This study involved 405 participants selected from a study base of 7,683 individuals living in the South 24-Parganas district of West Bengal, India, who were identified in a 1995 through 1996 cross-sectional survey (Guha Mazumder et al. 1998
). Participants had been selected for a case–control study of skin lesions caused by arsenic involving 192 cases and 213 controls (Haque et al. 2003
), whose primary sources of drinking water contained < 500 µg/L InAs [n
= 4,185 (2,160 females and 2,025 males)].
Cases had positive skin-lesion classifications, with either hyperpigmentation (mottled, dark-brown pigmentation bilaterally distributed on the trunk), keratoses (diffuse thickening of palms or soles, with or without nodules), or both conditions at the time of the survey. Of the 7,683 individuals in the base population, we randomly selected controls from a subset of 4,185 individuals, with drinking water concentrations of arsenic < 500 µg/L; controls were matched to cases by sex and by age within 4 years. The study protocol was approved by the institutional review boards of the Institute of Post Graduate Medical Education and Research, Kolkata, India, and the University of California–Berkeley, Berkeley, California. Informed consent was obtained from all participants.
In this investigation, we assessed dietary intake and measured blood micronutrients and methylated arsenic species in urine samples for pooled cases and controls. Cases with skin lesions were included after adjusting for possible distortion of associations between the nutritional factors and arsenic methylation patterns by including an indicator variable for skin lesions in the analyses.
In India, socioeconomic status is commonly measured by type of dwelling, which is correlated with household economic status (Mishra et al. 1999
). In this study, we determined socioeconomic status based on the materials used to construct the house where the respondent lived. We considered three types of houses: pucca houses, built with high-quality materials such as bricks or concrete; semipucca houses, constructed partly with clay and bricks; and kacha (mud houses). We classified educational status as nonformal (participant never attended school), primary education (up to 4 years of education), high school education (between 8 and 12 years of education), and beyond high school education or tertiary education (> 12 years of formal education).
Assessment of dietary intake and blood micronutrients.
We ascertained food intake for each participant with a detailed questionnaire based primarily on 24-hr recall. The methods used for dietary assessment have been described elsewhere (Mitra et al. 2004
). In brief, the most senior woman, who in this population directed the preparation of food for the family, was interviewed and questioned about each meal from lunch the previous day through breakfast on the day of the interview. The volume of each cooked food was assessed by asking the senior woman to estimate these volumes using standard cups and plates. Standard-sized spoons were used to assess the intake of sugar and oil. We asked about weekly consumptions of meat, fish, eggs, milk, and fruit, because these items were not typically consumed on a daily basis. The 1-week intake of these food items was then divided by 7 to compute the mean intake per day. We calculated total 24-hr intake of the following nutrients using a spreadsheet program based on food composition tables (listed here in alphabetical order): animal fat, animal protein, calcium, carbohydrate, carotene, fiber, folate, iron, niacin, phosphorus, retinol, thiamin, vegetable fat, vegetable protein, vitamin B6
, vitamin B12
, vitamin C, and zinc (Mitra et al. 2004
). We adjusted each dietary variable for individual total calorie intake by dividing total daily dietary intake by total calorie intake.
Field team physicians interviewed participants using a structured questionnaire, conducted a general examination, and obtained blood samples from each participant when they were visited in their homes. We have previously presented detailed information concerning storage and analysis of blood samples (Chung et al. 2006
). In brief, nonfasting blood samples were collected and stored in an ice chest in the field. Aliquots were prepared within 24 hr, frozen at –20°C in India, and later transported to the United States on dry ice, where they were stored at –70°C until laboratory analysis. Pacific Biometrics (Seattle, WA, USA) conducted most of the serum and plasma analyses for the micronutrients and biochemical indicators or in some instances arranged for them to be done at a different laboratory. Plasma measurements included homocysteine, glutathione, cysteine, methionine, vitamin B6
, retinol (vitamin A), alpha-tocopherol (vitamin E), alpha-carotene, beta-carotene, lycopene, lutein-zeaxanthin, and beta-cryptoxanthine (Chung et al. 2006
). Serum measurements included glucose, cholesterol, vitamin B12
(cobalamin), folate, transthyretin, and selenium.
Measurement of urinary arsenic and creatinine.
