Association between in vitro fertilization success rate and ambient air pollution: a possible explanation of within-year variation of in vitro fertilization success rate.

OBJECTIVE: To evaluate patterns in air pollution concentrations and in vitro fertilization (IVF) success rates using data from a large, long-term clinical database.
METHODS: We conducted a retrospective cohort study investigating South Korean women who pursued IVF and embryo transfer (IVF-ET) between 2011 and 2017. Hourly concentrations of air pollutants measured at 318 air quality monitoring sites in South Korea between 2011 and 2017 were obtained from the National Institute of Environmental Research. Monthly trends in pregnancy rates and concentrations of air pollutants were assessed.
RESULTS: A total of 34,427 IVF-ET cycles in 18,478 patients were analyzed. The mean age of women at the time of IVF-ET was 36.6 years. The clinical pregnancy rate in the IVF-ET cycle was 30%. Analysis of pregnancy failure rates by month showed that IVF-ET failure rates tended to be higher in March and April and lower in July and August. Concentrations of air pollutants including particulate matter (PM) less than 10 μm in diameter, PM less than 2.5 μm in diameter, sulfur dioxide, nitrogen dioxide, and carbon monoxide were highest in March and April and lowest between July and September.
CONCLUSION: Within-year variations were similar between IVF-ET failure rates and air pollution concentrations based on analysis of a large, long-term database. Specifically, IVF-ET success rates were highest when PM concentrations were lowest. Further studies are warranted to examine the mechanisms accounting for the association between IVF success and air pollutant exposure.

Introduction

For decades, there has been growing awareness that ambient air pollution compromises human health. Particulate matter (PM) and ground-level ozone (O3) are the most problematic pollutants in terms of harm to human health, followed by benzopyrene (an indicator for polycyclic aromatic hydrocarbons [PAHs]) and nitrogen dioxide (NO2) [1]. Of particular concern is airborne PM less than 2.5 μm in diameter (PM2.5), which has a higher surface area to mass ratio than PM less than 10 μm in diameter (PM10) and can directly reach the lung alveoli. PM2.5 has been shown to activate multiple pathophysiologic processes, which, in turn, may contribute to various health problems such as cardiovascular disease [2], stroke [3], respiratory diseases [4,5], and infertility [6].

There are several mechanisms by which air pollution may impact human health. Specifically, PAHs and heavy metals (e.g., Cu, Pb, and Zn) contained in PM may affect health by disrupting endocrine activity [7,8]. Additionally, PM contains compounds with estrogenic, antiestrogenic, and antiandrogenic activities that can affect gonadal steroidogenesis and gametogenesis [7]. Air pollutants can also induce DNA modification through the formation of DNA adducts, leading to changes in gene expression or the appearance of epigenetic mutations or modifications such as an alteration of DNA methylation [9,10]. These mechanisms can affect all health functions, especially reproduction, in which gonadal endocrine function plays a critical role and genetic material is passed down to offspring.

Previous studies have provided evidence that both short- and long-term exposure to particulate air pollution significantly impact female reproductive function. Acute preconception exposure to diesel exhaust particles and chronic exposure to PM2.5 present in ambient air were implicated in the disruption of inner cell mass and trophectoderm lineage segregation at the blastocyst stage [11,12]. Defective embryonic development after implantation resulted in increased numbers of implantation failures, decreased numbers of viable fetuses, and higher rates of miscarriage [13,14]. Available literature demonstrates poor reproductive outcomes such as preeclampsia and preterm delivery in humans after exposure to traffic pollution and diesel exhaust during pregnancy [6].

Several epidemiologic studies have investigated the association between air pollution and in vitro fertilization (IVF) treatment outcomes; however, these studies included small numbers of patients in limited geographic areas [15-17]. In addition, the largest study only analyzed PM10 as the primary exposure, which has less health impact compared to that of PM2.5 [17].

