Ambient Air Pollution and Pregnancy-Induced Hypertensive DisordersNovelty and Significance
A Systematic Review and Meta-Analysis
Pregnancy-induced hypertensive disorders can lead to maternal and perinatal morbidity and mortality, but the cause of these conditions is not well understood. We have systematically reviewed and performed a meta-analysis of epidemiological studies investigating the association between exposure to ambient air pollution and pregnancy-induced hypertensive disorders including gestational hypertension and preeclampsia. We searched electronic databases for English language studies reporting associations between ambient air pollution and pregnancy-induced hypertensive disorders published between December 2009 and December 2013. Combined risk estimates were calculated using random-effect models for each exposure that had been examined in ≥4 studies. Heterogeneity and publication bias were evaluated. A total of 17 articles evaluating the impact of nitrogen oxides (NO2, NOX), particulate matter (PM10, PM2.5), carbon monoxide (CO), ozone (O3), proximity to major roads, and traffic density met our inclusion criteria. Most studies reported that air pollution increased risk for pregnancy-induced hypertensive disorders. There was significant heterogeneity in meta-analysis, which included 16 studies reporting on gestational hypertension and preeclampsia as separate or combined outcomes; there was less heterogeneity in findings of the 10 studies reporting solely on preeclampsia. Meta-analyses showed increased risks of hypertensive disorders in pregnancy for all pollutants except CO. Random-effect meta-analysis combined odds ratio associated with a 5-µg/m3 increase in PM2.5 was 1.57 (95% confidence interval, 1.26–1.96) for combined pregnancy-induced hypertensive disorders and 1.31 (95% confidence interval, 1.14–1.50) for preeclampsia. Our results suggest that exposure to air pollution increases the risk of pregnancy-induced hypertensive disorders.
There is increasing evidence that exposure to ambient air pollution contributes to the risks of cardiovascular events in adults and the elderly.1 Maternal exposure to air pollution during pregnancy has been associated with increased risk of adverse birth outcomes such as low birthweight and preterm birth.2,3 Concern has recently been raised that ambient air pollution exposure may also increase the risk of hypertensive disorders during pregnancy.4
Gestational hypertension and preeclampsia are the most common complications of pregnancy affecting 2% to 10% of pregnancies after 20 weeks of pregnancy.6,7 Gestational hypertension is usually diagnosed based on a systolic blood pressure of ≥140 mm Hg or a diastolic blood pressure of ≥90 mm Hg in previously normotensive women.7 Preeclampsia is characterized by hypertension and protein in the urine.6,8 In addition to maternal mortality, preeclampsia and related conditions are a leading cause of maternal morbidity, perinatal death, placental abruption, preterm birth, and child growth restriction.6–8
The objective of this study was to systematically review epidemiological studies investigating associations between ambient air pollution and pregnancy-induced hypertensive disorders and to perform a meta-analysis of these studies.
Studies were identified using the electronic bibliographic databases PubMed and Web of Science using the following search terms (preeclampsia, eclampsia, pregnancy hypertension, gestational hypertension, or HELLP) and (air pollution, traffic, or criteria air pollutants: particulate matter [PM], nitrogen oxides [NOX], carbon monoxides [CO], or ozone [O3]). We also searched the reference lists of identified articles for additional publications. We limited the search to peer-reviewed articles published in English between December 2009 and December 19, 2013. Full articles published in other languages and abstracts were excluded (Figure 1).
A total of 17 studies were considered relevant to be included in the systematic review and meta-analysis. If not reported, we requested from the authors effect estimates for continuous exposure. Additional unpublished information was received for a Swedish study,9 which was included in our analyses. We excluded 1 small (n=2707) study conducted in Iran from the meta-analysis because only effect estimates of high versus low exposure to CO were reported.10
Finally, in the case of 2 publications that had overlapping study populations,11,12 we selected the results from the study that had the largest sample size.12 In a sensitivity analysis, we included the other study11 instead, because the 2 studies used different exposure assessment methods.
To allow comparison of effect estimates between different studies, the reported odds ratios (OR) and 95% confidence intervals (CIs) were converted to correspond to common exposure increments (Methods S1 in the online-only Data Supplement).
