Associations of mixed exposure to ambient air pollution and environmental factors with the risk of miscarriage in women undergoing assisted reproductive technology: a retrospective cohort study.

In this study, we evaluated 2654 participants who underwent fresh embryo transfer. The mean age of the participants who experienced miscarriage was 31.99 ± 3.91 years, and the average BMI was 21.79 ± 3.09 kg/m 2 (Table  1 ). Table 1 Characteristics of 2654 participants who underwent fresh embryo transfer according to miscarriage status (mean ± SD a or N (%)) Maternal Characteristics Overall ( n  = 2654) Miscarriage ( n  = 306) Female age (years) 30.95 ± 3.63 31.99 ± 3.91 BMI b 21.48 ± 2.84 21.79 ± 3.09 Duration of infertility (years) 3.51 ± 2.34 3.59 ± 2.43 Stimulation protocol (%) GnRH agonist 2281 (86.0) 248 (81.0) GnRH Antagonist 360 (13.6) 56 (18.3) Mild stimulation 11 (0.4) 2 (0.7) Natural cycle 1 (0.0) 0 (0.0) Dosage of gonadotropic (U) 2637.45 ± 885.08 2649.97 ± 868.59 Duration of gonadotropic use (days) 12.26 ± 2.13 11.95 ± 2.11 Endometrial thickness (mm) 11.69 ± 2.98 11.37 ± 1.73 Fertilization method (%) ICSI 531 (20.0) 56 (18.3) IVF 2122 (80.0) 250 (81.7) Number of oocytes retrieved 10.51 ± 4.85 9.94 ± 4.87 Availability of high-quality embryos 1.91 ± 0.33 1.90 ± 0.32 Number of embryos transferred 1.80 ± 0.40 1.72 ± 0.45 Type of embryos transferred (%) Blastosphere 113 (4.3) 15 (4.9) Embryo 2540 (95.7) 291 (95.1) The season of surgery spring 572 (21.6) 75 (24.5) summer 801 (30.2) 95 (31.0) full 757 (28.5) 78 (25.5) winter 523 (19.7) 58 (19.0) Year 2018.79 ± 1.88 2018.94 ± 2.11 The level of exposure in period 3 PM 2.5 (µg/m 3 ) 22.93 ± 3.18 22.59 ± 5.01 PM 10 (µg/m 3 ) 40.98 ± 6.32 40.84 ± 7.53 NO 2 (µg/m 3 ) 21.04 ± 4.69 20.92 ± 5.78 CO (mg/m 3 ) 0.67 ± 0.12 0.67± 0.13 SO 2 (µg/m 3 ) 6.04 ± 1.94 6.08± 2.05 O 3 (µg/m 3 ) 86.38 ± 8.85 87.13 ± 12.49 Temperature (°C) 20.28 ± 1.78 21.01 ± 4.61 Relative humidity 77.54 ± 2.43 77.76± 3.40 NDVI 0.68 ± 0.04 0.68 ± 0.05 Wind speed (m/s) 2.02 ± 1.25 2.09 ± 1.31 Surface pressure (hPa) 971.14 ± 19.27 970.78 ± 20.05 UV intensity (J/m 2 ) 1,591,202.21 ± 171,835.17 1,637,762.10 ± 362,019.61 SD a : standard deviation; BMI b : body mass index (kg/m 2 ) Characteristics of 2654 participants who underwent fresh embryo transfer according to miscarriage status (mean ± SD a or N (%)) SD a : standard deviation; BMI b : body mass index (kg/m 2 ) We calculated pollutant exposures for participants who underwent fresh embryo transfer during period 1 and for miscarriage patients during period 3. Fig.  2 and Supplementary Table 1 display the distribution of participants' exposure to each environmental factor. In period 1, the exposure levels of the environmental factors were as follows: PM 2.5 22.18 (11.17, 50.85) µg/m 3 , PM 10 40.56 (21.95, 77.15) µg/m 3 , NO 2 20.40 (7.97, 41.95) µg/m 3 , SO 2 5.80 (3.35, 40.24) µg/m 3 , CO 0.66 (0.36, 1.63) mg/m 3 , O 3 88.22 (37.84, 125.54) µg/m 3 , relative humidity 78.34 (60.77, 87.30) %, temperature 22.11 (9.47, 29.08) °C, NDVI 0.69 (0.48, 0.80), wind speed 1.56 (0.99, 8.62) m/s, surface pressure 970.73 (918.39, 1019.83) hPa, and UV intensity 1,679,460.74 (762,308.97, 2,561,579.65) J/m 2 . In period 3: PM 2.5 22.68 (17.03, 32.99) µg/m 3 , PM 10 40.64 (25.70, 58.10) µg/m 3 , NO 2 20.52 (12.01, 34.78) µg/m 3 , SO 2 5.73 (3.83, 17.26) µg/m 3 , CO 20.52 (12.01, 34.78) mg/m 3 , O 3 86.44 (54.01, 109.46) µg/m 3 , relative humidity 77.22 (71.97, 85.06) %, temperature 20.26 (16.14, 23.22) °C, NDVI 0.67 (0.52, 0.76), wind speed 1.59 (1.08, 6.52) m/s, surface pressure 972.44 (923.10, 1013.47) hPa, and UV intensity 1,585,940.98 (1,294,682.52, 2,010,410.40) J/m 2 . Fig. 2 Combined violin plot and box plot depicting the distribution of environmental factor exposure. Period 1: Exposure to environmental factors in nonpregnant patients from 85 days before oocyte retrieval to the day of oocyte retrieval; period 2: exposure to environmental factors in miscarriage patients from 85 days before oocyte retrieval to the date of miscarriage or delivery Combined violin plot and box plot depicting the distribution of environmental factor exposure. Period 1: Exposure to environmental factors in nonpregnant patients from 85 days before oocyte retrieval to the day of oocyte retrieval; period 2: exposure to environmental factors in miscarriage patients from 85 days before oocyte retrieval to the date of miscarriage or delivery Correlation analysis revealed robust correlations between PM 2.5 and PM 10 ( r  = 0.86, p  < 0.001), between PM 10 and NO 2 ( r  = 0.84, p  < 0.001), and between PM 2.5 and NO 2 ( r  = 0.77, p  < 0.001) across period 1 (Supplementary Fig. 2). The variance inflation factor (VIF) indicates that there is no high collinearity among the exposure