Carcinogenesis Advance Access originally published online on March 14, 2007
Carcinogenesis 2007 28(5):913-915; doi:10.1093/carcin/bgm034
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
‘Environment’ in cancer causation and etiological fraction: limitations and ambiguities
1 International Agency for Research on Cancer, 150 cours Albert-Thomas, 69008 Lyon, France
2 International Epidemiology Institute, 20850 Rockville, MD, USA
3 Mario Negri Institute, 20157 Milan, Italy
* To whom correspondence should be addressed. Tel: +33 4 72738441; Fax: +33 4 72738320; Email: boffetta{at}iarc.fr
 |
The notion of environmental cancer |
|---|
 Top The notion of environmental... Estimates of the fraction...A plea for cautionReferences |
|---|
The definition of environment refers to external physical conditions that may affect human health (1), but the word is used in the medical literature with different connotations, both in English and in other languages. This generates confusion in the medical literature, when the role of environment is discussed in the context of disease etiology. In particular, research on the role of environment in human carcinogenesis suffers from this ambiguity.
On the one hand, environment can encompass all non-genetic factors such as diet, lifestyle and infectious agents. In this broad sense, the environment is implicated in the causation of the majority of human cancers, as has been demonstrated since the 1960s (2). Such a broad meaning of the word environment is assumed when referring to ‘gene–environment interactions’. On the other hand, environmental factors can include only the (natural or man-made) agents and circumstances encountered by humans in their daily life, upon which they have no or limited personal control. In this sense, environmental factors are restricted to air, water, soil and food pollutants, including physical pollutants such as sources of ionizing radiation.
These ambiguities in the terminology and the inconsistencies in the use of the vocabulary by cancer researchers contribute to public confusion regarding the role of the environmental causes of cancer. A distinction relevant to cancer prevention may be made between factors related to personal behaviors, lifestyle (e.g. tobacco smoking and alcohol drinking), involuntary exposures, such as those linked to air, water, soil or food pollutants, and occupation. It would be preferable to abandon the term environment and to use terms such as ‘non-genetic’ or ‘modifiable’ determinants of disease (broad sense of environment) and ‘pollutants’ (narrow sense).
|
Estimates of the fraction of cancer attributable to environmental factors |
|---|
|
TopThe notion of environmental...Estimates of the fraction...A plea for cautionReferences |
|---|
Since it is not appropriate to attribute all non-genetic cancers to ‘environmental’ causes, the questions remain of which fraction of cancers (either overall or of specific neoplasms) are attributable to environmental factors, on the basis of the available evidence. Attribution of cases of cancer—or any other disease—to a given factor assumes that causality has been demonstrated and that the removal of the factor would result in the elimination of a proportional number of cases, once the adequate time latency is taken into account. Several estimates of the proportion of cancers attributable to environmental factors have been reported. In particular, a recent World Health Organization report on environment and health concluded that 19% of all cancers are globally attributable to environmental factors, in the narrow sense defined above, but also including occupational exposures (3). Other reviews, not including occupational exposures, resulted in estimates in the range of 1–3% of all cancers (4–6). These estimates have to be taken with great caution, as they are subject to important methodological limitations. They depend on both the magnitude of the estimated cancer risk entailed by the exposure and the estimated frequency or extent (i.e. dose) of exposure. Errors, such as those due to bias or confounding, in either the strength of the association of the frequency or level of exposure would affect the corresponding estimate of attributable risks. We briefly discuss in the following paragraphs the main sources of error. However, while these limitations may well leave a range of uncertainty of a factor of three or four, they can hardly justify differences of an order of magnitude.
Lack of correspondence between risk and exposure prevalence data
Several risk estimates used in the calculation of environment-attributable cancer derive from meta- or pooled-analyses of epidemiological studies that combine results generated in different populations. The case of non-occupational exposure to asbestos is a typical example of this problem. Studies on residential asbestos have been conducted in circumstances of high exposure (e.g. residence in a mining area or near a processing plant) to maximize their ability to detect an elevated risk (Table I). A meta-analysis of results of the 10 studies reported in Table I resulted in an overall relative risk of mesothelioma equal to 3.5 (95% confidence interval 1.8–7.0) (5). In order to calculate an attributable fraction, one needs an estimate of the proportion of the population experiencing circumstances of exposure comparable with those experienced by the populations included in the studies listed in Table I. Few estimates are available of the proportion of the population experiencing non-occupational exposure to asbestos. In the case of Europe, the figure of 5% has been included in a World Health Organization report on air quality (17). However, this estimate is based on a broader definition of exposure to asbestos as air pollutant as compared with that used in the studies listed in Table I. The combination of a relative risk of 3.5 (which incidentally ignores the possibility of confounding by employment in the asbestos industry, a source of bias that was not excluded in all the studies listed in Table I) and an exposure frequency of 5% results in a fraction of mesothelioma attributable to non-occupational exposure to air pollution equal to 11%, corresponding to 174 cases/year among men and 30 cases/year among women in a country like Great Britain (18); however, it is highly probable that this represents an overestimate of the true effect of non-occupational exposure to asbestos.