Few estimates were available for trimester-specific exposure, and thus combined estimates were only estimated for the full pregnancy mean to secure sufficient number of studies for meta-analyses. Three studies13–15 did not report the effect of the exposure during the entire pregnancy, and so for these 3 studies we included the effect estimates related to exposure during the first trimester.
Four studies only reported categorical effect estimates associated with exposure to residential proximity to major roads and traffic density,9,16–18 and for these studies we estimated the combined risk estimates related to the highest versus the lowest exposure category (Methods S2).
We examined all studies on pregnancy-induced hypertensive disorders (ie, gestational hypertension and preeclampsia as separate outcomes or a combination of these conditions), irrespective of the definition used, to secure sufficient number of studies for random effect meta-analyses.19 Second, we restricted the meta-analyses to studies on preeclampsia because there is controversy whether gestational hypertension and preeclampsia are a spectrum or different diseases.6–8 There were an insufficient number of studies on gestational hypertension to analyze this as a separate outcome.
To identify potential influential studies and examine the robustness of the findings of the meta-analysis to the exclusion of these studies, we conducted sensitivity analyses by repeating meta-analyses after removing 1 study at a time and comparing the combined estimates with and without that study.
Publication bias was evaluated using Funnel plots and Egger tests.20 In the absence of bias, the funnel plot should resemble a symmetrical inverted funnel with larger studies being close to the pooled estimate, and smaller studies more widely, but evenly, scattered around the pooled estimate. The Egger test provides a formal statistical evaluation of whether there is asymmetry in the study results that is related to their standard error. Stata S.E. version 12.1 was used for the statistical analyses (StataCorporation, TX).
Description of the Study Characteristics
We identified 17 studies published between 2009 and 2013 for the systematic review that are summarized in Table S1 in the online-only Data Supplement.10–18,21–27 All studies with the exception of 1 case–control study28 were prospective cohorts or birth record–based retrospective cohorts. The studies were based on 298 to 468 517 pregnancies occurring during 1996 to 2008. Nine studies were conducted in the United States, 5 in Europe, and 3 in Iran, Japan, and Australia (Table S1). Most studies excluded multiple pregnancies, but not all.21,23 The applied inclusion and exclusion criteria varied between studies and so did the availability of potential confounders (Table S1).
Outcome Definition and Prevalence
Ten of the 17 studies evaluated preeclampsia,9,11,12,16,21–25 5 studied a combination of gestational hypertension and preeclampsia without separating them,15,17,26–27 4 considered gestational hypertension,10,13,16,24 and one evaluated a combination of gestational hypertension and preeclampsia among women with preterm births.18 Few studies further evaluated subgroups of preeclampsia according to severity9,11,15 or early and late onset.21 The definition of outcomes varied among the studies (Table S2). In some studies, preeclampsia included the more severe outcomes such as eclampsia,11,12,24,26,27 and the definitions used were not described in detail in all studies.25,27
The reported prevalence for gestational hypertension ranged from 3.6%12 to 6.0%,13 preeclampsia ranged from 1.2%21 to 4.0%,22 and the combination of gestational hypertension and preeclampsia ranged from 2.7%14 to 4.7%.27
Air Pollution Exposure
Half of all the included studies estimated the pollutant concentration in ambient air at the maternal home address at birth using environmental models (land-use regressions or dispersion),9,12,21–24 whereas routine air pollutant data from central air–pollution-monitoring stations were used in the other half of the studies.10,12,15,18,25–27 Most commonly, exposures were assigned using the monitoring station nearest to the maternal residence at the time of birth. Distance between the monitors and participant’s home addresses as well as the number and density of monitors differed across studies.
Seven studies evaluated PM2.5, 6 NO2, 5 PM10, CO, or O3, and 4 evaluated NOX (Table S1). The mean pregnancy exposure (µg/m3) to PM2.5 and NO2 ranged from 10.123,27 to 17.312 and from 23.022 to 55.7,21 respectively. The ranges of the other air pollutant concentration also differed among studies (Table S3).
