Association between body mass index and arsenic methylation in three studies of Bangladeshi adults and adolescents

Abstract

Background.

Water-borne arsenic (As) exposure is a global health problem. Once ingested, inorganic As (iAs) is methylated to mono-methyl (MMA) and dimethyl (DMA) arsenicals via one-carbon metabolism (OCM). People with higher relative percentage of MMA (MMA%) in urine (inefficient As methylation), have been shown to have a higher risk of cardiovascular disease and several cancers but appear to have a lower risk of diabetes and obesity in populations from the US, Mexico, and Taiwan. It is unknown if this opposite pattern with obesity is present in Bangladesh, a country with lower adiposity and higher As exposure in drinking water.

Objective.

To characterize the association between body mass index (BMI) and As methylation in Bangladeshi adults and adolescents participating in the Folic Acid and Creatine Trial (FACT); Folate and Oxidative Stress (FOX) study; and Metals, Arsenic, and Nutrition in Adolescents Study (MANAS).

Methods.

Arsenic species (iAs, MMA, DMA) were measured in urine and blood. Height and weight were measured to calculate BMI. The associations between concurrent BMI with urine and blood As species were analyzed using linear regression models, adjusting for nutrients involved in OCM such as choline. In FACT, we also evaluated the prospective association between weight change and As species.

Results.

Mean BMIs were 19.2/20.4, 19.8/21.0, and 17.7/18.7 kg/m² in males/females in FACT, FOX, and MANAS, respectively. BMI was associated with As species in female but not in male participants. In females, after adjustment for total urine As, age, and plasma folate, the adjusted mean differences (95% confidence) in urinary MMA% and DMA% for a 5 kg/m² difference in BMI were −1.21 (−1.96, −0.45) and 2.47 (1.13, 3.81), respectively in FACT, −0.66 (−1.56, 0.25) and 1.43 (−0.23, 3.09) in FOX, and −0.59 (−1.19, 0.02) and 1.58 (−0.15, 3.30) in MANAS. The associations were attenuated after adjustment for choline. Similar associations were observed with blood As species. In FACT, a 1-kg of weight increase over 2 to 10 (mean 5.4) years in males/females was prospectively associated with mean DMA% that was 0.16%/0.19% higher.

Discussion.

BMI was negatively associated with MMA% and positively associated with %DMA in females but not males in Bangladesh; associations were attenuated after plasma choline adjustment. These findings may be related to the role of body fat on estrogen levels that can influence one-carbon metabolism, e.g. by increasing choline synthesis. Research is needed to determine whether the associations between BMI and As species are causal and their influence on As-related health outcomes.

Background

Over 140 million individuals globally are exposed to water arsenic (As) above the World Health Organization (WHO) safety standard of 10 μg/L.¹ In Bangladesh, 19 million people are exposed to water As above 50 μg/L, the country safety standard.² This is concerning as inorganic As (iAs), the form of As in drinking water, is a carcinogen linked to cancers of the lung, bladder, and skin. Arsenic can also lead to adverse cardiovascular, neurological, pulmonary, and other health-related outcomes.^3–5

After ingestion, the methylation of iAs leads to the formation of monomethyl (MMA) and dimethyl (DMA) arsenicals, in a process that facilitates urinary As (uAs) elimination and reduces As body burden. The majority of studies use urine As (uAs) instead of blood As (bAs) as biomarkers of exposure because blood levels are much lower than in urine and analysis of bAs requires more sensitive methods. Additionally, because DMA is rapidly excreted from blood to urine, it is difficult to capture the full extent of As methylation in blood. However, bAs may more closely reflect tissue As exposure compared to uAs.⁶ Of the As species, MMA in its trivalent form is believed to be the most toxic. In epidemiological studies, the partial methylation of iAs that results in higher MMA% in urine has been associated with a greater risk for cardiovascular disease (CVD), skin lesions, and cancers of the bladder, lung, and skin.^7–13 Conversely, lower MMA% and higher DMA% in urine has been associated with increased risk of diabetes-related outcomes in populations from the US, Mexico, and Taiwan.^14–17 Several previous studies have found that higher body mass index (BMI) is associated with lower MMA% and higher DMA% in urine in populations from the US, Mexico, and Central Europe.^18–20 The association of As methylation with BMI, however, has not been investigated in populations from rural Bangladesh, where adiposity levels tend to be lower compared to other countries. Understanding factors that influence individuals’ As methylation capacities across diverse populations are important for understanding As-related health outcomes. Known factors that reduce As methylation efficiency include smoking, older age, betel nut use, alcohol consumption (not relevant for Bangladeshi populations as they rarely consume alcohol),²¹ and folate deficiency.²²

Arsenic methylation is facilitated by one-carbon metabolism (OCM, Figure 1), a biochemical pathway that produces the methyl donor S-adenosylmethionine (SAM). Methyltransferases require SAM to transfer methyl (−CH₃) groups to numerous substrates including As.²³ OCM is heavily dependent on folate and choline. Choline can be converted to betaine (a.k.a. trimethylglycine), which can donate a methyl group to various substrates including As, forming the methylated product and dimethylglycine (DMG). When folate levels are low, choline and betaine become alternative sources of methyl groups for OCM.²⁴

Figure 1.

One-carbon metabolism (OCM) and nutrients relevant for this study.

The objective of this study was to characterize the cross-sectional association between BMI and As methylation in Bangladeshi adults and adolescents, a population characterized by lower nutritional status and lower levels of adiposity, particularly in males, compared to previous populations in the US and Mexico that have evaluated this relationship. Given the importance of OCM nutrient status (measured as plasma folate, choline, betaine and DMG levels) in As methylation, we further evaluated whether the relationship between BMI and As methylation is influenced by OCM nutrients. We hypothesize that higher BMI is associated with higher As methylation efficiency, as indicated by lower MMA% and higher DMA% in both urine and blood, and that OCM nutrient status influences this association such that lower folate and other OCM-related nutrients will be associated with lower As methylation efficiency and lower BMI. In two studies with repeated weight measures available (FACT and FOX), we also evaluated the prospective association between individual-level weight change and As methylation.

Methods

Study Populations

This study leveraged data and samples previously collected from participants in three studies conducted in Araihazar, Bangladesh including the Folic Acid and Creatine Trial (FACT); Folate and Oxidative Stress (FOX) study; and Metals, Arsenic, and Nutrition in Adolescents Study (MANAS). As shown in Figure 2, these three studies recruited participants from the Health Effects of Arsenic Longitudinal Study (HEALS), a prospective cohort study which recruited and has followed 12,000 participants since 2000.²⁵

Figure 2.

Participants recruited from HEALS for the three Bangladeshi studies analyzed with corresponding sample sizes. Complete metabolite data refers to participants that had measures for all predictor and outcome variables used in analyses.

