Mercury Exposure and Antinuclear Antibodies Among Females of Reproductive Age in the United States: NHANES

Emily C. Somers; Martha A. Ganser; Jeffrey S. Warren; Niladri Basu; Lu Wang; Suzanna M. Zick; Sung Kyun Park

Disclosures

Environ Health Perspect. 2015;123(8):792-798. 

In This Article

Methods

Study Population

NHANES is conducted by the Centers for Disease Control and Prevention (CDC), National Center for Health Statistics (NCHS) (CDC/NCHS 2015b). It uses a stratified, multistage probability cluster design, with oversampling of selected subpopulations, to obtain a representative sample of the civilian, noninstitutionalized U.S. population. NHANES protocols were approved by the NCHS Institutional Review Board, and informed consent was obtained. In our study we used data from three cycles of continuous NHANES data (1999–2004) (CDC/NCHS 2015a). Participation rates were 76% for 1999–2000, 80% for 2001–2002, and 76% for 2003–2004 (CDC/NCHS 2013). The eligible population for our analysis included female participants 16–49 years of age who completed a physical examination with biospecimen collection for ANA and mercury assessment.

From a total of 5,984 females 16–49 years of age in NHANES 1999–2004, 1,932 were included in the one-third subsample with ANA assessment, of whom 1,354 had available ANA data. Hair mercury was available for one cycle (1999–2000), total blood mercury for three (1999–2004), and urinary mercury for two (1999–2002). For hair, blood, and urinary mercury, respectively, samples sizes were 452, 1,352, and 804 (after excluding 16, 2, and 29 persons with missing data).

ANAs

As detailed elsewhere (CDC/NCHS 2012), standard methodology for ANA screening was used, involving indirect immunofluorescence with HEp-2 substrate for detection of IgG antibodies to cellular antigens. Titers to which fluorescence remained positive (serial dilution range, 1:80–1:1,280) and staining patterns were determined for positive specimens. ANA patterns refer to indirect immunofluorescence patterns (e.g., speckled, nucleolar, homogeneous) reflecting the anatomic distribution of intracellular antigens, and thus different nuclear components. A variety of different ANAs can give rise to a given pattern. Follow-up immunoprecipitation was used for identification of specific antigens from a standard panel.

Mercury Exposure Assessment

Three types of biomarkers for mercury exposure were used: hair (organic), total blood (organic and inorganic), and urine (predominantly inorganic/elemental). One-centimeter hair segments were utilized (approximating exposure during the preceding 2.5 months). Standard methodology for mercury measurement was used, as described elsewhere (CDC/NCHS 2005, 2007). In brief, for hair, cold vapor atomic fluorescence spectrometry following analyte extraction was used. For blood and urine, flow-injection cold vapor atomic absorption spectrometry (PerkinElmer Flow Injection Mercury System-400) was used in NHANES 1999–2002, and inductively coupled plasma mass spectrometry (ICP-MS; PerkinElmer ELAN 6100) was used in 2003–2004. Limits of detection (LODs) for hair mercury varied by batch and ranged between 0.011 and 0.027 ppm (method detection limit); 6% of the females in our study had hair mercury levels below the LOD. LODs for total blood mercury varied according to cycle and batch, ranging between 0.14 and 0.2 μg/L. Of the females in our study, 7.4%, 6.5%, and 7.5% had total blood mercury levels below the LOD for the three cycles, respectively. We did not separately investigate the inorganic fraction of blood mercury because of the large proportion < LOD (97.4%, 95.1%, and 77.2% for the three cycles). The LOD for urinary mercury was 0.14 μg/L, with 13.3% and 14.3% of the participants having urinary mercury levels < LOD for the two cycles, respectively.

