Extreme ambient heat and outcomes of assisted reproduction in the continental United States, 1996 to 2018

Abstract

Background:

While extreme ambient heat has been associated with several adverse pregnancy and birth outcomes little is known regarding the link between extreme ambient temperatures and human fertility.

Objective:

To evaluate the associations between extreme ambient heat and assisted reproductive technology (ART) outcomes.

Methods:

Our analysis includes 1,087,379 patients in the continental United States who underwent 1,648,907 autologous oocyte, fresh embryo ART cycles reported to the National ART Surveillance System (1996–2018). Daily ambient temperature was derived by linking patient residential ZIP-code to a spatially refined gridded meteorology dataset. Three extreme heat metrics were created using different combinations of duration and intensity of hot days (mean daily temperature exceeding the ZIP-code-specific 98^th percentile), as well as a continuous measure of temperature during folliculogenesis (90 days prior to cycle start) and ovarian stimulation. We estimated the adjusted odds ratio of cycle cancellation and live birth associated with extreme heat using generalized estimating equations, accounting for correlation within patients and clinics.

Results:

Extreme heat during folliculogenesis and ovarian stimulation had no or modest associations with cycle cancellation or live birth; however, there was significant heterogeneity by region. Patients in the West with ≥15 days of extreme heat during folliculogenesis had 21% (confidence interval [CI] 5–41%) greater odds of cycle cancellation and 28% (CI 20–36%) lower odds of live birth compared to patients with no extreme heat in that region. Most associations were null in other regions. Nationally, a 10°F higher ambient temperature during folliculogenesis and ovarian stimulation was associated with a modest (1–2%), but statistically significant, increased odds of cycle cancellation and decreased odds of live birth.

Conclusions:

We did not find strong evidence for an association between extreme ambient heat and ART outcomes.

Graphical Abstract

Introduction.

Climate change has been recognized as potentially the greatest global threat of the 21^st century (Watts et al. 2015). The most immediate and direct impact of a changing global climate on human health is seen in the steady increase in global average temperature, and the increased frequency, intensity, and duration of extremes of heat (Watts et al. 2019). While high ambient temperatures have been associated adverse effects on pregnancy and birth outcomes (Bekkar et al. 2020), little is known regarding the link between ambient temperature and human fertility.

Demographic studies suggest that hot weather is followed by a significant decline in birth rates 8–10 months later (Barreca et al. 2018; Cho 2020), yet the drivers of this association are unclear. A study of all births in the US from 1931–2010 found that days above 80°F (26.7 °C) were associated with a large decline in birth rates 8–10 months later (Barreca et al. 2018). Moreover, while temperature at the time of conception had no effect on birth rates, hot weather in the two weeks prior had a large impact, suggesting a stronger impact of heat on reproductive health as opposed to sexual activity (Barreca et al. 2018). In men, there is a well-documented relationship between environmental heat exposures, elevated scrotal temperatures, and disrupted spermatogenesis (Durairajanayagam et al. 2015; Santi et al. 2018; Zhou et al. 2020). In women, mature follicles, prior to ovulation, are also significantly cooler than other smaller follicles, suggesting that temperature may also play a critical role during oogenesis (Grinsted et al. 1985; Hunter and Einer-Jensen 2005).

While animal literature strongly supports a link between heat stress induced by high ambient temperatures and reduced fertility, largely mediated through effects on oocyte quality (Boni 2019; Khan et al. 2020; Roth 2017; van Wettere et al. 2021), studies directly addressing this question in humans are lacking. The literature on the influence of season and ambient temperature on the day of oocyte retrieval or embryo transfer on assisted reproductive technology (ART) outcomes remains inconclusive likely due to small samples and limited temperature variation because of focus on a single clinic, in one climate region, over a limited time (Chang et al. 2005; Chu et al. 2022; Correia et al. 2022; Du et al. 2023; Farland et al. 2020; Leathersich et al. 2023; Xiao et al. 2018). Only a handful of studies have investigated ambient temperature exposures during relevant windows of gametogenesis in relation to ART outcomes, with largely mixed findings (Cheng et al. 2024; Geng et al. 2023; Matsumoto et al. 2022; Wu et al. 2023; Zhao et al. 2019). The largest study to date, which included 15,217 women undergoing ART at six fertility clinics in northern China, found no association between ambient air temperature and pregnancy outcomes (Wu et al. 2023). However, like most previous studies, their analysis was largely focused on ambient temperature, modelled continuously, rather than the occurrence of extreme heat events which may elicit more robust reproductive consequences.

