Oxford Textbook of Public Health
Paolo Boffetta, Paul Brennan, and Rodolfo Saracci
Principles of primary and secondary cancer prevention
The global burden of neoplasms
Distribution, causes, and prevention of individual neoplasms
Cancer of the oral cavity and pharynx
Cancer of the oesophagus
Cancer of the stomach
Cancer of the intestine
Cancer of the liver and biliary tract
Cancer of the pancreas
Cancer of the respiratory tract
Neoplasms of the bone and soft tissue
Cancer of the skin
Cancer of the breast
Cancer of the female genital organs
Cancer of the male genital organs
Cancer of the urinary organs
Cancer of the nervous organs
Cancer of the endocrine glands
Neoplasms of the lymphatic and haematopoietic organs
Overview of the causes of human cancer
Obesity and physical exercise
Occupational and environmental agents
Perinatal and growth factors
Ionizing and non-ionizing radiation
Medical procedures and drugs
Neoplasms include a family of diseases, several hundreds of which can be distinguished in humans by localization, morphology, clinical behaviour, and response to therapy. Whether considered from a biological, clinical, or public health viewpoint, it is the malignant and invasive nature of many of these diseases which is of most importance.
Benign neoplasms represent localized growths of tissue with predominantly normal characteristics: in many cases they cause minor symptoms and are amenable to surgical therapy. Benign tumours, however, can become clinically very important when they occur in organs in which compression is possible and surgery cannot be easily performed (e.g. the brain), and when they produce hormones or other substances with a systemic effect (e.g. adrenaline produced by benign phaeochromocytoma). Relatively little is known about the distribution and causes of most benign neoplasms and, with the exception of benign brain neoplasms, they will not be discussed further in this chapter.
Malignant neoplasms are characterized by progressive growth of tissue with structural and functional differences with respect to the normal tissue. In many cases, the alterations can be so important that it becomes difficult to identify the tissue of origin. A peculiarity of most malignant tumours is the ability to migrate and colonize other organs (process of metastasization) via blood and lymph vessel penetration. The presence and extension of metastases are often the critical factors to determine the success of therapy and the survival of cancer patients.
The pace of growth of malignant neoplasms varies widely, and subclinical neoplasms are often found at autopsy of individuals who have died from other causes. It is assumed that for epithelial neoplasms, 10 or more years elapse between the beginning of the transformation process and the clinical diagnosis. For other types of neoplasms (e.g. sarcomas, lymphomas, leukaemias) this time might be shorter. The long process of carcinogenesis justifies the efforts to develop and apply screening approaches for early detection of subclinical neoplasms in healthy individuals.
At the molecular level, the process of malignant transformation is characterized by alterations in several genes that are responsible for the control of the replication cycle of the cell and other regulatory functions. Many oncogenic genes have been identified, and the distribution of their alterations varies among different neoplasms. However, neoplasms which are morphologically and clinically identical often include different genetic alterations, suggesting that the malignant transformation may result from the accumulation of genetic damage through different pathways.
Most malignant neoplasms arise from epithelial tissues and are defined as carcinomas or cancers. In practice, however, the terms ‘malignant neoplasm’, ‘malignant tumour’, and ‘cancer’ are used interchangeably. Neoplasms are classified according to the International Classification of Diseases—Oncology (WHO 1990) into topographical categories (according to the organ where the neoplasm arises) and morphological categories (according to the characteristics of the cells). More and more often, neoplasms are characterized at the clinical level according to phenotypic aspects (e.g. presence of receptors, production of proteins) and genetic alterations (e.g. mutation in a given gene).
Knowledge about the causes and the possible preventive strategies for malignant neoplasms has greatly advanced during the last century. This has been largely based on the development of cancer epidemiology, which has in turn benefited from the establishment of population-based cancer registries in many areas of the world (see below). In parallel to the identification of the causes of cancer, primary preventive strategies have been developed. Secondary preventive approaches have also been proposed and in some cases their effectiveness has been evaluated. A careful consideration of the achievements of cancer research, however, suggests that the advancements in knowledge about the causes of cancer have not been followed by an equally important reduction in the burden of cancer. Part of this paradox is explained by the long latency occurring between exposure to carcinogens and development of the clinical disease. Changes in exposure to risk factors are therefore not followed immediately by changes in disease occurrence. The main reason for the gap between knowledge and public health action, however, rests with the cultural, societal, and economic aspects of exposure to most carcinogens.
The identification of environmental and genetic determinants of cancer relies on two complementary approaches, the epidemiological and the experimental. The epidemiological approach has produced both general and specific evidence for the role of different types of agents in cancer causation. The evidence of a more general nature derives from the observations of considerable variation (often 10-fold to 100-fold) of the incidence rates of most cancers in different populations, defined according to geographical area. Table 1 reports the ratio of the 80th to the 20th percentile of the ranking of country-specific incidence rates of selected countries, as estimated by Ferlay et al. (1998). For all gender-specific rates, the ratio is above 2, and for several it is close to 10. This comparison is based on stable figures, but masks ever larger variations among very-high-risk and very-low-risk areas, which for many neoplasms may reach 100- or even 1000-fold differences. Variations are also shown within countries or according to other characteristics such as ethnic group, religion, or social class. For instance, when contrasted with other religious groups, the Mormons of Utah and the Seventh-Day Adventists of California exhibit low rates for cancers of the respiratory, gastrointestinal, and genital systems. This marked variation in rates according to different axes of exploration is unlikely to be explained chiefly by concomitant genetic variations, and indicates the role of environmental (including lifestyle) determinants.
Table 1 Ratio of the 20th and 80th percentile in the ranking of country-specific estimated age-standardized incidence rates of selected cancers
In addition, the observation of changes of incidence in migrant groups after they have moved to a new living environment suggests a major role of non-genetic factors. Migrants are selected members of a population and they are likely to be in several aspects different from both their population of origin and from the population of the receiving countries. This demands caution when interpreting data from migrant studies, particularly when dealing with minor differences in rates. However, studies like those of Japanese migrants to Hawaii show an assimilation of their cancer incidence rates to the pattern of Caucasians, with the emergence of large differences with respect to the rates in Japan, which appears difficult to explain solely on the basis of selective factors.
