PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Environ Res. Author manuscript; available in PMC 2017 May 1.
Published in final edited form as:
PMCID: PMC4821760
NIHMSID: NIHMS763157

Particulate air pollution, fetal growth and gestational length: the influence of residential mobility in pregnancy

Abstract

Background

It remains unclear as to whether neglecting residential mobility during pregnancy introduces bias in studies investigating air pollution and adverse perinatal outcomes, as most studies assess exposure based on residence at birth. The aim of this study was to ascertain whether such bias can be observed in a study on the effects of PM10 on risk of preterm birth and fetal growth restriction.

Methods

This was a retrospective study using four pregnancy cohorts of women recruited in Connecticut, USA (N=10,025). We ascertained associations with PM10 exposure calculated using first recorded maternal address, last recorded address, and full address histories. We used a discrete time-to-event model for preterm birth, and logistic regression to investigate associations with small for gestational age (SGA) and term low birth weight (LBW).

Results

Pregnant women tended to move to areas with lower levels of PM10. For all outcomes, there was negligible difference between effect sizes corresponding to exposures calculated with first, last and full address histories. For LBW, associations were observed for exposure in second trimester (OR 1.09; 95% CI: 1.04 – 1.14 per 1μg/m3 PM10) and whole pregnancy (OR 1.08; 95% CI: 1.02 – 1.14). For SGA, associations were observed for elevated exposure in second trimester (OR 1.02; 95% CI: 1.00 – 1.04) and whole pregnancy (OR 1.03; 95% CI: 1.01 – 1.05). There was insufficient evidence for association with preterm birth.

Conclusion

PM10 was associated with both SGA and term LBW. However, there was negligible benefit in accounting for residential mobility in pregnancy in this study.

Keywords: residential mobility, pregnancy, preterm birth, fetal growth, exposure misclassification

Background

Epidemiological studies indicate that exposures to particulate matter air pollution may have adverse effects on pregnancy outcomes1, 2, with fetal growth and gestational length among the outcomes commonly investigated. Most often, ground-level measurements from a government monitoring network are used to derive exposure at a single residential address, usually recorded at delivery. However, as approximately 9% – 32% of women move during pregnancy there is potential for a high degree of exposure misclassification3. Patterns of residential mobility among pregnant women are largely unknown but studies indicate that moving is more likely among mothers who are younger 46, have lower parity 4, 5, 7, and have lower socioeconomic status 4, 5, all of whom have greater risk of delivering smaller babies and delivering preterm. Exposure misclassification might be minimal if women tend to move short distances (median, <10km)5, 8, 9. In a New York cohort, whole pregnancy exposure to particulate matter with aerodynamic diameter ≤10 μm (PM10) was essentially unchanged when based on residence recorded by maternal interviews (20.11 μg.m−3) compared to that based on the residential location recorded at delivery (20.09 μg.m−3)5. In a UK cohort, annual PM10 derived using the residential location at delivery was highly correlated with that derived using residential locations throughout pregnancy (Pearson r=0.88)10. In contrast, in another study, estimated PM10 exposure based on address at delivery compared to complete residential history differed by more than one standard deviation in 16% of pregnancies10. Consequently, it remains unclear as to whether final effect estimates on preterm birth and fetal growth restriction are biased by ignoring residential mobility in the derivation of PM10 exposure5, 1115.

The aim of this study was to compare effect estimates of particulate matter (PM10) exposure on fetal growth and gestational length, with and without accounting for residential mobility using four large pregnancy cohorts in Connecticut, between 1988 and 2008.

Methods

Study design and setting

This was a retrospective study using four pregnancy cohorts of women recruited in Connecticut, USA (N=10,025). Women were interviewed 2–4 times in pregnancy. Women were recruited at <25 weeks gestation for the Asthma in Pregnancy study16 (AIP; 1996–2000; N=2,255) and the Pink and Blue study17 (PAB; 2005–2008; N=2,645) of depression in pregnancy. Women were recruited at <16 weeks gestation for the Nutrition in Pregnancy study18 (NIP; 1996–1999; N=2,344) and Environmental Tobacco Smoke study19 (ETS; 1988–1991; N=2,781). Further details of the cohorts have been published previously1619.