factors (Supplementary Table 2). Logistic regression analysis revealed that, in Period 1, PM 2.5 ( aOR  = 1.43, 95% CI : 1.05 ~ 1.95) and NO 2 ( aOR  = 1.28, 95% CI : 1 ~ 1.64) increased the miscarriage risk, whereas temperature ( aOR  = 0.6, 95% CI : 0.38 ~ 0.95) and UV intensity ( aOR  = 0.36, 95% CI : 0.13 ~ 0.99) decreased the miscarriage risk. In Period 2, PM 2.5 ( aOR  = 1.49, 95% CI : 1.08 ~ 2.06) and NO 2 ( aOR  = 1.3, 95% CI : 1.01 ~ 1.67) increased the miscarriage risk, and temperature ( aOR  = 0.58, 95% CI : 0.36 ~ 0.92) decreased the miscarriage risk. In Periods 3 and 4, temperature ( aOR  = 1.36, 95% CI : 1.25 ~ 1.48; aOR  = 1.78, 95% CI : 1.47 ~ 2.14), relative humidity ( aOR  = 1.35, 95% CI : 1.19 ~ 1.54; aOR  = 1.15, 95% CI : 1.02 ~ 1.31), the NDVI ( aOR  = 1.22, 95% CI : 1.02 ~ 1.46; aOR  = 1.32, 95% CI : 1.1 ~ 1.57) and UV intensity ( aOR  = 1.36, 95% CI : 1.21 ~ 1.54; aOR  = 1.51, 95% CI : 1.27 ~ 1.79) increased the miscarriage risk, and PM 2.5 ( aOR  = 0.76, 95% CI : 0.61 ~ 0.95) decreased the miscarriage risk (Fig.  3 ). The other detailed data are displayed in Table  2 . Fig. 3 Associations between environmental factor exposure and miscarriage. a : 85 days before oocyte retrieval to the day of oocyte retrieval; b : 85 days before oocyte retrieval to the day of β-hCG; c : 85 days before oocyte retrieval to the date of miscarriage or delivery; d : the date of embryo transplantation to the date of miscarriage or delivery Table 2 Effects of single environmental factors on miscarriage Environmental factors OR (95% CI) p value Miscarriage a Miscarriage b Miscarriage c Miscarriage d PM 2.5 Q 1.43(1.05 ~ 1.95)  1.49(1.08 ~ 2.06)  0.86(0.65 ~ 1.14) 0.76(0.61 ~ 0.95)    0.024   0.016   0.286   0.016 PM2.5 C 1.16(0.98 ~ 1.38) 1.19(0.99 ~ 1.42)  0.96(0.82 ~ 1.14) 0.83(0.71 ~ 0.96)   0.09   0.063 0.659   0.014 PM 10 Q 1.06(0.84 ~ 1.35) 1.06(0.83 ~ 1.35) 1.13(0.93 ~ 1.38) 1.08(0.87 ~ 1.34)  0.606 0.636 0.208 0.463 PM10 C 1.03(0.88 ~ 1.19) 1(0.86 ~ 1.16) 1.14(0.98 ~ 1.33) 1.07(0.92 ~ 1.24)    0.718 0.99 0.09 0.374 NO 2 Q 1.28(1 ~ 1.64) 1.3(1.01 ~ 1.67) 1.11(0.87 ~ 1.41) 0.96(0.77 ~ 1.19) 0.048   0.043 0.397   0.684 NO 2 C 1.09(0.93 ~ 1.27)  1.09(0.93 ~ 1.27) 1.01(0.88 ~ 1.17) 0.96(0.83 ~ 1.1) 0.282 0.308 0.883   0.534 CO Q 1.02(0.89 ~ 1.17) 1.04(0.91 ~ 1.18)  1.03(0.93 ~ 1.14) 1.03(0.9 ~ 1.18) 0.804 0.59 0.54 0.673 CO C 1.02(0.9 ~ 1.16) 1.02(0.9 ~ 1.17)  0.97(0.85 ~ 1.1) 1.01(0.89 ~ 1.16)  0.794 0.711 0.6 0.83 O 3 Q 1.05(0.91 ~ 1.22)  1.05(0.91 ~ 1.23) 1.11(0.96 ~ 1.28) 1.12(0.97 ~ 1.3)    0.494   0.49 0.171     0.123 O 3 C 1.07(0.94 ~ 1.22)  1.04(0.91 ~ 1.18) 1.01(0.88 ~ 1.15) 1.03(0.91 ~ 1.17) 0.301 0.582 0.919 0.646 SO 2 Q 1.08(0.89 ~ 1.3)  1.06(0.88 ~ 1.28) 1.14(0.96 ~ 1.36) 0.95(0.81 ~ 1.12)   0.448   0.535   0.149   0.557 SO 2 C 1.04(0.93 ~ 1.18)  1.03(0.92 ~ 1.16) 1.05(0.94 ~ 1.18) 0.96(0.86 ~ 1.07) 0.468   0.619   0.381   0.448 Temperature Q 0.6(0.38 ~ 0.95)  0.58(0.36 ~ 0.92)  1.36(1.25 ~ 1.48) 1.78(1.47 ~ 2.14) 0.031 0.021  <  0.001  <  0.001 Temperature C 0.77(0.63 ~ 0.94) 0.77(0.63 ~ 0.93) 1.32(1.17 ~ 1.5) 1.38(1.18 ~ 1.6) 0.012 0.008  <  0.001 <  0.001 Relative humidity Q 1.08(0.86 ~ 1.35) 1.1(0.87 ~ 1.38)  1.27(1.04 ~ 1.54) 0.99(0.83 ~ 1.19) 0.493 0.416 0.02 0.928 Relative humidity C 1.09(0.75 ~ 1.58) 1.06(0.92 ~ 1.22) 1.35(1.19 ~ 1.54) 1.15(1.02 ~ 1.31)   0.633 0.392  <  0.001 0.026 NDVI Q 0.9(0.73 ~ 1.11) 0.9(0.73 ~ 1.11) 1.22(1.02 ~ 1.46) 1.32(1.1 ~ 1.57) 0.333 0.332 0.032 0.003 NDVI C 0.95(0.82 ~ 1.1) 0.97(0.84 ~ 1.12)  1.12(1 ~ 1.26) 1.37(1.21 ~ 1.55) 0.501 0.701 0.058  <  0.001 Wind speed Q 0.33(0.08 ~ 1.35) 1.02(0.99 ~ 1.05) 1.01(1 ~ 1.03) 1.01(0.99 ~ 1.03)   0.123   0.146   0.179   0.231 Wind speed C 1.04(0.92 ~ 1.17) 1.01(0.9 ~ 1.14)  1(0.9 ~ 1.12) 1.02(0.91 ~ 1.14) 0.537   0.842   0.962   0.776 Surface pressure Q 5.93(0.02 ~ 2171.84)  1.03(0.94 ~ 1.12) 1(0.98 ~ 1.02) 0.99(0.95 ~ 1.03) 0.555 0.577   0.753   0.66 Surface pressure C 1.02(0.9 ~ 1.15) 1.04(0.92 ~ 1.17)  0.83(0.74 ~ 0.93) 0.89(0.79 ~ 0.99) 0.65 0.52 0.002 0.039 UV intensity Q 0.36(0.13 ~ 0.99) 0.71(0.5 ~ 1.01) 1.36(1.21 ~ 1.54) 1.51(1.27 ~ 1.79) 0.047 0.059 <  0.001 <  0.001 UV intensity C 0.82(0.69 ~ 0.98) 0.86(0.72 ~ 1.03)  1.22(1.09 ~ 1.37) 1.22(1.07 ~ 1.39) 0.031 0.105   <  0.001   0.003 Q Impact of changes in the interquartile range of exposure factors on miscarriage; C Impact of each 10 µg or 10 mg increase in exposure factors on miscarriage a Period 1; b Period 2; c Period 3; d Period 4 Associations between environmental factor exposure and miscarriage. a : 85 days before oocyte retrieval to the day of oocyte retrieval; b : 85 days before oocyte retrieval to the day of β-hCG; c : 85 days before oocyte retrieval to the date of miscarriage or delivery; d : the date of embryo transplantation