The Influence of Ambient Air Pollution during Folliculogenesis, Endometrial Development, and Spermatogenesis on Fertility: A Study among Patients using Vitrified Donor Oocyte In Vitro Fertilization

Abstract

Background:

Air pollution is associated with reduced fertility; however, the independent effects of air pollution exposure during folliculogenesis, endometrial development, and spermatogenesis on human reproduction remains unclear.

Objective:

We investigated the associations between air pollution during critical windows and clinical outcomes of in vitro fertilization (IVF) using patients undergoing vitrified donor oocyte IVF.

Methods:

We included 551 non-identified oocyte donors and 1,353 recipients who underwent 2,533 embryo transfers at Reproductive Biology Associates in Atlanta, Georgia (2008-2019). Daily ambient exposure to air pollutants was estimated using spatiotemporal models linked to residential addresses and averaged over folliculogenesis (90 days before stimulation) and during stimulation in donors, three months and three weeks prior to transfer in recipients, and spermatogenesis (72 days prior to oocyte thaw) in male partners. Multivariable generalized estimating equations were used to estimate adjusted odds ratios (aOR) and 95% confidence intervals (CI) per interquartile range increase in pollutants in relation to clinical IVF outcomes.

Results:

Higher carbon monoxide (CO) exposure during folliculogenesis (aOR= 0.86 95% CI 0.77, 0.96) and ovarian stimulation (aOR=0.88 95% CI 0.79, 0.98) in donors was associated with lower odds of positive pregnancy test. Higher elemental carbon (EC) exposure during folliculogenesis was associated with lower odds of live birth (aOR=0.87 95% CI 0.79, 0.96) and higher odds of pregnancy loss (aOR=1.16 95% CI 1.02, 1.32). During ovarian stimulation, higher coarse particulate matter (PM₁₀) exposure was associated with lower odds of live birth (aOR=0.90 95% CI 0.80, 1.00) and higher odds of pregnancy loss (aOR= 1.18 95% CI 1.02, 1.36). Among male partners, higher PM₁₀ exposure during spermatogenesis was associated with higher odds of pregnancy loss (aOR=1.45 95% 1.09, 1.92). Air pollution exposures in the recipient were generally not associated with IVF outcomes.

Conclusions:

Preconception air pollution exposure during both folliculogenesis and spermatogenesis may adversely impact IVF outcomes.

Introduction

Air pollution is a common environmental exposure across the globe, accountable for more than 7 million premature deaths annually.¹ In recent years, interest in the potential reprotoxic effects of air pollution exposure has grown.² Air pollution exposure has been associated with lower probability of pregnancy and live birth and higher risk of miscarriage among the general population and subfertile patients using in vitro fertilization (IVF).² However, the specific pollutants, critical exposure windows, and biological mechanisms underlying these associations remain unclear. Patients undergoing IVF offer several advantages to investigating these research questions as it is possible to 1) directly observe many early reproductive outcomes that cannot be observed in couples attempting to conceive naturally and 2) more clearly define etiologically relevant windows of air pollution exposure because the timing of key events (e.g., day of fertilization) are certain.

A systematic review and meta-analysis of air pollution exposures during multiple periods of IVF showed that ambient residential exposure to particulate matter less than or equal to 2.5μm (PM_2.5), particulate matter less than or equal to 10μm (PM₁₀), sulfur dioxide (SO₂), and nitrogen dioxide (NO₂) was associated with lower biochemical pregnancy rates, ambient NO₂, SO₂, and carbon monoxide (CO) exposure was associated with lower clinical pregnancy rates, and ambient PM_2.5 and CO exposure was associated with lower live birth rates.³ Moreover, the review concluded that the 85-day period before initiation of controlled ovarian stimulation was the most sensitive time window of air pollution exposure, indicating that preconception air pollution exposures may be particularly harmful for IVF outcomes.⁴ While this seemingly implies that air pollution exposure to oocytes are perhaps the most important, the majority of studies included in the review and meta-analysis only included fresh IVF cycles, which means that the 85-day preconception period may encompass folliculogenesis, endometrial development, and spermatogenesis. This greatly limits our ability to disentangle the independent effects of air pollution exposures on specific pathways including oocyte quality, sperm quality, and uterine receptivity.

Two studies have utilized autologous frozen embryo transfer (FET) cycles to examine the independent associations of air pollution exposures to the oocyte and endometrium on clinical outcomes of IVF.⁵ In the first study, Wu and colleagues evaluated ambient exposure to six air pollutants (PM_2.5, PM₁₀, O₃, NO₂, CO, and SO₂) during the 85-day period prior to oocyte retrieval and the 30-day period prior to embryo transfer in relation to the outcomes of 17,676 FET cycles. The authors observed that higher exposure to SO₂ in the 85-day period prior to oocyte retrieval and the 30 days prior to FET was associated with decreased odds of pregnancy and live birth, suggesting both exposure windows may be important. However, no associations were observed between SO₂ exposures and IVF outcomes among fresh embryo transfers. In the second study, Leathersich et al. explored associations between air pollution during various periods prior to oocyte retrieval and embryo transfer and live birth among 3,659 FETs.⁶ Their study found that PM_2.5 exposure in the 3 months and PM₁₀ exposure in the 2 weeks before oocyte retrieval were negatively associated with live birth following FET. No associations were observed for air pollution exposures surrounding the time of embryo transfer. To date, no studies have evaluated the independent effects of paternal air pollution exposure on IVF outcomes as typically all male/female couples resided at the same address and the time window of spermatogenesis largely overlapped with the window of folliculogenesis and ovarian stimulation.

