Public Health Strategies for Western Bangladesh That Address Arsenic, Manganese, Uranium, and Other Toxic Elements in Drinking Water

Background More than 60,000,000 Bangladeshis are drinking water with unsafe concentrations of one or more elements. Objectives Our aims in this study were to evaluate and improve the drinking water testing and treatment plans for western Bangladesh. Methods We sampled groundwater from four neighborhoods in western Bangladesh to determine the distributions of arsenic, boron, barium, chromium, iron, manganese, molybdenum, nickel, lead, antimony, selenium, uranium, and zinc, and to determine pH. Results The percentages of tube wells that had concentrations exceeding World Health Organization (WHO) health-based drinking water guidelines were 78% for Mn, 48% for U, 33% for As, 1% for Pb, 1% for Ni, and 1% for Cr. Individual tube wells often had unsafe concentrations of both Mn and As or both Mn and U. They seldom had unsafe concentrations of both As and U. Conclusions These results suggest that the ongoing program of identifying safe drinking water supplies by testing every tube well for As only will not ensure safe concentrations of Mn, U, Pb, Ni, Cr, and possibly other elements. To maximize efficiency, drinking water testing in Bangladesh should be completed in three steps: 1) all tube wells must be sampled and tested for As; 2) if a sample meets the WHO guideline for As, then it should be retested for Mn and U; 3) if a sample meets the WHO guidelines for As, Mn, and U, then it should be retested for B, Ba, Cr, Mo, Ni, and Pb. All safe tube wells should be considered for use as public drinking water supplies.

Residents who drink water contaminated with As are at risk for developing dermatologic dis eases, skin cancers, and internal cancers and for adverse pregnancy outcomes and increased mortality (Ashraf et al. 2004;Frisbie et al. 2005). The Bangladeshi government, non governmental organizations, and the scientific community have responded by instituting widespread drinking water testing for As, as well as education programs designed to inform the populace about the dangers of drinking Ascontaminated water. As a result, approxi mately 5,000,000 of the country's 10,000,000 tube wells have been tested for As [UNICEF (United Nations Children's Fund) 2007], and increasing numbers of villagers are becoming aware of the health risks associated with drink ing Ascontaminated water (Parvez et al. 2006).
Routine testing of drinking water for As is crucial for promoting public health in Bangladesh. However, two nationalscale surveys of tube well water for other toxic ele ments revealed that As, manganese (Mn), ura nium (U), boron (B), barium (Ba), chromium (Cr), molybdenum (Mo), nickel (Ni), and lead (Pb) are found at concentrations that exceed World Health Organization (WHO) healthbased drinking water guidelines (BGS/ DPHE 2001;Frisbie et al. 2002). Our 2002 study was prompted, in part, by clinical observations that certain As patients had more severe symptoms than would be expected given the levels of As in their drinking water, suggesting possible synergistic effects from other toxins, such as antimony (Sb), as well as deficient quantities of beneficial elements such as selenium (Se) and zinc (Zn).
Although there is much ongoing research about the distribution of As in the geologic materials of the region (Bhattacharya et al. 2002), the distribution of the other toxic elements commonly found in the region's drinking water has received much less atten tion. An essential question for those charged with ensuring public health is whether drink ing water with As concentrations that meet national or WHO criteria can be designated as safe without further testing for other toxic elements. It is crucial for public health policy to determine whether the concentrations of other commonly occurring toxic elements are correlated with the concentration of As. If they are, then the current practice of testing every tube well for As only might be sufficient to identify safe drinking water supplies. If the concentrations of these other toxicants are not positively correlated with the concentration of As, then testing every tube well for As alone will not identify safe drinking water supplies. Drinking water must be safe with respect to As and all other toxic elements.

Sample collection, preservation, and analyses.
We collected groundwater samples from four neighborhoods in western Bangladesh ( Figure 1). Western Bangladesh was chosen for this study because it has some of the wid est ranges of groundwater As concentrations in the country, according to our two national scale surveys (Frisbie et al. 1999USAID 1997). Therefore, it is a region where both drinking water testing and treatment for As are important public health strategies. We selected these neighborhoods at random within this region.