Finally, changes in incidence rates in time, particularly when they take place over a few decades are incompatible with a genetic explanation, as changes in the genes of a population pool require much longer intervals. Recorded incidence rates are affected by diagnostic changes and mortality rates are, in addition, affected by changes in treatment effectiveness; however, marked trends like the one for lung cancer (Fig. 1) are most likely to reflect real changes in cancer rates, indicating the operation of environmental factors.
Fig. 1 (a) Age-adjusted cancer death rates, by site, in the United States, 1930 to 1996 for men. (b) Age-adjusted cancer death rates, by site, in the United States, 1930 to 1996 for women. The uterus cancer death rates are for the uterine cervix and uterine corpus combined. (Source: ACS 2000.)
Specific evidence produced by analytical studies (case–control and cohort) has shown the causal role of specific exposures in the aetiology of several malignant neoplasms. One limitation of the epidemiological approach may prove of critical importance in trying to detect comparatively small increases in risk, for example from chemicals polluting the environment. This is that even in the best conditions it is impossible to identify confidently by epidemiological means an increase in risk smaller than perhaps 10 per cent (and serious problems arise in the interpretation of increases below 50 per cent) as the biases inherent in any observational study are of at least this order of magnitude.
Genetic determinants of cancer have also been demonstrated. As discussed below, several inherited conditions carry a very high risk of one or several cancers. High-penetrance genes are identified through family-based and other linkage studies. These conditions, however, are relatively rare and explain only a small proportion of human cancers. Genetic factors, however, are likely to play an important role in interacting with environmental exposure to determine individual cancer risk.
The experimental approach to identification of carcinogens includes at a more basic level the investigation of mechanisms of carcinogenesis. A result of general relevance has been the demonstration that chemical carcinogens are usually metabolized to reactive molecules with high affinity for electron-rich molecules, like DNA, with which they form combinations (‘adducts’) (Wild and Pisani 1997). These have been characterized in structure and properties, but the specific, critical lesions directly responsible for the carcinogen-induced malignant transformation have not yet been fully identified. However, there is growing evidence that the action of genotoxic carcinogens, possibly through DNA adduction, results in mutation in proto-oncogenes and tumour suppressor genes (Albertini and Hayes 1997). Not all carcinogens, however, act through DNA damage, and other well-established mechanisms involve non-genotoxic mechanisms, such as enhanced cellular proliferation.
The identification of carcinogens via the laboratory relies on three types of tests: (a) long-term (often lifetime) carcinogenicity tests in experimental animals, most commonly rodents (mice, rats, hamsters); (b) short-term tests assessing the effects of chemical agents on a variety of endpoints belonging to three general classes: DNA damage, mutagenicity, and chromosome damage; (c) mechanistic tests, aimed at identifying the intermediate steps in the compound-specific carcinogenic process.
These tests are valuable to the extent that such effects may reflect underlying events in the carcinogenic process. Indeed, consistent positivity in tests measuring DNA damage, mutagenicity, and chromosomal damage is usually regarded as indicating potential carcinogenicity of the tested agent. Results of laboratory tests constitute useful supporting evidence when adequate epidemiological data for the carcinogenicity of an environmental agent exist (e.g. vinyl chloride), but they become all the more essential when the epidemiological evidence is non-existent or inadequate in quality or in quantity. In the latter case, although no universally accepted criteria exist to translate data automatically from long-term animal tests or short-term tests in terms of cancer risk in humans, an evaluation of the risk can be made on a judgmental basis using all available scientific evidence. This policy has been applied by the International Agency for Research on Cancer (IARC) in a systematic programme of evaluation of the carcinogenic risk of chemicals to humans. Within this programme of IARC monographs, agents are classified into group 1 (established human carcinogen), 2A (probable human carcinogens), 2B (possible human carcinogens), and 3 (not classifiable as to carcinogenicity to humans) (IARC 1972–2001). Agents are commonly classified as group 1 when the evidence of their carcinogenicity in humans, derived from epidemiological studies, is considered sufficient, and as group 2A when the evidence in humans is limited and the agent is an experimental carcinogen. Agents in group 2B include mainly experimental carcinogens for which the human evidence is inadequate or non-existent. Between 1972 and 2001, 78 volumes presenting evaluations (and re-evaluations) for 869 chemical, physical and biological agents and groups of agents, as well as exposure circumstances such as occupations, have been published (IARC 1972–2001). A total of 87 agents have been classified as group 1, 63 as group 2A, and 235 as group 2B. The complete list of agents, with their evaluations can be found on the IARC web site (http://www.iarc.fr/).
Principles of primary and secondary cancer prevention
Many determinants of malignant neoplasms—including ultraviolet radiation, ionizing radiation, tobacco smoking, alcohol drinking, a number of viruses and parasites, and a number of chemicals, industrial processes, and occupational exposures—are sufficiently well established to constitute logical priorities for preventive action. Two more reasons add weight to this priority: some of the agents are responsible for sizeable proportions of the cancers occurring today, and for most agents it is in principle feasible to reduce or even to eliminate exposure completely. If this is taken as the objective of preventive action, some practical points are helpful in guiding such action.
Firstly, none of the data sets allowing a relatively accurate definition of dose (exposure)–response relationships for cancer in humans (such data are available for selected agents, such as tobacco smoking, radon, aflatoxin, and dioxin; Cardis et al. 1999) indicate divergence from linearity at low dose, i.e. in the portions of the curves which are of key interest when exposures can be controlled to low levels but, for whatever reason, cannot be completely avoided. From a practical viewpoint, therefore, in assessing the effect of reducing exposure to carcinogens, it is reasonable to build and use a linear dose–response relationship with no threshold.
