The human health effects of unconventional oil and gas development (UOGD): A scoping review of epidemiologic studies

Objective Unconventional oil and gas development (UOGD, sometimes termed “fracking” or “hydraulic fracturing”) is an industrial process to extract methane gas and/or oil deposits. Many chemicals used in UOGD have known adverse human health effects. Canada is a major producer of UOGD-derived gas with wells frequently located in and around rural and Indigenous communities. Our objective was to conduct a scoping review to identify the extent of research evidence assessing UOGD exposure–related health impacts, with an additional focus on Canadian studies. Methods We included English- or French-language peer-reviewed epidemiologic studies (January 2000–December 2022) which measured exposure to UOGD chemicals directly or by proxy, and where health outcomes were plausibly caused by UOGD-related chemical exposure. Results synthesis was descriptive with results ordered by outcome and hierarchy of methodological approach. Synthesis We identified 52 studies from nine jurisdictions. Only two were set in Canada. A majority (n = 27) used retrospective cohort and case–control designs. Almost half (n = 24) focused on birth outcomes, with a majority (n = 22) reporting one or more significant adverse associations of UOGD exposure with: low birthweight; small for gestational age; preterm birth; and one or more birth defects. Other studies identified adverse impacts including asthma (n = 7), respiratory (n = 13), cardiovascular (n = 6), childhood acute lymphocytic leukemia (n = 2), and all-cause mortality (n = 4). Conclusion There is a growing body of research, across different jurisdictions, reporting associations of UOGD with adverse health outcomes. Despite the rapid growth of UOGD, which is often located in remote, rural, and Indigenous communities, Canadian research on its effects on human health is remarkably sparse. There is a pressing need for additional evidence. Supplementary information The online version contains supplementary material available at 10.17269/s41997-024-00860-2.


Introduction
Unconventional oil and gas development (UOGD, sometimes referred to as "fracking" or "hydraulic fracturing") is an industrial process to extract methane gas and/or oil deposits primarily from shale or "tight" rock (Environmental Protection Agency, 2023).The technique first involves a pad preparation phase with clearing land and transporting material to the site.Next, a shaft is drilled vertically down 3-4 km into the ground-past the fresh and saline water aquifers, and horizontally for a further thousands of metres ("the spud" or drilling phase) (Rasmussen et al., 2016).This is followed by the hydraulic fracturing or "fracking" phase.In this phase, large amounts of fluid-most commonly water mixed with sand and chemical additives-are pumped along the well shaft under high pressure creating micro-fractures of the shale or tight rock, thereby freeing trapped oil and gas and starting the final production phase (US EPA, 2013;Water Resources Mission Area, 2019).During the production phase, the internal pressure of the rock formation causes fluid to return to the surface through the wellbore.This fluid is known as "flowback" or "produced water" and may contain the injected chemicals plus naturally occurring materials such as brines, metals, radionuclides, and hydrocarbons (Brown, 2014;Srebotnjak, 2018).The flowback is typically stored on site in tanks or open pits or surface impoundments before treatment, disposal, or recycling (Brown, 2014).
Multiple jurisdictions have imposed UOGD bans or moratoria out of concerns for environmental and health impacts (AIDA, 2019).Within Canada, there has been substantial interprovincial policy variation, with some provinces declaring moratoria or bans (Minkow, 2017), and others investing heavily in expansion (Schmunk, 2018).Canada is a major methane gas producer globally, with Alberta and British Columbia as the largest producers (Statista, 2014).Notably, most UOGD in Canada occurs in remote and rural communities (Natural Resources Canada, 2016) where Indigenous people are more likely to reside (Government of Canada, 2022).Indigenous communities living in rural and remote locations also rely on the land for food and traditional medicines, and land and water are embedded into peoples' livelihood and identity (Poirier & Neufeld, 2023).Disparities in environmental exposures among Indigenous, Black, and other racialized communities have been documented in numerous settings (Hoover et al., 2012;Johnston et al., 2020;Kaufman & Hajat, 2021;Waldron, 2020), including in the context of UOGD, and there is growing recognition of environmental racism and environmental injustice as determinants of health (Waldron, 2020).