Therefore, the aim of this study was to evaluate the associations of extreme ambient heat exposures during folliculogenesis and ovarian stimulation with ART outcomes using all fresh non-donor cycles performed in the US between 1996 and 2018. We hypothesized that patients who experienced an extreme heat event during folliculogenesis and ovarian stimulation would have a higher risk of cycle cancelation and lower probability of clinical pregnancy and live birth. Moreover, we hypothesized that these effects would likely depend on climate region of residence and month of cycle start due to differential acclimatization and access to heat mitigation measures.

Methods.

Study Design.

We used data from the U.S. Centers for Disease Control and Prevention (CDC) National ART Surveillance System (NASS) (CDC 2021). NASS includes information on nearly all (98%) ART cycles performed in the US and contains information on patient demographics, obstetrical and medical history, ART treatment procedures, and resultant pregnancies and births. To verify accuracy of reporting, a random sample of fertility clinics is selected annually for data validation. Discrepancy rates are typically <6% for all fields. Epidemiological research using NASS data is approved by the Institutional Review Board at the CDC. Cycles that were initially eligible for our analysis include all fresh embryo, autologous oocyte ART cycles that were started between 1996 and 2018 (the most recent year for which the high resolution temperature data was available) (n=1,947,515). Detailed exclusion criteria are provided in Figure S1. In brief, we excluded cycles from non-US citizens, patients residing outside the continental US, and patients with missing or erroneous ZIP codes, unstimulated ART cycles, cycles with unknown pregnancy outcomes, cycles with missing or implausible dates, cycles with <6 or >20 days of stimulation, and cycles with missing temperature data. Our final sample size included 1,648,907 cycles from 1,087,379 patients collected over 21 years.

Exposure Assessment.

Information on daily ambient mean (T_mean), minimum (T_min), and maximum (T_max) temperature were derived by linking the patient’s residential ZIP code and relevant dates from the NASS database (e.g., the date of cycle start and date of oocyte retrieval) to the High-resolution Urban Meteorology for Impacts Dataset (HUMID) meteorology dataset covering the conterminous United States (Newman et al. 2024). The average size of a ZIP code in the conterminous US is 82.25 square miles, although it can vary from 0.1 to 10,000 square miles. Residential ZIP codes were linked to Census defined ZIP code tabulation areas (ZCTA) from the corresponding year’s US Census shapefile to define the spatial boundaries. Patients undergoing ART prior to 2000 were linked to the 2000 US Census shapefile. The HUMID dataset is an hourly, national 1 km resolution dataset of near-surface temperature and humidity that spans 1981–2018. In brief, HUMID uses the High Resolution Land Data Assimilation System (HRLDAS)(Newman et al. 2024) to synthesize one-eighth degree hourly meteorological data from the North American Land Data Assimilation Phase 2 (NLDAS) for computation of the surface and subsurface energy and moisture fluxes (Cosgrove et al. 2003; Xia et al. 2012). NLDAS is used to drive research-grade land-surface models such as HRLDAS, which includes urban energy balance models to explicitly model urban impacts on surface meteorological conditions (Cosgrove et al. 2003; Mitchell et al. 2004). HUMID uses the HRLDAS with the single layer urban energy balance modeling configuration to model natural and urban landscapes and includes bias corrected temperatures using thousands of in situ observations. The HUMID thus accounts for the fine-scale impacts of heterogeneous land surface types on weather (e.g., rural versus urban surfaces and within-city spatial heterogeneity such as less dense suburban versus densely populated areas) (Frumkin 2002; Newman et al. 2024), which is a substantial improvement from previous research utilizing temperature measurements from weather stations, which may not fully capture exposures where people live, and coarser spatio-temporal models of temperature which cannot account for fine-scale land surface heterogeneity.

An extreme heat day was defined as a day where T_max exceeded the ZIP code specific 98^th percentile calculated across 1996–2018. Several other extreme heat definitions were also evaluated using T_min and T_mean. Three critical time windows of exposure were considered for this analysis: 1) the 3 months prior to treatment initiation (the antral phase of follicle development), 2) the first week of ovarian stimulation, and 3) the second week of ovarian stimulation (among cycles lasting 14 days or longer, n=472,128 cycles, 29%) (Broekmans et al. 2010). Periods two and three were chosen as they represent the two weeks prior to ovulation, which is the final stage of follicular development when follicles are selected and grow/mature into dominant follicles. During each of these periods we summed the total number of extreme heat days and the total number of times there were ≥2 and ≥3 consecutive extreme heat days. We also averaged the daily ambient temperature values over these three windows.

Outcome Assessment.

Our analysis focused on two primary outcomes - cycle cancellation and live birth - evaluated per initiated cycle. We also included clinical pregnancy as a secondary outcome. A cycle cancellation included any cycles that were discontinued before planned oocyte retrieval. A clinical pregnancy was defined as cycles in which a gestational sac(s) was seen on ultrasound (typically conducted 6 weeks after embryo transfer). A live birth was defined as the delivery of one or more live-born infants.