Secondly, in this formulation, dose means total dose cumulated over time. This implicitly assumes that the carcinogenic effect is equally dependent on the dose rate (dose per unit of time) and on duration of exposure which, multiplied together, give total dose. However, both experimental and epidemiological data indicate that this is not the case. For lifelong exposures in regular smokers, for example, the incidence rate (of lung or bladder cancer in smokers) depends more strongly on duration of exposure, increasing with the fourth power of it, than on dose rate, increasing only with the first or second power of it (Peto 1977). Therefore, for exposures in adult life, and if other biological effects (e.g. genetic) can be excluded, it may be preferable if more people are exposed for shorter periods than fewer people for longer periods.
Furthermore, experimental and epidemiological investigations suggest that the carcinogenic process may be represented as a succession of stages, taking place in the time span from first exposure to a carcinogenic agent to the appearance of clinical cancer (Joint IARC/IPCS/CEC Working Group 1983). In its simplest form, as identified in mouse skin carcinogenesis experiments, the multistage process reduces to two stages: an irreversible ‘initiation’ stage inducing malignant cells, and a ‘promotion’ stage which propagates these cells into a malignant growth. A third stage of ‘progression’, characterized by an increased rate of growth and metastases, as well as an increase in chromosomal changes in the cell, has also been observed. Formal statistical multistage models of carcinogenesis have provided a useful framework to interpret on a common basis of (postulated) mechanism both experimental and epidemiological observations. As the stages are assumed to occur in a specific sequence, some may be described as ‘early’ and some as ‘late’. Epidemiological observations indicate that, for example, smoking has both an early stage effect, as indicated by the existence of a minimum interval of several years before an increase in risk of lung cancer becomes manifest, and a late stage effect, as indicated by the decrease in risk (with respect to non-quitters) soon after stopping smoking. The identification of carcinogens acting mainly at early or late stages would have consequences on the preventive strategies to apply.
The attribution of causality to environmental agents (as done when, for instance, smoking was said to be the cause of some 30 per cent of all cancers; see below) is complicated by the interactive effects of several agents. This is particularly relevant when considering the relative effectiveness of removing (or reducing) exposure to one of two (or more) jointly-acting agents. To illustrate this point, Table 2 presents the results of a large cohort study in insulator workers in the United States and Canada (Hammond et al. 1979), classified according to their exposure to asbestos and to their smoking habits. Due to the multiplicative interaction of asbestos and smoking, the group exposed to both agents has the highest lung cancer rates, greatly in excess not only to those not exposed to either agent, but also to those exposed only to asbestos (the excess rate in the latter group is 47.1, while it is 590.3—that is 12.5 times higher—in the group exposed to both agents). Superficially it could be concluded that, since smoking is by far the dominant factor, its removal will correspondingly be far more effective as a preventive measure than asbestos removal. However, this is not true. In fact, if smoking is removed the fraction, or percentage, of the excess rate removed will be
Table 2 Mortality rate (per 100 000) from lung cancer among American insulator workers, according to exposure to asbestos and smoking
[(590.3 – 47.1)/590.3] = 0.92,
and if asbestos is removed it will be
[(590.3 – 113.13)/590.3] = 0.81.
This means that smoking is more effective than asbestos removal, but only by a factor of 0.92/0.81 = 1.14, and not by a factor of 12.5. The result is not surprising if one considers that the bulk of the effect in subjects exposed to both agents is due to their positive interaction (synergism) and thus is removed whichever of the two agents is eliminated. In practical terms, whenever a positive interaction (synergism) occurs between two (or more) hazardous exposures, there is an enlarged possibility of preventive action; the effect of the joint exposure can be attacked in two (or more) ways, each requiring the removal or reduction of one of the exposures; moreover, the larger the size of the interaction relative to the total effect, the more these ways of attack tend to become equal in effectiveness.
It has been mentioned that a number of chemicals, groups of chemicals, and industrial processes have been found, when proper account has been taken of all evidence (laboratory and epidemiological), to be ‘probably’ carcinogenic for humans, and that a large number of other chemical agents have been demonstrated to be carcinogenic in animals, with no data available in humans (‘possible’ human carcinogens) (IARC 1999). However, these agents should be considered, for practical purposes, as if they entail a risk of cancer for humans and actions minimizing exposure taken accordingly.
Finally, reducing exposure to carcinogens can be implemented in two major ways: by elimination of the carcinogen or its substitution with a non-carcinogen, or by impeding by various means the contact between the carcinogen and people. Reduction of exposure depends in each case on technical and economic considerations. Generally, however, methods of exposure reduction that minimize the number of decisions involved are to be preferred; for instance, change to an innocuous material in preference to having each user exercise caution in using it.
Given the limitations still constraining the primary prevention of many cancers, early detection needs to be considered as a secondary and alternative option, based on the reasonable expectation that the earlier the diagnosis and the stage at which a malignancy is discovered, the better the prognosis. This implies that an effective treatment for the disease exists and that the less advanced the cancer at the preclinical stage, the better the scope for treatment, and the better the prognosis. This latter aspect cannot be taken for granted. It is important to keep early detection separate from secondary prevention. The former aspect belongs to the domain of ordinary medical care, while the latter is conducted in asymptomatic populations through large-scale (mass screening) programmes.
Before a screening programme can be adopted on a large scale, a number of requirements need to be fulfilled in addition to the fundamental one (already mentioned) that an effective treatment for the preclinical malignant lesion must be available also on a sufficient scale. These requirements have been extensively discussed in publications specifically addressing the issue of evaluating screening programmes for cancer (Miller 1996).