To circumvent these limitations, our objective was to utilize a novel and emerging model of human fertility- vitrified donor oocyte IVF- to evaluate how air pollution influences human reproduction. Because vitrified oocytes are obtained from non-identified (formerly known as “anonymous”) young, female donors months to years prior to the female recipient’s IVF cycle, there is no correlation in exposure between the donor and the female recipient or the donor and the recipient’s male partner. This allows for the independent investigation of how air pollution exposure in the oocyte and sperm affects IVF outcomes and allows for an investigation of how air pollution affects oocyte quality and endometrial receptivity independently.

Methods

Study Population

This retrospective cohort study included non-identified vitrified oocyte donors who underwent oocyte retrieval at Reproductive Biology Associates (RBA) in Atlanta, Georgia, USA between 2008 and 2019 and the recipient couples who underwent embryo transfers with the vitrified oocytes during the same time. Institutional Review Board (IRB) approval for this study was provided by Emory University (00080463). Oocyte donors were screened according to the clinical protocols outlined by the American Society for Reproductive Medicine.⁷ Females were eligible for oocyte donation if they were between 21 and 31 years old, had a body mass index (BMI) between 18 and 27 kg/m², reported regular menses, and had no medical contraindications to surgery. In certain cases, at physician discretion, candidates with values beyond those outlined above were permitted to donate oocytes. Further, donors were screened for infectious diseases, underwent psychological evaluations, completed a questionnaire on demographic and lifestyle characteristics, and had general and gynecological exams conducted. Among recipients, the primary reasons for utilizing oocyte donation were advanced reproductive age, genetic or chromosomal disorders, or ovarian insufficiency.

There was a total of 662 oocyte donors who underwent 1,082 retrieval cycles and 1,587 recipients (including intended parents and gestational carriers) who underwent 3,067 embryo transfers in the RBA data (Figure S1a). After application of exclusion criteria and linking oocyte donors with complete air pollution information to oocyte recipients with complete information, our primary analytic sample included 551 donors and 1,353 recipients (2,533 fresh and frozen embryo transfers). For our analyses aimed at isolating the independent associations of air pollution exposures during spermatogenesis on IVF outcomes, we further excluded fresh embryo transfers (since the time window for spermatogenesis and endometrial development were too concordant) and frozen transfers that included embryos not created from fresh male partner sperm. This analysis included a total of 534 recipients and 793 frozen embryo transfers (Figure S1b).

Air Pollution Exposure Assessment

Residential addresses extracted from oocyte donor and recipient medical records were geocoded using ArcGIS Streetmap USA. Any changes in donor residential addresses between oocyte retrieval cycles were recorded with the new address and year of move. Recipient residential addresses were collected at first treatment cycle. Male/female recipient couples who used fresh male partner sperm were assumed to reside at the same residential address. If a gestational carrier was utilized, information on their residential address was collected from available medical record data.

Ambient air pollution exposures were estimated by fusing results from both the Community Multiscale Air Quality (CMAQ) model and a fine-resolution research line-source (R-LINE) dispersion model, along with observations. This joint model exposure assessment approach has been previously applied to the RBA cohort and described elsewhere.⁸ In short, CMAQ is a chemical transport model that uses inputs from local emission data, meteorological factors, and photochemical properties of pollutants in the atmosphere to produce hourly volume-averaged pollutant concentrations for fixed three-dimensional 12 km² spatial resolution grids.⁹ We calculated daily averages of six criteria air pollutants and four particulate matter constituents: CO, elemental carbon (EC), ammonium (NH₄), NO₂, nitrate (NO₃), nitrogen oxides (NOx), O₃, organic carbon (OC), PM_2.5, and PM₁₀. Daily average concentrations of all air pollutants were bias-corrected with observations from station monitors and land use variables,¹⁰ then linked to oocyte donors and recipients using geocoded addresses.

Within the Atlanta metro area hourly estimates from CMAQ were combined with daily average concentrations of CO, NO_x, and PM_2.5 from traffic simulated using the R-LINE model, again bias-corrected using station monitors and traffic data from 2002-2019, to produce 250 m spatial resolution estimates.^10–12 For oocyte donors and recipients residing within the 29-county Atlanta region, daily estimates of CO, NO_x, and PM_2.5 from the R-LINE model were preferentially linked using geocoded addresses and the remaining criteria air pollutants were linked using daily estimates from the coarser resolution CMAQ model.

Daily average pollutant estimates were then averaged over two biologically relevant periods for the oocyte donors: folliculogenesis (estimated as the 90-day period before beginning ovarian stimulation) and controlled ovarian stimulation (8-13 days) and two periods for the oocyte recipients: three weeks prior to embryo transfer (roughly corresponding to the period of endometrial preparation) and three months prior to the initiation of endometrial preparation. In male partners who provided fresh sperm, we estimated average air pollution exposures during spermatogenesis (72 days prior to oocyte thaw/fresh semen sample). An overview of these exposure windows as well as the timeline for vitrified donor oocyte IVF is shown in Figure S2.

Outcome Assessment

Controlled ovarian stimulation was performed in the donors using injectable gonadotropins with an antagonist protocol with agonist trigger. Mature oocytes were vitrified 39-40 hours after trigger, using the “minimum volume” method.¹³ Months to years later, recipients received a cohort of vitrified oocytes that were fertilized with sperm provided by the male partner(s) or sperm donor using intracytoplasmic sperm injection (ICSI), 2-3 hours after oocyte warming. Embryos were cultured until cleavage or blastocyst stage. Pre-implantation genetic testing was not performed on any of the embryos. The highest quality embryos were transferred first as part of a fresh embryo transfer cycle and any supernumerary embryos with adequate quality were cryopreserved for later use. The female recipient or a gestational carrier underwent a standard endometrial preparation of estrogen and progesterone for approximately three weeks to prepare for embryo transfer. Our main analysis included both fresh and frozen embryo transfer cycles.

