Congenital Anomalies Associated with Oil and Gas Development and Resource Extraction: A Population-Based Retrospective Cohort Study in Texas

Abstract

Background:

Oil and gas extraction-related activities produce air and water pollution that contains known and suspected teratogens. To date, health impacts of in utero exposure to these activities is largely unknown.

Objective:

We investigated associations between in utero exposure to oil and gas extraction activity in Texas, one of the highest producers of oil and gas, and congenital anomalies.

Methods:

We created a population-based birth cohort between 1999 and 2009 with full maternal address at delivery and linked to the statewide congenital anomaly surveillance system (n=2,234,138 births, 86,315 cases). We examined extraction-related exposures using tertiles of inverse distance-squared weighting within 5 km for drilling site count, gas production, oil production, and produced water. In adjusted logistic regression models, we calculated odds of any congenital anomaly and 10 specific organ sites using two comparison groups: 1) 5 km of future drilling sites that are not yet operating (a priori main models), and 2) 5–10 km of an active well.

Results:

Using the temporal comparison group, we find increased odds of any congenital anomaly in the highest tertile exposure group for site count (OR: 1.25; 95% CI: 1.21, 1.30), oil production (OR: 1.08; 95% CI: 1.04, 1.12), gas production (1.20; 95% CI: 1.17, 1.23), and produced water (OR: 1.17; 95% CI: 1.14, 1.20). However, associations did not follow a consistent exposure-response pattern across tertiles. Associations are highly attenuated, but still increased, with the spatial comparison group in the highest tertile exposure group. Cardiac and circulatory defects are strongly and consistently associated with all exposure metrics.

Significance:

Increased odds of congenital anomalies, particularly cardiac and circulatory defects, were associated with exposures related to oil and gas extraction in this large population-based study. Future research is needed to confirm findings, examine specific exposure pathways, and identify potential avenues to reduce exposures among local populations.

Background

In the United States, 3% of infants are diagnosed with at least one congenital anomaly (1), and congenital anomalies account for 20% of infant mortality cases (2). Some specific congenital anomalies, such as Down syndrome, have well-established hereditary links, accounting for about 20% of congenital anomaly cases (3). However, etiologies of most congenital anomalies remains unknown (1). Environmental factors (e.g., radiation) are emerging as causative agents (3), but few environmental teratogens have been studied in human populations (4).

Oil and gas extraction processes produce known and suspected teratogens that may pose a risk to human health and developing fetuses (5–7). An estimated 17.6 million Americans live within 1.6 km of an active oil or gas site (8), and this population is likely to expand in the future (9). Air and water pollution from this industry has been consistently associated with teratogens, including diesel particulate matter, benzene, toluene, naphthalene, and formaldehyde (10–14). Recent work has shown that criteria pollutants can be elevated as far as 5 km from the drilling site (15), and there is some evidence of increases in the concentrations of radioactive particles out to 20 km (14,16). Toxicological literature identifies several chemicals used for hydraulic fracturing, a subset of oil and gas extraction exposures, as endocrine disruptors at levels that could harm human health (17–19). These pollutants can cause chronic inflammation and oxidative stress in the mother, which may lead to an increased risk of adverse birth outcomes, including congenital anomalies (20–22). This industry also produces substantial noise pollution from the ongoing extraction activities (23), and extraction activities are associated with increased airborne particle radioactivity (14). Although little literature has examined each exposure pathway’s independent influence on human health (24,25), we hypothesize that the totality of these pathways could be associated with adverse infant health outcomes.

An emerging body of evidence examines the influence of natural gas drilling activity on pregnancy and infant health outcomes (26–32), which is a small subset of the overall oil and gas industry. The few studies that have investigated associations between drilling-related exposures and congenital anomalies show mixed results in settings across the United States (33–37). To date, little literature on drilling-related activity and congenital anomalies include oil extraction in their exposure metrics, an important co-exposure in many regions that may pose similar risks to developing fetuses (30,38,39).

We created a population-based retrospective birth cohort (n=2,234,138) from 1999 to 2009 in Texas, the state with the highest on-shore oil and gas production in the country (40), using geocoded birth certificate information with linkage to the state birth defects registry. Approximately 4.5 million Texans live within 1.6 km of an active oil or gas drilling site (8), providing an ideal setting to leverage the largest single state population exposed to this industry in the United States, and these birth years align with a rapid period of expansion in the state’s oil and gas industry. We expand upon existing literature by examining subtypes of congenital anomalies that have not been studied to date, as well as investigating variation in exposure type (e.g., oil production, gas production, water produced) in separate and combined models, which provides evidence for what parts of the drilling industry may be hazardous to developing fetuses.