Any screening test under consideration (i.e. a relatively simple and rapid test aimed at the presumptive identification of preclinical disease) must be capable of correctly identifying cases and non-cases. In other words, both sensitivity and specificity should be high, approaching 100 per cent. While high sensitivity is obviously important, given that the very purpose of screening is to pick up, if possible, all cases of a cancer in its detectable preclinical phase, it is specificity that plays a dominant role in the practical utilization of the test within a defined population, that is in a screening programme. As the prevalence of a preclinical cancer to be screened in well-defined populations is often in the range of 1 to 10 per 1000, if a test is used with a specificity of 95 per cent, then 5 per cent of results will be false-positives. In other words, for every case which will turn out at the diagnostic work-up to be a true cancer (assuming 100 per cent sensitivity), there will be between five and 50 cases falsely identified as such and ultimately found not to be cancers. This situation is likely to prove unacceptable due to too high psychological and economic costs. One solution is an increase in specificity, for example by developing better tests or combinations of tests, or by changing the criterion of positivity of a given test to make it more stringent (this necessarily decreases sensitivity). In addition, one might select populations with relatively high prevalence of the cancer (‘high-risk’ groups), so as to increase the number of the true positives. Whatever the group on which the programme operates, additional requirements are that the test is safe, easily and rapidly applicable, and broadly acceptable to the population to be examined (e.g. sigmoidoscopy may not be). It has also to be cheap, but what is or is not cheap is better evaluated within a cost-effectiveness analysis of different ways of preventing a cancer case or death, an issue not further discussed here.
If these requirements are met, still nothing is known about the possible net benefit in outcome deriving from the screening programme (in fact, screening test plus diagnostic work-up plus treatment, as applied in a given population). To evaluate benefit, several measures of outcome can be assessed. An early one, useful but not sufficient, is the distribution by stage of the detected cancer cases which, if the programme is ultimately to be beneficial, should be shifted to earlier, less invasive stages of the disease in comparison with the distribution of the cases discovered through ordinary medical care. A second measure of outcome is the survival of cases detected at screening compared with the survival of cases detected through ordinary medical care. This is a superficially attractive but usually equivocal criterion, to the extent that a screening may only advance the time of diagnosis (and therefore the apparent survival time), without postponing the time of death (‘lead-time bias’). A final outcome (and the main test of the programme) is the site-specific cancer mortality in the screened population compared with the mortality in the unscreened population.
Correct unbiased comparison of this outcome, and thus unbiased measure of the effect of the screening programme, can only be made within the framework of a randomized controlled trial, in which two groups of subjects are randomly allocated to the screening programme and to no screening (i.e. receiving only the existing medical care system) or to two alternative screening programmes, for instance, entailing different tests or different intervals between periodical examinations. Unfortunately, largely due to pressures to adopt on a large scale screening programmes hoped to be effective, a situation has often arisen where withholding screening to a group has been regarded as unethical, thus preventing the conduct of a proper experiment. Very few randomized trials evaluating the effectiveness of screening programmes are available. Comparisons made through non-randomized experiments (which also are infrequent for the same reason), or through observational studies of what has happened following the introduction of a screening programme in a population, are liable to bias.
Four types of bias are peculiar to the assessment of screening programmes. Firstly, earlier detection only moves forward the time of a patient’s diagnosis, without postponing the time of death (lead-time bias): should this be the case the result would only be a longer period of morbidity for the patient. Secondly, because of self-selection, people who elect to receive early detection may be different from those who do not: for instance, they may belong to better educated classes, be generally healthier and health conscious, and this could produce a longer survival independent of any effect of early detection. Thirdly, cancers with longer preclinical phases, which may mean less biological aggressiveness and better prognosis, are more likely to be intercepted by a programme of periodical screening than cancers with a short preclinical phase and a rapid aggressive clinical course (length bias). Finally, because of criteria of positivity adopted to maximize yield of early cases, a number of lesions which in fact would never become malignant growths are included as ‘cases’, thus falsely improve the survival statistics (overdiagnosis bias).
As observational studies, however, may represent the only source of information to evaluate screening programmes, they need to be carefully considered. Among purely observational situations, one which may provide particularly useful information is that arising when a relatively small, closed population (with no migration in or out) is saturated over a short interval (a few years) with the screening programme to be evaluated. If good ordinary diagnostic and treatment services are available before, during, and after the introduction of the screening, and the mortality rate from the cancer can be regarded as predictable, a clear change in mortality would represent good scientific evidence for effectiveness of the screening programme.
The global burden of neoplasms
The number of new cases of cancer which occurred worldwide in 2000 has been estimated at about 10 100 000 (Table 3) (Ferlay et al. 2001). Of these, 5400 000 occurred in men and 4700 000 in women. About 4700 000 cases occurred in developed countries (North America, Japan, Europe including Russia, Australia, and New Zealand) and 5400 000 in developing countries. Among men, lung, stomach, colorectal, prostate, and liver cancers are the most common malignant neoplasms (Fig. 2), while breast, colorectal, cervical, lung, and ovarian cancers are the most common neoplasms among women (Fig. 3).
Table 3 Estimated number of new cases of cancer (incidence) and of cancer deaths (mortality) in 2000, by gender and geographical area
Fig. 2 Estimated number of new cancer cases (× 1000) during 2000 in men. (Source: Ferlay et al. 2001.)
Fig. 3 Estimated number of new cancer cases (× 1000) during 2000 in women. (Source: Ferlay et al. 2001.)
The number of deaths from cancer was estimated at about 6200 000 in 2000 (Table 3) (Ferlay et al. 2001). No global estimates of survival from cancer are available: data from selected cancer registries suggest wide disparities between developed and developing countries for neoplasms with effective but expensive treatment, such as leukaemia, while the gap is narrow for neoplasms without an effective therapy, such as lung cancer (Kosary et al. 1995; Sankaranarayanan et al. 1998; Berrino et al. 1999) (Fig. 4). The overall 5-year survival of cases diagnosed during 1985 to 1989 in European Union countries was 41.0 per cent (Berrino et al. 1999).
Fig. 4 Five-year relative survival from cancer in selected populations.
One complementary approach in assessing the global burden of neoplasms is to estimate the loss in disability-adjusted life years (DALYs). This indicator weighs the years of life with disability and adds them to the years lost because of premature death. A recent estimate for 1990 resulted in about 70 000 000 DALYs lost worldwide because of malignant neoplasms, of which 48 000 000 occurred in developing countries and 22 000 000 occurred in developed countries. Lung cancer was responsible for 8 900 000 DALYs, stomach cancer for 7 700 000, liver cancer for 6 600 000, leukaemia for 4 600 000, and breast cancer for 4 200 000 (Murray and Lopez 1996).
