We recruited recreational anglers for participation in the study from May through November 2006 in two ways: “in- person” participants (n = 225) were recruited by study personnel at boat launches and fishing tournaments in coastal Louisiana; “web-based” participants (n = 438) took an Internet-based version of the same survey, which was promoted through a variety of media outlets targeted to Louisiana anglers (web sites, newspapers, and radio shows). Criteria for inclusion in the study were age ≥ 18 years, Louisiana residence, and at least one recreational fishing trip in the past 3 months. Both in-person and web-based participants provided informed consent before beginning the survey.
Of the 663 anglers recruited through both methods, 129 were excluded because they did not meet the inclusion criteria detailed above (n = 66) or because they did not complete the interview (n = 63; primarily web-based participants). After exclusions, the sample included 534 anglers (in-person n = 196; web-based n = 338).
Survey design and administration
The survey instrument [see Supplemental Material, Appendix A (doi:10.1289/ehp.1002609)] was developed by combining fish consumption questions, modified from a semiquantitative food frequency questionnaire (FFQ) used in the Nurses’ Health Study (Hu et al. 2002
), with an additional set of questions to characterize anglers’ fishing practices, fish consumption during the 3 months before the survey, sources of consumed fish (recreational vs. commercial), and demographic information. In-person surveys were administered by interviewers trained to query anglers in a standard, consistent manner. The web-based survey was identical to the in-person survey in content but was formatted so that it could be self-administered.
The research protocol, survey instrument, and consent procedures were reviewed and approved by the Harvard School of Public Health (HSPH) Human Subjects Committee before recruitment.
Hair sample collection and analysis
At the completion of the survey, all participants were asked to submit a hair sample for mercury (Hg) analysis. Overall, 402 of 534 eligible anglers (75%) provided a sample: 181 in-person (response rate = 92%) and 221 web-based participants (response rate = 65%).
The hair samples of the in-person participants were collected by the interviewer. A bundle of hairs approximately 3 mm in diameter was cut from the occipital region of the head using stainless steel scissors and tied with unwaxed dental floss to mark the proximal end.
After completing the online survey, web-based participants were mailed a kit containing detailed instructions and materials for cutting a hair sample. All instructions and materials were identical to those used for in-person hair sample collection. Web-based participants were asked to mail their sample back to HSPH within 30 days using a preaddressed, stamped envelope.
Two centimeters of the proximal end of each sample were analyzed for total Hg by thermal decomposition, amalgamation, and atomic absorption spectrophotometry [EPA method 7473 (U.S. EPA 2007
); Milestone Direct Mercury Analyzer; Milestone Inc., Shelton, CT, USA). MeHg makes up the majority of total Hg in hair (80–90%; Institute of Medicine 2007
), and total Hg in hair is a reliable biomarker of MeHg intake from fish consumption (Grandjean et al. 2002
). Precision and accuracy of this method were confirmed through repeated analysis of standards of known concentration; additional details on analysis, quality control, and detection limits are provided in the Supplemental Material [see Section 1 (doi:10.1289/ehp.1002609)]. Four hair samples were excluded because of insufficient sample size, leaving a final n
of 398 (in-person n
= 177; web-based n
= 221) for all analyses involving hair Hg concentration. Within 6 months of the sample collection, each participant was mailed a letter that contained the results of their analysis, along with guidance on how to interpret that result and further information on MeHg in fish.
Fish consumption and Hg dose
The survey’s recall period (3 months) was chosen to approximately coincide with the exposure period represented by the hair biomarker: 1–3 months before the survey [see Supplemental Material, Section 2 (doi:10.1289/ehp.1002609)]. We assessed anglers’ fish consumption over this time period using two approaches. In the “overall” approach, anglers were asked to indicate the frequency that best fit their actual finfish and shellfish consumption from the following choices: never, once a month or less, once a week, three times a week, once a day, more than once a day. In the analysis we combined the two lowest categories as well as the two highest categories because of low numbers in those groups. In the “species-specific” approach, anglers were asked to indicate the frequency that best fit their actual consumption of each of 28 common recreational and commercial fish, from the following choices: never, once in the past 3 months, once a month, once a week, three times a week, once or more a day. Participants were also prompted to report consumption of any fish types not included in the list. Each participant’s consumption frequency for each species was converted to a number of meals per day, based on a recall period of 91 days (3 months), and summed to give participants’ total species-specific fish consumption.