to the date of miscarriage or delivery Effects of single environmental factors on miscarriage Q Impact of changes in the interquartile range of exposure factors on miscarriage; C Impact of each 10 µg or 10 mg increase in exposure factors on miscarriage a Period 1; b Period 2; c Period 3; d Period 4 We analyzed the interactive effects of air pollutants and environmental factors on the risk of miscarriage (Supplementary Table 3–6). In period 3, there were significant antagonistic effects between temperature and PM 2.5 (RERI = −2.36, p-interaction  < 0.001), PM 10 (RERI = −1.71, p-interaction  < 0.001) and NO 2 (RERI = −2.94, p-interaction  < 0.001). Relative humidity was significantly negatively correlated with SO 2 (RERI = −1.2, p-interaction  =  0.002 ) and positively correlated with O 3 (RERI = 1.57, p-interaction  <  0.001 ). The NDVI was significantly negatively correlated with PM 2.5 (RERI = −2.1, p-interaction  < 0.001), CO (RERI = −1.1, p-interaction  =  0.007 ), and SO 2 (RERI = −0.38, p-interaction  =  0.002 ) and positively correlated with O 3 (RERI = 1.66, p-interaction  <  0.001 ). Wind speed was significantly negatively correlated with PM 2.5 (RERI = −0.74, p-interaction  =  0.005 ), PM 10 (RERI = −0.83, p-interaction  =  0.004 ), NO 2 (RERI = −0.73, p-interaction  =  0.007 ), CO (RERI = −1, p-interaction  =  0.001 ), and O 3 (RERI = −2.83, p-interaction  <  0.001 ). Surface pressure was significantly negatively correlated with O 3 (RERI = −1.42, p-interaction  <  0.001 ) and positively correlated with NO 2 (RERI = 0.49, p-interaction  =  0.009 ) and PM 2.5 (RERI = 0.82, p-interaction  <  0.001 ). UV intensity was significantly negatively correlated with PM 2.5 (RERI = −4.48, p-interaction  <  0.001 ), PM 10 (RERI = −2.39, p-interaction  <  0.001 ), NO 2 (RERI = −3.85, p-interaction  <  0.001 ), CO (RERI = −0.86, p-interaction  =  0.011 ), and SO 2 (RERI = −0.67, p-interaction  =  0.035 ). The results of period 4 were similar to those of period 3. The QGC regression model indicated a significant increase in the risk of miscarriage in periods 3 ( aOR  = 4.07, 95% CI : 2.8 ~ 5.9, P  <  0.001 ) and 4 ( aOR  = 1.99, 95% CI : 1.35 ~ 2.94, P  =  0.015 ) (Supplementary Table 7 and Fig.  4 ), whereas the WQS regression model (seed=2022) indicated a nonsignificant increase in the risk of miscarriage (Supplementary Fig. 3 and Supplementary Table 7). Additionally, we modified the seeds (seed = 2021) to verify the robustness of the results. This finding is similar to the results of the QGC in period 3 ( aOR  = 3.65, 95% CI : 1.74 ~ 7.69, P  <  0.001 ) and period 4 ( aOR  = 2.55, 95% CI : 1.18 ~ 5.53, P  =  0.018 ) (Supplementary Fig. 4 and Supplementary Table 7). The BKMR regression model has similar results. In period 4, when the levels of mixed exposure to environmental factors increased within the lower ( P  < 0.05) quantile range, the risk of miscarriage increased accordingly (Fig.  5 ). The BKMR model did not identify any significant nonlinear associations between environmental factors and miscarriage risk (Supplementary Fig. 5). Fig. 4 Weights of positive or negative effects for each environmental factor on the risk of miscarriage were assessed via a quantile-based g-computation (QGC) regression model. a  85 days before oocyte retrieval to the day of oocyte retrieval; b  85 days before oocyte retrieval to the day of β-hCG; c  85 days before oocyte retrieval to the date of miscarriage or delivery; d  date of embryo transplantation to the date of miscarriage or delivery Fig. 5 Mixed effects of environmental factors on the risk of miscarriage analyzed by BKMR. a  85 days before oocyte retrieval to the day of oocyte retrieval; b  85 days before oocyte retrieval to the day of β-hCG; c  85 days before oocyte retrieval to the date of miscarriage or delivery; d  date of embryo transplantation to the date of miscarriage or delivery Weights of positive or negative effects for each environmental factor on the risk of miscarriage were assessed via a quantile-based g-computation (QGC) regression model. a  85 days before oocyte retrieval to the day of oocyte retrieval; b  85 days before oocyte retrieval to the day of β-hCG; c  85 days before oocyte retrieval to the date of miscarriage or delivery; d  date of embryo transplantation to the date of miscarriage or delivery Mixed effects of environmental factors on the risk of miscarriage analyzed by BKMR. a  85 days before oocyte retrieval to the day of oocyte retrieval; b  85 days before oocyte retrieval to the day of β-hCG; c  85 days before oocyte retrieval to the date of miscarriage or delivery; d  date of embryo transplantation to the date of miscarriage or delivery To increase the robustness and credibility of the findings, sensitivity analyses were