Systematic (Bamber et al., 2019) and scoping reviews (Deziel et al., 2020;Wright & Muma, 2018) have been published on the human health effects of UOGD over the past 5 years.However, as a relatively new area of research, more studies are being published annually.Our primary objective was to conduct a scoping review to update the available evidence on the health effects of UOGD-related chemical exposures and to identify knowledge gaps (Munn et al., 2018).Considering the rapid growth of UOGD in Canada (Atkinson et al., 2016), we additionally sought to identify Canadian studies.We limited our approach to a scoping review (Tricco et al., 2018) without meta-analysis or systematic assessment of study bias based on the substantial heterogeneity of exposures, outcomes, and methodological approaches.

Data sources and searches
We defined UOGD as directional vertical and horizontal drilling for long distances combined with the injection of fluids containing chemicals and proppants (for example, silica) with enough pressure to fracture shale formations thereby releasing oil or gas or both.We excluded coal seam gas studies because the extraction technique is often too different to make meaningful comparisons between exposures.
A biomedical librarian (MDW) conducted comprehensive searches in MEDLINE, and Embase (OVID) for all published studies in English or French from January 2000 through December 2022, with the most recent search completed on 10 January 2023.Our search concept included the various terms deployed for UOGD AND (population health OR pregnancy OR physical health OR Indigenous) and was combined with a search for UOGD-related toxicology studies (to be reported elsewhere).The search strategy is detailed in Online Resource 1.We also hand-searched the Physicians, Scientists, and Engineers for Healthy Energy citation database of oil and gas research health folder, to identify other eligible studies (PSE, 2023).

Study selection
We included epidemiology studies which measured exposure to UOGD chemicals directly or by proxy.Using a similar approach to Bamber et al. (2019), we included studies where health outcomes were plausibly caused by UOGD-related chemical exposure.We excluded studies on UOGD and traffic accidents, occupational injury, anxiety, or depression where the hypothesized causal pathway for these outcomes was less likely related to chemical exposure and more likely related to indirect pathways such as income, industrial safety practices, and community disruption.We further excluded studies with no comparison group or reference population and studies that assessed the association of UOGD on climate change, seismicity, air/water/soil quality, animal health, community disruption, or socioeconomic impacts (Fig. 1).
Title and abstract screening and full text review were carried out by two pairs of reviewers (MMcG, MF, AA, TT) to determine eligibility for study inclusion.Disagreements at either of these stages were resolved through discussion with a fifth reviewer (LAR) until consensus was reached.Due to the wide heterogeneity of outcomes, exposures, and methodological approaches, we did not apply a formal bias tool to evaluate the quality of studies.

Data extraction
Data for selected articles were independently extracted in duplicate by two pairs of reviewers (MMcG, MF, AA, TT) into an electronic data capture form designed for this purpose (REDCap) (Harris et al., 2009).Extracted data included the following: first author, publication year, journal, funding source, study objective, study design, geographic location, study dates, details of exposed and reference populations (sampling method, sample sizes, % response rate), exposure measurements (types, data sources, dates), covariables (types, data sources, dates), health outcome measurements (types, data sources, dates), statistical analysis methods, and effect estimates (including 95% confidence intervals (CI) or p-values) as reported by authors in their respective publications.When an element of a study was unclear, the corresponding author was contacted for clarification.

Data synthesis
Given the variation in exposure and outcome definitions, data synthesis was descriptive.We followed PRISMA guidelines for scoping reviews (Tricco et al., 2018).First, we produced a high-level summary of all reviewed studies (first author, study setting, study population and time period, population source and sampling method, exposure measures, outcome measures, and study findings) grouped by hierarchy of epidemiologic study design (retrospective cohort, case-control, cross-sectional, and ecologic) and alphabetical order of first author.We identified a study as having a significant effect (harmful or protective) for a given health outcome when there was a statistically significant association reported between one or more exposure levels and one or more health outcomes (i.e., a reported p-value < 0.05 and/or effect estimate where the 95% CI did not cross one for relative risk or zero for absolute risk).
