To assess associations between unconventional natural gas development (UGD) and perinatal outcomes.
We conducted a retrospective birth cohort study among 158,894 women with a birth or fetal death from November 30, 2010-November 29, 2012 in the Barnett Shale, in North Texas. We constructed three UGD-activity metrics by calculating the inverse distance-weighted sum of active wells within three separate geographic buffers surrounding the maternal residence: ≤½, 2, or 10-miles. We excluded women if the nearest well to her residence was >20 miles. Metrics were categorized by tertiles among women with ≥1 well within the respective buffer; women with zero wells ≤10 miles (the largest buffer) served as a common referent group. We used logistic or linear regression with generalized estimating equations to assess associations between UGD-activity and preterm birth, small-for-gestational age (SGA), fetal death, or birthweight. Adjusted models of fetal death and birthweight included: maternal age, race/ethnicity, education, pre-pregnancy body mass index, parity, smoking, adequacy of prenatal care, previous poor pregnancy outcome, and infant sex. Preterm birth models included all of the above except parity; SGA models included all of the above except previous poor pregnancy outcome.
We found increased adjusted odds of preterm birth associated with UGD-activity in the highest tertiles of the ½- (odds ratio (OR) = 1.14; 95% confidence interval 1.03, 1.25), 2- (1.14; 1.07, 1.22), and 10-mile (1.15; 1.08, 1.22) metrics. Increased adjusted odds of fetal death were found in the second tertile of the 2-mile metric (1.56; 1.16, 2.11) and the highest tertile of the 10-mile metric (1.34; 1.04–1.72). We found little indication of an association with SGA or term birthweight.
Citation: Whitworth KW, Marshall AK, Symanski E (2017) Maternal residential proximity to unconventional gas development and perinatal outcomes among a diverse urban population in Texas. PLoS ONE 12(7): e0180966. https://doi.org/10.1371/journal.pone.0180966
Editor: Jaymie Meliker, Stony Brook University, Graduate Program in Public Health, UNITED STATES
Received: April 24, 2017; Accepted: June 23, 2017; Published: July 21, 2017
Copyright: © 2017 Whitworth et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: Birth and fetal death records are owned by the TX Department of State Health Services, Center for Health Statistics. Researchers may access these data by submitting a request through the Center for Health Statistics and obtaining Institutional Review Board approval at the TX DSHS (https://www.dshs.texas.gov/irb/applirb.shtm). Data regarding unconventionally drilled gas wells used in the study are owned by DrillingInfo; subscriptions to this data source can be obtained at www.drillinginfo.com.
Funding: This study was funded by the National Institute of Environmental Health Sciences (NIEHS), Grant number 1 R03 ES023954-01. AKM was supported by an Occupational Epidemiology Traineeship funded by Grant No. T42OH008421 from the National Institute for Occupational Safety and Health (NIOSH)/Centers for Disease Control and Prevention (CDC) to the Southwest Center for Occupational and Environmental Health, a NIOSH-funded Education and Research Center.
Competing interests: The authors have declared that no competing interests exist.
Advancements in horizontal drilling and hydraulic fracturing have increased access to previously untapped natural gas reserves in shale formations. Unconventional natural gas development (UGD) is associated with several potential environmental hazards. The hydraulic fracturing process involves injecting a pressurized mixture of sand, water, and proprietary fracking fluid into wellbores, fracturing the rock and unlocking trapped hydrocarbons . Fracking fluid may contain compounds that are known or possible human carcinogens, regulated under the Safe Drinking Water Act, or classified as hazardous air pollutants . Ground and surface water contamination can occur from migration of fluids through failed well casings, leakage from open pit storage, and improper disposal or treatment of wastewater [3–7]. Further, many compounds found in fracking fluid and wastewater have been indicated for their reproductive or developmental toxicity [8, 9]. Multiple air pollutants including volatile organic compounds (VOCs) (e.g., toluene, benzene), polycyclic aromatic hydrocarbons (e.g., naphthalene, benzo(a)pyrene), nitrogen oxides, ozone, and particulate matter have also been detected near unconventional drilling sites [6, 10–14]. In addition to potential chemical exposures, individuals living in communities near UGD may experience noise and light pollution, noxious odors, and increased psychosocial stressors [15, 16]. Non-chemical stressors can contribute to allostatic load, reducing overall health and wellbeing , and potentially increase susceptibility to chemical stressors .