A positive pregnancy test was defined as a serum β- human chorionic gonadotropin (hCG) level >6 mIU/mL, generally assessed 10 days after fresh or frozen embryo transfer. A clinical pregnancy was defined as the presence of an intrauterine gestational sac confirmed by ultrasound, which occurs at approximately 6 weeks. A live birth was defined as the birth of a neonate on or after 24 weeks gestation. These outcomes were evaluated among all embryo transfer cycles and defined as 1 if the outcome was achieved and 0 otherwise. Pregnancy loss was defined as any positive pregnancy test that did not result in a live birth (e.g. biochemical loss, ectopic pregnancy, miscarriage, elective termination, or stillbirth). Spontaneous abortion was defined as any clinical pregnancy that resulted in a loss before 20 weeks gestation (e.g. miscarriage).

Covariate Assessment

Age, race, and education were extracted from donor medical charts. Age and race were extracted from recipient and gestational carrier medical records. Estimates of median family income, median home value, and percent of households living below the federal poverty line at the census tract level were derived from the American Community Survey 5-Year Data (2011-2015; overlapping with the midpoint of study years) and linked to donors and recipients using geocoded addresses. Daily ambient temperatures were estimated at a 4 km² spatial resolution using the Parameter-elevation Regressions on Independent Slopes Model (PRISM).¹⁴ Using daily average dewpoint and maximum temperatures, we calculated average daily relative humidity using the Magnus approximation,¹⁵ and averaged daily maximum temperature and relative humidity over the same two periods for oocyte donors and recipients and during spermatogenesis in male partners.

Statistical Analysis

Descriptive statistics summarizing the demographic, reproductive, and residential information for both oocyte donors and recipients were presented as counts and percentages or medians and 25^th and 75^th percentiles for categorical and continuous variables, respectively. Spearman correlation coefficients were used to measure correlations of average air pollutants during and between exposure periods for donors (i.e., folliculogenesis and ovarian stimulation) and recipients (i.e., three months prior to endometrial preparation and three weeks prior to embryo transfer). We used multivariate generalized estimating equations with a binomial distribution, logit link function, and robust standard errors to estimate the adjusted associations between an interquartile range (IQR) increase in each pollutant during each exposure period in relation to positive pregnancy test, clinical pregnancy, live birth, pregnancy loss, and spontaneous abortion. We applied a marginal modeling approach to account for the potential non-nested clustering of outcomes for donors who had contributed oocytes to multiple recipients and recipients who underwent more than one embryo transfer.¹⁶ In order to compare effect estimates, the IQR of pollutant exposures were standardized across all exposure periods as follows: CO (0.5 ppm), NOx (30 ppb), NO₂ (9 ppm), PM_2.5 (3.0 μg/m³), EC (0.3 μg/m³), NH₄ (0.5 μg/m³), NO₃ (0.8 μg/m³), O₃ (0.01 ppm), OC (1.2 μg/m³), and PM₁₀ (5.0 μg/m³).

We determined potential confounding using a priori knowledge and directed acyclic graphs. Donor exposure models were adjusted for donor age (21-23, 24-26, 27-29, ≥30 years), education (completed high school or currently enrolled in or attended some college, completed college degree or technical school, pursuing or completed advanced degree), race (White, Black, Other), year (continuous) and season (Winter [Dec-Feb], Spring [Mar-May], Summer [Jun-Aug], Fall [Sept-Nov]) of oocyte retrieval cycle, percent of households in the oocyte donor’s census tract living below the federal poverty line (continuous), and average relative humidity during folliculogenesis and controlled ovarian stimulation (continuous). Recipient exposure models were adjusted for age (25-34, 35-44, ≥45 years), race (Black, White, Other), year (continuous) and season (Winter [Dec-Feb], Spring [Mar-May], Summer [Jun-Aug], Fall [Sept-Nov]) of embryo transfer, percentage of households in the recipient’s census tract living below the federal poverty line (continuous), and average relative humidity during the corresponding time window (continuous). Male partner exposure models (i.e., during spermatogenesis) were adjusted for male partner age (continuous), year (continuous) and season of oocyte thaw (Winter [Dec-Feb], Spring [Mar-May], Summer [Jun-Aug], Fall [Sept-Nov]), percent of households in the recipient male partner’s census tract living below poverty (continuous), and average relative humidity during spermatogenesis (continuous). Due to the high correlations between maximum temperature and air pollutants (specifically, NO₃, O₃, and PM₁₀; Figure S3), models were not adjusted for maximum temperature to avoid multicollinearity.

All covariates had low (≤5%) missingness (e.g. donor and recipient race, donor education, census tract sociodemographic measures, and recipient age) and were singly imputed using the most common category or the median value for categorical and continuous variables, respectively. All statistical analyses were conducted using SAS software, Version 9.4 and R Statistical Software, Version 4.3.0.^17,18

Sensitivity Analyses

To test the robustness of our findings, we further adjusted our multivariable models for maximum temperature and other pollutants. In multipollutant exposure models, PM₁₀, PM_2.5, and PM_2.5 components (EC, NH₄, NO₃, and OC) were additionally adjusted for NOx and O₃; CO, NOx, and NO₂ were adjusted for O₃ and PM_2.5; and O₃ was adjusted for NOx and PM_2.5. We also estimated the adjusted associations between ambient air pollution exposures during all exposure periods in relation to live birth, stratified by fresh or frozen embryo transfer cycles, which was hypothesized to be a potential effect mediator. To evaluate the potential for non-linear relations between the air pollutants and live birth, we used restricted cubic splines, which compared the linear model to a model including linear and cubic spline terms using a likelihood ratio test.