We further grouped studies by health outcomes, and calculated summary statistics when two or more studies reported on the same health outcome.The direction of effect estimates for each outcome is summarized descriptively in Table 2 using arrows, including the proportion of studies reporting harmful and protective effects for a given outcome, and the number of distinct geographic settings in which these outcomes were studied.Supplementary materials provide detailed descriptions of each study's exposure and outcome measurement, covariates, data sources, analytic approach, main study results, and conclusions as reported by authors in the abstract (Online Resources 2-3).For the sake of brevity, outcomes less extensively studied were also described in more detail in supplementary material (Online Resource 4).
Many studies used a cumulative exposure measure of UOGD activity based on number of active wells weighted by proximity to residence (inverse distance weighted [IDW]) within defined geographic radius/buffer zones (Table 1; Online Resource 2).More recent studies further refined the IDW measure by development phase, well depth, and production volumes (Elliott et al., 2018;Koehler et al., 2018;McAlexander et al., 2020;Rasmussen et al., 2016;Tang et al., 2021;Tustin et al., 2017;Walker Whitworth et al., 2018).Several studies incorporated upwind/downwind or uphill/downhill directionality for air and water exposure measurement (Hill & Ma, 2022;Johnston et al., 2021;Li et al., 2022).Some researchers included flaring events (Cushing et al., 2020;Koehler et al., 2018;Willis et al., 2020), compressor engine activity (Koehler et al., 2018), and conventional oil and gas extraction as separate exposure covariates (Apergis et al., 2021;Elser et al., 2021;Schuele et al., 2022;Willis et al., 2020), or examined annual air (Apergis et al., 2021;Blinn et al., 2020;Brown et al., 2019;Hill, 2018;Hill & Ma, 2022;Hu et al., 2022;Li et al., 2022;McKenzie et al., 2017McKenzie et al., , 2019a;;Tran et al., 2021;Willis et al., 2018) and water contamination (Hill, 2018 Outcome reported as having an effect if effect estimate is statistically significant (i.e., a reported p-value < 0.05 and/or effect estimate where the 95% CI did not cross one for relative risk or zero for absolute risk); and reported as being suggestive of an effect when there is evidence of a trend but not statistically significant (i.e., a p-value was between 0.05 and 0.1, and/or the lower CI was between 0.9 and 1.0, or authors reported a large effect magnitude) *** "Any adverse" includes study outcomes included in the columns and others not included and falling under the same category ↑ = statistically significant increased risk; ↓ = statistically significant decreased risk; ↗/↘ = suggestive of an effect but non-significant (i.e., a p-value was between 0.05 and 0.1, and/or the lower CI was between 0.9 and 1.0, or authors reported a large effect magnitude); ↔ = nonsignificant Ma, 2022) alongside UOGD exposure metrics and/or model covariates.Two studies directly measured drinking water and/or air pollutants at participants' residences (Elliott et al., 2018;Steinzor et al., 2013) (Table 1; Online Resource 2).Outcome and covariate data were drawn from secondary administrative/clinical health records, registries, laboratory markers, biometrics, and self-report (Table 1; Online Resource 3).Health outcomes for epidemiologic studies included birth-related (fetal growth, preterm birth, birth deformities, etc.), respiratory (predominantly asthma), and cardiovascular outcomes, cancer, self-reported symptoms, all-cause/cause-specific hospitalizations, and mortality (Tables 1 and 2; Online Resource 3).
