The effect of climate anomalies and ambient air pollution on spontaneous miscarriage in Ghana: Evidence from a national cross-sectional survey

Abstract

Background:

Climate anomalies and ambient air pollutants (AAPs) are linked to adverse pregnancy outcomes, but their impact on spontaneous miscarriage, particularly in sub-Saharan Africa, remains underexplored. This study examines how preconception exposure to climate anomalies, AAPs, and vegetation (measured by the Enhanced Vegetation Index [EVI]) affects spontaneous miscarriage among Ghana women, focusing on EVI’s role as an effect modifier.

Methods:

Data from the 2017 Ghana Maternal Health Survey were analyzed using Bayesian hierarchical linear and nonlinear clog-log models to assess the relationship between environmental exposures and spontaneous miscarriage. Meteorological data came from the Climatic Research Unit’s gridded Time Series (1960–2017), while AAP and vegetation data were sourced from NASA’s Goddard Earth Sciences Data. The analyses include interaction terms between EVI and environmental exposures.

Results:

Preconception mean temperature anomalies and carbon monoxide (CO) were strongly associated with increased miscarriage risk (hazard ratio = 4.286 for temperature anomalies and 1.013 for CO). Other AAPs, such as PM_2.5 and ozone, were confounded by climate anomalies and maternal factors. EVI modified the impact of environmental exposures, particularly reducing the effects of temperature anomalies and PM_2.5 on miscarriage risk at higher EVI levels.

Conclusions:

This study underscores the significant role of climate anomalies and AAPs in increasing spontaneous miscarriage risk in Ghana. The effect modification by EVI highlights vegetation’s potential to mitigate the impact of environmental stressors. These findings stress the need for policies addressing air pollution and climate change to protect maternal health, especially in vulnerable regions.

What this study adds.

This study provides novel evidence linking preconception exposure to climate anomalies and ambient air pollution with increased risk of spontaneous miscarriage in Ghana. It is among the first to assess multiple exposures, temperature, precipitation, PM_2.5, CO, NO₂, SO₂, Ozone, and vegetation, simultaneously. The findings highlight that mean temperature anomalies and CO consistently elevate miscarriage risk, while vegetation modifies these effects, potentially offering protective benefits. By integrating environmental, maternal, and household factors, the study advances understanding of how climate change and air pollution jointly influence reproductive health in sub-Saharan Africa, emphasizing the need for targeted environmental and public health interventions.

Introduction

Climate change is a pressing global issue, with human activities driving significant alterations in the Earth’s climate system. The global mean surface temperature has risen by approximately 1.1°C since preindustrial times, primarily due to anthropogenic factors.¹ This warming trend has led to widespread changes in climatic conditions worldwide. The increase in global temperature has resulted in more frequent and intense extreme weather events, including heatwaves, floods, and hurricanes.^1,2 These events have not only caused substantial loss of life and property damage but also severely impacted natural ecosystems.¹ Moreover, climate change has emerged as a major public health concern, affecting human health and well-being in numerous ways.^3–6 Children are particularly vulnerable to the impacts of climate change. Nearly 1 billion children face heightened risks from climate-related hazards, while an alarming 2 billion are exposed to air pollution levels exceeding World Health Organization guidelines.⁷ Sub-Saharan Africa (SSA) stands out as a region of particular concern due to its acute vulnerability to climate change impacts.^7,8 A staggering 98% of the region faces high or extreme climate risks, exacerbating existing challenges such as food insecurity and malnutrition.⁸

Increasing research highlights a significant link between extreme weather, environmental factors, and the health of mothers and children.^9,10 Pregnant women and developing fetuses are particularly vulnerable to these environmental risks due to the physiological and psychological changes that occur during pregnancy.⁹ For example, exposure to high temperatures can negatively affect the placenta, disrupt developing cells, and impair the vascular system.¹¹ Such climate-sensitive stressors, including extreme heat and air pollution, can lead to pathological outcomes like placental dysfunction, epigenetic modifications, endocrine disruptions, and oxidative stress, all of which threaten fetal growth and survival.^9,11,12

