Sample collection We collected samples and conducted interviews in June 2005, in the rainy season; the maximum temperature during the survey period was around 38 °C. The study area was one block in Chunakhali village (24°36’N, 88°12’E), Chapai Nawabganj district, Rajshahi division, Bangladesh. This typical poor rural village has a population of about 2500. The block, where about 160 people live, is severely affected by arsenic and many people are afflicted with skin lesions associated with chronic arsenic poisoning. In addition, there are no arsenic treatment utilities or safe water distribution systems in the block. Water from dug wells contains much less arsenic than that from tubewells, but whereas some people have a dug well in or near their home, others have to walk up to a couple of hundred meters to get this less arsenic-contaminated water. After consulting with the local block leader and a dermatologist, we selected 77 people from 18 families in this block as our target subjects. We selected families representative of the block both financially and geographically and that had at least one family member with arsenical skin lesions. All the selected families agreed to participate the survey and many of them reported that they had changed their drinking water source from the arsenic-contaminated tubewells to a less-contaminated source from several months to several years previously. We collected water from the present drinking water source of each family and also from previous drinking water sources when they were still available, in polyethylene bottles. Ideally, we sampled water directly from the source, but in some cases, we sampled water from a container in a house. In the latter case, we collected the sample after shaking the container vigorously. Water samples were transported to the laboratory by air, and then 1% v/v nitric acid (ultrapure grade; Kanto Chemical Co., Inc., Tokyo, Japan) was added to the samples, which were kept in a dark container at 4 °C until analysed. We collected food samples by the duplicate portion sampling method (Tsuda et al., 1995; WHO/IPCS, 2000). We selected one adult from each family as the respondent, who submitted his or her duplicate meals (breakfast, lunch, and supper) for one day. Each cooked item was collected in a separate plastic bag and weighed. We estimated food intakes of non-respondent family members in relation to the food intake of the respondent from interview results, and these food intakes were used for the estimation of arsenic intakes via cooking water of the non-respondents. Food samples are typically composited by meal or by day (WHO/IPCS, 2000), but we composited the food separately into the following four categories: cooked rice, cereal, solid food, and liquid food. Cereal was defined in this study as cereals that the people ate for breakfast, and did not include rice or rice gruel. Cereal was mainly of two types: one, called “Kalai rooti” locally, was a variety of chapati made from lentil powder and rice flour, and the other was oat powder mixed with water, sugar, and salt. Solid food was defined as the solid portion of meals such as fried vegetables, excluding staples such as cooked rice and cereal. Liquid food was the liquid portion of soup, but solid ingredients in soups were categorised as solid food. Spices and lentils that could not be separated using a spoon and fork were treated as liquid food. Uncooked rice was also collected for comparison. After collection and arrangement of the food samples, solid food samples were homogenised and then transported to the laboratory by air. In the laboratory, solid food, rice, and cereal samples were freeze-dried for two to three days, and the water contents were estimated by measuring the weight of the samples before and after freeze-drying. The samples were stored in individual plastic bags and kept in desiccators. Liquid food samples were stored at –80 °C until analysed. Estimation of direct water consumption rate To quantify the daily water intake by direct drinking, we used a method similar to the Cup Method (Watanabe et al., 2004), in which direct water intake is estimated by asking the respondent how many cups of water are consumed in a day. At the first visit, the cup used for drinking water was identified and the capacity of the cup was measured. Each subject self-recorded the number of cups drunk (the “water diary” method) (Shimokura et al., 1998; Levallois et al., 1998). Recording sheets were provided to the subjects, who were asked to mark the sheet every time they drank water from their own cup. About 24 h later, the sheets were collected, and the number of cups marked was multiplied by the capacity of that subject's cup to estimate the water consumption rate. Water diary data were obtained for 65 of the 77 subjects (84%). Beverages made with water, such as tea, were not considered in this study because they are not common drinks for the poor village people. Most of the poor village people take tea in a small cup, and the amount of water used is negligible, not more than one cup per day. Analytical method for total arsenic concentration Arsenic in water was quantified by inductively coupled plasma–mass spectrometer (ICP-MS; HP-4500; Agilent Technologies, Inc., Palo Alto, CA, USA). The instrumental parameters were as follows: RF power, 1200 W; RF matching, 1.8 V; sample skimmer cone in Ni; monitoring masses; 75 (As) and 77 (ArCl+); plasma flow rate, 16 L/min; auxiliary flow rate, 1.1 L/min; nebulizer flow rate, 1.2 L/min. Yttrium (Y; m/z = 89) was used as the internal standard. Dried samples of rice, cereal, and solids were finely ground in a mill (A11 basic; IKA Werke GmbH & Co. KG, Staufen, Germany). Then the ground solid samples and liquid samples were digested using a microwave digestion system (ETHOS TC; Milestone S.r.l., Bergamo, Italy) by the following procedure. A 0.5-g (dry weight) portion (solid sample), or a 1.0-g (wet weight) portion (liquid sample) was weighed into a PTFE vessel, and 3 mL of nitric acid and 2 mL of hydrogen peroxide (ultrapure grade; Kanto Chemical Co., Inc., Tokyo, Japan) were added. The basic program of the microwave digester was as follows: increase the temperature from room temperature to 210 °C over 30 min, hold at that temperature for 15 min, and cool down to room temperature over 10 min; maximum power was 1000 W. Times and temperatures were modified slightly, depending on the sample type. The digested solution was made up to 50 mL with ultrapure water and filtered through a 17 0.45-μm membrane filter before injection into the ICP-MS instrument. The validity of the analysis was confirmed with the Standard Reference Materials (SRM) Rice Flour (SRM1568a) and Typical Diet (SRM1548a), purchased from the National Institute of Standards and Technology (NIST), Gaithersburg, MD, USA. The certified arsenic concentration and our observed concentration (mg/kg dry wt) of the Rice Flour were 0.29 ± 0.03 and 0.26 ± 0.01 (n = 3, mean ± standard deviation [SD]), and those of the Typical Diet were 0.20 ± 0.01 and 0.21 ± 0.00 (n = 3), respectively. The certified and the observed values were thus in good agreement.