Data collection
The GuLF STUDY is a prospective cohort study of individuals who worked for at least one day on, or who trained for but did not work on, the response and clean-up of the 2010 Deepwater Horizon oil spill. Participants in the study include individuals who completed mandatory oil spill safety training in order to take part in the oil spill response and clean-up as well as government workers and oil professionals. Details of the study design and participant enrollment have been described previously [3].
Recruitment for the GuLF STUDY began in March 2011, 11 months after the start of the oil spill, and continued through May 2013. Eligible participants were at least 21 years old and lived in the United States (US) at the time of enrollment. From a list of 62,803 presumably unique names with sufficient contact information, a total of 32,608 participants were enrolled into the study. Of these, we excluded 999 individuals who completed a Vietnamese language abbreviated version of the questionnaire which did not collect complete information on oil spill clean-up jobs, leaving 31,609 participants for this analysis.
During the enrollment telephone interview, conducted in English or Spanish, participants were asked to provide demographic and lifestyle information, details about tasks they performed during the oil spill response and clean-up, and information on their personal health history, including any diagnoses of MI. If the participant reported ever having an MI, they were asked the month and year of their first diagnosis. If they could not recall the month and year, they were asked their age at diagnosis. We excluded 495 study participants (366 clean-up workers, 129 non-workers) who reported an MI occurring prior to the oil spill, and 5 participants who reported an MI with unknown timing, resulting in a final sample size of 31,109.
Information on oil spill clean-up activities was assessed retrospectively during the enrollment telephone interview, 1–3 years after the spill. Participants were asked their home residence location, work locations during the oil spill response and clean-up, dates of oil spill work, and detailed information on their tasks, and contact with oil, dispersants and other chemicals during the oil spill. The study protocol was approved by the Institutional Review Board (IRB) of the National Institute of Environmental Health Sciences and this analysis was approved by the IRB at the University of North Carolina at Chapel Hill.
Defining outcomes, exposures, and covariates
The outcome of interest was any incident self-reported diagnosis of a first nonfatal MI. MI considered in this analysis were those occurring after an individual began clean-up work, or, for non-workers, from the start of the oil spill on April 20, 2010 until the enrollment interview. Therefore, the risk period for an incident MI was the three-year period following the start of the oil spill.
In this analysis we evaluated the complex and varied exposures that workers faced during the oil spill. This included participation in clean-up work and duration of clean-up work, clean-up tasks, heat stress, and exposures to crude oil and burning oil during the clean-up work. To consider the impact of community level stress resulting from the spill, we assessed associations with living in, or adjacent to, a county or parish with coastline oiled during the spill. We considered both coastal and adjacent counties because these areas were most likely to have been impacted socioeconomically by the oil spill. Living in, or adjacent to, a county oiled during the spill was also a predictor of post-spill adverse mental health outcomes [26].
The exposures of interest in this analysis were defined as participation in clean-up work (yes, no); duration of oil spill clean-up work (categorized based on the distribution of work duration as 1–30 days, 31–90 days, 91–180 days, > 180 days); highest exposure clean-up job (response, operations support work, clean-up on the water, clean-up on land, decontamination, administrative support – classified hierarchically (highest to lowest) by likely level of exposure to total hydrocarbons as an indicator of exposure to oil spill chemicals [19]); maximum overall total hydrocarbon (THC) exposure during oil spill work (< 0.30 ppm, 0.30–0.99 ppm, 1.00–2.99 ppm, ≥3.00 ppm) as estimated from a job exposure matrix [3, 19] as described below; potential work exposure to burning crude oil (yes, no), which was derived from work task and location information; ever having had to stop clean-up work activities due to heat (yes, no); and residential proximity to the oil spill (“direct or indirect proximity”: resident of a county or parish that had coastline oiled from the spill or was adjacent to a county with oiled coastline; “away from the spill”: elsewhere in the Gulf states or broader US). All exposure data were self-reported or were derived from self-reported information, aside from maximum THC exposure estimates which also incorporated information on hydrocarbon concentrations from personal exposure monitors.
