Study Design. This mercury exposure project was an environmental health component of ongoing community-driven projects led by the Canadian North Helicobacter pylori (CANHelp) Working Group in western Canadian Arctic communities (www.canhelpworkinggroup.ca). The CANHelp Working Group formed during 2006-2008 in response to concerns raised by community leaders about H. pylori infection and gastric cancer risk. This research program is a collaborative effort, linking northern Canadian Indigenous communities, their health care providers and regional health authorities with investigators from multiple disciplines at the University of Alberta 12,13. At the invitation of community leaders, cross-sectional projects were established to describe the community health burden from H. pylori infection and associated disease and address community concerns. In each community, a planning committee made up of community members guided the conduct of each project and ensured that research activities were culturally appropriate and in keeping with community priorities.
Person, Place and Time. The mercury exposure project was conducted within three CANHelp Working Group community projects. The first of these projects launched in 2007 in the hamlet of Aklavik, Northwest Territories (NT) (2006 census population=590, ~92% identifying as Gwich’in [Athabascan First Nation] or Inuvialuit [Inuit])14,15. Projects began in 2010 in Old Crow, Yukon (YT) (2011 census population=245, ~85% identifying as Vuntut Gwich’in16,17, and in 2012 in Fort McPherson, NT (2011 census population=844, ~90% Tetlit Gwich’in)18. Participation in the mercury exposure project was open to all residents of these three communities during September-November 2016. Recruitment activities involved radio announcements, social media posts, flyers on community message boards, and directly contacting participants of CANHelp Working Group projects for which current contact information was available.
Informed Consent. All participants received an information sheet that outlined the study objectives, methods, information to be collected, benefits and potential risks of participation, and confidentiality protection measures. Following review of this document, each consenting participant filled out a consent form, confirming they had received enough information about the project and agreed to participate. Project information sheets and consent forms were reviewed and approved by the Research Ethics Board at the University of Alberta and have been published previously19.
Choice of Tissue for Biomarker Analysis. Evidence suggests that hair is the biological medium best suited for measuring MeHg exposure1,7,20–32. Hair from the scalp is a commonly selected matrix for biomonitoring of MeHg exposure, because MeHg accounts for approximately 80% of the total mercury found in hair and can be measured directly1,7,20–32. Practical advantages to collecting hair samples relative to urine and blood include: chemical stability; simple and non-invasive sampling; ease in storing, transporting and archiving specimens; and relatively low cost21,33–36.
Exposure Time Window. Among healthy individuals, estimates of the scalp hair growth rate range from 0.6 to 3.36 cm/month, with an average of 1 cm/month33–35. The concentration of mercury measured in hair reflects exposure over the growth period of the sampled hair, which depends on hair length. According to input from local planning committees, residents of participating communities consume the greatest amount of aquatic species, on average, during the spring and summer seasons. For this reason, hair sample collection took place during the fall season (September-November).
Hair Sample Collection. Procedures for collecting hair samples were adapted from protocols outlined by the United States Centers for Disease Control (CDC) for use in the National Health and Nutrition Examination Survey (NHANES)37. We collected all hair samples from the occipital region of the scalp using stainless steel shears, obtaining a minimum of 120 mg of hair from each participant to allow for duplication of the laboratory measurements for quality assurance/quality control (QA/QC) purposes. Given that hair length determines the exposure period represented in the strand, we also used a ruler to measure hair length (in cm) before transferring samples into a zip-closable plastic bag and applying a label specifying the sample ID number, collection date, sample weight and hair length. Additionally, we recorded information on use of permanent hair treatments, including hair dye or permanent waves, and time since the most recent treatment.
Laboratory Analysis of Samples. The collected hair samples were analyzed by the University of Alberta Biogeochemical Analytical Service Laboratory (BASL). This lab has been accredited by the Canadian Association for Laboratory Accreditation (CALA) as meeting ISO/IEC 17025 standards for the performance of specific tests. MeHg was measured in the full-length of each hair sample using gas chromatography inductively coupled plasma-mass spectrometry (GC-ICP-MS)38,39. Quality control methods employed by the lab included the use of reference material 1AEA-085 for MeHg, total mercury and other trace elements in hair. Single point calibration was applied, and the calibration standard was analyzed in 4 replicates. The relative standard deviation for the ratio of Hg isotope 201:202 was considered acceptable if the value was less than 5%. If the value was greater than 5%, the calibration was repeated. Instrument and method blanks and a second source reference material were also used to monitor contamination with MeHg, accuracy and instrumental drift during analysis. These were incorporated into the analysis at a frequency of 1 per batch of approximately 30 samples. The instrument was re-calibrated if the second source reference material measurements were outside of the 80-120% recovery range. Additionally, water samples were spiked with a known quantity of enriched MeHg isotope (CH3201Hg) as an internal standard. Finally, laboratory duplicates were performed at a frequency of 1 per 5 samples. For added quality assurance, we divided approximately 10% of the samples into duplicates and submitted them to the lab as separate individuals. Lab personnel were blinded to all participant characteristics, including age, sex and the amount and types of aquatic species consumed.
