From 1989 to 1991, all men and women 45–74 years of age from selected communities in Arizona and Oklahoma were invited to participate in the SHS (Lee et al. 1990
). In North and South Dakota, a cluster sampling technique was used. Of all the individuals invited, 62% agreed to participate. Participants were similar to nonparticipants in age, body mass index (BMI), and self-reported frequency of diabetes. Women were more likely to participate than men. Starting in 1998, the Strong Heart Family Study (SHFS) recruited persons who were ≥ 18 years of age and extended family members of the original SHS participants to participate in a study of the genes that contribute to cardiometabolic risk in American Indian populations (North et al. 2003
). For the SHFS, families who had at least five living siblings, including three original SHS participants, were invited; and parents, spouses, offspring, spouses of offspring, and grandchildren were enrolled to build extended pedigrees. The SHFS genotyped genome-wide STR markers in all participants.
Urine metals, including the arsenic species inorganic arsenic, MMA and DMA, were measured in 3,974 individual persons who participated in the SHS baseline visit (1989–1991) (Scheer et al. 2012
). For the present analysis, we excluded 1 participant who was missing total arsenic, and 1 participant missing inorganic arsenic concentrations. We further excluded 222 participants whose %iAs, %MMA, or %DMA were below the limit of detection. We also excluded 5 participants who were missing information on smoking, 9 participants missing information on alcohol consumption, 16 participants missing BMI measurement, 1 participant missing level-of-education information, and 5 participants missing urine creatinine, leaving a sample size of 3,714 SHS participants. Among those, 2,907 SHS participants had at least 1 relative within the cohort, allowing heritability analysis and 487 were also SHFS participants with STR marker information for the linkage analysis.
The 13 participating tribes, the Indian Health Service (IHS) Institutional Review Board (IRB), and the IRBs of the participating institutions approved the SHS and SHFS protocol and consent forms. All participants provided written and oral informed consent at enrollment into the SHS. The present heritability and linkage study was covered by the original SHS consent form because arsenic is a potential cardiovascular risk factor. In addition, the present study was specifically approved by the SHS Publications and Presentations Committee and by the participating tribes.
. Spot urine samples were collected in 1989–1991, frozen within 1–2 hr of collection, and stored at –80ºC at the Penn Medical Laboratory, MedStar Health Research Institute (Hyattsville, MD, and Washington, DC, USA) (Lee et al. 1990
). In 2009, up to 1.0 mL of urine from each participant was transported on dry ice to the Trace Element Laboratory of the Institute of Chemistry–Analytical Chemistry, Karl-Franzens University (Graz, Austria). There, total arsenic concentrations in the urine samples were measured by inductively coupled plasma mass spectrometry (ICPMS) (Agilent model 7700× ICPMS; Agilent Technologies, Waldbronn, Germany), and arsenic species were determined by high performance liquid chromatography (HPLC; Agilent model 1100) coupled to ICPMS and served as the arsenic selective detector (HPLC/ICPMS). The analytical methods used to determine urine arsenic concentrations have been described in detail (Scheer et al. 2012
). The limits of detection were 0.2 µg/L for total arsenic, and 0.1 µg/L for iAs, MMA, DMA, and for arsenobetaine plus other cations. Participants with iAs, MMA, and DMA below the limits of detection (5.3% for iAs, 0.7% for MMA, 0.03% for DMA) were excluded from this analysis because it is not possible to evaluate arsenic metabolism with undetectable urine arsenic biomarkers. An in-house reference urine and the NIES No. 18 Human Urine from the National Institute for Environmental Studies (Ibaraki, Japan) were analyzed together with the samples. The interassay coefficients of variation for the in-house reference urine for total arsenic, iAs, MMA, DMA, and arsenobetaine plus other cations were 4.4%, 6.0%, 6.5%, 5.9%, and 6.5%, respectively.
Urine arsenic data were transmitted to the Texas Biomedical Research Institute (previously known as Southwest Foundation for Biomedical Research) where they were transferred to the pedigree data management system PEDSYS (Dyke 1999
Demographic and lifestyle assessment
. Baseline sociodemographic, lifestyle, and anthropometric information was obtained through interview and physical examination conducted by trained nurses and medical assistants (Lee et al. 1990
). The standardized in-person questionnaire included sociodemographic data (age, sex, education) and smoking status (never, current, former). Body mass index (BMI) was estimated by dividing measured weight in kilograms by measured height in meters squared.
Short tandem repeat markers
. DNA from white cells was isolated and stored at the Texas Biomedical Research Institute. We genotyped nearly 400 STR markers, spaced, on average 10 cM (centimorgans) apart (range, 2.4–24.1 cM) [For details, see Supplemental Material, Table S3 (http://dx.doi.org/10.1289/ehp.1205305
)], using ABI PRISM Linkage Mapping Set-MD10 version 2.5 (Applied Biosystems, Foster City, CA, USA). Individual polymerase chain reaction (PCR) products were loaded into an ABI PRISM 377 Genetic Analyzer for laser-based automated genotyping. Genotypes were assigned using the Genotyper software system (Applied Biosystems). Pedigree relationships were verified and likely genotyping errors were detected using PREST software version 3.02 (Pedigree Relationship Statistical Tests, http://utstat.toronto.edu/sun/Software/Prest/prest3.02/
) (McPeek and Sun 2000
; Sun et al. 2002
) and SimWalk2 (Sobel et al. 2002
). Mendelian inconsistencies and unlikely double recombinants in marker genotypes were removed with an overall blanking rate of < 1% in the total study population. The average heterozygosity was 0.69 for Arizona, 0.74 for Oklahoma, and 0.76 for North and South Dakota.