|
This form of bias, due to the combination of risk estimates calculated on small groups of heavily exposed individuals and prevalence data based on loose definitions of exposure, leads to the inclusion of cancer cases caused by other factors in the estimate of the attributable fraction. In the case of occupation-attributable risk of cancer, relative risks are often derived from studies conducted among workers experiencing high exposure levels, whereas the proportion of exposed workers is derived from surveys based on a broader definition of exposure, thus also including workers at low (or no) excess risk. This bias, leading to an overestimation of the number of cancers attributable to occupation, may lead to totally unrealistic estimates, discussed in detail in the comprehensive review of the causes of cancer published by Doll and Peto in the early 1980s (4). They showed that the application in an Occupational Safety and Health Administration (OSHA) document (19) of a relative risk of five for respiratory cancer observed in a Norwegian cohort of refinery workers heavily exposed to nickel in the first part of the 20th century (20) to
1 400 000 USA workers exposed to (any level of) nickel in 1972 would have implied a nickel-related excess of 15 000 lung cancers during the following decades. They also estimated that if indeed 15 000 lung cancers were caused by occupational exposure to nickel, an additional 5 000 sinonasal cancers would have also occurred, since the number of excess sinonasal cancers in studies of nickel refinery workers is approximately one-third of that of excess lung cancers. However, in 1973–1977, there were only 274 deaths per year in the USA due to sinonasal cancer, and 158 in women. It is unlikely that more than half of the total sinonasal cancers in men are due to occupational exposures (including agents other than nickel), and the number of 5 000 excess cases due to nickel is ten times higher than any estimate based on more reasonable assumptions. This type of error, due to a correct calculation based on wrong combinations of exposure data and risk estimates, has continued to affect subsequent studies.
Residual confounding in risk estimates
Results of individual studies included in meta- and pooled-analyses may also be subject to bias and residual confounding. Evaluations of causality based on epidemiological studies include an assessment of the possibility that bias or confounding explains the association observed between a specific agent and the disease of interest. However, the conclusion that an association cannot be explained by bias, confounding or chance is usually a qualitative statement, that does not imply that the magnitude of the association is not affected by bias or confounding. Residual confounding by tobacco smoking is a common concern in the interpretation of results regarding respiratory carcinogens occurring either as pollutants or in the workplace. Evidence from occupational studies conducted in North America suggests that, in the case of cigarette smoking, it is unlikely that the confounding relative risk (i.e. the relative risk due to confounding by cigarette smoking rather than the true effect of the occupational exposure) is >1.3 (21). This argument has been used to dismiss confounding by smoking [see for example the evaluation of cancer risk associated with employment as painter made within the framework of the International Agency for Research on Cancer Monographs on the basis of results of occupational studies (22)]. While valid in qualitative terms, this argument does not imply that confounding is not occurring, and in the case of modestly elevated relative risks, as those found for most pollutants, a confounding relative risk in the order of 1.3 may well be important. For example, an increased risk of lung cancer was detected among Chinese women exposed to heavy levels of cooking and heating fumes from use of smoky coal in poorly ventilated houses (23). It is unclear whether a similar effect exists in circumstances of lower exposure, such as those entailed by the use of fossil fuels in Europe or North America; in any case, the relative risk in these countries is probably small, in the order of 1.2 (24), i.e. within the realm of possible residual confounding by smoking. In case of low relative risks, restriction to unexposed individuals, as in the case of studies of involuntary smoking conducted among lifetime non-smokers (25), is a more convincing argument against confounding than the results of analyses including some approach to ‘control for’ confounding via stratification or modeling.
Relevance of past to present exposure circumstances
There is evidence that exposure level to many pollutants has decreased over time and the quality of the exposure itself (e.g. the composition of complex mixtures such as air pollution) has changed. These changes are largely due to the control measures implemented in many countries following the suspicion of a cancer risk. For example, levels of dichloro-diphenyl-trichloroethane (DDT) in human fat, serum or milk have decreased in many countries more than one order of magnitude between the 1970s and the 1980s (26). For most human carcinogens, risk estimates reflect the effect of exposure levels that occurred two to three decades before. The evidence of a risk following exposure to pollutants comes mainly from studies conducted between the 1970s and the 1990s (see for example the studies on non-occupational exposure to asbestos listed in Table I), and the risks are the result of exposures that occurred between the 1940s and the 1970s. The relevance of risk functions derived from the observation of the effect of exposures in the past to current risks is debatable. Future risks caused by current exposures should be modeled on the basis of recent exposures. This was made, for instance, for mesothelioma in Great Britain (18) and Europe (27), and suggested that, after a peak early this century, the number of mesothelioma deaths is expected to decline rapidly after 2010.
