Chronic arsenic poisoning associated with groundwater contamination has been reported from many developing countries, where poor nutritional status is concomitantly found. Although it has been suggested that poor nutritional status affects the toxicity and metabolism of arsenic,^1–3 few systematic reports dealing with this issue exist. A case–control study conducted in Bangladesh showed that malnourished individuals are more often found among patients with arsenicosis than among the non-exposed population⁴; whether the decrease was due to exposure or to clinical manifestations could not be determined. Another case–control study in West Bengal (India) detected some differences between cases and controls in the intake of some nutrients; the difference in the arsenic exposure level between cases and controls and the long interval between the diagnosis and nutritional evaluation made the interpretation of the results difficult.⁵ An extensive study conducted in West Bengal reported heightened susceptibility to arsenic toxicity among individuals with lower body weight⁶; because the groups with different exposure levels were recruited from different areas, the modifying effects (or lack of effects) of nutritional status could have been mediated by other unidentified (environmental) factors such as dietary habit or socioeconomic status. There are some other studies^7,8 describing the effects of nutrition on arsenic toxicity, with very limited quantitative information.

In 1999, groundwater pollution by arsenic was reported from lowland Nepal. We conducted a community-based, comprehensive survey in a “hot spot” of Terai in 2003 and recently published a summary report⁹ describing the enhanced susceptibility of men relative to women, which we also reported on in a Bangladesh community.¹⁰ Our survey was conducted in three closely located communities sharing similar natural and socioeconomic environments, but with considerable within-community variations in exposure levels, which enabled us to evaluate the contribution of nutritional factors to the symptoms of chronic arsenic toxicity. We conducted anthropometry to calculate the body mass index (BMI = body weight (kg)/height² (m²)) as an index of nutritional status, and examined the water consumption of the villagers to estimate arsenic intake from water. On the basis of these data, we report the possible modifying effects of nutritional status on arsenic toxicity in terms of skin symptoms. Part of the basic data such as arsenic concentrations in tubewell water and in urine, and the prevalence of skin symptoms have been published elsewhere as a short communication.⁹

MATERIALS AND METHODS

Study sites and population

This study targeted three communities, the same as we reported elsewhere,⁹ located close to each other, in the Nawalparasi district in western Terai. Most of the residents are of Chaudhari ethnicity (caste) and almost all of them were farmers. These communities share a similar environment and lifestyle, hence, they were analysed as a whole in this paper. The field investigations were conducted from December 2002 to February 2003 and in July–August 2003. The population covered was 1705 (893 men and 812 women) in 249 households at the time of the first survey. The subjects of this study had voluntarily visited our health examination site set up in each community for dermatological examination (n = 1343), had been living in the communities for >6 months, were >15 years of age and for whom anthropometrical measurements (n = 539: 248 men and 291 women) were available. Pregnant women and physically handicapped people were excluded.¹¹ Informed consent was obtained from each participant before the investigations. The study protocol was approved by the ethics committee of the Graduate School of Medicine, University of Tokyo, Tokyo, Japan, and the local authorities of the Nawalparasi District Headquarters.

Dermatological examination

The presence or absence of arsenic-induced skin symptoms, including melanosis, leucomelanosis and keratosis, was examined by one of us (AA), who is a doctor with ample experience in diagnosing arsenicosis cases in Bangladesh^12,13 and was blinded to the exposure level and anthropometric data of the participants.

Anthropometric measurements

To assess nutritional status, the height and body weight of the participants were measured with conventional methods.¹⁴ The BMI values were calculated from height and weight, and each individual was classified as “underweight” (BMI <18.5), “normal” (18.5⩽ BMI <25), “overweight” (25⩽ BMI <30) or “obese” (BMI ⩾ 30) individuals as defined by the World Health Organization (WHO),¹⁵ where the “underweight” category is regarded as chronic energy deficiency.¹⁵