Research indicates that gestating women are more susceptible to heat stress than their nongestating counterparts due to their compromised thermoregulation and homeostasis.^9,13 Numerous studies show that exposure to extreme heat, especially during the first trimester, can adversely affect pregnancy outcomes, leading to stillbirths, low birth weights, and both induced and spontaneous abortions.^14–16 For instance, Davenport et al¹⁷ found that an increase in the number of days with temperatures exceeding 95 and 105°F heightened the risk of spontaneous and induced abortions and reduced the likelihood of healthy births. A study conducted in Hungary similarly reported that exposure to high temperatures shortly after conception increased the rate of pregnancy loss.¹⁶ Other research has also established associations between extreme heat exposure, preterm birth, and low birth weight.^18,19

Ambient air pollutants can further impact pregnancy outcomes through mechanisms such as endothelial dysfunction, endocrine disruption, immunological dysfunction, systemic inflammation, and oxidative stress.⁹ Pregnancy creates a particularly susceptible state to toxins present in air pollution due to heightened cell proliferation, organ development, and changes in fetal metabolism.²⁰ A systematic review from 2020 found that the majority of studies reported associations between exposure to PM_2.5 or ozone and increased risks of preterm birth (79% of studies) and low birth weight (86% of studies).¹⁵ Specifically, a study by Hao et al²¹ in Kansas linked ozone exposure during pregnancy to preterm birth and lower birth weight. Additionally, research in SSA found that an increase of 10 μg/m³ of PM_2.5 was associated with a 12% rise in the probability of pregnancy loss.²²

While existing studies have advanced our understanding of extreme weather events’ impact on pregnancy and birth outcomes, the conceptualization of climate change in this research remains limited. Most studies focus on isolated extreme weather events, failing to contextualize them within broader local climate patterns. In contrast, climate anomalies provide a more nuanced perspective by capturing deviations from standard weather patterns, offering insights into how temperature and precipitation variability affect health outcomes over time.^23–25 While climate change sets the broader context, climate anomalies capture specific, measurable deviations from expected weather patterns that are highly relevant for health outcomes. In this context, climate anomalies refer to measurable departures from long-term climatic averages, including temperature extremes and unusual precipitation patterns. In Ghana, recent analyses have documented significant warming trends and shifts in temperature distributions over the past 4 decades. For example, decadal temperature anomalies from 1981 to 2020 shifted from predominantly negative values in the 1980s and 1990s to strongly positive anomalies ranging between 0.2 and 0.9°C in more recent decades.²⁶ Furthermore, there is a significant research gap in examining the interplay between climate anomalies, ambient air pollution, and birth outcomes, often overlooking the potential confounding effects of vegetation coverage, which can influence both climatic factors and air quality. Understanding these complex relationships is crucial for developing effective adaptation strategies in the face of changing climatic conditions. By elucidating how sustained climatic variability affects maternal and infant health, we can better inform public health interventions and policies aimed at minimizing adverse health impacts in vulnerable populations.

In light of this critical gap, this study seeks to examine the effects of preconception climate anomalies and ambient air pollution exposures on spontaneous miscarriage among women in Ghana. Specifically, the study aims to:

1. examine the effect of preconception temperature and precipitation anomalies, and ambient air pollution exposures on spontaneous miscarriage; and

2. investigate the effect modification of vegetation (Enhanced Vegetation Index [EVI]) on the association between climate stressors (temperature, precipitation, and ambient air pollution) and spontaneous miscarriage.

Data and methods

To achieve the objectives of this study, we used a big data framework integrating demographic, health, climate, and air pollution data. The demographic data came from the 2017 Ghana Maternal Health Survey (GMHS), a nationally representative survey conducted by the Ghana Statistical Service.²⁷ The GMHS targeted women aged 15–49 years, focusing on maternal and child health, birth histories, and perceptions of antenatal and obstetric care. The data collection for the 2017 GMHS took place from 15 June to 12 October 2017. A two-stage stratified sampling method was employed, using data from the 2010 Ghana Population and Housing Census as the sampling frame. First, each of the 10 administrative regions was divided into urban and rural areas, creating 20 sampling strata. In the first stage, 900 census enumeration areas (EAs) (466 urban and 434 rural) were selected with probability proportional to size, based on the number of households recorded in the 2010 census. Within each selected EA, a household listing was carried out between January and April 2017, providing the sampling frame for the second stage. In the second stage, 30 households were randomly chosen from each EA based on the household listing or database, thereby yielding a total of 27,000 households.²⁷

Meteorological data were obtained from the Climatic Research Unit’s gridded Time Series covering the period from 1901 to 2022, with a focus on data from 1960 to 2017.²⁸ Air pollution and vegetation data were sourced from NASA’s Goddard Earth Sciences Data and Information Services Center and the Atmospheric Composition Analysis Group.^29,30 The air quality data, including monthly PM_2.5 levels, covered the period from 1998 to 2021 with varying spatial resolutions. To ensure consistency, we resampled the data using bilinear interpolation.³¹ A detailed description of the data, measures, and analytical approach can be found in the Supplementary Material; https://links.lww.com/EE/A372.

