Article Text

Download PDFPDF

Lung cancer and exposure to arsenic in rural Bangladesh
  1. M G Mostafa1,
  2. J C McDonald2,
  3. N M Cherry3
  1. 1
    National Institute of Cancer Research and Hospital, Dhaka, Bangladesh
  2. 2
    National Heart and Lung Institute, Imperial College, London, UK
  3. 3
    University of Alberta, Edmonton, Alberta, Canada
  1. Nicola Cherry, 12–102 Clinical Sciences Building, Edmonton, Alberta T6G 2G3, Canada; nicola.cherry{at}


Background: Arsenic is a known carcinogen but the risk of lung cancer from the widespread contamination of drinking water in rural Bangladesh has not been estimated.

Objectives: To determine whether estimated exposure of villagers in Bangladesh to arsenic in drinking water differed between those with lung cancer and those with non-malignant lesions.

Methods: Data were obtained from 7286 subjects who underwent lung biopsy in 2003–2006 at a diagnostic centre taking referrals from throughout Bangladesh. Analysis was limited to 5372 people living in villages for the last 10 years who reported using tube well water. Of these, 3223 with a primary lung tumour were enrolled as cases and 1588 with non-malignant lesions as referents in an unmatched analysis. Arsenic exposure was estimated by average concentrations for each of 64 districts. Logistic regression was used to test the effects of age, arsenic and smoking on risk and to investigate relationship to cell type.

Results: Male cases were older than referents and more likely to smoke, to smoke >20 units/day and to smoke bidi—small, hand-rolled cigarettes. Odds ratios for lung cancer increased steadily with mean arsenic concentration, but the confidence interval excluded 1.0 only at concentrations >100 μg/l (OR 1.45, 95% CI 1.16 to 1.80). This trend was seen only in smokers where the increased risk at >100 μg/l was 1.65 (95% CI 1.25 to 2.18). A similar trend was seen in women smokers. Squamous cell lung cancer was more frequent in smokers and, having adjusted for smoking, in districts with arsenic concentrations >100 μg/l.

Conclusions: Among Bangladeshis who smoke, those whose drinking water is contaminated with arsenic at concentrations >100 μg/l are at increased risk of lung cancer. With high levels of exposure misclassification and short latency of exposure, the study cannot estimate or exclude the likely long term risk in non-smokers and at lower arsenic concentrations.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

The International Agency for Research on Cancer concluded in 2004 that arsenic in drinking water was probably causally related to lung cancer in man.1 However, the studies cited were limited both as regard geographical area (Taiwan, Chile) and in information on potential confounders.2 3 The largest populations exposed to high concentrations of arsenic through drinking water live in Bangladesh and West Bengal (India), but the risk of lung cancer in these areas has not yet been estimated and may differ because of other water borne contaminants, smoking habits and possibly genetic factors. Confirmation of a causal relationship with ingested arsenic is important scientifically, in that only inhaled carcinogens have previously been related to primary lung cancer. It is also a major public health issue in a region with very limited health care capacity. The interaction between arsenic and tobacco use is uncertain: previous studies of environmental (ingested) and occupational (inhaled) arsenic suggest a synergistic effect, with tobacco and arsenic having a risk that is greater than additive.24 The present study addresses these questions further in patients with lung opacities investigated in a single diagnostic centre in Dhaka, Bangladesh. Our objective is to determine whether subjects whose biopsy indicated a primary lung cancer were more likely to have lived in a village with high arsenic contamination than those whose biopsy result was benign, and to assess the role of tobacco use in this relationship.



The Anowara Diagnostic Center, established by one of us (MGM), has kept records of all patients investigated since 2003. The present analysis is limited to those examined during the period January 2003–July 2006 and evaluated by cytology following needle biopsy for lesions seen on chest radiograph. All subjects in the study were referred to the clinic by their doctor, having already had at least one course of antibiotics and failing to respond. The clinic was not established to diagnose arsenic related disease but rather to assist in the diagnosis of radiographically determined lesions of uncertain aetiology. Patients referred to this clinic routinely provided much personal information before the result of the biopsy was known, including systematic recording of data on residence, the use of tube well water as a source of drinking water and the use of tobacco.