conducted. According to Supplementary Table 9, after additional adjustment for different cities, compared with the original logistic model, PM 2.5 was associated with a reduced risk of miscarriage in period 3 ( aOR  = 0.67, 95% CI : 0.48 ~ 0.93, P  =  0.018 ) and period 4 ( aOR  = 0.67, 95% CI : 0.52 ~ 0.85, P  <  0.001 ). Further stratified analyses by urban and rural areas (Supplementary Table 10) indicated that NO 2 and PM 10 increased the risk of miscarriage only among urban women, whereas the NDVI (aOR = 1.19, 95% CI: 1.02 ~ 1.38, P = 0.03) increased the miscarriage risk exclusively among urban women in period 3. In contrast, PM 2.5 increased the risk of miscarriage only among rural women in periods 1 and 2. Moreover, considering that differences in miscarriage timing could lead to variations in exposure windows, stratified analyses were performed according to miscarriage timing (Supplementary Table 11). The results revealed that temperature ( aOR  = 0.8, 95% CI : 0.63–1, P = 0.048; aOR  = 0.77, 95% CI : 0.62–0.95, P = 0.016) and UV intensity ( aOR  = 0.3, 95% CI : 0.1–0.93, P = 0.037) were associated with a decreased risk of early miscarriage. The previously observed effects of the NDVI, relative humidity, and surface pressure on promoting miscarriage disappeared. In contrast, PM 2.5 , PM 10 , temperature, relative humidity, wind speed, surface pressure, UV intensity, and the NDVI are associated with an increased risk of second-trimester miscarriage. The Cox regression model revealed that temperature, relative humidity, NDVI and UV intensity increased the risk of miscarriage in period 3, but PM 2.5 decreased the risk of miscarriage in period 4 (Supplementary Fig. 7). Stratified analyses by maternal age (Supplementary Table 12) revealed that among women younger than 35 years, during period 3, PM 2.5 and wind speed were associated with a reduced risk of miscarriage, whereas temperature, relative humidity, NDVI, surface pressure, and UV intensity were associated with an increased risk. Notably, surface pressure was associated with a decreased risk of miscarriage in period 4. The DLNMs revealed that both the two-dimensional and three-dimensional exposure–response curves revealed nonlinear associations between ambient temperature and miscarriage risk but nonsignificant lagged associations. At lower temperatures (−20 °C—0 °C) and higher temperatures (25 °C—30 °C), the relative risk of miscarriage increased; however, the relative risk of miscarriage did not vary significantly with increasing lag duration (Supplementary Fig. 9). We identified 116 genes that are strongly linked to both exposure to environmental factors and miscarriage (Supplementary Fig. 10). Using the bioinformatics instrument GeneMANIA, a comprehensive network analysis was conducted to investigate the potential interconnections and regulatory mechanisms of these genes in intricate biological contexts (Supplementary Fig. 11 A). The analysis revealed that the majority of these genes (61.99%) were coexpressed. In addition, 12.48% of the genes presented physical interactions, 11.44% presented colocalization, 8.92% presented predictive interactions, 2.67% presented pathway associations, 1.30% presented genetic interactions, and 1.20% presented shared protein structural domains. We examined a protein‒protein interaction (PPI) network with 106 nodes and 666 edges via MCODE analysis to identify crucial proteins associated with environmental factors and miscarriage, where each gene circle's size corresponds to its degree of criticality. In the concentric circle visualization, the outermost circle represents the 33 most crucial genes identified by MCODE analysis, whereas the middle circle corresponds to the 13 genes deemed to have secondary importance (Supplementary Fig. 11B). Further analysis via cytoHubba identified 10 hub genes, including CXCL10, CD68, CD163, CD8A, TGFB1, FN1, CSF3, NCAF1, IL6, and IL1B, which were subsequently subjected to Gene Ontology (GO) and pathway analysis (Supplementary Fig. 11 C). GO analysis includes biological processes, molecular functions, and cellular components. Biological processes include biological regulation, regulation of biological processes, and response to stimuli. The main molecular functions included binding, protein binding, and signaling receptor binding. In terms of cellular components, these genes were associated with extracellular regions, cellular anatomical entities, the extracellular space, etc. (Supplementary Fig. 11D). Pathway analysis highlighted the participation of these genes in several major signaling pathways, with the most prominent being extracellular matrix organization, the PI3K‒Akt signaling pathway, and cytokine signaling in the immune system (Supplementary Fig. 11E).

