Self-reported oil spill exposure and birth outcomes among southern Louisiana women at the time of the Gulf oil spill: The GROWH Study

Abstract

Background:

The chemical, physical, economic, and social effects of a major oil spill might adversely affect pregnancy health.

Objectives:

To examine the relationship between oil spill exposure and birth outcomes in a cohort of women living near the Gulf of Mexico at the time of the 2010 oil spill

Methods:

Between 2012 and 2016, 1375 women reported their exposure to the oil spill, and at least one livebirth. Five hundred and three had births both before and after the oil spill. Indicators of oil spill exposure included self-reported financial consequences, direct contact with oil, traumatic experiences, loss of use of the coast, and involvement in litigation. Birth outcomes were low birthweight (LBW; birthweight <2500 g) and preterm birth (PTB; >3 weeks early). Women who were not pregnant at the time of the interview (n=1001) self-reported outcomes, while women who were pregnant (n=374) primarily had them abstracted from medical records (n=374). All pregnancies prior to the oil spill were considered unexposed; those after the oil spill were considered exposed or unexposed depending on interview responses. Generalized estimating equations were used to control for clustering within women, with control for confounders.

Results:

The most common type of exposure was economic (49%), but 302 women (22.0%) reported some degree of direct contact with the oil. Associations between most indicators of oil spill exposure and pregnancy outcomes were null, although when all pregnancies were examined, associations seen were with high levels of contact with oil for LBW (adjusted Odds Ratio [aOR] 2.19, 95% CI, 1.29-3.71) and PTB (aOR 2.27, 1.34-3.87).

Discussion:

In this community-based cohort, we did not find associations between report of exposure to the oil spill and birth outcomes. Research incorporating specific biomarkers of oil spill exposure and stress biomarkers would be valuable, to allow for assessing both perceived and actual exposure, especially when direct toxicant exposure is minimal.

1. Introduction

The Deepwater Horizon disaster was the largest marine oil spill in history (Mascarelli, 2010). The effects of the oil spill included not only direct health effects of environmental contamination, but also the anxiety and concern about adverse environmental and health effects, and the economic effects of the fishing closures and the drilling moratorium. The Institute of Medicine, currently the National Academy of Medicine (IOM (Institute of Medicine), 2010), as well as local community groups, identified pregnant women as a particular population of concern. The location and restrictions around the spill itself meant that relatively few people outside the clean-up workers were directly adjacent to the spill (Goldstein et al., 2011). However, the social and financial effects were extensive, due to the drilling and fishing moratoria and effects on tourism, and there was widespread concern about secondary environmental exposures, such as oil and tar balls on beaches (Jonsson, 2010), and exposure via seafood (Rotkin-Ellman et al., 2012).

An oil spill may expose people directly exposed to oil, volatile organic compounds (VOCs), tar balls, or fine particulate matter (PM) (Nance et al., 2016), whose effects may be modified by use of dispersants (Afshar-Mohajer et al., 2019; National Academies of Science, 2020). The combination of evaporation, burning, and clean-up operation emissions can also result in air pollution in the form of increased ozone, peroxyacetyl nitrate, NO_x, carbon monoxide, and black carbon (Middlebrook et al., 2012). Fish and shellfish may take up contaminants, and oil compounds may enter the food supply (Rotkin-Ellman et al., 2012). For the Deepwater Horizon spill specifically, the drilling moratorium and closure of the fishing grounds were financial strains on many families, while social factors such as concern about the environmental destruction and involvement in litigation were also sources of stress. Several studies indicate associations between oil spill exposure and worsened mental health, especially among those with significant direct exposure to the oil spill, financial effects, or disruption of life (Laffon et al., 2016).