Measures

Outcome

The primary outcome of this study is spontaneous abortion, defined as pregnancies that resulted in miscarriage within the 5 years preceding the 2017 GMHS survey. Participants were asked about the cause of their miscarriages, and those classified as spontaneous were recorded as the outcome. Other pregnancy outcomes—such as live births, stillbirths, and miscarriages attributed to accidents, dietary issues, external harm, or other reasons—were coded as nonspontaneous miscarriages or no miscarriages. Ultimately, the outcome variable is binary, indicating whether a spontaneous miscarriage occurred among women between 2012 and 2017.

Exposures

The exposures of interest include temperature anomalies, precipitation anomalies, ambient air pollution, and vegetation coverage. Temperature and precipitation anomalies were calculated as deviations from historical averages for each location, focusing on the 12 months before conception. Data on ambient air pollution included PM_2.5, carbon monoxide (CO), nitrogen dioxide (NO₂), sulfur dioxide (SO₂), sulfate (SO₄), and ozone (O₃), with 12-month average exposure calculated before pregnancy.

Modifiers and confounders

In this study, we treated vegetation coverage, measured by the EVI, as a potential effect modifier. Thus, we assessed whether vegetation modifies the association between climate anomalies (temperature and precipitation anomalies) or ambient air pollution and the risk of spontaneous miscarriage. Given that our EVI data were annual means, we computed a weighted average for the corresponding exposure years if the 12-month period spanned two different years.^32–34 If the entire 12-month period fell within a single year, we used that annual mean.

We included several maternal and household factors as potential confounders in our analysis. These include maternal education, household wealth—assessed through a wealth quintiles variable based on consumer goods ownership, access to drinking water, sanitation facilities, and flooring materials—and the use of clean cooking fuels such as electricity, LPG, natural gas, and biogas. Additionally, we considered access to improved drinking water, proper sanitation, and the place of residence (urban or rural).

Analysis

We used a Bayesian hierarchical modeling approach to examine the effects of climate anomalies and ambient air pollution on spontaneous miscarriage. This approach accounts for the hierarchical nature of the data, with observations nested within sample clusters (enumeration areas) and regions. We employed the Integrated Nested Laplace Approximations framework, a computationally efficient method for Bayesian inference using latent Gaussian models.³⁵ Given the binary nature of our outcome (spontaneous miscarriage status), we applied a complementary log–log link function due to the rarity of the event. This model relaxes the symmetric assumption of typical link functions; the response curve of the complementary log–log model is asymmetrical, with a heavier tail as it approaches one.³⁶ We fitted single-exposure models to examine the relationships between climate anomalies, ambient air pollution, and spontaneous miscarriage. The first model considered only the exposures, while the second adjusted for maternal and household factors. These models were designed to examine the effects of individual exposures and the role of covariates such as maternal education, household wealth, and access to clean cooking fuels.

We then explored multiple-exposure models to assess the cumulative effects of climate anomalies, air pollution, and vegetation (EVI). Interaction models were also utilized to assess how vegetation (EVI) modifies the effects of climate anomalies and ambient air pollutants on spontaneous miscarriage.

To ensure robust findings, we conducted three sensitivity analyses. First, we examined nonlinear relationships between exposures and spontaneous miscarriage. Second, we examined region-specific effects, assuming independent residuals across administrative regions. Finally, we investigated the interaction between rural–urban residence and the exposure variables. By modeling how the effects of climate anomalies, air pollution, and vegetation differ between rural and urban populations, we aimed to understand how the rural–urban context might modify the impact of environmental factors on miscarriage. All statistical analyses and result visualization were performed in R (R Core Team, Vienna, Austria). A detailed description of the analysis and the results of these sensitivity analyses are provided in Supplementary Material; https://links.lww.com/EE/A372.