To eliminate the potential impact of multiple visits during the study period on the outcome. This retrospective cohort study evaluated 10,709 infertile women who first underwent intracytoplasmic sperm injection (ICSI) or in vitro fertilization (IVF) cycles between January 2015 and December 2022 at Fujian Maternal and Child Health Hospital. Participants who met the following exclusion criteria were excluded: (1) missing characteristic or treatment information; (2) endometrial thickness < 7 mm on the transfer day; and (3) frozen embryo transfer after oocyte retrieval. None of the participants used oocytes or semen from donors. The Fujian Medical University Institutional Review Board approved the research protocol (Approval number: 2015035; Approval date: March 9, 2015). All the participants consented to the collection of environmental data. An informed consent form was signed by each participant who was enrolled in the research. Fig.  1 shows a flowchart that illustrates how individuals are included and excluded. Fig. 1 Flowchart of participants included in this study Flowchart of participants included in this study The operation process strictly followed the requirements of the relevant guidelines [ 20 ]. Published studies have described the surgical process and the definitions of related clinical outcomes in detail [ 21 – 25 ]. In brief, ovulation induction in the IVF/ICSI procedure was achieved via controlled ovarian stimulation. The protocol involves the following: controlled ovarian hyperstimulation (COH), follicular ovulation induced by β-hCG injection via transvaginal ultrasonography, oocyte retrieval via transvaginal ultrasonography, and transfer of the embryo to the uterine cavity after fertilization. According to previous studies [ 26 ], biochemical pregnancy was adjudicated when serum hCG levels were ≥ 10 mIU/mL on the 14th day after embryo transfer. Biochemical loss was confirmed when no gestational sac was detected through ultrasound 5 weeks after embryo transfer. Clinical pregnancy was defined as an intrauterine gestational sac confirmed by ultrasound on day 35 after embryo transfer, and miscarriage was defined as clinical pregnancy loss before 12 weeks of pregnancy. Second trimester miscarriage was defined as clinical pregnancy loss at 12–28 weeks of pregnancy. Six environmental elements (temperature, relative humidity, wind speed, surface pressure, UV intensity, normalized difference vegetation index NDVI and six air pollutants (PM 2.5 , PM 10 , NO 2 , CO, O 3 , SO 2 ) are representative of environmental factors. The China Air Quality Online Monitoring and Analysis Platform (AQI) website's real-time dataset ( https://www.aqistudy.cn/ ) provided detailed information on air pollutants. The China Meteorological Data Service Centre ( https://data.cma.cn/ ) provided the average daily temperature, relative humidity, wind speed, surface pressure, UV intensity during the same period for Fujian Province. The National Tibetan Plateau’s China regional 250 m normalized difference vegetation index dataset (2000–2023) (10.11888/Terre.tpdc.300328) provided the NDVI data. Throughout the study duration, the assessment of exposure level is detailed in the published research [ 22 ]. 8-h moving average O 3 concentration (containing at least six valid hourly values) within a calendar day. It has been widely used in O 3 health effects research [ 27 ]; thus, we collected 8-h O₃ levels during the day, whereas 24-h concentrations were used for the other environmental factors. Individual exposure average level to environmental factors throughout the study period was calculated by administrative cities where infertile women lived and number of days in different phase during ART. It takes approximately 85 days for folliculogenesis to occur under COH [ 28 ]; therefore, period 1 is defined as 85 days before oocyte retrieval to the day of oocyte retrieval [ 29 ]; period 2 is defined as 85 days before oocyte retrieval to the day of β-hCG [ 9 ]; period 3 is defined as the time from 85 days before oocyte retrieval to the date of miscarriage or delivery [ 9 ]; and period 4 is defined as the time from the date of embryo transplantation to the date of miscarriage or delivery (Supplementary Fig. 1). The Comparative Toxicological Genomics Database (CTD) is extensively mined for genes associated with air pollutants and miscarriage, while the GeneCards database looks for genes associated with environmental factors to gain a better understanding of the relationship between environmental factors and miscarriages. A gene interaction network of the common genes linked to miscarriage and environmental factors were explored by the GeneMANIA platform. Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) enrichment were analyzed to further understand the roles of these genes. The above relevant genes retrieved from the CTD and GeneCards database were used to further analyze protein interactions to speculate on the possible causes of miscarriage due to ambient air pollutants. Protein‒protein interaction (PPI) networks were created using the STRING database and shown with Cytoscape software (version 3.10.3) in order to clarify the networks of interactions between related proteins and their possible functions in certain biological processes. The MCODE (degree cutoff = 2, node score cutoff = 0.2, k-core = 2, and max. depth = 100) and cytoHubba (top nodes: 10, algorithm: MCC) plugins in Cytoscape were applied to identify major functional clusters and key genes from complex PPI networks. The descriptive statistics are provided on the characteristics of participants in fresh embryo transfer, based on miscarriage vs live birth. In the case of continuous variables, the mean ± SD will be employed. For categorical variables, n (%) will be used to display. Considering that changes in the units of different variables may have different meanings, the environmental factors (continuous variables) were subjected to quantile normalization, binary binning and quantile binning transformation. The multiple factors adjusted logistic regression model were conducted to analyze the correlation between individual exposure