As pregnant women are a relatively small proportion of the population and do not normally work in environmental disaster clean-up, few studies have addressed the effects on this group of environmental disasters, outside of Chernobyl and the environmental effects of 9/11 (e.g., (Ericson and Kallen, 1994; Spratlen et al., 2018). One study in Nigeria found a doubling of risk of neonatal mortality in areas near oil spills (Bruederle and Hodler, 2019). A study after the Hebei Spirit (South Korea) oil spill found higher rates of skin and abdominal pain among pregnant women but does not appear to have assessed birth outcomes (Kim et al., 2009). Considering more general exposures to petroleum, a review of exposure to oil and gas extraction activities and pregnancy outcomes reported that several relatively high-quality studies have found a relationship between unconventional oil/gas activities (i.e., hydraulic fracturing) and preterm birth (PTB), while the studies of low birthweight (LBW) were insufficient to draw conclusions (Balise et al., 2016). Another study found that flaring from such activities was associated with PTB but not fetal growth (Cushing et al., 2020). The objective of this analysis was to estimate the association between aspects of exposure to the Gulf oil spill and adverse birth outcomes.

2. Material and Methods

The Gulf of Mexico oil spill occurred between April and September 2010; the GROWH (Gulf Resilience on Women’s Health; U19 ES020677) Study was conducted between 2011 to 2016 (Harville et al., 2018; Lichtveld et al., 2016). Data were available on 1638 participants recruited from prenatal, health, and Women, Infants, and Children (WIC) clinics; day care centers; and community events and gathering places in southeastern Louisiana (targeting West Bank of Jefferson and Orleans Parishes, and Lafourche, Plaquemines, St. Bernard, and Terrebonne Parishes). Eligibility criteria included: aged 18–45 at enrollment, living in the Gulf area during the oil spill, and, if pregnant, carrying a singleton gestation. 1375 women had information on at least one oil spill exposure and birth outcome with information about pregnancy timing relative to the spill, while 263 either had no pregnancies or were missing data on exposure or outcome. Excluded women had lower ages and BMIs; were less likely to be smokers, Black, or White; and were less likely to reside in a coastal ZIP code (Table S1).

2.1. Exposures

Oil spill experience was measured with questions from several sources, including questions about: (1) a participant’s involvement in work on the clean-up and contact with oil, taken from the Gulf Long-Term Follow-Up Study (National Institute of Environmental Health Sciences); (2) direct exposure to the oil spill, taken from studies performed after the Exxon Valdez spill (Palinkas et al., 1993); and (3) the social and economic effects of the oil spill, from a previous study (GUMBO, R03 NR012052). Factor analysis was used to see if the patterns of grouping of similar response questions matched the underlying hypothesized latent constructs, and results were then grouped for power: financial/income consequences (dichotomized as 0-1 indicators/more and categorized as 0-1 indicator/2-3 indicator/4+ indicators); direct contact with oil (handled wildlife, came into contact with oil during cleanup, spent time in an area were oil or oily materials were used, came into contact with oil during activities such as fishing or hunting, or reported coming into contact with oil between April 2010 and April 2011; categorized as none or ≤ 1 month of exposure/1 indicator or > 1 exposure/2+ indicators); oil spill-related trauma (damage to people or own property; dichotomized as any/none, categorized as yes to both, yes to one, none); loss of use of the coast (damage to areas where one or one’s family fishes, boats, or goes to the coast or beach); and involvement in litigation (Table S2). In addition, a summary variable was created across categories for total exposure to the oil spill (sum of the above individual experiences coded as 3-level variables: financial, direct contact, trauma, loss of use, and litigation, weighted equally; theoretical range was 0 to 10; range in this sample was 0 to 9; recategorized to 0, 1-2, 3-4, and 5+).

2.2. Outcomes

Each woman was also asked for a reproductive history including up to 8 pregnancies. Questions for participants included date of/age at each pregnancy, outcome of each pregnancy (livebirth, stillbirth, miscarriage, induced abortion, ectopic/molar/other), birthweight of each child, and whether the pregnancy was early, late, or on time (and, if not on time, by how much). LBW was defined as birthweight <2500 g and PTB as birth more than 3 weeks early. If a woman was still pregnant at the time of the interview, her medical records were abstracted for birthweight and gestational age (characteristics of women by outcome are provided in Table S3)

2.3. Analytic methods

Each pregnancy (based on estimated last menstrual period and delivery date) was determined to have occurred before (prior to 20 April 2010), or during or after the oil spill (on or after 20 April 2010 through 2016). If the precise date of the start or end of pregnancy was not known and it was estimated to have occurred within 6 months of the oil spill, it was omitted from the analysis because it could not be categorized as exposed or unexposed (n=41 pregnancies; usually this occurred when a woman reported age at a pregnancy rather than date). Results were analyzed with generalized estimating equations, using each woman as a cluster, with a logit link; zip code was also incorporated to allow for any spatial autocorrelation. Multiple imputation was used to account for missing covariate data (166 observations with missing data for covariates included in modeling; 10 imputations with Markov chain Monte Carlo methods).

