1
Introduction: Why Molecular Epidemiology?
Chris Wild,1 Seymour Garte2 and Paolo Vineis3
1University of Leeds, UK, 2University of Pittsburgh Cancer Institute, USA, and 3Imperial College London, UK
Physicians, public health workers, the press and the public at large are increasingly preoccupied with ‘environmental risks’ of disease. What are the causes of Alzheimer’s disease? Are the causes environmental or genetic, or a mixture of the two? Does exposure to particles from incinerators or traffic exhausts cause cancer? Epidemiology has traditionally tried to answer such important questions. For example, the associations between tobacco smoking and lung cancer, between chronic hepatitis B virus infection and liver cancer, and between aromatic amines and bladder cancer are now considered to be ‘causal’, i.e. there is no doubt about the causal nature of these relationships. This is because the same positive observations have been made in a large number of settings, the association is very strong, it is biologically plausible, there is a dose-response relationship, and we cannot explain away the association in terms of bias or confounding. In the case of aromatic amines, strong animal evidence was available before observations in humans.
But not all issues of causality in human disease from environmental exposures are so clear. Consider two examples. Is there a ‘causal’ association between dietary exposure to acrylamide and cancer in humans? Are polycyclic aromatic hydrocarbons (PAHs) a cause of lung cancer? These examples are obviously much more difficult to resolve than the previous ones. In the case of cigarettes, a simple questionnaire proved accurate enough to allow a reasonable estimation of exposure, while in the case of aromatic amines there were rosters in the chemical industries that allowed unequivocal identification of exposure to single agents for all workers. In contrast, estimation of the intake of acrylamide from French fries and other sources, using dietary questionnaires, is an almost desperate enterprise. For PAHs, the sources of exposure are multiple (e.g. diet, air pollution from car exhaust, heating, industrial pollution, specific occupations), making it almost impossible to have an estimate of total PAH exposure based on a questionnaire. One can measure PAHs in ambient air through air sampling (in the work environment or with a personal monitor); however, the level of PAHs in the air is only indirectly associated to the amount of PAHs that actually enter the body and end up binding to DNA. The ability of PAHs to reach DNA and bind to it depends on individual metabolic capabilities (which are in part genetically determined), involving a number of different enzymes and pathways. Finally, the ability of PAHs to lead to heritable changes in DNA (mutations) is additionally related to the individual’s ability to repair DNA damage.
For these reasons, starting at least in 1982 with a paper by Perera and Weinstein but probably before with a paper by Lower (Vineis 2007), ‘molecular epidemiology’ was introduced into the practice of cancer research. A simple definition is that:
’#x2026; it entails the inclusion in epidemiologic research of biologic measurements made at the molecular level - and is thus an extension of the increasing use of biologically based measures in epidemiologic research’ (McMichael 1994).
This corresponds to one of the first (if not the first) definition:
Advanced laboratory methods are used in combination with analytic epidemiology to identify at the biochemical or molecular level specific exogenous and/or host factors that play a role in human cancer causation (Perera and Weinstein 1982).
However, the terminology has been criticized by some:
The term ‘molecular epidemiology’ may suggest the existence of a sub-discipline with substantive new research content. Molecular techniques, however, are directed principally at enhancing the measurement of exposure, effect, or susceptibility, and not at formulating new etiologic hypotheses. As techniques of refinement and elaboration, the integration of molecular measures into mainstream epidemiologic research can offer higher resolution answers in relation to disease causation’ (McMichael 1994).
In this book, many examples are drawn from cancer epidemiology but there is also reference to other chronic degenerative conditions, including vascular disease, diabetes and neurodegenerative disease, which share some of the challenges with cancer in terms of establishing causality. In contrast, we have not included specific emphasis on infectious disease, where the term ‘molecular epidemiology’ was introduced many years earlier than in the cancer epidemiology field.