Covariates.

Individual-level covariates routinely collected as part of the NASS data included patient, age, race and ethnicity, height, weight, smoking status, gravidity, parity, number of prior miscarriages, number of prior ART cycles, infertility diagnosis/reason for ART and treatment characteristics such as use of intracytoplasmic sperm injection, day of embryo transfer, and number of embryos transferred. Frequency of missing data was generally low (<1%) except for BMI (~53%), which was only reported starting in 2007, and race/ethnicity (~26%). We obtained the estimated proportion of the population living in poverty for each ZIP code via the 5-year American Community Survey estimates (2010–2014), as a proxy for neighborhood socioeconomic status (United States Census Bureau 2015). This period was chosen as it represents the mid-point of our included study years. We also classified each patient’s state of residence into nine climatically consistent regions (Northeast, Upper Midwest, Ohio Valley, Southeast, South, Northern Rockies and Plains, Southwest, Northwest, and West) according to the National Centers for Environmental Information (NCEI 2025).

Statistical Analysis.

We examined the distributions of patient and treatment characteristics for all initiated cycles using counts and percentages as well as medians and interquartile ranges. Multivariable generalized estimating equations (GEE) with binomial distribution and logit link function were used to model the associations between extreme heat exposures and ART outcomes. These models allow women to contribute multiple ART cycles while accounting for within-person correlations in outcomes. These models also account for clustering by fertility clinic. Results are presented as adjusted odds ratios (aORs) and 95% confidence intervals (CIs) for women exposed to extreme heat compared to unexposed women. In the models evaluating mean ambient temperature during folliculogenesis and ovarian stimulation in relation to ART outcomes, results are presented as the aORs and 95% CI for a 10°F increase in mean ambient temperature, assuming a monotonic/linear relationship. Confounding was assessed based on biological relevance and descriptive statistics from our study population. Since the exposure contrasts in our study were temporal (e.g., all women in a zip code were assigned the same exposure for a given day), individual level factors associated with ART outcomes should not be confounders because they do not vary systematically with daily changes in ambient temperature. Our final models adjusted for patient age, patient age squared, month of cycle start, year of cycle start, climactic zone, percent of households below the federal poverty line, and infertility diagnoses.

Since there is evidence that exposure to higher temperatures may be more detrimental to health outcomes outside of the peak summer months and in certain climactic regions due to acclimatization,(Lee et al. 2014) we evaluated effect modification according to the month of cycle start and region of residence by adding cross product terms to the final multivariable model. Extreme heat events were only observed May through October, so our models stratified on month focused on cycles started July through November for extreme heat exposure during folliculogenesis and cycles started May through October for extreme heat exposure during the first week of ovarian stimulation to ensure an adequate number of exposed cycles were present for models to converge.

Results.

Among the 1,087,379 patients included in our analysis, the majority were nulliparous (76%) with no history of prior ART treatments (82%) (Table 1). Patients resided in every continental US state with the most common being California (12%) and New York (11%). Of the 1,648,907 initiated cycles, 10% (n=161,089) were cancelled prior to oocyte retrieval, largely due to low response (n=137,422; 85%), 34% resulted in a clinical pregnancy, and 28% resulted in a live birth (Table 2).

Table 1.

Characteristics of included patients who underwent an autologous oocyte, fresh embryo assisted reproductive technology (ART) cycle, National ART Surveillance System (1996 to 2018).

Characteristic	N (%)
Total number of patients	1,087,379
Age, years
18–29	156,749 (14%)
30–34	369,960 (34%)
35–37	235,880 (22%)
38–40	194,762 (18%)
41–43	106,423 (10%)
44–59	23,605 (2%)
Body Mass Indexa, kg/m2 (collected 2007–2018, n=515,306)
<18.5	13,667 (3%)
18.5–24.9	271,714 (53%)
25–29.9	125,238 (24%)
30–34.9	60,008 (12%)
≥35	44,679 (9%)
Race/Ethnicitya
Non-Hispanic White	516,691 (48%)
Non-Hispanic Black	50,273 (5%)
Asian	73,783 (7%)
Hispanic	55,360 (5%)
Otherb	107,180 (10%)
Missing	284,092 (26%)
Number of prior pregnancies
0	566,054 (52%)
1	254,931 (23%)
≥2	264,582 (24%)
Number of prior live births
0	818,595 (76%)
1	187,293 (17%)
≥2	78,544 (7%)
Prior ART treatments
0	886,677 (82%)
1	94,439 (8%)
2	51,452 (5%)
≥3	53,148 (5%)
Infertility diagnosis/Reason for ARTc
Male factor	390,913 (36%)
Endometriosis	134,349 (12%)
Tubal factor	211,496 (19%)
Ovulatory disorder	165,828 (16%)
Diminished ovarian reserve	208,466 (20%)
Uterine factor	51,393 (5%)
Other reason	139,415 (13%)
Unexplained	142,557 (13%)
Climate region of residenced
Northeast	367,831 (34%)
Upper Midwest	73,039 (7%)
Ohio Valley	161,360 (15%)
Southeast	158,204 (15%)
Northern Rockies and Plains	11,151 (1%)
South	103,947 (10%)
Southwest	40,628 (4%)
Northwest	30,264 (3%)
West	140,955 (13%)

Abbreviations: ART, assisted reproductive technology; BMI, body mass index.

