The objective of this study was to conduct a systematic review to provide summarized evidence on the association between maternal exposure to particulate air pollution and birth weight (BW) and preterm birth (PTB) after taking into consideration the potential confounding effect of maternal smoking.
We systematically searched all published cohort and case-control studies examining BW and PTB association with particulate matter (PM, less than or equal to 2.5μm and 10.0 μm in diameter, PM2.5 and PM10, respectively) from PubMed and Web of Science, from January 1980 to April 2015. We extracted coefficients for continuous BW and odds ratio (OR) for PTB from each individual study, and meta-analysis was used to combine the coefficient and OR of individual studies. The methodological quality of individual study was assessed using a standard protocol proposed by Downs and Black. Forty-four studies met the inclusion criteria.
In random effects meta-analyses, BW as a continuous outcome was negativelyassociated with 10 μg/m3 increase in PM10 (-10.31 g; 95% confidence interval [CI], -13.57 to -3.13 g; I-squared=0%,
While this systematic review supports an adverse impact of maternal exposure to particulate air pollution on birth outcomes, variation in effects by exposure period and sources of heterogeneity between studies should be further explored.
Maternal exposure to air pollution during pregnancy is associated with increased risk of adverse birth outcomes such as low birth weight (BW) and preterm delivery [
Several case-control and cohort studies have revealed a positive association between maternal exposure to ambient air pollution and a range of adverse pregnancy outcomes [
This study aimed to conduct a systematic review to provide summarized evidence of the associations between PM (less than or equal to 2.5 μm in diameter. PM2.5 and PM10) and BW and PTB, by providing summary estimates of effect by gestational period and taking into consideration the potential confounding effect of maternal smoking.
We performed a systematic literature search in PubMed (
After full review, data extraction from relevant studies was performed independently by two investigators, using standard template. The template included information on study design, location, dates of data collection, data sources, sample size, descriptive information on study subject characteristics, outcome frequency, distribution of exposure, method of exposure characterization, statistical analysis methods, ES estimates, and covariates examined jointly with air pollution. We extracted the smoking adjusted and smoking unadjusted effect measures separately from the original studies. As studies did not all include or adjusted for the same gaseous pollutants, single pollutant models fully adjusted for other covariates were used. Data were analyzed using Excel 2007 (Microsoft, Redmond, WA, USA) and Stata version 13.0 (Stata Corp., College Station, TX, USA).
In meta-analyses, we selected the results that were based on the larger number of observations for the same outcome, pollutant, and population. We selected the study of Morello-Frosch et al. [
To assess the methodological quality of articles that were included for the quality comparison across studies, we used the checklist developed by Downs and Black [
In order to facilitate comparisons of ES from different studies, all risk estimates were converted to a common exposure unit of 10 μg/m3 increase in PM10 or PM2.5. Effect estimates were grouped by gestational period (trimester-specific and whole pregnancy). Most effect estimates relating to PTB were expressed as adjusted OR, however, when relative risks were reported, we converted these to ORs using the approximation approach described by Zhang and Yu [
We performed sensitivity analyses and subgroup analyses to evaluate potential sources of heterogeneity across the included studies. Sensitivity analyses were performed by removing each study one-by-one. Subgroup analyses were based on the possible confounding effects of smoking and the methodological quality of the studies (high or low).
