Acute gastrointestinal symptoms associated with oil spill exposures among U.S. coast guard responders to the Deepwater Horizon oil spill

Abstract

Purpose:

Research investigating gastrointestinal (GI) symptoms from oil spill-related exposures is sparse. We evaluated prevalent GI symptoms among U.S. Coast Guard responders deployed to the Deepwater Horizon oil spill cleanup.

Methods:

Crude oil (via skin contact, inhalation, or ingestion routes), combined crude oil/oil dispersant exposures, other deployment exposures, deployment characteristics, demographics, and acute GI symptoms during deployment (i.e., nausea/vomiting, diarrhea, stomach pain, and constipation) were ascertained cross-sectionally via a post-deployment survey (median time between deployment end and survey completion 185 days) (N = 4885). Log-binomial regression analyses were employed to calculate prevalence ratios (PRs) and 95 % confidence intervals (CI). Effect modification was evaluated.

Results:

In adjusted models, responders in the highest (versus lowest) tertile of self-reported degree of skin contact to crude oil were more than twice as likely to report nausea/vomiting (PR=2.45; 95 %CI, 1.85–3.23), diarrhea (PR=2.40; 95 %CI, 2.00–2.88), stomach pain (PR=2.51; 95 %CI, 2.01–3.12), and constipation (PR=2.21; 95 %CI, 1.70–2.89). Tests for trend were statistically significant (p < 0.05). Results were similar for crude oil exposure via inhalation and ingestion. Higher PRs for all symptoms were found with combined crude oil/dispersant exposure than with crude oil exposure alone.

Conclusions:

These results indicate positive associations between self-reported crude oil and combined crude oil/oil dispersant exposures and acute GI symptoms.

Introduction

In April 2010, the Deepwater Horizon (DWH) oil-drilling platform in the Gulf of Mexico exploded and sank. This was followed by an oil spill that released an estimated 210 million gallons of crude oil, making this the largest marine oil spill in history [1,2]. Cleanup efforts included the deployment of over 8500 U.S. Coast Guard (USCG) members and the use of approximately 2 million gallons of oil dispersants (specifically, Corexit 9527A and 9500) [3].

USCG responders were potentially exposed to crude oil, oil dispersants, and high ambient heat [4]. A 2016 review by Laffon, et al reported that oil spill cleanup workers experience adverse mental health, physiologic, and genotoxic/endocrine effects [5]. Given the random nature of oil spills and limited research in this area, relatively little is known about the human health effects of oil spills and cleanup efforts. Moreover, less is known about the human health risks from a spill of this size and the compounding effects of various exposures.

The oil released from the DWH oil spill is known to contain volatile organic compounds (e.g., benzene, toluene, and xylene), polycyclic aromatic hydrocarbons (e.g., acenaphthene and phenanthrene), and heavy metals (e.g., nickel, cobalt, chromium, and lead) [6]. The oil dispersants used in the DWH cleanup also contain constituents which may be hazardous to humans (e.g., Corexit 9527A contains 2-butoxyethanol, propylene glycol, and organic sulfuric acid salt; Corexit 9500 contains petroleum distillates, propylene glycol, and organic sulfuric acid salt) [7,8]. Among the potential crude oil- and dispersant-related exposures of DWH responders, lead and 2-butoxyethanol are associated with acute gastrointestinal (GI) health symptoms (i.e., abdominal pain, nausea/vomiting, diarrhea, and/or constipation) in humans and animals [9, 10]. There have been reports of xylene, naphthalene, toluene, nickel, and benzene being associated with GI symptoms in humans depending on the route of exposure [11–15].

Media reports following the DWH oil spill chronicled Gulf coast residents with GI symptoms like nausea and diarrhea [16–18]; however, there is sparse epidemiological literature investigating the GI health risks from crude oil and oil dispersant exposures [19–24]. Additionally, combined exposure to crude oil and oil dispersants on GI symptoms have not been studied. The DWH Oil Spill Coast Guard Cohort (CG-DWH) was established to investigate potential acute and longer-term health symptoms associated with oil spill response exposures among USCG responders [15]. Here, we employed cross-sectional survey data to evaluate potential associations between oil spill exposures and prevalence of four acute GI symptoms (nausea/vomiting, stomach pain, diarrhea, and constipation).