Distribution, causes, and prevention of individual neoplasms
This section includes a systematic review of the descriptive epidemiology of the most important malignant neoplasms. It also includes an overview of the current state of knowledge about the risk factors and the strategies for primary and secondary prevention. A global approach is used which excludes important local aspects of the descriptive epidemiology, aetiology, and prevention of neoplasms. For a more detailed review, the reader is referred to a textbook of cancer epidemiology and prevention edited by Schottenfeld and Fraumeni (1996). All incidence and mortality rates are standardized to the world population.
Cancer of the oral cavity and pharynx
Tumours of the oral cavity and the pharynx comprise a broad group of different malignant neoplasms. Cancers of the oral cavity, oropharynx (including the tonsil), and hypopharynx are usually considered as a group, because of shared descriptive characteristics and risk factors, while neoplasms arising from the lip, nasopharynx, and salivary glands are treated separately, because of differences in aetiological factors.
An estimated 454 000 new cases of cancers of the oral cavity and pharynx occurred worldwide in 2000, 70 per cent of which were in developing countries (Ferlay et al. 2001). The estimated number of deaths is 244 000 (187 000 in developing countries) (Ferlay et al. 2001).
Oral cavity, oropharynx, and hypopharynx (Blotet al. 1996)
The incidence of cancers of the oral cavity and pharynx varies over 20-fold between high-risk areas (e.g. France, Switzerland, Italy, Spain, India, African-Americans) and low-risk areas (e.g. China, Nordic countries, the United Kingdom). The highest rates in the world among men are recorded in France, those in women are from India. In all populations, rates in men exceed those in women by a factor of 4 to 10. This cancer is very rare before the age of 30 years; above that age, the rate increases logarithmically up to about 60 years of age, after which the rate of increase declines. Incidence rates have increased in Europe and the Americas during the last decades (Coleman et al. 1993). When looking at subsites within the oral cavity and the pharynx, cancer of the oropharynx and hypopharynx account for as many or more cases than cancer of the oral cavity in high-risk European populations, while cancers of the tongue, floor of the mouth, and other parts of the oral cavity represent the majority of cases in India and the United States.
Consumption of tobacco smoke is the most important risk factor for cancer of the oral cavity and pharynx (Blot et al. 1996). In Western populations, smoking represents the main use of tobacco, and relative risks of oral cancer among smokers, as compared with non-smokers, are of the order of 3 to 10. The risk is higher for heavy smokers, long-term smokers, and smokers of black tobacco or high-tar cigarettes. Cigar and pipe smoking also poses a risk, while stopping smoking is followed by a decrease in risk. In India, chewing of products containing tobacco is the main risk factor for oral cancer, although smoking of cigarettes and ‘bidis’ (consisting of 0.3 g of tobacco hand-rolled in a dried tembuzni leaf) also contribute to the risk. In the United States and Europe, use of tobacco snuff has also been implicated in the genesis of oral cancer.
Consumption of alcoholic beverages increases the risk of oral and pharyngeal cancer (Blot et al. 1996). Relative to abstainers and very light drinkers, the risk in heavy drinkers is of the order of 10. While it is clear that all types of alcoholic beverages increase the risk, it is unclear whether there is a quantitative difference in their carcinogenic effect. There is convincing evidence that the effects of tobacco smoking and alcohol drinking are multiplicative, i.e. that the effect of exposure to both is close to the product of their individual effects (Fig. 5). The combined effect of tobacco smoking and alcohol drinking accounts for almost 80 per cent of cancers of the oral cavity and pharynx in Western populations. Similarly, tobacco chewing and smoking and their combination are responsible for a large proportion of these cancers in India.
Fig. 5 Risk of oral cancer, tobacco smoking, and alcohol drinking. (Source: Blot et al. 1988.)
A diet rich in fruits and vegetables protects from developing cancer of the oral cavity and pharynx in several populations (WCRF 1997). This effect seems to be independent of the effect of tobacco and alcohol. Among the micronutrients that might be implicated in the protective effects exerted by fresh fruits and vegetables are vitamins A, C, E, and B12. A role of b-carotene has been suggested in chemopreventive trials of preneoplastic lesions but not confirmed in trials with clinical cancer as outcome. The role of other dietary factors is unclear, although high consumption of maté, a herbal tea consumed in South America, might also represent a risk factor.
Human papilloma virus DNA has been frequently detected in preneoplastic and neoplastic lesions of the oral cavity; however, its aetiological role has not yet been fully established (IARC 1995). Other viruses are suspected of playing a role in oral carcinogenesis, including herpes simplex 1 and Epstein–Barr viruses (Blot et al. 1996).
Poor oral hygiene and ill-fitting dentures are likely to represent additional risk factors for oral cancer. The use of mouthwash with high alcohol content has also been associated with oral cancer. Several occupations have been sporadically reported to entail an increased risk of oral and pharyngeal cancer. The evidence is somewhat consistent only for employment as a waiter and bartender, probably reflecting an increased consumption of alcohol and tobacco, and for exposure to asbestos.
The role of genetic susceptibility in oral carcinogenesis is probably modest. High-risk families have been reported only occasionally. A role is, however, likely for low-penetrance factors, such as increased sensitivity to mutagens and polymorphism of enzymes implicated in the metabolism of alcohol (alcohol dehydrogenase and aldehyde dehydrogenase).
The descriptive epidemiology of lip cancer is complicated by misclassification with lesions from adjacent areas of the skin and oral mucosa. This neoplasm is rare in most populations and occurs mainly in the lower lip. The main known causes are solar radiation, smoking of cigarettes, cigars and pipes, and exposure to polycyclic aromatic hydrocarbons. Fishermen and farmers are at increased risk. Infection with herpes simplex virus 1 might also play a role (Blot et al. 1996).
