The sociodemographics and the range of As concentrations in water [As]w and urine [As]u of the respondents are summarized in Table 1. Approximately 56 % of the subjects were female; more than 55 % of the houses had corrugated tin walls and roofs. The households were mainly engaged in farming and wage labor. Though a considerable number of the households belonged to people falling into the businessman category, the parents were actually engaged in a very small-scale business such as vegetable selling and shop keeping. Bangladesh is a developing country with the majority of households falling into the low-income stratum. This study’s participating households belonged to low to low–medium strata. More than 50 % of the respondents were either illiterate or had only received primary education.
The mean As concentration in tube-well water was 71.7 μg/L, which is above both the WHO and Bangladesh recommended maximum tolerable contamination levels, whereas the mean urinary As concentration was 205.3 μg/L, which is above the noneffective level (137 μg/L). In general, the results in Table 1 indicate that most of the respondents came from families with poor socioeconomic status, and thus, they are consuming highly As-contaminated tube-well water.
ANOVA was performed to investigate the influence of socioeconomic status on As contamination. The results indicated that [As]u and [As]w significantly differed by parental income [F(2,309) = 5.03, p < 0.01 and F(2,309) = 5.01, p < 0.01, respectively]. This suggests that sociodemographic conditions have an influence on As exposure, and individuals with higher income can take preventive measures, such as using a water filter, changing water source, etc. However, there was no significant relationship between As contamination ([As]u and [As]w) and other socioeconomic indicators such as education, sanitation, and house type.
The predicted means and standard deviations of the IQ percentile and SC score for the three urinary As groups are presented in Table 2. The IQ in the high-[As]u group did not differ from that in the medium-[As]u group. The percent distributions of the IQ grades for the three [As]u groups (low, medium, and high; Fig. 1) indicated that a very small percentage of the respondents from the high-[As]u group possessed above-average intellectual capacity (>grade III), with most having average or below-average IQ grades. In contrast, a comparatively higher percentage of the respondents from the low-[As]u group possessed above-average intellectual capacity. One-way ANOVA was applied to assess the differences in mean and variance, and the results indicated that the IQ percentile significantly differed among the [As]u groups [F(2,309) = 7.7, p < 0.01]. A post hoc analysis for multiple comparison revealed that high (p < 0.05) and medium (p < 0.01) levels of [As]u significantly lowered the mean IQ percentile compared with the low [As]u level.
The mean SC scores (Table 2) differed among the [As]u groups (low, medium, and high). The percent distribution of SC for the [As]u groups in Fig. 2 illustrates that a higher percentage of children who were averagely socially competent scored between 31 and 45. A very small percentage (approximately 4–5 %) of the children from the high-[As]u group possessed high social competence (scored above 45). Comparatively, a high percentage of the children from the high-[As]u group scored below 30, which indicates poor social competence. One-way ANOVA revealed that SC significantly differed in the [As]u groups [F(2,309) = 14.1, p < 0.001]. A multiple-comparison post hoc analysis clarified that a significant reduction in SC score was found in the medium- (p < 0.001) and high-[As]u (p < 0.01) groups in comparison with the low-[As]u group. Moreover, a significant effect of [As]u on the SC score [F(2,306) = 12.4, p < 0.001] was identified after controlling for the socioeconomic indicators.
The relationship between the individual urinary ([As]u) and drinking water ([As]w) As concentrations was analyzed, and the results are presented in Fig. 3. A positive correlation (Pearson r = 0.47, p < 0.01) was identified between [As]w and [As]u; both variables were converted into their respective logarithmic values. It is apparent that, in the lower [As]w range (up to 10 μg/L), the contribution of tube-well water to As intake is small in comparison with other unknown sources. Moreover, the results indicate a relatively constant [As]u value up to [As]w of 10 μg/L. When [As]w exceeds 10 μg/L, the deviation also increases rather than remaining constant. This suggests the existence of other sources of As intake.
We further investigated the effect of As concentration in water [As]w on IQ and SC. The mean IQ percentile (50.5 ± 24.3) presented in Table 3 for the level 1 (L1) [As]w group was higher than that of other groups. However, the mean IQ scores among the L2, L3, and L4 [As]w groups were nearly the same. It was found that water As concentration ([As]w) had a significant influence on IQ [F(3,308) = 5.4, p < 0.01]. A planned contrast indicated that the mean IQ score of the L1 group significantly differed from the L2 and L4 groups at p < 0.05 and p < 0.01, respectively. Moreover, the mean IQ percentile in the L4 group was significantly lower than in the L1 group (p < 0.01). Finally, controlling for socioeconomic indicators such as parental education, occupation, and income, the ANCOVA revealed that there was a significant effect of [As]w on IQ [F(2,306) = 6, p < 0.01]. This indicates that consumption of a high concentration of As through groundwater significantly lowered the mean IQ of the children. On the contrary, a similar trend in mean SC scores was found among the [As]w groups. SC significantly differed in the [As]w [F(3,308) = 3.4, p < 0.05] groups, and the mean SC score in group L4 significantly differed from the L1 group (p < 0.05). However, no significant effect of [As]w on SC was found after controlling for the socioeconomic indicators.
To identify other sources of As, food consumption patterns were assessed by FFQ and the 24-h recall method. According to the 24-h recall method (Fig. 4a), 62 % of the respondents consumed rice and 32 % consumed daal three times per day. The FFQ results (Fig. 4b) indicated that 100 % of respondents consumed rice 5–7 days per week, followed by nonleafy vegetables (90 %) and then daal (61 %).
The amount of water used in cooking was measured using a food preparation survey on site (Table 4). Rice, the major Bangladeshi food, is usually cooked with either “fixed water” or “excess water.” In the former process, water is added at approximately 2.5 times the volume by weight of rice. This rate is much higher than the 1.3 used in Japan . In the excess water process, water is added at approximately three to four times the volume by weight of rice, and when the rice has been boiled, the excess water is discarded. In our study area, most of the households cooked rice with “excess water.” As shown in Table 4, an average of 1,165 g rice was boiled with 3,827 mL water initially. After discarding the excess water, 3,312 mL water (approximately 2.8 times the volume by weight of rice) was absorbed in the cooked rice. Curry is the most common cooking recipe, including either meat, fish, egg or vegetables. A large amount of water is also required when cooking curry. Another common recipe is daal soup with different types of pulses/lentils. Water is not discarded after cooking either curry or daal soup. For daal soup, the ratio of lentils to used water is approximately 1/6, whereas the ratio for vegetables is 1/0.8. Based on the data obtained from 24-h recall and the FFQ, the weekly consumption of water from cooked foods was estimated (Table 4). As shown in the rightmost column, the major source of water consumption was cooked rice, followed by daal soup. The contribution of water from different types of curry and chapati was minor because they were not often consumed.