environmental factors and miscarriage (0 = live birth, 1 = miscarriage) in ART. To accurately evaluate the effects of environmental factors, such as air pollution and temperature, on miscarriage rates in ART and the magnitude of these effects, we included a range of variables as covariates, including potential confounders [ 30 ]. Guided by previous comparable studies[ 9 ], the covariates included residence (urban or rural), female age, duration of infertility, infertility diagnosis (PCOS, DOR, tubal factor infertility, endometriosis), number of oocytes retrieved, BMI, fertilization technique (IVF or ICSI), duration of gonadotropin use, gonadotropin dosage, number of embryo transferred, type of embryo transferred (blastocyst or morula), stimulation protocol, year of surgery, and season of surgery. To quantify the false-positive risk, we performed multiple-hypothesis-testing correction (E-value). To analyze the independent effects of ambient air pollutants and environmental factors on miscarriage and the interaction between them, a Poisson regression model containing an interaction term for one air pollutant and one environmental factor was fitted to explore whether there were synergistic or antagonistic effects. Another model was fitted to include only one air pollutant and one environmental factor as separate exposure variables, without considering their interactions, adjusting for the same covariates, to analyze the independent effects of single air pollutant and environmental factor on miscarriage. The final level of interaction was assessed by relative excess risk due to interaction (RERI). RERI = RRA + B + -RRA + B–RRA-B + + 1. RRA + B + indicates the relative risk when both factors are high level, RRA + B- indicates the relative risk when only factor A is high level, and RRA-B + indicates the relative risk when only factor B is high level. RERI reflects the extent to which the joint effect of the two factors deviates from the additive effect. RERI = 0 indicates that there is no interaction between the two factors; RERI > 0 indicates that there is a synergistic interaction, i.e., the two factors have no additive interaction. RERI = 0 indicates that there is no interaction between the two factors; RERI > 0 indicates that there is a synergistic interaction; RERI < 0 indicates an antagonistic interaction. To investigate the effect of mixed exposure on miscarriage, we applied weighted quantile sum (WQS) regression model, quantile g-computation (QGC) regression model and Bayesian kernel Regression (BKMR) regression model. The QGC regression model is designed to assess the joint impact of high-dimensional mixtures on a single outcome, combining the inferential simplicity of weighted quantile sum (WQS) regression model with the flexibility of g-computation [ 31 , 32 ]. Unlike the direction homogeneity required by the WQS regression model, the QGC regression model offers flexibility by assigning positive or negative weights to individual independent variables, depending on the direction of their independent effects. The R package “gWQS” was employed to construct the WQS model, with contaminants categorized into quartiles. The dataset was randomly partitioned into 40% for training and 60% for validation. A bootstrap procedure was performed with 1,000 iterations, and a logit link function was applied for binary outcomes. QGC regression model were constructed using the qgcomp.noboot function from R software package “qgcomp” and contaminants were scored using the six-quartile classification. Since QGC mainly assesses the overall effect through quantile coding and a weighted index and is relatively poor at dealing with nonlinearities, the BKMR model, which depends on the “BKMR” R package, was used in addition to the QGC model to analyze the linear relationship between exposure factors and miscarriage, as well as the combined effect of mixed exposures on miscarriage. Normalized data were used to fit the BKMR model via the kmbayes function, with the number of iterations set to 35,000, variable selection turned on, and nodes specified for subsequent analysis. On the basis of the correlations among the exposure factors, PM 2.5 , PM 10 , NO 2 were assigned to “1” group; CO, SO 2 , O 3 were assigned to “2” group; temperature and relative humidity were assigned to “3” group; NDVI, wind speed, surface pressure, and UV intensity were classified into groups 4, 5, 6, and 7, respectively. The OverallRiskSummaries function was used to assess the overall effect of the mixture on health outcomes. The PredictorResponseUnivar function was used to analyze the exposure–response curves for a single variable in relation to the outcome, with the levels of the other variables fixed at the median. This revealed the nonlinear relationship between the exposure factor and the outcome. Furthermore, we estimated the kernel hyperparameter r of the BKMR model across different exposure windows (trace plots) to evaluate the smoothness of the mixture effect functions. In the exploration of the nonlinear relationship between environmental factors and miscarriage, we utilized the "rms" package in R to fit a logistic regression model with restricted cubic splines (RCS). Additionally, to investigate the short-term effects of environmental factors on miscarriage using distributed lag nonlinear models (DLNMs), we employed the "dlnm" and "splines" packages. A cross-basis function for daily temperature was constructed using natural cubic splines (3 degrees of freedom) for both the cumulative-response and lag-response dimensions, with a maximum lag of 28 days [ 33 ]. The df that minimizes the Akaike Information Criterion (AIC) were selected for the lag functions. This cross-basis was included in a generalized linear model (GLM) with a binomial family. Predicted relative risks were visualized using three-dimensional surface plots and two-dimensional contour plots, enabling the assessment of both delayed and nonlinear associations between environmental factors and miscarriage. All analyses were conducted in R version 4.4.0, all statistical tests were two-sided, and p < 0.05 was considered statistically significant.

