Association Between Ambient Fine Particulate Matter and Maternal Vitamin D Levels During The Second Trimester of Pregnancy

Background: Evidence for an association between the amount of ambient ne particulate matter (PM) in the atmosphere and vitamin D status of pregnant women is limited. We aimed to examine the independent association between ambient ne PM and maternal levels of serum 25-hydroxyvitamin D (25OHD) during the second trimester and to explore possible modications to the association by meteorological factors. Methods: 27,768 pregnant women presenting for prenatal examination who were tested for serum 25OHD concentration during the second trimester between January 1, 2016, and 31 December, 2020, were included in this retrospective analysis. Exposure to PM was evaluated based on daily average PM with an aerodynamic diameter of ≤ 2.5 μm (PM 2.5 ) and PM with an aerodynamic diameter of ≤ 10 μm (PM 10 ). Corresponding meteorological data for daily average atmospheric temperature, atmospheric pressure, relative humidity, sunshine duration, and wind speed, were collected. Results: Vitamin D deciency (<12 ng/mL), inadequacy (12-20 ng/mL), and adequacy ( ≥ 20 ng/mL) were respectively present in 23.5%, 41.3%, and 35.2% of pregnant women during their second trimester. The maximum cumulative effects of PM 2.5 occurred at lag 45 days, and the maximum cumulative effects of PM 10 occurred at lag 60 days. In crude models, 45-day moving daily average PM 2.5 concentrations were negatively associated with 25OHD levels (β, −0.200; 95% CI, −0.206 to −0.193), as were 60-day moving daily average PM 10 concentrations (β, −0.142; 95% CI, −0.146 to −0.137). After adjusting for temporal and meteorological factors, the effect values were drastically reduced (adjusted β of PM 2.5 , −0.032; 95% CI, −0.046 to −0.018; adjusted β of PM 10 , −0.039; 95% CI, −0.049 to −0.028). Conclusions: Our study showed an independent negative association between ambient ne PM in the atmosphere and maternal serum 25OHD levels during the second trimester of pregnancy. However, the effect values were small after adjusting for temporal and/or meteorological factors, which indicates that ambient ne PM may have a limited inuence on maternal serum 25OHD levels.


Introduction
The key fat-soluble nutrient vitamin D has multiple functions, and its de ciency is thought to be a risk factor in skeletal health and various non-skeletal conditions, such as schizophrenia, skin disorders, certain types of cancer, type 2 diabetes, and infections [1]. For pregnant women, su cient vitamin D stores need to be maintained to provide for both themselves and their fetus [2]. Research continues to provide evidence of an association between maternal Vitamin D de ciency (VDD) and a higher risk of numerous undesirable pregnancy outcomes, such as a low body weight, respiratory tract infections, and neurocognitive developmental problems in newborns and preeclampsia and high blood pressure in mothers [3][4][5][6][7][8]. Therefore, VDD is a key clinical issue for pregnant women and deserves research attention. Throughout the world at present, VDD during pregnancy is still a frequent occurrence. In one study of pregnant women in Finland, 77.4% were found to suffer from VDD during their rst trimester [9]. While in, the USA, Ginde et al. [10] reported 46%, 32%, and 18% of women with VDD in the rst, second, and third trimesters, respectively, and there were comparable observations among British [11], Indian [12], and Chinese [13] women.
For the most part, factors determining the vitamin D status during pregnancy do not differ from those important for the general population, such as the number of melanocytes in the skin, the amount of exposure of the skin to the sun, the amount of adipose tissue in the body, geographical latitude, dietary factors, and intake of supplementary vitamins [14][15][16][17][18]. An indoor lifestyle can also increase the risk of developing a VDD [19]. Moreover, the amount of PM in the atmosphere has been associated with circulating levels of 25-hydroxyvitamin D (25OHD), which is a key marker for vitamin D status, in humans [20][21][22][23], including pregnant women [24].
It is not yet clear what mechanisms coordinate the links between maternal vitamin D and air pollution levels. Because vitamin D is a relatively scarce resource in food, humans generally rely on the vitamin D produced by cells in the skin on exposure to natural solar ultraviolet (UV) B radiation [25,26]. Some scholars have hypothesized that the reduction in the amount of solar UVB radiation reaching ground level by ambient ne particulate matter (PM) is the main mechanism linking PM to circulating 25OHD levels in humans [23,24]. In an investigation into the relationship between certain pollutants and ground-level UVB intensity, PM with an aerodynamic diameter of ≤10 μm (PM 10 ) was shown to be a signi cant negative predictor of solar UVB radiation, but the impact of PM 10 was very small [27]. In addition, vitamin D status correlated with temporal and meteorological factors. During their second trimester, the 25OHD levels of women in eastern China demonstrated variation with season and air temperature [28]. Another study reported 25OHD levels were positively correlated with the prior month's temperature [29]. Thus, when exploring the independent association between ambient ne PM and vitamin D status, meteorological factors may need to be taken into account.
Thus, our study objective was to examine the independent association between ambient ne PM and maternal serum 25OHD levels during the second trimester in pregnant women and to explore if meteorological factors have an impact on this association.