Results of the few epidemiologic studies of the association between maternal residential proximity to UGD and perinatal outcomes are equivocal [19–21]. Although exposure timing is a critical consideration in such studies, only one of these previous studies limited analyses to UGD-activity occurring specifically during the gestational period . The remaining two studies captured all UGD-activity during the year of the child’s birth [19, 20]. Additionally, previous studies have been conducted among mostly white, mostly rural populations. Our study includes women living in the Barnett Shale, one of the oldest and most developed shale plays in the United States. UGD-activity in the Barnett Shale is concentrated in and around the Dallas-Fort Worth metroplex, the fourth largest metropolitan statistical area in the nation and home to a heterogeneous population . Our goal was to assess the association between maternal residential proximity to UGD-activity and perinatal outcomes, considering timing of UGD-activity relative to pregnancy. Due to the dearth of data informing: (a) the most relevant distances within which to capture impacts of chemical and non-chemical stressors related to UGD and (b) implications of such decisions on health effect estimates, we also examined the characterization of proximity to UGD-activity according to several distance criteria.
Materials and methods
This retrospective birth cohort study included women with a singleton birth or fetal death from November 30, 2010-November 29, 2012 in the 24-county Barnett Shale area (Archer, Bosque, Clay, Comanche, Cooke, Coryell, Dallas, Denton, Eastland, Ellis, Erath, Hill, Hood, Jack, Johnson, Montague, Palo Pinto, Parker, Shackelford, Somervell, Stephens, Tarrant, Wise, and Young counties), in North Texas. Birth and fetal death records were obtained from the Texas Department of State Health Services (TXDSHS) for 166,966 births and 866 fetal deaths. This study was approved by the Committee for Protection of Human Subjects at The University of Texas Health Science Center at Houston and the TXDSHS IRB.
We corrected implausible birthweight for gestational age combinations for live births, according to methods previously described [23–25]. For live births, we estimated the conception date by subtracting gestational age from the child’s birth date. For fetal deaths, we estimated the conception date by subtracting the last menstrual period (LMP)-based estimate of gestational age from the date of death; for records missing the LMP-based estimate (n = 223), we used the clinical estimate. We defined small-for-gestational age (SGA; yes/no) as birthweight for gestational age ≤10th percentile of the sex-specific weight for age distribution in our study sample. Preterm birth (yes/no) was defined as a live birth delivered before 37 completed weeks gestation. We identified fetal deaths from death records (yes/no). We obtained birthweight (g) from birth records and treated it as a continuous outcome.
Street-level geocoded location of maternal residence at birth was provided by TXDSHS  for the majority of records and we manually geocoded physical addresses for the remaining records using ArcMap (v. 10.2.1; ESRI, Redlands CA). We obtained the following covariates from birth/fetal death records: maternal age (≤20, 21–25, 26–30, 31–35, >35 years), education (<high school, high school graduate, some college, college graduate), parity (0, ≥1), smoking during pregnancy (yes/no), race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, other), pre-pregnancy body mass index (BMI: ≤18.5 kg/m2, 18.5–24.9 kg/m2, 25.0–29.9 kg/m2, 30.0–34.9 kg/m2, ≥35.0 kg/m2), infant sex, and previous poor pregnancy outcome (i.e., previous perinatal death, intrauterine growth restriction, pregnancy termination, preterm birth, or SGA; yes/no). We also constructed the Adequacy of Prenatal Care Utilization Index (inadequate, intermediate, adequate, adequate plus, unknown) which captures timing of first prenatal visit and frequency of visits . The ‘unknown’ category includes women for whom date of first visit or number of visits was missing, but for whom records indicated prenatal care was received. The ‘adequate plus’ category indicates receipt of more than the recommended number of visits (i.e., one visit/month for weeks up to 28, two visits/month for weeks 29 to 36, and weekly visits from 37 weeks on, as outlined in Kotelchuck ), presumably due to high-risk pregnancies. Finally, we calculated the exact geodesic line-distance from the residence to the nearest major roadway as a proxy for traffic-related air pollution (<300m, ≥300m) .
We obtained UGD data from DrillingInfo (www.drillinginfo.com), a commercial site which maintains a national database of oil and gas well locations and characteristics (updated twice monthly)  on May 12, 2015. We identified unconventional (i.e., horizontal/directionally drilled) gas wells in the Barnett Shale with either spud (i.e., earliest known date ground was broken in the process of well development), completion (i.e., date when installation of the well casing, pumping mechanism, and hydraulic fracturing were completed)  or production dates between January 1, 2010–November 29, 2012. Wells can be completed more than once, often to stimulate production, and multiple completion dates may be reported . In this case, we retained the most recent date. We captured active wells beginning January 1, 2010 to characterize UGD-activity for the entire pregnancy for all births in the cohort. We did not include wells that had a permit date but no record of other activity. In total, we identified 14,351 unique active UGD wells.