Full descriptions of these studies have previously been reported.^25–28 Briefly, all study participants resided in Araihazar, a region of Bangladesh 25-km² southeast of Dhaka, where well water concentrations of naturally occurring As widely varied. The Columbia University Institutional Review Board and the Ethical Committee of the Bangladesh Medical Research Council approved all study protocols. The eligibility criteria for HEALS participants were: (1) married adults aged 20–65 years old, (2) area residents for at least 5 years, and (3) drinking water primarily from the same household well for at least 3 years. Starting in 2009, participants in FACT were randomly selected from HEALS participants who had been drinking from a household well for at least 1 year with water As > 50 μg/L. In 2008, FOX participants were selected from HEALS participants based on well-water As concentrations <10 μg/L (n = 76), 10–100 μg/L (n = 104), 101–200 μg/L (n = 86), 201–300 μg/L (n = 67), and >300 μg/L (n = 45). MANAS participants, enrolled between 2012 and 2016, included a random selection of 14–16-year old adolescents whose mothers were HEALS study participants. Across these three studies, exclusion criteria were: 1) women who were pregnant, 2) participants taking nutritional supplements, and 3) participants with known diabetes, cardiovascular, or renal disease, or other known health issues. All three studies had a wide range of water As exposures with BMI and As species measures, allowing us to study the cross-sectional relationship between BMI and As methylation.

Demographic Characteristics.

Age, sex, history of smoking and betel nut use, and sociodemographic indices such as education and land ownership were acquired during baseline interviews.

Arsenic Measurements.

Tube well water As

Tube well water As concentrations were measured at Columbia University’s Lamont Doherty Earth Observatory using graphite furnace atomic absorption (GFAA). The limit of detection (LOD) was 5 μg/L, and FOX samples with concentrations below the LOD were reanalyzed using inductively coupled plasma-mass spectrometry (ICP-MS) as described.²⁹ In FACT and MANAS, no samples were below the LOD.

Total Urinary As

Total Urinary As was measured in the Columbia University Trace Metals Core Laboratory using GFAA with a Perkin-Elmer Analyst 600 graphite furnace system as previously described.³⁰ The LOD for uAs was 2 μg/L. None of the samples were below the LOD.

Total Blood As.

FACT and FOX venous blood samples collected in EDTA vacutainer tubes at baseline were used for bAs analyses using a PerkinElmer Elan DRC II ICP-MS with an AS93+ autosampler.⁶ In MANAS, blood As was not measured.

Urinary and Blood As Species.

As described previously, arsenite (As^III), arsenate (As^V), MMA, and DMA were measured in blood and urine using HPLC coupled to dynamic reaction cell inductively coupled plasma MS.³¹ This method cannot differentiate between reduced and oxidized forms of MMA and DMA. While this chromatography can separate As^V from As^III, samples can oxidize during sample processing, so these are added and reported as a single variable reflecting total iAs. All inter- and intra- assays were well below 5% for most species measured in the three studies with some exceptions (Table A.1).

OCM and Other Metabolite Measurements.

FACT and FOX venous blood samples collected at enrollment were used for OCM measures (plasma folate, choline, betaine, and DMG). Plasma folate concentrations were analyzed using radioprotein binding assays (MP Biomedicals) as described by Hall et al. 2013.³² Plasma betaine, choline, and DMG concentrations for FACT and FOX studies were analyzed in the laboratory of Marie Caudill at Cornell University as previously described³³ using Liquid Chromatography Tandem Mass Spectrometry (LC–tandem MS) following previously described methods³⁴ and modifications.³³ Plasma choline, betaine, and DMG were not measured in MANAS.

BMI Categories.

BMI was computed based on weight and height measured by the interviewers following a standardized protocol. For FACT and FOX, height was measured at the HEALS baseline visit in 2000–2008, while weight was measured at the HEALS baseline as well as at the FACT (2009–2010) and FOX (2008) baselines. For MANAS, height and weight were measured at the same baseline visit in 2012–2016. For adults, BMI (weight in kg / height in m²) was categorized based on WHO international standards, with underweight: BMI < 18.5 kg/m², normal weight: 18.5 kg/m² ≤ BMI < 25.0 kg/m², overweight: BMI ≥ 25 kg/m², and obese: BMI ≥ 30 kg/m².³⁵ For adolescents, BMI was categorized based on WHO growth curves for children for males and females separately. As the BMI cutoffs for the normal and overweight categories corresponding to the 5^th and 85^th percentiles, respectively³⁶, were similar for those aged 14–16 years (the ages of the study participants in MANAS), we used the percentile cutoffs for those aged 15 years and 6 months old.³⁷ Due to small numbers of participants considered obese in all three studies, overweight and obese categories were combined.

Lean Body Mass.

Lean body mass (LBM) was computed for males and females separately using the Boer formulas based on weight in kilograms and height in centimeters as follows:

$$\text{For\hspace{0.17em}men}:\text{LBM}=\left(0.407\times \text{weight}\right)+\left(0.267\times \text{height}\right)-19.2$$

$${\text{For\hspace{0.17em}woman}:\text{LBM}=\left(0.252\times \text{weight}\right)+\left(0.473\times \text{height}\right)-\mathrm{48.3.}}^{38}$$

Statistical Analyses.

All variables were assessed for normality. Statistical analyses were conducted using R software (version 1.1.383; R Project for Statistical Computing). Percentages of iAs, MMA, and DMA, (iAs%, MMA%, DMA%) were used to estimate As methylation levels, and the percentage values were calculated by dividing each species concentration by the sum of the three species and multiplying the corresponding value by 100.^39,40 Two methylation indices, the Primary Methylation Index (PMI), calculated as MMA/iAs, and the Secondary Methylation Index (SMI), calculated as DMA/MMA, were calculated to assess As methylation efficiency. All analyses were stratified by sex, as sex-related differences in As methylation are known.^20,41,42 The main analyses were cross-sectional based on concurrent BMI and arsenic species measured at the same visit. Welch’s two-sample t-tests were used to compare demographic characteristics, As measurements, and concurrent BMI between male and female participants. Correlation matrices with Pearson correlation coefficients were used to estimate univariate correlations between total As, iAs%, MMA%, DMA%, plasma OCM nutrients, and BMI (Figures in Appendix (A), A.1 & A.1.2). Additionally, given that As species sum to 100%, we graphically described the distribution of As species by concurrent BMI category using a triplot (a diagram with three axes) and compared the compositional means of iAs%, MMA%, and DMA% in female and male participants separately (Figures A.2, A.2.1, & A.2.2).

It was not feasible to pool the participants from all three cohorts due to the differences in study design and recruitment. BMI was used as a predictor of As methylation efficiency in the main linear regression models to estimate the mean differences in iAs%, MMA%, DMA%, PMI, and SMI for each cohort separately; BMI was measured at the same visit as the As species. In FOX and FACT, analyses were run for both bAs and uAs species. In MANAS, analyses were only conducted for uAs species. Models were adjusted for log-transformed total As (uAs for urine analyses and bAs for blood analyses), age, and log-transformed plasma folate. Folate is an important OCM nutrient for As methylation and although it did not influence the coefficients compared to the unadjusted models, it was included to assess the impact of other OCM-related nutrients versus folate-adjusted models.

To evaluate the role OCM metabolites play in the association between As methylation and BMI, in addition to plasma folate (log-transformed), we adjusted linear regression models for plasma betaine/DMG ratio, and plasma choline.