Methods

Study population

The CG-DWH cohort has been described elsewhere [15]. Of 8696 USCG responders to the DWH spill, 4855 (56 %) completed a post-oil spill response survey, described previously [15], which queried information about deployment-related GI symptoms and oil spill exposures (crude oil/oily water, oil dispersant, engine exhaust), as well as oil personal protective equipment (PPE) use, and deployment information (timing, duration). The median time between deployment end dates and survey completion was relatively short (185 days). This study was approved by the Institutional Review Boards of the Uniformed Services University and the USCG.

Exposure assessment

We evaluated associations between self-reported crude oil/oily water (hereafter referred to as “crude oil”) and oil dispersant exposures with self-reported GI symptoms during deployment to the DWH oil spill response. Frequency of exposure to crude oil assessment has been described previously [25]. Briefly, exposure to crude oil in general and through direct skin contact, inhalation, and ingestion was assessed via questionnaire. General exposure, based on all routes of exposure combined, and ingestion alone were evaluated using a binary variable (ever/never). Skin contact was evaluated in tertiles (described below). Inhalation was evaluated using a 5-point Likert scale (never, rarely, sometimes, most of the time, all of the time).

Degree of skin contact exposure was calculated based on responses to 1) “Which part of your body was typically in direct contact with the oil or oily water?” and 2) “How often was any part of your body submersed in the spilled crude oil or oily water?” The Lund-Browder chart [26,27] was used to estimate the percent of body surface area exposed to crude oil. This percentage was used as a degree of exposure and the percent body surface area was then multiplied by the frequency of reported skin contact and categorized into tertiles of exposure for analysis.

We also evaluated the effects of combined self-reported exposure to crude oil and oil dispersants. Responders were asked about frequency of personally handling, applying, or coming in contact with oil dispersants and answered on the same 5-point Likert scale as described above, and responses were re-categorized into a binary (ever/never) exposure variable.

Outcome assessment

Nausea/vomiting, stomach pain, diarrhea, and constipation during deployment were queried in the survey. A 3-point scale was used to ascertain frequency of experiencing a given symptom (i.e., most of the time, sometimes, never). Reponses to these questions were recategorized to create binary variables (i.e., ever/never experiencing the symptom).

Statistical analyses

We employed adjusted log-binomial regression analyses to calculate adjusted prevalence ratios (aPRs) and 95 % confidence intervals (CIs) [28]. We selected this method because odds ratios from logistic regressions generally overestimate PRs for non-rare outcomes, and the log-binomial models have shown to better estimate the PRs [29]. When log-binomial models did not converge, we used the COPY method (SAS macro) with 1,000,000 copies [30]. To identify covariates included in the final regression models, we evaluated associations with both exposure (ever/never crude oil exposure) and outcome (any GI symptom) with the variable included. If there was an association (p ≤ 0.05) between the covariate and both the exposure and outcome, we included the covariate in the final model. Models for the crude oil exposure metrics (general exposure, degree of skin contact, inhalation, and ingestion) were adjusted for age (years), sex (male, female), employee class [active duty (AD), Selected Reserve], self-reported engine exhaust exposure [frequency of inhalation of exhaust fumes from a car, boat, trailer, or other source of exhaust; categorized as never, low (rarely/sometimes), and high (most/all of the time) [31,32]], self-reported oil dispersant exposure (never/rarely, sometimes, most/all of the time) [32,33], and ambient heat exposure [derived from hourly heat index (HI) temperatures relevant to each responder’s deployment period, combined with self-reported time spent outdoors, categorized as: low time outdoors/median HI < 95°F, low time outdoors/median HI ≥ 95°F, high time outdoors/median HI < 95°F, high time outdoors/median HI ≥ 95°F[25,34,35]]. Cut points for categorical variables were based on prior studies of the DWH Cohort population [25,31–35]. We conducted tests for linear trend (p_trend) by including exposures with > 2 categories as pseudo-continuous variables in the log-binomial regression models.

We also conducted analyses investigating the combined effects of exposure to crude oil and oil dispersant. We compared the exposed responders (1. crude oil exposure without dispersant exposure or 2. crude oil exposure with dispersant exposure) to responders with neither exposure and adjusted for age, sex, employee class, exhaust exposure, and ambient heat exposure. Six participants who reported exposure to oil dispersants but not crude oil were excluded from the crude oil/oil dispersant combination analyses.