Results

This retrospective cohort study included 551 oocyte donors and 1,353 oocyte recipients (including 41 gestational carriers) who underwent 2,533 embryo transfers (Figure S1a). The median (25^th, 75^th percentile) age of oocyte donors was 25.0 (23.0, 27.0) years, most (72.1%) reported White race, 89.3% were never smokers, and 43.4% had completed a college degree or technical school (Table 1). In our analytic sample, most oocyte donors (66.2%) underwent a single retrieval cycle, 18.9% underwent two, and 14.9% underwent between three and six retrieval cycles. The majority of oocyte donors (89.3%) contributed oocytes to more than one recipient.

Table 1.

Demographic, reproductive, and residential characteristics of vitrified oocyte donors at first oocyte retrieval.

Donors (n=551)	N (%) or Median (25th, 75th)
Demographic Characteristics
Age (years)	25.0 (23.0, 27.0)
Race/Ethnicity
White	397 (72.1)
Black	63 (11.4)
Asian/South Asian/Southeastern Asian	28 (5.1)
Hispanic/Latino	27 (4.9)
Other	30 (5.4)
Missing	6 (1.1)
Smoking status
Never smoker	492 (89.3)
Past smoker	37 (6.7)
Current smoker	11 (2.0)
Missing	11 (2.0)
BMI (kg/m2)	22.4 (20.8, 24.4)
Missing	1 (0.1)
Highest level of education completed
Completed high school or attended some college	37 (6.7)
Currently enrolled in college	154 (28.0)
Completed college degree or technical school	239 (43.4)
Pursuing or completed advanced degree	108 (19.6)
Missing	13 (2.4)
Census-level Characteristics
Median household income	$58,839 (45,346, 75,246)
Median value of owner-occupied houses	$179,400 (133,700, 246,100)
Missing	3 (0.5)
Percent of households living below poverty level	12.6 (7.8, 20.8)
Retrieval information
Year of Retrieval
2008-2010	130 (23.6)
2011-2013	165 (30)
2014-2016	165 (30)
2017-2019	91 (16.5)
Season of Retrieval
Spring (Mar-May)	148 (26.9)
Summer (Jun-Aug)	164 (29.8)
Autumn (Sept-Nov)	143 (26)
Winter (Dec-Feb)	96 (17.4)
Number of recipients per donor a
1	270 (10.7)
2-5	1335 (52.7)
6-15	928 (36.6)
Total retrieval cycles per donor a
1	365 (66.2)
2	104 (18.9)
3-6	82 (14.9)
Estimated ambient residential air pollution exposure
3 months before ovarian stimulation
NOx (ppb)	32.4 (19.9, 49.9)
CO (ppm)	0.5 (0.2, 0.7)
PM2.5 (μg/m3)	9.4 (8.3, 11.4)
EC (μg/m3)	0.6 (0.4, 0.8)
NH4 (μg/m3)	0.5 (0.3, 0.8)
NO2 (ppm)	18.5 (12.2, 22.9)
NO3 (μg/m3)	0.8 (0.5, 1.4)
O3 (ppm)	0.04 (0.03, 0.05)
OC (μg/m3)	2.3 (1.8, 2.9)
PM10 (μg/m3)	17.1 (14.9, 20.2)
During ovarian stimulation
NOx (ppb)	29.4 (17.4, 50.0)
CO (ppm)	0.5 (0.2, 0.8)
PM2.5 (μg/m3)	9.5 (7.8, 11.5)
EC (μg/m3)	0.6 (0.4, 0.8)
NH4 (μg/m3)	0.4 (0.3, 0.8)
NO2 (ppm)	17.2 (11.4, 22.7)
NO3 (μg/m3)	0.7 (0.4, 1.2)
O3 (ppm)	0.04 (0.03, 0.05)
OC (μg/m3)	2.3 (1.6, 3.1)
PM10 (μg/m3)	17.1 (14.5, 20.5)

^aIn the analytic sample.

The median (25^th, 75^th) age of oocyte recipients (inclusive of both intended parents and gestational carriers) was 42.2 (38.9, 44.8) years, most (69.9%) reported White race, 76.2% were nulliparous, and 84.9% were undergoing donor oocyte IVF for the first time (Table 2). Half of oocyte recipients (50.7%) underwent one embryo transfer, 41.8% underwent two or three transfers, and 7.5% underwent four or more during our study period. Blastocyst-stage (94.5%), single embryo (71.3%) transfers were most common (Table S1). Across all transfer cycles, 68.9% resulted in a positive pregnancy test, 59% resulted in a clinical pregnancy, and 47% resulted in a live birth (Table S1). In our analysis focused on air pollution exposures during spermatogenesis, there were 534 recipient male partners linked to 739 frozen embryo transfers. The median (25^th, 75^th) age and BMI of male partners was 41.0 (37.0, 45.0) years and 27.6 (25.1, 30.1) kg/m², respectively (Table S2).

Table 2.

Demographic, reproductive, and residential characteristics of oocyte recipients and gestational carriers at first embryo transfer