Methods

Study Population

We acquired statewide vital statistics data from the Texas Department of State Health Services for all births from January 1, 1999 through December 31, 2009, where we have access to maternal residences at delivery geocoded to the address (n=3,713,654). Vital statistics data is linked to the Birth Defects Registry, an active surveillance system, for infants with a congenital abnormality diagnosis before their first birthday. We restrict our analysis to singleton births with data in bounds for maternal age (11–64 years old), gestational age (25–43 weeks), and birth weight (500–5,000 grams) (n=130,166 excluded) and residential address outside of 5 km of a current or future oil and gas extraction site or outside of 10 km from a current oil and gas extraction site (n= 1,349,350 excluded). In total, these criteria yield 2,234,138 infants for this analysis.

Oil and Gas Extraction Exposure Assessment

Oil and gas extraction data is from Enverus DrillingInfo (41), a proprietary database of drilling sites in the United States. We linked maternal addresses to drilling sites that were active between January 1^st, 1985 and June 30^th, 2019 and located within 10 km of a maternal residence at delivery. Resource extraction sites can remain active for over thirty years (42,43); therefore, sites that started prior to our study period may contribute to ongoing exposures. By including active drilling sites in the period after our births occurred, we limit our analysis to places that are viable for the oil and gas industry.

For each unique drilling site, we determine the distance between every active drilling site (i.e., single well), where individual drilling sites are classified using the American Petroleum Institute (API) identifiers that correspond at a one-to-one ratio with drilling site boreholes. We also calculated a separate metric of intensity in a 5 km buffer for each maternal residence and calculate the average monthly oil production, average monthly gas production, and average monthly produced water that is attributed to a given drilling site over its lifetime. Each metric represents a separate exposure, though the pathways are highly correlated, by which drilling intensity may influence congenital anomaly risk. For instance, drilling sites with more oil production generally flare more often (44,45), while drilling sites with more produced water may be engaging in hydraulic fracturing (46). We classified a drilling site as active for a specific residence if the infant’s delivery was after the spud date or first production date and there was not a reported final date of production in the database. We then calculate inverse distance-squared weighted (IDW) metrics for the number of drilling sites within 5 km of a maternal residence:

$${\mathit{\text{Count IDW}}}_{5km}={\sum }_{i=1}^n\frac{1}{d_i^2},$$

where i is a unique drilling site within the buffer distance, d is the distance in kilometers from the maternal residence to the unique drilling site, and n is the total count of drilling sites within the specified buffer of the maternal residence. Additional IDW_5km metrics are weighted by the average monthly oil production, average monthly gas production, and average monthly produced water within 5 km of a maternal residence:

$${\mathit{\text{Production IDW}}}_{5km}={\sum }_{i=1}^n\frac{p}{d_i^2},$$

where i is a unique drilling site within the buffer distance, p is the average monthly production or produced water for a unique drilling site, d is the distance in kilometers from the maternal residence to the unique drilling site, and n is the total count of drilling sites within the specified buffer of the maternal residence. Within our IDW_5km metrics (count and production), we create tertiles of exposure levels.

We also incorporate the nearest primary or secondary road (e.g. highways, interstates) using the 2000 census road shapefile (47) to account for traffic-related air pollution exposures (48).

Congenital Anomaly Outcome Assessment

Our outcome assessment comes from two distinct state-based administrative sources: Texas Vital Statistics database and Birth Defects Registry. Using the Vital Statistics data, we first identify all births with geocoded maternal residences to the address level, where we capture detailed information on maternal sociodemographic and clinical characteristics. We then link the Vital Statistics data to the Birth Defects Registry, where we identify congenital anomaly cases in our birth population. This registry contains all cases of physician-diagnosed chromosomal and structural congenital anomalies for infants born in Texas whose maternal residence is also in the state. Infants without a congenital anomaly diagnosis constitute our comparison group. This linkage process was previously published (49).

We group each of our congenital anomalies by the primary diagnostic site (Supplemental Table 1)(49). Anomalies that do not fit into a category are included in “any monitored congenital anomaly”. Given ongoing work on infants diagnosed with multiple congenital anomalies (50), we implement a category for infants with diagnoses at more than one primary diagnostic site.

Covariate Assessment

Sociodemographic and clinical characteristics of the mother-infant dyad were extracted from the vital statistics database. We also integrated data on neighborhood characteristics via the decennial United States Census, where births before 2005 were linked to the 2000 information and births in 2005 and after were linked to the 2010 information. In addition, we calculated the distance between the geocoded maternal residence and the nearest highway using the 2010 census road shapefile as a proxy for traffic-related air pollution.

Statistical Analysis

We use logistic regressions with robust standard errors to test our hypothesis that oil and gas extraction activity is associated with increased odds of congenital anomalies. For each exposure, we examine each group of congenital anomalies in separate models. First, we assess the influence of the nearest active drilling site on the odds of congenital anomalies within 10 km of an active drilling site, where we create 1km bins from 0–10 km. Reported coefficients from this model set are relative to maternal residences in the 9–10 km group. Second, we investigate the associations between IDW_5km metrics and odds of a congenital anomaly. Reported coefficients from this model set are relative to maternal residences that are not exposed to active resource extraction.