Study Population
We designed a single-hospital-based cross-sectional retrospective observational study to test our hypotheses. All prenatal examination data were collected from the hospital medical database in A liated Kunshan Hospital of Jiangsu University located in Kunshan, eastern China, at 31.2°N latitude and approximately 30 kilometers from Shanghai. This is the largest hospital that provides prenatal care in all the districts of Kunshan. Consecutively, 35,476 pregnant women at the 15-20th week gestation (during the second trimester) who visited our institution for prenatal examination and were tested for serum 25OHD concentrations between January 1, 2016, and 31 December, 2020, were included in this study. We excluded 7,708 women for the following reasons: those (1) with a chronic metabolic disease with consequences for the metabolism of vitamin D (n = 5,389); (2) with a high-risk pregnancy (n = 1,087), or (3) who had lived in Kunshan for less than 1 year (n = 1,232). Finally, 27,768 pregnant women were included in this analysis (Additional le 1: Figure S1).
The study protocol, submitted for review by the ethical committee at the A liated Kunshan Hospital of Jiangsu University (approval No. 2020-03-046-K01), was approved, and it complied with the Declaration of Helsinki. Patient information was initially documented for hospital's quality improvement purposes. Data analyzers were blinded to the identity of patients. Written informed consent was not required due to the anonymous data gathering and observational nature of the study.

Measurement of Vitamin D
In humans, the most abundant form of vitamin D in the blood is 25OHD, and serum levels are used to reliably estimate a patient's vitamin D status. All pregnant women in this study were in their 15-20th week of gestation (during the second trimester of pregnancy) and were asked to fast before blood samples were taken. Serum 25OHD concentrations were measured immediately using an automated electrochemiluminescence immunoassay on a Roche Cobas 8000/e602 analyzer (Roche Diagnostics, Mannheim, Germany). There is no medical consensus regarding the status cut-off values of 25OHD concentrations. However, the Institute of Medicine (National Academy of Sciences, Washington, DC, USA) and the National Osteoporosis Society (Bath, England) concur that, for bone health, a serum 25OHD level (2.5 nmol/L 25OHD = 1 ng/mL 25OHD) of < 30 nmol/L (12 ng/mL) is de cient, 30-50 nmol/l (12-20 ng/mL) is insu cient for certain individuals, while >50 nmol/L (20 ng/mL) is su cient for most people [30,31]. The time of blood collection was included in the analysis. Seasons were de ned as: Spring, March-May; Summer, June-August; Autumn, September-November; Winter, December-February.

