Environmental Mercury Exposure and Blood Pressure Among Nunavik Inuit Adults

In the fall of 2004, the Qanuippitaa? health survey was conducted in the 14 Nunavik communities. The target population included all of the permanent residents, except for non-Inuit households and individuals living full-time in public institutions. The survey plan was a complex, 2-stage stratified random sampling. The first stage involved selecting a stratified random sample of private Inuit households with proportional allocation. The community was the only stratification variable used. Because home addresses (civic numbers) in some municipalities are consecutive, the survey frame was sorted first by home addresses, followed by a systematic draw of a predetermined number of households to avoid selection of 2 immediate neighbors. Because many Inuit regularly move from one house to another, it was decided to sample households instead of individuals. To obtain a good representation of each community, a proportional allocation of sample units corresponding with the size of each village was chosen. In the second stage, all of the eligible individuals were asked to participate according to the survey steps or instruments. A total of 670 Inuit households were eligible, and 521 agreed to participate, which corresponds with a household response rate of 77.8%. Among 1330 adults ≥18 years of age who were eligible for the clinical section, 889 agreed to participate. Taking into account the household response rate, the participation rate for the clinical session was estimated at 52% (77.8×889/1330). A complete set of data was obtained from 806 adults ≥18 years of age. We excluded 74 individuals who were treated for systemic hypertension during the data collection period. Thus, the final sample was composed of 732 individuals. Each individual who agreed to participate in the survey signed a consent form. The study protocol was approved by the ethics committee of Laval University.

All of the information was gathered onboard the Amundsen research icebreaker. Questionnaires were used to collect information regarding age, sex, smoking, alcohol consumption, physical activity, and socioeconomic status (measured as total income). During a clinical session, blood samples and anthropometric and physiological measurements were taken. The height of participants was obtained using a rigid squared and measuring tape as they stood barefoot on a hard surface with their back against a wall. This forced participants to stand upright. Waist circumference (WC) was measured after exhalation with the tape placed horizontally to where the abdomen curves in. If a subject’s waist was not sufficiently defined, he or she was measured at roughly the location of the last floating rib.¹⁶ Hip circumference was assessed by placing the measuring tape horizontal to the hips at the pubic symphysis and the most prominent part of the buttocks. All of the foregoing measurements were recorded to the nearest centimeter. Waist:hip ratio (WHR) and body mass index [BMI=weight/height²] were calculated. BP was measured according to the Canadian Coalition for High Blood Pressure technique using mercury sphygmomanometers, 15-in stethoscopes, and cuffs sized to the subjects’ arms.¹⁷ Before having their BP taken, subjects had to have rested for 5 minutes and not eaten or smoked for ≥30 minutes. Each subject had 3 BP readings, and means of SBP and DBP were calculated using the last 2 readings. SBP and DBP intraindividual variations were 1.9% and 2.4%, respectively. PP (PP=SBP−DBP) was calculated.

Mercury, lead, and selenium were determined in whole blood, which is a biomarker of recent exposure. Determination was performed by inductively coupled plasma mass spectrometry. For mercury and lead determinations, samples were diluted 20-fold in a solution containing ammonium hydroxide before analysis. The detection limit for mercury, selenium, and lead was 0.5 nmol/L, 0.1 μmol/L, and 1 nmol/L, respectively, and each run of samples included a standard. The interassay variabilities for the mercury, lead, and selenium measurements were 2.1%, 2.8%, and 6.1%, respectively. Analyses were performed by the Institut national de santé publique du Québec Human Toxicology Laboratory, which is accredited International Organization for Standardization 17025 by the Standards Council of Canada. This laboratory is also an international leader in analytic toxicology applied to human and environmental studies and a reference institution for interlaboratory comparison programs in heavy metals measurements. Concentrations of triglycerides, low-density lipoprotein (LDL) cholesterol, and high-density lipoprotein (HDL) cholesterol were determined according to methods of the Lipid Research Clinics (US Department of Health). Determination of insulin was performed with a Roche Modular Analytics E170 (Elecsys module) autoanalyzer using a commercial double-antibody radioimmunoassay. Insulin sensitivity was estimated from the HOMA model as follows: fasting insulin×fasting glucose/22.5.¹⁸ The fatty acid composition of the erythrocyte membranes was measured after membrane purification, chloroform/methanol lipid extraction, and methylation of fatty acids, followed by capillary gas-liquid chromatography using a DB-23 column (39.0×0.25 mm ID×0.25 μm thickness) in a Hewlett-Packard gas chromatograph.

Description of all of the variables was carried out, and those not normally distributed were log-transformed. Participants taking medication for systemic hypertension at the moment of data collection were excluded from the analysis. We tested the interaction between mercury and selenium, as well as between mercury and DHA and EPA. Pearson correlation and simple regression were used to study the relationship between mercury levels and BP parameters, whereas multiple regressions were carried out to control for confounders. Variables considered as potential confounders were age, sex, HDL cholesterol, LDL cholesterol, triglycerides, insulin sensitivity, physical activity, socioeconomic status (measured as total income), smoking, alcohol consumption, n-3 fatty acids (DHA and EPA), selenium, and lead levels. We also verified the confounding effect of obesity (WC, BMI, and WHR) by using independent models that included one of the obesity indices and the other potential confounders. To assess the confounding effect, the change-in-estimate method was used. All of the variables were included in the initial models and considered as confounders if they modulated the regression coefficient ≥10%. We also used the 1-way ANOVA to test differences in PP by quartiles of mercury distribution. Means were also adjusted for the confounding factors detected in the multiple regression models. A P<0.05 was considered of statistical significance. Analyses were performed using SAS version 9.1 and SUDAAN version 9.3 software (SAS Institute).