Oil and gas development is a complex exposure for epidemiologic analysis as the industry brings economic and environmental changes at the same time (30,51,52), thus finding a reasonable counterfactual comparison group is often challenging (15,26,31,53,54). Therefore, we constructed two distinct counterfactual groups that come with their own unique strengths and weaknesses, which allows us to assess the robustness of our risk estimates. First, we assess maternal residences with future (but no active) drilling sites within 5km as a comparison group. This comparison group restricts our evaluation to areas with proven drilling potential, which reduces the potential for residual confounding from unmeasured population or community characteristics. In sensitivity analyses, we implement a spatial comparison group that contains maternal residences within 5–10 km of the nearest active drilling site. This comparison group restricts our models to areas that are adjacent to active drilling activity, where the pregnancies may be exposed to economic changes from the industry but likely not exposed to the highest levels of pollution. By implementing both comparison groups, we can alleviate some degree of concerns regarding unmeasured confounding in our risk estimates.

For each of the models discussed above, fully adjusted models contain the following variables: infant sex (male, female), gestational age in weeks (continuous), birth weight in grams (continuous), maternal age (continuous), maternal race and ethnicity (white non-Hispanic, black non-Hispanic, Hispanic, other/missing),(55) maternal education (less than high school, high school, some college, bachelors, more than bachelors, missing), maternal smoking (yes, no, missing), maternal alcohol usage (yes, no, missing), prenatal care initiated (yes, no, missing), census tract unemployment (continuous), census tract percent White population (continuous), and census tract median household income in United States dollars (continuous), and distance to nearest highways in meters (continuous). All models also contain separate indicators for birth year and maternal residential county to reduce potential for unmeasured spatial and temporal confounding (e.g., changes in regional population). Missing categorical covariates were coded as a separate category, while records with missing continuous covariates were removed.

We conducted statistical analyses in Stata 16.1 (56).

Results

Descriptive Statistics

We provide the spatial distribution of oil, gas, and water production (Figure 1) and details on the mean number of extraction sites for each tertile of exposure (Supplemental Table 2). When we examined the sociodemographic characteristics by IDW_5km tertiles of site count, we find that many characteristics such as gestational length, infant sex, maternal age, maternal educational attainment, and prenatal care are similar among our tertiles compared to the spatial and temporal comparison groups (Table 1). However, the compositions of maternal race and ethnicity, smoking during pregnancy, distance to the nearest highway, and census tract characteristics are not similar among the tertiles nor the comparison groups. For instance, the percentage of women who smoked during pregnancy is 7.8, 8.0, and 9.2 among our tertiles, respectively, but the percentage of smokers in our spatial and temporal comparison groups is only 6.7 and 7.0, respectively.

Figure 1:

Spatial distribution of oil and gas drilling in Texas, 1985–2019. Data are from Enervus DrillingInfo

Table 1:

Cohort demographics for inverse distance-squared weight metrics for active extraction site counts within 5km of a maternal residence

Characteristic	Temporal Comparison	Spatial Comparison	Tertile 1	Tertile 2	Tertile 3
Total births (count)	254,557	837,078	380,834	380,831	380,838
Any monitored congenital anomaly (count)	9,516	31,296	14,395	14,768	16,340
Birth weight (g)	3,317	3,284	3,285	3,278	3,280
Gestational age (weeks)	38.7	38.5	38.5	38.5	38.5
Female sex (%)	48.8	48.9	49.0	48.8	48.9
Maternal age (mean)	26.2	26.3	26.2	26.1	26.1
Maternal race and ethnicity
White non-Hispanic (%)	39.7	32.6	35.8	36.0	41.4
Black non-Hispanic (%)	10.8	12.0	13.1	13.0	10.6
Hispanic (%)	45.4	51.0	46.8	47.3	44.7
Other/missing (%)	4.1	4.4	4.4	3.7	3.8
Maternal educational attainment
Did not complete high school (%)	33.3	32.9	30.2	30.3	29.7
Completed High school (%)	30.5	28.9	30.2	29.7	29.8
Some college (%)	17.4	19.5	21.1	22.2	23.1
Bachelor’s degree (%)	11.1	11.1	11.5	11.4	11.8
Post-graduate (%)	6.5	6.7	6.2	5.7	5.1
No prenatal care initiated (%)	1.3	1.1	1.1	1.3	1.3
Smoking during pregnancy (%)	7.0	6.7	7.8	8.0	9.2
Distance to highway (m)	1,309	1,478	1,497	1500	1,600
Census tract unemployment (mean %)	4.5	6.1	6.2	6.3	6.6
Census tract proportion white race (mean, %)	67.9	65.5	65.8	67.6	70.3
Census tract median household income (mean, $)	43,641	43,980	45,509	45,946	47,275

Nearest Active Drilling Site Analysis

In adjusted models, we found elevated odd ratios (ORs) of any monitored congenital anomaly for the coefficients 0–1 km and 1–2 km relative to the 9–10 km coefficient (Figure 2). However, this association is imprecise and associations are largely null or protective for the further distance groups (e.g., 7–8 km).