Considering this interaction, the results of the present study demonstrated that PM 10 , NO 2 , and CO increase the risk of miscarriage. In our study, temperature reduced the effects of air pollutants on miscarriage. Mixed exposure was positively associated with the risk of miscarriage. The enrichment of predicted genes suggests that air pollutants may induce miscarriage through inflammatory mechanisms.

This study highlights the significant role of environmental factors in adverse reproductive outcomes. In this retrospective cohort study, six air pollutants and six environmental factors were examined to approximate, as closely as possible, the effects of real-world environmental exposure on the risk of miscarriage following ART. These results indicate that PM 2.5 , NDVI and UV intensity are associated with an increased risk of miscarriage. Moreover, the analysis revealed a complex interactive effect between air pollutants and environmental factors on miscarriage. Furthermore, mixed exposure to both air pollutants and environmental factors is associated with an increased risk of miscarriage. A detailed analysis of genes revealed that genes associated with environmental factors and miscarriage are associated primarily with immune-inflammatory responses. The interaction analysis revealed that temperature was significantly negatively correlated with PM 2.5 , PM 10 and NO 2 . A previous study confirmed our findings [ 34 ], indicating that the highest levels of air pollution occurred in winter and that the lowest levels occurred in summer. However, heat exposure may impair fetal growth by reducing uterine blood flow and altering placental–fetal exchange [ 35 ]. Moreover, high temperatures may also exacerbate the adverse effects of PM 2.5 [ 36 ]. Sensitivity analysis revealed that the protective effect of PM 2.5 in period 4 was observed only among rural women and women younger than 35 years. This finding may be attributed to greater work-related stress and less balanced dietary patterns among younger women[ 37 ], as well as to potential confounding factors such as lower educational and socioeconomic levels among rural women. Future studies should pay particular attention to these factors. In addition, because transient population mobility during early and mid-pregnancy could not be avoided, city-wide averages were used to represent individual exposures, which may have introduced some measurement bias. However, with the exception of the NDVI, the associations between other environmental factors and miscarriage were consistent with findings from previous studies. Future research should continue to account for the issue of population mobility. In addition, regardless of whether a single exposure or mixed exposure was used, we found that the NDVI increased the risk of miscarriage. A study conducted in Guangdong, China, reported a similar result: an increase in green space intensified the negative impact of high temperatures on miscarriage rates [ 38 ]. However, the NDVI was significantly linked to an increased risk of miscarriage among urban women, whereas no such association was found among rural women in our study. This phenomenon may be linked to urban-specific factors, such as pollution from traffic and higher levels of work-related stress experienced by women. One study, which used NO₂ as a marker for traffic emissions, indicated that elevated levels of traffic-related air pollution were associated with a greater risk of miscarriage [ 39 ]. This aligns with our sensitivity analysis, where increased NO₂ exposure was correlated with a greater risk of miscarriage among urban women, whereas no significant association was found in rural women. Another study indicated that women living near major roads with heavy traffic had an increased risk of spontaneous miscarriage [ 40 ]. In addition, work stress is significantly associated with an increased risk of miscarriage[ 41 ]. Therefore, future studies should consider the social factors of the included population. Many studies have focused on the mixed effects of mixed exposure to a few environmental factors on reproductive outcomes [ 42 , 43 ], and these findings are similar to our results. However, the underlying mechanisms are currently unclear. Among our hub genes, CSF3, IL-6 and IL-1B are potent proinflammatory cytokines [ 44 , 45 ]. CXCL10 is categorized functionally as a Th1-chemokine. It binds to the receptor CXCR3 and regulates immune responses [ 46 ]. has also been implicated in various immune-related diseases[ 47 , 48 ]. FN1 may be associated with autophagy[ 49 ]. TGFB1, CD68, CD163, and CD8A are related to inflammation and immune responses[ 50 , 51 ]. Studies have shown that once fine particulate matter is deposited in the lungs, the organic chemicals and metals it contains can generate reactive oxygen species (ROS) through redox cycling, depletion of cellular thiols, or activation of lymphocytes [ 52 ]. Pulmonary inflammation and oxidative stress pathways respond most rapidly to ambient air pollution (within 24 h), subsequently triggering systemic inflammatory responses [ 53 ]. Ambient fine particulate matter has also been shown to be associated with placental inflammation [ 54 ]. In addition, ambient fine particulate matter carries some endocrine-disrupting chemicals, such as per- and polyfluoroalkyl substances [ 55 ], and these pollutants can induce oxidative stress, which in turn affects estrogen synthesis [ 56 ]. These findings suggest that air pollutants may increase the risk of miscarriage by promoting inflammatory or immune responses. Environmental factors may be involved in inflammatory mechanisms that influence reproductive outcomes, such as heat stress, which can lead to oxidative stress and the release of inflammatory markers [ 57 , 58 ]. UV-induced damage can result in inflammation through the COX-2 mechanism, mitogen-activated protein kinase (MAPK) signaling pathway or epidermal growth factor receptor (EGFR) pathway. High humidity exacerbates respiratory mucosal and lung inflammation [ 59 , 60 ]. In addition, inflammatory factors may directly affect embryo implantation and thus reproductive outcomes. In mice, IL-11 is necessary for blastocyst implantation and the decidualization of endometrial stromal cells [ 61 ]. In humans, IL-11 mRNA and protein are produced in the stroma and are limited to PR decidualized stromal cells during the late secretory phase, where they promote blastocyst implantation [ 62 ]. The expression of inflammatory genes may influence the results of IVF reproduction. However, whether environmental factors increase the risk of miscarriage by enhancing inflammatory responses remains unclear, and future mechanistic studies are needed