Given that the sum of iAs%, MMA% and DMA% equals 100%, interpretation of the individual species is complicated in that increases in the level of one species, could reflect a decrease of one or both of the other two species. Therefore, in sensitivity analyses, linear models were used to estimate the mean difference in BMI levels by As species, with adjustments for log-transformed total As, age and log-transformed plasma folate at baseline; and further adjustments for baseline plasma betaine/DMG ratio or baseline plasma choline levels in separate models. The leave-one-out (LOO) method has been used in the As methylation literature to address this issue by including two species in the model at a time and leaving the third out by fixing, or holding it constant in the model.¹⁵ In this way, for example, if we include MMA% and DMA% in the model, and we fix changes in iAs%, a positive coefficient observed for DMA% would reflect a corresponding decrease in MMA%. Therefore, in addition to models entering each species individually, we also implemented leave-one-out (LOO) models.

Further sensitivity analyses used LBM instead of BMI as the predictor in linear models adjusted for log-transformed total uAs, age, and log-transformed plasma folate, with additional adjustment for plasma betaine/DMG ratio and plasma choline in separate models. In additional sensitivity analyses (data not shown), models were further adjusted for water As exposure, smoking status, betel nut use, and sociodemographic characteristics (education and land ownership).

For FACT and FOX, which had two weight measures (at the HEALS baseline and at the FACT or FOX baseline), we also described individual-level changes in weight between the two visits. We analyzed the association between the change in weight (modeled as continuous and as categorical) with the change in iAs%, MMA%, and DMA%, modeled as the dependent variables in separate models.

Results

Participant Characteristics.

By design, roughly half of the study participants were female (Table 1). For the MANAS participants, water As levels were derived from the participants’ mothers at the HEALS baseline visit. Mean water As was above 100 μg/L in FACT and FOX; in MANAS mean (standard deviation (SD)) water As levels were 79 (±106) μL in males and 68 (±99) μL in females. PMI was significantly higher in males compared to females, with corresponding means (SD) of 1.13 (±0.40) and 0.875 (±0.38) in FACT; 0.97 (±0.35) and 0.73 (±0.34) in FOX; and 0.94 (±0.32) and 0.81 (±0.31) in MANAS. Conversely, SMI was significantly lower in males compared to females, with corresponding means (SD) of 5.33 (±2.08) and 7.76 (±3.46) in FACT; 4.77 (±2.03) and 6.86 (±3.31) in FOX; and 6.91 (±2.54) and 7.96 (±3.58) in MANAS. The mean concurrent BMI (SD) (kg/m²) for males and females respectively were 19.2 (±2.3) and 20.4 (±3.0) in FACT, 19.8 (±3.2) and 21.0 (±3.6) in FOX, and 17.7 (±2.6) and 18.7 (±3.0) in MANAS. The FOX study had the widest range of female BMIs (from 13.9 to 35.3 kg/m²). The mean (±SD) weight change (kg) between the HEALS visit and the FACT/FOX visit was 1.03 (2.72) and 1.62 (4.13) for males and females in FACT and 0.27 (1.90) and 0.77 (2.57) for males and females in FOX.

Table 1.

Participant characteristics by sex in all three Bangladeshi studies.

	FACT		FOX		MANAS
Characteristic	Males Mean (SD)	Females Mean (SD)	Males Mean (SD)	Females Mean (SD)	Males Mean (SD)	Females Mean (SD)
Urine Arsenic Data	N = 527		N = 342		N = 708
Sex%	52.2	47.8	48.9	51.1	46.3	53.7
Concurrent BMI (kg/m2)	19.2 (2.30)	20.4 (2.98)∔	19.8 (3.16)	21.0 (3.58)∔	17.7 (2.61)	18.7 (2.95)∔
Weight Change (kg)a	1.03 (2.72)	1.62 (4.13)	0.27 (1.90)	0.77 (2.57)*	--	--
Water Arsenic (μg/L)	147 (118)	160 (132)	144 (136)	137 (113)	79.1 (106)	67.7 (99.4)
Age (yr)	40.3 (8.32)	36.6 (7.26)∔	44.9 (8.79)	41.9 (7.67)∔	14.6 (0.65)	14.6 (0.66)
Smoking (% have ever smoked)	53.0	1.00∔	68.0	6.00∔	--	--
Education (% < 1 year)	43.0	41.0	32.0	47.0∔	3.00	1.00∔
Plasma Folate (nM/L)	17.1 (15.7)	16.5 (12.5)	11.5 (6.69)	14.4 (7.66)∔	15.8 (9.53)	15.0 (7.00)
Plasma Choline (nM/mL)	12.4 (2.68)	11.2 (2.45)∔	12.2 (3.38)	11.0 (3.20)*	-	-
Plasma Betaine (nM/mL)	53.2 (16.0)	37.9 (13.5)∔	53.3 (17.0)	41.7 (18.6)∔	-	-
Plasma DMG (nM/mL)	7.90 (7.30)	6.97 (6.73)	6.39 (14.5)	4.97 (4.06)	-	-
Total Urinary Arsenic (μg/L)	147 (150)	155 (130)	213 (191)	221 (263)	76.7 (108)	68.9 (115)
Urinary Arsenic Species
iAs%	14.0 (4.2)	13.9 (4.7)	17.2 (4.4)	17.9 (6.2)	13.7 (5.33)	15.2 (9.98)*
MMA%	14.9 (4.4)	11.1 (3.7)∔	16.0 (5.0)	12.0 (4.3)∔	11.9 (3.43)	10.6 (3.53)∔
DMA%	71.1 (6.7)	75.1 (6.6)∔	66.8 (7.3)	70.1 (8.0)∔	74.4 (6.87)	74.2 (10.1)
PMI	1.13 (0.40)	0.875 (0.38)∔	0.97 (0.35)	0.73 (0.34)∔	0.94 (0.32)	0.81 (0.31)∔
SMI	5.33 (2.08)	7.76 (3.46)∔	4.77 (2.03)	6.86 (3.31)∔	6.91 (2.54)	7.96 (3.58)∔
Blood Arsenic Data	N = 593		N = 293		-
Sex%	50.9	49.1	50.2	49.8	-	-
Total Blood Arsenic (μg/L)	10.5 (6.4)	9.4 (7.4)	17.6 (10.4)	14.6 (7.5)*	-	-
Blood Arsenic Species					-	-
iAs%	27.0 (3.6)	27.2 (4.2)	29.5 (4.2)	29.8 (4.4)	-	-
MMA%	44.8 (4.3)	43.9 (4.5)*	40.5 (5.1)	37.6 (5.7)∔	-	-
DMA%	28.1 (5.2)	28.9 (5.0)	30.0 (5.1)	32.7 (5.7)∔	-	-

Two-sample t-test was used to compare characteristics between males and females within the same study.

^*Two-sample t-test with p-value < 0.05

^∔Two-sample t-test with p-value < 0.001

^-Study did not measure characteristic

^aWeight change was calculated using an earlier measure of weight obtained from participants when they were recruited in HEALS

In both the FACT and FOX studies, females had lower choline, betaine, and DMG levels than males. Females also had lower folate levels in FACT and MANAS but they were higher in FOX with mean (SD) 14.3 (7.50) nM/L in females and 11.3 (6.56) nM/L in males. In FACT participants, folate levels were lower and choline levels were higher among participants with increasing BMI categories (Table A.2.1).