We constructed three separate exposure metrics by generating geographic buffers around each residence, at distances of ½, 2, and 10 miles. One-half mile was chosen on the basis of a previous risk assessment , 10 miles was chosen to be consistent with prior studies [19, 20], and 2 miles was chosen as an intermediate. For each woman’s residence, we calculated the inverse distance-weighted (IDW) sum of active UGD wells within each buffer according to: where ‘a’ indicates buffer distance, ‘i’ is a given well in the specified buffer, ‘d’ is the exact geodesic line distance between that well and the residence, and ‘n’ is the total number of wells in the specified buffer. Because women living very far from UGD wells likely differ from women living near UGD activity, women for whom the nearest well was >20 miles from the residence were excluded. We then categorized each of the three metrics by tertiles among women with ≥1 well within the respective buffer. To enable comparison of effect estimates across the three metrics, we chose a common referent group for all analyses: women with zero wells ≤10 miles of her residence (given exclusion of women for whom the nearest well was >20 miles, this group effectively represents women for whom the nearest well is ≥10 miles but >20 miles away).
We used logistic regression to examine the relation between each UGD-activity metric and preterm birth, SGA, and fetal death and linear regression to examine the relation with birthweight. Given potential correlation among women within census-tracts, we applied generalized estimating equations to all models, assuming an exchangeable correlation structure and treating census tract as a random effect. We included maternal age, pre-pregnancy BMI, and maternal race/ethnicity in all models a priori. We identified additional covariates separately for each outcome. Covariates which were statistically significantly (p<0.05) associated with the respective outcome were included in the final adjusted model. In this way, a common set of variables were included in all adjusted models for each outcome, irrespective of the UGD-metric, facilitating comparison across metrics. In addition to the a priori variables, education, parity, smoking status, infant sex, previous poor pregnancy outcome, and the Adequacy of Prenatal Care Utilization Index were included in models of birthweight and fetal death. The preterm birth models did not include parity and SGA models did not include previous poor pregnancy outcome.
Given the association of maternal residential distance to the nearest major roadway with pregnancy outcomes in previous studies , we conducted a sensitivity analysis including this variable in the adjusted models. In a second sensitivity analysis, we controlled for season of conception, categorized as October-March or April-September, given the weather patterns in Texas. All analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC) or ArcGIS version 10.2 (ESRI, Redlands, CA).
Fig 1 outlines the study process and exclusions. Briefly, through the process of cleaning gestational age [23–25], we excluded 28 (<1%) births missing both LMP- and clinical-based estimates of gestational age as well as 185 (<1%) births with estimated gestational age <22 or >44 completed weeks (Fig 1). A total of 227 births were excluded due to implausible/improbable gestational age estimates. We excluded five fetal deaths with no estimate of gestational age. Among records with a street-level geocode, 1,149 (<1%) were located outside the study area and were also excluded. An additional 5,764 (3.5%) subjects were excluded because the nearest UGD well was >20 miles from the residence. The final sample was 158,894: 158,104 live births and 790 fetal deaths.
Note: GA = Gestational Age; TX DSHS = Texas Department of State Health Services.
Women in this study were mostly young (31.5% were >30 years) with at least a high school education (79.1%) and did not smoke during pregnancy (95.6%) (Table 1). Hispanics (39.7%) comprised the largest racial/ethnic group, followed by non-Hispanics whites (37.4%) and Blacks (16.1%). More than one-third (36.2%) of women had less than adequate prenatal care and 18.1% received adequate plus care, suggesting higher risk pregnancies.
The proportion of women with ≥1 active UGD well near her residence during pregnancy varied by distance within which UGD was captured: 15.9% at ½ mile, 45.1% at 2 miles, and 75.8% at 10 miles. The median number of proximal wells during pregnancy increased with buffer size: three ≤½ mile, 28 ≤2 miles and 413 ≤10 miles. This divergence was more apparent at the extremes of the distribution: maximum wells ≤10 miles of the residence during pregnancy was >2,000, versus 32 wells ≤½ mile. We observed similar patterns in the distribution of IDW metrics (Table 2).
Crude associations between UGD-activity and preterm birth were largely null for each UGD-metric. After adjustment, we found increased odds of preterm birth associated with UGD-activity in the highest tertiles of the ½- (odds ratio (OR) 1.14; 95% confidence interval (CI) 1.03, 1.25), 2- (OR 1.14; 95% CI 1.07, 1.22), and 10- (OR 1.15; 95% CI 1.08, 1.22) mile metrics (Table 3). The highest odds of preterm birth were found among women classified in the second tertile of the ½- mile metric compared to women with zero wells ≤10 miles of her residence (OR 1.21, 95% CI 1.09, 1.33).
We found indication of weak inverse associations between UGD-activity and SGA within each distance in the crude models (Table 3). Little evidence of an association was observed in adjusted models, though, as results from adjusted models were attenuated (Table 3).