We performed this study to generate a hypothesis about the possible association between air pollution and IVF treatment outcome. To accomplish this, we evaluated patterns in air pollution concentrations and IVF success rates using a large, and long-term clinical database.

Materials and methods

1. Data

Our study population comprised infertile women who underwent 1 or more fresh IVF and embryo transfer (IVF-ET) cycles at the Fertility Centre CHA Hospital between January 2011 and December 2017. This study included all fresh IVF cycles in women aged 20–44 years performed during the study period. Donor oocyte IVF cycles were excluded from this study. Patients' baseline serum anti-Müllerian hormone (AMH) level, body mass index (BMI), and additional laboratory data were obtained from medical records.

2. Fresh in vitro fertilization and embryo transfer procedures

For fresh IVF-ET, controlled ovarian stimulation was initiated with a daily injection of gonadotrophins that was individualized for the woman's age, ovarian reserve, and previous ovarian response to gonadotrophins during stimulation cycles. The dose of gonadotrophin was adjusted after 4 days of stimulation according to each patient's ovarian response, as assessed through ultrasonographic monitoring of follicular growth and serum estradiol levels. Administration of cetrorelix acetate (Cetrotide®; Merck–Serono, Seoul, Korea) was started on the day when the lead follicle reached 13–14 mm in diameter and continued until at least 2 follicles reached 18 mm in diameter, at which point recombinant human chorionic gonadotropin (hCG, Ovidrel®; Merck–Serono) was administered to trigger ovulation. Transvaginal ultrasound-guided oocyte retrieval was performed 35–36 hours after hCG administration. Patients underwent either a cleavage stage or a blastocyst transfer at 3–5 days after oocyte retrieval. The number of embryos transferred was determined based on the embryo quality and the age of the patient, in accordance with the national guideline from the Ministry of Health and Welfare in Korea [18]. Starting on the day of oocyte retrieval, luteal support with daily transvaginal or intramuscular progesterone was provided until either a negative pregnancy test (serum β-hCG level <5 mIU/mL) or the pregnancy reached 8 weeks of gestation. Clinical pregnancy was defined as the confirmation of a gestational sac in the uterine cavity by transvaginal ultrasound at 5–6 weeks gestation.

3. Air pollutant exposure assessment

Hourly concentrations of PM10, NO2, sulfur dioxide (SO2), carbon monoxide (CO), and O3 measured at 318 air quality monitoring sites in South Korea during 2011–2017 were obtained from the National Institute of Environmental Research. A pilot analysis of the data showed consistent air pollutant concentrations among the monitoring sites. Finally, we calculated the daily average concentrations of the air pollutants across the 318 sites. Raw data for PM2.5, and consequently its daily averages, were missing for nearly 40% of the time points from 2015 to 2017, whereas raw data for all pollutants other than PM2.5 were missing for less than 5% of the time points and their daily averages were available for the entire research period (2011–2017).

4. Statistical analyses

Unlike the air pollutant concentrations, it was difficult to identify the within-year variations in IVF failure rates from the raw data. Let Y(t) be an observed value, such as failure rate or PM10 concentration, at t∈[0,T) within the ith year (i=1,⋯,N). To uncover within-year variations possibly contained in Y(t), we assumed a simple additive structure:

where ε was a year-specific effect indexed by i, α(t) denoted the within-year variation, and ε(t) was random error from N(0,σ2) at a given t. Here, T was a given constant, such as T=12 or 365 for monthly or daily reports Y(t), respectively. Our primary goal was to estimate and compare the within-year variations α(t) for both failure rate and air quality.

To this end, we first estimated the year-specific effect, determined by , and the error variance, determined by . We then computed.

Notice that Z(t) was a standardized estimate of α(t)+ε(t). Finally, the seasonal variation α(t) was estimated by applying B-spline smoothing to Z(t) [19].