Combined Estimates of Air Pollution and the Risk of Hypertensive Disorders in Pregnancy
The effect estimates of individual studies and the combined effect estimates for pregnancy-induced hypertensive disorders associated with PM2.5 and NO2 are presented in Figure 2. For all pollutants, heterogeneity as measured by the I2 was high, and there were highly significant Q tests when all pregnancy-induced hypertensive outcomes were combined (Table). For most pollutants, the heterogeneity among studies tended to be smaller when studies on preeclampsia only were considered (Table; Figure 2).
Results from the individual studies on PM2.5 consistently reported increased risk of pregnancy-induced hypertensive disorders ranging from 14%13 to 398%27 associated with each 5-µg/m3 increase in PM2.5 (Figure 2). Likewise for NO2, a consistent pattern of increased risks between 1%21 and 85%27 associated with each 10-µg/m3 increase in NO2 were reported (Figure 2). Increased risks were also observed in most, but not in all, of the individual studies on NOX, PM10, and O3 (Figure S1).
Pregnancy-induced hypertensive disorders were statistically significantly associated with a 5 µg/m3 increment in PM2.5 (OR=1.57; 95% CI, 1.26–1.96; Table; Figure 2), a 10-µg/m3 increment in NO2 (OR=1.20; 95% CI, 1.00–1.44; Table; Figure 2), and a 10-µg/m3 increment in PM10 (OR=1.13; 95% CI, 1.02–1.26; Table). When all outcomes were combined, there were also marginally statistically significant associations with NOX and O3 but not with CO (Table). The forest plots of the other pollutants are presented in Figure S1.
Residential Proximity to Traffic, Traffic Density, and Hypertensive Disorders in Pregnancy
One study reported effect estimates associated with increments in traffic density12 and 4 reported effect estimates of categorical proximity to major road.9,17,18,24 The individual studies have reported increased ORs for pregnancy-induced hypertensive disorders among women exposed to high traffic density,9,12 and among women living within 200 to 250 m from major roads,17,18 although not for all of the associations evaluated.16,17
The heterogeneity among the studies reporting effects of low versus high exposure to traffic indicators was not statistically significant. The random-effect meta-analysis suggested that pregnancy-induced hypertensive disorders were statistically significantly associated with higher exposure to traffic density (Table; Figure S1).
The results of our sensitivity analyses showed that our combined estimates were generally robust to the exclusion of each study (Table S4). However, for NO2 when an influential study27 was removed there was no longer evidence of heterogeneity between the individual studies and the combined estimate for NO2 reduced to 1.08 (95% CI, 1.02–1.13) from 1.20 (95% CI, 1.00–1.44). For PM2.5, removal of this study also decreased the combined OR point estimate. For NOX, removal of 1 study10 changed the heterogeneity P value of 0.02 to 0.82.
Funnel plots for each pollutant did not exhibit a notable asymmetry, suggesting absence of a notable publication bias (Figure S2). For all exposures, the P values of Egger tests were statistically nonsignificant (P>0.05), indicating that any asymmetries in the funnel plots may be because of chance; however, for some associations the number of studies and, therefore, the statistical power of Egger test was limited.
To our knowledge, this is the first systematic review and meta-analysis of available evidence on a possible association between exposure to air pollution and hypertensive disorders in pregnancy. We identified and reviewed 17 large studies, including >1 000 000 pregnancies in total. Our meta-analysis showed statistically significantly increased risks for hypertensive pregnancy disorders in association with exposure to PM2.5, NO2, and PM10 during pregnancy when all studies were combined. Exposure to PM2.5 and NO2 were also associated with significantly increased risk for preeclampsia.