^aAmount of missing data: 53% for BMI (only reported from 2007 onwards), 26% for race/ethnicity, and <1% for gravidity, parity, prior ART treatments and infertility diagnoses.

^bOther races included Native Hawaiian or Other Pacific Islander, American Indian or Alaska Native, and Two or More Races.

^cInfertility diagnoses are not mutually exclusive.

^dNortheast includes Connecticut, Delaware, Maine, Maryland, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, and Vermont; Upper Midwest includes Iowa, Michigan, Minnesota, and Wisconsin; Ohio Valley includes Illinois, Indiana, Kentucky, Missouri, Ohio, Tennessee, and West Virginia; Southeast includes Alabama, Florida, Georgia, North Carolina, South Carolina, and Virginia; Northern Rockies and Plains includes Montana, Nebraska, North Dakota, South Dakota, and Wyoming; South includes Arkansas, Kansas, Louisiana, Mississippi, Oklahoma, and Texas; Southwest includes Arizona, Colorado, New Mexico, and Utah; Northwest includes Idaho, Oregon, and Washington; West includes California and Nevada.

Table 2.

Characteristics of included autologous oocyte, fresh embryo transfer assisted reproductive technology (ART) cycles, National ART Surveillance System (1996 to 2018).

Characteristic	N (%)
Total number of initiated cycles	1,648,907
Total number of cancelled cycles	161,089
Low response	137,422 (85%)
High response	6,445 (4%)
Illness	1,197 (<1%)
Othera	16,025 (10%)
Clinical pregnancy (% per initiated cycle)	562,097 (34%)
Live birth (% per initiated cycle)	463,798 (28%)
Number of oocyte retrievals	1,487,818
Number of oocytes retrieved, median ± IQR	11 ± 9
Semen source
Partner	1,421,956 (96%)
Donor	61,081 (4%)
Mixed	3,619 (<1%)
Use of ICSI	1,030,540 (70%)
Total number of transfers	1,310,575
Number of embryos transferred
1	209,933 (16%)
2	569,377 (43%)
≥3	530,788 (41%)
Stage of embryo transfer
Cleavage (days 2–3)	750,229 (57%)
Blastocyst (days 5–6)	394,117 (30%)
Other	166,229 (13%)
Outcome of embryo transfer
Not pregnant	628,224 (48%)
Live birth	463,798 (35%)
Biochemical pregnancy loss	105,007 (8%)
Spontaneous abortion	89,104 (7%)
Ectopic pregnancy	11,042 (<1%)
Therapeutic abortion	5,696 (<1 %)
Stillbirth	3,432 (<1%)
Maternal death prior to birth	67 (<1%)
Unknown	4205 (<1%)

Abbreviations: ART, assisted reproductive technology; ICSI, intracytoplasmic sperm injection; IQR, interquartile range.

^aOther cancellation reasons included: withdrawal due to personal, financial, psychological, family, or other reasons (8%); inadequate endometrial response (<1%); abnormal estradiol levels (<1%); no sperm retrieved (<1%); and unknown or unspecified reason (<1%).

The mean (range) zip code-specific 98^th percentile used to define extreme heat was 81°F (27.2°C) (55–105°F; 12.8–40.6°C) for T_mean, 70°F (21.1°C) (42–91°F; 5.6–32.8°C) for T_min, and 94°F (34.4°C) (64–122°F; 17.8–50°C) for T_max (Figure S2). Since the window of folliculogenesis covers 90 days, it was not uncommon for a cycle to have at least one (33%), two (25%), or three (17%) consecutive days >98^th percentile (Table 3). In the first week of ovarian stimulation, only 7%, 4%, and 2% of cycles experienced 1, 2, or 3 consecutive days >98^th percentile, respectively. The median (25^th, 75^th percentile) ambient T_max averaged across folliculogenesis and ovarian stimulation were 71.3°F (21.8°C) (54.6, 82.1°F; 12.6, 27.8°C) and 71.4°F (21.9°C) (55.5, 83.5°F; 13.1, 28.6°C).

Table 3.