Results

The total sample for this study included 19,189 pregnancy outcomes collected from interviewed or surveyed women who provided complete data on exposure, maternal, and household characteristics. Summary statistics (Table 1) reveal that 9.4% of these pregnancies ended in spontaneous miscarriage. The majority of the women interviewed had a junior high school education or lower. Furthermore, 84.1% of the women lived in households that used unclean cooking fuels, such as biomass, coal, and charcoal. In terms of water and sanitation, 86.8% had access to improved drinking water sources, while 54.0% lived in households with improved sanitation facilities. Notably, most women resided in rural areas (55.0%). Regarding environmental exposures, the mean PM_2.5 exposure during the 12 months before conception or throughout pregnancy was 29.7 µg/m³, while the mean concentrations of carbon monoxide (CO) and ozone were 155.6 ppbv and 264.4 Dobson units, respectively.

Table 1.

Summary statistics for the study variables (N = 19,189)

	Frequency (%)/mean (SD)
Spontaneous miscarriage
No	17,391 (90.6%)
Yes	1,798 (9.4%)
Maternal and household characteristics
Highest level of education
No formal education	6,270 (32.7%)
Primary or middle school	3,406 (17.7%)
Junior high school	5,957 (31.0%)
High school	2,328 (12.1%)
Higher education	1,228 (6.4%)
Clean cooking fuel—household
Yes	3,248 (16.9%)
No	15,941 (84.1%)
Improved source of drinking water—household
Yes	16,647 (86.8%)
No	2,542 (13.2%)
Improved sanitation—household
Improved sanitation	10,354 (54.0%)
Unimproved sanitation	3,046 (15.9%)
Open defecation	5,789 (30.2%)
Wealth—household
Richest	2,747 (14.3%)
Richer	3,252 (16.9%)
Middle	3,250 (16.9%)
Poorer	3,782 (19.7%)
Poorest	6,158 (32.1%)
Place of residence—household
Urban	8,630 (45.0%)
Rural	10,599 (55.0%)
Exposures
Mean temperature anomaly	0.8 (0.3)
Maximum temperature anomaly	0.7 (0.4)
Precipitation anomaly	−0.2 (0.4)
PM2.5 (µg/m³)	29.7 (7.0)
Carbon monoxide (ppbv)	155.6 (14.8)
Nitrogen dioxide (log)	2.1 (13.0)
Sulfur dioxide (µg/m³)	0.7 (0.4)
Ozone (Dobson units)	264.4 (3.2)
Enhanced Vegetation Index (EVI)	0.3 (0.1)

Effect of climate anomalies and ambient air pollutants

The results of the clog-log regression analysis examining the associations between climate anomalies and ambient air pollutants with spontaneous miscarriage are illustrated in Figure 1. The analysis indicates that mean temperature anomalies, maximum temperature anomalies, and precipitation anomalies were significantly associated with spontaneous miscarriage. Specifically, the mean temperature anomaly was linked to an increased hazard ratio [HR = 2.415, credible interval (CrI): 2.116, 2.756] for spontaneous miscarriage, even after controlling for maternal and household factors (Model 2). Similarly, maximum temperature anomaly was associated with an increased HR (HR = 1.863, CrI: 1.661, 2.009). In contrast, precipitation anomaly was associated with a decreased HR (HR = 0.690, CrI: 0.614, 0.775) for spontaneous miscarriage after adjustment for maternal and household factors. Regarding ambient air pollutants, both PM_2.5 and Ozone showed strong associations with spontaneous miscarriage. In the adjusted model (Model 2), each additional µg/m³ of PM_2.5 was linked to a 1.7% increase in the HR (HR = 1.017, CrI: 1.008, 1.027), while each 1 Dobson unit increase in Ozone was associated with a 4.0% increase in HR (HR = 1.040, CrI: 1.025, 1.055). Although SO₂ was initially associated with an increased HR (HR = 1.040, CrI: 1.025, 1.055), this association dissipated after adjusting for maternal and household factors. Notably, NO₂ was linked to a decreased HR (HR = 0.964, CrI: 0.958, 0.970).

Figure 1.

Posterior mean (hazard ratio) and 95% credible interval of the effect of preconception climate anomalies, ambient air pollution, and vegetation (Enhanced Vegetation Index) on spontaneous miscarriage. Model 1: unadjusted model; Model 2: adjusts for maternal and household factors.