to verify this hypothesis. Additionally, experimental or clinical intervention studies could investigate whether reducing inflammation mitigates the increased risk of miscarriage associated with adverse environmental exposures. The strengths of this study are that, to the best of our knowledge, no study has incorporated such a large number of environmental factors. in the fields of environmental factors and reproductive health. The broad range of environmental factors included in our study offers a more accurate simulation of real-world effects on the number of oocytes retrieved and reproductive outcomes. Moreover, the specific environmental conditions of Fujian Province, China, characterized by lower air pollutant concentrations, higher temperatures, and greater NDVIs, provided representative exposure levels for the study. We also employed a QGC regression model to examine the positive or negative effects of a single factor under mixed exposure to accurately identify favorable and harmful factors. Additionally, we explored potential mechanisms through the identification of common genes associated with air pollutants and miscarriage, offering a theoretical foundation for the development of personalized targeted prevention strategies or further validation of these mechanisms. However, this study has several limitations. First, the registration information was recorded at the time of receiving the cycle. Therefore, determining the exposure levels of participants in different regions other than those recorded during registration before oocyte retrieval and after embryo transfer is difficult. Second, we are unable to account for potential confounding effects since we were unable to gather parity, prior loss, smoking/passive smoke, alcohol, occupation, and individual socioeconomic position (SEP) data, such as household income and education. Third, the duration of sunshine and seasonal changes are strongly correlated with the UV intensity. However, data on the hourly UV intensity could not be collected since the UV intensity statistics reflect daily averages. The same restrictions apply to weather conditions and exposure levels to other air contaminants. Finally, our research focused mainly on patients who underwent fresh embryo transfer. This is because when frozen embryos are used, personal information (such as age and physical condition) as well as the quality of the frozen ovum may change, thereby affecting the research results. Therefore, the results of this study are applicable only to patients undergoing fresh embryo transfer. Future studies should broaden the study population, enhance exposure assessment through remote sensing technologies and sampling methods, and validate the underlying mechanisms in animal and cellular models.

The advent of industrialization has given rise to a series of environmental issues that have become a matter of grave concern for public health [ 1 ]. In recent years, an increasing number of studies have focused on the relationship between air pollution and reproductive health. However, existing studies have reached inconsistent conclusions [ 2 – 6 ]. Moreover, conclusions about the effects of individual air pollutants on reproductive outcomes are inconsistent [ 7 – 9 ]. People are exposed to a complex variety of environmental factors, not just air pollution, and reproductive outcomes are the result of a combination of effects. Although ART is the primary treatment for infertile patients, infertile patients often have poor sperm or oocyte quality, which may negatively impact reproductive outcomes. Therefore, exploring the effects of real-world air pollution combined with other environmental factors on reproductive outcomes is necessary. Studies have shown that some environmental factors not only affect the concentration and spatial‒temporal distribution of air pollutants [ 10 , 11 ] but also directly affect reproductive outcomes, such as temperature [ 12 ] or atmospheric pressure [ 13 ], which play a more important role than does air pollution in association with poor birth outcomes. A multicenter study in China revealed that particulate matter (PM 2.5) , sulfur dioxide (SO 2 ) and ozone (O 3 ) were adversely correlated with reproductive outcomes after assisted reproductive technology (ART) under the influences of air temperature, relative humidity, and wind speed [ 14 ]. However, this study was limited by relatively few air pollutants. Greenness, which plays an important role in environmental conditions, is also related to various aspects of reproductive health, including androgen and progesterone levels [ 15 ], preterm birth (PTB) [ 16 ] and low birth weight (LBW) [ 17 ]. Previous studies have paid little attention to the impact of green space on pregnancy rates and miscarriages. Although a higher annual average UV index has also been shown to be positively associated with higher rates of LBW and PTB [ 18 ], very few studies on the effects of UV radiation on reproductive outcomes exist to provide reliable evidence. Importantly, although studies have examined the effects of mixed exposure to multiple factors on reproductive outcomes, research on the interactions between these factors is urgently needed. Existing evidence suggests that the NDVI and PM2.5 interact in relation to mortality and that temperature, the NDVI, and PM 2.5 interact in the development of hypertension[ 19 ]. However, studies investigating the interactions between environmental pollutants and other environmental factors in relation to miscarriage are scarce. This gap in knowledge is critical for environmental management and the prevention of adverse reproductive outcomes. In conclusion, the number and types of pollutants included in the current literature are still inadequate in relation to the situations in which people are exposed in the real world and how they interact with each other. Consequently, the present study sought to assess the mixed effects of comprehensive environmental factors, encompassing six air pollutants and six environmental factors, on the risk of miscarriage, as well as their individual contributions to these effects. To elucidate the potential mechanism underlying the relationship between environmental factors and miscarriage, genetic prediction analyses were also performed to screen for genes that are strongly associated with both. This research more fully elaborates the influence of environmental factors on reproductive health. The results provide a scientific basis for further related interventions and policy-making to ameliorate the adverse effects of the environment on health.

Supplementary Material 1. Supplementary Material 2. Supplementary Material 1. Supplementary Material 2.