Stratified analyses

To test for effect modification, we stratified analyses by: sex; employee class; PPE use; timing of a responder’s deployment with respect to the capping of the DWH well head on July 15th, 2010 [pre- (entire deployment pre-capping of the well), peri- (deployment started pre-capping and continued until post-capping), post-capping (entire deployment post-capping of the well)]; ambient heat exposure (low, high); deployment length (<30 days, ≥30 days); time between deployment end and survey completion date (<6 months, ≥6 months); and heat exposure (low time outdoors and median HI < 95°F, low time outdoors and median HI ≥ 95°F, and high time outdoors and median HI <95°F vs high time outdoors and median HI ≥ 95°F). We included a multiplicative interaction term between the overall ever/never crude oil exposure variable and each stratifying variable. Effect modification was considered statistically significant if the interaction p-value (p_int) < 0.05.

Sensitivity analyses

To account for the potential influence of ambient heat exposure on our findings we conducted a sensitivity analysis excluding responders in the highest category of ambient heat exposure (high time outdoors/median HI ≥ 95°F). Similarly, to investigate the potential influence of the 2010–2011 influenza season, we carried out a sensitivity analysis excluding responders with end dates of deployment after December 1, 2010.

We also conducted sensitivity analyses to account for pre-existing GI conditions. Because data on pre-existing conditions was available for AD members only, we restricted our analyses to AD responders (n = 3102). We then utilized health encounter data from the Military Health System Data Repository (MDR), a comprehensive database of all health encounters [15]. MDR data are coded using the International Classification of Diseases, 9th edition, and our query of pre-existing conditions/prior diagnoses was based on health encounters between October 1, 2007 until the date of a responder’s start of DWH response in 2010. We identified conditions relevant to each of the acute GI symptoms evaluated in our study, a priori (Supplemental Table 1). AD USCG personnel who had either one inpatient or two outpatient medical encounters for these conditions were considered to have prevalent GI conditions upon deployment and were excluded from these analyses. All analyses were conducted using SAS Version 9.4 (SAS Institute, Cary, NC, USA).

Results

The responders included in this study were generally young (43.8 % between ages 25–34), male (85.0 %), white (77.1 %), AD (63.9 %), and enlisted (72.7 %). The most common GI symptom was diarrhea (12.8 %), followed by stomach pain (9.4 %). Over half of responders reported high time outdoors (50.6 %) (Table 1).

Table 1.

Baseline characteristics of study population (N = 4855).

Characteristic	N	(%)
Age (years)
< 25	827	17.0
25 – 34	2128	43.8
35 – 50	1738	35.8
≥ 50	162	3.3
Sex
Male	4127	85.0
Female	728	15.0
Race
White	3741	77.1
Black or African American	203	4.2
Asian, AI/AN, NH/PI	182	3.7
Other	242	5.0
Unknown	487	10.0
Duty Status
Active Duty	3102	63.9
Selected Reserve	1753	36.1
Education
≤ High school	2740	56.43
> High school - ≤ Bachelor’s level degree	1775	36.56
> Bachelor’s level degree	214	4.41
Other or not specified	126	2.60
Grade
E1-E5	2233	46.0
E6-E9	1311	27.0
O1-O4, W2-W4	1128	23.2
O5-O10	183	3.8
Gastrointestinal Symptoms
Nausea/Vomiting	300	6.2
Diarrhea	622	12.8
Stomach Pain	454	9.4
Constipation	325	6.7
Heat Exposure Index
Low time outdoors and median HI < 95° F	1479	30.5
Low time outdoors and median HI ≥ 95° F	919	18.9
High time outdoors and median HI < 95°F	1380	28.4
High time outdoors and median HI ≥ 95°F	1077	22.2

Abbreviations: AI/AN = American Indian/Alaskan Native; E = enlisted; HI = heat index; O = commissioned officer; NH/PI = Native Hawaiian/Pacific Islander; W = warrant officer