Nasopharynx (Yu and Henderson 1996)
The geographical distribution of cancer of the nasopharynx shows high-risk areas in southern China, Southeast Asia, northern Africa, and the Arctic region. Migrant populations from these regions are also at increased risk. It is a rare neoplasm in most Western populations. The male to female ratio is of the order of 2 to 3. In China, the rates are highest in the 45- to 54-year age group, while in low-risk populations a peak is suggested around the age of 20 years.
Infection with Epstein–Barr virus is causally associated with the development of nasopharyngeal carcinoma, although cofactors, which have not been fully understood, are necessary to the virus to exert its action.
Consumption of Chinese-style salted fish, in particular in childhood, has been associated with increased risk of nasopharyngeal cancer, in particular among Cantonese populations from Guangzhou, Hong Kong, and the United States, but also in non-Cantonese populations from southern China and Taiwan. Consumption of other types of salted fish might represent a risk factor in Southeast Asia and the Arctic. Other preserved foods used as weaning food in different areas of China have also been associated to nasopharyngeal cancer: chung choi (a salted root), salted shrimp paste, salted eggs, and preserved fruits. The high rates in northern Africa might be due to consumption of dried mutton, touklia (a spiced mixture of peppers), or harissa (a hot sauce). It is unclear whether the carcinogenic agents in Chinese-style salted fish and other preserved foods are nitrosamines, bacterial mutagens, or other genotoxic substances. Low intake of fresh fruits and vegetables has also been associated with increased risk of nasopharyngeal cancer: in high-risk populations, however, low fruit and vegetable intake is often closely associated with high intake of preserved food, making it difficult to separate the effect of the two factors.
Additional suspected risk factors for nasopharyngeal cancer are occupational exposure to formaldehyde and wood dust, and drinking water rich in nickel, while studies on exposure to fumes and smoke from burning of organic matter, on consumption of Chinese herbal beverages, and on the use of Chinese nasal oil are inconclusive. A moderate association (relative risks of the order of 3 to 5) with tobacco smoking has been detected in several populations. Chronic inflammatory conditions of the ear, nose, and sinuses might act as predisposing factors. Familial aggregation of nasopharyngeal cancer has been shown in China and other high-risk areas: the possible contribution of shared dietary and environmental factors, however, has not been excluded. In addition, there is evidence that certain HLA genotypes increase the risk of this neoplasm. The role of genetic polymorphism to metabolic enzymes is not clear.
Geographical and temporal comparisons of tumours of the salivary glands are complicated by the inclusion of neoplasms with benign or intermediate clinical behaviour in the series of cases. Incidence rates, however, tend to be low, with high rates (of the order of 2 per 100 000) in Alaska and northern Canada. Men experience higher rates than women, but the ratio rarely exceeds 2. Apart from a role of ionizing radiation, the causes of cancer of the salivary glands are largely unknown: among the suspected causes are low vitamin A intake, hormonal or occupational factors, and infection with Epstein–Barr virus. Tobacco smoking does not seem to play a role (Blot et al. 1996).
Avoidance of tobacco smoking, chewing, and snuffing and avoidance of excessive alcohol drinking would represent the main preventive measures for cancer of the oral cavity and pharynx. Additional benefits might be obtained from increase in fruit and vegetable intake and improvement of oral hygiene. Avoidance of excessive exposure to solar radiation would represent the main preventive approach for lip cancer. In populations at high risk of nasopharyngeal cancer from China and possibly other countries, avoidance of salted fish and other preserved food, in particular as weaning food, should be recommended.
Oral inspection aimed to identify preneoplastic lesions might be an effective approach for secondary prevention of oral cancer. The inspection can be performed by medically certified professionals, but also, in particular in high-risk areas from developing countries such as India, by specifically trained health workers. Large-scale preventive trials are on-going, that should provide evidence in favour or against this approach (Sankaranarayanan et al. 2000).
Cancer of the oesophagus (Muñoz and Day 1996)
There are two main histological types of oesophageal cancer: squamous cell carcinoma and adenocarcinoma. The former occurs mainly in the upper and middle third of the organ, while adenocarcinoma occurs in the lower third. Squamous cell carcinoma is the predominant type in most human populations, in particular in populations at very high risk. In this section, oesophageal cancer is discussed as a whole, although most data refer to squamous cell carcinoma. Peculiar aspects of adenocarcinoma are discussed in below.
The geographical distribution of oesophageal cancer is characterized by very wide variations within relatively small areas. Very high rates (over 50 per 100 000) are recorded in both genders from northern Iran and the provinces of eastern China, Shanxi, Henan, and Jiangsu, in certain areas of Kazakhstan and among men from Zimbabwe. Intermediate rates in men (10 to 50 per 100 000) occur in eastern Africa, southern Brazil, the Caribbean, most of China (with the exception of southern provinces, such as Yunnan, Guizhou, Hunan, and Guangxi), regions of central Asia, northern India, southern Europe, as well as in black Americans. Ethnic factors are suggested by the facts that populations at higher risk in central Asia are of Turkish or Mongolian origin, but not of Caucasian origin, and that black Americans experience in both genders two- to threefold higher rates than white people. In men, rates are two- to 10-fold higher than in women. In many high-risk areas, a decrease in the incidence of oesophageal cancer has occurred during recent decades. The opposite pattern has been shown in low-risk populatons, such as northern Europeans and white people in the United States. In the latter country, the increase was mainly accounted for by an increase of adenocarcinoma of the lower oesophagus (see below), while the incidence of squamous cell carcinoma remained stable.
Tobacco smoking is an important risk factor for oesophageal cancer (Muñoz and Day 1996). The risk in heavy smokers, relative to non-smokers, is of the order of 4 to 8. A linear dose–response relationship has been shown for the duration of smoking and average consumption. Quitting smoking substantially reduces the excess risk. Smoking of black tobacco, high-tar, and hand-rolled cigarettes, as well as of pipe smoking, might exert a stronger effect that smoking of other products. Chewing of tobacco-containing products represents an important risk factor in India and southern Africa, but its role has not been confirmed in central Asia. In the latter region, use (smoking and eating) of opium might be (or at least might have been in the past) a reason for the high incidence rates.
