Residential proximity to oil and gas development and risk of gestational diabetes mellitus.

Pregnancy Study Online (PRESTO) is an internet-based prospective cohort study of pregnancy planners residing in the United States and Canada. 37 Enrollment began in 2013 and is ongoing as of 2025. Recruitment is primarily through social media advertisements. Eligible participants are aged 21–45 years old, assigned female at birth, reside in the U.S. or Canada, and trying to conceive without the use of fertility treatment. After completing an eligibility screener and providing informed consent, eligible participants are invited to complete a baseline questionnaire on sociodemographic characteristics, lifestyle factors, and medical histories and follow-up questionnaires every 8 weeks for up to 12 months to update information on time-varying covariates and pregnancy status. Participants who conceived during the study period completed an early pregnancy questionnaire (median: 5 weeks’ gestation [interquartile range, IQR: 3, 7 weeks]), a late pregnancy questionnaire (median: 29 weeks’ gestation [IQR: 28, 30 weeks], and a postpartum questionnaire approximately 6 months after their reported due date (median: 6 months after delivery [IQR: 6, 8 months]). This analysis is restricted to participants whose pregnancies progressed beyond 28 weeks’ gestation, corresponding to the timing of screening for gestational diabetes per standard clinical guidelines (i.e., 24–28 weeks). 29 The institutional review board of the Boston University Medical Campus approved this study. Exposure to oil and gas development was derived from the Oil and Gas Infrastructure Mapping (OGIM) database. 38 , 39 Briefly, this database contains harmonized information on oil and gas sites from state and provincial sources; we supplemented these data with additional public information. 40 Important data include site locations, key dates (e.g., spud date, completion date, first production date), production type (e.g., oil, gas), and drilling type (e.g., horizontal, directional, vertical). Once a site is drilled, we consider that site active until 30 years after the first date of known activity, aligning with industry estimates of the maximum exposure duration. 41 Due to inconsistent data quality across locations (e.g., drilling type), we were unable to disaggregate some components in our analysis. Participants provided full address information during participation, which we geocoded to the street level. We assigned individual-level exposure metrics at estimated date of conception within a 20 km buffer around each residence, which aligns with recent literature on the environmental and health impacts of oil and gas development. 19 Exposure measures were highly correlated throughout the pregnancy period (e.g., Spearman correlation of 0.97 for exposure at conception vs late pregnancy [~32 weeks gestation]). We selected conception as a temporal anchor for our exposure measures, ensuring that exposures preceded the development of gestational diabetes. Our measures included 1) distance to the nearest active oil or gas development site (km), and 2) an inverse distance-squared weighted (IDW) sum of active oil or gas development sites, representing the density of local industry activity that quantifies exposure intensity. This methodology, consistent with recent literature, 19 , 42 captures the spatial and temporal components of oil and gas development, enabling us to examine how exposure to the oil and gas industry varies over space and time around each participant’s residence. 14 , 41 , 42 We ascertained gestational diabetes via self-report on the late pregnancy questionnaire and the postpartum questionnaire. To improve outcome ascertainment, we also linked cohort data with birth registries for select states where gestational diabetes diagnosis is recorded (e.g. California, Florida, Massachusetts, New York [excluding New York City], Ohio, Pennsylvania, Texas). 