Sampling and determination of arsenic in tubewell water and urine

Details of the sampling and arsenic determination of both tubewell water and urine have been described elsewhere.^9,16 Water samples from all tubewells in the three communities (n = 146; “age” of tubewells = 10 (8) years; mean (SD)) were collected, immediately acidified with HCl and maintained at 4°C until analysis. Spot urine samples were collected from 106 couples (husbands and wives), frozen, transported to the laboratory at the University of Tokyo, and stored at –80°C until analysis. Arsenic concentrations of the tubewells (As_tw) and the urine samples (As_u) were analysed by atomic absorption spectrophotometry coupled with hydride generation (for water samples (SOLAAR 969AA, Thermo Elemental, Cambridge, UK) and for urine samples (ZL-4100, Perkin Elmer, Norwalk, Connecticut, USA)). Before atomic absorption spectrophotometric measurement, the urine samples were wet-ashed by heating with acid mixture. All the samples were found to contain arsenic above the detection limits (1 and 3 μg/l for water and urine, respectively). The accuracy of the assay was confirmed by inclusion of reference materials (for details, see Watanabe et al¹⁰). The results fell within the respective certified ranges. The creatinine concentration of the urine samples was determined spectrophotometrically.¹⁷ The chemicals and the reagent kit used were obtained from Wako Pure Chemical Industries (Osaka, Japan).

Water consumption measurement and calculation of daily arsenic dose

The amount of water drunk by the villagers was measured for 24 h in both winter (January) and summer (August) seasons, to take into account seasonal climatic variations. For this purpose, we provided a 1-litre polyethylene terephthalate water bottle to a voluntary subset of participants (total n = 180; n = 45 per sex per season) and requested them to drink water only from the bottle and refill the bottle when empty. The number of refills and the amount of water that remained were checked after 24 h to calculate daily water intake. The details and reliability of this method have been described elsewhere.¹⁶

Using the intake data, sex-specific and season-specific regression lines of the water intake on body weight were calculated, from which two season-specific water intakes were estimated for all participants (n = 539). Then, taking into account the relative duration of the seasons, annually averaged daily water intake (DWI; per person) was calculated as

DWI=(3×(water intake in winter))+9×(water intake in summer))/12.

Finally, the daily arsenic intake (DAI) from drinking water (per unit body weight) was computed for each subject as (As_tw×DWI)/BW. Although contribution from potential arsenic sources other than water (for example, food) may not be negligible, arsenic concentration of water should be the primary determinant of total arsenic intake.¹⁶

Statistical analysis

Analysis of variance and χ² test were used appropriately. For these tests, SPSS V.10.0 statistical software was used. Significance was set at p<0.05. The statistical difference in the prevalence was evaluated by Poisson distribution and 95% confidence intervals (CI), using CI Analysis V.2.0.0 software.¹⁸ To adjust for the possible confounding effects of age, the prevalence was also analysed by the nominal logistic regression model using JMP software (v.6.0, SAS Institute). In this model, sex, BMI (categorised as 1 for BMI<18.5, others as 0) and age (categorised into five groups: <20, 20–29, 30–39, 40–49, ⩾50 years) were included as independent variables.

RESULTS

Arsenic exposure indices

A summary of the arsenic exposure indices has been reported elsewhere.⁹ Briefly, the mean (SD) As_tw value was 403 (229) (range 3–1072) μg/l; 97.9% and 87.6% of the tubewells exceeded the WHO and Nepal Interim Standard (NIS) limits (10 and 50 μg/l, respectively) and >30% exceeded 500 μg/l. The mean As_u values were 744 (411) (range 33–1763) and 937 (565) (range 16–2746) μg/g creatinine for men and women, respectively. As_u significantly and positively correlated with As_tw both in men (r = 0.74, p<0.001) and women (r = 0.69, p<0.001), and As_u values of husbands were significantly correlated with those of wives (p< 0.001); data not shown), suggesting that the main source of arsenic intake was related to the household, presumably tubewell water. In this area, household members rarely used tubewells of other households.

Estimation of arsenic intake from water

For men and women, the means of DWI were 67 (7) and 63 (10) ml/kg/day, respectively, and those of DAI were 31 (17) and 27 (17) μg/kg body weight/day, respectively. Both indices showed a significant sex difference. The DAI far exceeded the provisional tolerable daily intake of 2.1 μg/kg body weight/day (Food and Agriculture Organization/WHO, 1989). The DAI and the As_u were significantly and positively correlated in both sexes (fig 1).