The first set of models includes only pregnancies after the oil spill. In order to address possible biases due to correlated over-reporting or unmeasured confounding by woman, several additional analyses were run. Women were categorized based on their reported exposure to the oil spill, with an interaction between pregnancies occurring prior to and those occurring after the oil spill. If exposure was equally or more predictive of outcomes prior to the oil spill (i.e., the interaction was not significant), it was considered that any associations were likely due to unmeasured confounding. Thus, the pregnancies before the oil spill serve as negative controls (Lipsitch et al., 2010). A sensitivity analysis was also performed stratifying the analysis by time since the oil spill (less than and greater than 2 years); in most cases the sample size was too small to allow for model convergence, but where results were possible they have been added to the text. A separate analysis also incorporated all pregnancies before the oil spill as unexposed, with those after the oil spill are categorized as exposed or unexposed as defined above. All analysis was conducted using SAS version 9.4 (SAS Institute Inc.).

The study was approved by the Institutional Review Boards of Tulane University, Ochsner, and Louisiana WIC, and all participants provided written informed consent (Tulane IRB # 239911).

3. Results

Analysis was limited to women with a history of at least one birth or who were pregnant at the time of the interview (Table 1). The analyzed population was about two-thirds black, predominantly low-income, and of high BMI. Approximately one-third were exposed to the oil spill in some way, with 302 (22.0%) reporting some degree of direct contact with the oil, and about 10% being significantly affected by the spill (3+ indicators of exposure). The 503 women who had pregnancies both before and after the spill were similar in profile to the overall cohort, with the exception of being older on average and all having two or more births.

Table 1.

Participants in the GROWH study with information on birth outcomes, oil spill exposure, and timing of pregnancy.

	all women (n=1375)	birth before and after oil spill (n=503)
	N (%)	N (%)
age
18-25	385 (28.5)	80 (16.3)
>25-30	391(28.9)	180 (36.6)
>30-35	297 (22.0)	145 (29.5)
>35	279 (20.6)	87 (17.7)
parity at interview
nulliparous	65 (4.8)
multiparous	1297 (95.2)	503 (100.0)
race
black	866 (63.3)	331 (66.3)
white	377 (27.6)	129 (25.9)
some other race*	125 (9.1)	39 (7.8)
married or living with partner
yes	547 (40.6)	202 (41.1)
no	799 (59.4)	289 (58.9)
income
<$15K	623 (47.0)	239 (48.9)
$15K-35K	443 (33.4)	165 (33.7)
>=$35K	260 (19.6)	85 (17.4)
smoker
current	364 (26.6)	131 (26.2)
former	72 (5.3)	23 (4.6)
never	931 (68.1)	347 (69.3)
BMI
<=20	73 (5.5)	30 (6.3)
20-25	295 (22.4)	99 (20.7)
>25-30	334 (25.4)	133 (27.8)
>30	615 (46.7)	216 (45.2)
pregnant at time of interview
yes	374 (27.2)	157 (31.2)
no	1001 (72.8)	346 (68.8)
ZIP code of residence
coastal	491 (36.9)	184 (37.6)
non-coastal	839 (63.1)	305 (62.4)
any exposure to oil spill
yes	1000 (68.5)	324 (64.4)
no	460 (31.5)	179 (35.6)
financial loss
yes	667 (48.8)	222 (44.4)
no	700 (51.2)	278 (55.6)
direct contact with oil
yes	302 (22.0)	99 (19.7)
no	1073 (78.0)	404 (80.3)
direct impact to people or property
yes	79 (5.8)	29 (5.8)
no	1291 (94.2)	473 (94.2)
loss of use of coast
yes	560 (41.0)	174 (37.4)
no	807 (59.0)	327 (65.3)
litigation
yes	342 (24.9)	122 (24.3)
no	1030 (75.1)	380 (75.7)
total number of indicators of experience
0	388 (28.2)	158 (31.4)
1	367 (26.7)	134 (26.6)
2	449 (32.7)	156 (31.0)
3+	171 (12.4)	55 (10.9)
time between oil spill (may include more than one pregnancy per woman)
>5 years before	924 (30.9)	361 (24.8)
2-5 years before	440 (14.7)	287 (19.7)
within 2 years before	335 (11.2)	201 (13.8)
2 years after	448 (15.0)	228 (15.6)
>2 years after	843 (28.2)	381 (26.1)