Unadjusted PRs, aPRs, and 95 % CIs estimating the associations between crude oil exposures and acute GI symptoms are presented in Table 2. Ever (vs never) being exposed to crude oil via any route was associated with nausea/vomiting (aPR=1.82; 95 %CI, 1.35–2.48), diarrhea (aPR=1.72; 95 %CI, 1.41–2.11), stomach pain (aPR=1.76, 95 %CI, 1.39–2.25), and constipation (aPR=1.46; 95 %CI, 1.12–1.93). There were statistically significant associations between crude oil exposure via degree of skin contact and nausea/vomiting (aPR_{T3 vs. T1} =2.05; 95 %CI, 1.54–2.74; p_trend<0.01), diarrhea (aPR_{T3 vs. T1} =2.05; 95 %CI, 1.70–2.48; p_trend<0.01), stomach pain (aPR_{T3 vs. T1} =2.13; 95 % CI, 1.70–2.68; p_trend<0.01), and constipation (aPR_{T3 vs T1} =1.89; 95 % CI, 1.43–2.48; p_trend<0.01). Results were similar for the inhalation exposure metric for the highest versus lowest levels (“all of the time” versus “never”) for nausea/vomiting (aPR =3.33; 95 %CI, 2.07–5.21; p_trend<0.01), diarrhea (aPR =2.32; 95 %CI, 1.67–3.17; p_trend<0.01), and stomach pain (aPR =2.47; 95 %CI, 1.66–3.58; p_trend<0.01), though for constipation there was not a clear exposure-response (aPR =1.85; 95 % CI, 1.04–3.11; p_trend=<.01). Although only 163 people reported ever exposure to crude oil via ingestion, ingestion was positively associated with nausea/vomiting (aPR_{ever vs. never}=1.54; 95 %CI, 1.05–2.17), diarrhea (aPR_{ever vs. never}=2.03; 95 %CI, 1.64–2.49), and stomach pain (aPR_{ever vs. never}=1.81; 95 %CI, 1.35–2.35). Additionally, 97 % of responders who reported ingesting crude oil reported at least one GI symptom (N = 158).

Table 2.

Unadjusted and adjusted¹ prevalence ratios and 95 % confidence intervals for the associations between crude oil exposures and gastrointestinal symptoms during the DWH oil spill response (N = 4885).

Exposure Measure	Nausea/Vomiting (n = 300)						Diarrhea (n = 622)						Stomach Pain (n = 454)						Constipation (n = 325)
	N*	PR	95 % CI	aPR	95 % CI	p-trend†	N*	PR	95 % CI	aPR	95 % CI	p-trend†	N*	PR	95 % CI	aPR	95 % CI	p-trend†	N*	PR	95 % CI	aPR	95 % CI	p-trend†
Crude oil exposure via any route (ever/never)
Never	72	1.00	–	1.00	–		151	1.00	–	1.00	–		112	1.00	–	1.00	–		99	1.00	–	1.00	–
Ever	228	2.63	(2.04–3.43)	1.82	(1.35–2.48)	–	471	2.59	(2.18–3.10)	1.72	(1.41–2.11)	–	342	2.54	(2.07–3.13)	1.76	(1.39–2.25)	–	226	1.90	(1.51–2.40)	1.46	(1.12–1.93)	–
Body surface area weighted skin contact 2
Tertile 1	125	1.00	–	1.00	–		258	1.00	–	1.00	–		192	1.00	–	1.00	–		159	1.00	–	1.00	–
Tertile 2	62	1.84	(1.36–2.46)	1.41	(1.03–1.92)		136	1.96	(1.61–2.37)	1.51	(1.23–1.85)		92	1.78	(1.40–2.24)	1.39	(1.08–1.79)		62	1.45	(1.08–1.91)	1.29	(0.95–1.72)
Tertile 3	113	3.03	(2.38–3.87)	2.05	(1.54–2.74)	< 0.01	228	2.97	(2.52–3.49)	2.05	(1.70–2.48)	< 0.01	170	2.97	(2.45–3.60)	2.13	(1.70–2.68)	< .01	104	2.19	(1.72–2.77)	1.89	(1.43–2.48)	< .01
Inhalation Exposure
Never	77	1.00	–	1.00	–		170	1.00	–	1.00	–		126	1.00	–	1.00	–		106	1.00	–	1.00	–
Rarely	69	1.83	(1.33–2.52)	1.55	(1.10–2.18)		180	2.17	(1.78–2.64)	1.62	(1.32–2.00)		131	2.13	(1.68–2.69)	1.66	(1.29–2.14)		95	1.83	(1.40–2.40)	1.51	(1.13–2.01)
Sometimes	90	3.27	(2.45–5.20)	2.40	(1.71–3.37)		162	2.67	(2.18–3.26)	1.74	(1.39–2.18)		115	2.56	(2.01–3.25)	1.71	(1.30–2.25)		74	1.95	(1.46–2.60)	1.43	(1.02–1.98)
Most of the time	37	3.60	(2.45–5.20)	2.48	(1.62–3.75)		66	2.91	(2.23–3.75)	1.84	(1.38–2.43)		48	2.86	(2.07–3.87)	1.88	(1.32–2.63)		34	2.41	(1.64–3.43)	1.91	(1.25–2.85)
All of the time	27	5.56	(3.63–8.22)	3.33	(2.07–5.21)	< .01	44	4.10	(3.03–5.40)	2.32	(1.67–3.17)	< 0.01	34	4.28	(2.99–5.93)	2.47	(1.66–3.58)	< .01	16	2.39	(1.39–3.81)	1.85	(1.04–3.11)	< .01
Swallowed crude oil
Never	273	1.00	–	1.00	–		556	1.00	–	1.00	–		411	1.00	–	1.00	–		303	1.00	–	1.00	–
Ever	27	2.85	(1.93–4.00)	1.54	(1.05–2.17)	–	66	3.42	(2.76–4.14)	2.03	(1.64–2.49)	–	43	3.01	(2.25–3.89)	1.81	(1.35–2.35)	–	22	2.09	(1.35–3.04)	1.46	(0.94–2.12)	–