To quantify anglers’ Hg intake, fish Hg concentration data were gathered from a variety of sources. These ranged from regionally specific monitoring databases (e.g., Louisiana Department of Environmental Quality 2010
; U.S. EPA 2000
) to federal databases maintained by the U.S. EPA (2003)
and FDA (2006)
. Additional details on the databases and fish Hg concentration values used in the Hg dose calculations are presented in the Supplemental Material [see Section 3 and Table 1 (doi:10.1289/ehp.1002609)]. In general, only data on total Hg in fish were available, so these values were used to quantify the total Hg doses of the participants. Speciation studies have shown that 90–100% of total Hg in most finfish is MeHg (Bloom 1992
). Thus, in most cases total Hg intake via fish consumption can be used as a reasonable proxy for MeHg exposure, but to maintain the distinction between what is measured and what is relevant to health, we refer throughout to Hg intake or dose in contrast to MeHg exposure.
Two Hg dose metrics were calculated to explore how well questionnaire data predicted Hg levels in anglers’ hair. The first Hg dose metric, referred to as the “species-specific” dose, was calculated as follows:
(micrograms Hg per kilogram body weight per day), where for each fish type i
represents number of meals per day, C
represents Hg concentration (micrograms per gram), p
represents a standard portion size (129 g; U.S. EPA 1997
), and bw represents self-reported body weight (kilograms).
The second Hg dose metric, referred to as the “scaled” Hg dose, was constructed by calculating each participant’s average species-specific Hg dose per fish meal (representing a measure of the Hg richness of the fish diet) and then multiplying this by the overall fish consumption frequency. This scales the species-specific fish consumption down by the level of overall fish consumption while still retaining species-specific Hg information, and attempts to account for potential overreporting in the species-specific fish consumption variable. It is calculated as follows:
(micrograms Hg per kilogram body weight per day), where F represents overall fish consumption in meals/day and M represents species-specific fish consumption in meals/day.
We first explored the data using scatter plots and descriptive statistics to identify departures from normal distributions. The distribution of hair Hg concentrations was positively skewed, and a log-transformation of the variable was used in subsequent analyses. Variables describing fish consumption and Hg dose were also non-normally distributed, and nonparametric tests were used where necessary.
The relationships among anglers’ fish consumption, Hg doses, and hair Hg concentrations (considered here to be the most accurate indicator of MeHg exposure) were explored using multivariable linear regression. Three separate regression models were created, relating overall fish consumption, species-specific Hg dose, and scaled Hg dose to hair Hg. Age, sex, education, race/ethinicity, body mass index (BMI; calculated from self-reported height and weight), and survey type were included in each model as potentially important covariates. Calculated Hg dose metrics, age, and BMI were treated as continuous variables. Overall fish consumption, sex, race/ethinicity, and education were treated as categorical variables.
Residual plots for each model were examined to ensure that standard assumptions of linearity, normality, and homoskedasticity were met. Studentized residuals and Cook’s D
values were calculated to identify potential outliers and influential points in each model. Several influential points were identified, and the original surveys and laboratory output data for these points were reexamined and found to be error free. We compared results of our original linear regressions with robust regressions, which reduce the effect of extreme values by weighting points in proportion to their leverage (Hampel et al. 1986
). Our findings did not change when robust regression was used; thus, the results presented are those from the original linear regressions.
All analyses were performed using R version 2.6.2 (R Development Core Team 2008
). The level of statistical significance was set at α = 0.05.