In FACT, plasma choline was positively correlated with total uAs among females (r = 0.17; p < 0.01) but was null for males (r = 0.02; p = 0.71). Similarly, plasma choline was positively correlated with total bAs for females (r = 0.16; p < 0.01) but not males (r = 0.03; p = 0.59) (Figure A.1 and A.1.2). Plasma choline was negatively associated with uMMA% (r = −0.27; p < 0.01;r = −0.09; p = 0.13) and positively associated with uDMA% (r = 0.15; p = 0.02; r = 0.09; p = 0.15) among females and males, respectively, although these associations were only significant females. In blood, plasma choline was negatively associated with bInAs% (r = −0.14; p = 0.02; r = −0.04; p = 0.52) and bMMA% (r = −0.20; p < 0.01; r = −0.10; p = 0.08) in females and males, respectively, but only significant in females, while it was positively and significantly associated with bDMA% (r = 0.27; p < 0.01; r = 0.11; p = 0.05) among both females and males. In contrast to FACT, among FOX participants plasma choline was negatively associated with total uAs (but was not associated with bAs) among both males and females. Similar to FACT, among FOX participants plasma choline was negatively associated with uMMA% and positively associated with uDMA%, but only achieved statistical significance among males. The patterns of association with As species in blood for male FOX participants were similar in direction to those in urine among male participants, but did not reach statistical significance.

BMI and As Methylation.

In all three studies, the correlation plots (Figure A.1a–e) show that female participants’ concurrent BMI was negatively correlated with iAs% and MMA% but positively correlated with DMA% in urine. However, this finding was not apparent in male participants in any of the three studies. For bAs species, the correlation with BMI was negative for MMA% and positive with DMA% in females from FACT (Figure A.1.2). In FOX, BMI was non-significantly positively correlated with bMMA% in females. By concurrent BMI categories, females in the underweight category had higher iAs%, higher MMA%, and lower DMA% compared to those in the obese and overweight category, with consistent patterns in FACT, FOX and MANAS (Figure 3). This is also illustrated in triplots in Figure A.2a–c. Male overweight and obese participants tended to have lower iAs%, lower MMA%, and higher DMA% than the other two categories in urine (Figure A.2.1a–c). Similar As methylation patterns were found in corresponding bAs triplots (Figure A.2.2a–d).

Figure 3.

Median (IQR) urinary inorganic arsenic (iAs) and mono-methyl (MMA) and dimethyl (DMA) arsenical percentages among female participants in the underweight, normal, and overweight and obese BMI categories in all three Bangladeshi studies.

In all three studies, female participants with higher BMI levels had lower MMA% and higher DMA% in urine after adjustment for log-transformed total As, age and log-transformed plasma folate (Table 2), although associations were only significant in FACT participants. For the FACT study, the adjusted mean difference (95% confidence) in MMA% and DMA% for a mean difference in 5-kg/m² BMI (e.g., a difference from 20 to 25 kg/m²) was −1.21 (−1.96, −0.45)% and 2.47 (1.13, 3.81)%, respectively. Results of all models were similar for bAs species in FACT participants, including the attenuation of the association between BMI and As species with choline adjustment (Figure 4). Consistently, higher BMI was positively associated with higher SMI (reflecting higher As methylation efficiency, i.e. higher DMA% and lower MMA%) in the three studies (significant in FACT and MANAS). The association between BMI and SMI was no longer significant following adjustment for plasma choline (Table 2).

Table 2.

Adjusted mean difference in urinary arsenic species (%), primary methylation index (PMI), and secondary methylation index (SMI) for a 5-kg/m² difference in BMI stratified by sex in all three Bangladeshi studies.

	FACT			FOX			MANAS
Outcome Variable	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3	Model 1
Males	N = 275			N = 167			N = 328
iAs%	0.03 (−1.04, 1.11)	0.02 (−1.06, 1.10)	0.02 (−1.07, 1.11)	0.15 (−0.89, 1.19)	0.12 (−0.94, 1.17)	0.16 (−0.89, 1.20)	−0.54 (−1.63, 0.56)
MMA%	−0.44 (−1.58, 0.70)	−0.40 (−1.54, 0.74)	−0.34 (−1.49, 0.81)	0.06 (−1.14, 1.26)	0.19 (−1.01, 1.40)	−0.04 (−1.23, 1.15)	0.06 (−0.66, 0.78)
DMA%	0.41 (−1.34, 2.16)	0.37 (−1.38, 2.13)	0.32 (−1.45, 2.08)	−0.21 (−1.90, 1.49)	−0.31 (−2.02, 1.40)	−0.12 (−1.82, 1.58)	0.48 (−0.95, 1.91)
PMI	−0.02 (−0.12, 0.08)	−0.02 (−0.12, 0.08)	−0.02 (−0.12, 0.08)	−0.02 (−0.10, 0.07)	−0.01 (−0.10, 0.08)	−0.02 (−0.11, 0.06)	0.05 (−0.02, 0.11)
SMI	0.14 (−0.39, 0.68)	0.13 (−0.41, 0.67)	0.11 (−0.43, 0.66)	−0.06 (−0.55, 0.43)	−0.09 (−0.58, 0.40)	−0.02 (−0.50, 0.46)	0.23 (−0.31, 0.77)
Females	N = 252			N = 175			N = 380
iAs%	−1.26 (−2.21, −0.31)	−1.28 (−2.23, −0.33)	−1.30 (−2.28, −0.32)	−0.77 (−2.08, 0.54)	−0.77 (−2.09, 0.55)	−0.72 (−2.13, 0.69)	−0.99 (−2.71, 0.73)
MMA%	−1.21 (−1.96, −0.45)	−1.19 (−1.94, −0.44)	−0.86 (−1.61, −0.10)	−0.66 (−1.56, 0.25)	−0.55 (−1.44, 0.34)	−0.60 (−1.57, 0.38)	−0.59 (−1.19, 0.02)
DMA%	2.47 (1.13, 3.81)	2.47 (1.13, 3.82)	2.16 (0.78, 3.53)	1.43 (−0.23, 3.09)	1.32 (−0.34, 2.98)	1.31 (−0.48, 3.10)	1.58 (−0.15, 3.30)
PMI	−0.04 (−0.12, 0.03)	−0.04 (−0.12, 0.03)	−0.01 (−0.09, 0.06)	0.00 (−0.07, 0.07)	0.01 (−0.06, 0.08)	0.01 (−0.07, 0.08)	−0.01 (−0.06, 0.05)
SMI	0.83 (0.11, 1.55)	0.82 (0.10, 1.54)	0.50 (−0.22, 1.22)	0.61 (−0.09, 1.31)	0.52 (−0.16, 1.21)	0.5 (−0.25, 1.26)	0.84 (0.23, 1.45)

Note: All models are adjusted for BMI, log-transformed sum of iAs metabolites (As^III, As^V, MMA, and DMA), age and log-transformed plasma folate at baseline. Model 2 indicates model 1 is further adjusted for plasma betaine/DMG ratio at baseline. Model 3 indicates model 1 is further adjusted for only plasma choline levels at baseline. *Bolded values have significant p-values < 0.05.