¹All models were adjusted for age, sex (male, female), employee class (Active Duty, Select Reserve), exhaust exposure (none, low, high), dispersant exposure (none, any), and heat exposure (low time outdoors & median HI<95°F, low time outdoors & median HI 95 ≥ °F, high time outdoors & median HI<95°F, high time outdoors & median HI 95 ≥ °F)

²Tertile variable based on what percentage of body-surface area was exposed to crude oil, and how often.

Abbreviations: CI = confidence interval; PR = prevalence ratio; aPR = adjusted prevalence ratio

^*Number with the GI symptom

^†P-trend presented for adjusted analyses

Fig. 1 presents the combined effects of self-reported crude oil and oil dispersant exposures, showing higher aPRs, though overlapping 95 % CIs, for the combined crude oil/dispersant exposure metric (i.e., “oil ever + dispersant”) than for the crude oil only metric (i.e., “oil only”): nausea/vomiting (aPR_oil+disp = 3.51; aPR_{oil only} = 1.89), diarrhea (aPR_oil+disp = 2.36; aPR_{oil only} = 1.71), stomach pain (aPR_oil+disp = 2.65; aPR_{oil only} = 1.74), and constipation (aPR_oil+disp = 2.15; aPR_{oil only} = 1.50).

Fig. 1.

Adjusted prevalence ratios (aPR) and 95 % CI for the associations between combination of self-reported crude oil and oil dispersant and GI symptoms among DWH USCG responders (N = 4855). Note: Circles indicate aPRs and dashes represent 95 % CIs.

Sensitivity analyses

Sensitivity analyses excluding responders with the highest ambient heat exposure (high time outdoors/median HI ≥ 95°F) are presented in Supplemental Table 2. Results for each of the acute GI symptoms were similar to overall results (Table 2). Table 3 presents results of sensitivity analyses excluding responders with pre-existing medical diagnoses related to the acute GI symptoms. When we restricted to AD service members and when we further restricted to AD service members without pre-existing GI conditions, the general pattern of statistically significant increasing prevalence with increasing levels of exposures remained.

Table 3.

Adjusted¹ prevalence ratios and 95% confidence intervals for the associations between crude oil exposures and gastrointestinal symptoms during the DWH oil spill response among all active duty responders (N=3102) and among active duty responders without preexisting medical diagnoses related to nausea/vomiting (N=2853), diarrhea (N=2982), stomach pain (N=2840), and constipation (N=2970).