43 We prioritized diagnoses from birth registry data; otherwise, we relied on diagnoses via self-report on the late pregnancy questionnaire and, if necessary, the postpartum questionnaire. For our analysis, gestational diabetes was categorized as a binary variable (i.e., yes/no). The gold standard for gestational diabetes diagnosis involves an oral glucose tolerance test performed around 24–28 weeks of gestation, 31 and these diagnostic results are typically reflected in birth records. To quantify validity, we compared the self-reported gestational diabetes from the questionnaire to the birth record, a proxy for the diagnostic test. Among 493 participants where we had self-reported and birth registry data, sensitivity was 0.79 and specificity was 0.97. Among participants who enrolled between June 2013 and July 2024, 8,129 participants reported pregnancies that progressed beyond 28 weeks of gestation. We excluded participants who were still actively participating (n=1,208, 13.1%), reported a pre-existing diabetes diagnosis at baseline (n=57, 0.6%), or had a residential address that could not be geocoded to the street level (n=579, 6.2%). Our final analytic sample included 6,285 participants. We evaluated baseline characteristics of our analytic sample according to exposure status. Approximately 8.4% of participants had missing data on gestational diabetes, and 5.8% had missing data on gestational age at delivery. Missing covariate data (<5%) were multiply imputed via fully conditional specification methods, where we created 20 data sets and statistically combined the standardized parameter estimates and standard errors. 44 We used log-binomial regression models to estimate risk ratios (RRs) and 95% confidence intervals (CIs) for distance and intensity analyses. In addition, we used restricted cubic splines to assess non-linearity in this association. For the distance analyses, we categorized participants into distance-based groups: 0 to <5 km, 5 to <10 km, 10 to <15 km, 15 to <20 km, and ≥20 km. For the intensity analyses, we grouped participants who lived within 20 km of at least one active site into tertiles of low, medium, and high site density. Reference groups for all analyses contained the participants who resided ≥20 km from active oil and gas development. We selected covariates based on prior literature and a directed acyclic graph ( Supplemental Figure 1 ). In adjusted models, we included the following covariates: age at baseline (<25, 25–29, 30–34, 35–39, ≥40 years), enrollment year (2013–2024), pre-pregnancy smoking (yes/no), and geographic region (U.S.: Northeastern, Southern, Midwestern, Western; Canada). Because oil and gas development may influence socioeconomic conditions (e.g., household income, educational attainment) and siting decisions are shaped by structural factors (which may be reflected in race and ethnicity), 20 , 45 these variables may be along the causal pathway to gestational diabetes. Therefore, we did not adjust for these factors in our primary models ( Supplemental Figure 1 ). To examine effect measure modification by adiposity, we stratified models by pre-pregnancy body mass index (BMI, <25.0, 25.0–29.9, ≥30.0 kg/m 2 ). We also examined potential effect measure modification by parity in stratified models (nulliparous, parous). Due to small sample sizes in some categories, stratified models for BMI and parity are only adjusted for age, enrollment year (two-year categories), and geographic region. We conducted sensitivity analyses to evaluate the consistency of our findings. First, to examine the influence of our selected comparison group, we excluded participants who lived >50 km from oil and gas development. Second, to minimize confounding by well-established risk factors, we ran a successive series of models where we excluded those with polycystic ovary syndrome (PCOS), those who smoked prior to pregnancy, those who drank alcohol during preconception, multiple gestations, and those who had a history of anxiety or depression before pregnancy. Third, we accounted for potential exposure misclassification by excluding those known to have resided less than one year at their residential address at the time of conception. Fourth, we restricted the analysis to participants who had seen a primary care physician in the previous year to account for healthcare utilization. Fifth, we implemented three separate models that are further adjusted for education (high school or less, some college or college degree, graduate school), annual household income (<$50,000; $50,000–99,999; $100,000–149,999; ≥$150,000) and physical activity categorized by weekly metabolic equivalent task (MET) hours (<10, 10–19, 20–39, ≥40 MET-hours/week). Spatial exposure measures were derived using R (R Foundation for Statistical Computing, Vienna, Austria Version 4.2.2), while geocoding and statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, NC, USA).