Results

We analyzed the outcomes of 34,427 IVF-ET in 18,478 patients by applying machine learning approaches. The machine learning approaches refer to a collection of statistical methods that uncover informative signals in large-scale data without relying on stringent model assumptions. Table 1 summarizes the clinical characteristics of the treatment cycles in the study population. The mean patient age at the time of IVF-ET was 36.6 years. The average BMI was 21.4 kg/m2 and the mean serum AMH level was 3.0 ng/mL. An average of 9 oocytes was retrieved. The clinical pregnancy rate (CPR) per transfer was 29.6% (10,177/34,427 cycles). There were no significant seasonal differences in the patients' basal characteristics, AMH level, number of ET cases, CPR, and the percentage of poor responders (age ≥38 years or serum AMH ≤1.0 ng/mL).

Table 1

Characteristics of patients and in vitro fertilization and embryo transfer cycles at a fertility center in Seoul, Korea, during 2011–2017 (n=34,427)

Characteristics		Values	P-value
Age (yr)		36.6±4.3	0.035
	Spring	36.6±4.3
	Summer	36.5±4.3
	Fall	36.5± 4.3
	Winter	36.7±4.3
Body mass index (kg/m2)		21.41±3.03	0.615
	Spring	21.45±3.06
	Summer	21.38±3.02
	Fall	21.37±3.01
	Winter	21.43±3.03
Anti-Müllerian hormone (ng/mL)		2.77±2.87	0.322
	Spring	2.82±2.90
	Summer	2.77±2.89
	Fall	2.72±2.79
	Winter	2.78±2.90
No. of oocytes retrieved		9.16±6.93	0.102
	Spring	9.29±7.09
	Summer	9.03±6.84
	Fall	9.12±6.87
	Winter	9.19±6.92
Seasonal distribution of oocyte retrieval or frozen embryo transfer (%)
	Spring	26.1
	Summer	25.7
	Fall	25.0
	Winter	23.2
Clinical pregnancya) rate by season of oocyte retrieval or frozen embryo transfer (%)		29.6	0.738
	Spring	29.6
	Summer	29.9
	Fall	29.5
	Winter	29.1
Distribution of poor respondersb) (%)			0.565
	Spring	40.0
	Summer	39.1
	Fall	39.4
	Winter	40.0

Values are shown as mean±standard deviation or percentages.

a)Visualization of gestational sac on trans-vaginal ultrasound between 5th and 6th week of gestation; b)Age ≥38 years or anti-Müllerian hormone ≤1.0 ng/mL.

The average daily concentrations of PM10, PM2.5, NO2, SO2, CO, and O3 were 47.8 μg/m3, 25.2 μg/m3, 0.024 parts per million (ppm), 0.005 ppm, 0.514 ppm, and 0.026 ppm, respectively (Table 2). All pollutants exhibited distinct seasonal variations (Fig. 1A). The concentrations of most pollutants were lowest in summer and highest in winter, except for that of O3, which was lowest in fall and highest in spring. The IVF treatment failure rates had no distinct within-year variations during the 7-year study period (Fig. 1B).

Table 2

Concentrations of air pollutants by season at a fertility center in South Korea during 2011–2017

Characteristics		Concentration	P-value
PM10 (μg/m3)		47.8065±11.7959	<0.001
	Spring	59.7015±6.6690
	Summer	37.2329±6.4223
	Fall	40.1755±6.4689
	Winter	53.6688±9.0348
PM2.5 (μg/m3)		25.7007±5.1526	0.003
	Spring	28.8735±3.8556
	Summer	21.1215±4.1021
	Fall	23.0524±4.3771
	Winter	29.6003±2.1043
NO2 (ppm)		0.0241±0.0045	<0.001
	Spring	0.0255±0.0020
	Summer	0.0183±0.0021
	Fall	0.0242±0.0034
	Winter	0.0286±0.0025
SO2 (ppm)		0.0050±0.0010	<0.001
	Spring	0.0053±0.0004
	Summer	0.0043±0.0004
	Fall	0.0042±0.0005
	Winter	0.0061±0.0009
CO (ppm)		0.5142±0.1057	<0.001
	Spring	0.4978±0.0406
	Summer	0.4077±0.0188
	Fall	0.4952±0.0692
	Winter	0.6613±0.0668
O3 (ppm)		0.0264±0.0083	<0.001
	Spring	0.0362±0.0052
	Summer	0.0304±0.0061
	Fall	0.0219±0.0045
	Winter	0.0179±0.0033

Data are shown as mean±standard deviation.