	Nausea/Vomiting								Diarrhea
	All active duty (n=171)				Active duty without pre-existing diagnoses (n=156)				All active duty (n=346)				Active duty without pre-existing diagnoses (n=329)
Crude Oil Exposure Measure	N*	aPR	95% CI	p-trend	N*	aPR	95% CI	p-trend	N*	aPR	95% CI	p-trend	N*	aPR	95% CI	p-trend
Crude oil (Ever/Never)
Never	38	1.00	–		36	1.00	–		85	1.00	–		82	1.00	–
Ever	133	1.75	(1.18–2.67)	–	120	1.69	(1.12–2.62)	–	261	1.55	(1.20–2.04)	–	247	1.48	(1.13–1.95)	–
Body surface area weighted skin contact 2
Tertile 1	70	1.00	–		64	1.00	–		142	1.00	–		135	1.00	–
Tertile 2	30	1.15	(0.74–1.76)		28	1.22	(0.77–1.90)		70	1.39	(1.04–1.83)		65	1.35	(1.01–1.80)
Tertile 3	71	2.13	(1.48–3.07)	<0.0001	64	2.08	(1.42–3.06)	0.0002	134	2.05	(1.60–2.63)	0.0001	129	2.04	(1.59–2.64)	<0.0001
Inhalation Exposure
Never	41	1.00	–		39	1.00	–		98	1.00	–		95	1.00	–
Rarely	31	1.27	(0.78–2.05)		26	1.15	(0.69–1.92)		82	1.32	(0.98–1.77)		77	1.27	(0.94–1.71)
Sometimes	49	2.28	(1.46–3.59)		46	2.30	(1.45–3.68)		91	1.62	(1.21–2.17)		85	1.52	(1.12–2.05)
Most of the time	27	2.60	(1.54–4.36)		25	2.64	(1.52–4.51)		42	1.57	(1.08–2.25)		41	1.51	(1.03–2.17)
All of the time	23	4.08	(2.31–7.07)	<0.0001	20	3.91	(2.13–7.00)	<0.0001	33	2.13	(1.42–3.13)	<0.0001	31	2.03	(1.34–3.01)	0.0005
Swallowed crude oil/oily water
Never	154	1.00	–		141	1.00	–		308	1.00	–		294	1.00	–
Ever	17	1.90	(1.14–2.96)	–	15	1.84	(1.06–2.93)	–	38	2.26	(1.66–2.98)	–	35	2.19	(1.59–2.93)	–
	Stomach Pain								Constipation
	All active duty (n=254)				Active duty without pre-existing diagnoses (n=215)				All active duty (n=157)				Active duty without pre-existing diagnoses (n=143)
Crude Oil Exposure Measure	n*	aPR	95% CI	p-trend	n*	aPR	95% CI	p-trend	n*	aPR	95% CI	p-trend	n*	aPR	95% CI	p-trend
Crude oil (Ever/Never)
Never	59	1.00	–		54	1.00	–		41	1.00	–		38	1.00	–
Ever	195	1.65	(1.20–2.29)	–	161	1.43	(1.02–2.02)	–	116	1.73	(1.15–2.65)	–	105	1.59	(1.05–2.48)	–
Body surface area weighted skin contact 2
Tertile 1	104	1.00	–		87	1.00	–		73	1.00	–		64	1.00	–
Tertile 2	51	1.34	(0.95–1.86)		42	1.33	(0.91–1.91)		26	1.09	(0.68–1.70)		24	1.16	(0.71–1.84)
Tertile 3	99	2.07	(1.54–2.79)	<0.0001	86	2.08	(1.51–2.88)	<0.0001	58	2.05	(1.39–3.03)	0.006	55	2.16	(1.43–3.25)	0.0003
Inhalation Exposure
Never	65	1.00	–		60	1.00	–		45	1.00	–		42	1.00	–
Rarely	62	1.49	(1.04–2.13)		50	1.29	(0.88–1.89)		39	1.56	(0.99–2.45)		33	1.38	(0.85–2.23)
Sometimes	70	1.89	(1.32–2.72)		56	1.58	(1.07–2.32)		40	1.84	(1.15–2.97)		37	1.73	(1.06–2.83)
Most of the time	31	1.75	(1.11–2.71)		27	1.64	(1.01–2.61)		23	2.28	(1.30–3.89)		23	2.25	(1.28–3.90)
All of the time	26	2.65	(1.63–4.21)	<0.0001	22	2.38	(1.40–3.92)	0.0009	10	1.89	(0.86–3.79)	0.0004	8	1.50	(0.63–3.19)	0.02
Swallowed crude oil/oily water
Never	231	1.00	–		196	1.00	–		148	1.00	–		135	1.00	–
Ever	23	1.86	(1.22–2.69)	–	19	1.82	(1.14–2.74)		9	1.24	(0.60–2.25)	–	8	1.19	(0.54–2.22)	–

^*Number of exposed with the acute GI symptom.