Little evidence of association between UGD-activity and fetal death was observed in crude models (Table 3). Though the estimate was imprecise, we found increased adjusted odds of fetal death among women classified in the highest tertile of UGD-activity for the ½-mile metric (OR 1.27, 95% CI 0.82, 1.97) and in the 2nd tertile of the 2-mile metric (OR 1.56, 95% CI 1.16, 1.58). We observed the strongest association between UGD-activity and fetal death using the 10-mile metric. We observed increased adjusted odds of fetal death among women in each tertile of the 10-mile UGD-activity metric: 1st tertile OR 1.26 (95% CI 0.99, 1.60), 2nd tertile OR 1.22 (95% CI 0.95, 1.57), 3rd tertile OR 1.34 (95% CI 1.04, 1.72).
Crude models between UGD-activity and birthweight revealed positive associations (Table 4). However, after adjusting for confounders, we found negative associations that were only moderate in strength. For example, compared to women with zero wells ≤10-miles of her home, we found an 8.20 g decrease (95% CI -18.36, 1.96) and 7.75 g decrease (95% CI -15.94, 0.44) in average birthweight among infants of women in the 2nd tertile of the ½- and 2-mile metrics, respectively. Infants born to women classified in the 1st and 3rd tertiles of the 10-mile metric had birthweights, on average, 7.36 g (95% CI -14.79, 0.08) and 6.56 g (95% CI -13.68, 0.56) less than infants of women in the referent group.
We found evidence of a moderate positive association between maternal residential proximity to UGD-activity and increased odds of preterm birth and a suggestive association with fetal death. Not surprisingly, we found that the characterization of UGD was dependent upon the distance within which activity was defined.
In contrast to our findings, a Colorado (CO) based study  reported an inverse association between UGD and preterm birth (3rd tertile OR 0.91, 95% CI 0.85, 0.98) as well as a positive association with birthweight (3rd tertile b 22, 95% CI 15, 29). The authors of that study defined UGD-activity based on active wells ≤10 miles of the maternal residence at any time during the child’s birth year. The authors also restricted the analysis to women in rural areas, given concerns of confounding by other sources of air pollution. A similar restriction was not feasible in the present study because the majority of UGD in the Barnett Shale occurs in urban/suburban areas. However, we did not observe meaningful changes in effect estimates when adjusting for proximity to the nearest major roadway, a marker for traffic-related air pollution.
Two studies of UGD and birth outcomes in Pennsylvania (PA) have conflicting results. Stacy et al.  employed an activity metric similar to that in the CO study : all wells ≤10 miles of the maternal residence during the child’s birth year were included. In the second PA study, Casey et al.  included all wells in the state, regardless of their distance from the woman’s residence, but only considered UGD wells active during a woman’s pregnancy. Stacy et al.  found increased odds of SGA among women classified in the highest versus lowest UGD-activity quartile (OR 1.34; 95% CI 1.10, 1.63), but no association with preterm birth. In contrast, Casey et al.  reported increased odds of preterm birth among women in the 2nd (OR 1.3, 95% CI 1.0, 1.8), 3rd (OR 1.6, 95% CI 1.1, 2.4), and 4th (OR 1.9, 95% CI 1.2, 2.9) UGD-activity quartiles, but no association with SGA. Stacy et al.  also found a 21.8 g decrease (p = 0.02) in birthweight among infants born to women in the highest versus lowest UGD-activity quartile. Though Stacy et al.  included gestational age in their birthweight models, we chose not to adjust for gestational age given its potential to act as a collider and thus, bias estimates of effect .
Given key differences in drilling characteristics (including density and distribution of wells in urban/rural areas) as well as population characteristics, it is difficult to directly compare our results with previous studies’ findings. For example, there were 509 total active wells in the Stacy et al.  study and the most highly ‘exposed’ women had ≥ six wells ≤10 miles of her home. Casey et al.  noted the greatest density of UGD near women’s homes was 122 wells ≤20 km. The maximum number of wells near women’s homes was not reported by McKenzie et al. . Our study included >13,000 UGD wells and the median number of wells ≤10 miles of the residence of the most highly ‘exposed’ women was 1,188. Additionally, the racial/ethnic makeup of women in this study is quite different from previous studies, which have included primarily (73–97%) non-Hispanic white women [19–21]. Women in our study represent a more diverse population: 39.7% Hispanic, 16.1% Black, and 37.4% non-Hispanic white. Lastly, women in our study are from primarily urban/suburban areas compared with more rural populations in previous studies.