Participants

We excluded women with at least one address that could not be geocoded (N=182). We sequentially excluded records with multiple gestations (N=165), missing sex (N=55), and records with missing gestational age (N=1) or gestational age > 42 weeks (N=35), which resulted in a study population of 9,587 singleton pregnancies. Women were not explicitly asked for their residential histories. Day of residential move was ascertained in the course of cohort follow-up from the point of contact at recruitment to the post-partum interview.

Outcome variables

Preterm birth (PTB) was defined as birth before 37 completed weeks of gestation. Period of gestation was obtained from the birth certificate record. This was the clinical best estimate of gestational age, based on ultrasound or last menstrual period if ultrasound was not available. Births were classified as small for gestational age (SGA) if birth weight was <10th centile for gestational age and sex20. Term low birth weight (LBW) was defined as a birth with at least 37 weeks of completed gestation attaining a birth weight <2,500g.

Exposure variables

Daily PM10 measurements from the US Environmental Protection Agency (EPA) monitoring network were obtained for all monitors within 100km of participants’ residential addresses. We calculated exposure using measurements from monitors within circular “buffer” radii of 20km, 40km, and 100km from the residential address. At each residential location and gestational week of pregnancy we calculated the 7-day average PM10 concentration using (i) measurements from the closest monitor to the residential location within the buffer distance, and (ii) the inverse distance weighted (IDW) average of measurements from all monitors within the buffer distance. These weekly means were then used to compute average PM10 concentrations for each trimester (< week 14, weeks 15–26, > week 26) and for the whole pregnancy. By definition, pregnancies are not at risk of preterm birth after gestational week 36. For this reason, only measurements prior to either birth or gestational week 36 (whichever was earlier) were included in the calculation of third-trimester and whole-pregnancy exposures for the preterm birth analyses. To ascertain the effects of acute exposure on the risk of preterm birth, we calculated mean PM10 exposure for the week of delivery and the 6-week period prior to delivery. We calculated exposures using (i) the address at recruitment (first address), (ii) the address at delivery (last address), and (iii) all addresses updated throughout pregnancy (updated addresses).

Final study population

The full study population of 9,587 singleton pregnancies was used to investigate SGA. Preterm birth was investigated separately with (i) the full study population (N=9,587), and (ii) further restriction to vaginal deliveries (N=7,334 remaining). By definition, assessment of risk of term LBW required restriction to term births (N=8,997 remaining). Characteristics of the study population are shown in Table 1.

Table 1
Maternal Characteristics at Study Entry

Statistical methods

For all models, adjustment was made for the following variables ascertained at study entry: maternal age (<20, 20–24, 25–29, 30–34, 35–39, 40+ years), race/ethnicity (White, African American, Hispanic, Asian, other), marital status (married, single, divorced/separated), highest level of educational attainment (did not complete high school, high school, post-secondary, graduate and above), parity (0, 1, 2, >=3), pre-pregnancy weight (kg), an indicator for smoking (tobacco), and an indicator for alcohol consumption (beer, wine, liquor). To adjust for temporal confounding by unmeasured factors such as long-term trends and seasonal factors we included an adjustment term for year and season of conception. For Term LBW, we also adjusted for final gestational age (weeks) due to accumulating evidence that perinatal outcomes continue to vary along the gestational age continuum from 37 weeks21. Logistic regression was used to calculate odds ratios (OR) for associations between PM10 exposure and term LBW and SGA. For PTB, a discrete time-to-event model22 was used to calculate hazard odds ratios (HOR). This model allowed calculation of prospective risk estimates that ensured comparisons were restricted to only pregnancies at risk at each week of gestation. Pregnancies entered the risk set at gestational week 20, were followed until the earlier of birth or gestational week 36 inclusive, and were censored thereafter. More specifically, we modelled Hi(t), the hazard of preterm birth for pregnancy i at week t, as:

logit(Hi(t))=logit(P(Yi(t)=1|Yi(t1)=0))=h(t)+γZi+βXi(t)+εit

where Y(t) is an indicator for birth at time t, h(t) are the week-specific intercepts, Z is the matrix of adjustment variables with corresponding parameter estimate vector γ, X(t) is the time-varying PM10 exposure with parameter estimate β, and epsilon denotes the residuals.

