Skip to main content

Archived Comments for: Disease proportions attributable to environment

Back to article

  1. Advancing methods for environmental burden of disease calculations

    Colin L Soskolne, University of Alberta

    30 November 2007

    In response to the thoughtful commentary by Saracci and Vineis, "Disease proportions attributable to environment", I am inspired to hearken back to first principles, some more and some less relevant to their commentary:

    1. Any single summary measure can reflect only what it is: a single summary measure. As with a mean expressed alone, questions of the distribution of the parts are lost. Age, gender, social class, and other factors are submerged in any such single measure.

    2. The scientist reporting overall summary measures, in bearing the above two points in mind, ought to consider also reporting additional measures to give some insight into what underlies and thereby explains the overall summary measure.

    3. The strengths and limitations of any measure need to be recognized, and the reader ought to be reminded of these every time they are used.

    4. Definitional questions are fundamental to being able to compare one experience at one time with an experience at another time. "Compare like with like", as Saracci and Vineis note.

    5. In science, enough methodological detail needs to be provided to allow the interested reader to assess the validity of the work, and to replicate it if desired.

    6. In providing summary measures, the assumptions being made and the definitions being used need to be explicit. The reader disagreeing with the assumptions should then be able to recalculate any estimates based on his/her own variation on the assumptions, and which might or might not include certain classes of exposure and/or outcome.

    7. The concern about blending experiences globally is real. We might therefore ask that if this kind of statistic is so important for public policy, why are we not pressing for funding to conduct global assessments in some randomly stratified way in which resources are put into being more thorough in a few countries that will render more precision in the estimates? From such a base, generalizations to the global burden could perhaps be more reliably estimated.

    8. The contribution of such a commentary serves to identify the great need for methodological advances, and also of the need for funding to enable this type of work.

    9. Care over the focus of such work needs to be considered. The burden of disease attributable to this or that putative cause depends on past experience. When we consider, as alluded to in the commentary, global change phenomena, prediction is also critical; the burden of disease through, for instance, endocrine disruption and other influences, needs to be considered in advancing this field methodologically. What we see today based on historic exposures and trends, could be the tip of tomorrow's expanding iceberg.

    Thanks to Drs Saracci and Vineis for their stimulating commentary.

    Colin L. Soskolne

    Competing interests

    None declared

  2. Widening a welcome debate

    Anthony (Tony) J. McMichael, Australian National University, Canberra, Australia

    2 January 2008

    The paper by Saracci and Vineis (‘Disease proportions attributable to environment’) underscores important conceptual and methodological issues, and confusions, that arise in the vexed matter of estimating population disease fractions attributable to ‘the environment’. It is as easy to under-estimate as it is to over-estimate.

    That estimation has always posed a challenge to epidemiologists. Most environmental exposures impinge upon whole areas or communities; they are often difficult to measure at the individual level (the ‘dose’ received’); and, anyway, there may be little difference between individual exposure levels within an exposed community. For related reasons, aetiological inferences about shared environmental exposures are often more difficult to draw than for epidemiological studies in which individual exposures (cigarette smoking, alcohol consumption, oral contraceptive use, etc.) can be measured and related to individual health outcomes.

    Further, the population attributable fraction for an ambient environmental exposure may be under-estimated because of the lack of a true non-exposure counter-factual. All members of modern urban populations are exposed to some level of air pollution; there is no equivalent of the non-smoker or teetotaler ‘referent’ category. Likewise for sunlight and skin cancer.

    There is, too, a larger issue here. This field of enquiry is necessarily moving beyond a restricted consideration of direct-acting toxicological exposures in air, water, soil and food. Such exposures, due to classical ‘pollutants’, remain important – particularly in societies undergoing rapid industrialization and urbanization. However, by dint of sheer scale and type of contemporary economic activity, human societies are now doing more than causing local environmental pollution; they are disrupting and changing whole natural systems, both biophysical and ecological. Climate change, the best-known example, is part of a larger syndrome of human-induced environmental change that includes stratospheric ozone depletion, nitrification of ecosystems, loss of soil fertility, depletion of freshwater supplies, exhaustion of many great fisheries and accelerating loss of biodiversity. The actual and potential population health consequences of these environmental changes, around the world, are huge.