Fish and Marine Mammal Consumption Data. We designed a population-appropriate Food Frequency Questionnaire (FFQ) focused on fish and marine mammal consumption in the past year. Community input guided the selection of included fish species and incorporation of familiar names and descriptions for locally harvested fish, to ensure respondents had a clear understanding of each FFQ item. Planning committee members identified Beluga Whale (D. leucas) as the only regularly consumed marine mammal. The FFQ measured consumption frequencies as average number of times each type of fish or whale was consumed per week (we will refer to food consumption events as “meals”). The FFQ did not include portion size to reduce the burden on participants and because validation studies have shown that attempting to ascertain portion size does not appreciably improve overall characterization of diet, because most people have poor recall of portion sizes40. To capture seasonal variability in consumption, the FFQ asked respondents to specify the time of year in which they typically harvest each aquatic species they reported consuming. The FFQ then asked respondents the typical number of meals per week of each species during the time of year they are harvested. Since it is common for community members to preserve harvested fish by drying, freezing or smoking the meat, the FFQ asked respondents to report the frequency of consuming each species during other parts of the year. Given the potential for preparation methods to alter the bioavailability of mercury in consumed fish, the FFQ also asked participants to specify how they typically prepared each type of fish/whale for eating and the parts they consumed 41. Most participants were able to identify the specific species they consumed; pictures were available for those who were unsure. The potential for the overall composition of an individual’s diet and intake of specific nutrients to directly or indirectly influence the toxicokinetic properties of MeHg has been described in the scientific literature42. For this reason, the FFQ collected data on other dietary components, including average weekly intake of: fruit, fresh fruit juice, raw and cooked vegetables, fresh or packaged milk, and yogurt.
Exposure Definition. Fish/whale consumption constituted the source of mercury exposure examined for this analysis. The structure of the FFQ permitted the creation of separate variables representing the usual frequency of consuming each reported species in units of average meals per week in each of the four seasons. We estimated the average number of fish/whale meals/week across seasons. The number of seasons incorporated in the average was determined by each participant’s hair length, with 3 cm corresponding to a single season.
Outcome Definition. The outcome for this analysis was the MeHg concentration measured in hair samples in units of μg/g on a continuous scale. Guidelines generated by Health Canada for interpreting the degree of risk associated with hair-mercury levels provide perspective for interpreting values43: hair mercury concentrations ≤6 μg/g are considered acceptable for adult males and females who are not pregnant or breastfeeding43; among children under 12 years and women who are pregnant, breastfeeding or of reproductive age, concentrations ≤2 μg/g are considered acceptable43.
Statistical Analysis. The goal of the statistical analysis was to estimate the association between fish/whale meals per week and hair-MeHg concentration in the study population. We constructed a multivariable linear regression model to estimate beta coefficients and 95% confidence intervals (CIs) as measures of the association between characteristics of interest and hair-MeHg concentration (μg/g). To confirm whether fish/whale consumption frequency could be modeled as a continuous variable, the linearity of the relationship of the continuous form of this variable to hair-MeHg concentration was assessed using a lowess plot (bandwidth: 0.80). Visual inspection of the lowess plots representing the locally weighted regression of MeHg concentration on exposure variables for each season showed that the relationships were not sufficiently linear to justify modeling exposures as continuous variables. Given lack of linearity, the fish/whale consumption variable was converted to a categorical format. A categorical format was the chosen alternative to modeling fish/whale consumption as continuous as a widely accepted approach that results in effect estimates that are easy to interpret, relative to transformations of continuous values. When possible, category boundaries were defined so that there was no more than a two-fold increase in number of servings within a category40. The purpose of this was to generate categories within which the effect of interest did not vary substantially40,44.
Variable Selection. We used purposeful selection, as proposed by Hosmer and Lemeshow (2000), to identify the best set of adjustment variables 45. This method follows a change-in-estimate approach, with variable selection decisions based on the extent to which each potential covariate influences the magnitude of exposure effects of interest: All potential covariates were included in a multivariable model and subsequently removed one at a time. If the coefficient of any independent variable changed by ≥10% with the removal of a given covariate, the removed variable was included in the final model45. Variables considered for inclusion in the model were: age, sex, community, use of permanent hair treatments, the proportion of consumed fish/whale species harvested from the ocean or local rivers, the proportion of consumed species usually prepared by cooking (versus eaten raw, dried or smoked), and other dietary frequencies, including fruits and vegetables, dairy products or regular use of dietary supplements.
Bias Analysis. Given the potential for MeHg measurement error to produce outcome misclassification, we conducted a quantitative bias analysis using the variation in measured hair-MeHg concentrations of duplicated samples. We calculated the percent change between duplicate analyses of the same participant’s hair. For participants with more than 2 samples analyzed, we used the 2 values that differed the most for the percent change calculation. For participants with more than 2 measurements, the largest difference between measured concentrations was used. To quantify the extent to which measurement error influenced inferences drawn from this analysis, we adjusted the initially measured MeHg values in two ways. First the overall mean percent change and the proportion of the repeated measurements that increased or decreased in value were used to estimate the magnitude of measurement error and frequency of change in either direction in the entire study population. Second, the mean percent change between repeated measurements and the proportion that increased or decreased were stratified by participant characteristics to apply stratum-specific estimates of the magnitude and direction of measurement error to corresponding subsets of participants, selecting at random the participants assigned increasing or decreasing MeHg concentrations. All analyses were repeated using the adjusted MeHg concentrations as outcome variables.