The goals of the new discipline of molecular epidemiology are the same as those suggested by the few examples above. First of all, to contribute to better estimation of exposure, including ‘internal’ exposure, through the measurement of end-points, such as chemical metabolites and adducts (e.g. haemoglobin adducts for acrylamide, DNA adducts for PAHs). Second, genetic susceptibility, emerged as became an important subject for enquiry, since it became clear that between exposure and effect there was a layer of metabolic reactions, including activation, deactivation and DNA repair, which affected the dose-response relationship in a fashion analogous to other susceptibility factors, such as age, sex and nutritional status. A further goal of epidemiology is to reduce disease burden by identification of risk factors for disease. It took a long time to discover the association between aromatic amines and bladder cancer, and thus hundreds of workers died from causes that could have been avoided. One limitation therefore of cancer epidemiology is the long latency period (decades) after exposure starts and before the disease is clinically diagnosed. For this reason, epidemiologists have been searching for early lesions that could be reasonably used as surrogates of the risk of cancer. Chromosome aberrations, gene mutations and, more recently, gene expression and epigenetics have been introduced as intermediate markers in the pathway that leads from exposure to overt disease, thus adding to the categories of exposure and susceptibility biomarkers mentioned above. The goal of all these efforts in biomarker development is to allow faster and earlier detection of disease in individuals, as well as to shorten the time needed to identify possible human carcinogens.
Molecular epidemiology studies will normally employ a number of tools for the measurement of exposure, susceptibility and disease, e.g. questionnaires, job-exposure matrices, data from environmental monitoring, routinely collected health data and biomarkers. The biomarkers have certain properties which influence their application. For example, some biomarkers of exposure may only relate to the recent past, whilst the level of others may be affected by the presence of disease (reverse causation). Consequently, careful consideration of the properties of the biomarkers is needed in relation to how they are to be applied from the point of study design through to data analysis. This book therefore begins (Chapters 2-5) by discussing study design, particularly in light of the complex interplay between the environment and genes, thus laying a foundation for the subsequent parts of the book.
Molecular epidemiology has had several success stories (see Table 1.1) in all three categories of biomarkers - exposure, susceptibility and early response. Biomarkers of exposure can make a significant contribution to establishing disease aetiology. For example, aflatoxins were structurally identified in 1963 and shown to be liver carcinogens in animals shortly afterwards. However, despite evidence from ecological studies, it was only with the development of biomarkers of individual exposure that convincing evidence of the association with human liver cancer risk was obtained in a prospective cohort study in China, published in 1992 (IARC 1993). In the case of Helicobacter pylori and gastric cancer, the time between identification of the pathogen and evaluation as a human carcinogen by IARC was around 10 years; this was in no small part due to the early availability of serum markers (antibodies to bacterial antigens) to establish exposure status (IARC 1994). The latter example also demonstrates the value of being able to establish long-term past exposure to putative risk factors, something which has been easier with infectious agents than with chemical exposures. The development and validation of biomarkers of exposure to environmental risk factors therefore remains an outstanding challenge to molecular epidemiology (Vineis 2004; Wild 2005). The principles of biomarker development, validation and application are discussed by Vineis and Garte and by Nieuwenhuijsen in Chapters 6 and 7, followed by a number of examples of categories of biomarker described in chapters by Hecht, Phillips, and Berwick and Albertini (Chapters 8-10).
Table 1.1 Discoveries that support the original model of molecular epidemiology1
Exposure/biologically effective dose |
DNA adducts | PAHs, aromatic compounds | Tang et al. 2001 |
| AFB1 | Ross et al. 1992 |
Albumin adducts | AFB1 | Wang et al. 1996 Gong et al. 2002 |
Haemoglobin adducts | Acrylamide Styrene 1,3-Butadiene | Hagmar et al. 2005 Vodicka et al. 2003 Albertini et al. 2001 |
Preclinical effect (exposure and/or cancer) |
Chromosome aberrations | Lung Leukaemia Benzene | Bonassi et al. 2004 Smith et al. 2005 Holeckova et al. 2004 |
HPRT | PAHs 1,3-Butadiene | Perera et al. 200... |