Figure 2:

Associations between nearest active spud to a maternal residence and odds of congenital anomalies. Model coefficients are relative to the maternal addresses whose nearest active drilling site is 9–10km from their residences. All models contain include birth year (indicator), county of maternal residence at delivery (categorical), infant sex (male, female), gestational age (continuous), birth weight (continuous), maternal age (continuous), maternal race and ethnicity (white non-Hispanic, black non-Hispanic, Hispanic, other), maternal education (less than high school, high school, some college, bachelors, more than bachelors), smoking (yes, no, missing), prenatal care initiated (yes, no, missing), distance to nearest highways in meters (continuous), census tract unemployment rate (continuous), census tract median household income (continuous), and census tract percent white population (continuous).

Inverse Distance-squared Weighted Metric Analysis

Our subsequent models examine site-specific congenital anomalies for IDW_5km tertiles of site count, where we leverage separate models for a temporal and spatial comparison group (Table 2). Using the temporal comparison group, we found that IDW_5km tertiles of site count are associated with increased ORs of 1.19 (95% CI: 1.15, 1.22) for tertile 1, 1.24 (95% CI: 1.19, 1.28) for tertile 2, and 1.25 (95% CI: 1.21, 1.30) for tertile 3 for any monitored defect. For cardiac and circulatory anomalies, we observed ORs of 1.14 (95% CI: 1.08, 1.21), 1.20 (95% CI: 1.12, 1.27), and 1.20 (95% CI: 1.13, 1.28) across tertiles. These associations are attenuated, but still elevated, using a spatial comparison group. For instance, using the spatial comparison group, we found that IDW_5km tertiles of site count are associated with increased ORs of 1.01 (95% CI: 0.98, 1.03) for tertile 1, 1.04 (95% CI: 1.02, 1.06) for tertile 2, and 1.05 (95% CI: 1.03, 1.08) for tertile 3 for any monitored defect. When we used a spatial comparison group for cardiac and circulatory anomalies, we observed ORs of 1.02 (95% CI: 0.98, 1.05), 1.06 (95% CI: 1.02, 1.10), and 1.06 (95% CI: 1.02, 1.10) across tertiles. While we find evidence of associations between IDW_5km site count and >1 congenital anomaly, eye and ear, gastrointestinal, genitourinary, and chromosomal anomalies using the temporal comparison group, these associations are not robust to the spatial comparison group.

Table 2:

Associations between inverse distance-squared weight metrics for active drilling site counts within 5km of a maternal residence and odds of congenital anomalies

Comparison Group	Congenital Anomaly	Total Births	Reference Group		Tertile 1		Tertile 2		Tertile 3
			Cases	Odds Ratio	Cases	Odds Ratio	Cases	Odds Ratio	Cases	Odds Ratio
Temporal (Future Drilling Sites within 5km)	All Monitored Defects	1,397,060	9,516	1	14,395	1.19 (1.15, 1.22)	14,768	1.24 (1.19, 1.28)	16,340	1.25 (1.21, 1.30)
	>1 Sites	1,346,666	884	1	1,195	1.08 (0.97, 1.21)	1,252	1.17 (1.04, 1.31)	1,294	1.19 (1.10, 1.33)
	Cardiac and Circulatory	1,359,504	2,654	1	4,728	1.14 (1.08, 1.21)	4,911	1.20 (1.12, 1.27)	5,170	1.20 (1.13, 1.28)
	Central Nervous System	1,343,754	423	1	444	0.97 (0.81, 1.15)	440	1.02 (0.85, 1.23)	406	1.00 (0.82, 1.21)
	Eye and Ear	1,342,663	117	1	168	1.42 (1.04, 1.93)	158	1.37 (0.99, 1.91)	179	1.53 (1.10, 2.12)
	Gastrointestinal	1,346,346	736	1	1,156	1.12 (0.99, 1.26)	1,224	1.18 (1.04, 1.33)	1,189	1.10 (0.97, 1.24)
	Genitourinary	1,348,914	1,434	1	1,832	1.05 (0.96, 1.15)	1,834	1.12 (1.02, 1.23)	1,773	1.12 (1.01, 1.23)
	Musculoskeletal	1,344,974	562	1	724	0.98 (0.85, 1.13)	758	1.00 (0.86, 1.16)	889	1.11 (0.96, 1.29)
	Oral Clefts	1,344,109	394	1	556	1.10 (0.93, 1.31)	560	1.07 (0.90, 1.28)	558	1.02 (0.85, 1.22)
	Respiratory	1,342,577	126	1	151	0.99 (0.72, 1.36)	128	0.88 (0.63, 1.24)	131	0.96 (0.69, 1.35)
	Chromosomal	1,343,977	348	1	501	1.18 (0.99, 1.40)	516	1.25 (1.04, 1.49)	571	1.32 (1.10, 1.59)
Comparison Group	Congenital Anomaly	Total Births	Reference Group		Tertile 1		Tertile 2		Tertile 3
			Cases	Odds Ratio	Cases	Odds Ratio	Cases	Odds Ratio	Cases	Odds Ratio
Spatial (Active drilling sites at 5–10 km)	All Monitored Defects	1,979,581	31,296	1	14,395	1.01 (0.98, 1.03)	14,768	1.04 (1.02, 1.06)	16,340	1.05 (1.03, 1.08)
	>1 Sites	1,909,268	2,745	1	1,195	0.97 (0.91, 1.04)	1,252	1.04 (0.97, 1.12)	1,294	1.06 (0.98, 1.15)
	Cardiac and Circulatory	1,927,385	9,794	1	4,728	1.02 (0.98, 1.05)	4,911	1.06 (1.02, 1.10)	5,170	1.06 (1.02, 1.10)
	Central Nervous System	1,905,159	1,087	1	444	0.94 (0.84, 1.06)	440	0.99 (0.88, 1.12)	406	0.94 (0.83, 1.08)
	Eye and Ear	1,903,620	333	1	168	1.09 (0.90, 1.33)	158	1.04 (0.84, 1.29)	179	1.15 (0.92, 1.43)
	Gastrointestinal	1,908,976	2,625	1	1,156	0.97 (0.90, 1.04)	1,224	1.02 (0.95, 1.10)	1,189	0.96 (0.88, 1.04)
	Genitourinary	1,912,405	4,184	1	1,832	0.96 (0.91, 1.02)	1,834	1.02 (0.96, 1.08)	1,773	1.02 (0.96, 1.09)
	Musculoskeletal	1,906,943	1,790	1	724	0.91 (0.83, 1.00)	758	0.93 (0.85, 1.03)	889	1.02 (0.92, 1.12)
	Oral Clefts	1,905,768	1,312	1	556	0.95 (0.86, 1.06)	560	0.93 (0.83, 1.03)	558	0.88 (0.78, 0.99)
	Respiratory	1,903,496	304	1	151	1.13 (0.92, 1.38)	128	1.01 (0.81, 1.27)	131	1.09 (0.86, 1.38)
	Chromosomal	1,905,572	1,202	1	501	0.96 (0.86, 1.07)	516	1.01 (0.90, 1.13)	571	1.06 (0.94, 1.19)