Maternal blood volume, heart rate, stroke volume, and cardiac volume increase along the course of pregnancy to accommodate the needs of the developing fetus.5 In healthy pregnancies, the blood pressure falls during the first trimester, reaching its lowest levels in midpregnancy, and then it increases to prepregnancy levels by term, whereas for women who develop gestational hypertension or preeclampsia blood pressure is stable during the first half of pregnancy and then increases until delivery.28 A few studies have examined associations between air pollution and blood pressure in pregnant women.24,29–31 Differences in design and exposure periods make comparisons between these studies difficult; however, the findings from these studies suggest that exposure to air pollutants during pregnancy increases blood pressure. For instance, the first-trimester exposure to PM1024,31 and second-trimester exposure to PM2.5 have been associated with increased blood pressure in late pregnancy.30 Moreover, short-term exposure to NO2 has been associated with increased blood pressure in pregnant women.29
Air pollution particles inhaled can trigger enhanced maternal oxidative stress, lipid peroxidation, inflammation, and changes in the blood system, damage vascular endothelium thereby decreasing placental blood flow, disrupt transplacental oxygenation, and cause placental oxidative stress, inflammation, and imbalance between angiogenic placental growth factors and antiangiogenic proteins (placenta growth factor, vascular endothelial growth factor, soluble fms-like tyrosine kinase-1, and soluble endoglin),1,4,32 which play significant roles in the development of hypertensive pregnancy disorders.6,8 For example, placental dysfunction has been proposed to have a critical role in the pathogenesis of preeclampsia.6,8 Maternal–fetal immune maladaptation, oxidative stress, and placental ischemia/hypoxia may contribute to placental dysfunction, which results in the release of antiangiogenic factors and other inflammatory mediators from placenta. PM2.5 exposure has been associated with elevated blood levels of markers of endothelial dysfunction (ICAM-1, VCAM-1),33 which are also shown to increase in preeclampsia.6 Exposure to PM2.5 has also been linked to the release of cytokines including interleukin-6,34 which are reported to be involved in the pathogenesis of preeclampsia.35
Gestational hypertension or preeclampsia and related outcomes were examined separately or in combination (Table S2). Although these hypertensive disorders of pregnancy may be part of a common spectrum of conditions associated with raised blood pressure during pregnancy,6–8 different diagnostic criteria were applied by studies to ascertain the cases. This variation in diagnostic criteria together with differences in definition of outcomes (eg, inclusion of all pregnancy-induced hypertensive disorders together or separately or inclusion of and the fact that a mixture of mild and severe conditions as well as early- and late-onset conditions separately or as a whole) as well as potential variation in the completeness of case ascertainment could also have contributed to our observed heterogeneity between the studies.
Early-onset (diagnosed between weeks 20 and 34 of pregnancy) and late-onset (diagnosed after week 34 of pregnancy) preeclampsia have been reported to have different prevalence and prognosis and possibly different pathogenesis and risk factors.36 To date, there has been only 1 study reporting on the impact of air pollution on early- and late-onset preeclampsia separately,21 and more studies are required to replicate their findings.
Exposure Assessment and Exposure Contrasts
Half of the studies on air pollutants relied on data from central air-monitoring stations10,12,14,15,25–27 that may not adequately capture spatial contrasts in air pollutant levels and mostly accounts for variation of pollutant levels over time. The other half assessed air pollution using models9,11–13,21,22,24 that are also subject to uncertainty because of limitations in the input data. Traffic density and proximity to major roads are rather crude estimates of air pollution when compared with air pollutant measurements or sophisticated prediction models.
Our meta-analysis was restricted to effect estimates of commonly studied air pollutants; however, we are not able to rule out whether these pollutants are surrogates for the effect of other unmeasured pollutants like ultrafine particles.
All of our reviewed studies relied on ambient levels of air pollutants at residential address overlooking the potential variation between outdoor and indoor levels of pollutants. Moreover, no studies took into account maternal time activity and exposure in other microenvironments (eg, in workplace or during commuting). Information on residential mobility during pregnancy was available in only 1 study population.16,24 Almost all studies estimated exposure at the address at time of birth; having lived at other addresses during pregnancy would result in misclassification of exposure. Furthermore, among the studies using exposure levels during the entire pregnancy, only one21 truncated the exposure levels at the time of diagnosis to ensure that the exposure precedes the outcome. Other studies using exposure levels during the entire pregnancy could have been affected by exposure misclassification as they have included exposure levels after diagnosis of preeclampsia/pregnancy-induced hypertension in their assessed exposure levels. However, most of these factors are likely to create nondifferential misclassification of exposure, which would make the estimation more conservative (ie, bias toward null).