Associations of extreme ambient heat exposure (based on T_max>98^th percentile) during folliculogenesis and ovarian stimulation with risk of cycle cancellation, clinical pregnancy, and live birth following assisted reproductive technology (ART), National ART Surveillance System (1996–2018).

		Cycle Cancellation	Clinical Pregnancy	Live Birth
Folliculogenesis (90 days prior to start)	N of cycles (%)	aOR (95% CI)	aOR (95% CI)	aOR (95% CI)
No. days Tmax>98th
0 days	1104432 (67%)	1.0 (REF)	1.0 (REF)	1.0 (REF)
1–7 days	414625 (25%)	1.02 (1.00, 1.04)	0.99 (0.98, 1.00)	0.99 (0.98, 1.00)
8–14 days	97701 (6%)	1.02 (0.99, 1.04)	0.99 (0.97, 1.00)	0.99 (0.97, 1.01)
≥15 days	32139 (2%)	1.02 (0.98, 1.06)	1.02 (0.99, 1.04)	1.03 (1.00, 1.06)
No. of times Tmax>98th for 2 consecutive days
0	1240033 (75%)	1.0 (REF)	1.0 (REF)	1.0 (REF)
1–7	357294 (22%)	1.02 (1.01, 1.04)	1.01 (0.99, 1.02)	1.00 (0.99, 1.01)
≥8	51580 (3%)	0.99 (0.96, 1.02)	1.03 (1.01, 1.05)	1.04 (1.01, 1.06)
No. of times Tmax>98th for 3 consecutive days
0	1373613 (83%)	1.0 (REF)	1.0 (REF)	1.0 (REF)
1–7	253796 (15%)	1.00 (0.98, 1.02)	1.01 (1.00, 1.02)	1.01 (1.00, 1.02)
≥8	21498 (1%)	1.03 (0.98, 1.08)	1.02 (0.99, 1.06)	1.03 (1.00, 1.07)
First Week of Ovarian Stimulation
No. days Tmax>98th
0 days	1543596 (94%)	1.0 (REF)	1.0 (REF)	1.0 (REF)
1–2 days	75035 (5%)	0.99 (0.96, 1.01)	0.99 (0.97, 1.01)	0.98 (0.96, 1.00)
≥3 days	30276 (2%)	1.02 (0.98, 1.06)	1.01 (0.98, 1.03)	1.00 (0.97, 1.02)
Tmax>98th for 2 consecutive days
No	1585818 (96%)	1.0 (REF)	1.0 (REF)	1.0 (REF)
Yes	63089 (4%)	1.00 (0.98, 1.03)	1.01 (1.00, 1.03)	1.01 (0.99, 1.02)
Tmax>98th for 3 consecutive days
No	1614394 (98%)	1.0 (REF)	1.0 (REF)	1.0 (REF)
Yes	34513 (2%)	0.99 (0.96, 1.03)	1.03 (1.00, 1.05)	1.02 (0.99, 1.05)
Second Week of Ovarian Stimulation a
No. days Tmax>98th
0 days	436257 (92%)	1.0 (REF)	1.0 (REF)	1.0 (REF)
1–2 days	24036 (5%)	1.03 (0.95, 1.11)	0.99 (0.96, 1.02)	0.97 (0.94, 1.01)
≥3 days	11835 (3%)	1.02 (0.91, 1.13)	0.96 (0.92, 1.01)	0.96 (0.92, 1.00)
Tmax>98th for 2 consecutive days
No	450163 (95%)	1.0 (REF)	1.0 (REF)	1.0 (REF)
Yes	21965 (5%)	0.96 (0.88, 1.04)	0.99 (0.95, 1.02)	0.96 (0.93, 1.00)
Tmax>98th for 3 consecutive days
No	459936 (97%)	1.0 (REF)	1.0 (REF)	1.0 (REF)
Yes	12192 (3%)	0.94 (0.85, 1.05)	0.97 (0.93, 1.01)	0.96 (0.92, 1.00)

Abbreviations: aOR, adjusted odds ratio; ART, assisted reproductive technology; CI, confidence interval; REF, reference group; T_max, maximum temperature.

Models were adjusted for patient age (continuous), patient age squared (continuous), month of cycle start (Jan, Feb, Mar, Apr, May, Jun, Jul, Aug, Sep, Oct, Nov, Dec), year of cycle start (continuous), climactic zone (Northeast, Upper Midwest, Ohio Valley, Southeast, Northern Rockies and Plains, Southwest, Northwest, West), percent of households below poverty (quintiles + missing indicator), and infertility diagnoses (eight indicator variables for male factor, endometriosis, ovulatory disorder, tubal factor, diminished ovarian reserve, unexplained, uterine factor, and other infertility diagnosis).