Models include birth year (indicator), county of maternal residence at delivery (categorical), infant sex (male, female), gestational age (continuous), birth weight (continuous), maternal age (continuous), maternal race and ethnicity (white non-Hispanic, black non-Hispanic, Hispanic, other), maternal education (less than high school, high school, some college, bachelors, more than bachelors), smoking (yes, no, missing), prenatal care initiated (yes, no, missing), distance to nearest highways in meters (continuous), census tract unemployment rate (continuous), census tract median household income (continuous), and census tract percent white population (continuous). Reported coefficients from this model set are relative to maternal residences that are not exposed to active resource extraction.

We then examined site-specific congenital anomalies for IDW_5km tertiles of oil and gas production and produced water with a temporal comparison group (Table 3). We found that IDW_5km tertiles of oil production are associated with increased odds of any monitored defect and cardiac and circulatory, respectively, while we also found that IDW_5km tertiles of gas production are associated increased odds of any monitored defect and cardiac and circulatory defects. For IDW_5km tertiles of produced water, we found associations with increased odds of any monitored defect using the temporal comparison group. For cardiac and circulatory anomalies, we observed increased odds across tertiles. Results are highly attenuated by the spatial comparison group for most congenital anomaly sites, and the confidence intervals are often imprecise.

Table 3:

Associations between inverse distance-squared weighted metrics for production-related measures within 5km of a maternal residence and odds of congenital anomalies