Taking into account the complex sampling design used in this study, population weights were incorporated into all of the statistical analyses. Weighting participant answers takes into account the probability of selecting each individual as induced by the design of the survey, the rates of nonresponse, and differences observed between the sample and the population. The weight corresponds with the number of persons in the entire population who are represented by the respondent and are adjusted for age, sex, and community. Thus, the estimates generated by using population weights could be generalized to the entire population of Nunavik. To determine variances in the estimates, the bootstrap technique was used.^19–21 This technique provides precision measurements for estimates obtained from a complex sample design. Details on methodology are available in the methodologic report of the survey.²⁰

With a response rate to the clinical session of 52%, a selection bias could have been introduced if individuals who did not agree to participate differed from those included in the analyses. However, the weighting method used in this study took into account differences in age, sex, and community between respondents and nonrespondents, which minimized the risk of a selection bias. Another potential limitation of our study is the use of casual BP measurement instead of ambulatory BP measurement. Ambulatory BP measurement provides a profile of BP away from the medical environment and shows BP behavior over a 24-hour period during the patient’s usual daily activities. However, in the context of our study, 24-hour ambulatory BP measurements were not possible, because data collection was carried out on board an icebreaker, and, in some communities, the research team stayed <24 hours. Knowing this limitation, we tried to minimize the bias caused by the stress of being in a doctor’s office, and we observed only slight intraindividual variations in SBP and DBP (1.9% and 2.4%, respectively). Furthermore, the method used for measuring BP is in agreement with recent European³⁶ and American³⁷ guidelines. Another possible limitation is that level of physical activity takes into account only leisure-time physical activity. This measure may not represent the real level of physical activity, because domestic, work-based, transportation-based and other activities were not considered. In our study, leisure-time physical activity was not a confounder in BP models. However, it is possible that a residual confounding bias attributed to physical activity other than leisure time affects the results. Another limitation of this study is that individual salt intake could not be assessed, because only a 24-hour dietary recall was used. However, taking into account information on food consumed the day before the survey, we described sodium intake by sex and groups of age. Sixty-one percent of the population had excessive sodium intake (≥2300 mg/d). The median intake of sodium was higher in men than in women, and men aged 30 to 49 years had the highest sodium intake among all sex/age groups.³⁸ Not controlling for sodium intake could result in residual confounding bias. However, we compared BP by quartile of sodium intake, and no significant differences were observed in SBP, DBP, and PP means. Also, mercury was inversely correlated with sodium intake (Spearman r=−0.11; P=0.0064). Consequently, it is unlikely that the association observed between mercury and BP is attributable to sodium intake. We also deal with the limitation of cross-sectional designs, where it is not possible to establish a cause-effect relationship. However, to minimize the bias attributed to the temporality, we excluded the hypertensive participants, which avoids the overestimation of the association between mercury and BP. Hopefully, the associations obtained in this cross-sectional study could be tested in a cohort study that will be conducted in this population in a near future. This study will allow us to determine whether mercury exposure is associated with the development of hypertension.

Strengths of our study include the minimization of the information and confounding bias, which increases the internal validity of the study. Mercury determination was carried out by the Institut national de santé publique du Québec Human Toxicology Laboratory, which is accredited International Organization for Standardization 17025 by the Standards Council of Canada. This laboratory is also an international leader in analytic toxicology applied to human and environmental studies and a reference institution for interlaboratory comparison programs in heavy metals measurements. Also, in this population, we have reported previously that mercury exposure is stable over time. For example, the Pearson r correlation coefficient was 0.74 (P<0.0001) between maternal blood and last-trimester hair concentrations,³⁹ suggesting that blood mercury is a good indicator of exposure for the last months. To minimize the confounding bias, we took into account traditional BP risk factors (age, sex, obesity [BMI, WC, and WHR], insulin sensitivity [measured as HOMA-IR], cholesterol [HDL and LDL cholesterol], and triglycerides levels, smoking habits, alcohol consumption, physical activity, and socioeconomic status). We also considered lead exposure, which was correlated with mercury and could influence BP.⁴⁰ In addition, we included n-3 fatty acids and selenium levels, because their blood levels are high among Nunavik Inuit adults,^2,41 especially considering that a protector effect of these substances on BP has been suggested.^24,25 However, the associations between mercury and BP parameters remained significant, which may suggest that the negative impact of mercury on BP may be stronger than the possible protective effect of n-3 fatty acids and selenium.

The results of the present study obtained in a highly mercury-exposed population help us to better understand the cardiovascular mercury toxicity and to assess the risks/benefits of fish consumption. This aspect is very important for Nunavik Inuit, because traditional food is an important element of their culture. Moreover, this study constitutes the baseline of a cohort study that will allow us to determine whether mercury exposure is associated with the development of hypertension. Finally, the observation that WC, in comparison with BMI, was a better predictor of SBP is of interest in light of the refinement needed in obesity assessment in clinical practice.^42,43

Even in the setting of BP within the normal range and after controlling for fish nutrients (n-3 fatty acids and selenium) and other confounding factors, mercury is associated with higher BP and PP among Nunavik Inuit adults.