- P Cullinan, senior lecturera,
- S Acquilla, lecturerb,
- V Ramana Dhara, visiting scientist, on behalf of the International Medical Commission on Bhopalc
- a Department of Occupational and Environmental Medicine, Imperial College (National Heart and Lung Institute), London SW3 6LR
- b Department of Epidemiology and Public Health, University of Newcastle upon Tyne, Newcastle upon Tyne NE2 4HH
- c Agency for Toxic Substances and Disease Registry, E-31, 1600 Clifton Road, Atlanta 30333, Georgia, USA
- Accepted 15 November 1996
Objective: To examine the role of exposure to the 1984 Bhopal gas leak in the development of persistent obstructive airways disease.
Design: Cross sectional survey.
Setting: Bhopal, India.
Subjects: Random sample of 454 adults stratified by distance of residence from the Union Carbide plant.
Main outcome measures: Self reported respiratory symptoms; indices of lung function measured by simple spirometry and adjusted for age, sex, and height according to Indian derived regression equations.
Results: Respiratory symptoms were significantly more common and lung function (percentage predicted forced expiratory volume in one second (FEV1), forced vital capacity (FVC), forced expiratory flow between 25% and 75% of vital capacity (FEF25-75), and FEV1/FVC ratio) was reduced among those reporting exposure to the gas leak. The frequency of symptoms fell as exposure decreased (as estimated by distance lived from the plant), and lung function measurements displayed similar trends. These findings were not wholly accounted for by confounding by smoking or literacy, a measure of socioeconomic status. Lung function measurements were consistently lower in those reporting symptoms.
Conclusion: Our results suggest that persistent small airways obstruction among survivors of the 1984 disaster may be attributed to gas exposure.
Many people died as a result of exposure to gas after the 1984 Union Carbide disaster in Bhopal but long term effects remain unclear
In this study respiratory disease attributable to gas exposure was detected in adult survivors
The frequency of symptoms decreased with decreasing exposure (as estimated by distance of home from the plant)
Lung function showed similar trends, although the number of subjects was much smaller
Much of the disease is probably due to irreversible obstruction in small airways
In the early hours of 3 December 1984 an explosion at the Union Carbide India pesticide plant in Bhopal, Madhya Pradesh, India, resulted in about 27 tonnes of toxic gas being dispersed over the city. The exact nature of the gas remains disputed, but most of it was probably methyl isocyanate, a pungent gas of low boiling point (39°C) and high vapour pressure (348 mm Hg at 20°C).1 There is little non-anecdotal information on the subsequent path of the gas plume, though attempts to model its dispersion using the scanty meteorological data available suggest that it travelled slowly southwards from the plant.2 About a quarter of the city's million inhabitants are believed to have been exposed.
The acute effects of heavy exposure were consistent with inhalation of a volatile (and deeply penetrating), highly irritative aerosol; large numbers of people died rapidly with bronchial necrosis or pulmonary oedema.3 Within months of the disaster clinical, radiological, and pathological evidence suggested survivors had persistent airflow obstruction associated with an obliterative bronchiolitis.4 Subsequent case reports have proposed other pulmonary outcomes including interstitial fibrosis.5 However, there have been no published systematic studies of the long term respiratory effects of exposure to the gas leak.
We describe findings from a survey conducted 10 years after the disaster at the request of community groups in Bhopal. An earlier paper documented the wide variety of symptoms reported by those exposed to the gas leak.6 Here we present a detailed analysis of respiratory symptoms and function among a random sample of survivors in which the role of gas exposure has been assessed by examining its relation with disease frequency.
Subjects and methods
To ensure that we selected subjects with different levels of exposure we created four concentric zones of 2, 4, 6, and 8 km radius centred on the Union Carbide factory site using a map of the city. We also examined an area outside the city but of similar socioeconomic composition to the urban zones to act as an unexposed control zone. From each of these areas we surveyed a sample of the current resident population by selecting houses using random numbers from two electoral wards on either side of a straight line south from the factory site. We invited (after random selection from a hat) one of the adults (18-60) present who had been resident in Bhopal in 1984 to complete a questionnaire.
The previously piloted structured questionnaire was administered in Hindi by trained interviewers and inquired into details of exposure to the gas leak, current and past health, and socioeconomic factors. Place of residence at the time of the gas leak was confirmed from official records held by each of those interviewed. Subjects were defined as dyspnoeic if they reported breathlessness either when walking on level ground or on climbing hills. Literacy was defined as the completion of at least primary education. Men with low incomes were those with no paid work or a monthly wage of less than 500 rupees (about £10).
We invited a 20% random sample of those interviewed to undergo spirometry using a rolling seal spirometer (Ohio 822, National Institute for Occupational Safety and Health, United States). After instruction from an experienced doctor subjects carried out at least two acceptable and reproducible manoeuvres7 before and after inhalation of 200 µg of salbutamol. The instrument was calibrated and checked for leaks before and after each day's session. Measurements of forced expiratory volume in one second (FEV1), forced vital capacity (FVC), FEV1/FVC, and forced expiratory flow between 25% and 75% of vital capacity (FEF25-75) were corrected for body temperature, pressure, and saturation and expressed as proportions of predicted values by using regression equations for Central Indian adults provided by Udwadia et al. 8
We analysed symptoms and lung function measurements according to whether the subject reported exposure to the gas leak. More detailed exposure-response relations were examined by estimating exposure expressed as the distance from the factory of the central point of the electoral ward in which the subject was living at the time of the disaster; this index was grouped before analysis according to roughly equal sized categories of increasing exposure.