A cohort study in Oklahoma (> 500,000 newborns) reported a progressive decrease in birthweight the closer an individual lived to UOGD wells (Apergis et al., 2019).Similarly, a large study in Pennsylvania (> 1,000,000 newborns) reported a 38 g decrease in birthweight associated with residence within 1 km of a well during pregnancy (Currie et al., 2017).These results were consistent with another study in Pennsylvania that reported an inverse association of birthweight with UOGD-related contamination of public drinking water sources (Hill & Ma, 2022), after adjustment for individual-level socioeconomic status (SES) variables, smoking status, month and year of birth, and child sex (Hill & Ma, 2022).A cohort study in British Columbia detected lower birthweights with increased exposure to oil and gas wells; the associations were only significant in the second or third quartiles (and not the fourth quartile) compared to the first (Caron-Beaudoin et al., 2021).The authors discuss this pattern as a possible non-linear non-monotonic dose response related to endocrine disruption.This study did not include SES as a covariate which could have led to some bias in effect estimates due to confounding.Casey et al. (2016) used electronic clinical health records to study fetal growth outcomes in Pennsylvania and adjusted for a wide range of potential confounders (including primary care provider status, smoking status during pregnancy, pre-pregnancy BMI, parity, antibiotic orders during pregnancy, and receipt of medical assistance).The authors detected a decrease in birthweight in the highest quartile of exposure compared to the first (Q4 versus Q1, β − 31 g, 95% CI − 57, − 5) that lost statistical significance after adjusting for year of birth (Q4 versus Q1, β − 20 g, 95% CI − 15, 16).Whitworth et al. (2017) examined these outcomes in Texas also adjusting for a wide range of clinical confounders (including pre-pregnancy BMI, adequacy of prenatal care utilization, and previous poor pregnancy outcome) and reported no effect between exposed and unexposed in adjusted models.
There were methodological differences in the treatment of gestational age.For example, while some only included birthweight data for term births (≥ 37 weeks gestation) in their models (Caron-Beaudoin et al., 2021;Casey et al., 2016;Currie et al., 2017;Stacy et al., 2015;Tran et al., 2020;Willis et al., 2021), others adjusted for gestational age as a covariate (Cushing et al., 2020;Erickson et al., 2022;Hill, 2018;McKenzie et al., 2014;Schuele et al., 2022;Willis et al., 2021), while others did not account for gestational age in their models (Hill & Ma, 2022;Whitworth et al., 2017).Despite these differences in exposure metrics, control groups, and statistical models, a majority of studies identified a decrease in birthweight with UOGD exposure.
Proximity to UOGD activity was also associated with low birthweight (< 500 g) in a majority of studies examining this outcome (Table 2).In contrast, one study set in Colorado reported an increase in average birthweight and decreased odds of low birthweight associated with higher UOGD exposure (McKenzie et al., 2014).The authors of the study pointed to the lack of adjustment of SES, prenatal care, and pregnancy complications that may explain these discordant results.
Small for gestational age (SGA) (birthweight < 10th percentile for gestational age) was another frequently examined outcome.In a cohort study in rural Alberta, Canada, living within 10 km of one or more wells was associated with an increased risk ratio (RR) of SGA (RR 1.12, 95% CI 1.03-1.23)(Cairncross et al., 2022).To better account for well density, the study also examined the risk of living within 10 km of > 100 wells compared to 1-24 wells and reported a higher risk ratio of SGA (RR 1.65, 95% CI 1.10-2.48).However, certain individual-level factors such as smoking and SES were not included as covariates because these variables were unavailable in the provincial administrative dataset.Five other studies (Hill, 2018;Schuele et al., 2022;Stacy et al., 2015;Tran et al., 2020Tran et al., , 2021) ) reported an association between UOGD proximity and SGA (Table 2).Cohort studies in British Columbia (with > 5000 women) (Caron-Beaudoin et al., 2021) and Pennsylvania (with > 10,000 women) (Casey et al., 2016) and two studies in Texas with > 23,000 (Cushing et al., 2020) and > 150,000 (Whitworth et al., 2017) women reported no association between UOGD exposure and SGA.