Seventyone samples were collected from 67 randomly selected tube wells in these four neighborhoods. We collected a total of 18 sam ples from 17 tube wells in each of three neigh borhoods (Bualda, Fulbaria, and Jamjami). We were denied access at one sampling loca tion; therefore, 17 samples were collected from 16 tube wells in the fourth neighbor hood (Komlapur). To the extent possible, the sampled tube wells in each neighborhood were distributed at 500m intervals along perpen dicular axes that radiated in four equal lengths from the center (Figure 1). Two samples were collected from the centermost tube well in each neighborhood. We averaged the results for each analyte from each of these four centermost tube wells. One sample was collected from each of the remaining tube wells. The northings and eastings of these tube wells were measured using a Global Positioning System 12 Channel Personal Navigator (Garmin International, Olathe, KS, USA).
We used established collection, preserva tion, and storage methodologies to ensure that each sample was representative of ground water quality [American Public Health Association (APHA) et al. 2005;Frisbie et al. 2005]. Accordingly, all sampled tube wells were purged by pumping vigorously for 10 min immediately before sample collection. All samples were collected directly into polyeth ylene bottles and were not filtered. Samples were analyzed immediately after collection with pH paper, preserved by acidification to pH < 2 with 5.0 M hydrochloric acid (prod uct no. 101256J; BDH Laboratory Supplies, Poole, UK), and stored in icepacked cool ers. The temperature of all stored samples was maintained at 0-4°C until immediately before analysis at laboratories in Dubai, France, and Vermont.
Samples were shipped to Dubai and ana lyzed for As by the arseno molybdate method (Frisbie et al. 2005). The samples were then shipped to France and analyzed for Ba, Cr, Mn, Mo, Ni, Pb, Se, U, and Zn by induc tively coupled plasma mass spectrometry (PlasmaQuad PQ2+ Spectrometer; Fisons/ VG Analytical, Manchester, UK) (APHA et al. 2005). Finally, the samples were shipped to Vermont and analyzed for B by the azo methine H method, iron (Fe) by flame atomic absorption spectroscopy (210VGP Atomic Absorption Spectrometer; Buck Scientific, East Norwalk, CT, USA) (APHA et al. 2005), and Sb by graphite furnace atomic absorption spectroscopy (210VGP; Atomic Absorption Spectrometer) (APHA et al. 2005).
Interviews. The depth, age, and number of users were determined for each tube well by interviewing its owner or a principal user at the time of groundwater sampling. The interview was conducted in Bangla using a list of standard questions.
Mapping and statistics. Contour maps were drawn by hand, using linear interpola tion, to show the concentrations of As, B, Ba, Cr, Fe, Mn, Mo, Ni, Pb, Sb, Se, U, and Zn, as well as pH, depth of tube well, age of tube well, and number of users per tube well, for each of the four neighborhoods. We interpreted these maps visually to help make hypotheses about the effects of geology on the distributions of these elements in ground water. We used standard methods of linear regression to test these hypotheses (Neter et al. 1985;Snedecor and Cochran 1982).

Results and Discussion
The distributions and health risks of toxic elements. All 71 groundwater samples from Bualda, Fulbaria, Jamjami, and Komlapur were analyzed for every toxic element that has ever been found to exceed WHO healthbased guidelines in Bangladesh's drinking water: As, B, Ba, Cr, Mn, Mo, Ni, Pb, and U (BGS/ DPHE 2001;Frisbie et al. 2002). In this study, we found concentrations of As, Cr, Mn, Ni, Pb, and U that exceeded WHO health based drinking water guidelines. Conversely, we found that B, Ba, and Mo levels did not exceed these guidelines. In addition, we ana lyzed all samples for Fe, Sb, Se, Zn, and pH (Tables 1, 2). A list of all of these elements follows, with elements arranged from the most to the least significant health risk in this study. We then summarize the toxicity of these ele ments and review the rationale for WHO healthbased drinking water guidelines.