PM Exposure and Meteorological Data Assessment
We received data for the daily average concentration of PM with an aerodynamic diameter of ≤2.5 μm (PM 2.5 ; μg/m 3 ) and PM 10 [32].
Because the serum 25OHD half-life is about 15 days [33], the cumulative effects of PM 2.5 and PM 10 on maternal serum 25OHD levels during the second trimester of pregnancy were approximated using the moving-average lag structures (blood collection day [day 0] was not included) as follows: 0-3 days (3-day moving daily average PM concentration), 0-7 days (7-day moving daily average PM concentration), 0-15 days (15-day moving daily average PM concentration), 0-30 days (30-day moving daily average PM concentration), 0-45 days (45-day moving daily average PM concentration), 0-60 days (60-day moving daily average PM concentration), 0-75 days (75-day moving daily average PM concentration), and 0-90 days (90-day moving daily average PM concentration). When adjusting for the in uence of meteorological factors, the corresponding moving average lag structures for the meteorological factors were calculated.

Statistics
The summary statistics for the characteristics of pregnant women were expressed as frequencies (proportions) for categorical variables and as the means (standard deviation [SD]) or medians (Q1-Q3) for continuous variables. We also conducted univariate logistic regression analysis to evaluate the association between the characteristics of pregnant women and maternal serum 25OHD concentrations during the second trimester. The daily average meteorological variables and daily average PM 2.5 , PM 10 , and 25OHD concentrations were screened for correlations by Pearson's.
Using generalized estimating equations (GEE), we evaluated the relationships between the cumulative effects of PM on maternal serum 25OHD levels. The cumulative effects of PM were divided into four levels on the basis of quartiles. After adjusting for year and age at blood collection (adjust I) or year, season, and age at blood collection (adjust II), β-values (95% con dence interval [CI]) of the maternal serum 25OHD levels were calculated based on group trends. As part of the sensitivity analysis, we also calculated the β-values (95% CI) for the 25OHD levels when using the concentrations of PM 2.5 and PM 10 as continuous variables.
Controlling for the in uence of meteorological covariances by applying multivariate linear regression analysis allowed us to evaluate independent associations between the maximum cumulative effects of PM and maternal serum 25OHD. We calculated the results of the unadjusted or minimally adjusted analysis and those from fully adjusted analysis. First, collinearity diagnosis of covariances was performed using variance in ation factor (VIF) analysis (the variable average atmospheric temperature was rst removed due to VIF >10). Then, a judgement on whether to adjust covariances was made using the following principles: Criteria 1, the covariate is added to the basic model or removed from the full model and the matched odds ratio (OR) is changed by at least 10%; Criteria 2: Criteria 1 or a covariate Pvalue of <0.1 in the univariate model [34]. As part of the sensitivity analysis, we transformed 25OHD quantitative variables into dichotomous qualitative variables (1 = vitamin D de ciency and inadequacy [<20 ng/mL]; 0 = adequacy [≥20 ng/mL]), then the OR and 95% CI for maternal vitamin D de ciency and inadequacy (<20 ng/mL) associated with a 10 mg/m 3 increase in PM 2.5 or PM 10 was determined.
Non-linear relationships were additionally identi ed via a generalized additive model (GAM), and on nding a non-linear correlation, the threshold effect in terms of the smoothing curve was calculated using a two-piecewise linear regression model. When a clear ratio was apparent in the smoothing curve, the recursive method was applied to automatically calculate the turning point at which to use the maximum likelihood model [35]. In addition, to test the robustness and potential variation in the different subgroups, we repeated the subgroup analyses while stratifying by season, age, and meteorological factors. The age threshold was derived from the turning point calculated by the GAM followed by an inspection of the modi cation and interaction of the subgroups with the likelihood ratio test.
All statistical analyses were performed using the Empower Stats (www.empowerstats.com, X&Y solutions, Inc., Boston, MA, USA) and R software version 3.6.3 (http://www.r-project.org). A P-value <0.05 was set as the signi cance threshold. Table 1 displays the data on the pregnant women and their vitamin D statuses. This analysis included 27,768 individuals with a mean age of 28.86 (SD, 4.27) years. The mean (SD) and median (Q1-Q3) values of maternal serum 25OHD concentrations during the second trimester of pregnancy were 17.7 (7.9) ng/mL and 16.0 (12.0-22.0) ng/mL, respectively. Vitamin D de ciency, inadequacy, and adequacy were present in 23.5%, 41.3%, and 35.2% of women, respectively. Univariate logistic regression analysis revealed that maternal 25OHD concentration was positively associated with maternal age and showed seasonal variation, with the peak in September and the nadir in February.