In contrast, McKenzie et al. reported a decreased odds of preterm birth, in line with a protective effect on low birthweight (McKenzie et al., 2014).This study did not adjust for SES (described earlier).A large ecologic study across 28 US states also reported a decreased risk of preterm birth with exposure to UOGD (Schuele et al., 2022).Exposure to wells was not significantly associated with preterm birth in rural populations in California (OR 1.17, 95% CI 0.64-2.12),and was associated with a decrease in preterm birth in urban populations (OR 0.65, 95% CI 0.48-0.87)(Tran et al., 2021).Notably, while UOGD does take place in California, most wells are a result of conventional oil and gas production.
One study looking at preterm birth by UOGD phase reported that the drilling phase-specific IDW yielded stronger associations with preterm birth compared to the production-specific phase (Walker Whitworth et al., 2018).The authors further found that the greatest risk for extreme preterm birth (< 28 weeks) was associated with residence in the top third of UOGD activity (OR 2.00 (1.23-3.24)and 1.53 (1.03-2.27))for drilling and production, respectively (Walker Whitworth et al., 2018).A study in Texas (Cushing et al., 2020) reported the association between UOGD flaring and preterm birth was only significant with high (OR 1.50, 95% CI 1.23-1.83)but not low flaring (OR 0.82, 95% CI 0.61-1.04).A strength of this study was its subgroup analysis by race; a disproportionate exposure to flaring was identified in Hispanic populations.

Other maternal/infant health outcomes
One cohort study (McKenzie et al., 2014) and one case-control study (Tang et al., 2021) found increased odds of neural tube defects associated with UOGD exposure.Another case-control study reported increased odds of congenital heart defects (McKenzie et al., 2019a).A large retrospective study reported increased risk of all congenital abnormalities (Willis et al., 2023), while another reported no effect of UOGD on birth defects (Ma et al., 2016) (Tables 1 and  2).One study reported an association of gestational hypertension and eclampsia with residential proximity to UOGD activity using a difference in differences approach allowing for counterfactual comparisons (Willis et al., 2022).Further details on these studies are provided in Online Resource 4.

Asthma and other respiratory outcomes
Several studies reported significant associations between UOGD activity and asthma (Table 2).A case-control study in Pennsylvania examined the association of phase-specific exposure activity metrics and asthma exacerbations by severity level (mild, moderate, and severe) (Rasmussen et al., 2016).The authors reported higher risk of all types of asthma exacerbation irrespective of exposure phase.The magnitude of OR for mild asthma exacerbations and the production phase was 4.4 (95% CI 3.8-5.2) and most of the models described linear dose-response patterns across increasing exposure quartiles.Another case-control study in Pennsylvania (using the same dataset but incorporating the four phases of well development and UOGD-related compressor engines) reported an adjusted OR (95% CI) of 3.69 (3.16-4.30)for asthma mild exacerbations and residential location in the highest UOGD activity quartile compared to the lowest, after adjustment by several individual-level covariates, compressor station activity (air pollution surrogate), weather estimates, and community-based socioeconomic measures (Koehler et al., 2018).
A cross-sectional study in California compared lung capacity measures and self-reported wheezing between residents living within 1000 m of an active oil well versus an idle well, and living near (< 200 m) an active well versus further away (> 200 m) (Johnston et al., 2021).The odds of wheezing increased among those living near an active versus idle well (OR 2.58, 95% CI 1.19-5.59),but was not significantly increased for those living near compared to those living further away.However, there was a consistent decrease in forced expiratory volume during first second (FEV1) and forced vital capacity (FVC) measures regardless of the reference group.The study was also unique in that it examined the impact of living upwind or downwind of a drilling well, showing a decrease in FEV1 and FVC among those living downwind and less than 200 m compared to those living upwind and more than 200 m from wells.

Cardiovascular and cerebrovascular outcomes
A case-control study with 12,330 participants in Pennsylvania (McAlexander et al., 2020) reported a significant association with heart failure hospitalizations.Additionally, a cross-sectional study (McKenzie et al., 2019b) and ecological studies reported associations with cardiovascular (Apergis et al., 2021;Denham et al., 2021;Jemielita et al., 2015;Peng et al., 2018) and cerebrovascular outcomes (Hu et al., 2022).Further details on these studies are provided in Online Resource 4.