Arsenic. Chronic As poisoning is the most significant health risk caused by drink ing water from these four neighborhoods. Arsenic concentrations ranged from < 7 µg/L to 590 µg/L, with 33% of tube wells exceeding the 10 µg/L WHO drinking water guideline (Table 1; WHO 2004WHO , 2006. Drinking water with 10 µg/L As has been associated with three extra deaths per 5,000 people from skin cancer (WHO 1996a(WHO , 1996b and 10 extra deaths per 5,000 people from bladder, liver, or lung can cer (Morales et al. 2000). In addition to these cancers, chronic As poisoning has been asso ciated with melanosis, leuko melanosis, kera tosis, hyper keratosis, and non pitting edema in Bangladesh (Frisbie et al. 2005).
Manganese. Mn concentrations ranged from 160 µg/L to 2,400 µg/L, with 78% of tube wells exceeding the 400 µg/L WHO healthbased drinking water guideline (Table 1) (WHO 2004(WHO , 2006. Mn is required for human nutrition; however, the accumula tion of Mn may cause hepatic encephalopa thy in humans (Layrargues et al. 1998). The chronic ingestion of Mn in drinking water is associated with neurologic damage in humans (Kondakis et al. 1989;WHO 1996aWHO , 1996b. The WHO guideline for Mn in drinking water was calculated using the no observed adverse effects level (NOAEL) for these neu rologic effects in humans and laboratory ani mals (WHO 1996b(WHO , 2004. As worldwide life expectancy increases, chronic neurologic dis eases such as parkinsonian disorders associated Table 1. The average concentrations of toxic elements in the groundwater of Bualda, Fulbaria, Jamjami, and Komlapur, the WHO health-based drinking water guidelines for these toxicants, and the percent of tube wells exceeding these guidelines.

Average
WHO health-based Percent of unsafe Element concentration (µg/L) guideline (µg/L) tube wells a  with Mn exposure are likely to increase, espe cially in developing countries (Dorsey et al. 2007;Ferri et al. 2005;He et al. 2005). Thus, high intake of Mn by Bangladeshis may increase parkinsonian dis orders associated with Mn exposure. Uranium. U concentrations ranged from < 0.2 µg/L to 10 µg/L, with 48% of tube wells exceeding the 2 µg/L WHO health based drinking water guideline (Table 1). This WHO guideline was calculated using the low est observed adverse effects level (LOAEL) for kidney lesions in male laboratory rats (WHO 1998a(WHO , 1998b. The carcinogenic effect of U in drinking water at natural isotopic abundance ( 238 U at 99.2830%, 235 U at 0.7110%, and 234 U at 0.0054%) has not been adequately studied in humans and experimental animals (Weast et al. 1983;WHO 1998aWHO , 1998b. The first study on humans of the effects of chronic U ingestion from drinking water showed adverse kidney function, with the proxi mal tubule as the site of toxicity (Zamora et al. 1998). Later, a much larger study on exposure of humans to U in drinking water revealed nephro toxic effects even at low con centrations without a clear threshold (Kurttio et al. 2002). In another study, the same authors found that people who drank water with elevated concentrations of U had indica tions that, in addition to kidneys, bone may be another target of toxicity (Kurttio et al. 2005).
Lead. Pb concentrations ranged from < 0.2 µg/L to 17 µg/L, with 1% of tube wells exceeding the 10 µg/L WHO healthbased drinking water guideline (Table 1) (WHO 2004(WHO , 2006. The WHO drinking water guide line for Pb was calculated using the lowest meas urable retention of Pb in the blood and tis sues of human infants (WHO 1996a(WHO , 1996b). Pb is a "possible human carcinogen" because of inconclusive evidence of human carcino genicity and sufficient evidence of animal carcino genicity. Oral exposure to Pb has been found to increase the incidence of renal tumors in laboratory rats, mice, and hamsters (WHO 1996a(WHO , 1996b(WHO , 2004(WHO , 2006. In addition, Pb also causes many non carcinogenic disorders in humans, including, but not limited to, "neuro toxicity, developmental delays, hypertension, impaired hearing acuity, impaired hemoglobin synthesis, and male reproductive impairment" [U.S. Environmental Protection Agency (EPA) 2008]. The effects of Pb on the central nervous system of fetuses, infants, children up to 6 years of age, and pregnant women can be especially serious (WHO 1996a(WHO , 1996b.