For outcome of BW, there were seven studies that examined PM10 exposure and 11 studies using PM2.5, with nine studies providing estimates for both pollutants. Similarly, for PTB, there were five studies relying on PM2.5, nine studies on PM10, and four studies on both pollutants. For PTB studies, the reported mean exposure levels ranged from 5.1 μg/m3 [
We used Downs and Black checklist to assess the methodological quality of studies related to BW and PTB (
The results of the pooled analyses are illustrated in
Sixteen studies were included in the primary meta-analysis for the change in BW associated with a 10 μg/m3 increment in PM10, by gestational period. We found the pooled smoking-adjusted ES during pregnancy (ES, -10.31 g; 95% CI, -13.57 to -7.05 g; I-squared = 0.0%,
A forest plot for PTB associated with a 10 μg/m3 increment in PM10 is presented in
With some noted exception, overall, we observed that meta-analysis estimates were stable, excluding a particular study did not change the summary point estimates much. For example, the pooled estimated reduction in smoking adjusted BW during pregnancy with a 10 μg/m3 increase in PM10 increased from -6.46 g (95% CI, -14.20 to 1.28 g) to -10.31 g (95% CI, -13.57 to -7.05 g) after removing the study by Geer et al. [
We did not detect a statistically significant publication bias based on the Egger’s test (
This systematic review has updated current scientific evidence and shows that decrease in BW per 10 μg/m3 increase in particulate matter (PM2.5 or PM10) during pregnancy; a decreased BW estimated with smoking-adjusted studies was consistently higher than the weight estimated with smoking unadjusted studies. We found the pooled estimates of decrease in BW of approximately 10 g for PM10 and 22 g for PM2.5 without evidence of publication bias or heterogeneity (PM10), after taking into consideration the confounding effect of smoking. In addition, our results of combined smoking-adjusted and-unadjusted studies suggest that maternal exposures to PM10 and PM2.5 during pregnancy are associated with 23% and 14%, respectively, excess risk of PTB.
We identified 13 previous reviews linking air pollution, BW and PTB [
Our pooled estimates of decrease in BW of 10 g and 22 g for each 10 μg/m3 increase in PM10 and PM2.5, respectively, is comparable with a recent coordinated international analysis [
We found overall or pooled smoking-adjusted ES for BW was stronger than that without adjustment for smoking in most cases. Combined smoking-unadjusted ES indicated a decrease in BW associated with a 10 μg/m3 increase in PM10 exposure during pregnancy (-8.17 g; 95% CI, -10.99 to -5.36 g). The association was stronger when adjusted for smoking (-10.31 g; 95% CI, -13.57 to -7.05 g). The strength and direction of this association was comparable with a previous fully adjusted (vs. unadjusted) random effects meta-analyses for term BW (-8.9 g; 95% CI, -13.2 to -4.6 g) [
We found that the pooled OR of PTB during entire pregnancy is higher than that of previous review. Sapkota et al. [
We found no evidence of publication bias based on contourenhanced funnel plot for PM and BW. Therefore, we expect that any publication bias, if present, is minimal in our review. Thus, some difference in the ESs between this meta-analysis and previous one may be due to publication bias. A previous study that demonstrated a comparison between meta-analyses and large multicenter analyses of the associations between air pollution and mortality has suggested that difference in ESs reported in meta-analyses could be attributable to publication bias [
Woodruff et al. [
Although we realized that the countries where studies were conducted and the study design might also be sources of heterogeneity, they were not analyzed in the review due to the limited number of studies conducted in different countries. Though we recognized that several sensitivity analyses were conducted in relation to race or other factors, stratified analyses were not performed based on these categories due to the limited number of studies, particularly when divided by exposure period. We also aware that the use of effect estimates based on associations with ambient levels of pollutants as a surrogate for personal exposure levels may have resulted some exposure misclassification. Other limitation includes the fact that none of the included studies provided the precise information on the timing of smoking during pregnancy. Hence, future large cohort studies with sufficient data and detailed information on timing of smoking during pregnancy and other potential confounding factors as well as reliable exposure data are required for a better understanding of the association between PM and the risk of adverse birth outcomes.
This systematic review and meta-analysis, which are based on 44 studies of particulate pollution and birth outcomes, revealed that 10 to 22 g decrease in BW was linked to maternal exposure to particulate pollution (PM10 or PM2.5) during pregnancy, after adjustment for smoking. Likewise, pooled smoking-adjusted OR for PTB (0.97; 95% CI, 0.86 to 1.08; I-squared = 57.9%,
The authors would like to thank Professor Nick Black and Eun Hee Ha for their helpful response to our queries. This study was funded by Inha University and Korea Centers for Disease Control and Prevention (#2014ER270501).
The authors have no conflicts of interest associated with material presented in this paper.