The results from the multiple-exposure models are summarized in Table 2. In the first model, which included only climate anomalies, mean temperature anomaly was associated with a significantly increased HR (HR = 7.730, CrI: 5.095, 11.729), while maximum temperature anomaly was associated with a reduced HR (HR = 0.331, CrI: 0.231, 0.474). In the model examining multiple ambient air pollutants and EVI, PM_2.5, CO, and SO₂ were linked to increased HRs for spontaneous miscarriage, whereas NO₂ was associated with a decreased HR. In the combined model (Model 3), which accounted for climate anomalies, ambient air pollutants, and EVI, mean temperature and precipitation anomalies remained associated with increased HRs for spontaneous miscarriage. CO and SO₂ continued to show increased HRs, while PM_2.5, NO₂, Ozone, and EVI were associated with reduced HRs. Finally, in the comprehensive model adjusting for all factors, including climate anomalies, ambient air pollutants, EVI, maternal, and household factors (maternal education, household wealth, household cooking fuel type, household access to improved drinking water, household sanitation facilities, place of residence), mean temperature and precipitation anomalies, CO, and EVI were associated with increased HRs for spontaneous miscarriage, whereas PM_2.5, NO₂, and ozone were linked to reduced HRs.

Table 2.

Effect of climate anomalies and ambient air pollutants on spontaneous miscarriage risk

Exposures	Model 1	Model 2	Model 3	Model 4
Mean temperature anomaly	7.730 (5.095, 11.729)		4.225 (2.605, 6.875)	4.286 (2.626, 7.007)
Maximum temperature anomaly	0.331 (0.231, 0.474)		0.952 (0.592, 1.525)	0.876 (0.539, 1.421)
Precipitation anomaly	0.908 (0.801, 1.029)		1.444 (1.239, 1.684)	1.446 (1.236, 1.691)
PM2.5 (µg/m³)		1.024 (1.013, 1.035)	0.965 (0.951, 0.978)	0.973 (0.958, 0.988)
Carbon monoxide (ppbv)		1.012 (1.006, 1.019)	1.011 (1.005, 1.018)	1.013 (1.005, 1.020)
Nitrogen dioxide		0.957 (0.951, 0.964)	0.954 (0.948, 0.961)	0.955 (0.948, 0.961)
Sulfur dioxide (µg/m³)		1.627 (1.297, 2.038)	1.245 (1.006, 1.552)	1.110 (0.879, 1.409)
Ozone (Dobson units)		0.997 (0.981, 1.013)	0.967 (0.948, 0.985)	0.971 (0.952, 0.990)
Enhanced Vegetation Index (EVI)	0.471 (0.236, 0.939)	0.443 (0.214, 0.923)	0.320 (0.162, 0.638)	1.071 (0.499, 2.317)

Model 1 = adjusts for other climate anomalies and vegetation; Model 2 = adjusts for other air pollutants and vegetation; Model 3 = adjusts for climate anomalies, air pollutants, and vegetation; Model 4 = adjusts for climate anomalies, air pollutants, vegetation, maternal and household factors (maternal education, household wealth, household cooking fuel type, household access to improved drinking water, household sanitation facilities, place of residence). Statistically robust (significant) results are presented in bold.

Sensitivity analyses (Appendix E, Supplementary Material; https://links.lww.com/EE/A372) indicated that when mean temperature anomaly and maximum temperature anomaly were included in the same model, the statistical significance of the maximum temperature anomaly diminished, despite both variables independently showing positive associations with the hazard of spontaneous miscarriage when modeled separately.

Effect of climate anomalies and ambient air pollutants modified by vegetation

Table 3 presents the modifiable effect of EVI on the association between climate anomalies, ambient air pollutants, and spontaneous miscarriage. In the unadjusted model, significant interactions were observed between EVI and exposures (mean temperature anomaly, maximum temperature anomaly, precipitation anomaly, PM_2.5, NO₂, and O₃), though no clear pattern emerged. Generally, the interaction between low EVI and exposures showed lower HRs, followed by high EVI, with medium EVI exhibiting the highest HRs—except for O₃, NO₂, and precipitation anomaly. For precipitation anomaly, high EVI was associated with a lower HR for spontaneous miscarriage compared with low and medium EVI.

Table 3.