We assessed differences in proportions between categorical variables by Χ2 statistics and Χ2 tests for linear trend using egret and epi-info 6 statistical software. Associations between lung function and estimated exposure (expressed continuously by distance of ward from the factory) were examined by linear regression.
A total of 454 subjects were interviewed (no selected subjects declined an interview) and 74 (82% of those eligible) underwent spirometry. Table 1 shows the distribution of subjects by estimated exposure and potentially confounding variables. Only five of the 14 subjects in the control area attended for lung function testing. There were no important or consistent differences in the age, sex distribution, socioeconomic status, or reporting of cough, sputum, or breathlessness between those who did and did not attend for testing. One person was unable to perform satisfactory lung function measurements.
There were close relations between sex and low income (most women were not in paid work) and smoking (only two women reported ever smoking). All but two of those who reported ever having smoked were current smokers. The proportions of illiterate subjects and low income men in the high exposure categories tended to be greater than in the non-exposed groups, but the trends were not consistent. There was no such gradient with smoking. Although numbers were small, similar trends were discernible among those undergoing spirometry. Smoking was not significantly associated with literacy (Χ2 = 0.5, P = 0.480).
Of those who reported exposure to the gas leak, 190 (54%) described a cough and 54 (15%) phlegm production for at least three months each year. Two hundred and ninety four (84%) reported dyspnoea and 30 (9%) wheeze in the past 12 months. These frequencies were significantly higher (P<0.001) than those in people who were unexposed (16 (16%), 3 (3%), 55 (54%), and 1 (1%) respectively), and the differences remained when literate subjects or those who had never smoked were examined separately.
A more detailed analysis revealed consistent and significant trends in each symptom across six categories of estimated exposure (fig 1). These trends were equally strong (P<0.001 for each group) in men and women, and more so in those aged less than 35 years, the median age of the group. All symptoms were reported more often by cigarette smokers but were still common in non-smokers; after those who had ever been smokers were excluded the associations between symptoms and reported or estimated exposure remained (table 2). Similar results were obtained when literate subjects or men with low incomes (data not shown) were examined alone, though the numbers in each group were small and the gradients less consistent.
Before we analysed lung function and exposure among those tested we collapsed the four estimated exposure categories based on distance from the plant. All indices of pulmonary function were lower in those reporting exposure to the gas and among those in the higher categories of estimated exposure (table 3, fig 2). When distance was analysed as a continuous variable FEF25-75 was significantly reduced (P = 0.004) but no association was found for the other indices of lung function (P>0.05). We also divided measurements of function into equal quartiles; seven (35%) of those in the 0-2 km exposure category had an FEF25-75 result in the lowest quartile (<67% predicted), compared with nine (29%), three (18%), and zero in the 2-6 km, 6-10 km, and >10 km categories respectively. The equivalent proportions were six (29%), eight (26%), four (24%), and zero for FEV1/FVC ratio; seven (33%), seven (23%), four (23%), and one (20%) for FEV1; and six (29%), four (19%), five (29%), and one (20%) for FVC. Stratification according to smoking habit or socioeconomic factors (literacy or, in men, low income) weakened but did not abolish the relations with reported exposure. There were too few subjects to adjust simultaneously for these factors and for estimated exposure.
Lung function values among those reporting respiratory symptoms were consistently reduced (table 4) as they were among cigarette smokers. Only two subjects (both exposed to gas) had an increase in FEV1 of 200 ml (and at least 15%) after administration of salbutamol.
We have shown an excess of respiratory symptoms and a reduction in mean lung function among those reporting exposure to the Bhopal gas leak in 1984. We have also shown a direct gradient in effects according to an estimate of the intensity of gas exposure. We studied people still resident in Bhopal 10 years after the gas leak and thus their symptoms cannot represent the acute or long term lethal effects of gas exposure in the intervening decade. The symptoms reported, and differences in lung function detected, are compatible with chronic airflow limitation and particularly with disease of the small airways.
Several features suggest that the relation between the symptoms and exposure to the gas leak was causal rather than a chance association. Firstly, our findings are plausible: the physical characteristics of methyl isocyanate suggest that airborne droplets would have vaporised after inhalation and penetrated deep into the bronchial tree. Secondly, there was a high internal consistency in responses to our questionnaire and between these and the results of spirometric testing. Thirdly, we have shown a dose-response effect in both symptoms and lung function. In the absence of a validated respiratory symptom questionnaire for these circumstances and of an objective measurement of abnormal lung function we relied on internal comparisons to assess the relation between gas exposure and outcome.