Congenital Anomaly	Total Births	IDW5km Oil Production			IDW5km Gas Production			IDW5km Produced Water
		Tertile 1	Tertile 2	Tertile 3	Tertile 1	Tertile 2	Tertile 3	Tertile 1	Tertile 2	Tertile 3
Temporal Comparison
All Monitored Defects	1,397,060	1.10 (1.07, 1.13)	1.08 (1.05, 1.11)	1.08 (1.04, 1.12)	1.14 (1.11, 1.18)	1.13 (1.10, 1.17)	1.20 (1.17, 1.23)	1.14 (1.11, 1.17)	1.14 (1.10, 1.17)	1.17 (1.14, 1.20)
>1 Site	1,346,666	1.11 (1.01, 1.21)	1.08 (0.97, 1.20)	1.10 (0.98, 1.23)	1.05 (0.95, 1.16)	1.07 (0.97, 1.18)	1.16 (1.05, 1.28)	1.06 (0.97, 1.17)	1.07 (0.97, 1.17)	1.15 (1.04, 1.27)
Cardiac and Circulatory	1,359,504	1.10 (1.05, 1.15)	1.05 (1.00, 1.11)	1.06 (1.00, 1.13)	1.07 (1.02, 1.13)	1.08 (1.03, 1.14)	1.15 (1.09, 1.21)	1.11 (1.06, 1.16)	1.10 (1.05, 1.16)	1.10 (1.05, 1.16)
Central Nervous System	1,343,754	1.11 (0.95, 1.30)	1.20 (1.00, 1.44)	1.18 (0.97, 1.43)	1.20 (1.02, 1.41)	1.16 (0.98, 1.36)	1.24 (1.05, 1.47)	1.15 (0.99, 1.34)	1.08 (0.92, 1.27)	1.19 (1.01, 1.40)
Eye and Ear	1,342,663	1.03 (0.80, 1.32)	0.88 (0.66, 1.17)	1.02 (0.75, 1.38)	1.07 (0.81, 1.40)	1.01 (0.77, 1.33)	1.23 (0.94, 1.61)	1.19 (0.92, 1.55)	1.24 (0.96, 1.61)	1.22 (0.93, 1.60)
Gastrointestinal	1,346,346	1.04 (0.94, 1.14)	1.11 (1.00, 1.23)	1.10 (0.98, 1.24)	1.11 (1.00, 1.22)	1.10 (0.99, 1.21)	1.08 (0.97, 1.19)	1.01 (0.92, 1.12)	1.10 (1.00, 1.21)	1.07 (0.97, 1.19)
Genitourinary	1,348,914	1.03 (0.96, 1.11)	1.10 (1.01, 1.19)	1.12 (1.02, 1.22)	1.06 (0.98, 1.15)	1.09 (1.01, 1.18)	1.17 (1.08, 1.27)	1.06 (0.98, 1.15)	1.13 (1.05, 1.22)	1.12 (1.03, 1.21)
Musculoskeletal	1,344,974	1.11 (1.00, 1.25)	1.01 (0.88, 1.15)	1.08 (0.94, 1.24)	1.03 (0.91, 1.16)	1.01 (0.89, 1.14)	1.16 (1.03, 1.31)	1.12 (1.00, 1.26)	1.05 (0.93, 1.18)	1.08 (0.96, 1.22)
Oral Clefts	1,344,109	1.04 (0.91, 1.19)	0.86 (0.74, 1.00)	0.87 (0.74, 1.03)	0.92 (0.79, 1.06)	1.01 (0.87, 1.16)	0.97 (0.84, 1.13)	1.00 (0.87, 1.14)	0.97 (0.84, 1.12)	0.98 (0.85, 1.13)
Respiratory	1,342,577	0.98 (0.74, 1.32)	1.05 (0.78, 1.42)	0.96 (0.69, 1.32)	1.10 (0.83, 1.46)	1.13 (0.84, 1.52)	1.06 (0.78, 1.42)	0.93 (0.71, 1.23)	1.00 (0.75, 1.32)	1.04 (0.78, 1.37)
Chromosomal	1,343,977	1.17 (1.02, 1.35)	1.13 (0.96, 1.34)	1.05 (0.87, 1.26)	1.12 (0.96, 1.31)	1.07 (0.91, 1.25)	1.11 (0.95, 1.30)	1.11 (0.96, 1.29)	1.05 (0.91, 1.22)	1.19 (1.02, 1.38)
Spatial Comparison
All Monitored Defects	1,979,581	1.03 (1.01, 1.06)	1.02 (0.99, 1.04)	1.02 (0.99, 1.05)	1.04 (1.02, 1.07)	1.03 (1.00, 1.05)	1.08 (1.05, 1.10)	1.04 (1.02, 1.07)	1.04 (1.01, 1.06)	1.06 (1.04, 1.09)
>1 Site	1,909,268	1.05 (0.97, 1.13)	1.00 (0.92, 1.08)	1.01 (0.92, 1.11)	0.98 (0.91, 1.06)	1.00 (0.93, 1.09)	1.10 (1.01, 1.19)	1.00 (0.92, 1.07)	1.01 (0.93, 1.09)	1.09 (1.00, 1.18)
Cardiac and Circulatory	1,927,385	1.06 (1.02, 1.10)	1.02 (0.98, 1.06)	1.03 (0.98, 1.08)	1.02 (0.98, 1.07)	1.03 (0.99, 1.07)	1.09 (1.04, 1.13)	1.05 (1.01, 1.10)	1.05 (1.00, 1.09)	1.04 (1.00, 1.09)
Central Nervous System	1,905,159	0.98 (0.87, 1.12)	1.08 (0.94, 1.23)	1.05 (0.91, 1.22)	1.05 (0.92, 1.19)	1.00 (0.88, 1.15)	1.06 (0.92, 1.22)	1.04 (0.92, 1.18)	0.97 (0.85, 1.11)	1.04 (0.90, 1.19)
Eye and Ear	1,903,620	0.97 (0.78, 1.20)	0.91 (0.72, 1.14)	1.06 (0.83, 1.36)	1.00 (0.80, 1.24)	0.93 (0.74, 1.17)	1.10 (0.88, 1.38)	1.06 (0.86, 1.31)	1.10 (0.88, 1.36)	1.06 (0.84, 1.33)
Gastrointestinal	1,908,976	0.98 (0.90, 1.06)	1.01 (0.93, 1.09)	0.99 (0.90, 1.08)	1.00 (0.93, 1.08)	1.00 (0.92, 1.09)	0.98 (0.90, 1.07)	0.95 (0.88, 1.02)	1.02 (0.94, 1.10)	1.00 (0.92, 1.09)
Genitourinary	1,912,405	0.99 (0.93, 1.05)	1.00 (0.94, 1.07)	1.02 (0.95, 1.09)	0.98 (0.92, 1.04)	1.02 (0.96, 1.09)	1.10 (1.03, 1.17)	0.98 (0.92, 1.04)	1.05 (0.99, 1.12)	1.05 (0.98, 1.12)
Musculoskeletal	1,906,943	1.04 (0.95, 1.14)	0.93 (0.84, 1.03)	1.00 (0.90, 1.12)	0.96 (0.87, 1.05)	0.93 (0.84, 1.03)	1.05 (0.94, 1.16)	1.04 (0.94, 1.14)	0.96 (0.87, 1.06)	0.97 (0.88, 1.08)
Oral Clefts	1,905,768	0.99 (0.88, 1.10)	0.83 (0.74, 0.94)	0.85 (0.74, 0.98)	0.87 (0.78, 0.98)	0.91 (0.81, 1.03)	0.87 (0.77, 0.99)	0.92 (0.82, 1.03)	0.89 (0.79, 1.00)	0.89 (0.79, 1.01)
Respiratory	1,903,496	1.03 (0.81, 1.31)	1.11 (0.88, 1.40)	1.03 (0.79, 1.34)	1.13 (0.90, 1.42)	1.15 (0.90, 1.46)	1.15 (0.90, 1.48)	1.00 (0.79, 1.26)	1.10 (0.86, 1.39)	1.15 (0.89, 1.48)
Chromosomal	1,905,572	1.05 (0.94, 1.18)	1.03 (0.91, 1.16)	0.95 (0.82, 1.09)	1.03 (0.91, 1.16)	0.97 (0.86, 1.10)	1.00 (0.88, 1.13)	1.01 (0.90, 1.13)	0.96 (0.85, 1.08)	1.06 (0.94, 1.21)