Figure 4.

Adjusted mean difference in urinary and blood arsenic species (%) for a 5- kg/m² difference in BMI in (A) FACT and (B) FOX. All models were adjusted for log-transformed total As (uAs for urine species and bAs for blood species), age and log-transformed plasma folate. Models 2 and 3 were further separately adjusted for plasma betaine/DMG and choline, respectively.

Sensitivity analyses with As methylation as the predictor of BMI resulted in similar patterns of association between BMI and As methylation profiles such that higher BMI was associated with lower iAs%, lower MMA%, and higher DMA% as compared to individuals with lower BMI (Table A.3). Furthermore, adjustment for plasma choline attenuated the association between both MMA% and DMA% with BMI. In LOO models where iAs% was fixed, these associations were consistent in FACT and FOX participants and became significant in MANAS participants. Results were also consistent in LOO models that additionally adjusted for betaine/DMG. However, the significant association between BMI and As species in FACT participants was attenuated with adjustment for plasma choline levels in LOO models when iAs% was fixed.

Additional sensitivity analyses using LBM instead of BMI as the predictor found that adjustment for choline in conventional models attenuates the association between LBM and MMA% in males and females in FACT (Table A.4).

BMI was not associated with uAs and/or bAs species in males in any of the three studies (data not shown).

Greater increases in weight between the HEALS baseline visit and the FACT enrollment visit (mean (SD) 5.4 (3.1) years) were associated with higher DMA% at the FACT visit. In continuous models adjusted for time since HEALS baseline visit (in years) and age, a 1-kg increase in weight was associated with a mean difference in 0.16 (95%CI −0.144, 0.458) % and 0.19 (−0.008, 0.395) % units for DMA% in males and females respectively. After similar adjustment, the mean difference (SD) in DMA% comparing weight change quartiles 1 (< −0.75 kg), 3 (1–3 kg), and 4 (>3 kg) to quartile 2 (−0.75–0.99 kg) were 0.50%, 0.11% and 1.70% in FACT male participants (p-trend 0.44) and −1.10%, 1.00% and 0.67% in FACT female participants (p-trend 0.08). Weight changes in FOX were small and therefore the impact in As metabolism in FOX could not be evaluated.

Discussion

In these three studies from Bangladesh, concurrent BMI was positively associated with As methylation efficiency among females but not males. These findings are consistent with previous reports linking higher BMI and other diabetes-related outcomes to lower MMA% and higher DMA%.^14–19,39 In a study in Central Europe, the findings were also consistent, although stronger in men compared to women.²⁰ Other studies have found no statistically significant association between BMI and As methylation efficiency.^41,43,44 The inconsistences across studies may be due to population differences in adiposity, small sample sizes, and/or other factors such as choline nutritional status.

In FACT and FOX, we found that associations between As methylation and BMI in females were attenuated after adjustment for plasma choline.

We found that adjustment for plasma choline attenuated the association between BMI and As methylation—supporting earlier findings that choline is an important confounder of this association. This was supported by further sensitivity analyses with LBM as the predictor, as there was a clear attenuation of the association between LBM and MMA% following adjustment for choline. These findings with LBM are also consistent with analyses conducted in the SHS that found a positive association between more precise measures of adiposity and lean body mass (e.g., percent body fat, percent lean mass, and weight circumference) and As methylation efficiency.⁴⁵

Sex-related differences in As methylation are well known^20,41,42 and may be related to sex-related differences in OCM.⁴⁶ Estrogen drives endogenous choline synthesis, as the gene encoding an important enzyme in choline synthesis, phosphatidyl ethanolamine N-methyl transferase (PEMT), has 8 estrogen response elements. For this reason, premenopausal women in most populations have higher plasma choline and lower dietary choline requirements than men. Surprisingly, our female participants had lower choline levels than males in both FACT and FOX. While the reason for this is unclear, it may have been related to recent pregnancies and/or breastfeeding, both of which have been shown to deplete maternal choline stores.^47–49 In addition to being positively correlated with As methylation efficiency, BMI was positively correlated with plasma choline concentrations in females from FACT and FOX (Figure A.1), a finding that may be driven by estrogen (Figure A.3). Collectively, these findings suggest that BMI may be driving higher estrogen levels among females,⁵⁰ leading to increased choline synthesis which increases As methylation efficiency. In addition to our cross-sectional analyses here, in a pilot study, we found that choline supplementation increased As methylation (Hall and Gamble, unpublished data).

The interpretation of the study findings may be further complicated by reported effects of As exposure on increasing choline concentrations in the plasma, liver, and kidney of rodents.^51,52 While the mechanism underlying this is unclear, it may be through estrogenic effects of As.⁵³ Data on estrogen levels in observational studies are particularly complex, given the dynamic hormonal changes due to the menstrual cycle. As we previously reported, total bAs and total uAs concentrations were positively correlated with plasma choline at the baseline visit of FACT, and plasma choline decreased over time in the placebo group following provision of water As-removal filters; these observations were specific to women.³³ While speculative, it is possible that the effect of As exposure on increasing choline in women results in more efficient As methylation. The data available, however, do not allow for a definitive conclusion. While it is known that there is metabolic crosstalk between choline, estrogen, energy metabolism and adiposity,⁵⁴ longitudinal data is limited and needed to evaluate the temporality of the observed associations and whether the effects of As exposure on choline are additive to those of adiposity (Figure A.3).

The lack of association between BMI and As methylation efficiency in Bangladeshi male participants could be related to low levels of adiposity in this population of rural males who are physically very active. In this setting, higher BMI is likely related to higher muscle mass rather than higher adiposity. In other populations in Mexico, the US and Central Europe, higher BMI was related to higher As methylation efficiency in males.^18–20 Adipose tissue plays a role in sex hormone metabolism in males, with the enzyme aromatase, expressed in adipose tissue, mediating the conversion of androgens to estrogens.⁵⁰ In the Strong Heart Study, the association between diabetes-related outcomes and As methylation efficiency was markedly attenuated after adjustment for choline metabolites in both males and females.⁵⁵ Further research is needed to test whether the influence of adipose tissue in estrogen and choline synthesis is a major driver of the association between BMI and As methylation efficiency. The positive association between As methylation efficiency and BMI are contrary to numerous other studies indicating that more efficient methylation of As, i.e. lower MMA% and higher %DMA, is associated with reduced risk for other health outcomes including CVD, skin lesions, and cancers of the bladder, lung, skin and breast.^{7,10,12,13,56–59} While the associations between concurrent BMI and As metabolites are cross-sectional and the direction of the association is unclear, the finding that higher weight gain was prospectively associated with DMA% in FACT supports that higher adiposity, in some manner (again, possibly through estrogen), facilitates As methylation, and not that efficient As methylation increases adiposity.