The data were analysed as an unmatched case–referent study, limited to men and women in the clinic series who reported that they had lived for the previous 10 years in a village rather than a township or city, were aged 18 years or more and had habitually drunk water from a tube well. Cases were subjects whose cytology indicated to MGM a primary lung malignancy. Referents were those with non-malignant lesions. Patients with lung metastases from other primary sites, with malignancy of the pleura, a lung sarcoma or a haematological malignancy with lung manifestations, were excluded from the analysis.


In 1998 the government of the People’s Republic of Bangladesh commissioned the British Geological Survey (BGS) to assess the extent and severity of arsenic contamination of drinking water in rural Bangladesh. The report on this extensive work indicated that 27% of tube wells exceeded the Bangladesh standard of 50 μg/l, with highest values in the south and south-east of the country.5 Detailed information tabulated for all 3534 wells measured by the BGS provides the mean concentration for each of 64 of the 67 administrative districts in Bangladesh: three districts in the Chittagong hill tracts were not included in the well survey. For our study, the subject was assigned the mean concentration, as published by the BGS, of the administrative district of his or her usual address.

Other factors

Sex and age at biopsy were recorded together with tobacco use (ever/never), number of cigarettes or bidi smoked/day and the tobacco products used. Among cases, cell type was recorded by MGM and classified for this study as adeno (including “large cell favouring adeno” and adeno-squamous), squamous, small cell, non-small cell, large cell or other.

Statistical methods

Following descriptive cross-tabulation, logistic regression was used to assess the influence of age, smoking and arsenic concentration on case status. The main effect for arsenic, having allowed for smoking and other confounders, was estimated in a multivariate logistic model for smokers and non-smokers and the interaction between arsenic and smoking assessed. The relationship between cell type, smoking and arsenic concentration was also examined, using logistic regression among those with lung cancer. The analysis of smoking and lung cancer was largely restricted to men as the quality of smoking data for women was uncertain (see Discussion).


The project was assessed and approved by the Health Research Ethics Board of the University of Alberta.


A total of 7286 subjects with lung biopsy were recorded between January 2003 and July 2006. All 64 districts covered by the BGS survey were represented in this patient series. Subjects aged less than 18 years (n = 149) were excluded (fig 1). Of those who were 18 years or over, 4025 were found to have a primary lung tumour, 2384 a non-malignant lesion (1508 inflammatory, 597 tubercular and 279 “other” (including sarcoidosis)) and 728 other pulmonary malignancies. More than 75% (5504/7137) had lived in a village for the past 10 years: the proportion who reported using a tube well for drinking water was 84.0% and exceeded 97% in those living in villages (5372/5504). The analysis reported here was thus restricted to the 3223 (2811 male) subjects with a primary lung cancer and 1588 (1183 male) subjects with benign lesions. The 561 (426 male) individuals with other lung malignancies were not studied further.

Figure 1 Selection of cases and referents.

The characteristics of the male cases and referents are shown in table 1. Cases were more likely than referents to be 50 or more years of age and to have been a smoker. Among the smokers, cases were more likely than referents to have smoked more than 20 cigarettes or bidi per day. Although many of the men reported smoking both cigarettes and bidi, male referents who smoked were more likely than cases to report smoking cigarettes, whereas cases were more likely than referents to have reported smoking bidi.

