April 25, 2014

expert reaction to The Environment and Health Atlas for England and Wales

Researchers at the UK Small Area Health Statistics Unit (SAHSU) launched an atlas that maps health outcomes and environmental agents at neighbourhood level across England and Wales. These comments accompanied a press briefing.

Prof David Coggon, Professor of Occupational and Environmental Medicine, University of Southampton, said:

“The main aim of this atlas is to map geographical differences in cancer incidence and mortality within England and Wales that might provide clues to preventable environmental causes of disease.  Because the focus is on variation in disease rates, results are presented principally as relative risks (i.e. the ratio of the frequency of cases in a local area to that in the country as a whole).  This is entirely appropriate, but when interpreting the findings it is important to bear in mind that the absolute risk of an outcome will depend also on its underlying frequency.  Thus, for example, a 20% increase in the occurrence of a common disease such as lung cancer represents a bigger absolute elevation of risk than a doubling of a much rarer disorder such as mesothelioma.

“The atlas differs from predecessors in its finer level of spatial resolution, with the use of statistical smoothing to compensate for chance fluctuations when only a small number of cases is expected in an area.  In general, this works well, although because of the smoothing, the maps for mesothelioma do not pin-point so clearly the locations of the asbestos-using industries that have been responsible for the marked variation in its occurrence. 

“The sources of data are the most reliable that are available, but they do have limitations, many of which the authors identify.  For example, environmental levels of potentially hazardous chemical and physical agents may be only a poor proxy for the personal exposures of individuals.  Thus, in the case of pesticides, individual exposure is influenced also by use of pesticidal products in the home and garden, and by residues in food.  Exposures from these other sources are extremely low, but they often outweigh any exposures of rural residents from spray drift.

“Limitations of the data on mortality and cancer incidence include:

• their assessment by place of residence at the time of death or diagnosis of cancer, with no allowance for a possible latent interval between relevant exposures and the manifestation of disease;

• the possibility of chance variation in the occurrence of cases (which can only be partially addressed by the statistical analysis); and

• the likelihood of residual confounding by differences in exposure to non-environmental causes of disease such as smoking and diet, which are not fully characterised by a simple measure of deprivation.

“These unavoidable shortcomings do not invalidate the analyses presented, but they are a reason for caution in interpretation.  Even where geographical differences are not due simply to chance, they will not necessarily reflect a hazard in the environment.  Thus, it would be naïve to assume that a person’s health would be improved by moving from a place with higher disease rates to one where the disease was less common.  People who wish to optimise their health will do better to focus on their personal lifestyle, eating a healthy diet, taking regular exercise, and avoiding smoking, excessive alcohol and unnecessarily risky behaviours such as dangerous driving.”

Prof Jon Ayres, Professor of Environmental and Respiratory Medicine, University of Birmingham, said:

“This atlas represents a considerable amount of hard work pulling together a wide range of datasets of varying quality to assess any possible associations between a health outcome and an environmental exposure.  However, association does not infer that the relationship is causal and, in my opinion, that has not been sufficiently stressed in the introduction to this work.  This is crucially important – the atlas, which is to be welcomed, can thus be seen as a hypothesis generating tool not as a means of proof for certain environment outcome pairings.  Having said that, there are some associations where the message is clear and correct (e.g. air quality and respiratory and cardiovascular outcomes; mesothelioma and asbestos).

“Some of the exposure datasets are surrogates for true personal exposure and these are areas where most caution needs to be shown.  Further, it is very important to realise that, where a causal association exists, data such as this is only true at a population level and cannot be used to assess individual risk, nor should this be used as a means for deciding where one should decide to live.”

Prof Paul Pharoah, Professor of Cancer Epidemiology, University of Cambridge, said:

“The Atlas comes across as a data “tour de force” and is a tremendous achievement in describing in detail the geographical distribution of a wide range of environmental exposures and health outcomes.

“One of the three explicit primary aims of the Atlas is: “To help in development of hypotheses to understand and explain variability in disease risk that may relate to the environment, life- style factors and/or location”.  This statement emphasises the fact that it would be wrong to imply any causal association between any of the environmental exposures and any of the health outcomes described in the Atlas.  Indeed the authors do not attempt to drawn any such conclusion and where a link between a specific exposure and health outcome is mentioned, this is on the basis of results from other research.  What these data should do is help researchers identify important hypotheses that should be tested using other research designs.

“It would be inappropriate to draw conclusions between any association of environmental exposures and disease rates at the population level and assume that the association was in any way causal – we cannot say that any of the environmental exposures considered caused the development of any of the diseases measured.  There are two important reasons why such a conclusion would be inappropriate:

i) One does not have information about any specific environmental exposure in an individual and the occurrence of a specific health outcome in the same individual (because exposure and outcome are measured at the population level).

ii) A correlation between a specific environmental exposure and specific health outcome may simply be due to the fact that the environmental exposure is correlated with another important (causal) exposure.  When a causal association is inappropriately inferred from such data it is known as the ‘ecological fallacy’.

“In terms of the authors controlling for confounding factors, the disease rates in each area have been adjusted for deprivation which is a proxy for socio-economic status, but it is an imperfect proxy.  Furthermore, there are many other potentially important confounders that may be important at the individual level, including diet and smoking and other individual behavioural factors.

“The health risks are measured as relative risks.  That is the risks in one area are compared to another and the population average.  This is entirely appropriate as a way of understanding variation across the country.  However relative risks do not provide any information about the actual individual (absolute) risk in any one area, that is the actual disease rate in an area.  The relative risks can be adjusted for the age-structure and area-based socio-economic deprivation.  Furthermore, by using the same relative risk scale for each health outcome, it is possible to compare directly the patterns of variation by area for the different health outcomes.

“When mapping the data, the authors used a process called smoothing.  The smoothing was done in an attempt to reduce the effects of random variation for rare diseases in small areas.  To explain this simply, assume a disease was so rare that only one case would be expected to occur in a given region/population in one year.   In one such region you might get two cases and one case may occur in another region just on the basis of chance.  It would be inappropriate to conclude that the disease rate was really twice as much in the first region as in the second.

“Overall, it is important for policy makers to understand the role of environmental risk factors for disease which may affect health at the population level, and for this they will find this atlas useful.  However, for individuals, there are multiple risk factors that are likely to have greater effects on the risk of major diseases than environmental exposures, and, individuals can change their own exposure to these risk factors.  In other words, people should adopt a healthy diet and lifestyle as far as possible.  This atlas does not enable anyone to judge their individual absolute risk.  People should definitely not use this atlas to decide where to live.”

‘The Environment and Health Atlas for England and Wales’ by A.L. Hansell et al. published by Oxford University Press on Friday 25 April 2014.