The current study has several limitations. First, the three studies were not designed and statistically powered with the intention of relating BMI to As methylation profiles. The similar directions of the associations across studies is informative as this is suggestive that the findings are robust, despite differences in the study designs, and that they are primarily observed among females. Also, several of our main findings, in particular the positive association between BMI and SMI was statistically significant in two studies at different life stages (adults and adolescents). Second, BMI is not an ideal proxy for adiposity, as body weight is a poor indicator of body fat and muscle mass, and, in Bangladesh, men tend to be markedly leaner than females at any given BMI. Third, we do not have actual measures of estrogen levels, rendering our interpretation somewhat speculative. Finally, the current analyses were primarily cross-sectional. Although we found a prospective association between weight change and DMA% in FACT, it is critical to validate these results in other longitudinal studies. The assessment of forward causation using a prospective study design with repeated measures of As species and OCM nutrients, as well as diabetes-related outcomes would be particularly useful. Strengths of this study include the availability and replication of the findings in 3 different study populations, the availability of both urine and blood As species, and the adjustment of the associations with measures of OCM-related nutrients.

The findings of this study indicate that higher BMI is cross-sectionally and prospectively associated with more efficient As methylation in women from rural Bangladesh. Understanding factors that influence As methylation and corresponding health outcomes, such as OCM nutrients, is important for determining why individuals vary in their ability to methylate As, and what interventions would likely help to mitigate the effects of As and potentially prevent adverse health outcomes.

Highlights.

• Higher BMI is cross-sectionally and prospectively associated with more efficient As methylation in women from rural Bangladesh

• The associations were attenuated after adjustment for choline

• Findings may be related to the influence of body fat on estrogen levels; estrogen stimulates choline synthesis and choline, through conversion to betaine, can serve as an alternative methyl donor for As methylation

Appendix A

Table A.1.

Intra- and inter-assay CVs for urine and blood arsenic measurements in FACT, FOX, and MANAS.

Arsenic Measure	FACT		FOX		MANAS
	Intra-assay CV (%)	Inter-assay CV (%)	Intra-assay CV (%)	Inter-assay CV (%)	Intra-assay CV (%)	Inter-assay CV (%)
Total uAs	3.1	5.4	3.8	5.1	4.2	4.1
iAs	2.7	4.7	4.5	10.6	1.8	2.9
MMA	2.8	3.9	1.5	3.5	2.9	2.9
DMA	0.6	1.3	0.6	2.8	0.6	1.1
Total bAs	2.7	5.7	3.2	5.7	-	-
iAs	2.9	4.4	11.5	23.2	-	-
MMA	2.5	4.4	3.6	2.9	-	-
DMA	2.4	4.7	2.6	3.5	-	-

Note: In MANAS blood As was not measured.

Table A.2.1.

Female participant characteristics by BMI category in all three Bangladeshi studies.

	FACT			FOX			MANAS
Characteristic	Underweight (<18.5 kg/m2) Mean (SD)	Normal (18.5 – 24.9 kg/m2) Mean (SD)	Obese & Overweight (≥ 25 kg/m2) Mean (SD)	Underweight (<18.5 kg/m2) Mean (SD)	Normal (18.5 – 24.9 kg/m2) Mean (SD)	Obese & Overweight (≥ 25 kg/m2) Mean (SD)	Underweight (<16.7 kg/m2) Mean (SD)	Normal (16.7 – 23.9 kg/m2) Mean (SD)	Obese & Overweight (≥ 24 kg/m2) Mean (SD)
Urine Arsenic Data	N = 252			N = 175			N = 380
% of subjects	24.2	66.3	9.5	25.7	62.9	11.4	21.3	73.7	5.0
Water Arsenic (μg/L)	171 (160)	153 (120)	173 (131)	143 (116)	146 (114)	77.0 (79.1)	58.3 (101)	71.1 (101)	58.3 (63.8)
Age (yr)	37.1 (6.81)	36.8 (7.46)	34.0 (6.63)	43.6 (7.29)	41.7 (7.68)	38.7 (7.71)	14.4 (0.49)	14.7 (0.68)	14.6 (0.62)
Plasma Folate (nM/L)	18.0 (21.0)	16.2 (8.35)	14.6 (6.58)	13.9 (7.25)	14.5 (8.14)	14.6 (5.84)	15.8 (6.22)	14.8 (7.17)	13.9 (7.66)
Plasma Choline (nM/mL)	10.8 (2.42)	11.1 (2.41)	12.4 (2.50)	10.1 (2.51)	10.9 (3.03)	13.6 (4.21)	-		-
Plasma Betaine (nM/mL)	41.9 (11.8)	36.4 (13.8)	38.1 (14.0)	44.2 (17.8)	40.5 (19.6)	42.6 (14.8)	-		-
Plasma DMG (nM/mL)	6.91 (5.36)	6.85 (6.35)	8.01 (11.3)	5.60 (5.87)	4.55 (2.81)	5.84 (4.84)	-		-
Total Urinary Arsenic (μg/L)	164 (154)	157 (124)	114 (96.1)	247 (285)	222 (267)	156 (175)	72.7 (119)	68.4 (118)	59.8 (43.0)
Urinary Arsenic Species
iAs%	14.5 (5.1)	13.8 (4.6)	12.1 (3.2)	18.0 (6.5)	18.3 (5.9)	15.6 (7.2)	15.7 (12.3)	15.2 (9.3)	13.3 (9.7)
MMA%	12.0 (4.2)	11.0 (3.5)	9.5 (2.5)	13.3 (5.0)	11.7 (4.0)	10.4 (3.4)	10.5 (3.4)	10.7 (3.6)	9.2 (3.0)
DMA%	73.5 (6.6)	75.2 (6.6)	78.4 (4.7)	68.7 (8.7)	70.0 (7.4)	74.0 (8.2)	73.8 (8.8)	74.1 (9.6)	77.5 (9.7)
Blood Arsenic Data	N = 291			N = 146			-
% of subjects	25.1	66.0	8.9	26.7	63.7	9.6	-		-
Total Blood Arsenic (μg/L)	10.7 (7.6)	9.1 (7.4)	8.1 (5.7)	14.3 (6.5)	14.8 (8.1)	14.5 (6.6)	-		-
Blood Arsenic Species							-		-
iAs%	27.5 (4.1)	27.1 (4.2)	26.9 (4.4)	30.1 (4.4)	29.5 (3.8)	31.0 (7.4)	-		-
MMA%	44.1 (3.8)	44.0 (4.5)	42.2 (5.3)	37.3 (5.9)	37.7 (5.4)	37.6 (7.0)	-		-
DMA%	28.3 (4.9)	28.9 (4.9)	30.9 (5.4)	32.6 (6.5)	32.9 (5.4)	31.4 (6.0)	-		-

Table A.2.2.

Male participant characteristics by BMI category in all three Bangladeshi studies.