PM10, particulate matter less than 10 μm in diameter; PM2.5, particulate matter less than 2.5 μm in diameter; NO2, nitrogen dioxide; SO2, sulfur dioxide; CO, carbon monoxide; O3, ground-level ozone; ppm, parts per million.

Fig. 1

Within-year variations in air pollutant concentrations (A) and in vitro fertilization treatment failure rates (B) throughout the study period (January 2011–December 2017).

PM10, particulate matter less than 10 μm in diameter; PM2.5, particulate matter less than 2.5 μm in diameter; SO2, sulfur dioxide; CO, carbon monoxide; NO2, nitrogen dioxide.

To investigate the possible association between CPR and air pollutant concentrations, we compared the within-year variation in CPR to that of the 5 different air pollutants, which had nearly identical seasonal patterns. The CPRs of total IVF treatment cycles were highest in July and August and lowest in February and March (Fig. 2); this pattern was the exact opposite of the air pollutant pattern. Assessment of the correlations between within-year variations in CPR and air pollutants revealed significant correlations between the within-year variations of CPR for IVF-ET cycles and the concentrations of NO2, PM10, and PM2.5 (Table 3).

Fig. 2

Comparisons of within-year variations in total in vitro fertilization failure rates and air pollutant concentrations.

Table 3

Correlation coefficient between estimated within year variation between in vitro fertilization (IVF) success rate and concentration levels of sulfur dioxide (SO2), carbon monoxide (CO), nitrogen dioxide (NO2), particulate matter less than 10 μm in diameter (PM10), and particulate matter less than 2.5 μm in diameter (PM2.5) along with P-values in parentheses

	SO2	CO	NO2	PM10	PM.2.5
IVF success rate	0.481 (0.114)	0.558 (0.059)	0.787 (0.002)	0.719 (0.008)	0.758 (0.004)

Discussion

The present cohort study of 34,427 IVF-ET cycles in South Korea between January 1, 2011, and December 31, 2017, provided compelling evidence that exposure to PM air pollution was associated with decreased pregnancy rates in women undergoing fresh IVF-ET treatments, as there were similar within-year variations between IVF-ET failure rates and air pollution levels. To our knowledge, this is the largest study to report a possible association between air pollution and IVF outcome. Previous studies have reported positive associations between exposure to PM2.5, PM10, and suspended particulates and IVF outcomes [11,12,17,19,20]. However, most of these studies included small numbers of patients and the exposure periods and sizes of PM investigated were inconsistent among the studies.

In our study, although there was no significant within-year variation in CPR, the CPR of IVF-ET tended to be higher in summer. Previous reports investigating within-year variation in IVF-ET outcomes showed inconsistent results. Wood et al. [21] conducted a 4-year retrospective analysis and reported a significant improvement in IVF outcomes in cycles performed in summer months. Stolwijk et al. [22] reported improved pregnancy rates from November to February. Finally, a recent study failed to demonstrate any significant seasonal influence on IVF outcomes [23]. In previous reports, the possible etiology of seasonal variances in IVF success rates was purported to be the length of day, which might suggest the effects of vitamin D exposure and melatonin on the female reproductive tract [24]. However, no study has assessed the possible role of air pollutants in the within-year variation of IVF outcomes. In our study, the patterns of within-year variation in IVF outcomes were opposite to those of the PM concentrations, indicating that the CPR of IVF cycles was lower when the air quality was worse. Several confounding factors should be considered in interpreting this possible association between air pollution and IVF outcome. First, the severity of air pollution could have differed among the patients' residences. However, our analyses revealed no significant differences in PM concentrations among regions across the country. Likewise, the IVF treatment outcomes did not differ by region. Second, as mentioned above, there is a possible seasonal effect of day length and vitamin D exposure. Finally, various clinical factors can influence IVF success, including patient age; the cause of infertility; and the quality of the oocyte, sperm, or embryo. Although it is important to adjust for all possible confounding factors to accurately access the association between 2 variables, this retrospective study compared the overall patterns of the variables based on a large clinical dataset. Additionally, as this is the first study to examine the relationship between within-year variation in IVF outcomes and air pollution, we expect this to be a hypothesis-generating study that provides pilot data for future studies investigating the possible association between air pollution and IVF outcome. It is hard to render a definite conclusion based on our results; however, our findings support the hypothesis that air pollution could be a major factor influencing IVF success rates. Based on the large number of patients included, the long study period, and the wide geographical region included in our study, our results showing concordance between seasonal variation in IVF outcomes and air pollution warrant additional studies.