Mortality
Evidence from cohort and ecological studies suggests increased mortality rates among populations living proximal to various UOGD exposure measures (Apergis et al., 2021;Denham et al., 2021;Hu et al., 2022;Li et al., 2022).Further details on these studies are provided in Online Resource 4.

Discussion
This review includes 52 studies of which over half were not included in previous reviews (Bamber et al., 2019;Deziel et al., 2020).Almost one half examined the association of living in proximity to UOGD and birth outcomes, with many using cohort and case-control study designs in a variety of settings.Other studies examined respiratory outcomes, cardiovascular outcomes, cancer, self-reported symptoms, hospitalizations, and mortality.Studies are set in a growing number of diverse geographic regions in the United States and two regions in Canada.Overall, the studies suggest evidence of detrimental health effects related to living in proximity to UOGD.However, some knowledge gaps remain.
To the best of our knowledge, this review is the first published on this topic with a stated goal to focus on Canadian studies.Despite a number of published biomonitoring studies (Caron-Beaudoin et al., 2018, 2019, 2022;Claustre et al., 2023), we only identified two Canadian epidemiologic studies that met our study inclusion criteria.This "evidence of absence" is concerning given the country's almost 20-year history of UOGD, the industry's continued expansion, and the wells' frequent location on the territories of Indigenous communities already disproportionately impacted by health and economic disparities due to the ongoing effects of colonization (FNHA, 2018).
Our review builds on prior published reviews on this topic (Bamber et al., 2019;Deziel et al., 2020).Like Bamber et al., we restricted our search to outcomes more likely related to chemical causal pathways.Those authors reviewed 20 studies and concluded that despite study limitations, there were modest findings of adverse health impacts with several studies focusing on birth outcomes.In a scoping review published one year later, Deziel et al. (2020) reviewed 29 articles, excluding outcomes based on self-report but including outcomes related to non-chemical causal pathways (sexually transmitted infections (STIs) and mental health outcomes).They concluded that the available research points to a growing body of evidence of health effects in communities living in proximity to oil and gas development (Deziel et al., 2020).
Our review identified a number of studies reporting adverse effects of UOGD exposure on birth outcomes, most of which were retrospective longitudinal cohort or case-control studies, reducing concerns of reverse causation.Both impaired fetal growth and preterm birth have been associated with adverse cardiovascular, metabolic, neurodevelopmental, and respiratory sequelae in later life (Crump, 2020).Although fewer in number, other cohort and case-control studies identified higher rates of neural tube, congenital heart defects, any congenital anomaly, lower infant health index, and fetal/infant mortality.
In addition to birth outcomes, an increasing number of studies report higher rates of asthma exacerbation.Considering the irritant gas emissions from UOGD, this association is not surprising.Further investigation of the most prevalent airway disease, chronic obstructive pulmonary disease (COPD), is warranted.Some evidence from case-control and cohort studies also suggest an increased risk of childhood acute lymphocytic leukemia, hospital admission for heart failure, and mortality.The relatively large effect estimates observed in case-control studies on cancer warrant further investigation in future studies despite the difficulties in examining the latent effects of UOGD exposure and cancer outcomes.