Alcohol drinking is also an important risk factor for oesophageal cancer and, in Western populations, its effect seems to be stronger than that of tobacco smoking (Muñoz and Day 1996). It is unclear whether there are differences in the carcinogenic potency of different alcoholic beverages. A reduction in the excess risk of oesophageal cancer is suggested after quitting alcohol drinking, and the decrease seems to be stronger than in the case of quitting tobacco smoking. The effect of alcohol is independent of the effect of tobacco, and the interaction between the two exposures fits well a multiplicative model (Fig. 6). Taken together, tobacco smoking and alcohol drinking account for 90 per cent or more of the cases of oesophageal cancer in western Europe and North America: this proportion, however, is lower in developing countries, in particular in areas at very high risk.
Fig. 6 Risk of oesophageal cancer, tobacco smoking, and alcohol drinking. (Source: Tuyns et al. 1977.)
Intake of large amounts (more than 1 litre/day) of hot maté is an important risk factor for oesophageal cancer in southern Brazil, Uruguay, and northern Argentina. It is unclear, however, whether the effect is due to components of maté or to the high temperature: studies from other areas suggest that intake of hot beverages (e.g. hot tea in Iran, Singapore, and Japan, hot coffee in Puerto Rico, and hot drinks or soups in Hong Kong) increases the risk of oesophagitis and oesophageal cancer, although the evidence is less consistent than in the case of maté.
Dietary factors are likely to play an important role in the aetiology of oesophageal cancer. Reduced intake of fresh fruits and vegetables represents a risk factor. A similar effect has been suggested for low intake of fresh or frozen meat or fish, for low intake of dairy products, and for high intake of barbecued meat. The available data do not allow the full establishment of the potentially preventive role of specific micronutrients from fruits and vegetables, and the results of chemopreventive trials with retinol, riboflavin, vitamin E, zinc, and selenium have failed to show a clear benefit.
In several areas of China, intake of pickled vegetables has been associated with an increased risk of oesophageal cancer. The active carcinogens might be mycotoxins or N-nitroso compounds. Mycotoxins, including fumonosin B1, have also been detected in mouldy corn from high-risk areas in China and southern Africa. In addition, in Japan, eating of bracken fern has been associated with an elevated oesophageal cancer risk. The elucidation of dietary factors implicated in oesophageal carcinogenesis, in particular of the possible role of mycotoxins and N-nitroso compounds (including endogenously formed nitrosamines), would represent an important step in the understanding and prevention of this disease.
Among the other environmental agents suspected to cause oesophageal cancer are infection by human papilloma virus, occupational exposure to asbestos, silica, and combustion fumes, and ionizing radiation.
In addition to Barrett’s oesophagus (see below), other chronic conditions increase the risk of developing oesophageal cancer. Patients suffering from Plummer–Vinson syndrome, a sideropenic dysphagia due to deficit of iron, riboflavin, and other vitamins, had an increased incidence of oesophageal and hypopharyngeal cancers. Oesophageal cancer risk is also increased among coeliac disease patients, possibly because of nutritional deficiencies.
A familial aggegration of oesophageal cancer has been occasionally shown, with joint segregation of a gene responsible for keratosis palmaris et plantaris (tylosis). Studies of families without the tylosis gene have not provided conclusive evidence of an important role of other high-penetrance genetic susceptibility factors in oesophageal cancer. However, low-penetrance genes, including those encoding for enzymes involved in the metabolism of tobacco and alcohol, may play a role in individual susceptibility to this neoplasm.
Adenocarcinoma mainly occurs in the lower third of the oesophagus. Its incidence is higher in Western countries, in white people, and in high social class individuals and has sharply increased in the last decades in most Western countries. In countries such as Scotland, it represents the main type of oesophageal cancer. Barrett’s oesophagus, a columnar metaplasia of the epithelium, is strongly associated with subsequent development of adenocarcinoma. The main risk factor for Barrett’s oesophagus and oesophageal adenocarcinoma is persistent reflux oesophagitis. The associations between adenocarcinoma of the oesophagus and tobacco smoking and alcohol drinking are weak. A protective role of high intake of fruits and vegetables and low intake of salty food has been suggested (Muñoz and Day 1996).
Control of tobacco smoking and elevated alcohol drinking remains the main preventive approach in reducing the burden of squamous cell oesophageal cancer in Western populations. Improved diet, in particular increase in consumption of fresh fruits and vegetables, is likely to represent an additional important preventive step. The incomplete understanding of the role of opium derivates, mycotoxins, and other factors complicates the elaboration of preventive strategies in many high-risk regions, although decrease in intake and temperature of maté might be important in South America.
Chemoprevention with micronutrients might represent an additional strategy, although the results of the early trials do not strongly suggest any candidate for population-based actions.
Secondary prevention has been attempted in high-risk areas through endoscopy: also in this case, however, the available evidence does not justify activities at the population level.
Cancer of the stomach
Stomach cancer was the third most frequent cancer worldwide in 2000, accounting for approximately 876 000 new cases or 9 per cent of the global cancer burden (Ferlay et al. 2001). High incidence areas, with rates above 25 per 100 000 in men and 15 per 100 000 in women, are found in central and eastern Europe, Portugal, eastern Asia, and parts of South America. The highest observed rates are found in Japan with an incidence rate in 1990 of 78 per 100 000 in men and 33 per 100 000 in women. Low-incidence areas include eastern and northern Africa, North America, and south and Southeast Asia. The rates are approximately twice as high among men as among women and are also two to three times higher among groups with low socio-economic status. Stomach cancer is very rare before the age of 30 years, after which an exponential increase is observed with increasing age.
It has been shown that migrants tend to maintain the high risk of their home country; their offspring tend to acquire a risk closer to their host country. However, the most striking feature of the epidemiology of stomach cancer is the dramatic decline in its incidence which has been observed in most countries over the past 60 years. The decline is apparent for both sexes, although it appears to have occurred earlier in countries which currently have a low risk. This continuous dramatic decline, as well as the results from migrant studies, suggest a strong environmental influence on the disease which is generally believed to be dietary.