Errors in quantitative exposure–response models
In addition to residual confounding, risk estimates based on temporal exposure contrasts within the same population can be quantitatively wrong because of errors in the estimated dose–response relationship resulting from underestimation of past exposure levels. As discussed above, levels of exposure to many pollutants have decreased over time, and estimates of dose–response relationships consist essentially of comparisons of past (high level) and current (low level) exposure circumstances. Past exposure levels might be underestimated as a result of either low sensitivity of measurements available for the past or reconstruction of past exposure based on knowledge of current levels. Figure 1 reports the results of a hypothetical study comparing no exposure (reference category) with high exposure (relative risk 3.7 for 30 units of exposure in Figure 1). If exposure is expressed on a continuous scale, a dose–response relationship can be derived, upon which risk at low exposure level can be estimated. However, if the quantitative estimate of the ‘high exposure’ category is underestimated (relative risk 3.7 for 20 units in Figure 1), the slope of the dose–response relationship will be overestimated, and so will be the risk estimate at low doses. This effect of misclassification of high exposures, mainly belonging to the past, on the slope of the dose–response relationship has been discussed by Siemiatycki et al. (28) in the case of retrospective assessment of occupational exposure to asbestos. Similar considerations may apply to risk estimates for pollutants, e.g. drinking water contaminants.
|
Lack of consideration of effect modifiers
Even if the risk estimates for pollutants are unbiased and consistent with the exposure prevalence data, they may not be directly applicable to any population since the effect of any risk factor depends, in addition to its intrinsic carcinogenic properties, on the presence and level of other factors (effect modifiers), whose distribution varies across populations. Age, time factors related to exposure (e.g. age at starting, stopping or changing level of exposure) and genetic susceptibility factors (e.g. polymorphisms) are examples of such effect modifiers.
|
A plea for caution |
|---|
|
TopThe notion of environmental...Estimates of the fraction...A plea for cautionReferences |
|---|
We have discussed the ambiguity in the use of the term environment to denote a constellation of non-genetic causes of cancer as well as methodological problems in the estimation of attributable risks with respect to pollutants. Such ambiguous terminology should be avoided and caution should be exercised in providing and interpreting estimates of attributable and avoidable numbers of cancers; these figures should be used simply as benchmarks to assess the relative importance of various risk factors. Although attributable risks derived from epidemiological results might often be underestimated because of the lack of sensitivity inherent in the discipline, we have presented several types of error that would lead to a bias in the opposite direction and that are less widely appreciated. A systematic combination of these errors and bias in a single direction (3) may well lead to estimates of cancers attributable to pollutants one order of magnitude larger than the range of reasonably accepted estimates (4–6).
|
Acknowledgments |
|---|
This work was partly supported by Environmental Cancer Risk, Nutrition and Individual Susceptibility, a network of excellence operating within the European Union 6th Framework Program, Priority 5: ‘Food Quality and Safety’ (Contract No 513943).
Conflict of Interest Statement: None declared.
|
References |
|---|
|
TopThe notion of environmental...Estimates of the fraction...A plea for cautionReferences |
|---|
- Merriam-Webster's Collegiate Dictionary (2006) 11th edn (Merriam-Webster, Springfield, MA).
- Doll R. (1969) The geographical distribution of cancer. Br. J. Cancer 23:1–8.[Medline]
- Prüss-Üstün A, et al. (2006) Preventing Disease through Healthy Environments. Towards an Estimate of the Environmental Burden of Disease(WHO, Geneva).
- Doll R, et al. (1981) The Causes of Cancer(Oxford University Press, Oxford).
- Boffetta P, et al. (2003) Contribution of environmental factors to cancer risk. Br. Med. Bull. 68:71–94.
[Abstract/Free Full Text] - Boffetta P. (2006) Human cancer from environmental pollutants: the epidemiological evidence. Mutat. Res. 608:157–162.[ISI][Medline]
- Botha JL, et al. (1986) Excess mortality from stomach cancer, lung cancer, and asbestosis and/or mesothelioma in crocidolite mining districts in South Africa. Am. J. Epidemiol. 123:30–40.