We deliberately used a crude estimate of gas exposure, relying on addresses at the time of the leak, in the belief that these would be readily recalled and less liable to misclassification than more complex indices. Though place of residence at the time of the leak was established objectively, self reported symptoms are open to recall bias. However, the associations between the frequency of reported symptoms and estimated exposure were striking, and the consistency of the symptom gradients across narrow (2 km) bands of estimated exposure supports our belief that these findings are not wholly explained by differential reporting. Similar gradients were found for the results of lung function testing; there were no obvious differences between those invited for testing who did and did not attend, but the variability in response between exposure groups and the small numbers involved necessitate cautious interpretation. To account for the different age, sex, and height distributions across exposure categories we adjusted the measurements using regression equations derived from an Indian population. Despite this the derived figures should be treated as relative rather than absolute values.
The issue of confounding by other factors is unavoidable. We addressed this by stratifying analyses according to various potential confounding variables. Although the exposure relations tended to be weaker after stratification, this exercise suggested that a proportion of the observed relations can be attributed directly to gas exposure. In an earlier study three years after the gas leak, Andersson et al reported a relation between estimated exposure and respiratory symptoms which could not be accounted for by age or cigarette smoking.9 We remain aware, however, of the difficulties of completely expressing the effect of confounding variables–particularly those which reflect socioeconomic factors such as low birth weight.10
Previous studies of respiratory morbidity in Bhopal survivors have not examined disease beyond two or three years after the disaster. Rastogi et al examined lung function in 783 exposed volunteers and detected severe respiratory impairment (reductions in FEV1 or FVC, or both) in 2.4%, mainly of a mixed obstructive-restrictive nature.11 They did not attempt to correlate their findings with a measure of exposure. Kamat and colleagues followed 113 subjects who had been initially admitted to hospital.12 Over two years they found no improvement in respiratory symptoms or in basic spirometry, but 42% of the group had evidence consistent with deteriorating small airways function. It is difficult to estimate how much of this could be attributed to gas exposure. Bronchoalveolar lavage in 36 subjects one to two and a half years after the leak showed increased cellularity (macrophages or neutrophils, or both) in those who were severely exposed (as assessed by immediate morbidity), a finding which could not be wholly attributed to smoking13 and which the authors suggested was consistent with a subclinical alveolitis.
Our long term findings are consistent with small airways obstruction and with the obliterative bronchiolitis reported among early survivors.4 Similar effects have been reported following accidental inhalation of other irritant gases.14 15 Long term obstructive changes in spirometry have also been described in seven Finnish miners exposed to high concentrations of (mainly) sulphur dioxide, in whom measurements of lung function before exposure were available; mean FEV1 was 13% below predicted after four years with a smaller proportional reduction in FVC.16 Persistent obstruction after exposure to an irritant is probably the result of repair processes after initial epithelial and basement membrane damage, the level of obstruction being determined by the site of maximal bronchial deposition, which is determined largely by the physicochemical characteristics of the irritant. Irritant induced asthma (or reactive airways dysfunction syndrome) has also been described after high intensity exposures17 and was reported in four of the Finnish miners.16 Although we did not conduct formal tests of histamine or metacholine reactivity, the lack of response to salbutamol in all but two of our subjects suggests that this is not a common outcome in survivors of the Bhopal gas leak.
The dearth of rigorous examinations of the relation between exposure to the gas and subsequent morbidity is one of the lesser tragedies of the Bhopal disaster. One result is that the establishment of appropriate models of health care and the equitable distribution of available compensation have been hampered. We did not attempt formal measurement of exercise capacity, but several subjects reported being too breathless to undertake paid employment. Though there was little evidence in this group that the airflow obstruction was reversible by simple bronchodilators, many were taking prescribed (and often expensive) medicines inappropriately–for example, haphazard courses of systemic steroids and a variety of (mainly oral) bronchodilators.
Our findings suggest the need for further, locally initiated studies of the extent of persistent airflow obstruction attributable to gas exposure and for controlled trials of effective treatments. This small study, carefully planned but completed in nine days, is a model for low cost and low technology research into the long term effects of the gas leak. Similar studies might usefully concentrate on respiratory disease in those who were infants at the time of the disaster as well as effects on other (for example, ocular and neurological) systems.
Other members of the International Medical Commission on Bhopal were R Bertell (Canada), T Calender (USA), I Eckerman (Sweden), J Havens (USA), B Heinzow (Germany), J Jaskowski (Poland), C Sathymala (India), L Titov (Byelorussia), G Tognoni (Italy), M Verweij (Netherlands), W Zhengang (China).
We thank Satinath Sarangi, Bhopal Gas Peedith Mahila Udyag Sanghatana, Gas Peedith Stationary Karamchari Sangh, and Zahreeli Gas Kand Sangharsh Morcha, Dr Isobel Gillis, Dr Eugene Milne, Mr Adrian Cook, and Professor A J Newman Taylor for their help. The spirometer was kindly loaned by the division of respiratory diseases, National Institute for Occupational Health, Morgantown, USA.
Funding: The commission was funded by Christian Aid (UK), Greenpeace and Bread for the World (Germany), United Council of Churches in Christ (USA), and other environmental and religious charities.
Conflict of interest: None.