    For that reason, at least, it would be inappropriate to relabel the environment as either ‘non-genetic’ or as ‘pollutants’ (as suggested by Boffetta et al). The word ‘non-genetic’ both trivializes and confuses. The word ‘pollutants’ invokes a now inadequate twentieth-century view of the full scope and content of the environmental risks to health that bear on current and future generations of humans.

    AJ McMichael

    National Centre for Epidemiology and Population Health

    The Australian National University



    Competing interests


  3. Authors' response to Anthony McMichael

    rodolfo saracci, IFC-National Research Council,Pisa,Italy

    3 January 2008

    We had only tangentially alluded in the final section of our paper to the existence of special problems when trying to infer disease causation- the first step towards prevention- from widespread environmental exposures, and "a fortiori" from exposures of a global nature.

    Implied in McMichael's comments are new challenges hard to address with the standard methodological armamentarium of epidemiologists.There will never be- unless theoretical physics opens magic doors to "parallel universes" - evidence on the population effects of global climatic changes from randomized trials nor even from observational comparisons of concurrent worlds ; and no help can come from playing with nomenclatures of the factors involved. Materially and metaphorically global environmental changes "fall from the sky" on people and on people's health , nurturing ancestral collective fears in contemporary guise.These cannot be dispelled unless the scientific response is firmly grounded on innovative approaches to causal inference, risk modelling, risk evaluation and uncertainty weighing in environmental health, as pioneered by McMichael's work.

    Rodolfo Saracci

    Paolo Vineis

    Competing interests

    No conflict of interest to be declared

  4. Pollutants, man-made products and human activities are probably major determinants of human cancer

    Nicolas van Larebeke, Ghent University. Study centre for Carcinogenesis and primary preventioin of cancer

    11 January 2008

    Comment on the paper by Saracci & Vineis.

    We agree with all considerations of Saracci &Vineis. However, while emphasising the importance of the notion “environment” and of preventive measures at the collective level, their paper does not elaborate on another important issue: the major impact that human activities, man-made substances and diverse pollutants have on cancer incidence. And indeed not only on cancer incidence, but probably also on the incidence and prevalence of many other “diseases of civilisation” such as atheromatosis, diabetes, Parkinson’s Disease and possibly also on the metabolic syndrome and on diseases related to immunological reactions such as astma, allergies and some auto-immune diseases. The publications, the most important of which is probably the International Agency for Research on Cancer (IARC) report “Attributable causes of cancer in France in the year 2000”, stating that pollutants cause only a few percent or even only 0.1 % of all cancers in an industrialized country such as France, take into account only a tiny fraction of the insights and knowledge that is available on the subject. The IARC report for instance is only based on the impact of agents considered by IARC to be proven human carcinogens. However, it is certain that we can only in exceptional circumstances prove that something is carcinogenic to human beings, as experiments on humans are ethically unacceptable, and as all of us are exposed to hundreds if not thousands of agents that can contribute to carcinogenesis. Lack of sensitivity and negative confounding limit the discerning power of our epidemiologic studies.

    Cancer, probably the most fundamental disease of multicellular organisms, is a multicausal disease and rests on the one hand on accumulation of mutations and on the other on disturbance of inter- and intracellular signalling. Probably any increase in entropy, both in terms of DNA sequence and in terms of gene expression, increases risk of cancer. Many agents intervene as necessary but not sufficient etiologic factors in most, probably all, cases of cancer.

    There are no reasons to think that the many carcinogenic agents surrounding us, mostly in very low concentrations or intensities, have no effect.

    First, lets consider genotoxic agents.

    Most mutations are endogenous, but a slight increase in mutation rate has far reaching consequences. If mutation frequency would have been 10 times the observed one, life on earth would not have evolved further than organisms of the complexity of drosophyla. The large increase in cancer risk observed during the 20th century rests probably on only a limited increase in mutation frequency. There is a substantial load of endogenous damage and exogenous genotoxic agents causing the same damage as endogenous ones probably increase mutation rate proportionally to the increase in the corresponding damage. Endogenous DNA damage is swiftly and efficiently repaired by DNA repair mechanisms that are permanently active, which is essential for survival of the cell.