^aOnly includes cycles with a duration of ovarian stimulation of 14 days or longer (n=472,128, 29% of cycles)

Overall, there was little evidence for an association between extreme ambient heat, defined as total or consecutive days >98^th percentile of T_max, during folliculogenesis or ovarian stimulation and odds of cycle cancellation, clinical pregnancy, or live birth (Table 3). Similar results were found using T_mean and T_min to define extreme ambient heat (Tables S1, S2). The aORs of cycle cancellation for a 10°F (~5.6°C) increase in mean ambient T_max, T_mean, and T_min during folliculogenesis was 1.00 (95% CI 0.99, 1.01) (Figure S4), 1.01 (95% CI 1.00, 1.02), and 1.01 (95% CI 1.00, 1.02). The aORs of live birth for a 10°F (~5.6°C) increase in mean ambient T_max, T_mean, and T_min during folliculogenesis were 0.98 (95% CI 0.97, 0.98) (Figure S4), 0.98 (95% CI 0.97, 0.99), and 0.98 (95% CI 0.98, 0.99). Nearly identical effect estimates were seen for mean ambient temperatures during the first week of ovarian stimulation and odds of cycle cancellation and live birth (Figure S5).

The associations between extreme ambient heat exposure during folliculogenesis and ART outcomes varied by climactic region of residence (Figure 1, Table S3). In the West, exposure to 1–7, 8–14, and ≥15 days when T_max exceeded the 98^th percentile was associated with 1.09 (95% CI 1.05, 1.14), 1.11 (95% CI 1.03, 1.20), and 1.21 (95% CI 1.05, 1.41) times the odds of cycle cancellation and 0.97 (95% CI 0.94, 0.99), 0.89 (95% CI 0.85, 0.94), and 0.72 (95% CI 0.64, 0.80) times the odds of live birth compared to patients with no exposure to extreme heat days. A similar, positive association between extreme ambient heat exposure and cycle cancellation was also seen in the Northwest (aOR: 1.59, 95% CI 1.15, 2.19 for ≥15 days of extreme heat versus none). In contrast, the aOR of live birth for exposure to ≥15 days of extreme heat during folliculogenesis was 1.09 (95% CI 1.04, 1.15) in the Northeast, 1.12 (95% CI 1.04, 1.20) in the Ohio Valley, and 1.15 (95% CI 1.07, 1.23) in the South. Results were largely null in the other climate regions. Associations between extreme heat exposures during ovarian stimulation and ART outcomes were less variable across climate regions and generally indicative of no association (Figure S3). The aOR of cycle cancellation per 10°F (~5.6°C) increase in mean ambient T_max during folliculogenesis was highly variable across region, ranging from 0.90 (95% CI 0.88, 0.93) in the West to 1.11 (95% CI 1.06, 1.16) in the Upper Midwest (Figure S4). The association between mean ambient T_max during folliculogenesis and live birth was largely negative (e.g. had an aOR below one) for seven out of 9 climate regions and was statistically significant for five regions. Results were similar (e.g. six out of 9 climate regions had an aOR below 1 and four were statistically significant), albeit attenuated, for mean ambient T_max during ovarian stimulation in relation to live birth (Figure S5).

Figure 1.

Association of number of days T_max>98^th percentile during folliculogenesis with cycle cancellation (Panel A) and live birth (Panel B) following assisted reproductive technology (ART) by US climate region (1996–2018).

Abbreviations: ART, assisted reproductive technology; CI, confidence interval. Models were adjusted for patient age, patient age squared, month and year of cycle start, percent of households in zip code below poverty, and infertility diagnoses. Northeast includes Connecticut, Delaware, Maine, Maryland, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, and Vermont; Upper Midwest includes Iowa, Michigan, Minnesota, and Wisconsin; Ohio Valley includes Illinois, Indiana, Kentucky, Missouri, Ohio, Tennessee, and West Virginia; Southeast includes Alabama, Florida, Georgia, North Carolina, South Carolina, and Virginia; Northern Rockies and Plains includes Montana, Nebraska, North Dakota, South Dakota, and Wyoming; South includes Arkansas, Kansas, Louisiana, Mississippi, Oklahoma, and Texas; Southwest includes Arizona, Colorado, New Mexico, and Utah; Northwest includes Idaho, Oregon, and Washington; West includes California and Nevada.

In general, the associations between extreme heat during folliculogenesis and ART outcomes stratified by month were largely null, consistent with the overall analysis; however, exposure to ≥15 days of extreme heat was positively associated with live birth in all months except August (with aORs ranging from 1.06 to 1.24) (Figure S6). There were no appreciable differences in the effects of extreme heat during ovarian stimulation on cycle cancellation or live birth by month (Figure S7). The associations between mean ambient T_max during folliculogenesis and odds of cycle cancellation showed some variability across month (ranging from 0.98 in January, July, and September to 1.07 in December) (Figure S8). There were no differences in the effect of mean ambient T_max during folliculogenesis on live birth by month. Higher mean ambient T_max during ovarian stimulation was associated with higher odds of cycle cancellation in cycles started August through December but lower odds of cycle cancellation in cycles started March through July (Figure S9). There were no apparent differences in the effects of mean ambient T_max during ovarian stimulation on live birth by month.

