Bulletin of the World Health Organization

Risk of arsenic-related skin lesions in Bangladeshi villages at relatively low exposure: a report from Gonoshasthaya Kendra

Corbett McDonald a, Rezaul Hoque b, Nazmul Huda c, Nicola Cherry d

Introduction

Arsenic is a human carcinogen and skin pathogen; the evidence has been documented by the International Agency for Research on Cancer on several occasions.1 A potentially serious threat to public health that has become evident during the last 30–40 years is natural arsenic contamination of drinking water, notably in South America and Asia and more recently in Bangladesh and West Bengal (India). Toxic manifestations have been primarily keratotic skin lesions; more threatening, however, are internal cancers, for which there is less certain evidence.2 A major national disaster was evidenced by the publication in 2001 of a systematic survey of nearly 4000 wells3 showing that half Bangladesh’s administrative districts had average arsenic concentrations above – often well above – 50μg/L, a level known to be hazardous. These averages may be a poor reflection of the problem, as levels varied enormously within and between villages by several orders of magnitude. Thus it is virtually impossible to estimate what proportion of Bangladesh’s rural population is at risk even at the national standard of 50μg/L, let alone above the WHO recommended guideline of 10μg/L. Given this extremely varied pattern of exposure, there is need for epidemiological data over a wide range of concentrations.

There have been five recent prevalence studies: one in West Bengal,4 later used in a case-control analysis,5 and four in Bangladesh,6–9 where exposures estimated individually were related to risk of skin lesions. Two studies6–7 reported very low risks below 50μg/L, rising to prevalences of 20–30% at higher concentrations. The two remaining studies by Ahsan et al.8 and Rahman et al.9 published in 2006 were both large and from circumscribed areas south-east of Dhaka, where arsenic contaminations are high. The former study, in Araihazar, reported a systematic increase in prevalence odds ratios (OR) compared to those in drinking water containing <8.1μg/L, of 1.91 at 8.1–40μg/L, rising to 5.39 at > 175.1μg/L. Rahman et al. did not present comparable data on exposure-response in Matlab, but the overall prevalence in adults was only 4/1000 after an average of some 20 years’ tube-well use, and a mean exposure of 167μg/L. The apparent difference between the findings in these two studies and others is addressed below.

Research objectives

Uncertainties over levels of risk at relatively low arsenic concentrations seriously concern Gonoshashthaya Kendra (GK), a large NGO long known internationally for innovations in health care.10 GK provides comprehensive services for the entire population of over a million in some 600 villages spread across much of the country, excepting the divisions of Khulna in the south and Sylhet in the north-east. Only 3% of GK’s villages are in districts with average concentrations above 50μg/L, and half in districts below 11μg/L. GK is thus well-placed to study the effects of relatively low average exposures, although such studies are complicated by the wide variations within and between villages. GK is also concerned that the risk is not confined to skin lesions, and in particular about possible effects in pregnancy and internal cancers, both of which it has begun to investigate. First, however, it was thought important to assess the risk of skin lesions at levels prevailing in the villages for which it is responsible and which are typical of much of Bangladesh.

Methods

Selection of villages, together with training and supervision of paramedics, was described in a preliminary report.11 Cases were defined as women with one or more nodules or characteristic skin thickening on palms or soles, recorded and graded by specially trained village paramedics. No account was taken of ulceration or lesions elsewhere, and questions of diagnosis or causation were deliberately avoided. Prevalence rates by age were calculated against average concentrations reported by the National Hydrochemical Survey (NHS)3 based on 129 wells tested in the upazillas (subdistricts) where the study villages were located.

To assess relative risk within villages, each of the 176 women with skin lesions identified in the initial survey was matched on age (+/– 5 years) with one unaffected referent (randomly chosen within age strata) in the same village. At a visit by one of this paper’s authors in 2005–6 to the 27 villages with one or more case-control pairs, tube wells currently used by subjects were identified and a record made of how long each woman had used this water source. Photographs were taken of the lesions in all reported cases for later evaluation, and three water samples from each well were tested using the Arsenator equipment used in the NHS, which provided a digital readout of arsenic concentration. In the NHS, a comparison was made between the Arsenator and British Geological Survey (BGS) laboratory results based on some 250 samples in the Manduri village survey.3 No evidence was found of any systematic difference between the two sets, details of which are presented in the NHS.

The same survey procedure was followed in the study of two large Rajshahi villages, A and B. As these villages were new to GK, a census of all residents was the first step. Of the total population listed (n = 11 670), 11 021 (94%) were examined by a paramedic, including children under five if the mothers had noticed any skin abnormalities. This survey sought to assess prevalence in males and females over the full age range. Several young children were seen, but none showed significant signs. A comprehensive study of arsenic levels in some 1400 tube wells in these two villages is in progress, with sites of each well and house addresses of each case, as determined by the Global Positioning System. These findings will be reported separately.

