Commentary: Assessing the effects of environmental pollution when people know that they have been exposed

BMJ 1997; 314 doi: (Published 01 February 1997) Cite this as: BMJ 1997;314:343
  1. R T Mayon-White, consultant public health physiciana
  1. a Directorate of Public Health and Health Policy, Oxfordshire Health Authority, Oxford OX3 7LG


    Acute episodes of massive air pollution with harmful effects are fortunately rare, but this means that no one is very experienced at investigating such risks. The immediate effects may be blast and burns injuries, acute chemical toxicity as at Bhopal in 1984, irritation causing acute respiratory and conjunctival symptoms, or the sight and smell of smoke and fumes causing fear. The suddenness and speed at which the pollution arises, usually in an explosion or a fire, give little chance of collecting data before exposure. Epidemiological studies into the longer term effects have to be done after the pollution has dispersed and when public opinion has formed about the expected health effects. Investigations may be made more difficult if there is little toxicological knowledge of the long term effects of a single exposure or if the pollutants are not fully identified.

    So the investigation into the long term effects of a relatively brief exposure to airborne chemicals is an exploration, not a well trodden path. However, if the dramatic circumstances of incidents like Bhopal are set aside, the epidemiological task is the familiar problem of deciding whether there is an association between an environmental factor and ill health, and if any relation is cause and effect.

    The study by Cullinan et al is a good example of what can be achieved with modest resources. Cross sectional studies are usually much cheaper than cohort studies in which a population (or sample) is followed for years to assess the effects of differing levels of exposure. Without detailed information on where the toxic gas plume went and where subjects were at the time, the distance between the Union Carbide factory and the home of the subject is a simple way of grading exposure, and better than the subjects' own impressions. A random sample was essential, but a bias could occur if people with respiratory symptoms were more or less likely than others to move away from the area. In some situations the subjects might be asked if they stayed indoors, as another way of getting different degrees of exposure.

    In any study where both the subjects and the investigators are aware of their exposure, and the possible illness, the investigators need objective measurements of health: the lung function tests in the Bhopal study. As shown here, symptoms should not be dismissed as subjective and biased as sceptics might suppose. If symptoms can be compared with physiological or pathological tests in a subset of subjects the strength of large numbers studied by questionnaire can be exploited. In assessing the statistical association for cause and effect the strength of the association, the specificity of the medical condition, the biological gradient, the biological plausibility, and the coherence of the evidence are clear. The results fit with earlier studies in Bhopal but there are, fortunately, no other similar incidents which would test the consistency of the association. The relation with time cannot be assessed by one cross sectional study, and no experimental evidence or analogies apply. I am convinced by the evidence that the gas from Bhopal has caused chronic obstructive airways disease and I value this paper as a useful example of how to look for the damage that may be caused by a chemical disaster.

    View Abstract