4. Discussion.

Using national data on fresh, autologous IVF cycles performed in the continental USA between 1996 and 2018, we found little to no association between extreme ambient heat and ART outcomes in most climate regions. One notable exception was in the West, where extreme heat exposure during folliculogenesis was associated with higher probability of cycle cancellation and lower probability of live birth. Higher average ambient temperature exposures were negatively associated with ART outcomes, but effect sizes were extremely small suggesting limited clinical relevance.

There have been five studies examining ambient temperature exposures averaged over relevant windows prior to and during the ART cycle in relation to clinical outcomes of ART with varied results (Cheng et al. 2024; Geng et al. 2023; Matsumoto et al. 2022; Wu et al. 2023; Zhao et al. 2019). To date, the largest study included 15,127 women undergoing their first fresh embryo transfer cycle in Northern China (2015–2020) (Wu et al. 2023). Women were linked to daily temperature data from the nearest monitoring station based on residential address and averaged over four periods. Average ambient temperature exposure (modelled in quartiles) was not associated with risk of biochemical pregnancy, clinical pregnancy, or live birth across any of the time windows; however, for average temperature exposure in the 85 days prior to oocyte retrieval, there was a suggestion of lower likelihood of live birth among women in quartiles 2, 3, and 4 versus quartile 1. A second study focused on 5,264 women undergoing 1,428 fresh and 3,836 frozen embryo transfer cycles in Hubei Province, China (2017–2020) (Cheng et al. 2024). Daily temperature was derived from Google Earth Engine and the ERA5-Land dataset, linked to patient’s home address, and averaged over seven time periods in the fresh cycles (three in frozen). Average ambient temperature exposure during ovarian stimulation was positively associated with live birth but only during cold weather (defined as <14°C) but not during comfortable (14–27°C) or hot (>27°C) weather. Average temperature exposure in the 85 days prior to oocyte retrieval was not associated with any outcomes in this study. None of the temperature exposures were associated with ART outcomes in frozen embryo transfer cycles. A third study focused on 3,452 women undergoing their first fresh (n=2,294) or frozen (n=1,158) embryo transfer in Shanghai, China (2016–2020) with similar linkage of gridded daily temperature data to patient address and eight time periods of exposure (Geng et al. 2023). While higher ambient temperature exposure in the three months prior to ovarian stimulation was positively associated with clinical pregnancy during cold weather (defined as <12°C) there was no association in comfortable (12–25°C) or hot weather (>25°C). None of the average ambient temperature exposures measured prior to embryo transfer were associated with live birth. In contrast to these largely null results, a retrospective study from Hong Kong (n=860 patients and 1,029 fresh embryo transfers) 2005–2016 found that higher mean ambient temperature during ovarian stimulation was associated with higher odds of pregnancy (Zhao et al. 2019), while a retrospective study from Osaka, Japan (n=555 infertile couples) 2017–2020 found that higher mean ambient temperature during ovarian stimulation was associated with lower probability of clinical pregnancy and live birth following single blastocyst transfer (Matsumoto et al. 2022).

Our results, which largely found no association between ambient temperature exposures prior to embryo transfer and ART outcomes, are thus consistent with the majority of existing studies. As our analyses stratified by climactic region highlight, associations between ambient temperature and ART outcomes are likely heavily influenced by underlying climate which may encompass differences in acclimatization and resources to adapt to heat (Lee et al. 2014). For example, our findings that extreme heat exposure was adversely associated with ART outcomes in the Western US states could reflect the lower prevalence of residential air conditioning (AC) in this region (Beall and McNary 2022). It is also possible that the protective associations we observed in some regions could be due to patients changing their behavior (e.g., remaining inside when it is hot or increasing air conditioning use) which would obscure the correlation between ambient and personal temperature exposure. Additional reasons for discrepant results could be due to differences in types of ART cycles, ART practice patterns, ambient temperature assessment and characterization, sample sizes, patient characteristics, and the underlying neighborhood environment.