Because urine is the primary route of arsenic excretion, the proportions of the various arsenic metabolites in urine are commonly used as an indicator of the degree to which a person can methylate ingested InAs (National Research Council 2001
). Spot urine samples were collected from each participant and stored frozen for later analysis at the Trace Organics Analysis Laboratory (University of Washington, Department of Environmental and Occupational Health Sciences, Seattle, Washington). The urinary concentrations of arsenic were measured using hydride generation/cryogenic concentration, atomic fluorescence detection, using an Excalibur AFS detector (PS Analytical, Inc., Deerfield Beach, FL). In this technique, InAs, MMA, and DMA were reduced to the corresponding arsine in a batch reactor, using sodium borohydride in 5-mL samples. The volatile reduction products (arsenic, methyl arsine, and dimethylarsine) were removed by sparging with helium. Entrained arsines were concentrated in a chromosorb-filled cryogenic trap at liquid nitrogen temperatures until all arsine-forming arsenic in the sample had reacted. The cryotrap was then allowed to warm, and the collected arsines were separated on the basis of differential volatilization. The separated volatile arsenic species were detected with a hydrogen microburner combustion cell to convert arsines to elemental arsenic. To prevent interference by other compounds, each urine sample was acidified with 2 M HCl and allowed to sit for at least 4 hr prior to speciation analysis. Total arsenic was determined by atomic fluorescence spectrometry with flow injection analysis (Excalibur AFS, PS Analytical) on a portion of the urine sample that had been mineralized with hydrochloric acid and peroxide; the result was compared with the sum of the species detected. If a significant amount of arsenic remained undetected, additional digestion or assay for arsenobetaine was performed. Detection limits for InAs, MMA, and DMA were 0.5, 1, and 2 µg/L, respectively. Concentrations below the detection limit were set at one-half the detection limit. The methods detected both oxidation states for each form of arsenic (inorganic, monomethyl, dimethyl) and provided a single result for each methylation level. Urinary arsenic concentrations were measured in micrograms per liter of urine; percentages of InAs (InAs%), MMA (MMA%), and DMA (DMA%) were calculated using the sum of arsenic species detected as the denominators. The creatinine concentrations in urine were determined kinetically by a clinical autoanalyzer using the Jaffe reaction (Larsen 1972
Statistical methods. The relationships between demographic variables and urinary arsenic methylation patterns were first assessed in simple stratified analyses. Nutritional factors and urine creatinine were divided into tertiles, and we compared arsenic methylation patterns in the highest tertile of each factor with patterns in the lowest tertile. We conducted t-tests of the differences in mean urinary InAs% between the highest and lowest tertiles of each nutritional factor. Nutritional factors were then ranked, starting with the nutritional factor associated with the largest absolute mean difference in InAs% between its highest and lowest tertiles. Nutritional variables with the highest mean differences were then entered into a series of multiple linear regression analyses, as explained below. This series of steps was repeated for MMA% and DMA%.
In the multiple linear regression models, InAs%, MMA%, and DMA% were treated as response variables, and tertiles of selected dietary or micronutrient variables were the main explanatory variables. The lowest tertile of the dietary or micronutrient variable was treated as the reference category, and dummy variables were created for the middle and highest tertile for each nutritional variable. We used step-up methods, starting with the strongest predictors found in the univariate analyses. The beta coefficients corresponding to the highest tertile of the diet or micronutrient gives an effect measure of the potential impact of each nutrient on the specific urinary methylated arsenical. For example, a beta coefficient of –5 in the model of InAs% for a particular nutrient would indicate 5% difference (e.g., 20% InAs% for the high tertile compared with 25% InAs% for the low tertile), with adjustment for the effect of all other variables specified in the model. This represents the independent effect of the nutrient on the specified urinary methylated arsenical. In addition to nutritional factors, the other variables in the model were age, sex, housing, education, presence or absence of skin lesions, body mass index (BMI), and total urinary arsenic. Housing and education were modeled as categorical variables, whereas age, BMI, and urinary arsenic were modeled as continuous variables.
The important advantages of using tertiles of nutritional factors in the analyses, rather than continuous variables, include the fact that there is no implicit assumption about the functional form of any dose–response relationship. In addition, outliers often have undue influence on regression analyses, and dividing into tertiles avoids this problem, as outliers are merely one member of the tertile to which they belong. However, we repeated the analyses that produced the key findings using continuous variable regression and obtained similar results. p-Values were based on chi-square tests of trend for the following ordinal variables: categories of age, education, type of housing, urinary arsenic, and BMI ().
Distribution of the percentage of urinary total arsenic in inorganic form (InAs%), monomethylated (MMA%), and dimethylated (DMA%), according to other factors for the 405 participants.