Interaction of climate anomalies and ambient air pollutants with EVI categories on spontaneous miscarriage risk

		Unadjusted	Adjusted
Exposure	EVI categories	HR (95% CrI)	HR (95% CrI)
Mean temperature anomaly	Low	2.315 (1.959, 2.733)	2.535 (2.148, 2.991)
	Medium	2.480 (2.141, 2.874)	2.437 (2.099, 2.830)
	High	2.340 (1.966, 2.786)	2.262 (1.904, 2.688)
Maximum temperature anomaly	Low	1.801 (1.539, 2.107)	1.961 (1.677, 2.293)
	Medium	1.882 (1.643, 2.157)	1.835 (1.597, 2.107)
	High	1.853 (1.572, 2.183)	1.791 (1.522, 2.106)
Precipitation anomaly	Low	0.913 (0.734, 1.135)	0.877 (0.701, 1.098)
	Medium	0.690 (0.585, 0.812)	0.678 (0.575, 0.799)
	High	0.534 (0.434, 0.656)	0.555 (0.453, 0.679)
PM2.5 (µg/m³)	Low	1.015 (1.005, 1.025)	1.013 (1.002, 1.025)
	Medium	1.016 (1.006, 1.026)	1.010 (0.999, 1.022)
	High	1.012 (1.000, 1.024)	1.004 (0.992, 1.017)
Carbon monoxide (ppbv)	Low	0.998 (0.994, 1.002)	0.998 (0.993, 1.003)
	Medium	0.998 (0.994, 1.003)	0.998 (0.993, 1.002)
	High	0.997 (0.993, 1.002)	0.997 (0.992, 1.001)
Nitrogen dioxide	Low	0.967 (0.957, 0.978)	0.965 (0.954, 0.976)
	Medium	0.960 (0.952, 0.969)	0.960 (0.951, 0.968)
	High	0.965 (0.954, 0.977)	0.965 (0.954, 0.976)
Sulfur dioxide (µg/m³)	Low	1.103 (0.863, 1.417)	1.624 (1.278, 2.061)
	Medium	1.096 (0.903, 1.342)	1.409 (1.151, 1.734)
	High	1.010 (0.818, 1.263)	1.240 (0.995, 1.553)
Ozone (Dobson units)	Low	1.040 (1.025, 1.055)	1.040 (1.025, 1.056)
	Medium	1.040 (1.025, 1.055)	1.040 (1.025, 1.055)
	High	1.039 (1.024, 1.055)	1.040 (1.025, 1.055)

Adjusted models = adjusts for vegetation, maternal, and household factors (maternal education, household wealth, household cooking fuel type, household access to improved drinking water, household sanitation facilities, place of residence). Statistically robust (significant) results are presented in bold.

In the adjusted models, the effect of climate anomalies and air pollutants on spontaneous miscarriage was generally stronger at low EVI, except for O₃ and NO₂. Significant associations were found for all exposures except CO, with notable findings for SO₂ (high EVI), PM_2.5 (high and medium EVI), and precipitation anomaly (low EVI). For example, an increase in PM_2.5 was linked to a higher HR for spontaneous miscarriage at low EVI (HR = 1.013, CrI: 1.002, 1.025). Similarly, increases in mean and maximum temperature anomalies were associated with higher HRs for spontaneous miscarriage at low EVI (HR = 2.535, CrI: 2.148, 2.991 and HR = 1.961, CrI: 1.677, 2.293, respectively).

The results of the sensitivity analysis are are provided in the supplementary material; https://links.lww.com/EE/A372. For non-linearity, the findings indicate that mean temperature, maximum temperature, precipitation anomalies, and CO exhibit non/linear associations with spontaneous miscarriage (Supplementary Figures A1 and A2; https://links.lww.com/EE/A372). The remaining exposures demonstrated linear associations with spontaneous miscarriage. The region-specific effect (iid effect) is shown in Supplementary Figure B1; https://links.lww.com/EE/A372. Supplementary Figures C1–C9; https://links.lww.com/EE/A372 present the effects of preconception exposure on spontaneous miscarriage by rural–urban residency, while Supplementary Figures D1–D8; https://links.lww.com/EE/A372 illustrate the nonlinear effects of preconception exposure by vegetation, as measured using the EVI.