Nickel. Ni concentrations ranged from 0.5 µg/L to 570 µg/L, with 1% of tube wells exceeding the 70 µg/L WHO healthbased drinking water guideline (Table 1). This WHO guideline was calculated using the LOAEL in a study of oral exposure in fast ing patients (WHO 2006). Ni compounds are "carcinogenic to humans" by inhalation exposure. In contrast, the carcinogenic effects of Ni in drinking water for humans have not been adequately studied. Ni in drinking water did not increase the incidence of tumors in laboratory rats (WHO 1998a(WHO , 1998b(WHO , 2006.
Chromium. Total Cr concentrations ranged from < 0.5 µg/L to 100 µg/L, with 1% of tube wells exceeding the 50 µg/L WHO drink ing water guideline (Table 1) (WHO 2004(WHO , 2006. The International Agency for Research on Cancer (IARC) has cate gorized Cr(VI) as "carcino genic to humans" and Cr(III) as "not classifiable" (IARC 1987); however, the U.S. EPA (1996) listed total Cr in drinking water as having "inadequate or no human and animal evidence of carcinogenicity." The WHO has stated that the 50 µg/L drinking water guide line for total Cr is unlikely to cause significant health risks (WHO 1996a(WHO , 1996b. Boron. B concentrations ranged from < 50 µg/L to 440 µg/L, with no tube wells exceeding the 500 µg/L WHO healthbased drinking water guideline (Table 1) (WHO 2004(WHO , 2006. However, 5.3% of Bangladesh's tube wells exceeded this guideline in a nationalscale survey (BGS/ DPHE 2001).
Iron. Fe concentrations ranged from < 40 µg/L to 66,000 µg/L ( Table 1). The WHO has not established a healthbased drinking water guideline for Fe (WHO 2004(WHO , 2006. However, high body Fe stores and high dietary intakes of Fe are associated with hepato cellular carcinoma in humans (Marrogi et al. 2001) and mammary carcinogenesis in female SpragueDawley rats (Diwan et al. 1997). Bangladeshis ingest approximately 12%, 62%, and 26% of their dietary Fe from drinking water, eating rice, and ingesting soil, respectively; in Bangladesh, Fe is ingested at almost twice its recom mended dietary allow ance (Ortega et al. 2003).
Molybdenum. Mo concentrations ranged from 0.5 µg/L to 7.8 µg/L, with no tube wells exceeding the 70 µg/L WHO healthbased drinking water guideline (Table 1) (WHO 2004(WHO , 2006. In contrast, an unspecified per centage of Bangladesh's tube wells exceeded this WHO guideline in a nationalscale survey (BGS/ DPHE 2001).
Antimony. Sb concentrations ranged from < 0.5 µg/L to 6.2 µg/L, with no tube wells exceeding the 20 µg/L WHO healthbased drinking water guideline (Table 1) (WHO 2004(WHO , 2006. However, 81% of the samples with detectable concentrations of As had detectable concentrations of Sb (Table 2). Sb in drinking water has been reported to Table 2. Correlation coefficients (r) for the concentrations of toxic elements in tube well water from Bualda, Fulbaria, Jamjami, and Komlapur, along with characteris tics of these tube wells. modulate the toxicity of As (Gebel 1999). Therefore, it is possible that otherwise safe levels of Sb may magnify As toxicity. Sb trioxide (Sb 2 O 3 ) is "possibly carcino genic to humans" by inhalation exposure. In contrast, the effect of Sb in drinking water on cancer in humans has not been adequately studied. Sb in drinking water did not increase the incidence of tumors in laboratory mice and rats (WHO 1996a(WHO , 1996b(WHO , 2004(WHO , 2006. The WHO guideline for Sb in drinking water was calculated using the NOAEL for decreased water intake, food intake, and body weight in laboratory rats (WHO 2004(WHO , 2006.