Nutritional status of the participants

Of the 539 participants, approximately half (44% of men and 56% of women) were categorised as underweight (see Materials and methods section for definition). A few overweight (3% of both sexes) and no obese participants were found. The distribution of participants by BMI categories was significantly different between the sexes (χ² test; p<0.05), reflecting the higher prevalence of underweight among women. For further analyses, the participants were dichotomised into underweight and normal (including overweight) groups (table 1).

Underweight men had significantly higher As_tw (and As_u) than normal weight men, whereas such a difference was not found among women. In both sexes, the underweight groups had significantly higher DAI than the normal groups. In addition, there was a minor but significant age difference in the male groups. The youngest (<20 years) and oldest (>49 years) groups had significantly lower BMIs than other age categories for both sexes.

Effect of nutritional status on arsenic toxicity

The prevalence of arsenicosis among participants was 15% (83/539), being higher in men (23%; 57/248) than in women (9%; 26/291). The prevalence significantly increased with increasing exposure (table 2, upper).

Table 3 shows the effects of nutritional status (BMI) on the prevalence rates for each exposure level. In both sexes, the prevalence ratios were consistently larger than unity (>1) in the middle and higher tertiles, and significant differences were found in the middle tertile groups. The lower tertile groups of either sex did not show an increased prevalence in the underweight group. No significant difference in prevalence among the age categories was found for either sex, although those aged <30 years tended to show lower prevalence. With (nominal) logistic regression analyses on the three exposure level groups, significant regression was achieved only in the mid-tertile (χ² = 24.4, p<0.001) and high-tertile groups (χ² = 14.9, p<0.05). In the mid-tertile group, the effects of BMI (p = 0.023) and sex (p<0.001) were significant, whereas age was not. In the high-tertile group, only the effect of sex (p = 0.002) was significant. Thus, the significant effect of nutritional status on prevalence in the mid-tertile groups, as shown in table 3, remained after adjusting for age.

Effect of arsenic exposure on nutritional status

Figure 2 shows that the DAI and BMI were significantly and inversely correlated (p<0.001 for either sex). Significant inverse correlations were also found between As_tw and BMI (r = –0.24) or between As_u and BMI (r = –0.29) in men (n = 106) but not in women (n = 106). Both in men and in women, an approximately 30% increment in the prevalence of underweight was observed between the high-exposure tertile and the low tertile (table 1, lower).

DISCUSSION

This study examined the mutual interaction between nutritional status and arsenic toxicity in a rural, lowland Nepal population. The high prevalence of skin manifestation (15%) and malnutrition (“underweight”; 50%) in the surveyed population made such analyses of interaction meaningful. The prevalence of skin manifestation was lower in our previous report (7%), which included a younger subpopulation (<15 years) with negligible prevalence (<1%).⁹

The prevalence of malnutrition was in line with previous nationwide surveys exclusively examining women.^19,20

Our results showed that arsenicosis (skin symptoms) was more prevalent among those with lower BMI (low nutritional status) at similar exposure levels and, additionally, that arsenic exposure was associated with a decrease in BMI. This may be the first demonstration of dose-dependent suppression of BMI in a human population chronically exposed to arsenic, although a causal relationship cannot be established because of the cross-sectional nature of this study. Compared with a few preceding studies dealing with the interaction between nutrition and arsenic toxicity in human populations, this study describes the interaction in a well-defined and apparently homogeneous population in a quantitative manner. In the following text, we discuss the significance of malnutrition as a toxic manifestation of arsenic and its relationship with toxicity.

Although body weight reduction by high and acute exposure to arsenic, presumably a non-specific effect, has been reported,²¹ the effect of chronic exposure to arsenic on underweight/BMI has been scarcely reported. Intrapopulation variations in the factors that can influence BMI, such as dietary intake, physical activity, and socioeconomic status including education, should be minimal, because the three communities were closely located, sharing similar environment and lifestyle (see also Sudo et al²²). Thus, our results suggest that the lowered BMI could be a toxic manifestation of chronic exposure to arsenic at relatively high exposure levels—that is, at the mean DAI, 30 μg/kg/day for both sexes, 15 times higher than the provisional tolerable daily intake set by WHO/Food and Agriculture Organization.