Most previous studies investigating the association between air pollution and IVF outcomes evaluated the effect of acute, short-term exposure to air pollutants [12,17,25] and reported inconsistent results. Choe et al. [17] reported that exposure to air pollution for 9 to 11 days after ET was associated with a decreased probability of intrauterine pregnancy. Another study reported that exposure to high levels of particulate air pollution during the follicular phase, which usually lasts 2 weeks, was associated with early pregnancy loss in couples pursuing IVF [12]. Moreover, exposure to high NO2 concentrations between ovulation induction initiation and ET was associated with lower live-birth rates [26]. However, the exposure period of air pollutants is yet to be defined with respect to their effects on human fecundability. An animal study reported that intratracheally instilled carbon-14 passed through the air-blood barrier and was present at the highest concentration (0.1–1.0% of the instilled carbon-14) in the liver and spleen after 14 days [27]. A recent study using PAH labeled with radioactive iodine showed that PAH was mostly cleared within 48 hours after exposure, with less than 1% passing through the air-blood barrier or hepatobiliary system and accumulating in other internal organs [28]. A human epidemiologic study reported a significant negative impact of SO2 exposure during the second month before conception on fecundability [29]. Based on these reports, although the direct deposition rates of pollutants into the ovary and uterus are not known, it is not plausible to hypothesize that short, acute exposure to ambient air pollution during the IVF treatment cycle would affect the IVF treatment outcome. Additionally, the follicular and luteal phases are very short, continuous periods and analyzing the effect of pollution exposure during each of these phases separately may be irrelevant. It would likely be more informative to focus on the general pattern of IVF outcomes and air pollution levels during longer time periods to generate a hypothesis about the association between these 2 variables. Thus, we analyzed a large database without adjusting for any variables. The ascertainment bias may not be severe, which is commonly assumed in machine learning applications for large-scale data [30].

Our study has several limitations. Although we found similar patterns of IVF outcomes and air pollution levels, the association between these 2 variables should be further investigated by controlling for clinically important confounding factors. Furthermore, as most subjects in this retrospective study resided in a specific area, it is difficult to generalize our findings to a population-level conclusion based on our results. As the purpose of this study was to generate a hypothesis, it is impossible to elucidate any causal inference between air pollution IVF outcomes. Finally, we used the average concentrations of the air pollutants across the 318 sites rather than not the concentrations specific to the patients' residence. However, when we analyzed the concentrations of the air pollutants across the 318 sites, there were no significant differences between sites. Despite these limitations, this is the largest study analyzing data accumulated over 7 years. Moreover, this is the first study to assess the association between within-year variations in IVF success rates and air pollution levels, which generated an important hypothesis to serve as a basis for future studies.

In conclusions, this study, a distinct similarity between within-year variations in IVF failure rates and air pollution was found using a large, long-term database. The IVF success rate was highest in the summer when the PM concentration was lowest. Further studies to examine the mechanisms by which air pollutants affect human reproduction and the association between IVF success and air pollutant exposure are warranted.