Quality assessment of individual studies based on Downs and Black checklist
Effect size (ES) and 95% confidence interval (CI) of change in birth weight (g) per 10 μg/m3 PM2.5, by exposure period; size of shaded area around point estimate is proportional to weight in calculating pooled estimate. (A) Forest plot of pooled studies and (B) forest plot of pooled studies adjusting for smoking. PM2.5, particulate matter less than or equal to 2.5 μm in diameter.
Effect size (ES) (odds ratio) and 95% confidence intervals (CI) for preterm birth per 10 μg/m3 PM2.5, by exposure period. (A) Forest plot of pooled studies and (B) forest plot of pooled studies adjusting for smoking. PM2.5, particulate matter less than or equal to 2.5 μm in diameter.
Effect size (ES) (odds ratio) and 95% confidence intervals (CIs) for preterm birth per 10 μg/m3 PM10, by exposure period. (A) Forest plot of pooled studies adjusted for smoking and (B) forest plot of pooled studies not adjusting for smoking. PM10, particulate matter less than or equal to 10 μm in diameter.
Effect size (ES) (odds ratio) and 95% confidence interval (CI) of preterm birth per 10 μg/m3 PM10, by exposure period; size of shaded area around point estimate is proportional to weight in calculating pooled estimate. (A) Forest plot of pooled relatively better quality studies and (B) forest plot of pooled relatively low quality studies. PM10, particulate matter less than or equal to 10 μm in diameter.
Effect size (odds ratio) and 95% confidence interval (CI) of preterm birth per 10 μg/m3 PM10 exposure during third trimester or entire pregnancy and stratified by smoking status. PM10, particulate matter less than or equal to 10 μm in diameter.
Flow chart of selection of studies.
Effect size (ES) and 95% confidence interval (CI) of change in birth weight (g) per 10 μg/m3 PM10, by exposure period; size of shaded area around point estimate is proportional to weight in calculating pooled estimate. (A) Forest plot of pooled studies adjusting for smoking and (B) forest plot of pooled studies not adjusting for smoking. PM10, particulate matter less than or equal to 10 μm in diameter.
Effect size (ES) and 95% confidence interval (CI) of change in birth weight (g) per 10 μg/m3 PM10, by exposure period; size of shaded area around point estimate is proportional to weight in calculating pooled estimate. (A) Forest plot of pooled relatively better quality studies and (B) forest plot of pooled relatively low quality studies. PM10, particulate matter less than or equal to 10 μm in diameter.
Effect size (ES) (odds ratio) and 95% confidence intervals (CI) for preterm birth per 10 μg/m3 PM10, by exposure period. PM10, particulate matter less than or equal to 10 μm in diameter.
Contour-enhanced funnel plot for estimation of publication bias.
Characteristics of primary studies related to exposure to particulate matter and BW and preterm birth