As
Selenium. Se concentrations ranged from < 1 µg/L to 1 µg/L, with no tube wells exceed ing the 10 µg/L WHO guideline (Table 1) (WHO 2004(WHO , 2006. Se is needed for human nutrition. Se does not appear to cause can cer, with the exception of Se sulfide, which is not found in drinking water (WHO 1996a(WHO , 1996b. The NOAEL for Se in humans is 4 µg/ kg body weight per day. In this light, the WHO set the healthbased guideline for Se in drinking water at 10 µg/L (WHO 2004(WHO , 2006. Se prevents the cytotoxic effects of As (Biswas et al. 1999). Unfortunately, the food crops in Bangladesh are sometimes deficient in Se (Ortega et al. 2003), and the drinking water in Bangladesh is often deficient in Se . Therefore, it is possible that this lack of Se in food and drinking water might magnify As toxicity.
Zinc. Zn concentrations ranged from 2.6 µg/L to 88 µg/L (Table 1). Zn is needed by all living organisms. The provisional maxi mum tolerable daily intake for Zn in humans is 1,000 µg/kg body weight. In this light, the WHO concluded that a healthbased guide line for Zn in drinking water "is not required" (WHO 2004(WHO , 2006. In Bangladesh, the severity of chronic As poisoning may be magnified by a lack of dietary Zn Ortega et al. 2003). Zn promotes the repair of tissues dam aged by As (Engel et al. 1994). Food, not drinking water, is the major source of dietary Zn (WHO 1996a), but the agricultural soils, food crops, and diet in Bangladesh are often deficient of Zn (Brammer 1996;Ortega et al. 2003). Therefore, it is possible that this lack of Zn in soils, food, and drinking water may magnify As toxicity.
Ramifications for the monitoring, treatment, and distribution of drinking water. The average concentrations of toxic elements from all 67 tube wells sampled in this study are listed in Table 1. Thirtythree percent (22 of 67) of these tube wells exceed the WHO healthbased drinking water guideline for As of 10 µg/L ( Table 1).
Analysis of tube wells with unsafe concentrations of As. The average concentrations of toxic elements from the 22 tube wells with unsafe concentrations of As are listed in Table 3. That is, 59%, 14%, 5%, 5%, and 5% of these 22 tube wells had unsafe concentra tions of Mn, U, Pb, Ni, and Cr, respectively (Table 3). This suggests that drinking water wells with unsafe concentrations of As may also have unsafe concentrations of Mn, U, Pb, Ni, Cr, or possibly other elements.
In this neighborhoodscale study and in two nationalscale studies of Bangladesh, levels of As, Mn, U, Pb, Ni, Cr, B, Ba, and Mo were above WHO healthbased drinking water guidelines (Table 1) (BGS/DPHE Frisbie et al. 2002). In Bualda, increases in As concentration correlated with statistically sig nificant increases in concentrations of Mn, Pb, Ni, Cr, and B (Table 4). In Jamjami, increases in As concentration correlated with statisti cally significant increases in concentrations of Pb, Ni, and Ba (Table 4). In Komlapur, increases in As concentration correlated with statistically significant increases in Cr and Ba (Table 4). Finally, in the entire region, increases in As concentration correlated with statistically significant increases in Mn, Pb, Cr, B, Ba, and Mo (Table 2).
Almost all of the homescale drinking water treatment systems currently being used in Bangladesh have been designed to remove As but not these other toxic elements. The sta tistically significant increases in toxic elements in addition to As suggest that these treatment systems should be further evaluated for the removal of Mn, Pb, Ni, Cr, B, Ba, Mo, and possibly other elements.
Analysis of tube wells with safe concentrations of As. The average concentrations of toxic elements from the 45 tube wells with safe concentrations of As are presented in Table 5. Of these 45 tube wells 87% and 64% had unsafe concentrations of Mn and U, respectively (Table 5). In fact, 93% (42 of 45) of these tube wells had unsafe concentrations of Mn, U, or both Mn and U (Table 5). This suggests that drinking water wells with safe concentrations of As may have unsafe concen trations of Mn, U, or possibly other elements. Thus, the current practice of testing every tube well only for As will not identify drink ing water with safe concentrations of other toxic elements.