¹All models were adjusted for age group(<25, 25–34, ≥35 years), sex (male, female), exhaust exposure (none, low,high), dispersant exposure (none, any), and heat exposure (low time outdoors & median HI<95 °F, low time outdoors & median HI 95≥ °F, high time outdoors & median HI<95 °F, high time outdoors & median HI 95≥ °F).

²Tertile variable based on what percentage of body-surface area was exposed to crude oil, and how often. Abbreviations: CI = confidence interval; aPR = adjusted prevalence ratio.

Stratified analyses

Supplemental Table 3 presents results for stratified analyses. Risk estimates were somewhat higher for males than females, although the only significant difference was for diarrhea (males: aPR=1.79 (1.44–2.25); females: aPR= 1.56 (0.93–2.60); p_int= 0.04). There was evidence of statistically significant effect modification by PPE use for diarrhea (aPR_{no PPE use}=1.13, aPR_{PPE use}=2.17; p_int =0.003) and stomach pain (aPR_{no PPE use}=1.36, aPR_{PPE use}=2.07; p_int =0.03). Among responders who completed the survey 6 months or later from deployment end, aPRs for each GI symptom were slightly attenuated; however, there was no evidence of effect modification. Although aPRs in the highest ambient heat exposure groups appeared generally higher than those in the lower exposure groups, p_int-values were not statistically significant.

Discussion

Our results indicate positive associations between DWH oil spill cleanup exposures and acute GI symptoms in USCG responders. Elevated aPRs were found for all GI symptoms with crude oil exposure via any route and ingestion (ever vs never) and increasing aPRs were found with increasing crude oil exposure via dermal and inhalation routes in an exposure-response manner. We found elevated aPRs, though overlapping 95 % CIs, for acute GI symptoms for the combined crude oil/dispersant exposure compared to crude oil exposure alone. Interpretation of results did not meaningfully change in sensitivity analyses when we excluded responders in the highest ambient heat exposure category or when we restricted to AD responders and excluded AD responders with relevant pre-existing medical conditions. Stratified analyses indicated effect modification with crude oil exposure by PPE use, with significantly higher PRs for diarrhea and stomach pain found among responders reporting ever (vs never) PPE use.

Our findings are consistent with other studies [19–24], though several previous studies did not, or had limited ability to, control for confounding [20,21,24], had relatively small sample sizes [19,20], or focused on study populations of residents living near oil spills, as opposed to workers involved in oil spill clean-up [19,21–23]. Outcome assessment among these studies also varied, with one study evaluating reported symptoms in the 8-month time period following an oil spill [23], one evaluating four-week prevalence of symptoms after an oil spill [22], and another assessing GI symptoms during cleanup efforts [20]. Additionally, we were able to incorporate greater granularity for some exposure routes (i.e., a 5-point Likert scale rather than ever/never). Despite these differences, results were similar between our work and prior studies.

To our knowledge, no prior studies have evaluated combined exposures to crude oil and oil dispersants and associated GI symptoms. Kim et al., found that treating human fecal microbiota in vitro with crude oil plus COREXIT 9500 resulted in greater changes in fecal microbiota than did crude oil alone [36] by increasing intracellular oxidative stress, which has been associated with several GI conditions [37,38]. Because dispersants alter the properties of crude oil, it is unknown how this may impact human exposure and associated health outcomes.

Nausea/vomiting is a symptom of many illnesses, including influenza [39]. Although much of the cleanup efforts occurred prior to influenza season, some deployments lasted into the 2010–2011 winter months. In a sub-analysis excluding responders with deployment dates ending after 1 December, 2010 (n = 426), the associations between oil exposure and GI health symptoms were either unchanged or increased slightly (results not shown). The weekly percentage of US outpatient visits for influenza like illness did not exceed national baseline levels (2.5 %) until the week ending December 25, 2010 [40], hence the first of December was chosen as the cut-off to avoid seasonal influenza influencing these results.