First author | Year of publication | Location | Outcome | Pollutant | Design | Time period | Birth (n) | Exposure period | Exposure assessment | Exposure range (jig/m3) | Smoking adjusted |
---|---|---|---|---|---|---|---|---|---|---|---|
Basu et al, [ |
2004 | California, USA | CBW | PM2.5 | Cohort | Jan 1,2000 -Dec 31,2000 | 16693 | WP | MS | Range: 4.0, 34.0 | No |
Bell et al. [ |
2007 | Connecticut, Massachusetts, USA | CBW | PM2.5 | Cohort | Jan 1, 1999 - Dec 31, 2002 | 358504 | WP | MS | Mean (SD)=11.9 (1.6) | Yes |
PM10 | Mean (SD)=22.3 (5.3) | ||||||||||
Bell et al. [ |
2008 | Connecticut, Massachusetts, USA | CBW | PM2.5 | Cohort | Jan 1, 1999 - Dec 31, 2002 | 358504 | WP | MS | IQR=2.2 | Yes |
PM10 | IQR=7.4 | ||||||||||
Belletal.[ |
2010 | Connecticut, Massachusetts, USA | CBW | PM2.5 | Cohort | Jan 1,1999-Dec 31,2000 | 76788 | WPJS | MS | Mean (SD)=14.0 (2.13) | Yes |
Brauer et al. [ |
2008 | Vancouver, Canada | PTB | PM2.5 | Cohort | Jan 1, 1999 - Dec 31, 2002 | 70249 | WP | IDW | Mean (Min-Max)=5.1(1.0-7.6) | Yes |
PM10 | Mean (Min-Max)=12.5 (8.4-16.6) | ||||||||||
Chen et al. [ |
2002 | Nevada, USA | CBW | PM10 | Cohort | Jan 1,1991 - Dec 31,1999 | 39338 | WPJS | MS | Mean(SD)=31.53 (22.32) | Yes |
Currie et al. [ |
2009 | New Jersey, USA | CBW | PM10 | Cohort | Jan 1,1989-Dec 31,2003 | 312589 | TS | MS | Mean (SD)=2.97 (0.75) | Yes |
Darrow et al. [ |
2011 | Atlanta, USA | CBW | PM2.5 | Cohort | Jan 1, 1994 - Dec 31, 2004 | 406627 | TS | MS | Yes | |
PM10 | |||||||||||
Gouveia et al. [ |
2004 | Sao Paulo, Brazil | CBW | PM10 | Cohort | Jan 1,1997 -Dec 31,1997 | 179460 | TS | MS | Mean (Min-Max)=60.3 (25.5-153.0) | No |
Geer et al. [ |
2012 | Texas, USA | CBW | PM2.5 | Cohort | Jan 1, 1998 - Dec 31, 2004 | 1548904 | WP | MS | Mean (SD)=12.6 (1.0) | Yes |
PM10 | Mean (SD)=27.4 (4.1) | ||||||||||
Gehring et al. [ |
2011 | Netherlands | CBW, PTB | PM2.5 | Cohort | Jan 1,1996-Dec 31,1997 | 3853 | WPJS | LUR | Yes | |
Gray et al. [ |
2014 | North Carolina, USA | CBW | PM2.5 | Cohort | Jan 1,2002 - Dec 31,2006 | 457642 | WP | DFM | Mean (SD)=13.6 (1.7) | Yes |
Geer et al. [ |
2012 | Texas, USA | CBW | PM2.5 | Cohort | Jan 1, 1998 - Dec 31, 2004 | 1548904 | WP | MS | Mean (SD)=12.6 (1.0) | Yes |
PM10 | Mean (SD)=27.4 (4.1) | ||||||||||
Ha et al. [ |
2014 | Seoul, Korea | PTB | PM10 | Cohort | Jan 1,1998-Dec 31,2000 | 382100 | TS | MS | Mean (Min-Max)=66.21 (10.36-249.19) | No |
Hansen et al, [ |
2007 | Brisbane, Australia | CBW | PM10 | Cohort | Jan 1,2000 -Dec 31,2003 | 26617 | WPJS | MS | Mean (Min-Max)=19.6 (4.9-171.7) | No |
Hansen etal. [ |
2006 | Brisbane, Australia | PTB | PM10 | Cohort | Jan 1,2000 - Jun 30, 2003 | 28200 | WPJS | MS | Mean (Min-Max)=19.6 (4.9-171.7) | No |
Huynh et al. [ |
2006 | California, USA | PTB | PM2.5 | Case-control | Jan 1,1998 -Dec 31,2000 | 42692 | WP | MS | Mean (SD)=18.0 (5.2) | No |
Hyderetal. [ |