    Part of exogenous damage is far more mutagenic than endogenous damage. DNA adducts of aflatoxin or benzo(a)pyrene are associated with a substantial increase in cancer risk at adduct levels that are a hundred times below endogenous adduct levels. In addition, very low doses of some exogenous agents are more effective in inducing mutations and tumoral transformation than can be extrapolated from high dose experimental or occupational studies. That is due to the fact that such very low doses reach the cells off guard, that is without their relevant DNA repair mechanisms being induced maximally. Indeed, that repair mechanism directed towards exogenous damage are not permanently fully active is demonstrated by the existence of “adaptation”, and makes sense in terms of the cell using sparingly its resources. Such lightly damaged cells will in many instances show an increase in mutation rate rather than die as the mechanisms for apoptosis will not be induced and as general repair mechanisms will ensure the survival of the cell.

    Secondly, lets consider receptor binding substances that affect gene expression. Indeed, many man-made chemicals and many pollutants have biological effects through receptor binding, including hormone-like effects. Recently it has been shown that such substances can act even at very low doses. They can have effects that are identical to the endogenous physiological substances binding to the same receptor, but they can also have different effects. Endocrine disruption may be of particular importance during certain time windows in prenatal evolution. This insight has led to a new paradigm, the “developmental origins of human health and disease” paradigm.

    Where our observations are sensitive and precise enough supralinear dose-effect relationships, with relative more important effects per dose unit at low doses, were found in epidemiologic observations on humans, for instance concerning fine particle air pollution and also smoking.

    So it makes no sense , neither in scientific terms nor in terms of public health policy, to limit our assessment of causes of cancer to the relatively few proven human carcinogens .Only if we would be able to take into account the very many carcinogenic exposures that occur in real life would we be able to assess the real impact of pollutants on our cancer incidence. This would probably imply a huge combined biomonitoring and epidemiologic exercise that is beyond our present reach in view of the limited available budgets.

    Competing interests

    None declared

  5. Environmental disease burden – the confused case of UVR exposure

    Robyn Lucas, National Centre for Epidemiology and Population Health

    12 January 2008

    The commentary by Saracci and Vineis and earlier papers by Boffetta et al and Pruss-Ustun raise important issues in the estimation of the disease burden attributable to environmental factors. Saracci and Vineis note, correctly, the importance of a consistent definition of environment in comparing estimates of environmental burden of cancer. Consistency alone, however, is not enough – and it is not clear what is the value of Boffetta et al’s narrow definition of environment. From a public health perspective preventability is a key issue – in this regard it becomes important to include not just those exposures to “(natural or man-made) agents encountered by humans in their daily life, upon which they have no or limited personal control” but also those environmental factors over which individuals have considerable personal control and which are therefore amenable to change in exposure (and disease risk) at both individual and population levels.

    In our recent assessment of the global and regional disease burden due to ultraviolet radiation,1 we encountered further difficulties with the use of the population attributable proportion (PAP). In the calculation of PAP to assess the role of ultraviolet radiation (UVR) exposure in skin cancers, we found a wide disparity in PAP calculated from ecological (population-based) and case-control studies. The former give higher values than do the latter. We think this disparity results from several factors:

    1. the quality of the exposure assessment. In case-control studies past sun exposure is typically measured by recalled sunburns or time in the sun many years previously (with the likely long latency between exposure and skin cancer onset). Ecological studies typically compare skin cancer rates in countries with different ambient UVR (assuming similar exposure patterns) or compare fair-skinned populations with deeply pigmented populations under the same ambient UVR (assuming similar exposure patterns but different UV dose to underlying tissue consequent on the protective effect of skin pigmentation).