Recipients (n=1,353) and Gestational Carriers (n=41)	N (%) or Median (25th, 75th)
Demographic Characteristics
Age (years)	42.2 (38.9, 44.8)
Missing	5 (0.3)
Race
White	974 (69.9)
Black	187 (13.4)
Asian/South Asian/Southeastern Asian	123 (8.8)
Hispanic/Latino	53 (3.8)
Other	3 (0.2)
Missing	27 (1.9)
BMI (kg/m2)	23.7 (21.3, 27.3)
Missing	79 (5.6)
Smoking status
No	1264 (90.7)
Yes	56 (4)
Missing	74 (5.3)
Census-level Characteristics
Median household income	$74,678 (56,979, 99,357)
Missing	12 (0.8)
Median value of owner-occupied households	$241,700 (169,800, 355,200)
Missing	15 (1)
Percent of households living below poverty level	8.2 (4.6, 13.7)
Missing	12 (0.8)
Transfer Information
Year of embryo transfer
2008-2010	275 (19.7)
2011-2013	549 (39.4)
2014-2016	300 (21.5)
2017-2019	270 (19.4)
Season of embryo transfer
Spring (Mar-May)	337 (24.2)
Summer (Jun-Aug)	361 (25.9)
Autumn (Sept-Nov)	351 (25.2)
Winter (Dec-Feb)	345 (24.8)
Number of embryo transfers per recipient a
1	706 (50.7)
2-3	583 (41.8)
4-9	105 (7.5)
Number of prior births
0	1,062 (76.2)
1-3	332 (23.8)
Prior autologous IVF transfers b
0	750 (53.8)
1-9	574 (41.2)
Missing	68 (4.9)
Prior donor IVF transfers b
0	1,183 (84.9)
1-11	156 (11.2)
Missing	55 (3.9)
Uterine factor infertility	236 (16.9)
Fibroids	176 (12.6)
Recurrent pregnancy loss	90 (6.5)
Tubal factor infertility	253 (18.2)
PCOS or other ovulatory dysfunction	50 (3.6)
Estimated ambient residential air pollution exposure
During three months before endometrial preparation
NOx (ppb)	32.1 (20.6, 48.6)
CO (ppm)	0.5 (0.4, 0.7)
PM2.5 (μg/m3)	9.2 (8.2, 10.5)
EC (μg/m3)	0.6 (0.5, 0.8)
NH4 (μg/m3)	0.5 (0.3, 0.8)
NO2 (ppm)	18.9 (13.9, 23.5)
NO3 (μg/m3)	0.8 (0.5, 1.4)
O3 (ppm)	0.04 (0.03, 0.05)
OC (μg/m3)	2.2 (1.7, 2.7)
PM10 (μg/m3)	16.9 (14.9, 19.9)
Three weeks prior to endometrial preparation
NOx (ppb)	31.5 (18.9, 48.6)
CO (ppm)	0.5 (0.4, 0.7)
PM2.5 (μg/m3)	9.1 (7.8, 10.6)
EC (μg/m3)	0.6 (0.4, 0.8)
NH4 (μg/m3)	0.5 (0.3, 0.7)
NO2 (ppm)	19.1 (13.6, 23.7)
NO3 (μg/m3)	0.8 (0.5, 1.3)
O3 (ppm)	0.04 (0.03, 0.05)
OC (μg/m3)	2.1 (1.6, 2.7)
PM10 (μg/m3)	16.7 (14.3, 19.9)

^aThis includes 15 gestational carriers linked to more than one embryo transfer. Only 1 of the 15 gestational carriers were linked to more than one recipient couple.

^bAmong the oocyte recipient couples. Data on prior cycles were not collected for gestational carriers.

Within donors and recipients, the correlations across the ten air pollutants were generally positive and the strongest correlations (r > |0.6|) were observed between CO, NOx, and NO₂ as well as between PM_2.5, NH₄, and PM₁₀ (Figure S3). Maximum temperature was strongly correlated with NO₃ (range: −0.72 to −0.65) and O₃ (range: 0.65 to 0.77) in all exposure periods. Correlations between pollutants across exposure periods in donors and recipients (i.e., folliculogenesis and three months prior to endometrial preparation) were generally weak with the exception of particulate matter and some of its components, which were the pollutants with the strongest time trends in our study. The moderate correlations observed for PM_2.5 were attenuated after stratifying by year (Table S3).

During folliculogenesis, higher CO and NO_x exposure in oocyte donors was associated with lower odds of a positive pregnancy test in the recipients (aOR per IQR increase in CO= 0.86 95% CI 0.77, 0.96; aOR per IQR increase in NOx= 0.92 95% CI 0.84, 1.00) and higher EC exposure was associated with lower odds of clinical pregnancy (aOR per IQR increase= 0.90 95% CI 0.82, 0.98) and live birth (aOR per IQR increase= 0.87 95% CI 0.79, 0.96) (Figure 1 and Table S4). During ovarian stimulation, higher CO exposure was associated with lower odds of a positive pregnancy test (aOR per IQR increase in CO= 0.88 95% CI 0.79, 0.98) and higher PM₁₀ exposure was associated with slightly lower odds of live birth (aOR per IQR increase in PM₁₀= 0.90 95% CI 0.80, 1.00). In recipients, during the three weeks prior to embryo transfer (i.e., endometrial preparation), higher exposure to PM_2.5 (aOR per IQR increase= 1.17 95% CI 1.03, 1.33) and PM₁₀ (aOR per IQR increase= 1.12 95% CI 1.00, 1.24) was associated with higher odds of clinical pregnancy. No associations were observed between air pollution exposures during the three-month period prior to endometrial preparation and clinical outcomes.

Figure 1.

Associations between exposure to air pollution during folliculogenesis, controlled ovarian stimulation, three months prior to endometrial preparation, and three weeks prior to embryo transfer with positive pregnancy test, clinical pregnancy, and live birth outcomes.

Donor exposure models (during folliculogenesis and controlled ovarian stimulation) were adjusted for donor age, education, race, year and season of oocyte retrieval cycle, percent of households in the oocyte donor’s census tract living below the federal poverty line, and average relative humidity during folliculogenesis and controlled ovarian stimulation.

Recipient and gestational carrier exposure models (during the three months before endometrial preparation and the three weeks before embryo transfer) were adjusted for recipient age, race, year and season of embryo transfer, percentage of households in the recipient’s census tract living below the federal poverty line, and average relative humidity during the three months prior to endometrial preparation and three weeks prior to embryo transfer in respective models.

IQRs for pollutants were standardized for all exposure periods as follows: NOx (30.0 ppb), CO (0.5 ppm), PM_2.5 (3.0 μg/m³), EC (0.3 μg/m³), NH₄ (0.5 μg/m³), NO₂ (9.0 ppm), NO₃ (0.8 μg/m³), O₃ (0.01 ppm), OC (1.2 μg/m³), and PM₁₀ (5.0 μg/m³).