Both air and water contamination have been linked with UGD-activity, including UGD in the Barnett Shale. For example, air pollution models indicate urban drilling is a significant contributor to ambient ozone in the Barnett Shale , which may point to increased air toxics concentrations, given secondary formation of ozone through reactions between nitrogen oxides and VOCs . Investigators have also demonstrated pollutant migration and groundwater contamination related to natural gas production in the Barnett Shale area . In addition to chemical contamination, communities near UGD may be burdened by non-chemical stressors (see directed acyclic graph developed by Casey et al. ), which could affect health outcomes through altered allostatic load [37, 38]. In some areas, UGD occurs 24-hours a day and involves generator noise, increased truck traffic, noxious odors, and light pollution . The temporarily increased workforce can lead to transient population growth with accompanying demands for goods and services . UGD’s presence may also contribute to conflict and distrust, as well as division, within communities . Such “boomtown” psychosocial effects have been indicated in UGD risk assessments as potential drivers of adverse health outcomes [11, 15, 31].
Endocrine disruption has been suggested as a possible mechanism through which UGD-related contaminants may increase risk of adverse perinatal outcomes including stillbirth, preterm birth, and decreased birthweight [9, 41, 42]. Some air pollutants may also affect preterm birth via oxidative stress, endothelial dysfunction, or inflammation [43, 44]. The maternal stress response (resulting from either chemical or non-chemical stressors) may also result in ‘dysregulated parturition’ and an altered ‘pregnancy clock’, ultimately leading to preterm delivery [45, 46]. The mechanism through which exposure to UGD-related contaminants may result in fetal death is less clear. It has been posited that some air pollutants may be directly transported across the placenta resulting in hypoxia or immune-mediated injury of the fetus . Reduced oxygen-carrying capacity of maternal hemoglobin and alterations in transplacental function have also been suggested . Fetuses may be spontaneously terminated among women in particularly stressful circumstances [49, 50], providing an additional mechanism through which fetal death may be impacted among women near UGD who experience chemical and non-chemical stressors.
One of the strengths of this study was the large sample size, even when considering UGD-activity relatively close to women’s homes. Though it is a rare outcome, our large sample size allowed us to explore the association between UGD-activity and fetal death, which other studies have not considered. It is possible that women with preterm births or fetal deaths may be assigned lower values for UGD-activity than women with term births simply due to their shorter gestational periods. However, we anticipate that the result of any such bias would be toward the null. Because spatial autocorrelation can lead to biased estimates when using spatially derived data , we also assessed potential impact of clustering of women within census-tracts using GEEs with an exchangeable error structure. Impact of maternal residential mobility may also result in bias. However, though Canfield et al.  report that approximately 32% of women in Texas change residence between conception and delivery, Lupo et al.  found, among women who do move, assignment of area-level exposure was not largely impacted by the use of residence at conception versus residence at delivery.
Though the exposure metric utilized in this study was non-specific, our goal was not to examine risk related to any specific chemical. Rather, it was to examine effects of living near increased UGD-activity, which encompasses potential exposure to a multitude of chemical and non-chemical stressors, through a variety of pathways. Our use of proximity to UGD-activity assumes that women who reside near wells are more likely to be exposed (or to be exposed at higher levels) to these stressors than women living farther away. Still, there is some uncertainty regarding the ideal distance within which to capture UGD-activity.
Prior studies evaluating maternal residential proximity to UGD and birth outcomes focused on UGD ≤10 miles of the maternal residence. However, we posit it may be more plausible for UGD-activity to affect perinatal outcomes (via increased chemical and non-chemical stressors) at a much smaller distance. Many air toxics, such as benzene, are highly volatile in the atmosphere and undergo degradation relatively quickly –thus emissions of such pollutants are more likely to influence exposure of populations living near to, rather than far from, the source. McKenzie et al.  conducted an air pollution-focused risk assessment in Garfield County, CO and noted that residents ≤½-mile of a well pad had greater non-cancer health hazards than residents living further away. Results from a study near Dallas-Fort Worth indicated elevated modelled air toxics concentrations (e.g., acrolein, formaldehyde) near fence lines of wells (~600 feet), suggesting potential increased air pollution relatively near wells [12, 55]. To our knowledge, no published studies have evaluated personal exposure to UGD-related air pollution among individuals living varying distances from well sites. Nonetheless, a mechanism through which UGD-related chemical and non-chemical stressors increases risk of adverse health outcomes seems to have greater plausibility at more proximal distances. Thus, comprehensive exposure assessment studies are needed to inform relevant distances within which to best capture UGD-activity as it potentially relates to adverse health outcomes.