The Potential for Confounding and Residual Confounding
In our meta-analyses of pregnancy-induced hypertensive disorders, there was evidence of heterogeneity for most of the evaluated associations. Nevertheless, all studies showed increased ORs and, thus, the heterogeneity reflected differences in the magnitude rather than differences in the direction of the association. The heterogeneity of our findings may reflect genuine differences in the study settings, such as geographical location, land-use, traffic density, and composition, which could influence the variation in exposure to air pollution and potentially lead to different dose-response relationships. This heterogeneity could also be a consequence of specific biases or may arise from differences in study design, exposure assessment, and the definitions of outcome in the studies. The heterogeneity of our findings might also be related to differences in the degree of control for potential confounders. Maternal age, prepregnancy body mass index, preexisting and gestational diabetes mellitus, indicators of socioeconomic status (mostly maternal education), ethnicity, smoking, and parity were included as covariates in most, but not all, studies (Table S1). A high maternal body mass index has consistently been found to increase preeclampsia risk,6 but not all studies had information on body mass index.11–14,17,22,25–27 Indeed, not only do ambient air pollution levels vary largely over space but also possibly other parameters such as the prevalence of obesity and healthcare access. Residual confounding may occur if these risk factors are unevenly distributed across types of areas (rural and urbanized areas) and are not assessed or poorly taken into account. Also only 9 studies adjusted for season.11–14,17,21,23,27 Season may be a potential confounder because seasonal variation in air pollution is well known, and suspected risk factors of preeclampsia include temperature37 and vitamin D status.38
For each pollutant, however, the reported adjusted point effect estimates for preeclampsia were just slightly increased or decreased as compared with the crude models,12,13,16,23–25,27,30 and therefore, the effect of confounding on our combined effect estimates from meta-analysis of the reported effect estimates from adjusted models by the adjustment factors seems to be minimal.
Most studies included >10 000 pregnancies, with some exceptions.10,15,16,21,23,24 Large studies are needed to study relatively rare outcomes. Considering large areas also entails a potential increase in exposure contrasts, which is a priori desirable. However, greater population heterogeneity may also increase the potential for confounding.39 None of the large studies adjusted for type of area or other approaches to limit potential residual confounding because of area-associated factors.
Finally, mutual adjustment for factors such as other air pollutants3 and traffic-related noise,40 which could influence the effect of exposure to air pollution over time and in space, were also rarely taken into account. It is impossible to rule out the possibility of residual confounding, given the limitations in many of the studies included in our meta-analysis.
Strengths and Limitations
Meta-analysis can provide more precise findings than those obtained from individual studies and increases the statistical power, which is especially important for rare outcomes. This was the case in our meta-analysis, which was able to demonstrate a significant association between air pollution and hypertensive pregnancy disorders even though many of the individual studies could not.
Our findings were robust in the sense that excluding individual studies in most cases had little influence on our results and did not explain the heterogeneity. This was particularly true for the results for PM2.5, which is in agreement with previous studies, reporting that exposure to PM2.5 during pregnancy is most consistently associated with low birthweight.2,3 There was also no significant evidence of small study (ie, publication) bias in our study.
Our study has a few limitations that should be borne in mind when interpreting our findings. First, although the included studies were based on a large number of cases, ranging from 1385 to 19 139, there are still relatively few studies on air pollution and hypertensive disorders in pregnancy. Included in the meta-analysis are 3 studies relying on first-trimester exposure,13–15 and we cannot rule out that the inclusion of these studies could have introduced error (as the variance of the effect estimates can be influenced by the variability of the exposure of interest, which is smaller in a single trimester compared with whole pregnancy exposure), even if this error did not introduce bias. Moreover, a number of studies combined gestational hypertension with preeclampsia, and we were not able to include these studies in meta-analyses of pregnancy-induced hypertension as a separate outcome. Finally, the fact that not all women attend prenatal care might have resulted in underreporting of mild cases.
Our findings suggest an association between air pollution and pregnancy-induced hypertensive disorders, but the sources of heterogeneity between studies should be further explored. Large multicentered pooled analysis of the data could address this. Future studies need to take into account more factors, for example, temperature, maternal nutrition, and road noise, which may confound or modify the association between air pollution and pregnancy-induced hypertensive disorders.