Results
Pearson's correlation analysis was conducted to compare serum 25OHD concentration, meteorological variables, and air pollutant exposure. Figure S2 (Additional le 1) presents that, except PM 2.5 vs. sunshine duration (P-value = 0.821), all correlations among the variables were statistically signi cant (P-value < 0.001). There was a strong positive correlation between the daily average PM 2.5 concentration and PM 10 concentration, and the Pearson coe cient was 0.924. There was also a strong negative correlation between daily average atmospheric temperature and atmospheric pressure, and the Pearson coe cient was −0.892. There was a moderate negative correlation between daily sunshine duration and relative humidity, and the correlation coe cient was −0.683. Figure 1 shows periodic changes in the above indicators in terms of monthly average values over time. It can be seen that the variation in the average temperature of the last month was most consistent with the periodic variation of monthly average serum 25OHD concentrations. Similar periodic changes were seen for sunshine duration and relative humidity. However, monthly average PM 2.5 and PM 10 concentrations and atmospheric pressure showed periodic changes that were diametrically opposite to those of the monthly average serum 25OHD concentration. Figure 2 and Table S1 (Additional le 1) show the cumulative effects of PM 2.5 and PM 10 on maternal serum 25OHD levels during the second trimester of pregnancy. After we adjusted for year, season, and age at blood collection, the maximum cumulative effects occurred at lag 0-45 days of PM 2.5 and lag 0-60 days of PM 10 . When PM concentrations were used as continues variables, similar results were observed, and these are provided in Table S2 (Additional le 1).
The independent associations between the maximum cumulative effects of PM and serum 25OHD levels were further investigated, and concentration and the women's 25OHD levels (β, −0.032; 95% CI, −0.046 to −0.018; P-value <0.00001), and the 60-day moving daily average PM 10 concentration was negatively associated with 25OHD levels (β, −0.039; 95% CI, −0.049 to −0.028; P-value <0.00001). As part of the sensitivity analysis, the crude and adjusted OR for PM exposure's association with maternal vitamin D de ciency and inadequacy (<20 ng/mL) were determined, which are given in Table S3 (Additional le 1). We observed greater odds of maternal vitamin D de ciency and inadequacy (<20 ng/mL) with higher PM levels. In the fully adjusted Model 4, an increase in the PM 2.5 45-day moving daily average by 10 mg/m 3 was associated with a 12.4% (OR, 1.124; 95% CI, 1.071 to 1.180, P-value <0.00001) increase in the odds for maternal vitamin D de ciency and inadequacy (<20 ng/mL). In the fully adjusted Model 4, a 10 mg/m 3 increase in the PM 10 60-day moving daily average was associated with a 11.3% (OR, 1.113; 95% CI, 1.074 to 1.153, Pvalue <0.00001) increase in the odds of maternal vitamin D de ciency and inadequacy (<20 ng/mL).
In the subgroup analyses strati ed by season, we further investigated the role of season on the association between the maximum cumulative effects of PM and serum 25OHD levels. For PM 2.5 (Table   3), both linear and nonlinear effect values were higher in summer and autumn and lower in winter and spring (P-values for interaction <0.001). Figure 3A shows the different nonlinear associations between the 45-day moving daily average PM 2.5 concentration and maternal serum 25OHD levels strati ed by season. In addition, we calculated, using the two-piecewise linear regression model, the turning point of the adjusted smoothed curve. Speci cally, the difference between the two slopes was at its maximum in autumn, and the turning point (45-day moving daily average PM 2.5 concentration) was 31.11 μg/m 3 . For PM 10 (Table 3), linear regression analysis showed a signi cant interaction for season (interaction P-value <0.001), but the non-linear model showed this interaction was not signi cant (interaction P-value = 0.383). Figure 3B shows the different nonlinear associations between the 60-day moving daily average PM 10 concentration and maternal serum 25OHD levels strati ed by season. It is worth mentioning that, in autumn, the relationship between PM 10 and 25OHD levels was nonlinear, and the turning point was 38.10 μg/m 3 . Speci cally, when the 60-day moving daily average PM 10 concentration ranged from 32.32 μg/m 3 to 38.10 μg/m 3 , a stronger negative relationship was found between PM 10 and serum 25OHD levels (β, −1.121; 95% CI, −1.513 to −0.728; P-value <0.0001; number of pregnant women, 878).
The threshold effect analysis used to examine the associations between maternal age and serum 25OHD levels during the second trimester of pregnancy revealed a stronger positive relationship between age and serum 25OHD levels when the women were 15 to 25 years of age and a weaker positive relationship when they were 25 to 47. The results of four different adjusted models were robust (Additional le 1: Figure S3 and Table S4). We then categorized the pregnant women using a threshold of 25 years according to the results of the threshold effect analysis and further investigated the modi cation effect of age on the association between the maximum cumulative effects of PM and serum 25OHD levels. Table S5 (Additional le 1)shows that the interaction between age and PM 2.5 was not signi cant (linear interaction P-value = 0.083 and non-linear interaction P-value = 0.058), while Table S6 (Additional le 1) shows the interaction between age and PM 10 had a marginally signi cant effect on serum 25OHD levels (linear P-value for interaction = 0.045 and non-linear P-value for interaction = 0.038).
In the subgroup analyses strati ed by meteorological factors, Table 4 showed the associations between PM and serum 25OHD levels during the second trimester of pregnancy were generally modi ed by meteorological factors, although wind speed had no modi cation effect on PM 2.5 . Of particular interest, there was a positive association between PM concentration and 25OHD levels under low relative humidity. The result of this strati cation was the opposite to the nal conclusion.