We found evidence of an association between maternal residential proximity to UGD-activity and preterm birth and limited evidence of an association with fetal death among a diverse population of women living near the Barnett Shale. Though there may be differences in air pollution emissions during completion and production phases of UGD drilling [10, 56], we defined UGD-activity metrics without regard to drilling phase. Thus, we are presently working toward estimation of phase-specific UGD-activity metrics to inform potential differences in perinatal health risks related to drilling phase. Additionally, we are pursuing methods to improve the assignment of exposure given potential differences in time-at-risk between preterm births and fetal deaths compared with term births. Nonetheless, the lack of detailed exposure assessment data remains a critical gap in understanding potential health risks associated with UGD-activity. Exposure assessment studies would serve to quantify chemical and non-chemical stressors to which residents living near UGD are exposed, validate UGD-activity metrics like the one used in this and previous studies [19–21], and inform the most relevant distances within which to characterize chemical and non-chemical stressors. Future priority should be placed on obtaining such information to better characterize risk and inform epidemiologic studies.
S1 Table. Sensitivity analyses for preterm birth, SGA, and fetal death.
We acknowledge Drilling Info, Inc. and the DrillingInfo System as the source of well drilling data used in the analyses of this manuscript.
- 1. U.S. Environmental Protection Agency. The process of hydraulic fracturing 2013 [cited 2014 June]. https://www.epa.gov/hydraulicfracturing/process-hydraulic-fracturing.
- 2. US House of Representatives Committee on Energy and Commerce, Minorty Staff. Chemical used in hydraulic fracturing. Washington, DC. 2011.
- 3. Fontenot BE, Hunt LR, Hildenbrand ZL, Carlton DD Jr., Oka H, Walton JL, et al. An evaluation of water quality in private drinking water wells near natural gas extraction sites in the Barnett Shale formation. Environ Sci Technol. 2013;47(17):10032–40. pmid:23885945.
- 4. Ingraffea AR, Wells MT, Santoro RL, Shonkoff SB. Assessment and risk analysis of casing and cement impairment in oil and gas wells in Pennsylvania, 2000–2012. Proc Natl Acad Sci U S A. 2014;111(30):10955–60. pmid:24982144; PMC4121786.
- 5. Ferrar KJ, Michanowicz DR, Christen CL, Mulcahy N, Malone SL, Sharma RK. Assessment of effluent contaminants from three facilities discharging Marcellus Shale wastewater to surface waters in Pennsylvania. Environ Sci Technol. 2013;47(7):3472–81. pmid:23458378.
- 6. Burton TG, Rifai HS, Hildenbrand ZL, Carlton DD Jr., Fontenot BE, Schug KA. Elucidating hydraulic fracturing impacts on groundwater quality using a regional geospatial statistical modeling approach. Sci Total Environ. 2016;545–546:114–26. pmid:26745299.
- 7. Vengosh A, Jackson RB, Warner N, Darrah TH, Kondash A. A critical review of the risks to water resources from unconventional shale gas development and hydraulic fracturing in the United States. Environ Sci Technol. 2014;48(15):8334–48. pmid:24606408.
- 8. Elliott EG, Ettinger AS, Leaderer BP, Bracken MB, Deziel NC. A systematic evaluation of chemicals in hydraulic-fracturing fluids and wastewater for reproductive and developmental toxicity. J Expo Sci Environ Epidemiol. 2017;27(1):90–9. pmid:26732376.
- 9. Webb E, Bushkin-Bedient S, Cheng A, Kassotis CD, Balise V, Nagel SC. Developmental and reproductive effects of chemicals associated with unconventional oil and natural gas operations. Rev Environ Health. 2014;29(4):307–18. pmid:25478730.
- 10. Colborn T, Schultz K, Herrick L, Kwiatkowski C. An Exploratory Study of Air Quality Near Natural Gas Operations. Human and Ecological Risk Assessment: An International Journal. 2013;20(1):86–105.
- 11. Coons T, Walker R. Community health risk analysis of oil and gas industry impacts in Garfield County. Saccomanno Research Institute and Mesa State College: Grand Junction, CO. 2008.
- 12. Eastern Research Group. Fort Worth natural gas air quality study: Final report prepared for the City of Fort Worth 2011. http://fortworthtexas.gov/uploadedFiles/Gas_Wells/AirQualityStudy_final.pdf.
- 13. Macey GP, Breech R, Chernaik M, Cox C, Larson D, Thomas D, et al. Air concentrations of volatile compounds near oil and gas production: a community-based exploratory study. Environ Health. 2014;13(1):82. pmid:25355625; PMC4216869.
- 14. Zielinska B, Fujita E, Campbell D. Monitoring of emissions from Barnett Shale natural gas production facilities for population exposure assessment. Desert Research Institute. 2011:71.
- 15. Adgate JL, Goldstein BD, McKenzie LM. Potential public health hazards, exposures and health effects from unconventional natural gas development. Environ Sci Technol. 2014;48(15):8307–20. pmid:24564405.