^*includes those self-identifying as Asian, Hispanic with no other race, or Middle Eastern with no other race

The only associations seen among pregnancies after the oil spill (Table 2) were with high levels of contact with oil for LBW (aOR 2.19, 95% CI, 1.29-3.71) and PTB (aOR 2.27, 1.34-3.87), and for loss of use of the coast and PTB (aOR 1.52, 1.01-2.30). When limited to those with a birth before and after the spill (Table 3), the corresponding OR for LBW and high oil contact exposure was 1.88, 0.88-4.02 and for high levels of contact and PTB was 2.12, 0.90-4.99; for loss of use of the coast and PTB it was 1.71, 0.84-3.46. Associations were also examined with pregnancies that occurred before the oil spill, to assess possible unmeasured confounding or correlated measurement error, and the interaction with time was examined. Only the interaction for contact with oil and PTB was statistically significant (Table 3), although the associations for contact with oil were stronger for both outcomes for pregnancies occurring after the oil spill than before it. However, mid-level contact with oil was also associated with higher risk of PTB in the pre-oil spill analysis. When stratified by time relative to the oil spill, the ORs for high contact with oil were similar for pregnancies occurring within two years after the spill and those occurring later, for both LBW (< 2 years, aOR 2.01, 95% CI 0.91-4.41 and >2 years, aOR 2.30, 95% CI 1.20-4.41) and PTB (aOR 2.30, 95% CI 0.99-5.35 and aOR 2.25, 95% CI 1.09-4.65). Results were similar when pregnancies before the oil spill were included in the unexposed group (Table S3).

Table 2.

Self-reported exposure to the Gulf oil spill and birth outcomes in southern Louisiana women, 2011-2016 (n=960 women/1255 pregnancies)

	low birthweight				preterm birth
	unadjusted		adjusted		unadjusted		adjusted
	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
any oil spill exposure	0.99	(0.68, 1.44)	1.06	0.73-1.54	1.13	(0.74, 1.72)	1.12	0.72-1.74
oil money	0.75	(0.52, 1.08)	0.81	0.56-1.17	0.77	(0.52, 1.15)	0.76	0.50-1.14
trauma	1.19	(0.55, 2.58)	1.19	0.56-2.51	0.93	(0.40, 2.16)	0.84	0.34-2.07
coast	1.18	(0.82, 1.71)	1.31	0.90-1.90	1.50	(1.01, 2.23)	1.52	1.01-2.30
litigation	1.15	(0.76, 1.72)	1.15	0.76-1.72	0.72	(0.44, 1.18)	0.74	0.45-1.20
contact	1.00		1.00		1.00		1.00
1 indicator	0.88	0.40-1.95	0.92	0.42-2.01	1.02	0.50-2.07	1.02	0.49-2.12
2+ indicators	2.10	1.26-3.50	2.19	1.29-3.71	2.25	1.34-3.77	2.27	1.34-3.87
overall	1		1.00		1.00		1.00
1-2 indicators	0.99	0.62-1.56	1.01	0.63-1.62	0.95	0.56-1.63	0.95	0.54-1.65
3-5 indicators	0.85	0.53-1.39	0.94	0.58-1.54	1.04	0.61-1.78	1.01	0.60-1.83
>5 indicator	1.25	0.73-2.16	1.37	0.79-2.35	1.48	0.83-2.64	1.46	0.80-2.66

Table 3.