Discussion

Increasing evidence from existing studies highlights the impact of global climate change on health, particularly for vulnerable populations such as women and children. However, research simultaneously examining the effects of multiple climate-sensitive factors—such as temperature and precipitation anomalies, as well as ambient air pollution—on pregnancy outcomes, especially in SSA, remains limited. In this study, we investigate the effects of preconception exposure to climate anomalies and air pollution on the risk of spontaneous miscarriage among women in Ghana. Our findings indicate that 9.4% of pregnancies between 2012 and 2017 resulted in spontaneous miscarriage, as reported by the survey participants. The mean preconception exposure to particulate matter (PM_2.5) was approximately six times higher (29.7 µg/m³) than the World Health Organization’s recommended threshold of 5 µg/m³. The results suggest that preconception exposure to both mean and maximum temperature anomalies, as well as PM_2.5 and ozone levels, was associated with an increased rate of spontaneous miscarriage. These associations were further modified by maternal, household, and environmental factors, including ambient air pollutants and vegetation; however, the association between mean temperature anomalies and miscarriage remained robust across all models.

Existing research has established that exposure to extreme weather events, such as high temperatures, can adversely affect pregnancy outcomes, including low birth weight, preterm birth, induced abortion, and pregnancy loss. For example, Schifano et al¹⁴ found that a 10°C increase in maximum apparent temperature was associated with a 1.03–1.071 HR for preterm births during the 22nd–26th and 36th gestation weeks, respectively. Other studies in Spain, SSA, and other global regions have reported similar associations between temperature anomalies and adverse pregnancy outcomes.^12,16,17 Evidence suggests that pregnant women are more vulnerable to heat stress compared with nonpregnant individuals due to compromised thermoregulation and homeostasis.⁹ Extreme heat exposure may stimulate maternal antidiuretic hormone and oxytocin, leading to reduced uterine blood flow or altered fetal metabolism, both of which can contribute to negative pregnancy outcomes.⁹ Our study supports these findings by reinforcing the potential link between temperature anomalies and spontaneous miscarriage. Notably, this research goes beyond focusing solely on extreme temperatures; it conceptualizes climate anomalies as deviations from long-term climate trends, shedding light on the broader impact of climate change on reproductive health. Given the high correlation between mean and maximum temperature anomalies in our dataset, we conducted sensitivity analyses, which revealed that the association for maximum temperature anomaly was attenuated when both temperature measures were included in the same model. This suggests overlapping information between the two metrics, which should be considered when interpreting their respective associations.

Our study also examined the role of precipitation anomalies in pregnancy outcomes. The findings suggest that precipitation anomalies can significantly increase the risk of spontaneous miscarriage. Specifically, in a fully adjusted model, we observed that precipitation anomalies were associated with a 44.6% increase in the rate of spontaneous miscarriage (HR = 1.446; 95% confidence interval: 1.236, 1.691), while the association with maximum temperature anomalies was not significant. This suggests that the effects of precipitation on miscarriage may operate through direct or indirect pathways. Precipitation anomalies can influence pregnancy outcomes through various mechanisms. For instance, periods of extreme drought or heavy rainfall can exacerbate food insecurity, waterborne diseases, and air quality issues, all of which may stress the maternal body.^17,37 Such stress can lead to inflammatory responses, altered hormonal balances, and changes in uterine environment, all of which have been linked to adverse pregnancy outcomes, including miscarriage. Additionally, changes in precipitation patterns can disrupt the growth and distribution of vegetation, influencing air quality and exposure to environmental pollutants, which, as noted in our study, are also associated with an increased risk of miscarriage. Thus, while the direct effects of precipitation anomalies on miscarriage require further exploration, the broader environmental context, including changes in air quality and ecosystem stability, may amplify the risks associated with these climate disturbances.

The findings of this study also show that certain ambient air pollutants, such as PM_2.5, ozone, SO₂, and CO, are associated with an increased rate of spontaneous miscarriage. This finding supports knowledge from existing studies that show that ambient air pollutants have adverse effects on pregnancy outcomes.^14,15,22 A consistent and robust association was observed for CO exposure, even when adjusting for other ambient air pollutants, climate anomalies, and maternal and household factors. Evidence shows that many households in SSA (81%), including those in Ghana, heavily rely on biomass fuels and other unclean energy sources, contributing to elevated levels of CO emissions in the surrounding environment.³⁸ Carbon monoxide (CO) binds with hemoglobin in the blood, a process known as carboxyhemoglobinemia (COHb), which reduces the blood’s ability to carry oxygen. During pregnancy, this can lead to oxygen deprivation, potentially inducing oxidative stress in the maternal and fetal tissues. This stress is a known risk factor for adverse pregnancy outcomes, including miscarriage.