In response to this finding that Mn, U, and possibly other toxic elements commonly occur at unsafe concentrations even when As is at safe concentrations, we propose the fol lowing threestep testing program to provide safe drinking water in western Bangladesh, and possibly the entire country. This testing program is economical because it prioritizes the analysis of toxic elements, and analysis ends as soon as a sample is found to be unsafe for use as drinking water.  (WHO 1996a(WHO , 1998a. c The severity of chronic As poisoning in Bangladesh might be magnified by a lack of Se or Zn or both Ortega et al. 2003). Table 4. Correlation coefficients (r) for the concentration of As versus the concentrations of toxic elements in tube well water from each of the four neighborhoods in this study, along with the characteris tics of these tube wells. First, the toxicity and distribution of As relative to Mn, U, Pb, Ni, Cr, B, Ba, and Mo suggest that the current practice of sampling and testing every tube well in Bangladesh for As to find the safest sources of drinking water should remain the highest public health priority. Arsenic is expected to cause at least 150,000 extra cancer deaths during the life spans of the current population of Bangladesh (Frisbie et al. 2005). In contrast, the risk to public health in Bangladesh is smaller for Mn, U, Pb, Ni, Cr, B, Ba, and Mo WHO 1996bWHO , 1998b. Under condi tions of limited resources, testing of these toxic elements must be prioritized.

As
Second, the high concentrations of As, Mn, and U relative to Pb, Ni, Cr, B, Ba, and Mo suggest that if a sample meets the WHO guideline for As, it should be retested for Mn and U. This will identify tube wells with safe concentrations of As, Mn, and U for additional evaluation as a potential drinking water supply in these neighborhoods without the cost or delay of testing for all nine ele ments. For example, one tube well in Fulbaria, one tube well in Jamjami, and one tube well in Komlapur did not exceed WHO healthbased drinking water guidelines for As, Mn, and U.
Third, if a sample meets the WHO guide lines for As, Mn, and U, then it should be retested for Pb, Ni, Cr, B, Ba, and Mo. All tube wells that do not exceed WHO guide lines for these nine elements could be used as public drinking water supplies. For example, if the three tube wells that did not exceed WHO healthbased drinking water guidelines for As, Mn, and U also did not exceed any other WHO healthbased drinking water guidelines, they could supply safe drinking water to the residents of each neighborhood.
Testing only for As and then asking the owners of safe tube wells to share drinking water with their less fortunate neighbors has been a highly successful public health strategy in Bangladesh. More than 90% of western Bangladeshis share drinking water (Frisbie et al. 2005). The threestep testing program builds on this success by testing for all known toxic elements in Bangladesh's drinking water, not just As.
Unfortunately, no tube wells in Bualda met WHO guidelines for all elements; therefore, drinking water treatment will likely be required in this neighborhood. However, this testing strategy will help the residents of places like Bualda choose the safest tube wells for interim use until a treatment plant can be built.
All tube wells identified as safe by this threestep process should be used as public drinking water supplies. These safe tube wells must be periodically monitored for As, Mn, U, Pb, Ni, Cr, B, Ba, and Mo. If a tube well becomes unsafe, then an alternative drinking water supply must be identified or the unsafe water must be treated.
Our earlier nationalscale survey suggested that groundwater with unsafe levels of As, Mn, U, Pb, Ni, Cr, B, Ba, and Mo extends beyond Bangladesh's borders into the four adjacent and densely populated Indian states of West Bengal, Assam, Meghalaya, and Tripura . The present neighborhoodscale sur vey in western Bangladesh borders the West Bengal districts of Nadia and 24Parganas, where aquifers with similar charac teristics occur (Bhattacharya et al. 2002). Thus, we urge that a similar survey be done in West Bengal to investigate possible exposure to unsafe levels of Mn, U, Pb, Ni, Cr, B, Ba, and Mo in addition to As in drinking water.