Oil spill exposure and birth outcomes, women with pregnancies both before and after the oil spill, stratified by timing of pregnancy (n=503/1471 pregnancies)

	low birthweight					preterm birth
	post oil spill		pre oil spill		p for interaction	post oil spill		pre oil spill		p for interaction
	OR	95% CI	OR	95% CI		OR	95% CI	OR	95% CI
any oil spill exposure	0.94	0.55, 1.59	0.90	0.54, 1.50	0.81	1.55	0.74, 3.27	1.29	0.74, 2.23	0.77
income loss	0.78	0.45, 1.35	1.05	0.64, 1.71	0.44	1.04	0.52, 2.05	0.91	0.54, 1.52	0.89
trauma	1.52	0.56, 4.12	2.09	0.81, 5.44	0.79	0.87	0.18, 4.15	2.06	0.80, 5.29	0.24
loss of use of the coast	1.24	0.71, 2.17	0.93	0.55, 1.57	0.49	1.71	0.84, 3.46	1.11	0.65, 1.87	0.28
litigation	1.27	0.70, 2.29	1.59	0.96, 2.62	0.56	0.79	0.36, 1.75	1.40	0.84, 2.35	0.20
contact	1.00		1.00		0.12	1.00		1.00		0.02
1 indicator	1.24	0.38, 4.08	1.35	0.58, 3.15		1.03	0.28, 3.78	2.29	1.03, 5.09
2+ indicators	1.88	0.88, 4.02	0.84	0.40, 1.78		2.12	0.90, 4.99	0.83	0.39, 1.75
overall experiences					0.70					0.30
1-2 indicators	0.77	0.35, 1.66	0.90	0.49, 1.64		1.21	0.46, 3.14	1.53	0.78, 3.00
3-5	1.09	0.56, 2.11	0.86	0.43, 1.73		1.89	0.78, 4.57	1.25	0.66, 2.48
>5	1.12	0.54, 2.34	1.05	0.49, 2.24		1.47	0.52, 4.12	0.96	0.44, 2.08

Adjusted for age, gravidity, weight gain, BMI, race, income, education, smoking.

4. Discussion

Although pregnant women are considered a vulnerable population, few previous studies have addressed the adverse effects of oil spills on birth outcomes (Bruederle and Hodler, 2019). In this analysis of a community-based cohort, we did not find strong associations between report of social, economic, or physical exposure to the oil spill and birth outcomes. Most estimated associations were null or in the direction of protective association. The exception was contact with oil, which was associated with higher risk of LBW and PTB in some analyses. While the imprecision of the estimates and the measurement error inherent in self-report of exposure mean that no conclusion of effect can be drawn, this is worth following up in studies with more detailed measures of contact with oil compounds.

The most directly relevant previous study is from Nigeria, and indicated oil spills that occurred before conception were associated with a doubling of risk of neonatal mortality in nearby areas (Bruederle and Hodler, 2019; Nriagu et al., 2016). Both early birth and reduced fetal growth are strong risk factors for neonatal mortality, and some of the odds ratios we estimated were in the range of 2.0. However, our results are not as strong or definitive as the Nigerian study. In other environmental disasters, previous studies found that psychological effects were as strong as physical ones. Anxiety due to Chernobyl, but not the environmental threat itself, was associated with earlier births in a sample of Swedish women (Levi et al., 1989). The increased mental and emotional problems in children exposed to Chernobyl prenatally did not correlate with radiation dose, indicating that these deficits were likely due to difficulties of adaptation and relocation (Kolominsky et al., 1999). Post-9/11, post-traumatic stress was associated with reduced infant head circumference (Engel et al., 2005), while results of studies of environmental exposure were mixed, often finding effects only in subgroup analyses (Choi et al., 2008; Lederman et al., 2004; Perera et al., 2005). However, in the current analysis, effects of social and financial stress due to the oil spill were not found. This may be due to the fact that most of the population had few direct ties to industries affected by the oil spill and was fairly homogeneous with respect to income. Negative effects on fetal growth and birthweight have been seen fairly consistently after disaster, but associations with gestational age are very mixed (Harville et al., 2010).