The suggestive associations of extreme ambient heat exposure with higher odds of cycle cancellation and lower odds of live birth in Western states was in line with our initial hypothesis based on animal studies. In mammals, studies on both seasonal and induced heat stress indicate that ovarian follicles, beginning at the preantral phase, are impaired following heat exposure (Roth 2017). As such, around 2–3 months are required for the appearance of competent oocytes and restoration of fertility to normal levels (Aroyo et al. 2007; Roth et al. 2001). This suggests that heat-induced alterations, even at early stages of follicular development, can be expressed later as compromised oocyte maturation and developmental competence. The heat-induced impairment of oocyte maturation is due to a combination of adverse effects on the hypothalamic–pituitary–ovarian axis and ovary which leads to suppressed luteinizing hormone secretion, lower estradiol production by granulosa cells, reduced dominance of the preovulatory follicle, increased follicle stimulating hormone secretion, and a larger number of medium sized follicles (Roth et al. 2000; Wolfenson et al. 1995). Perturbations in the physiology of the follicle-enclosed oocyte during follicular development also lead to cellular and molecular alterations, reducing the oocyte’s competence for fertilization and development (Ferreira et al. 2016; Gendelman et al. 2010; Gendelman and Roth 2012; Roth and Hansen 2004). The most likely explanation for our null results in most climatic regions is the broad availability of indoor cooling systems in homes and offices during our time period and the limited time most adults spent outdoors (estimated at only 7–8% of the day) (Klepeis et al. 2001). If patients with higher exposure to extreme heat events are more likely to stay indoors and utilize air conditioning then this could even potentially explain the positive associations we observed with live birth in certain climate regions.

Our study was not without limitations. First, we used residence-based ambient temperature as a proxy for personal temperature exposure. We did not have information on daily activities (including time spent outdoors), residential mobility, occupational address, or use of air conditioning systems for patients and so exposure misclassification is likely. Yet, unlike many previous studies which relied on temperature measurements at an airport weather station, we minimized exposure misclassification by utilizing a highly spatially refined, gridded climate dataset, which more fully captures exposures in urban areas where people tend to live. Second, the availability of lifestyle and socioeconomic characteristics are limited in NASS data. However, our analytic approach, which primarily relies on within-location temporal contrasts of exposure, reduces the possibility of confounding by such individual-level factors since, to be confounders, these factors would have to vary in a way that is correlated with the occurrence of heat events. Because socioeconomic status and other maternal characteristics do not fluctuate with meteorology, this limitation is unlikely to bias our results; however, they could be important effect modifiers if they are associated with the ability to adapt to heat. It is also possible that other meteorological and environmental exposures could be confounding the associations as high humidity, air pollution, drought, and wildfires can co-occur with heat waves. This could potentially compound the measured effect or result in behavior change (e.g. staying indoors) that limits exposure. Of note, we did not adjust for ambient air pollution levels in our study as we hypothesized that this variable was more likely to be a mediator, rather than a confounder, of the association between ambient temperature and ART outcomes (Buckley et al. 2014). Our measure of cycle cancellation included cycles that were discontinued due to illnesses and other unrelated reasons; while these represented a minority of the cancelled cycles, it contributed to non-differential outcome misclassification, likely biasing our results towards the null. The analysis of mean ambient temperature exposure during folliculogenesis and ovarian stimulation in relation to ART outcomes assumed a monotonic/linear association between exposure and odds of the outcomes, which may not fully represent the full exposure-outcome relation. The GEE models may be biased if our data is not missing completely at random, which can occur when cluster size is informative of ART outcomes. Finally, it was beyond the scope of our current analysis to evaluate the impact of extreme heat following embryo transfer and during pregnancy on clinical endpoints as well as the impact of extreme heat prior to retrieval on outcomes of frozen embryo transfer (which could isolate the impacts of temperature during folliculogensis from that during implantation); however, future research is planned to address these important topics. The primary strength of our study was its large sample size, including almost all fresh autologous ART cycles in the continental US from 1996 to 2018, which uniquely allowed us to examine high intensity and long duration heat events and effect modification by climate region.

In summary, we observed that exposure to extreme ambient heat prior to or during ovarian stimulation for ART had limited association with probability of cycle cancellation or live birth. Results, however, were sensitive to stratification by region and timing, highlighting the importance of accounting for adaptation and acclimatization to ambient heat. The only region where a clear pattern of adverse associations between extreme heat exposure and ART outcomes emerged was in Western states, suggesting that targeted research in this region may be helpful to better understand this relationship and its clinical implications further.

Highlights.

• We evaluated 1.6 million in vitro fertilization cycles in the continental US

• Extreme ambient heat had limited associations with assisted reproduction outcomes.

• In the West, extreme heat was associated with lower probability of live birth.

• Most associations were null in other regions.

Data Sharing Statement:

Because the National Assisted Reproductive Technology Surveillance System (NASS) datasets are protected under the 308(d) Assurance of Confidentiality, all data files are considered confidential materials and safeguarded to the greatest extent possible. Due to these restrictions, access to NASS data files is limited to CDC staff and approved guest researchers. This limits our ability to directly share data and resources generated as a result of this proposal.