Bias

Multiple imputation using chained equations with 5 imputations and 5 iterations was used to minimize bias due to non-response from missing adjustment variables. Exposure (PM10) variables and outcome variables (birth weight and gestational length) were not imputed or used in the imputation. Smoking in pregnancy, alcohol consumption in pregnancy, maternal ethnicity, maternal education, marital status, pre-pregnancy weight and maternal age were imputed and used in the imputation to impute other variables. The variable “sex” is used to derive the outcome variable small for gestational age and sex (SGA), and consequently the variable “sex” was used to impute other variables but was not imputed itself. The variable “parity” and “study cohort indicator” are potential proxies for a time varying confounder and consequently were used to impute other variables but were not imputed themselves.

Results

Residential mobility

There were 1,061 (11%) women in the study population that moved during pregnancy. The median distance moved was 5km (IQR: 2km – 13km). Compared to women that did not move, women that moved did not have an elevated risk of delivering preterm (RR 1.03, 95% CI: 0.80, 1.31) or LBW (RR 0.90, 95% CI: 0.51, 1.48). However, SGA was more likely for women that moved (RR 1.40, 95% CI: 1.18, 1.67).

Exposure

There were 8,323 (87% of sample), 9,502 (99% of sample), and 9,587 (100% of sample) women who lived within 20km, 40km and 100km of a monitor during pregnancy. Overall exposure estimates were not sensitive to choice of buffer distance or exposure method (IDW vs closest monitor) (Tables S1 and S2). The median whole-pregnancy PM10 exposure was 22 μg.m−3 (IQR: 19–27 μg.m−3) using updated address histories (IDW, 20km buffer). For movers, the influence on exposure estimates of the decision to use first, last or updated address histories was sensitive to buffer distance and method. That is, we are less likely to observe differences in exposure attributable to using first, last and updated address histories with the IDW method as it uses measurements from multiple monitors which introduces greater variability in estimates. Similarly, although exposure misclassification might be reduced using smaller buffer sizes, there is an associated drop in sample size. Consequently, using the IDW method or small buffer distance (20km), exposure estimates were similar using first, last and updated address histories. However, using the closest monitor method (lower variance) and using larger buffer distances (40 km, greater sample size) we observed a small difference between mean whole pregnancy PM10 exposures calculated using the last address (l) and updated address (u) histories (l − u = −0.20 μg.m−3: 95% CI −0.37, −0.03 μg.m−3). Using a 40km buffer and the closest monitor method, whole pregnancy exposure using the last address was consistently less than that calculated using the first (f) address (l − f = −0.30 μg.m−3: 95% CI −0.59, 0.02 μg.m−3). That is, the magnitude of the difference was small but the direction of the effect was consistent, indicating women who moved tended to relocate to areas with lower levels of PM10.

Associations with pregnancy outcomes

For all outcomes, there was negligible difference between effect sizes corresponding to exposures calculated with first, last and updated address histories (Figures 14). That is, there was near complete overlap in the interval estimates (using first, last and updated address histories) for term LBW (Figure 1), SGA (Figure 2), PTB (Figure 3) and vaginal PTB (Figure 4). There was insufficient evidence that effect sizes differed by buffer distance (20km, 40km, 100km). However, statistically significant associations were observed using the 20km buffer but not the 40km or 100km buffer distances for SGA and term LBW. The IDW method resulted in less precise interval estimates (i.e. wider 95% CIs) than using the closest monitor method, with no observable difference (bias) between the point estimates. Consequently, we describe hereon the adjusted odds ratios and hazard odds ratios for increases (1 μg.m−3) in PM10 using updated address histories, a 20km buffer and the IDW method. For LBW, statistically significant associations were observed for elevated exposure in second trimester (OR 1.09; 95% CI: 1.04 – 1.14) and whole pregnancy (OR 1.08; 95% CI: 1.02 – 1.14) (Figure 1). Similarly, for SGA, associations were observed for elevated exposure in second trimester (OR 1.02; 95% CI: 1.00 – 1.04) and whole pregnancy (OR 1.03; 95% CI: 1.01 – 1.05) (Figure 2). There was insufficient evidence for an association between PTB and exposure to PM10 for cumulative exposure in trimesters, cumulative exposure over the whole of pregnancy, or exposure closer in the week preceding delivery or 6 weeks preceding delivery (Figure 3, Figure 4). The adjusted HOR of PTB for whole pregnancy exposure was 1.01 (95% CI: 0.98 – 1.04) and 1.00 (95% CI: 0.99 – 1.01) for elevated exposure during the week preceding delivery (Figure 3). The results were similar for PTB after restricting to vaginal deliveries. The adjusted HOR of PTB for whole pregnancy exposure was 1.01 (95% CI: 0.98 – 1.03) and 1.00 (95% CI: 0.99 – 1.01) for elevated exposure during the week preceding delivery (Figure 4).