Statistical procedures

Prevalence (%) was calculated by age and upazilla for the 53 villages in the initial survey, and by age and sex in the study of the two special villages; χ² statistics with tests for trend were calculated to investigate differences in prevalence by upazilla grouped by arsenic concentration. Conditional logistic regression was used to determine the relation with exposure (highest of three concentrations recorded) in the case-control study. Prevalence odds ratios associated with age, sex and village were calculated by logistic regression in the report on the special villages.

Results

These three surveys’ essential findings are summarized in Tables 1, 2 and 3. Table 1 shows the prevalence of skin lesions was low (0.37%) among 6448 women living in upazillas A-E (25 villages) with an average arsenic concentration of 5μg/L or less. It was 0.63% among 5547 women in upazillas F-K (21 villages), average concentration 16–50μg/L, but very much higher (6.84%) among 1710 women in upazilla L (7 villages) with an average concentration of 81μg/L. While the range of average concentrations in the first group was narrow (0–9μg/L), that in the three higher groups (F-L) was very wide indeed (0–166μg/L). Recorded data from 33 wells in the same union helped to narrow the range (see Table 1); however, only the values from all wells measured in each upazilla are used in the present paper.

The case-control study, potentially based on 176 pairs (352 women), was finally reduced to 155 pairs by the loss of 21 cases: 14 women had moved elsewhere, and seven were unwilling to participate or not available. The results in Table 2 correlate well with those on prevalence in Table 1, showing a threefold increase at over 50μg/L (P < 0.05), and some indication of an increase above 10μg/L. As cases and controls were matched for village and age, the relative risks show only the effect of differing exposure within and not between villages, and not in duration to the extent that age is a reliable surrogate. A further analysis of pairs aged over and under 40 years (not shown) suggests that the relative risks were similar in older and younger women.

Table 3, based on the two large villages, shows that the overall prevalence was over twice as high in Village B (3.2%) as in Village A (1.3%); it was fairly similar in men and women in Village B, but less so in Village A. Prevalence in both villages rose sharply with age, with rates somewhat higher in men than in women. Cases in children were very rare. A logistic regression, excluding those aged 5 years or less (and those with unknown age) confirmed (P < 0.001) the higher risk in village B (OR = 2.44, 95% CI: 1.74–3.41) and the increasing risk with age (OR = 1.034, 95% CI: 1.027–1.042): the observed slight increase in prevalence in men was not significant in this model (OR = 1.19, 95% CI: 0.87–1.62). In women over 30, in Village B, the average prevalence (5.2%) approximated that in upazilla L (Table 1). Of the 10 wells recorded in the upazilla in which these two villages A and B were located, only one showed a concentration above 1.1μg/L (57.8μg/L in Village A). In our ongoing survey, however, many wells in both villages had concentrations above 100μg/L.

Discussion

The main epidemiological problems in assessing the arsenic hazard in Bangladesh are in exposure estimation at the individual or household level, and in identifying potentially related disease. The commonly observed gradient with age in prevalence studies suggests that except perhaps in pregnancy, risk is determined by both arsenic concentration in drinking water and duration of consumption over at least 20 years. To assemble such data reliably for a large study population would be extremely difficult and, as mentioned earlier, was attempted by Haque et al.5 in West Bengal, where the mean arsenic concentration was 185μg/L (range 0–3400μg/L). A nested case control study limited to 21 villages in which the primary drinking-water sources contained < 500 μg/L had only limited success in estimating past exposure, but the odds ratio in relation to peak concentration of < 50μg/L was 2.4 at 50–99μg/L, a result close both to our own from a similar nested study (2.96) and to that reported by Ahsan8 (3.03) for exposures in the range 40–91μg/L.

Because the assessment of individual exposures is so difficult, we opted initially for an ecological approach for assessing prevalence, using data from the National Hydrochemical Survey.3 This survey of almost 4000 wells used a systematic grid resulting in one well tested each 37km², but amounted to only about 60 wells per district and 8 per upazilla for the calculation of averages. Although the geographical pattern of arsenic concentrations after statistical smoothing appeared clear, this obscured enormous local variations. For example, in three special survey areas concentrations ranged from < 3 to 2542μg/L. Even in the upazilla where Villages A and B are located, despite an average of 6μg/L, the only well tested in Village A had a level of 57.8μg/L. Thus the ecological approach, though useful, has limitations.