Table 1 Distribution of male cases and referents by age and smoking habit (men living in villages and drinking tube well water)

Among men living in villages and drinking tube well water, only 13.8% were in districts in which the estimated exposure met the WHO guideline of ⩽10 μg/l, but a further 47.8% met the Bangladesh guideline of ⩽50 μg/l; the remaining 38.4% of male patients lived in districts with concentrations above that level, the highest being 366 μg/l. Overall, for men, the proportion of cases increased with mean arsenic concentration in drinking water in their district, but the associated 95% CI excluded 1.0 only at >100 μg/l (OR 1.45, 95% CI 1.16 to 1.80) compared with 0–10 μg/l. In an analysis stratified by smoking habit and adjusted for age and tobacco product and quantity (table 2), the increase in odds ratios with arsenic concentration was seen only in men who smoked.

Table 2 Risk associated with arsenic concentration in male smokers and non-smokers living in villages and drinking tube well water

For the 817 women there was no clear trend between lung cancer and arsenic concentration in the drinking water of the district in which they lived. However, among the 93 women who smoked, 22/27 women living in districts with concentrations above 100 μg/l had lung cancer, giving an OR of 2.64 (95% CI 0.65 to 10.73) when contrasted with the 10/16 cases in female smokers in districts with concentrations of 0–10 μg/l. Only three female smokers lived in districts with concentrations of 51–100 μg/l: when these were combined with the 47 in districts with concentrations of 10–50 μg/l, the proportion of cases (33/50) gave an OR (contrasted with 0–10 μg/l) of 1.17 (95% CI 0.36 to 3.75). Adjustment, as in table 2, for age and smoking units, modified the estimates of risk to 1.06 (11–100 μg/l) and 3.22 (>100 μg/l). The pattern in smoking women, although based on small numbers, was thus essentially that shown in table 2 for men. In women, as in men, there was no such trend in the non-smokers (OR 0.69 at 11–100 μg/l and 0.98 at >100 μg/l).

Finally, the relationship between cell type, smoking and arsenic was explored in the 2755 male cases of lung cancer. Squamous cell lung cancer was much more frequent in smokers (50.5%, 1131/2239) than in non-smokers (39.0%, 201/516). Among non-smokers, no clear pattern with arsenic concentration was seen. In smokers, however, a higher proportion with squamous cell lung cancer was seen in districts with arsenic concentrations >100 μg/l. In an analysis adjusting for age and smoking, the OR for this cell type was 1.34 (95% CI 1.04 to 1.72) at arsenic concentrations >100 μg/l. No relationship was seen between small cell histology and either smoking or arsenic concentration (data not shown).


As cause of death is not reported or recorded in Bangladesh, no truly representative series of deaths from lung cancer can be obtained. Access to medical care of any kind is extremely limited and data from large urban hospitals would be highly biased socially and geographically. The diagnostic quality of records from the Anowara clinic thus affords a unique opportunity to investigate the relationship between arsenic in drinking water and lung cancer in an environment where high exposures are widespread but diagnosis and ascertainment of lung cancer problematic. For almost a decade patients from urban and rural areas throughout Bangladesh with a variety of illnesses have been referred to the clinic for an expert diagnostic opinion based on needle biopsy. The clinic’s main interest is serious respiratory disease, particularly lung cancer, thus allowing a relatively unbiased comparison of patients with and without histologically confirmed lung cancer with respect to residence, smoking history, source of drinking water and other characteristics, all recorded before the diagnosis was made. It is recognised, however, that these cases and referents do not represent random samples from base populations and that they cannot be used to estimate either the incidence or prevalence of lung cancer in Bangladesh.

Main messages

  • Arsenic contamination in drinking water from tube wells in rural Bangladesh is associated with lung cancer in males.

  • Having adjusted for smoking, squamous cell lung cancers are more likely in areas with arsenic concentrations above 100 μg/l.

  • In both men and women, the effect of arsenic in drinking water on lung cancer is greatest in those who smoke.

Policy implications

  • In a country with very limited capacity to provide safe water supplies, the health of future generations may depend on changes in smoking habits and avoidance of contaminated wells.