Figure 1
Adjusted log odds ratios for term LBW for a 1 μg.m−3 increase in PM10 in each trimester (T1, T2, T3) and whole pregnancy (P) ascertained with first address, last address and updated histories. Results presented for each buffer distance ...
Figure 2
Adjusted log odds ratios for SGA for a 1 μg.m−3 increase in PM10 in each trimester (T1, T2, T3) and whole pregnancy (P) ascertained with first address, last address and updated histories. Results presented for each buffer distance (20km, ...
Figure 3
Adjusted log hazard odds ratios for PTB for a 1 μg.m−3 increase in PM10 in each trimester (T1, T2, T3), whole pregnancy (P), week of birth (lag 0) and the 6-week period prior to birth (lag 05). Exposure was ascertained with first address, ...
Figure 4
Adjusted log hazard odds ratios for PTB for a 1 μg.m−3 increase in PM10 in each trimester (T1, T2, T3), whole pregnancy (P), week of birth (lag 0) and the6-week period prior to birth (lag 05). Exposure was ascertained with first address, ...

Discussion

Key results

We compared effect estimates of particulate matter (PM10) exposure on fetal growth and gestational length, with and without accounting for residential mobility using four large pregnancy cohorts in Connecticut and western Massachusetts, between 1988 and 2008. The results indicate that, at the levels of residential mobility observed in this study population (11%), the induced level of exposure misclassification for PM10 had a negligible influence on overall effect estimates. It is plausible that the influence on final effect estimates of moves over short distances is negligible, because residential exposure is intended to be proxy for exposure for time spent in the broader region about the residence. Interestingly, women who moved tended to move to areas of lower PM10 air pollution, characterized in this study by movement away from a city. This observation was not due to decreasing temporal trend because for each address, exposure was calculated for the same period. That is, exposure was calculated using the first address for each exposure period (trimesters and whole pregnancy), then calculated using the last address for each exposure period, and finally calculated using updated address histories for each exposure period. Therefore, we could compare spatial differences in results independently of temporal trend.

Interpretation

Exposure misclassification is induced by residential mobility during pregnancy when the time of ascertainment of the residential location is not temporally well aligned with the exposure period under investigation. We observed that such misclassification had little influence on the observed estimates of the effect of PM10 exposure on restricted fetal growth and gestational length in this study population. This provides greater credibility to past perinatal studies when the particulate matter exposure contrast under investigation is largely due to city-wide spatial comparisons and daily temporal comparisons1215, 23, 24. For fetal growth endpoints (SGA and term LBW), choice of method of exposure assessment (closest monitor vs IDW) and buffer size had greater influence on effect estimates and their precision than the extent of ascertainment of residential mobility in pregnancy.

Generalisability

A combination of factors contribute to the influence of residential mobility in pregnancy on effect estimates. Exposure misclassification increases with the fraction of women that move during pregnancy (11% in this study). An inherent assumption is that the process that governs residential mobility in pregnancy does not differ by outcome status (e.g. PTB vs term birth). The link between residential mobility and socioeconomic factors alone is sufficient reason to suggest that this assumption is often violated3, 25. However, this assumption is only important to these epidemiological investigations if exposure misclassification from inaccurate assessment of residential mobility leads to differential exposure misclassification. Although we did not observe differences in effect estimates obtained with and without accounting for residential mobility in this study, this result is not necessarily generalizable to other studies. For our study, a small fraction of women moved (11%), and women moved short distances (median 5km) relative to the spatial scale of exposure assessment (city-wide comparisons). Moreover, the exposure periods of interest for pregnancy outcomes (e.g., trimesters) are often long enough so that the recorded residential address is accurate for at least a portion of that exposure period. Naturally, greater misclassification is expected, for example, when residential address is recorded at delivery but used to calculate exposure near conception. The choice of the outcome itself influences the comparability in levels of residential mobility by outcome status. Inherently, preterm birth provides less opportunity to move than a term pregnancy. Finally, we note that it remains possible that the influence of residential mobility might not be observable overall but might be observable for sub-populations already at elevated risk of restricted fetal growth or gestational length e.g., socioeconomically disadvantaged groups. Assessment of this a posteriori hypothesis requires selecting a sample that targets such sub-populations, rather than the population representative samples used in this study.