Statistical analysis. Descriptive analysis. To evaluate urine arsenic methylation and excretion patterns, we computed the proportions of iAs, MMA, and DMA by dividing the concentration of each species by the sum of all three species and multiplying by 100, yielding %iAs, %MMA, and %DMA. The median [interquartile range (IQR)] percentages of urine arsenic species were reported for the overall population and according to age (≤ 55 years, > 55 years), sex (men, women), education (< 12 years completed, ≥ 12 years completed), BMI (< 30 kg/m2, ≥ 30 kg/m2), smoking (never, former, current), alcohol consumption (never, former, current), and study region (Arizona, Oklahoma, and North and South Dakota).
Heritability. We estimated the heritability of %iAs, %MMA, and %DMA using a general pedigree variance components decomposition-based method as conducted by the software Sequential Oligogenic Linkage Analysis Routines (SOLAR) version 4.4.0 (Almasy et al. 2012
). For a detailed description of the statistical methods used to estimate heritability, see Supplemental Material, pp. 2–3 (http://dx.doi.org/10.1289/ehp.1205305
). In brief, SOLAR incorporates the information contained in participant pedigrees to obtain maximum likelihood estimates for the proportion of unexplained variance due to additive genetic effects from polygenes (σ2g
) and the proportion of variance due to unmeasured environmental covariates, measurement error, and nonadditive genetic effects (σ2e
). Heritability (h2
) is defined as the proportion of unexplained variance in the observed distribution of the percent of each urine arsenic species that is attributable to additive genetic effects, or h2
). The p
-values for h2
are computed from a likelihood ratio test comparing the model in which the h2
component of the unexplained variance is estimated to a model in which h2
is constrained to be zero, following a 1/2:1/2 mixture of chi-square distributions with 1 degree of freedom (df) and a point mass at zero (Amos 1994
Initially, the percentage of each urine arsenic species was converted to an odds and introduced in a linear variance component model as a logit-transformed dependent variable adjusted for age, sex, age2, age × sex and age2 × sex, education (< 12 years of education, ≥ 12 years of education), BMI (< 30 kg/m2, ≥ 30 kg/m2), smoking status (never, former, current smokers), alcohol drinking status (never, former, current drinkers), region (Arizona, Oklahoma, and North and South Dakota), and total urine arsenic concentrations. To reduce kurtosis, residuals from the linear variance component model were transformed using an inverse Gaussian transformation and introduced as dependent variables in a second-stage variance component linear regression with no covariables. A household component of the variance was explored to account for “shared environment” among relatives living in the same household, but household component was not retained in the final model because it did not influence the heritability estimates (data not shown). We performed heritability analysis for the population as a whole, and stratified by region.
Linkage scan. We implemented a multipoint QTL linkage analysis based on variance components decomposition-based methods in SOLAR version 4.4.0 (Almasy and Blangero 1998
; Almasy et al. 2012
), building on the heritability models described above by including additional variance components for QTLs based on multipoint identity-by-descent (IBD) matrices [for details on model formulation, see Supplemental Material, pp, 2–3 (http://dx.doi.org/10.1289/ehp.1205305
)]. We computed multipoint IBD matrices using the software Loki (Heath 1997
; Heath et al. 1997
) by using STR marker map positions obtained based on the DeCode map (NHLBI 2012
). At each chromosomal location, SOLAR conducts likelihood ratio tests comparing a model that estimates the unexplained variance attributable to a potential QTL versus a model that constrains the unexplained variance attributable to the potential QTL to be equal to zero. The tests at each chromosomal location are reported as LOD [logarithm (to the base of 10) of the odds] scores in favor of genetic linkage with a QTL. LOD scores of 1.9 and 3.3 are considered suggestive and significant evidence, respectively, for a QTL (Lander and Kruglyak 1995
). To evaluate whether the linkage findings may have been due to chance, we conducted an adjustment analysis in order to calculate an empirical LOD score for each urine arsenic metabolite. The empirical LOD scores were computed by multiplying the original LOD scores with a correction constant. The correction constant is calculated when a fully informative marker, not associated with the phenotype, is simulated and goes through approximately 10,000 replicates, IBDs are calculated for this marker and the LOD score is computed for linkage of this phenotype to this marker. In this analysis, if the original LOD score is similar to the adjusted LOD score, then the possibility of findings being due to chance alone is low.
Sensitivity analyses. Several sensitivity analyses were conducted for both the heritability and the linkage scan. First, we adjusted for BMI as continuous rather than categorical. Second, we added an interaction term for BMI (categorical) × sex. Third, we further adjusted for urine selenium. Fourth, we accounted for urine dilution by adjusting for urine creatinine rather than by dividing by urine creatinine. Last, we repeated the analyses without accounting for urine dilution. The heritability and linkage scan analyses remained unchanged (data not shown).