Models include birth year (indicator), county of maternal residence at delivery (categorical), infant sex (male, female), gestational age (continuous), birth weight (continuous), maternal age (continuous), maternal race and ethnicity (white non-Hispanic, black non-Hispanic, Hispanic, other), maternal education (less than high school, high school, some college, bachelors, more than bachelors), smoking (yes, no, missing), prenatal care initiated (yes, no, missing), distance to nearest highways in meters (continuous), census tract unemployment rate (continuous), census tract median household income (continuous), and census tract percent white population (continuous). Reported coefficients from this model set are relative to maternal residences that are not exposed to active resource extraction.

Discussion

This study examines associations between multiple exposure metrics of oil and gas extraction and odds of specific congenital anomalies in the largest study (n=2,234,138) on this relationship to date. We identified that the odds of any monitored congenital anomaly were higher within 2 km of the nearest active drilling site to the maternal residence at delivery. We found consistent associations between well site count, gas production, and produced water metrics and increased odds of any monitored congenital anomaly, and a large and robust association with cardiac and circulatory anomalies specifically. Our associations with congenital anomalies are attenuated for oil production metrics. However, we did not find a consistent exposure-response pattern in our results, potentially due to heterogeneity in extraction activities at each location. We hypothesize that exposure pathways that may be responsible for this association include a variety of potential parts of the extraction process: fluids associated with various steps in the process, such as hydraulic fracturing (e.g., benzene, toluene) (11,46), pollutants from diesel truck traffic (e.g., diesel particulate matter) (14,15,57), flaring, the process where excess natural gas is combusted (e.g., polycyclic aromatic hydrocarbons) (45), and naturally occurring radioactive material that is pumped from the ground (e.g., ambient particle radioactivity in indoor and outdoor settings) (14,16).