Exposure assessment could be improved by including time activity and exposure in different microenvironments. Studies on more source-specific measures such as elemental composition of the particulate matter and ultrafine particles, polycyclic aromatic hydrocarbons, and volatile organic compounds are recommended. Further investigation on how exposure to air pollution might affect the placenta or the maternal blood system and lead to hypertensive disorders in pregnancy is needed. Experimental studies are needed to elucidate relevant biological pathways. More studies examining the associations between (short-term and long-term exposure) air pollution and blood pressure changes in pregnant women are recommended.
The pathogenesis of pregnancy-induced hypertensive disorders is not well understood. Our study provides strong evidence that exposure to air pollution increases the risk of these conditions. Our findings are of public health importance because if these exposures are avoided, they theoretically could result in the reduction of the incidence of such conditions, which are accompanied with considerable personal and social burdens. Although our observed combined risks were relatively small, the ubiquitous nature of air pollution exposure suggests that air pollution may have a large population-attributable risk for these disorders.
We thank Ebba Malmqvist for providing unpublished results, and we acknowledge the valuable contributions from David Martinez to recalculate the effect estimates to similar units and increments.
Sources of Funding
M. Pedersen holds a Juan de la Cierva postdoctoral fellowship awarded from the Spanish Ministry of Science and Innovation (JCI-2011–09479). P. Dadvand is funded by a Ramón y Cajal fellowship (RYC-2012–10995) awarded by the Spanish Ministry of Economy and Competitiveness.
The online-only Data Supplement is available with this article at http://hyper.ahajournals.org/lookup/suppl/doi:10.1161/HYPERTENSIONAHA.114.03545/-/DC1.
- Received March 13, 2014.
- Revision received March 28, 2014.
- Accepted May 21, 2014.
- © 2014 American Heart Association, Inc.
- Olsson D,
- Mogren I,
- Forsberg B
- Mobasher Z,
- Salam MT,
- Goodwin TM,
- Lurmann F,
- Ingles SA,
- Wilson ML
- Egger M,
- Davey Smith G,
- Schneider M,
- Minder C
- Pereira G,
- Haggar F,
- Shand AW,
- Bower C,
- Cook A,
- Nassar N
- van den Hooven EH,
- de Kluizenaar Y,
- Pierik FH,
- Hofman A,
- van Ratingen SW,
- Zandveld PY,
- Mackenbach JP,
- Steegers EA,
- Miedema HM,
- Jaddoe VW
- Xu X,
- Hu H,
- Ha S,
- Roth J
- Jedrychowski WA,
- Perera FP,
- Maugeri U,
- Spengler J,
- Mroz E,
- Flak E,
- Stigter L,
- Majewska R,
- Kaim I,
- Sowa A,
- Jacek R
- Thompson AM,
- Zanobetti A,
- Silverman F,
- Schwartz J,
- Coull B,
- Urch B,
- Speck M,
- Brook JR,
- Manno M,
- Gold DR
- Valensise H,
- Vasapollo B,
- Gagliardi G,
- Novelli GP
- Beltran AJ,
- Wu J,
- Laurent O
- Pedersen M,
- Siroux V,
- Pin I,
- Charles MA,
- Forhan A,
- Hulin A,
- Galineau J,
- Lepeule J,
- Giorgis-Allemand L,
- Sunyer J,
- Annesi-Maesano I,
- Slama R
Novelty and Significance
What Is New?
Exposure to ambient air pollution (PM2.5, NO2, and PM10) during pregnancy is associated with 57% to 13% increase in the risk of pregnancy-induced hypertensive disorders.
Exposure to ambient air pollution (PM2.5 and NO2) during pregnancy is associated with 31% to 7% increase in the risk of preeclampsia.
What Is Relevant?
Our findings extend the scope of etiologic studies of pregnancy-induced hypertensive disorders to environmental insults, which if avoided, could reduce or contribute to the reduction of the incidence of such conditions.
Our findings may have profound implications for public health, given the large burden associated with these conditions, and the ubiquity of air pollution in our society.
There are a number of shortcomings in the available body of evidence that need to be addressed in future studies.
Epidemiological evidence from >1 000 000 pregnancies suggests that exposure to air pollution increases the risk of pregnancy-induced hypertensive disorders.