Ambient heat is an important consideration. USCG response occurred during “above normal” temperatures [41], and heat-related illnesses can result in symptoms including nausea/vomiting [42]. The National Institute for Occupational Safety and Health reported heat as a primary exposure of concern during the DWH response and described heat stress as a significant hazard due to the PPE needed for some cleanup duties [4]. However, in an earlier study within this population of USCG responders, analyses stratified by PPE use found stronger associations between ambient heat exposure and heat-related symptoms in never PPE users (PR=2.23; 95 %CI: 1.61, 3.06) than PPE users (PR=1.64; 95 %CI: 1.14, 2.36) [34]. Because we lacked information on specific duties, exact PPE used by responders, and the appropriate PPE for the responders’ duties, we were limited in our ability to assess the true relationship between PPE use and heat-related symptoms in this population. A potential explanation for our findings is that PPE use made USCG responders feel hotter than they would have without PPE, contributing to symptoms such as nausea/vomiting. Although we adjusted for ambient heat exposure category and performed sensitivity analyses excluding responders in the highest category of ambient heat exposure, we cannot discount residual confounding. However, analyses stratified by ambient heat exposure showed limited evidence of effect modification by heat.

Stratified analyses indicated effect modification by PPE use in our study, with statistically significant increased PRs among ever (versus never) PPE users for diarrhea or stomach pain and suggestive increased prevalence of nausea/vomiting. Despite these findings, the data in our study pertaining PPE use is rather crude and we found that responders who reported ever wearing PPE reported greater exposure to crude oil (results not shown). Therefore, the likely explanation is that reported PPE use is a proxy for increased levels of, or opportunities for, crude oil exposure.

Our study has several limitations. First, exposure and outcome assessment were self-reported and may be subject to recall error. However, we did not find evidence of effect modification by timing between deployment end date and survey completion, instead finding similar patterns but slightly attenuated aPRs among responders with greater time between deployment end and survey completion end. Therefore, any potential recall error is likely to be non-differential. Second, given this study’s cross-sectional nature, these GI symptoms cannot be temporally related to the exposures. However, sensitivity analyses excluding those with pre-existing GI conditions produced similar results. Third, we conducted multiple comparisons across various exposures and several GI outcomes, so some results may be statistically significant due to chance. Most findings were confirmed in sensitivity analyses, however, reducing this risk. Fourth, not all USCG responders to the DWH oil spill participated in the survey (55.8 % response rate); however, differences in baseline characteristics between responders who did and did not complete the survey did not meaningfully differ [15]. Finally, our USCG population was predominantly young, white, and male, and results may not be generalizable to all oil spill responders due to potential selection bias.

A major strength is that because USCG members must maintain a certain level of medical readiness, the likelihood of existing comorbidities in this population was low and we were able to exclude USCG members with pre-existing GI conditions in a sensitivity analysis, lending further confidence to our findings. Second, our large study population enabled us to investigate effect modification by several factors and covariate data ascertained through questionnaires allowed us to control for potential confounding. Finally, we were able to assess crude oil exposure via multiple routes and with greater granularity for some routes of exposure than in most previous studies (i.e., 5-point Likert scale rather than ever/never).

Conclusion

Among USCG responders deployed to the DWH oil spill cleanup, self-reported crude oil exposure via skin contact, inhalation, or ingestion was positively associated with acute GI symptoms. Self-reported exposure to both crude oil and oil dispersants was associated with higher prevalence of GI symptoms than was exposure to crude oil alone. The relaxation of offshore drilling regulations [43] and expansion of drilling in increasingly harsh environments [44] as well as the limited understanding of human health consequences of crude oil exposure highlight the importance of continued study of oil spill exposures and associated health risks.

Abbreviations:

AD

Active Duty

CG-DWH

DWH Oil Spill Coast Guard Cohort

CI

confidence interval

DWH

Deepwater Horizon

GI

gastrointestinal

HI

heat index

ICD-9

International Classification of Diseases, 9th edition

MDR

Military Health System Data Repository

OR

odds ratio

PPE

personal protective equipment

PR

prevalence ratio

USCG

United States Coast Guard