive outcomes in women with polycystic ovarian syndrome [25].

Regarding ovarian function, a study did find a significant negative association between SO₂ exposure throughout the entire antral follicle development stage and AFC [26]. This suggests that during the follicular vascularization transition stage, SO₂ might be expedited in reaching follicles and granulosa cells, potentially contributing to impaired follicle development, hastened follicle loss, or altered ovarian function. The ovarian toxicity of environmental contaminants, characterized by meiotic disruption in oocyte formation resulting from genetic or epigenetic aberrations, can affect dormant follicles within the primordial pool that undergo growth stimulation and have the potential to develop into the antral stage [27]. Subsequently, these antral candidates are typically chosen to progress to the pre-ovulatory stage, marked by the transition from primary to pre-antral, during which the oocyte within the follicle enlarges, granulosa cells proliferate, and a layer of theca cells forms outside the granulosa cells, serving as a conduit for transporting steroids and growth factors to the granulosa cells to promote follicle vascularization [28]. Additionally, this bloodstream enables toxins to reach follicles, where they can suppress granulosa cell proliferation and growth. Furthermore, SO₂ toxicology in mammalian organ development involves cellular injuries caused by inflammation and oxidative stress, such as lipid peroxidation, DNA damage, protein oxidation, and interference with signal transduction [29].

While ozone (O₃) is traditionally considered a harmful air pollutant, emerging evidence suggests that under certain controlled clinical settings with specific dosing protocols, ozone therapy may exert anti-inflammatory and antioxidant effects. For instance, medical ozone therapy administered in controlled healthcare environments has shown potential benefits in conditions such as HIV/AIDS, fibromyalgia, and peripheral arterial disease [30–32]. Some studies have explored its use in reproductive medicine, including pelvic inflammatory disease and tubal infertility [31]. It is crucial to emphasize that these therapeutic effects are achieved through controlled medical administration and are not equivalent to ambient environmental ozone exposure, which remains harmful to human health. However, the negative association observed in our study between ozone exposure and infertility risk is counterintuitive and must be interpreted with caution. Ambient ozone levels are influenced by seasonal and geographic factors and may act as a proxy for lower traffic-related co-pollutants. Additionally, negative confounding or interaction with other pollutants cannot be ruled out. Further research is needed to clarify this unexpected finding. This negative trend may reflect negative confounding, as ozone concentrations are often higher in suburban areas with lower levels of traffic-related pollutants. It is also possible that ecological bias affected the results, given that area-level monitoring data may not precisely represent individual-level exposure. These findings highlight the need for future studies to incorporate more detailed exposure metrics and consider pollutant interactions to better understand the role of ozone in reproductive health.

This study possesses several notable strengths. Firstly, it is the inaugural investigation to reveal the association between long-term exposure to air pollution and female infertility. Secondly, our analysis encompassed a comprehensive range of pollutants and utilized a significantly large sample, bolstering the statistical robustness of the findings. Thirdly, we conducted stratified analyses considering potential confounding factors such as age, urbanization level, insurance amount, CCI score, ambient temperature, season, and lag0-2, along with a wide spectrum of demographic characteristics. Furthermore, in registry-based studies, a minimum follow-up threshold is often necessary to ensure valid exposure-outcome relationships. Our choice of a 5-year minimum follow-up reflects this principle and strengthens the internal validity of the analysis. While it may reduce inclusion of short-term registrants, the large sample size and nationwide coverage help preserve overall representativeness. Lastly, the population-based data used in this study is representative of the general population in Taiwan.

This study has several limitations. First, key reproductive and lifestyle factors such as smoking status, alcohol consumption, body mass index (BMI), marital status, and oral contraceptive use were not available in the NHIRD. Although proxy variables—such as comorbidities, insurance level, and urbanization—were included in the adjusted models, the potential for residual confounding remains. Second, clinical variables such as disease severity, laboratory data, and symptomatology were unavailable, limiting the granularity of outcome characterization. Third, both exposure and outcome misclassification are possible. Exposure was estimated based on insurance-registered residential postal codes, which may not reflect actual pollutant exposure due to mobility or address reporting errors. Although we incorporated updated residential records to reduce spatial misclassification, some inaccuracy may persist. Outcome misclassification may also occur, particularly in defining infertility, as the database does not capture fertility intention, menopausal status, male partner fertility status, or utilization of assisted reproductive technologies—all of which may influence diagnostic classification. Additionally, death was inferred from withdrawal from the National Health Insurance program, which may not be completely accurate. Fourth, the retrospective nature of the study limits causal inference, and unmeasured or unknown confounders cannot be excluded. Finally, while the use of a five-year follow-up period strengthens the exposure-outcome temporality, it may inadvertently exclude short-term insurance registrants, potentially introducing selection bias. Nonetheless, the large sample size and nationwide coverage enhance the generalizability of the findings. Despite these limitations, the results provide consistent epidemiologic evidence supporting the association between long-term exposure to SO₂, CO, PM₁₀, PM₂.₅, NOₓ, NO, NO₂, THC, NMHC, and CH₄ and increased risk of female infertility.