    2. differing magnitude of exposure range between “exposed” and “non-exposed”. In case-control studies, the “non-exposed” are less exposed (not unexposed) compared to those with high relative exposure. Ecological studies typically compare populations experiencing the extremes of UVR exposure, eg skin cancer in people born in Australia compared to that in migrants coming to Australia from low ambient UVR regions.

    3. intra-regional population differences that modify the exposure-disease association, eg skin pigmentation variation across Europe may explain the U-shaped relationship between latitude and melanoma risk.

    Further, while many man-made environmental exposures exhibit a linear dose-response function, this may not be true for some natural environmental exposures, including UVR. Thus in evaluating the environmental burden of cancers, UVR exposure contributes both to increased cancer burden (skin and eye cancers) and decreased cancer burden (postulated protective effect on several internal cancers through enhanced vitamin D production). Note also that Pruss-Ustun and Corvalan’s estimate of 19% is likely to be a substantial underestimation of cancers attributable to environment, since they did not include UVR in their assessed environmental factors and UV-induced skin cancer is the commonest cancer in many countries.

    Saracci and Vineis provide useful concluding statements on the values and risks of the population attributable proportion. One proviso is that in considering the PAP, there should be a clear statement of uncertainty in the estimates of the contributing parameters – both the attributable fraction and the population exposure patterns.


    1. Lucas RM, McMichael A, Smith W, Armstrong B. Solar Ultraviolet Radiation. Global burden of disease from solar ultraviolet radiation; Pruss-Ustun A, Zeeb H, Mathers C, Repacholi MH. Geneva: World Health Organization. 2006

    Competing interests


  6. Moving toward a new paradigm

    Richard W. Clapp, Boston University School of Public Health

    13 January 2008

    The Commentary by Saracci and Vineis raises some cogent points regarding the recent attempts to attribute proportions of illness and especially cancer due to environmental exposures. They note the problems of estimating etiologic fractions when the studies on which the estimates are based may be of groups of workers with particular circumstances of exposure. They also point out that many diseases, including cancers, have multiple causes or constellations of causes which means that the sum total of all attributable causes must exceed 100%. This has been noted by many authors, including Doll (1), and it begs the question why the various lists of attributable proportions are commonly forced to sum to a number that everyone agrees is false.

    We have argued (2) that the attempt to ascribe attributable proportions of causes for most cancers is a hopeless exercise and will inevitably lead to incomplete and misleading conclusions given the present state of knowledge. At this point, it is simply barking up the wrong tree, and it leads to endless arguments that shift the focus away from preventing exposures known or highly likely to cause cancer. At its worst, as Saracci and Vineis note, the discussion leads to focusing on genetically susceptible sub-populations and putting the responsibility on individuals for hazard prevention.

    Instead of continuing this endless debate, we call on our public health colleagues to shift the cancer control paradigm to hazard reduction on whatever scale possible. This means continued efforts to eliminate tobacco, continued vigilance about the quality and content of food, continued education about infectious agents that are carcinogenic, and renewed efforts to assure that industrial products are manufactured safely and that consumer products can be used and disposed of without causing cancer in workers and communities. Although obvious, it seems important to remember that the reason we have so much cancer in industrialized societies is because we are exposed to so many carcinogens. The point, therefore, is to reduce these exposures wherever possible, including at their source. This requires a chemicals policy at least as strong as the current REACH procedures in the EU, but ultimately will require a shift in thinking about the long-term human effects on the ecosystem.

    To those who say this is framing the issue too broadly, we would add our support to the points made by Tony McMichael in his previous comment (posted 02 January 2008). Unless we start to frame public health in such a broad scope, we will prolong “a now inadequate twentieth-century view of the full scope and content of the environmental risks to health that bear on current and future generations of humans.”

    Richard Clapp, D.Sc., MPH and Molly Jacobs, MPH

    1. Doll R. Epidemiological evidence of the effects of behavior and the environment on the risk of cancer. Recent Result Cancer Res. 1998;154:3–21.

    2. Clapp R, Howe G, Jacobs M. Environmental and Occupational causes of cancer re-visited. J Public Health Policy 2006;27:61-76.

    Competing interests

    None declared.