Most of the reviewed studies used surrogate exposure metrics, most commonly the IDW, due to the challenges of direct monitoring over large rural areas where UOGD is most common.This approach has been criticized for potential exposure misclassification (Wendt Hess et al., 2019).However, a growing number of studies are reporting correlation between IDW metrics and regional annual air pollutant emissions and/or UOGD-implicated chemicals in household and community reservoir water sources (Caron-Beaudoin et al., 2022, 2023;Claustre et al., 2023;Elliott et al., 2018;Hill & Ma, 2022).A recently published study linking water contaminants both to UOGD activity measures and to adverse birth outcomes (Hill & Ma, 2022) strengthens the evidence of a direct effect of UOGD exposure on adverse health outcomes.Cumulative exposures, as measured by the IDW approach, may be more reflective of "real life" exposure since these metrics capture aggregate exposure routes integrated over time.Exposure measurement is becoming increasingly sophisticated with more studies incorporating phase-specific metrics, flaring events, air and water pollution directional indicators, and adjustment for non-UOGD oil and gas development exposures in their models.Future exposure measurement should build on this multi-dimensional approach and also consider potential impacts of abandoned wells which have been identified as a growing concern (DiGiulio et al., 2023;Gross, 2023).Further examination of phase-specific contamination could better inform policies and regulations to protect communities from UOGD and other oil and gas development activities.Additionally, exposures associated with wildfire-triggered ignition of UOGD facilities (Cox, 2023;Gonzalez et al., under review -preprint available at https:// earth arxiv.org/ repos itory/ view/ 6253/) and radiation exposure from NORMs (found in wastewater brine) need to be examined.
A majority of studies used health services administrative data sources and included all individuals (versus sampling) to define a study population.Reference populations for cohort studies were usually defined as those in the lowest UOGD exposure category compared to other higher levels of exposure categories.For case-control studies that examined potentially rarer events, the study populations were typically nested within a cohort of individuals with a defined condition (for example, heart failure or asthma exacerbations).Cases were defined as case events and compared with non-exacerbation control events.Most studies adjusted for demographics, SES, and relevant comorbidities: individual-level SES was often measured using level of education and/or receipt of medical assistance, and community-level SES was measured using various community deprivation indices.Individual SES may be an important confounder, and the lack of its adjustment in some studies is a limitation.A majority of studies included smoking status as a covariable (n = 30; Online Resource 3), and with the exception of cross-sectional survey studies, a majority of studies conducted sensitivity analyses (n = 30, data not shown).Studies varied in their measurement of known potential confounding variables, such as clinical data not usually available in administrative health records (for example, body size), geographical settings (urban versus rural), and other exposure variables (for example, ambient temperature) (Online Resource 3).Last, many reviewed studies used retrospective data collected for purposes other than research, making results prone to bias from missing data or misclassification bias that may have spuriously driven effect estimates away from or towards the null.Future prospective studies can help overcome the limitations related to retrospective observational studies.
Most US studies included race/ethnicity as population descriptors and adjusted for race and SES in their modeling.Fewer described the distribution of racialized populations across UOGD activity exposure levels or reported on the independent effects of these variables in their models.Some studies described potentially disproportionate exposures to UOGD among racialized groups (Cushing et al., 2020;Tran et al., 2020Tran et al., , 2021)).Others explored effect modification by race (Cushing et al., 2020;Tang et al., 2021), with some evidence of higher effect magnitudes in Hispanic (Cushing et al., 2020), Black, and Asian (Schuele et al., 2022) populations.Few studies integrated community engagement methods into their study designs (Johnston et al., 2021;Steinzor et al., 2013) despite the increasingly recognized importance of grounding research processes in community-lived experience of Indigenous and other communities disproportionately affected by UOGD (Caron-Beaudoin & Armstrong, 2019;Garvie & Shaw, 2014;Hayward et al., 2021;Wing et al., 2008).No studies examined differential effects of UOGD in Indigenous populations.Future studies should consider the impact in systemically excluded populations in their research aims and methodological approach, and ensure meaningful engagement of affected communities throughout the research process.
The pathways linking UOGD exposure and health outcomes are still unclear.One hypothesized pathway is via increased exposure to environmental contaminants such as carbon monoxide (CO), nitrogen dioxide (NO 2 ), particulate matter (PM 2.5 , PM 10 ) (Ezani et al., 2018), and VOCs proximal to and downwind to areas from UOGD operations.VOCs have been detected at elevated environmental levels in the indoor tap water and air samples taken in the homes of pregnant women living proximal to UOGD drilling operations (Caron-Beaudoin et al., 2022).Exposure to these chemicals is known to induce cellular inflammation, oxidative stress, and alterations of placental tissue (Ferguson & Chin, 2017;Saenen et al., 2019), and has been implicated in lower neonatal birthweight in multiple studies (Stieb et al., 2012).Similarly, air pollutants such as CO, NO 2 , and PM 2.5 can induce airway inflammation (Silbajoris et al., 2011) and oxidative stress on human airways (Thangavel et al., 2022).