With regards to pregnancy loss, higher exposure to NO_x during folliculogenesis and ovarian stimulation was associated with lower odds of pregnancy loss (aOR per IQR increase during folliculogenesis= 0.89 95% CI 0.79, 0.99; aOR per IQR increase during ovarian stimulation= 0.88 95% CI 0.79, 0.98) and spontaneous abortion (aOR per IQR increase during folliculogenesis= 0.87 95% CI 0.75, 1.01; aOR per IQR increase during ovarian stimulation= 0.85 95% CI 0.73, 0.99) (Figure 2 and Table S4). During folliculogenesis, higher exposure to EC was associated with a higher odds of pregnancy loss (aOR per IQR increase= 1.16 95% CI 1.02 1.32). During ovarian stimulation, higher exposure to NH₄ was associated with slightly higher odds of spontaneous abortion (aOR per IQR increase= 1.25 95% CI 1.00, 1.57) and higher exposure to PM₁₀ was associated with higher odds of pregnancy loss (aOR per IQR increase= 1.18 95% CI 1.02,1.36) and spontaneous abortion (aOR per IQR increase= 1.23 95% CI 1.04,1.47). None of the air pollution exposures during the three months or three weeks prior to embryo transfer were associated with pregnancy loss or spontaneous abortion in the recipients.

Figure 2.

Associations between exposure to air pollution during folliculogenesis, controlled ovarian stimulation, three months prior to endometrial preparation, and three weeks prior to embryo transfer with pregnancy loss and spontaneous abortion.

Pregnancy loss was examined among a sub-sample of 1,745 embryo transfers that resulted in a positive pregnancy test. Pregnancy loss was defined as any positive pregnancy that did not result in a live birth (e.g. biochemical loss, ectopic pregnancy, miscarriage, elective termination & stillbirth). Spontaneous abortion was examined among the sub-sample of 1,495 cycles that resulted in a clinical pregnancy. Spontaneous abortion was defined as any clinical pregnancy that experienced pregnancy loss before 20 weeks’ gestation (e.g. miscarriage).

Donor exposure models (during folliculogenesis and controlled ovarian stimulation) were adjusted for donor age, education, race, year and season of oocyte retrieval cycle, percent of households in the oocyte donor’s census tract living below the federal poverty line, and average relative humidity during folliculogenesis and controlled ovarian stimulation.

Recipient and gestational carrier exposure models (during the three months before endometrial preparation and the three weeks before embryo transfer) were adjusted for recipient age, race, year and season of embryo transfer, percentage of households in the recipient’s census tract living below the federal poverty line, and average relative humidity during the three months prior to endometrial preparation and three weeks prior to embryo transfer in respective models.

IQRs for pollutants were standardized for all exposure periods as follows: NOx (30.0 ppb), CO (0.5 ppm), PM_2.5 (3.0 μg/m³), EC (0.3 μg/m³), NH₄ (0.5 μg/m³), NO₂ (9.0 ppm), NO₃ (0.8 μg/m³), O₃ (0.01 ppm), OC (1.2 μg/m³), and PM₁₀ (5.0 μg/m³).

Among the sub-sample of frozen embryo transfers that used fresh semen, higher PM₁₀ exposure during spermatogenesis was associated with higher odds of pregnancy loss (aOR per IQR increase in PM₁₀= 1.45 95% CI 1.09, 1.92) and slightly lower odds of live birth (aOR per IQR increase in PM₁₀= 0.81 95% CI 0.64, 1.02) (Figure 3 and Table S5). The associations of other pollutants averaged over spermatogenesis in relation to IVF outcomes were largely null with wide confidence intervals. Associations between oocyte donor and recipient air pollution exposures in this sub-sample were largely consistent with findings from the larger analytic sample, though often attenuated (Table S5).

Figure 3.

Associations between exposure to air pollution during spermatogenesis with positive pregnancy test, clinical pregnancy, and live birth (Panel A) and pregnancy loss and spontaneous abortion (Panel B) among the sub-sample of frozen embryo transfer cycles using fresh semen.

Models were adjusted for male partner age, year and season of oocyte thaw, percent of households in the recipient male partner’s census tract living below poverty, and average relative humidity during spermatogenesis. The sample size for Panel A was 793 embryo transfer cycles. The sample size for Panel B was 504 embryo transfer cycles that resulted in a positive pregnancy test for the outcome of pregnancy loss and 422 embryo transfer cycles that resulted in a clinical pregnancy for the outcome of spontaneous abortion. IQRs for pollutants were standardized for all exposure periods as follows: NOx (30.0 ppb), CO (0.5 ppm), PM_2.5 (3.0 μg/m³), EC (0.3 μg/m3), NH₄ (0.5 μg/m³), NO₂ (9.0 ppm), NO₃ (0.8 μg/m³), O₃ (0.01 ppm), OC (1.2 μg/m³), and PM₁₀ (5.0 μg/m³).

Temperature- and multipollutant-adjusted model estimates were consistent with our main results, though confidence intervals were wider, likely due to multicollinearity of maximum temperature with many of the air pollutants (Figure S4 and Table S6). In models evaluating effect modification by transfer type (i.e., fresh or frozen), the negative associations between CO, EC, and OC exposure during folliculogenesis and live birth were most pronounced among fresh embryo transfers (Figure S5 and Table S7) although most other effect estimates were similar in direction and magnitude comparing fresh and frozen embryo transfer cycles. There was no strong evidence suggesting non-linear associations between any of the air pollutants, during any of the five exposure periods, and live birth.