This analysis contributes to the small body of literature that examines the influence of oil and gas extraction on congenital anomaly risks (33–37). In Pennsylvania, Ma et al. leveraged an interrupted time series study design using an ecological study sample from birth certificate data, where they find no associations between zip code-level drilling exposure and congenital anomalies. Since many congenital anomalies are not identified at time of delivery (58,59), their outcome assessment is likely an underestimate of the true prevalence. In Colorado, McKenzie et al. use geocoded birth certificate data linked to the state congenital anomaly registry with similar IDW site count metrics and a spatial comparison group. Their study (n=124,842) observes associations in their highest tertiles of exposure with cardiac and circulatory defects (Odds Ratio: 1.3; 95% CI: 1.2, 1.5, based on 1,823 cases) and neural tube defects (Odds Ratio: 2.0; 95% CI: 1.0, 3.9, based on 59 cases). We confirm their findings for cardiac and circulatory defects. However, we do not corroborate their findings for neural tube defects. The study authors later published a re-analysis of the Colorado data regarding cardiac and circulatory defects with improved exposure assessment, where results are similar to their previous analysis and what we found here. In Oklahoma, Janitz et al. applied a similar study design to McKenzie et al. through geocoded birth certificate data linked to the state congenital anomaly registry (n=476,600), where they focus on critical congenital heart defects, neural tube defects, and oral clefts. Using IDW site count metrics and a spatial comparison group, their prevalence rate ratios show little evidence of an association between drilling activity and increased risk of these congenital anomalies. Most directly similar to our present study, Tang et al. uses a case control study design with statistical spatial smoothing to examine unconventional natural gas development and several congenital anomalies in Texas, including gastroschisis (n=2,179 cases), congenital heart defects (n=42,445 cases), neural tube defects (n=2,175 cases), and orofacial clefts (n=6,174 cases). Using a spatial comparison group with controls for temporal trends, the analysis shows an association with some subtypes of cardiac and circulatory defects and neural tube defects, but the results are not consistent. Overall, our results are largely in alignment with the findings in previous literature, thus adding to the body of evidence that in utero exposure to oil and gas extraction may influence the risk of congenital anomalies.

Our present analysis expands on the existing literature in key aspects. We use a pair of high-quality outcome data sources, a vital statistics database with an active surveillance registry, to create a population-based cohort of mother-infant dyads with full maternal address information. Given our large population with substantial exposure variation, we examined more subtypes of congenital anomalies with a finer spatial resolution. We also added new exposure production metrics to examine potential dissimilarities in exposure pathways among oil, natural gas, and water production from these drilling sites, and we showed that natural gas and water production (but not necessarily oil production) are associated with increased odds of congenital anomalies after accounting for drilling co-exposures. We noted that these exposure measures are all highly correlated, which precludes the separation of each pathway in our analyses. In addition, we demonstrated how the selection of a comparison group can change what conclusions should be drawn from our statistical models. Based on the variation in some results, future studies on the population health impacts of the drilling industry should strongly consider presenting multiple comparison groups.

We note some important limitations to consider when interpreting our results. First, we acknowledge that birth certificate data only contains maternal address at delivery. With a single timepoint for an address, we assume that the mother lived at this address throughout the pregnancy, including the first trimester which is likely the period of relevant exposure for congenital anomaly formation (60). Second, our congenital anomaly data is limited to liveborn children with diagnoses rendered before their first birthday. Therefore our analysis inherently excludes congenital anomalies that yielded spontaneous abortions, elective terminations, stillborn infants, and cases with diagnoses later in life, thus creating potential for under ascertainment of congenital anomaly cases. Third, our lack of evidence for associations between drilling exposures and specific congenital anomalies does not necessarily indicate that this association does not exist, especially for rarer anomalies. While our population-based study has more power to detect these associations than previous work, the rarity of these outcomes creates challenges for developing precise models in the full data set, let alone examining important subgroups (e.g., geological formations, lower socioeconomic status, secular time trends). Fourth, our exposure data source precludes straightforward assessment of other parts of the development process (e.g., pad preparation, hydraulic fracturing), which may also contribute towards etiologically relevant environmental exposures (52,57). In particular, produced water volumes are highest at the beginning and end of a site’s lifecycle (46), thus introducing exposure misclassification into our analysis. Fifth, the alignment of our exposure metrics with air pollution or water contamination data is unclear. Given the well-documented lack of air or water monitoring data near these drilling sites (39,45,61), policy-relevant proxy metrics such as distance to the nearest site, or even IDW calculations, are a reasonable method to provide preliminary evidence of how the drilling industry may affect population health. Simple distance metrics, as we presented here, are particularly helpful for developing health-protective zoning regulations (24,62,63). With these limitations in mind, our analysis provides a novel contribution to the literature on how drilling-related exposures may influence congenital anomalies, a substantial public health issue.

This analysis adds to the growing body of literature on the influence of the oil and gas industry on congenital anomalies. Within this analysis, we observed that drilling-related exposures within 5km of a maternal residence at delivery are associated with increased odds of a monitored congenital anomaly, as well as cardiac and circulatory defects. Congenital anomalies present a considerable burden on patients, their families, and the healthcare system, thus the prospect that in utero exposure to oil and gas drilling increases the risk of this outcome warrants further investigation.

Impact Statement:

About 5% of the U.S. population (~17.6 million people) resides within 1.6 km of an active oil or gas extraction site, yet the influence of this industry on population health is not fully understood. In this analysis, we examined associations between oil and gas extraction-related exposures and congenital anomalies by organ site using birth certificate and congenital anomaly surveillance data in Texas (1999–2009). Increased odds of congenital anomalies, particularly cardiac and circulatory defects, were associated with exposures related to oil and gas extraction in this large population-based study. Future research is needed to confirm these findings.

Data availability

The health data used in this study can be obtained for similar research purposes on request from the Texas Department of State Health Services, and the oil and gas development data can be obtained for similar research purposes from Enverus DrillingInfo. All other exposure data is referenced in the main text.