A further difficulty lies in the ascertainment of skin lesions. Whereas we relied on paramedics to describe objectively the results of simple inspection, other studies4–8 have used physicians exercising their varying levels of judgement, and in one9 (Matlab, Bangladesh) expert panels of physicians and dermatologists were used to reach consensus on the diagnosis of arsenical keratosis. This resulted in rejection of 70% of cases reported by the field workers. This alone may largely explain the differences between their rates and others, including ours, though it is worth noting that the average exposure in Matlab was 167μg/L, and only our upazilla L (Table 1) had values approaching that level, though with a wide range in values for the 10 wells tested (2–166μg/L). Also relevant is the fact that the average duration of tube-well use in Matlab was about 20 years, whereas most of the wells in upazilla L were reported to date from 1953, up to 50 years earlier than our survey.

An important purpose of the 53-village prevalence study was to evaluate the published ecological data3 and adequacy of the national standard of 50μg/L in assuring safety. Table 1 suggests that the number of cases in groups A-E was probably negligible, and in group L, in clear excess. The interpretation of groups F-I and J-K is more difficult. The prevalence rates are almost double the lowest group (P = 0.04), but the average concentrations in both are below 50μg/L. However, in these groups there is evidence that in 14 of 61 wells (23%) the level was above 50μg/L, suggesting a potential risk. The exposures used for classifying subjects for our case-control analysis were measured by Arsenator, and so were comparable to the BGS data used to classify the 53 villages in Table 1. A detailed analysis, not shown, of the 310 well measurements used for Table 2 indicated that most of the values above 50μg/L were from subjects in upazilla L, and only a few from upazillas F-K.

It was said that arsenic levels and skin lesion prevalence were higher in the upper socioeconomic sections of the Matlab populations.9 We therefore examined this question using GK’s plentiful data on social factors. It was not possible to do this for the villages individually, but we examined the main geographical groups to which our 53 villages belonged and selected as the best available index the proportion of pregnant women in the last two years classified as poor or very poor. The resulting proportions in the four arsenic concentration categories used in Table 1 were as follows: A-E 77%; F-I 86%; J-K 88%; L 78%. There was clearly no systematic trend. Other confounders such as smoking, occupation and sunlight were identified by Chen et al.12 as important in explaining the higher prevalence in men. This was not considered important in our 53-village study, as this was confined to adult women.

Our studies have both strengths and weaknesses, the most important of the former being that they have successfully evaluated risk in the total adult female population of 53 villages, randomly selected and widely scattered, at moderate levels of exposure. However, the skin lesions recorded by trained paramedics were not further validated. We are confident nevertheless that any excess risk in areas of the country where the average arsenic concentration was below 10μg/L is most unlikely. There is evidence of a small risk in areas with averages between 11 and 50μg/L, probably explained by wells in which 50μg/L was exceeded. Above 50μg/L, though our data at higher concentrations are scanty, the risk appears substantial. It is reasonably certain that the prevalence of skin lesions recorded in three of the seven villages in upazilla L was very high (102 cases in 772 women – 13%), a sample of whom were examined and confirmed by three of us (NC, NH and JCMcD). This level of risk was quite compatible with rates recorded in villages A and B of the supplementary survey in Rajshahi (Table 3).

If we are correct in these conclusions, it follows that public health policy in areas of the country at relatively low exposure should give priority to ensuring that the concentration of arsenic in water actually used for drinking is kept below 50μg/L. This would require reliable and systematic testing, retesting, marking and perhaps closing of all offending hand-pump tube wells, together with continued skin lesions surveillance. Such measures should be possible in most of the country, but in areas where the average concentration exceeds 50μg/L this policy may be inadequate or unfeasible, in which case more radical measures would be needed; these are beyond the scope of this paper. However, until it is known whether the lifetime ingestion of water containing arsenic below 50μg/L carries a significant risk of internal cancers or important adverse effects in pregnancy, no public health policy can be final. ■

Ethical issues

The study protocol was formally approved by the Ethics Committee of the Royal Brompton Hospital, London. In addition, research plans were approved by the GK Executive Committee, the body responsible for all health services in GK villages.

References

Affiliations

  • Department of Occupational and Environmental Medicine, National Heart and Lung Institute, Imperial College, London, UK.
  • Gonoshasthaya Kendra, Nayarhat, Savar, Dhaka, Bangladesh.
  • Institute of Child and Mother Health, Matuail, Dhaka, Bangladesh.
  • Community and Occupational Medicine Program, University of Alberta, Edmonton, Canada.
Share
Email Twitter Facebook Google Delicious LinkedIn
More...