Discussion
Based on our knowledge of the literature, this epidemiological study is the rst to focus on the independent relationship between PM and maternal serum 25OHD levels during the second trimester of pregnancy after adjusting for meteorological factors. We found vitamin D de ciency, inadequacy, and adequacy in 23.5%, 41.3%, and 35.2% of pregnant women during the second trimester of pregnancy, respectively. Prenatal exposure to higher levels of PM was associated with decreased maternal serum 25OHD concentrations. We also found that the maximum cumulative effects of PM 2.5 occurred at lag 45 days and the maximum cumulative effects of PM 10 occurred at lag 60 days. However, the effect values were drastically reduced after adjusting for temporal and/or meteorological factors. The results indicated that ambient ne PM has a limited in uence on maternal serum 25OHD levels.
Our conclusions are similar to those of Zhao et al., who reported an association between prenatal exposure to higher PM 2.5 and PM 10 levels and a decrease in circulating 25OHD concentrations in women in later stages of pregnancy [24]. In contrast to the analysis they conducted, we rst investigated the time window for the largest cumulative effect of PM at the individual level and then adjusted for temporal and meteorological factors. Although both effect sizes of PM 2.5 and PM 10 in the fully adjusted models were small (β-value of PM 2.5 = −0.032; β-value of PM 10 = −0.039, respectively), they were highly signi cant (Pvalue < 0.00001). The β-values showed two large drops, one after adjusting for season and the other after adjusting for atmospheric pressure. Due to the collinearity of atmospheric pressure and temperature, it can be inferred that season and atmospheric pressure/temperature may have been important confounders in the regression model. Thus, in some correlation analyses of PM and vitamin D levels, the independent role of PM may have been overestimated when there was no adjustment for meteorological factors.
Periodic changes in the seasons affect periodic changes in meteorological factors, such as temperature, humidity, and sunshine duration. Hence, unlike other studies that only adjusted for season, our study precisely adjusted for meteorological factor exposure at the individual level. . Thus, because PM can absorb and diffuse solar irradiation, some scholars have hypothesized that PM can indirectly reduce vitamin D formation by reducing UVB exposure [23,24]. However, other scholars have argued that there is signi cant spatiotemporal variation in the morphology, chemical makeup, density of PM, and it is di cult to determine its effect on UV radiation [48]. More recently, PM 10 was found to be a signi cant negative predictor of solar UVB radiation, but the effect of PM 10 was miniscule [27]. This conclusion can be used to further deduce that PM 10 has an effect on vitamin D levels, but the effect is small, which is consist with the results of our study. Combined with the above views, we propose a hypothesis (see Additional le 1: Figure S4) that PM and meteorological factors indirectly in uence the cyclical changes of vitamin D in pregnant women by impacting the level of personal UVB exposure. For example, in winter, when the temperature is lower and there are fewer hours of sunshine, the PM concentration is higher, but pregnant women spend less time outdoors and wear more clothes, which lead to lower solar UVB radiation exposure, reducing the synthesis of vitamin D.
In the subgroup analysis, with the exception of the low relative humidity, a negative effect between PM and 25OHD was evident in all subgroups considered and after careful adjustments. In the interaction analysis, season and most meteorological factors interacted with the association between PM and 25OHD. However, the mechanisms of their interactions were unclear and need to be further investigated in the future.
There were some limitations regarding our study. First, our results were obtained from a Chinese population of pregnant women during the second trimester and cannot be extrapolated to other populations. Second, the study was an analytical retrospective study and hence provides limited evidence that PM exposure and vitamin D outcomes were related, and the difference between cause and effect is uncertain. Third, demographic information was lacking (education level, vitamin D/calcium supplement use, outdoor activities, use of sun protection, and BMI, etc.) for the individuals whose test results were used. However, valuable insights can be gleaned from the study, as it involved the retrospective analysis of a large dataset from a prenatal examination population. We used precise adjustments at the individual level to reveal whether PM levels are independently associated with 25OHD concentration.