Another hypothesized pathway is through endocrine disruption, or the ability of certain chemicals, even at extremely low levels, to block and/or mimic sex and thyroid hormones (Kassotis et al., 2016;Vandenberg et al., 2012), thereby potentially disrupting normal gestational age and labour onset.Endocrine disruption as a potential causal pathway has resulted in growing awareness of non-linear and nonmonotonic dose-response relationships and the need for researchers to proactively recognize and characterize these in reporting of results (Vandenberg et al., 2012).Only one of the studies in our review made mention of non-linear dose response (Caron-Beaudoin et al., 2021).
A third hypothesized pathway is related to the mutagenic and carcinogenic properties of certain frack fluid chemicals (e.g., benzene, ethylene oxide), heavy metals (arsenic, beryllium), NORMs (Colborn et al., 2011;Xu et al., 2019), and air pollutants.These substances have been reported in higher concentrations in the air and water proximal to UOGD operations (Caron-Beaudoin et al., 2022;Garcia-Gonzales et al., 2019;Hill & Ma, 2022).Adding evidence for this pathway, three biomonitoring studies in British Columbia reported higher levels of a metabolite of benzene, a known carcinogen (Caron-Beaudoin et al., 2018), multiple trace toxicants (Claustre et al., 2023), and particulate air pollutants (Caron-Beaudoin et al., 2023) among pregnant women living close to UOGD activity.
Our review had several limitations.First, we only included health studies where the likely pathway was related to chemical exposure.Numerous studies looking at the effects of UOGD activity on other health outcomes (e.g., traffic accidents, sexually transmitted infections, mental health) were not included.Similarly, our review did not include studies focused on the social or economic changes related to UOGD and their impacts on health.Second, due to the heterogeneity of outcomes and exposure measurements, we did not conduct a systematic review or meta-analysis or systematically apply a formal bias assessment tool.Third, due to restricted resources, this review was limited to English-and French-language publications.Given that the top UOGD-producing countries after the USA are Russia, Iran, Qatar, and China, with Canada in sixth place (Statista, 2014), we may have missed studies published in other languages.Another limitation is possible publication bias given the potential for our review to amplify reporting and publication of positive versus negative findings.We have attempted to mitigate this by characterizing both harmful and protective effects and limiting these to reported effect estimates that reach statistical significance.

Conclusion
There is a growing body of research, across multiple jurisdictions, reporting adverse effects of unconventional oil and gas development exposure on human health, with an accumulating weight of evidence particularly in relation to birth outcomes and asthma.There is some evidence of disproportionately greater impacts in racialized populations with relatively little research focused on the differential exposure levels and effect modification by systemically disadvantaged populations.The absence of Canadian published research on health effects of UOGD is notable given the geographic relationship between UOGD and Indigenous communities, the considerable time over which UOGD has taken place, and a policy of continued expansion of this activity in several provinces.There is a pressing need for future research focused on the following: prospective and community-based studies; a focus on Indigenous, racialized, rural, and disproportionately disadvantaged populations; improved exposure assessment including measurement of phase-specific UOGD, flaring, abandoned wells, and non-UOGD exposures; impacts of wildfires and NORMS; and characterization of both linear and non-linear nonmonotonic dose-response effects.

Identification of studies via databases and registers
; Hill & **

Table 2
UNGD activity and health outcomes grouped by similar outcome measures and design a Oklahoma; b Pennsylvania; c British Columbia; d Texas; e Ohio; f California; g Colorado; h Alberta; i US states; j New York † All studies are survey of self-reported symptoms, grouped by symptom type(except McKenzie et al., 2019b)§ Hospitalization rates grouped by diagnostic categories * Anencephaly and spina bifida examined ** ALL: acute lymphoblastic leukemia