However, with the exception of NO₂, the direction and strength of associations for other ambient air pollutants were influenced by the inclusion of climate anomalies and maternal or household factors. For instance, while PM_2.5 was initially associated with a higher rate of spontaneous miscarriage when adjusted for maternal and household factors alone, the association reversed, showing a reduced rate when climate anomalies, other ambient air pollutants, and maternal and household factors were accounted for. This reversal of association may reflect the complex interplay between environmental pollutants and other confounding factors. Interestingly, NO₂ exposure appears to have a different effect. While it is generally considered harmful to health, our findings suggest that NO₂ exposure is associated with a lower rate of miscarriage. This unexpected inverse relationship may be due to the specific sources of NO₂ in the study area, which might be linked to urban pollution sources like vehicle and industrial emissions.³⁹ Such exposures may correlate with other protective factors, such as better access to healthcare or higher socioeconomic status, which could confound the direct association between NO₂ and miscarriage.

The effect of vegetation, as measured by the EVI, showed a notable pattern in this study. In all models except the fully adjusted one, higher EVI values—indicating denser or more vigorous vegetation—were associated with a reduced risk of spontaneous miscarriage. This suggests that vegetation may have a protective role, potentially improving air quality by enhancing carbon sequestration and filtering particulate matter, or by providing a more stable microenvironment that reduces heat stress and supports overall maternal health. Vegetation may also help mitigate the adverse effects of environmental pollutants known to negatively impact pregnancy outcomes. Interestingly, the fully adjusted model revealed a slight increase in the risk of miscarriage with higher EVI (HR = 1.071), suggesting that the protective effect of vegetation may be modified by other factors such as climate anomalies, ambient air pollution, and maternal and household confounders. Notably, EVI appeared to modify the impact of environmental exposures like mean temperature, maximum temperature, precipitation, and PM_2.5 on spontaneous miscarriage. Specifically, at higher EVI levels, the adverse effects of temperature anomalies and air pollutants were generally minimized, highlighting the potential of vegetation to buffer the impacts of these environmental stressors. These findings underscore the complex interplay between environmental factors and maternal health, with vegetation acting as both a protective and modifying factor in the risk of spontaneous miscarriage.

Study strengths and limitations

This study has several strengths, including its novel focus on the impact of climate anomalies and ambient air pollution on spontaneous miscarriage, a topic rarely explored. By combining climate data with environmental exposure assessments, the study provides valuable insights into how climate change and pollution may affect pregnancy outcomes in Ghana and SSA. Additionally, robust statistical modeling was used to control for various maternal, household, and environmental factors, enhancing the validity of the findings. However, there are limitations. The cross-sectional design limits the ability to infer causal relationships between climate, air pollution, and miscarriage risk. Longitudinal studies would better track environmental exposures over time and assess their impact on pregnancy outcomes. Another limitation is the lack of data on maternal diet, which plays a critical role in pregnancy outcomes and could interact with environmental exposures. Furthermore, while the study controlled for several factors, potential confounders such as psychosocial stress, healthcare access, and underlying health conditions (e.g., diabetes, hypertension) were not included, which could affect the results. The correlation analysis indicated a strong relationship between mean temperature anomaly and maximum temperature anomaly (r = 0.94, P > 0.001). Sensitivity analyses show that when both variables are included in the same model, the significance of maximum temperature anomaly diminished, although each independently increased the HR for spontaneous miscarriage (Appendix E, Supplementary Material; https://links.lww.com/EE/A372). Both were retained as they capture distinct aspects of temperature variation, but this high correlation may have introduced collinearity and reduced precision in individual estimates.

Conclusions

This study suggests that climate anomalies and ambient air pollution significantly impact spontaneous miscarriage rates among women in Ghana. Preconception exposure to mean temperature anomalies and carbon monoxide (CO) was strongly associated with increased miscarriage rates, while other pollutants were confounded by additional factors. Sensitivity analyses indicated nonlinear relationships and regional variations in these effects. These findings highlight the need for further research on the complex interactions between climate, pollution, and pregnancy outcomes. From a policy perspective, reducing exposure to pollutants like PM_2.5 and CO, alongside climate change mitigation efforts, is crucial. Interventions such as stricter air quality regulations, better monitoring, and climate-resilient healthcare can help reduce miscarriage risks, particularly for vulnerable groups in high-pollution or climate-vulnerable areas.

Conflicts of interest statement

The authors declare that they have no conflicts of interest with regard to the content of this report.