Among 6,285 participants from the United States (88.3%) and Canada (11.7%), 8.6% were diagnosed with gestational diabetes over follow-up ( Table 1 ). Participants in the sample were primarily non-Hispanic White (85.7%), attained a college degree or higher (82.3%), and had a primary care physician visit in the past year (87.1%). Characteristics of participants who resided within 5 km of active oil and gas development were generally similar to the full cohort, though the prevalence was slightly higher for a BMI ≥30 kg/m 2 (26.6% vs. 23.7%) and pre-pregnancy smoking (5.4% vs. 3.6%) compared with those who resided beyond 20 km. We also note that the prevalence of gestational diabetes was similar by source of information (i.e., birth registry vs. self-report on the questionnaire) ( Supplemental Table 1 ). For the distance measure in adjusted models, participants who resided within 5 km of active oil and gas development had a similar risk of gestational diabetes (RR: 1.07, 95% CI: 0.81, 1.40) compared with those who lived ≥20 km away from the nearest active oil or gas development site Table). Results were similar in farther distance groups ( Table 2 ), and patterns were similar in restricted cubic splines ( Figure 1 ). For the intensity exposure measure in adjusted models, participants who resided in the top tertile of exposure (i.e., the most exposure to active oil and gas development) also had a similar risk of gestational diabetes (RR: 1.10, 95% CI: 0.83, 1.45) compared with those who lived ≥ 20 km away from any oil or gas development ( Table 2 ). Restricted cubic splines likewise showed no clear patterns between higher exposure to oil and gas development and risk of gestational diabetes ( Figure 1 ). In models stratified by pre-pregnancy BMI ( Figure 2 ), the association between residence within 5 km of oil and gas development and gestational diabetes was relatively similar among individuals with BMI ≥30 kg/m 2 (RR: 1.24, 95% CI: 0.87, 1.76), 25–29.9 kg/m 2 (RR: 1.08, 95% CI: 0.61, 1.92), or <25.0 kg/m 2 (RR: 0.80, 95% CI: 0.45, 1.44). Similarly, the intensity exposure measure showed associations that were generally consistent across BMI categories, with little evidence of an overall association with gestational diabetes. In models stratified by parity ( Table 3 ), the association between residence within 5 km of oil and gas development and risk of gestational diabetes was similar in direction and magnitude for parous participants (RR: 1.10, 95% CI: 0.73, 1.65) and nulliparous participants (RR: 1.03, 95% CI: 0.71, 1.50) ( Table 3 ). Likewise, there was little evidence of effect modification by parity for analyses of highest category of intensity and risk of gestational diabetes. We implemented a wide range of sensitivity analyses. However, we note that precision was limited for selected subgroup analyses, reflected in the width of the confidence intervals, and that these results should be interpreted with caution. When we restricted analyses to participants who resided within 50 km of active oil or gas development, results were generally similar to the main models ( Supplemental Table 2 ). Associations restricted to participants who a) did not have a PCOS diagnosis, b) were non-smokers in the preconception period, c) did not drink any alcohol pre-pregnancy, d) had a singleton pregnancy, e) no history of anxiety or depression, f) did not move residences during pregnancy, and g) had seen a primary care physician in the previous year were generally similar in direction and magnitude to our main findings ( Supplemental Tables 3 – 9 ). In models further adjusted for education, income, and physical activity, results were also broadly consistent with our primary model specification ( Supplemental Tables 10 – 12 ).

Resource extraction operations for oil and gas have rapidly expanded across North America and are projected to continue steadily until at least 2050. 1 Approximately 20 million Americans live within 1 mile of an active extraction site, and this industry is also near many Canadian communities, 2 , 3 potentially placing large populations in the path of industrial hazards. 2 , 4 Oil and gas development (i.e., the industrial process of extracting oil and gas from the earth) produces a complex mixture of pollutants that can deposit into the air and water of surrounding communities. 5 – 8 Endocrine-disrupting chemicals emitted from oil and gas development (e.g., benzene, toluene, phthalates) are of particular concern due to their ability to interfere with hormonal and physiological functioning. 4 , 9 , 10 These chemicals may be released at varying stages of oil and gas extraction, including borehole drilling, hydraulic fracturing, and well production. 6 Air pollution near oil and gas development may also include benzene and particulate matter from increased vehicle traffic due to site construction, maintenance, and export of the product. 11 Exposure can be both short-term (during high-emission periods, such as drilling or re-fracturing) and long-term (due to ongoing emissions during the lifespan of a well). 12 Communities that host resource extraction may experience increased psychosocial stress, 13 – 16 which in turn can adversely affect reproductive health. 17 , 18 For example, one study reported associations between oil and gas development and increased symptoms of stress and worse mental health in the preconception period. 14 Previous research has linked residential proximity to oil and gas development with adverse reproductive health outcomes across a wide range of study settings in the U.S. and Canada 19 , 20 (e.g., hypertensive disorders of pregnancy, 21 reduced birth weight, 22 , 23 preterm delivery, 24 – 26 various birth defects 27 , 28 ). However, many important perinatal health conditions with plausible links to oil and gas development remain unstudied. Gestational diabetes mellitus (hereafter, gestational diabetes: glucose intolerance that is first recognized during pregnancy) affects approximately 7% of pregnancies in North America. 29 This condition is associated with immediate and long-term health risks for both pregnant individuals (e.g., pre-eclampsia, type 2 diabetes) and their infants (e.g., macrosomia, respiratory distress). 30 , 31 A wide range of environmental exposures, including endocrine-disrupting chemicals, has been implicated in disruptions to glucose metabolism and insulin resistance that characterize gestational diabetes. 32 Chronic and acute stress, such as rapid community changes from oil and gas development, 13 – 16 may also alter these metabolic processes. 33 , 34 However, an estimated 20–40% of cases are unexplained by traditional risk factors (e.g., polycystic ovary syndrome, obesity, advanced maternal age), 35 indicating that research on novel risk factors is necessary for future preventive efforts. Although no studies have directly examined oil and gas development and gestational diabetes, benzene (a prominent chemical emitted by the industry) has been linked to altered glycemic control and insulin resistance in non-pregnant populations. 36 The totality of this evidence suggests a plausible disease pathway between exposures related to oil and gas development and increased risk of gestational diabetes. Using integrated geospatial data in a large North American preconception cohort of pregnancy planners, this study examines the association between residential proximity to oil and gas development and risk of gestational diabetes.