2014 | Connecticut, Massachusetts., USA | CBW | PM2.5 | Cohort | Jan 1,2000 - DecG31,2006 | 628131 | WPJS | MS | Mean (Min-Max)=11.91 (4.02-19.97) | Yes |
Jalaludin et al. [ |
2007 | Sydney, Australia | PTB | PM2.5 | Cohort | Jan 1, 1998 - Dec 31, 2000 | 123840 | TS | MS | Mean (SD)= 9.0 (3.94) | Yes |
PM10 | Mean (SD)=16.3 (6.38) | ||||||||||
Jedrychowski et al. [ |
2004 | Krakow, Poland | CBW | PM2.5 | Cohort | Jan 1,2001 - Mar 31,2003 | 362 | WP | PM | No | |
Jedrychowski et al, [ |
2009 | Krakow, Poland | CBW | PM2.5 | Cohort | Jan 1,2001 - Feb 28, 2004 | 481 | TS | PM | Mean (SD)=43.83 (31.91) | No |
Jedrychowski etal. [ |
2010 | Krakow, Poland | CBW | PM2.5 | Cohort | Jan 1,2001 - Feb 28, 2004 | 481 | TS | PM | No | |
Kim et al. [ |
2007 | Seoul, Korea | CBW, PTB | PM10 | Cohort | May 1,2001 - May 31,2004 | 1514 | TS | MS | No | |
Kloog et al. [ |
2012 | Massachusetts.,USA | CBW | PM2.5 | Cohort | Jan 1,2000 - Dec 31,2008 | 572272 | WPJS | LUR | Mean PD)=9.6 (5.1) | Yes |
Lee et al. [ |
2013 | Pittsburgh, USA | PTB | PM2.5 | Cohort | Jan 1, 1997 - Dec 31, 2002 | 34705 | TS | MS | Yes | |
Leem et al. [ |
2006 | Incheon, Korea | PTB | PM10 | Cohort | Jan 1,2001 - Dec 31,2002 | 52113 | TS | OBK | Range=26.00,106.39 | No |
Madsen et al. [ |
2010 | Oslo, Norway | CBW | PM2.5 | Cohort | Jan 1, 1999 - Dec 31, 2002 | 25229 | WP | MS | Mean (SD)=12.6 (0.9) | Yes |
PM10 | Mean (SD)=25.5 (3.3) | ||||||||||
Mannes et al. [ |
2005 | Sydney, Australia | CBW | PM2.5 | Cohort | Jan 1, 1998 - Dec 31, 2000 | 151458 | TS | MS | Mean (Min-Max)=9.4 (2.4-82.1) | Yes |
PM10 | Mean (Min-Max)=16.8 (3.8-104.0) | ||||||||||
Medeiros and Gouveia [ |
2005 | Sao Paulo, Brazil | CBW | PM10 | Cohort | Jan 1,1998 -Dec 31,2000 | 311735 | TS | MS | No | |
Morello-Froschetal. [ |
2010 | California, USA | CBW | PM2.5 | Cohort | Jan 1,1996-Dec 31,2006 | 3303834 | WP,JS | MS | Mean (SD)=16.7 (5.5) | No |
PM10 | Mean(SD)=31.4(11.2) | ||||||||||
Parker et al. [ |
2005 | California, USA | CBW | PM2.5 | Cohort | Jan 1,2000 -Dec 31,2000 | 18247 | WRTS | MS | Mean (SD)=15.4 (5.1) | No |
Parker and Woodruff [ |
2008 | Entire USA | CBW | PM2.5 | Cohort | Jan 1,2001 - Dec 31,2003 | 401273 | WRTS | MS | Median (25th-75th)=13.5 (10.9-16.1) | Yes |
Pedersen et al, [ |
2013 | EU | CBW | PM2.5 | Cohort | Feb 11, 1994 - Jun 2, 2011 | 50151 | WP | LUR | Mean (5th -95th)=16.5 (8.8-24.9) | Yes |
PM10 | Mean (5th -95th)=25.4 (11.6-39.4) | ||||||||||
Pereira et al. [ |
2014 | Connecticut, USA | PTB | PM2.5 | Cohort | Jan 1,2000 - Dec 31,2006 | 61688 | WRTS | MS | Yes | |
Ritz et al. [ |
2000 | California, USA | PTB | PM10 | Cohort | Jan 1,1989 -Dec 31,1993 | 97158 | TS | MS | Mean (SD)=49.3(16.9) | Yes |
Ritz et al. [ |
2007 | California, USA | PTB | PM2.5 | Case-control | Jan 1,2003 - Dec 31,2003 | 58316 | WRTS | MS | Mean (range)=20.01 (13.80, 26.68) | Yes |
Rogers and Dunlop [ |