What is the practical importance of reduced BMI by arsenic? Firstly, in our population, exposure to arsenic substantially (a 1.9 and 1.4 times higher prevalence was observed in high and middle exposure tertiles than in the lowest tertile) exaggerated the malnutrition, a serious public health problem in such an area. Secondly, although the increase in the prevalence of skin symptoms by arsenic dose seems to be much larger in men than in women, the increments of underweight were similar in both sexes (table 2). Whether such endpoint dependence is associated with the possible existence of different “receptors” for different end points or is merely reflecting the inherent nature (sex difference) of different symptoms should be important in understanding the toxicology of arsenic. In this connection, it should be noted that either a biochemical or a behavioural explanation might be possible for the decrease in BMI. Arsenic can affect insulin-related sugar metabolism²³ or disturb glucocorticoid function,^24,25 both of which might compromise the energy metabolism. Alternatively, systemic disease status caused by exposure to arsenic^26–28 may lead to malnutrition (loss of body weight) through reduced food consumption.

On the other hand, skin symptoms of arsenic toxicity were more frequently observed among individuals with a lower BMI. This is in line with two other large-scale studies: one in West Bengal⁶ and another recent one in Bangladesh.²⁹

As BMI and prevalence of skin symptoms showed negative and positive correlations with exposure to arsenic, respectively, it is naturally expected that BMI and prevalence would be negatively correlated. If this is the case, it is expected that the higher the dose, the greater the prevalence ratio (table 2). In fact, the highest prevalence ratios were consistently obtained in the middle exposure tertiles rather than in the higher tertile groups of either sex (although not significant in women). Although this observation should be confirmed with much larger samples, it may be compatible with the view that poor nutritional status shifts the dose–response curve (between arsenic and skin lesions) to the left; if such a leftward shift occurs, the effect should be most evident at the middle range of the ordinal (sigmoid) dose–response curve (see fig 3). With higher dose, the effect of nutrition would be overwhelmed by the toxic effect of arsenic per se. In the West Bengal study,⁶ the effect of malnutrition was observed at their highest dose range, which was similar to that of our middle tertile group. Mitra et al⁵ found susceptibility to arsenic was enhanced at relatively low exposure levels (As_tw<500 μg/l), by reduced intake of some nutrients although they could not detect any effect of energy intake. Assuming that the water consumption of Mitra’s study participants was similar to our participants (approximately 65 ml/kgBW/day), the arsenic intake would be estimated as <32.5 μg/kg/day, which again coincides with our middle (and lower) exposure tertile. In the recent Bangladesh study,²⁹ the highest exposure group had a median As_tw of 255 (ranging from 175 to 864) μg/l, which also coincides with our middle exposure tertile. Taken together, there might be an arsenic exposure level at which nutritional status (either energy or specific nutrients) affects the toxic manifestation of arsenic. The prevalence of underweight was higher among those exhibiting skin symptoms than those not exhibiting them, and the prevalence ratio, as similar to table 3, was highest for the middle exposure tertile (data not shown).

The responsible nutrients and their mechanisms for toxicity modification need to be elucidated with further studies. The deficiency of several nutrients such as calcium, animal protein, folate and fibre (but not total energy) might increase the risk of skin lesions.⁵ Earlier experimental studies suggest that poor nutrition results in decreased methylation of arsenic owing to decreased supply of the methylation substrate, eventually leading to enhanced arsenic toxicity.³⁰ This is supported by a recent human study showing that a low dietary intake of protein, iron, zinc or niacin was associated with enhanced accumulation of monomethylarsonic acid in a US population,³¹ although the exposure level was much lower than those found in south Asian countries, including the present one.

In summary, our data showed that BMI reduction was also a quantitative end point of chronic arsenic toxicity, which was essentially devoid of the sex difference that was consistently observed in skin symptoms. The mutual interaction between poor nutrition and arsenic toxicity might create a vicious cycle between nutrition and toxicity, in which arsenic toxicity exaggerates malnutrition and malnutrition in turn exaggerates toxicity. Thus, such a cycle should be interrupted to alleviate the arsenic toxicity and to reduce the prevalence of malnutrition, which would be one of the most concerned public health problems in the rural communities of developing countries. Elucidation of the mechanisms responsible for the interaction might lead to a design of nutritional intervention to alleviate the effects of arsenic.