Conclusion
Our study revealed an independent negative association between ambient ne PM and maternal serum 25OHD levels during the second trimester of pregnancy. However, the effect values were drastically reduced after adjusting for temporal and/or meteorological factors, which indicates that ambient ne PM may have a limited in uence on maternal serum 25OHD levels.

Competing interests
The authors declare that they have no competing interests.

Ethics approval
This study complies with the Declaration of Helsinki and has been approved by the Ethics Committee of the First People's Hospital of Kunshan (no. 2020-03-046-K01).

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Authors' contributions
Ke Lu contributed to the study conception and design. Ya-qin Gong, Yun-yu Xia, Xiao-chun Wang, Lin Chen and Shan-jun Yan contributed to the acquisition of data. Ke Lu and Ya-qin Gong contributed to the analysis and interpretation of data. Chong Li drafted the manuscript, and Ke Lu revised it. All authors critically revised the manuscript, agree to be fully accountable for ensuring the integrity and accuracy of the work, and read and approved the nal manuscript. Tables   Adjusted for year, season, 45-day moving daily average atmospheric pressure, sunshine duration, relative humidity and wind speed except the subgroup variable.
Abbreviations: PM 2.5 , particulate matter with an aerodynamic diameter of ≤2.5 μm; 25OHD, 25-hydroxy vitamin D; CI, con dence interval. Adjusted for year, season, 60-day moving daily average atmospheric pressure, sunshine duration, relative humidity and wind speed except the subgroup variable.
Abbreviations: PM 10 , particulate matter with an aerodynamic diameter of ≤10 μm; 25OHD, 25-hydroxy vitamin D; CI, con dence interval.  Association among exposure to PM 2.5 (A) or PM 10 (B) and maternal serum 25OHD levels during pregnancy. The cumulative effects of PM were divided into four levels based on quartiles. After adjusting for year and age at blood collection (adjust I) or year, season, and age at blood collection (adjust II), the βvalues (95% CI) of maternal serum 25OHD levels during pregnancy were calculated based on the group trend. The dashed vertical line represents lag days corresponding to the maximum cumulative effects.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.