- 16. Korfmacher KS, Jones WA, Malone SL, Vinci LF. Public health and high volume hydraulic fracturing. New Solut. 2013;23(1):13–31. pmid:23552646.
- 17. McEwen BS. Stress, adaptation, and disease. Allostasis and allostatic load. Ann N Y Acad Sci. 1998;840(1):33–44. pmid:9629234.
- 18. Morello-Frosch R, Shenassa ED. The environmental "riskscape" and social inequality: implications for explaining maternal and child health disparities. Environ Health Perspect. 2006;114(8):1150–3. pmid:16882517; PMC1551987.
- 19. McKenzie LM, Guo R, Witter RZ, Savitz DA, Newman LS, Adgate JL. Birth outcomes and maternal residential proximity to natural gas development in rural Colorado. Environ Health Perspect. 2014;122(4):412–7. pmid:24474681; PMC3984231.
- 20. Stacy SL, Brink LL, Larkin JC, Sadovsky Y, Goldstein BD, Pitt BR, et al. Perinatal outcomes and unconventional natural gas operations in Southwest Pennsylvania. PLoS One. 2015;10(6):e0126425. pmid:26039051; PMC4454655.
- 21. Casey JA, Savitz DA, Rasmussen SG, Ogburn EL, Pollak J, Mercer DG, et al. Unconventional Natural Gas Development and Birth Outcomes in Pennsylvania, USA. Epidemiology. 2016;27(2):163–72. pmid:26426945; PMC4738074.
- 22. Texas Department of State Health Services, Center for Health Statistics. Texas population 2015 (projections) 2015 [cited 2016 September]. https://www.dshs.texas.gov/chs/popdat/ST2015.shtm.
- 23. Basso O, Wilcox A. Mortality risk among preterm babies: immaturity versus underlying pathology. Epidemiology. 2010;21(4):521–7. pmid:20407380; PMC2967434.
- 24. Talge NM, Mudd LM, Sikorskii A, Basso O. United States birth weight reference corrected for implausible gestational age estimates. Pediatrics. 2014;133(5):844–53. pmid:24777216.
- 25. Alexander GR, Himes JH, Kaufman RB, Mor J, Kogan M. A United States national reference for fetal growth. Obstet Gynecol. 1996;87(2):163–8. pmid:8559516.
- 26. Texas Department of State Health Services, Center for Health Statistics. Guidelines for Geocoding. https://www.dshs.texas.gov/chs/gis/.
- 27. Kotelchuck M. An evaluation of the Kessner Adequacy of Prenatal Care Index and a proposed Adequacy of Prenatal Care Utilization Index. Am J Public Health. 1994;84(9):1414–20. pmid:8092364; PMC1615177.
- 28. Wu J, Wilhelm M, Chung J, Ritz B. Comparing exposure assessment methods for traffic-related air pollution in an adverse pregnancy outcome study. Environ Res. 2011;111(5):685–92. pmid:21453913; PMC3114297.
- 29. DrillingInfo. DrillingInfo Online Manual. http://help.drillinginfo.com/robohelp/robohelp/server/DI_Main/projects/Drillinginfo%20Online%20Manual/MyNewProject.htm#Drillinginfo_Online_Help_System_010.htm
- 30. Wood R, Gilbert P, Sharmina M, Anderson K, Footitt A, Glynn S, et al. Shale gas: a provisional assessment of climate change and environmental impacts. Tyndall Center, University of Manchester, Manchester, England. 2011.
- 31. McKenzie LM, Witter RZ, Newman LS, Adgate JL. Human health risk assessment of air emissions from development of unconventional natural gas resources. Sci Total Environ. 2012;424:79–87. pmid:22444058.
- 32. Stieb DM, Chen L, Eshoul M, Judek S. Ambient air pollution, birth weight and preterm birth: a systematic review and meta-analysis. Environ Res. 2012;117:100–11. pmid:22726801.
- 33. Wilcox AJ, Weinberg CR, Basso O. On the pitfalls of adjusting for gestational age at birth. Am J Epidemiol. 2011;174(9):1062–8. pmid:21946386; PMC3243938.
- 34. Olaguer EP. The potential near-source ozone impacts of upstream oil and gas industry emissions. J Air Waste Manag Assoc. 2012;62(8):966–77. pmid:22916444.
- 35. World Health Organization. Air quality guidelines for particulate matter, ozone, nitrogen dioxide and sulfur dioxide: Global update 2005 (Summary of risk assessment). Geneva, Switzerland: World Health Organization; 2006.
- 36. Casey JA, Schwartz BS. The Authors Respond. Epidemiology. 2016;27(6):e37–8. pmid:27465525; PMC5488733.