A clearly increasing trend with exposure in the case–referent analysis supports a role for ingested arsenic, especially in smokers. The limitations of the study, however, must be considered. First, the patients were all from one clinic in Dhaka, the capital city. We know little about the referral patterns, but the service offered at the clinic is unique in Bangladesh as shown by the fact that patients were referred all districts (other than the hill tracts, which were also excluded from the BGS survey). All patients with a lung needle biopsy were referred on suspicion of malignancy but the referral pattern for those eventually assessed as benign may or may not have been similar to that for those found to have a malignancy. Whatever these uncertainties, however, there is no reason to suppose that the referral patterns were related to the mean arsenic concentration of the district of residence. Second, although the data collected by the clinic were extensive, it was not possible to validate them; the use of tube well water, for example, was based on the patient’s own report. This was also true of the smoking histories where social expectations, particularly for women, may well have led to substantial under-reporting. Third, the exposure assessment was limited; it was ecological in that it was based on measurements obtained on 50–70 wells per district. There is evidence of considerable variation in arsenic concentrations in wells within a district6 and the use of mean concentration as an estimate will result in misclassification of the exposure of many of those living within the region and a consequent underestimate of effect size. In addition, the data related to the current address rather than any previous one. We know that the subjects reported living in a village for the last 10 years and suspect from our other studies in Bangladesh that, for men at least, this is likely to be the village in which they were raised and drank the water. However, almost two thirds of wells measured by the BGS were put in place in 1990 or later with only 2% operational in 1970.7 If the latency for arsenic related lung cancer is several decades, the true risk for future illness will be much greater than described here. Finally, the likelihood that subjects within the referent series with non-malignant respiratory disease had ill-health caused or made-worse by exposure to arsenic810 or by smoking means that the estimates of risk are extremely conservative and that risks in non-smokers and at lower arsenic concentrations may have been missed.

There are thus several reasons for thinking that the results of our analyses substantially underestimate the level of risk in relation to both arsenic and smoking and limit our capacity to investigate the nature of any interaction between them. The extent to which subjects in the referent series had respiratory disease caused or made worse by smoking or arsenic exposure will have reduced the estimate of risk as will the misclassification of exposure resulting from the use of district mean arsenic concentrations and errors in reports of smoking. As such our analyses cannot preclude the possibility of substantial risk in non-smokers exposed to arsenic and at arsenic concentrations at or below 100 μg/l.

The relationship between squamous cell histology and smoking in men is strong in this study and is found at each level of arsenic concentration and each age group (data not shown). This gives added credence to the observed relationship between this cell type and concentrations of arsenic >100 μg/l, suggesting a specific tissue response. An excess of squamous (and also small cell) lung carcinomas was found among those with arsenic contaminated water in a study of 243 townships in Taiwan, but there was no information on tobacco use.11 The present study, with adjustment for smoking, suggests that ingested arsenic may have an independent effect on the cell type of the tumour: the Taiwanese finding of a higher proportion of small cell carcinomas in high arsenic areas was not seen in this population.

A substantial but unknown proportion of the Bangladeshi population is exposed to arsenic concentrations above 100 μg/l in their drinking water, and despite efforts to mark and discourage the use of highly contaminated wells, use of such water over many years can be anticipated. Our results, together with evidence from elsewhere,2 3 suggest that many Bangladeshis living in villages and drinking water from arsenic contaminated wells will be at increased risk of lung cancer. That this risk appears greatest in smokers, particularly of bidi, further underlines the need for reduction in smoking. In a country with very limited resources to ascertain the disease or provide safe municipal water supplies in some 100 000 villages, the health of future generations may depend on changes in smoking habits and avoidance of contaminated wells.


Sk Nazmul Huda and Ms Labin Rahman provided valuable assistance in accessing and checking the data.



  • Competing interests: None.

  • Funding: None.

  • Ethics approval: The project was assessed and approved by the Health Research Ethics Board of the University of Alberta.