[Abstract/Free Full Text] - Theriault GP, et al. (1978) Mesothelioma and asbestos in the province of Quebec, 1969-1972. Arch. Environ. Health 33:15–19.[ISI][Medline]
- Camus M, et al. (1998) Nonoccupational exposure to chrysotile asbestos and the risk of lung cancer. N. Engl. J. Med. 338:1565–1571.
[Abstract/Free Full Text] - Magnani C, et al. (1995) Pleural malignant mesothelioma and non-occupational exposure to asbestos in Casale Monferrato, Italy. Occup. Environ. Med. 52:362–367.[Abstract]
- Magnani C, et al. (2001) Increased risk of malignant mesothelioma of the pleura after residential or domestic exposure to asbestos: a case-control study in Casale Monferrato, Italy. Environ. Health Perspect. 109:915–919.[ISI][Medline]
- Magnani C, et al. (2000) Multicentric study on malignant pleural mesothelioma and non-occupational exposure to asbestos. Br. J. Cancer 83:104–111.[CrossRef][ISI][Medline]
- Newhouse M, et al. (1965) Mesothelioma of pleura and peritoneum following exposure to asbestos in the London area. Br. J. Ind. Med. 2:261–269.[Medline]
- Howel D, et al. (1997) Routes of asbestos exposure and the development of mesothelioma in an English region. Occup. Environ. Med. 54:403–409.[Abstract]
- Luce D, et al. (2000) Environmental exposure to tremolite and respiratory cancer in New Caledonia: a case-control study. Am. J. Epidemiol. 151:259–265.
[Abstract/Free Full Text] - Hansen J, et al. (1998) Environmental exposure to crocidolite and mesothelioma: exposure-response relationships. Am. J. Respir. Crit. Care Med. 157:69–75.[ISI][Medline]
- World Health Organization. (1987) Air Quality Guidelines for Europe (WHO Regional Publications, European Series No 23)(World Health Organization Regional Office for Europe, Copenhagen) pp. 182–199.
- Hodgson JT, et al. (2005) The expected burden of mesothelioma mortality in Great Britain from 2002 to 2050. Br. J. Cancer 92:587–593.[ISI][Medline]
- Bridbord K, et al. (1978) Estimates of the Fraction of Cancer in the United States Related to Occupational Factors(Bethesda, MD, US NCI, NIEHS, NIOSH).
- Pedersen E, et al. (1973) Cancer of respiratory organs among workers at a nickel refinery in Norway. Int. J. Cancer 12:32–41.[ISI][Medline]
- Siemiatycki J, et al. (1994) Are the apparent effects of cigarette smoking on lung and bladder cancers due to uncontrolled confounding by occupational exposures? Epidemiology 5:57–65.[ISI][Medline]
- International Agency for Research on Cancer. (1989) IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, Vol. 47. Some Organic Solvents, Resin Monomers and Related Compounds, Pigments and Occupational Exposures in Paint Manufacturing and Painting(Lyon, IARC) pp. 423.
- Boffetta P. (2004) Epidemiology of environmental and occupational cancer. Oncogene 23:6392–6403.[CrossRef][ISI][Medline]
- Lissowska J, et al. (2005) Lung cancer and indoor pollution from heating and cooking with solid fuels: the IARC international multicentre case-control study in Eastern/Central Europe and the United Kingdom. Am. J. Epidemiol. 162:326–333.
[Abstract/Free Full Text] - International Agency for Research on Cancer. (2004a) Tobacco smoke. Tobacco Smoke and Involuntary Smoking. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans(Lyon, IARC) Vol. 83: pp. 51–1187.
- International Agency for Research on Cancer. (1989) IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, Vol. 53. Occupational Exposures in Insecticide Application, and Some Pesticides(Lyon, IARC) pp. 208–216.
- Peto J, et al. (1999) The European mesothelioma epidemic. Br. J. Cancer 79:666–672.[CrossRef][ISI][Medline]
- Siemiatycki J, et al. (1998) Invited commentary: is it possible to investigate the quantitative relation between asbestos and mesothelioma in a community-based study? Am. J. Epidemiol. 148:143–147.
[Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  D. Mirabelli and E. Merler 'Environment' in cancer causation and etiological fraction: limitations and ambiguities Carcinogenesis, May 1, 2008; 29(5): 1083 - 1083. [Full Text] [PDF]
|
|
|
|
|
|
A. Pruss-Ustun and C. Corvalan Reply to the recent article by Boffetta et al. [28(5):913 915] on attribution of cancer to the environment Carcinogenesis, August 1, 2007; 28(8): 1849 - 1849. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||