Limitations

As this study was based on pregnancy cohorts, residential mobility prior to recruitment was not ascertainable. Therefore, our results are most relevant for differences in PM10 exposure and associated effects of this exposure after first trimester. This limitation is addressable in future pregnancy cohort studies by retrospective assessment of residential histories at recruitment. In some settings, residential histories can be more objectively ascertained via linkage to national health surveillance systems26. However, these systems tend to record residential location at the time of health service contact, not the time of the move. Diurnal activity patterns may also influence accuracy of exposure classification based on home address in relation to time spent at home. In one study, 3.5 more hours per day were spent at home by pregnant women who did not work, 2.6 more hours were spent at home by those with low income and 1.5 more hours were spent at home by those already at home with children27. It remains uncertain as to whether diurnal activity patterns changed after residential movement and whether this change was associated with both a consistent directional bias in PM2.5 exposure and risk of adverse perinatal outcome (SGA, LBW, PTB).

Conclusion

In general, pregnant women tended to move to areas with lower levels of PM10. The influence of residential mobility on effect estimates will be a function of the moving patterns in the population, or subpopulation, and the exposure of interest. In this study, there was negligible benefit in accounting for residential mobility in pregnancy as the observed effects of PM10 exposure on fetal growth and gestational length remained unchanged.

Highlights

  • PM10 exposure was associated with risk of low birth weight and small for gestational age, but not preterm birth.
  • We provide empirical evidence that there is negligible difference in effect estimates after accounting for residential mobility in pregnancy, and is generalizable to past studies and other settings for which the spatial variation in assessed exposure was regional (e.g., city-wide) and women tend move short distances.
  • Interestingly, women who moved during pregnancy tended to move to areas with lower levels of PM10 air pollution.
  • Choice of method of exposure assessment and buffer size had greater influence on effect estimates and their precision than the extent of ascertainment of residential mobility in pregnancy.

Supplementary Material

supplement

Acknowledgments

Source of funding: This work was supported by a National Health and Medical Research Council Early Career Fellowship grant [1052236 to GP], National Institute of Environmental Health Sciences grants [R01ES016317 and R01ES019587 to MLB, MB, and GP].

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Conflicts of interest: none declared.

No competing financial interests.

Ethical approval for this study was obtained from the Yale Human Investigation Committee.