	FACT			FOX			MANAS
Characteristic	Underweight (<18.5 kg/m2) Mean (SD)	Normal (18.5 – 24.9 kg/m2) Mean (SD)	Obese & Overweight (≥ 25 kg/m2) Mean (SD)	Underweight (<18.5 kg/m2) Mean (SD)	Normal (18.5 – 24.9 kg/m2) Mean (SD)	Obese & Overweight (≥ 25 kg/m2) Mean (SD)	Underweight (<16.8 kg/m2) Mean (SD)	Normal (16.8 – 23.1 kg/m2) Mean (SD)	Obese & Overweight (≥ 23.2 kg/m2) Mean (SD)
Urine Arsenic Data	N = 275			N = 167			N = 328
%	43.6	53.5	2.9	41.9	52.1	6.0	40.2	55.5	4.3
Water Arsenic (μg/L)	152 (121)	145 (119)	85.2 (39.1)	159 (134)	138 (142)	89.3 (87.1)	82.6 (111)	78.3 (105)	56.4 (65.2)
Age (yr)	40.8 (8.09)	39.6 (8.61)	43.3 (4.98)	46.0 (8.29)	44.1 (9.23)	43.7 (8.10)	14.5 (0.53)	14.7 (0.71)	14.5 (0.62)
Plasma Folate (nM/L)	18.2 (18.7)	16.0 (13.1)	19.9 (5.98)	11.6 (8.14)	11.5 (5.69)	11.0 (2.75)	18.1 (11.2)	14.2 (7.92)	15.3 (8.18)
Plasma Choline (nM/mL)	12.0 (2.49)	12.6 (2.78)	13.8 (3.07)	12.7 (3.77)	11.8 (3.08)	12.1 (2.76)	-		-
Plasma Betaine (nM/mL)	56.0 (17.3)	51.4 (14.5)	46.0 (16.1)	57.9 (18.7)	50.4 (14.9)	46.6 (14.4)	-		-
Plasma DMG (nM/mL)	7.53 (4.05)	8.31 (9.28)	5.83 (1.26)	8.21 (22.2)	5.09 (2.83)	4.97 (2.42)	-		-
Total Urinary Arsenic (μg/L)	148 (174)	149 (131)	119 (62.1)	247 (190)	191 (187)	163 (220)	65.4 (73.2)	82.6 (127)	106 (108)
Urinary Arsenic Species
iAs%	14.0 (4.1)	14.2 (4.2)	12.1 (5.5)	17.4 (4.1)	17.1 (4.7)	17.4 (4.8)	13.2 (4.6)	14.3 (5.8)	10.9 (3.7)
MMA%	14.9 (4.4)	14.9 (4.4)	13.7 (5.6)	15.9 (5.3)	16.2 (5.0)	14.0 (4.1)	11.6 (3.1)	12.2 (3.7)	10.9 (3.1)
DMA%	71.0 (6.7)	70.9 (6.5)	74.2 (9.9)	66.7 (7.2)	66.7 (7.5)	68.6 (7.1)	75.2 (6.1)	73.6 (7.3)	78.2 (6.1)
Blood Arsenic Data	N = 302			N = 147			-
%	43.0	54.3	2.7	44.2	50.3	5.5	-		-
Total Blood Arsenic (μg/L)	10.9 (6.8)	10.2 (6.0)	8.1 (3.6)	19.1 (11.9)	16.5 (8.4)	16.2 (13.2)	-		-
Blood Arsenic Species							-		-
iAs%	27.0 (3.7)	27.1 (3.6)	25.3 (3.1)	29.4 (3.6)	29.9 (4.7)	27.4 (4.3)	-		-
MMA%	45.1 (4.3)	44.8 (4.2)	41.9 (4.5)	40.9 (4.7)	40.2 (5.6)	40.7 (3.3)	-		-
DMA%	27.9 (5.4)	28.1 (5.0)	32.8 (5.8)	29.8 (4.9)	29.9 (5.3)	31.9 (4.0)	-		-

Table A.3.

Sensitivity analysis using linear models stratified by sex to estimate the mean difference in BMI (kg/m²) for 5% difference in urinary arsenic species stratified by sex in all three Bangladeshi studies.

	FACT			FOX			MANAS
Variable	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3	Model 1
Males	N = 275			N = 167			N = 328
Conventional Model
iAs%	0.01 (−0.32, 0.34)	0.01 (−0.32, 0.34)	0.01 (−0.32, 0.33)	0.08 (−0.49, 0.65)	0.06 (−0.50, 0.63)	0.09 (−0.48, 0.65)	−0.13 (−0.40, 0.14)
MMA%	−0.12 (−0.43, 0.19)	−0.11 (−0.42, 0.20)	−0.09 (−0.40, 0.22)	0.03 (−0.47, 0.52)	0.08 (−0.42, 0.58)	−0.02 (−0.52, 0.49)	0.03 (−0.38, 0.44)
DMA%	0.05 (−0.16, 0.25)	0.04 (−0.16, 0.25)	0.04 (−0.17, 0.24)	−0.04 (−0.39, 0.31)	−0.06 (−0.41, 0.28)	−0.03 (−0.38, 0.33)	0.07 (−0.14, 0.28)
Females	N = 252			N = 175			N = 380
Conventional Model
iAs%	−0.53 (−0.92, −0.13)	−0.54 (−0.94, −0.14)	−0.51 (−0.89, −0.12)	−0.25 (−0.68, 0.18)	−0.25 (−0.68, 0.18)	−0.20 (−0.60, 0.20)	−0.09 (−0.23, 0.06)
MMA%	−0.79 (−1.29, −0.30)	−0.79 (−1.28, −0.29)	−0.57 (−1.08, −0.07)	−0.45 (−1.07, 0.17)	−0.39 (−1.03, 0.24)	−0.35 (−0.93, 0.22)	−0.41 (−0.83, 0.01)
DMA%	0.51 (0.23, 0.78)	0.51 (0.23, 0.78)	0.43 (0.15, 0.70)	0.29 (−0.05, 0.62)	0.27 (−0.07, 0.60)	0.23 (−0.08, 0.54)	0.13 (−0.01, 0.28)
Leave One Out Models
iAs% (MMA% fixed)	−0.39 (−0.8, 0.01)	−0.41 (−0.81, 0.00)	−0.43 (−0.83, −0.03)	−0.23 (−0.66, 0.19)	−0.24 (−0.66, 0.19)	−0.19 (−0.59, 0.21)	−0.11 (−0.26, 0.04)
iAs% (DMA% fixed)	0.28 (−0.45, 1.00)	0.25 (−0.48, 0.98)	0.01 (−0.71, 0.74)	0.25 (−0.54, 1.04)	0.18 (−0.62, 0.97)	0.18 (−0.56, 0.92)	0.35 (−0.08, 0.77)
MMA% (iAs% fixed)	−0.67 (−1.18, −0.16)	−0.65 (−1.17, −0.14)	−0.43 (−0.95, 0.08)	−0.48 (−1.10, 0.13)	−0.41 (−1.04, 0.21)	−0.37 (−0.95, 0.21)	−0.46 (−0.88, −0.03)
MMA% (DMA% fixed)	−0.28 (−1.00, 0.45)	−0.25 (−0.98, 0.48)	−0.01 (−0.74, 0.71)	−0.25 (−1.04, 0.54)	−0.18 (−0.97, 0.62)	−0.18 (−0.92, 0.56)	−0.35 (−0.77, 0.08)
DMA% (iAs% fixed)	0.67 (0.16, 1.18)	0.65 (0.14, 1.17)	0.43 (−0.09, 0.95)	0.48 (−0.13, 1.10)	0.41 (−0.21, 1.04)	0.37 (−0.21, 0.95)	0.46 (0.03, 0.88)
DMA% (MMA% fixed)	0.39 (−0.01, 0.80)	0.41 (0.00, 0.81)	0.43 (0.03, 0.83)	0.23 (−0.19, 0.66)	0.24 (−0.19, 0.66)	0.19 (−0.21, 0.59)	0.11 (−0.04, 0.26)

Note: All models are adjusted for log-transformed sum of iAs species (AsIII, AsV, MMA, and DMA), age and log-transformed plasma folate at baseline. Model 2 indicates model 1 is further adjusted for plasma betaine/DMG ratio at baseline. Model 3 indicates model 1 is further adjusted for only plasma choline levels at baseline. *Bolded values have significant p-values < 0.05.