The relationships among As, Mn, and U. The results from Tables 3 and 5 suggest that Mn is often at unsafe concentrations in  (WHO 1996a(WHO , 1998a. c The severity of chronic As poisoning in Bangladesh might be magnified by a lack of Se or Zn or both Ortega et al. 2003 Bangladesh's tube well water. More than 50% of Bangladesh's area has groundwater with Mn concentrations greater than the WHO healthbased drinking water guideline . In addition, the contrast between 14% of tube wells with unsafe concentrations of U among the tube wells with unsafe con centrations of As (Table 3) and 64% of tube wells with unsafe concentrations of U among the tube wells with safe concentrations of As (Table 5) suggests that in western Bangladesh, drinking water with safe concentrations of U may have unsafe concentrations of As, whereas drinking water with safe concentra tions of As may have unsafe concentrations of U. In summary, the drinking water in these neighborhoods generally has unsafe levels of As and Mn, or U and Mn; however, it seldom (4%, 3 of 67 tube wells) has unsafe concentra tions of both As and U together. Figures 2-4 illustrate the relationships between As and Mn, U and Mn, and As and U.
The inverse trend between As and U may be caused by the variability that is characteristic of deltaalluvial plain deposits from the Bengal Delta Plain in Bangladesh and West Bengal, India. For example, in Jamjami the concentra tion of As decreases with depth (p = 0.002; Figure 2), and the concentration of U increases with depth (p = 0.04; Figure 4). Komlapur, to some extent, also shows these trends. In contrast, Bualda and Fulbaria show no trends between As and depth, and U and depth. The aquifers in Jamjami and possibly Komlapur contain medium to coarsegrained sand at depth that was deposited in former river chan nels (Alam et al. 1990). The ground water drawn into tube wells that are screened in these deposits may be under oxidizing conditions that remove As from ground water and release U into ground water. In contrast, the aquifers in all four neighborhoods have organicrich mud at all depths that was deposited in flood plains (Alam et al. 1990). The groundwater drawn into tube wells that are screened in these deposits may be under reducing con ditions that release As into groundwater and remove U from groundwater. Therefore, allu vial sediments of the Bengal Delta Plain make a complex threedimensional stratigraphy of medium to coarsegrained sand and organic rich mud deposits that may be responsible for the inverse trend between As and U. Other fac tors may also be controlling release of As and U. It is important to note that in areas where drilling deeper tube wells may access water with lower levels of As, the water from these deeper tube wells may contain increased levels of U, as we found in Jamjami and Komlapur.
Despite this inverse trend, 4% (3 of 67) of the tube wells in this study had unsafe concen trations of both As and U. This is important because the homescale drinking water filters that are being used in Bangladesh may not remove U. Also, up to 50% of Bangladesh's tube wells exceed the WHO healthbased drinking water guideline for U (BGS/ DPHE 2001). The water treatment filters used in Bangladesh typically oxidize soluble As(III) to insoluble As(V) to remove As by absorption or precipitation. However, this oxidation may convert insoluble U(IV) to soluble U(VI) and potentially increase the U concentration of the water after treatment. Alternatively, this oxida tion may keep dissolved U in the VI oxidation state and potentially cause no change in the U concentration of the water after treatment (Fairbridge 1972). Thus, these filters should be further evaluated for the removal of U.

Conclusions
In this neighborhoodscale study and in two nationalscale studies of drinking water tube wells in Bangladesh, concentrations of As, Mn, U, Pb, Ni, Cr, B, Ba, and Mo exceeded WHO healthbased guidelines (Table 1) (BGS/DPHE 2001;Frisbie et al. 2002). In the present study, 96% of the tube wells exceeded WHO healthbased guidelines for at least one of these toxic elements. The single greatest risk to public health is from As in drinking water.
Of the 67% of tube wells that had As concen trations below the WHO drinking water guideline, 87% had unsafe levels of Mn and 64% had unsafe levels of U (Table 5). Thus, testing for As alone is not sufficient to ensure safe drinking water. To address the threats to public health posed by the prevalence of multiple toxic elements, we have proposed a threestep drinking water testing program.
Of the 33% of tube wells that had As concentrations greater than the WHO drink ing water guideline, 59% also had unsafe levels of Mn, 14% had unsafe levels of U, 5% had unsafe levels of Pb, 5% had unsafe levels of Ni, and 5% had unsafe levels of Cr (Table 3). Thus, water treatment systems that have been designed solely for As removal may not pro vide safe drinking water and should be further evaluated for the removal of Mn, U, Pb, Ni, Cr, B, Ba, and Mo.