References

1. Sapkota A, Chelikowsky AP, Nachman KE, Cohen AJ, Ritz B. Exposure to particulate matter and adverse birth outcomes: a comprehensive review and meta-analysis. Air Quality, Atmosphere & Health. 2012;5:369–81.
2. Zhu X, Liu Y, Chen Y, Yao C, Che Z, Cao J. Maternal exposure to fine particulate matter (PM2. 5) and pregnancy outcomes: a meta-analysis. Environmental Science and Pollution Research. 2014:1–14.
3. Bell ML, Belanger K. Review of research on residential mobility during pregnancy: consequences for assessment of prenatal environmental exposures. Journal of exposure science & environmental epidemiology. 2012;22:429–38. [PMC free article] [PubMed]
4. Canfield MA, Ramadhani TA, Langlois PH, Waller DK. Residential mobility patterns and exposure misclassification in epidemiologic studies of birth defects. J Expo Sci Environ Epidemiol. 2006;16:538–43. [PubMed]
5. Chen L, Bell EM, Caton AR, Druschel CM, Lin S. Residential mobility during pregnancy and the potential for ambient air pollution exposure misclassification. Environmental research. 2010;110:162–8. [PubMed]
6. Raynes-Greenow CH, Nassar N, Roberts CL. Residential mobility in a cohort of primiparous women during pregnancy and post-partum. Australian and New Zealand journal of public health. 2008;32:131–4. [PubMed]
7. Miller A, Siffel C, Correa A. Residential mobility during pregnancy: patterns and correlates. Maternal and child health journal. 2010;14:625–34. [PubMed]
8. Hodgson S, Shirley M, Bythell M, Rankin J. Residential mobility during pregnancy in the north of England. BMC pregnancy and childbirth. 2009;9:52. [PMC free article] [PubMed]
9. Lupo PJ, Symanski E, Chan W, et al. Differences in exposure assignment between conception and delivery: the impact of maternal mobility. Paediatric and perinatal epidemiology. 2010;24:200–8. [PubMed]
10. Hodgson S, Lurz PW, Shirley MD, Bythell M, Rankin J. Exposure misclassification due to residential mobility during pregnancy. International journal of hygiene and environmental health. 2015;218:414–21. [PubMed]
11. Lee P-C, Roberts J, Catov J, Talbott E, Ritz B. First Trimester Exposure to Ambient Air Pollution, Pregnancy Complications and Adverse Birth Outcomes in Allegheny County, PA. Maternal and child health journal. 2013;17:545–55. [PMC free article] [PubMed]
12. Pereira G, Belanger K, Ebisu K, Bell ML. Fine particulate matter and risk of preterm birth in Connecticut in 2000–2006: a longitudinal study. American journal of epidemiology. 2013:kwt216. [PMC free article] [PubMed]
13. Pereira G, Bell ML, Belanger K, de Klerk N. Fine particulate matter and risk of preterm birth and pre-labor rupture of membranes in Perth, Western Australia 1997–2007: A longitudinal study. Environment international. 2014;73:143–9. [PMC free article] [PubMed]
14. Pereira G, Bell ML, Lee HJ, Koutrakis P, Belanger K. Sources of fine particulate matter and risk of preterm birth in Connecticut, 2000–2006: a longitudinal study. Environmental health perspectives. 2014;122:1117. [PMC free article] [PubMed]
15. Pereira G, Evans KA, Rich DQ, Bracken MB, Bell ML. Fine Particulates, Preterm Birth, and Membrane Rupture in Rochester, NY. Epidemiology. 2015;20:00. [PubMed]
16. Triche EW, Saftlas AF, Belanger K, Leaderer BP, Bracken MB. Association of asthma diagnosis, severity, symptoms, and treatment with risk of preeclampsia. Obstetrics & Gynecology. 2004;104:585–93. [PubMed]
17. Spoozak L, Gotman N, Smith MV, Belanger K, Yonkers KA. Evaluation of a social support measure that may indicate risk of depression during pregnancy. Journal of affective disorders. 2009;114:216–23. [PMC free article] [PubMed]
18. Bracken MB, Triche EW, Belanger K, Hellenbrand K, Leaderer BP. Association of maternal caffeine consumption with decrements in fetal growth. American journal of epidemiology. 2003;157:456–66. [PubMed]
19. Sadler L, Belanger K, Saftlas A, et al. Environmental tobacco smoke exposure and small-for-gestational-age birth. American journal of epidemiology. 1999;150:695–705. [PubMed]
20. Oken E, Kleinman KP, Rich-Edwards J, Gillman MW. A nearly continuous measure of birth weight for gestational age using a United States national reference. BMC pediatrics. 2003;3:6. [PMC free article] [PubMed]
21. Gouyon JB, Vintejoux A, Sagot P, Burguet A, Quantin C, Ferdynus C. Neonatal outcome associated with singleton birth at 34–41 weeks of gestation. International Journal of Epidemiology. 2010;39:769–76. [PubMed]
22. Chang HH, Reich BJ, Miranda ML. Time-to-Event Analysis of Fine Particle Air Pollution and Preterm Birth: Results From North Carolina, 2001–2005. American Journal of Epidemiology. 2012;175:91–8. [PubMed]
23. Bell ML, Ebisu K, Belanger K. Ambient air pollution and low birth weight in Connecticut and Massachusetts. Environmental Health Perspectives. 2007:1118–24. [PMC free article] [PubMed]
24. Ebisu K, Belanger K, Bell ML. Association between airborne PM2.5 chemical constituents and birth weight—implication of buffer exposure assignment. Environmental Research Letters. 2014;9:084007. [PMC free article] [PubMed]
25. Saadeh F, Clark M, Rogers M, et al. Pregnant and Moving: Understanding Residential Mobility during Pregnancy and in the First Year of Life using a Prospective Birth Cohort. Maternal and child health journal. 2013;17:330–43. [PubMed]
26. Hodgson S, Lurz PWW, Shirley MDF, Bythell M, Rankin J. Exposure misclassification due to residential mobility during pregnancy. International Journal of Hygiene and Environmental Health. 2015;218:414–21. [PubMed]
27. Nethery E, Brauer M, Janssen P. Time–activity patterns of pregnant women and changes during the course of pregnancy. Journal of Exposure Science and Environmental Epidemiology. 2009;19:317–24. [PubMed]