In this study of pregnancy planners in the United States and Canada, we found no consistent evidence that living near oil and gas development was associated with a higher risk of gestational diabetes. These results were consistent across a range of subgroup analyses and model specifications. Our findings add to the growing body of literature on residential proximity to oil and gas development and perinatal outcomes; however, most previous research finds some evidence of adverse health outcomes among those residing near oil and gas development. 19 – 28 For example, a quasi-experimental study of birth certificate records showed a 5% increased odds of gestational hypertension within 1 km of active oil and gas development in Texas. 21 Similarly, most studies on residential proximity to oil and gas development and adverse birth outcomes find at least one outcome with a consistent, elevated risk (22 out of 24 published papers). 19 , 20 We hypothesize that capturing the relevant exposure pathways between oil and gas development and risk of gestational diabetes may be more challenging than the other previously-studied outcomes. For instance, a primary route of exposure that could have implications for risk of gestational hypertension and adverse birth outcomes is air pollution from criteria pollutants, 46 which persists relatively far away from the extraction location. 47 In contrast, many of the most important exposure pathways for gestational diabetes, such as endocrine-disrupting chemicals via air or water, 32 are highly localized exposures. 42 The processes that emit these chemicals, such as hydraulic fracturing or chemical spills, are also more intermittent over the lifecycle of an oil or gas development site, 41 , 42 meaning that our broad proximity-based exposure measures may not fully account for the exposure pathways most relevant for this outcome. When interpreting the results of our study, there are several limitations to consider. First, our metric of exposure is a surrogate for the wide range of exposures that may be associated with oil and gas development, 42 largely due to limited available data at the multi-country geographic scale. 41 We also omit individual-level differences in time-activity patterns, such as participants’ time at work, that may change exposure assessment. When we restricted analyses to participants who had lived at their current residence for at least a year to account for some of this variation, we found similarly null effect estimates. Second, our outcome measure is self-reported diagnosis of gestational diabetes as opposed to a clinical test or physician diagnosis abstracted from medical records. 29 , 31 Although our data suggests that self-reported measures are highly valid, some degree of outcome misclassification is possible, likely in a nondifferential manner. Finally, our study population is unique: all participants were planning a pregnancy at enrollment and were recruited via the internet. 37 Although internet-based recruitment should not bias etiologic associations, 48 results from this cohort may not generalize to the broader population of reproductive-aged individuals due to higher interaction with medical providers higher educational attainment, and lower representation of individuals living in a rural setting. 49 , 50 Despite these limitations, our study has several important strengths, including its prospective design, large sample size, detailed covariate data, and extensive geographic coverage of oil and gas development across the United States and Canada, capturing a broad range of development activities and contexts.

This study provides insights into a novel exposure – oil and gas development - that could plausibly influence gestational diabetes, a highly prevalent pregnancy complication with long-term impacts on maternal-infant health. Our analysis found no appreciable association between residential proximity to oil and gas development and risk of gestational diabetes. This emerging research area warrants cautious interpretation and further investigation into the role of environmental determinants of pregnancy outcomes and metabolic health.

Supplementary data Supplementary data are available at IJE online