Table A.4.

Sensitivity analysis using linear models stratified by sex to estimate the mean difference in urinary arsenic species percentages, primary methylation index (PMI), and secondary methylation index (SMI) for a 5-unit difference in lean body mass (LBM) in FACT and FOX.

	FACT			FOX
Outcome Variable	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
Males	N = 275			N = 167
iAs%	−0.04 (−0.64, 0.55)	−0.05 (−0.64, 0.55)	−0.05 (−0.65, 0.55)	0.13 (−0.59, 0.85)	0.12 (−0.61, 0.84)	0.13 (−0.59, 0.86)
MMA%	−0.09 (−0.72, 0.54)	−0.09 (−0.72, 0.54)	−0.05 (−0.69, 0.58)	−0.23 (−1.06, 0.60)	−0.16 (−0.99, 0.67)	−0.26 (−1.08, 0.57)
DMA%	0.14 (−0.83, 1.10)	0.13 (−0.84, 1.10)	0.10 (−0.87, 1.08)	0.10 (−1.08, 1.27)	0.05 (−1.14, 1.23)	0.12 (−1.05, 1.29)
PMI	−0.01 (−0.06, 0.05)	−0.01 (−0.06, 0.05)	0.00 (−0.06, 0.05)	−0.03 (−0.08, 0.03)	−0.02 (−0.08, 0.04)	−0.03 (−0.09, 0.03)
SMI	−0.04 (−0.34, 0.26)	−0.04 (−0.34, 0.26)	−0.05 (−0.35, 0.25)	0.05 (−0.29, 0.39)	0.04 (−0.30, 0.38)	0.06 (−0.27, 0.40)
Females	N = 252			N = 175
iAs%	−0.47 (−1.23, 0.29)	−0.46 (−1.21, 0.30)	−0.47 (−1.23, 0.30)	−0.18 (−1.37, 1.00)	−0.18 (−1.38, 1.01)	−0.11 (−1.33, 1.12)
MMA%	−0.39 (−0.99, 0.22)	−0.40 (−1.00, 0.20)	−0.26 (−0.85, 0.33)	0.00 (−0.82, 0.83)	0.02 (−0.78, 0.83)	0.09 (−0.76, 0.93)
DMA%	0.86 (−0.22, 1.94)	0.86 (−0.22, 1.94)	0.73 (−0.35, 1.80)	0.18 (−1.33, 1.70)	0.16 (−1.35, 1.67)	0.02 (−1.54, 1.58)
PMI	0.00 (−0.06, 0.06)	0.00 (−0.06, 0.06)	0.01 (−0.05, 0.07)	0.02 (−0.04, 0.09)	0.02 (−0.04, 0.09)	0.03 (−0.04, 0.09)
SMI	0.32 (−0.25, 0.89)	0.33 (−0.25, 0.90)	0.20 (−0.35, 0.76)	0.06 (−0.58, 0.70)	0.05 (−0.58, 0.67)	−0.04 (−0.69, 0.62)

Note: All models are adjusted for LBM, log-transformed sum of iAs metabolites (As^III, As^V, MMA, and DMA), age and log-transformed plasma folate at baseline. Model 2 indicates model 1 is further adjusted for plasma betaine/DMG ratio at baseline. Model 3 indicates model 1 is further adjusted for only plasma choline levels at baseline. *Bolded values have significant p-values < 0.05.

Figure A.1.

Correlation matrices for female and male participants in FACT (A & B), FOX (C & D), and MANAS (E & F), respectively. Spearman Correlations between arsenic species, BMI, and OCM metabolites of interest. ΣAs is log-transformed. Abbreviations: Total uAs (sum of inorganic arsenic and its methylated species); uiAs (urinary inorganic arsenic); uMMA (urinary monomethylarsonic acid); uDMA (urinary dimethylarsinic acid); Folate (plasma folate); Choline (plasma choline); and BMI (body mass index).

Figure A.1.2.

Correlation matrices for female and male participants in FACT (A & B), FOX (C & D), respectively. Spearman Correlations between arsenic species, BMI, and OCM metabolites of interest. ΣAs is log-transformed. Abbreviations: Total bAs (sum of inorganic arsenic and its methylated species in blood); biAs (blood inorganic arsenic); bMMA (blood monomethylarsonic acid); bDMA (blood dimethylarsinic acid); Folate (plasma folate); Choline (plasma choline); and BMI (body mass index).

Figure A.2.

Triplots and compositional means of female participants’ uAs species by BMI levels in FACT (A), FOX (B), and MANAS (C). This triplot illustrates the distribution of uAs species in female participants from the lowest (lightest), to highest (darkest) BMI levels. The large circle, square, diamond, and triangle represent the uAs species compositional means for participants in the respective BMI groups. The compositional means represent the average values for each of the three uAs species. DMA% is shown along the left axis, MMA% along the right axis, and iAs% along the bottom axis. (A) FACT, (B) FOX, (C) MANAS

Figure A.2.1.

Triplots and compositional means of male participants’ uAs species by BMI levels in FACT (A), FOX (B), and MANAS (C). This triplot illustrates the distribution of uAs species in male participants from the lowest (lightest), to highest (darkest) BMI levels. The large circle, square, diamond, and triangle represent the uAs species compositional means for participants in the respective BMI groups. The compositional means represent the average values for each of the three uAs species. DMA% is shown along the left axis, MMA% along the right axis, and iAs% along the bottom axis. (A) FACT, (B) FOX, (C) MANAS

Figure A.2.2.

Triplots and compositional means of female and male participants’ bAs species by BMI levels in FACT (A & B) and FOX (B & C), respectively. This triplot illustrates the distribution of bAs species in male participants from the lowest (lightest), to highest (darkest) BMI levels. The large circle, square, diamond, and triangle represent the bAs species compositional means for participants in the respective BMI groups. The compositional means represent the average values for each of the three bAs species. DMA% is shown along the left axis, MMA% along the right axis, and iAs% along the bottom axis.

Figure A.3.

Conceptual diagram of known and proposed associations between adiposity, arsenic, and choline. Solid black arrows indicate known associations based on prior evidence or evidence provided by this research. Dashed blue arrows indicate proposed reasons for associations.