- 37. Kramer MS, Lydon J, Seguin L, Goulet L, Kahn SR, McNamara H, et al. Stress pathways to spontaneous preterm birth: the role of stressors, psychological distress, and stress hormones. Am J Epidemiol. 2009;169(11):1319–26. pmid:19363098.
- 38. Wallace ME, Harville EW. Allostatic load and birth outcomes among white and black women in New Orleans. Matern Child Health J. 2013;17(6):1025–9. pmid:22833335; PMC3504172.
- 39. Christopherson S, Rightor N. How shale gas extraction affects drilling localities: Lessons for regional and city policy makers. Journal of Town and City Management. 2012;2(4):1–20.
- 40. Jacquet JB. Review of risks to communities from shale energy development. Environ Sci Technol. 2014;48(15):8321–33. pmid:24624971.
- 41. Kassotis CD, Iwanowicz LR, Akob DM, Cozzarelli IM, Mumford AC, Orem WH, et al. Endocrine disrupting activities of surface water associated with a West Virginia oil and gas industry wastewater disposal site. Sci Total Environ. 2016;557–558:901–10. pmid:27073166.
- 42. Kassotis CD, Tillitt DE, Lin CH, McElroy JA, Nagel SC. Endocrine-Disrupting Chemicals and Oil and Natural Gas Operations: Potential Environmental Contamination and Recommendations to Assess Complex Environmental Mixtures. Environ Health Perspect. 2016;124(3):256–64. pmid:26311476; PMC4786988.
- 43. Li X, Huang S, Jiao A, Yang X, Yun J, Wang Y, et al. Association between ambient fine particulate matter and preterm birth or term low birth weight: An updated systematic review and meta-analysis. Environ Pollut. 2017;227:596–605. pmid:28457735.
- 44. Slama R, Darrow L, Parker J, Woodruff TJ, Strickland M, Nieuwenhuijsen M, et al. Meeting report: atmospheric pollution and human reproduction. Environ Health Perspect. 2008;116(6):791–8. pmid:18560536; PMC2430236.
- 45. Menon R, Bonney EA, Condon J, Mesiano S, Taylor RN. Novel concepts on pregnancy clocks and alarms: redundancy and synergy in human parturition. Hum Reprod Update. 2016;22(5):535–60. pmid:27363410; PMC5001499.
- 46. Brou L, Almli LM, Pearce BD, Bhat G, Drobek CO, Fortunato S, et al. Dysregulated biomarkers induce distinct pathways in preterm birth. BJOG. 2012;119(4):458–73. pmid:22324919.
- 47. Faiz AS, Rhoads GG, Demissie K, Kruse L, Lin Y, Rich DQ. Ambient air pollution and the risk of stillbirth. Am J Epidemiol. 2012;176(4):308–16. pmid:22811493; PMC3491970.
- 48. Siddika N, Balogun HA, Amegah AK, Jaakkola JJ. Prenatal ambient air pollution exposure and the risk of stillbirth: systematic review and meta-analysis of the empirical evidence. Occup Environ Med. 2016;73(9):573–81. pmid:27221104.
- 49. Saxton KB, Gemmill A, Catalano RA. Reproductive suppression follows threats to child survival. J Evol Biol. 2017;30(5):889–97. pmid:28267227.
- 50. Catalano R, Karasek D, Gemmill A, Falconi A, Goodman J, Magganas A, et al. Very low birthweight: dysregulated gestation versus evolutionary adaptation. Soc Sci Med. 2014;108:237–42. pmid:24593927; PMC4401996.
- 51. Cox LA Jr., Popken DA, Berman DW. Causal versus spurious spatial exposure-response associations in health risk analysis. Crit Rev Toxicol. 2013;43 Suppl 1:26–38. pmid:23557011.
- 52. 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(6):538–43. pmid:16736057.
- 53. Lupo PJ, Symanski E, Chan W, Mitchell LE, Waller DK, Canfield MA, et al. Differences in exposure assignment between conception and delivery: the impact of maternal mobility. Paediatr Perinat Epidemiol. 2010;24(2):200–8. pmid:20415777.
- 54. Agency for toxic substances and disease registry (ASTDR). Toxicological profile for Benzene. US Department of Health and Human Services: Atlanta, GA. 2007.
- 55. Armendariz A. Emissions from natural gas production in the Barnett Shale area and opportunities for cost-effective improvements. Report for Environmental Defense Fund. 2009.
- 56. New York State Department of Environmental Conservation. Revised draft supplemental generic environmental impact statement (SGEIS) on the oil, gas and solution mining regulatory program: Well permit issuance for horizontal drilling and high-volume hydraulic fracturing to develop the Marcellus shale and other low-permeability gas reservoirs. Albany, NY: NY SDEC, 2011.