Discussion

This retrospective cohort study conducted among non-identified vitrified oocyte donors and oocyte recipients observed that higher exposure to EC during folliculogenesis and PM₁₀ during controlled ovarian stimulation in oocyte donors was associated with lower probability of live birth and higher risk of pregnancy loss. Higher exposure to CO and NO_x during folliculogenesis and controlled ovarian stimulation was also associated with lower probability of a positive pregnancy test. In contrast, recipient exposure to air pollution in the three weeks to three months prior to embryo transfer was generally not associated with clinical outcomes. Among the male partners, PM_2.5 and PM₁₀ exposure during spermatogenesis was associated with a higher probability of pregnancy loss and spontaneous abortion.

This is the first study to utilize a novel non-identified vitrified oocyte donor model to evaluate the independent associations between oocyte donor and recipient air pollution exposures and fertility outcomes. A similar study, from Wu and colleagues, sought to parse out the differential impacts of PM_2.5, PM₁₀, O₃, NO₂, SO₂, and CO during folliculogenesis and endometrial development among 20,835 women in northern China who underwent autologous IVF with 11,787 fresh and 17,676 FET cycles.⁵ Among the FETs, higher exposure to SO₂ during all time periods, including the 85-day period before oocyte retrieval and the 30-day period before embryo transfer, was associated with lower probability of live birth; however, none of the other pollutants were associated with live birth. Among fresh embryo transfers, higher exposure to CO and O₃ during the 85-day period before oocyte retrieval was associated with a lower probability of live birth. The authors interpreted the discrepant findings between fresh and frozen embryo transfer cycles as suggestive that SO₂ may have an adverse impact on frozen embryos. This is in line with what was concluded by a different study from China which found that ambient SO₂ levels at the IVF clinic were associated with lower live birth rates following frozen, but not fresh, embryo transfers.¹⁹ Since we did not evaluate SO₂ levels in our study, it is hard to directly compare our results; however, it is notable that, similar to our findings, PM_2.5, PM₁₀, O₃, NO₂, and CO exposures in the recipient during the month prior to embryo transfer in FET were not associated with IVF outcomes. While we also observed a negative association between exposure to CO during folliculogenesis and pregnancy rates, consistent with their findings in fresh transfers, Wu and colleagues did not observe any associations with PM₁₀.

More recently, a retrospective cohort study conducted by Leathersich and colleagues utilized 3,659 autologous FETs contributed by 1,835 patients undergoing IVF in Western Australia (2013-2021) to evaluate the associations of city-wide average levels of PM_2.5, PM₁₀, NO, NO₂, SO₂, O₃ and CO prior to oocyte retrieval and embryo transfer on IVF outcomes. Their study found that higher exposure to PM_2.5 during folliculogenesis and PM₁₀ during ovarian stimulation was associated with lower odds of live birth. No associations were observed for PM_2.5 or PM₁₀ exposures prior to embryo transfer and no associations observed for the other gaseous pollutants.⁶ While our study did not observe any associations between PM_2.5 exposure during folliculogenesis and live birth, we did observe negative associations between EC exposure (a component of PM_2.5) during folliculogenesis and PM₁₀ exposure during ovarian stimulation and live birth, indicating agreement with their study findings. Moreover, we observed that PM₁₀ exposure during spermatogenesis was associated with lower odds of live birth. Given that spermatogenesis and folliculogenesis overlap in fresh embryo transfers, our study findings somewhat align to those of Leathersich and colleagues, though it would be important for more studies to isolate these periods to more clearly define which exposures are detrimental to fertility. Of important note, differences in findings between our study and these two previous studies may be due to a multitude of factors including the study population (autologous vs. donor oocyte patients), differences in geographies and exposure concentrations (i.e. median PM_2.5 value in Northern China=78.6 μg/m³ vs. Western Australia=8.0 μg/m³ vs. 9.4 μg/m³ within the RBA IVF cohort), statistical analyses (multi-pollutant vs. single pollutant models and choice of covariates), and exposure assessment (stationary monitor(s) in the Chinese and Australian studies vs. spatio-temporal models in our study).

Primarily, epidemiologic evidence on associations between air pollution exposures and IVF outcomes has been generated by studies conducted among women undergoing autologous IVF. A systematic review and meta-analysis of 14 observational studies found that during the 85 days before oocyte retrieval PM_2.5, PM₁₀, and NO₂ exposures were associated with a lower probability of biochemical pregnancy, CO exposure was associated with lower clinical pregnancy rates, and PM_2.5 and CO exposures were associated with lower live birth rates. Their results also suggested that this 85-day period preceding oocyte retrieval was perhaps the most sensitive window of air pollution exposures as it demonstrated the most consistent negative associations with IVF outcomes.³ Similar to the conclusions from the meta-analysis, our study also found that higher NOx (which includes NO₂) exposure during folliculogenesis and ovarian stimulation was associated with lower odds of a positive pregnancy test. Though imprecise, we also found a negative association between CO exposure and clinical pregnancy. While we did not observe any associations between PM_2.5 exposure during folliculogenesis or ovarian stimulation and these outcomes, EC (a component of PM_2.5) exposure during folliculogenesis and PM₁₀ exposure during ovarian stimulation were negatively associated with odds of live birth. The majority of studies included in the meta-analysis by Liu and colleagues included fresh transfer cycles only, which may account for some of the differences in findings with our study as some evidence has suggested that vitrification (of both oocytes and embryos) may increase susceptibility to the detrimental impacts of air pollution.¹⁹ More research is needed to clarify the associations between exposure to air pollution during oogenesis and IVF outcomes among donor oocyte and autologous IVF cycles, particularly regarding the possible heightened susceptibility of frozen embryos given their increasing usage over fresh transfers worldwide.