2006 | Georgia, USA | PTB | PM10 | Case-control | Apr1,1986-Mar 30,1988 | 325 | WP | DM | Median=7.84 | Yes |
Salam et al. [ |
2005 | California, USA | CBW | PM10 | Cohort | Jan 1,1975 -Dec31,1987 | 3901 | WRTS | MS | Mean (SD)=45.8 (12.9) | Yes |
Suhet al. [ |
2008 | Seoul, Korea | PTB | PM10 | Case-control | Jan 1,2003 -Mar 31,2007 | 235 | TS | MS | No | |
Suh et al. [ |
2009 | Seoul, Korea | PTB | PM10 | Cohort | Jan 1,1998 -Dec31,2000 | 374167 | TS | MS | No | |
Wilhelm and Ritz [ |
2005 | California, USA | PTB | PM2.5 | Cohort | Jan 1,1994 -Dec 31,2000 | 639710 | WRTS | MS | No | |
PM10 | |||||||||||
Wu et al. [ |
2009 | Los Angeles, USA | PTB | PM2.5 | Cohort | Jan 1,1997 -Dec31,2006 | 81186 | WP | DM | Mean(SD)=1.82 (1.33) | No |
Yang et al. [ |
2003 | Kaohsiung, Taiwan | CBW | PM10 | Cohort | Jan 1,1995 -Dec 31,1997 | 13396 | TS | MS | Tertile (33th-67th)=(62.43-100.44) | No |
Zhao et al. [ |
2015 | Lanzhou, China | PTB | PM10 | Cohort | Jan 1,2010-Dec31,2012 | 8969 | WRTS | IDW | Mean (SD)=142.1 (17.6) | Yes |
PM10, particulate matter less than or equal to 10 μm in diameter;PM2.5, particulate matter less than or equal to 2.5 μm in diameter;WP, whole pregnancy; TS;trimester specific.CBW;change in BW;PTB, preterm birth; MS, monitoring station; DM, dispersion model; LUR, land-use regression model;IDW, inverse-distance weighting; PM, personal monitoring; DFM, downscaling fusion model;OBK, ordinary block kriging;
, first trimester;
, second trimester;
, third trimester.
, group with higher exposure levels;
IQR, interquartile range;SD, standard deviation;Min, minimum;Max, maximum.
Summary of pooled estimates of effect for change in birth weight (BW) and preterm birth (PTB) in association with a 10 μg/m3 increase in maternal exposure to PM10 and PM2.5
Meta-analysis | PM10 | PM2.5 |
---|---|---|
Change in BW (g) | ||
Unadjusted (95% CI) | -8.17 (-10.99, -5.36) | NS |
n (I-squared, %) |
3 (35.2) | - |
Adjusted for smoking (95% CI) | -10.31 (-13.57, -7.05) | -22.17 (-37.93, -6.41) |
n (I-squared, %) |
5 (0.0) | 7 (92.3) |
Combined studies (95% CI) | -6.50 (-10.94, -2.5) | -13.88 (-15.7, -12.06) |
n (I-squared, %) |
8 (76.4) | 8 (47.5) |
Odds ratio for PTB | ||
Unadjusted (95% CI) | 1.04 (1.02,1.06) |
NS |
n (I-squared, %) |
4 (0.0) | - |
Adjusted for smoking (95% CI) | 0.97 (0.86,1.08) |
1.13 (0.98,1.28) |
n (I-squared, %) |
3 (57.9) | 5 (93.0) |
Combined studies (95% CI) | 1.23 (1.04,1.41) | 1.14 (1.06,1.22) |
n (I-squared, %) |
3 (0.0) | 7 (92.5) |
PM, particulate matte; PM10, PM less than or equal to 10 μm in diameter; PM2.5, PM less than or equal to 2.5 μm in diameter; CI, confidence interval; n, number of studies to conduct a meta-analysis; NS, not sufficient number of studies to conduct a meta-analysis.
The percentage of variability due to heterogeneity between studies.
Based on the third trimester exposure. All other estimates are based on entire pregnancy exposure.