A job exposure matrix was used to assign workers to maximum total hydrocarbon (THC) exposure categories based on their highest exposure task during the clean-up. The job exposure matrix (JEM) was created using data from personal exposure monitor measurements of VOCs, including THC, collected during the oil spill response and clean-up. Exposure ranges were determined and ordinal estimates for level of THC exposure were assigned to exposure groups defined by vessel or vessel type (for clean-up work on the water), job type and task, and time period of clean-up work and ultimately linked to data from participant questionnaires to estimate individual exposure levels. Levels of exposure to THC served as a proxy for oil spill chemical exposure and was assigned to each self-reported work task using the JEM. Most participants reported multiple work tasks during clean-up, with some reported during the same time period. For this analysis, workers were characterized by their maximum overall THC exposure across all self-reported jobs/tasks [19].
Information on other covariates including age, gender, ethnicity, cigarette smoking, height and weight, income, and education were ascertained via self-report during the enrollment telephone interview.
Estimating risk of heart attack
We used log binomial regression to estimate risk ratios (RR) for the association of each exposure with nonfatal MI, with separate models for each association of interest. Analyses focused on worker/non-worker status and residential proximity to the oil spill included the full cohort, while analyses of specific clean-up work characteristics (work duration, job type, THC exposure, burning oil exposure, heat exposure) were among workers only, using workers with the lowest exposure or no exposure as the referent group. Each regression model adjusted for age at enrollment (20–29, 30–39, 40–49, 50–59, 60–64, ≥65 years), gender (male, female), cigarette smoking (current, former, never), body mass index (BMI) (< 25, 25–29.9, ≥30 kg/m2), maximum education attainment (less than high school, high school diploma/GED, some college/2-year degree, 4+ year college graduate) and residential proximity to the oil spill (direct/indirect, away from the spill). We were unable to control for finer categories of smoking than current/former/never because there was a substantial amount of missing data for pack-years of smoking among former smokers. When modeling associations with residential proximity to the oil spill, stopping work due to heat, highest exposure job, and working near burning oil, we controlled for clean-up work duration in addition to the other confounders.
We evaluated the impact of heat stress on MI in several ways. We compared those who reported ever having to stop clean-up work due to heat to those who did not report this. We also carried out sensitivity analyses in which we adjusted for this measure of potential heat stress in the other exposure models to determine if adjusting for heat stress changed any associations with nonfatal MI. To differentiate potentially more acute heat stress-related MI events from any longer-term impacts of exposures on MI risk, we compared associations with nonfatal MI in two time periods - the active clean-up period up beginning April 20, 2010 until the end of 2010, and 2011 through the end of follow-up. Most clean-up activities had ceased by the end of 2010 and subsequent MI events were unlikely to be acutely due to clean-up-related heat stress.
Effect measure modification
We assessed the presence of effect measure modification by smoking or residential proximity to the oil spill, using stratified analyses and product terms. Since smoking is a strong risk factor for MI, associations between oil spill exposures and MI could be more apparent among nonsmokers or there could be synergistic effects between smoking and oil spill chemicals. We considered residential proximity to the oil spill as a potential effect measure modifier to evaluate the possibility of a synergistic effect among clean-up workers living in areas affected by the spill, who may have experienced socioeconomic stress related to the oil spill in addition to chemical/physical oil spill-related exposures. Smoking status was defined as current/former smoker vs. never smoker, and residential proximity to the spill remained coded as direct/indirect proximity to the spill vs. away from the spill. Estimates for the exposure/outcome relationships were generated using log binomial regression models after stratifying by the modifier. To assess effect measure modification on the multiplicative scale, a product term between the modifier and main exposure was included in each model. We tested the statistical significance of modification of the risk ratio on the multiplicative scale using the likelihood ratio test (LRT) to compare models including and not including the product term, with alpha = 0.05.