The mechanisms by which air pollution exposure in females affects IVF outcomes remain unclear, though animal studies have suggested some primary pathways. Carbon compounds such as CO and EC and diesel exhaust particles (which include PM_2.5 and NO_x) are environmental exposures that have been shown to exhibit estrogenic properties that affect the hypothalamus-pituitary-ovarian axis,^20,21 potentially adversely affecting folliculogenesis. Animal models have also suggested that higher air pollution, particularly PM_2.5, exposure (which includes EC as a constituent), leads to apoptotic granulosa cells and oocytes.²² Greater rates of apoptosis, likely occurring via ovarian oxidative stress and inflammation, may result in poorer quality oocytes, which in turn develop into poorer quality embryos with lower chance of survival.²³ While experimental evidence in animals has shown PM_2.5 exposure to be associated with morphological alterations to the uterus that may result in impaired embryonic implantation,^24,25 our study did not provide strong evidence to support this pathway. Further research is planned, however, to investigate additional windows of susceptibility in the recipients, including the peri-implantation window, 10 days following embryo transfer, which may lend additional insight into the role of ambient air pollution on endometrial receptivity.

Our non-identified vitrified oocyte donor model also enabled the assessment of the independent associations between air pollution exposures during spermatogenesis and IVF outcomes. Among the frozen embryo transfer cycles that utilized fresh partner semen, we observed a negative association between PM₁₀ exposure during spermatogenesis and odds of live birth. Moreover, higher exposure to PM_2.5 and PM₁₀ during spermatogenesis was associated with higher odds of pregnancy loss and spontaneous abortion. Currently, there is a wealth of literature linking air pollution exposure, in particular PM_2.5 and PM₁₀, to lower sperm concentration, count, and motility^26,27; yet, extending these findings to clinical outcomes has been challenging since no single or set of semen parameters has been shown to be predictive of IVF outcomes.^28–30 In our dataset, we lacked information on the quality of sperm that was used for donor oocyte insemination, which limited our ability to directly investigate if these parameters mediated the association. Beyond the influence of air pollution on traditional semen parameters, evidence also suggests that heightened PM exposure may influence sperm DNA fragmentation, aneuploidy, and methylation,^31,32 all of which could be pathways through which paternal air pollution exposure reduces the probability of live birth and increases the risk of pregnancy loss. Interestingly, our findings are concordant with a study from India, which found a higher frequency of spontaneous abortion, stillbirth, and neonatal death in wives of traffic policemen (n=259) compared to controls matched on age and socioeconomic status (n=200).³³ Given the limited evidence on paternal exposures to air pollution and clinical fertility endpoints, more research is needed to better elucidate these associations, particularly studies leveraging unique populations such as those with highly discordant maternal/paternal occupational exposure to air pollution or those undergoing donor-oocyte IVF.

The primary strength of this study was our ability to utilize a non-identified vitrified oocyte donor model to isolate the independent associations of air pollution on clinical IVF outcomes among oocyte donors, female recipients, and male partners. By virtue of the fact that our study was based in Atlanta, GA, our cohort included a racially and demographically diverse pool of oocyte donors and recipient couples (including gestational carriers) who resided in both urban and rural settings throughout the Southeastern United States. Our air pollution exposure assessment used a combination of a coarser 12 km² spatial resolution CMAQ model with a finer resolution 250 m spatial resolution R-LINE model to estimate exposures to ten ambient air pollutants, including PM_2.5 constituents. Additionally, we ran several sensitivity analyses to test the robustness of the findings from our main analysis.

Our study findings should be interpreted in light of study limitations. Our estimates of daily residential exposure to ambient air pollution may be subject to exposure misclassification, particularly since we lacked information on daily activities and occupational exposures. However, the spatio-temporal models we used represented the best available information on air pollution and incorporated all existing observations. Moreover, the use of outdoor ambient exposures is valuable because regulation typically focuses on these concentrations. Numerous studies have also documented moderate to strong correlations between estimated ambient and measured personal exposures.^34,35 It is possible that chance findings occurred in our analysis as we ran multiple regression models exploring associations between ten air pollutants during four exposure windows in relation to five outcomes without adjusting p-values. Our single pollutant model estimates may be confounded by other air pollutants as these seldom occur in isolation; however, our sensitivity analyses which further adjusted for other pollutants suggested similar associations. Future work is planned to further evaluate mixture modelling of air pollutants and their combined impact on IVF outcomes. The external validity of our findings may be limited as the oocyte donors were young and healthy whereas the recipient couples were sub-fertile, older, and of higher socioeconomic status; however, our donors and recipients had a wide range of air pollution exposures and a high percentage of racial minorities which enhances our generalizability. Previous investigations into environmental exposures such as smoking also confirm that results from donor oocyte populations can be generalizable to other infertile couples. Finally, since this was an observational study, there remains the possibility of residual or unmeasured confounding.

Conclusions

In this cohort of non-identified vitrified oocyte donors and recipient couples, we observed negative associations between ambient exposure to EC, CO, and NO_x during folliculogenesis in oocyte donors and clinical IVF outcomes. Male partner exposure to PM₁₀ and PM_2.5 during spermatogenesis was associated with lower odds of live birth and higher odds of pregnancy loss. Female recipient exposures to ambient air pollution prior to embryo transfer had very little associations with IVF outcomes. Our study findings suggest that air pollution (particularly traffic-related air pollution) exposures to oocytes during folliculogenesis may be particularly important and one of the most critical pathways underlying the observed associations between preconception ambient air pollution and reduced fertility. Paternal preconception exposure to air pollution during spermatogenesis may also play an important role in pregnancy loss, although additional research is needed.

Highlights.

• Air pollution is associated with reduced fertility, yet windows of susceptibility remain unclear.

• We examined air pollution and outcomes of vitrified donor oocyte in vitro fertilization (IVF).

• Air pollution exposures to the oocyte and sperm adversely impacted IVF outcomes.

• Exposure to air pollution in the female recipient was generally not associated with outcomes.

Data Sharing Statement:

Data will be shared upon reasonable request.