Pollution due to oil spills has many potential adverse health effects. While little mechanistic work has directly addressed pregnant women’s exposure to oil spills, air pollution, oxidative stress, and genetic damage have been linked to adverse birth outcomes, though not always conclusively (Li et al., 2019; Patelarou and Kelly, 2014; Protano et al., 2012; Sultana et al., 2017). These exposures have been found after oil spills: air benzene and PM2.5 levels were higher in the affected parishes after the oil spill (Nance et al., 2016), and coastal sediments near oil spills have shown long-term endocrine disruption potential (Liu et al., 2018). Long-term increased oxidative stress and genetic damage have been demonstrated post-oil spill in those exposed to Hebei Spirit and Prestige oil spills (Kim et al., 2017; Perez-Cadahia et al., 2008). Studies also indicate severe health effects of oil spills on other outcomes: many post-oil spill studies show adverse respiratory effects (Laffon et al., 2016), and there was an increased risk of nonfatal myocardial infarction among long-term workers on the Gulf oil spill clean-up (Strelitz et al., 2018). That said, risk assessment for children exposed to affected beaches in Louisiana was estimated to be quite low (Black et al., 2016), and most analyses of seafood and consumption risk analysis indicated no increased risk due to the spill (Wickliffe et al., 2018). Despite this, many women reported avoiding seafood or eating less fish, and may have replaced seafood with less healthy alternatives (Simon-Friedt et al., 2016), which could have influenced pregnancy health. Most of the women in this study lived in parishes adjacent to the coast but not in the coastal zip codes, nor did they work on the clean-up, and levels of exposure to the oil-spill-related pollutants were likely low.

Strengths of the study include the systematic assessment of reproductive history, recruitment at a range of community-based locations, and inclusion of a large number of women who may be vulnerable due to income, race, or geography. While analysis can never fully remove bias in an observational study, we performed several analyses to address possible biases. Presenting the results from before the oil spill addresses the possible bias due to overreporting; performing the analysis within-woman controls for potential unmeasured confounding; and stratifying by time allows for focusing on the time period when the strongest effects would be expected. Limitations include the lack of direct biological measures of exposure to oil spill chemicals and the fact that many of the included births took place substantially later than the oil spill (although this would be consistent with the Nigerian study that found the strongest risk with preconception exposure (Nriagu et al., 2016).) Most birth outcomes were self-reported. Overall, maternal recall of these outcomes is quite good (Harville et al., 2019; Rice et al., 2007; Troude et al., 2008), and women were asked pregnancy by pregnancy with absolute measures (“How much did the baby weigh?”) rather than potentially subjective ones (“Was the baby born too small?”). A concern could be that the attention paid to the oil spill caused over-reporting of complications; due to the objective outcome, the lack of an overall increase in adverse outcomes after the spill, and the results of the within-woman analysis, we do not think this is a major source of bias. We considered low birthweight, which is not an ideal outcome due to its incorporating both length of gestation and reduced fetal growth. However, we find that women remember birthweight more precisely than gestational age (Harville et al., 2019). As these women were predominantly already at high social and financial risk, and therefore at higher risk for adverse outcomes, the potential excess risk conferred by the oil spill may have been limited.

To conclude, this study found no effects of the social or economic consequences of the Gulf oil spill on adverse birth outcomes. Research incorporating specific biomarkers of oil spill exposure would be valuable (Huang et al., 2017). Such measures may be needed at the time of exposure, given the volatile nature of oil-related compounds, and would be assisted by programs such as Disaster Research Response 2 (DR2) (National Institute of Environmental Health Sciences, 2021) and other efforts to streamline the study design, approval, and funding possibilities post-disaster. It would also be useful to incorporate more stress biomarkers into environmental epidemiological research, to allow for assessing perceived and actual exposure, especially when direct toxicant exposure is minimal.

Abbreviations

LBW

low birthweight

PTB

preterm birth