Other Variables

Sociodemographic data were collected by self-administered questionnaires. Body mass index (BMI), calculated as kilograms of body weight divided by height in meters squared, was included due to the role of obesity in chronic inflammation. Serum cotinine (nanograms per milliliter), a marker of active and passive smoking, was measured by isotope dilution–high-performance liquid chromatography/atmospheric pressure chemical ionization tandem mass spectrometry; tobacco exposure has been linked to increased risk of autoimmune diseases (Costenbader and Karlson 2006). C-reactive protein (CRP), a nonspecific inflammatory marker, was quantified (milligrams per deciliter) by latex-enhanced nephelometry. Nutrient data were estimated based on a multiple pass, computer-assisted dietary interview of food and beverage consumption, with recall assessment of individual foods consumed in the previous 24 hr. We derived data on selenium (micrograms), eicosapentaenoic acid (20:5 n-3), and docosahexaenoic acid (22:6 n-3), all found in seafood; omega-3 fatty acid intake was calculated as eicosapentaenoic acid plus docosahexaenoic acid. Selenium potentially mitigates effects of mercury (Cuvin-Aralar and Furness 1991), and omega-3s have anti-inflammatory effects (Simopoulos 2002). Among participants who underwent the dietary interview, weekly seafood intake was estimated based on recall of fish/shellfish consumption in the previous 30 days. Serum polychlorinated biphenyls (PCBs) were measured by high-resolution gas chromatography/isotope dilution high-resolution mass spectrometry (CDC 2006). A summary measure for coplanar (dioxin-like) polychlorinated biphenyls (cPCBs), which included congeners with suspected immunotoxicity (Wolff et al. 1997), was calculated as the sum of the products of the concentration of each serum lipid–adjusted congener (PCBs 81, 105, 118, 126, 156, 157, 167, 169) and its corresponding 2005 World Health Organization-defined toxic equivalency factor (TEF) (Van den Berg et al. 2006). An alternate PCB summary measure was the sum of the lipid-adjusted values for the four most prevalent PCB congeners (118, 138, 153, 180) (Laden et al. 2010); three of these are noncoplanar and without defined TEFs to take into account. To address the potential for drug-induced autoimmunity, we assessed use within the past month of four prescription medications that have been implicated in this phenomenon (procainamide, hydralazine, carbamazepine, and minocycline) (Schoonen et al. 2010).

Statistical Analysis

To account for the complex, stratified, and multistage cluster sampling design, analyses were conducted using the survey packages of Stata (v.12; StataCorp) and R (v.2.11.1; R Foundation) to obtain appropriate estimates and standard errors. ANAs were measured in a one-third subsample, and we constructed and applied weights to our subsample according to NCHS analytic guidelines (Johnson et al. 2013). Values below the LOD for laboratory assays were handled as the LOD divided by the square root of 2. Hair and blood mercury were log-transformed due to their skewed nature, or handled as quantiles based on distributions in the study population. Two-sample t-tests for survey data and the Rao-Scott chi-square test were used for continuous and categorical data, respectively. p-Values < 0.05 were considered significant. Crude models included mercury as the independent variable; separate models were performed for each source of mercury (hair, blood, urine). Multivariable logistic regression was utilized to estimate adjusted odds ratios (ORs) for ANA positivity in association with mercury exposure. Model A included age, race/ethnicity, education, serum cotinine, and selenium; an indicator for the NHANES cycle was included in models combining data across cycles to account for potential methodological differences between cycles. Models B and C were further adjusted for omega-3 fatty acids and seafood intake, respectively, which have been suggested to negatively confound health effects of mercury (Budtz-Jørgensen et al. 2007; Guallar et al. 2002). Multivariable urinary mercury models adjusted for urinary creatinine to account for dilution of spot urine specimens. We performed sensitivity analyses adding BMI and CRP in Models A–C as potential confounders for all mercury types. Separate sensitivity analyses were performed, including the coplanar and prevalent PCB measures. For urinary mercury, we also conducted models excluding persons with impaired renal function [glomerular filtration rate (GFR) < 60 mL/min/1.73 m2] to account for potential reverse causation whereby chronic kidney disease may increase urinary mercury excretion. Piecewise continuous models were constructed, and linearity with the log-odds of ANA was examined by predicted probability plots with natural cubic splines with four degrees of freedom (three for hair). Multinomial logistic regression was utilized to examine ANA titer strength as the outcome, with negative ANA (< 1:80) as the base outcome, and low/moderate titer (1:80–1:640) and high titer (≥ 1:1,280) as the other outcome levels.

processing....