Diets high in fruits and vegetables have been consistently shown to have a protective effect against stomach cancer. This evidence appears to be particularly consistent for raw vegetables, citrus fruits, and possibly allium vegetables (onions, leeks, garlic, and so on), with decreased risks of approximately 50 per cent generally being reported. For salted and pickled vegetables, some increased risks have been observed, although the evidence is not conclusive.
Several studies have also examined specific dietary vitamins and minerals. Results from two cohort studies and six of eight case–control studies have shown that consumption of foods with high levels of carotenoid intake are associated with a decreased risk of stomach cancer. Four studies of serum carotenoid levels have also shown a decreased risk with higher levels. Similar consistent results of a decreased risk have also been observed for high vitamin C intake in a number of studies. However, while these may represent real associations it is also possible that carotenoids and vitamin C are simply markers for some other dietary constituent. Results for other dietary vitamins, including vitamin E and retinol, indicate no association. Four intervention trials have also been conducted involving nutrient supplements and stomach cancer. In one of these trials, which was conducted in a Chinese population known to be micronutrient deficient, a combination supplement of b-carotene, vitamin E, and selenium did result in a small (16 per cent) reduced risk of stomach cancer (Blot 1997).
Regarding beverages, no evidence has been found that black tea, coffee, or alcohol influence the risk of stomach cancer. Interestingly, a number of studies of green tea in Japan and China have suggested a protective effect with high levels of consumption. Such an association is biologically plausible, as polyphenol extracts of tea, especially of green tea, are known to have an anticarcinogenic effect in animals.
Worldwide there is a strong and consistent correlation with stomach cancer incidence and high consumption of salt and salted foods. A large number of studies that have examined this relationship have generally found a moderately increased risk of approximately twofold for frequent consumption of salt and salted foods. The relationship is also biologically plausible given that salt may lead to damage of the protective mucosal layer of the stomach. Other methods of food preservation, including curing and smoking foods, have also been found to be weakly associated with stomach cancer, although the evidence is not consistent.
Salting as well as other methods of food preparation, including smoking, curing, and pickling, have become far less common with the advent of refrigeration, both in the home and for industrial storage purposes. The advent in refrigeration has dramatically changed dietary habits in many parts of the world by also ensuring that fresh fruits and vegetables are available for much of the year. While it is not therefore possible to know which dietary agent or agents, if any, are directly responsible for the large decrease in stomach cancer incidence over the last 60 years, it is possible that this has been brought about by a fortuitous side-effect of the invention of refrigeration.
There is now strong evidence that some cases of stomach cancer are caused by infection with the Helicobacter pylori bacterium. The current evidence from at least 10 prospective studies indicates that the increased risk of gastric cancer associated with H. pylori is between two- and threefold. The biological plausibility of a causal association is also supported by a strong association between H. pylori and precancerous lesions, including chronic and atrophic gastritis and dysplasia. Given that the prevalence of infection is very high, especially in developing counties and among older cohorts, it is possible that H. pylori could explain over 50 per cent of all new cases of gastric cancer that occur, or over 5 per cent of all cancer cases globally. There are, however, still some uncertainties regarding this association. H. pylori is strongly associated with low socio-economic status, and it is unclear whether some of the observed increased risk of H. pylori may be due to other correlated factors such as poor diet. The extent to which different strains of H. pylori, for example those containing the cagA gene, have different carcinogenic potential is also unclear.
Another important cause of stomach cancer may be tobacco smoking. A combined analysis of 40 studies indicates that smokers may have a 50 to 60 per cent increased risk of stomach cancer, as compared with non-smokers. If this relationship is a truly causal one it would indicate that smoking is responsible for approximately 10 per cent of all cases, or 80 000 cases of stomach cancer annually worldwide.
Primary prevention of stomach cancer by dietary means is feasible by encouraging high-risk populations to increase consumption of fresh fruit and vegetables and decrease consumption of cured meats and salt preserved foods. Prevention may also be feasible through eradication of H. pylori infection. H. pylori vaccines, which protect against infection but also induce regression of current infection and associated lesions, have been developed in animal models, although a safe and effective vaccine for human populations requires further research. The sequencing of the genome for H. pylori is likely to speed up this process. Screening and early detection of stomach cancer have been developed in Japan with the use of X-ray photofluorography to identify possible early lesions, followed by gastroscopy. Screen-detected cases are more likely to be early stage localized disease and are likely to have a greater survival than other cases. The use of gastroscopy for screening and early detection may not, however, be cost-effective in populations outside high-incidence areas.
Cancer of the intestine
Cancer of the intestine is an important human neoplasm, in particular in developed countries. Most cancers of the intestine occur in the large intestine, while cancer of the small intestine is rare in most populations. Of colorectal cancers, approximately two-thirds originate from the colon and one-third from the rectum and the rectosigmoid junction. More than half of the cancers of the colon are located in the sigma and the caecum. Most cancers of the intestine are of adenocarcinoma type, i.e. originate from the glandular cells. Other histological types include carcinoids, sarcomas, and lymphomas.
When taken together, cancers of the colon and rectum accounted in 2000 for an estimated 940 000 new cases and 490 000 deaths worldwide (Ferlay et al. 2001). They represent the fourth most frequent malignant disease in terms of incidence and the third for mortality.
Small intestine (Schottenfeld and Islam 1996)
Age-standardized incidence rates of small intestinal cancer are in most populations below one case per 100 000 people in both genders. The highest rates (of the order of 2 per 100 000) are registered among black Americans and the Maori of New Zealand. The neoplasm is more common in men than in women, with a ratio of the order of 1.5 to 3. Its occurrence is correlated with the incidence of colon cancer but not stomach cancer.
Adenocarcinomas account for approximately 50 per cent of neoplasms of the small intestine. They originate mainly in the duodenum and proximal jejunum and are preceded by formation of adenoma. Various hereditary syndromes, such as familial adenomatous polyposis and Peutz–Jeghers syndrome, are characterized by multiple hamartomatous adenomas of the small intestine and, to a less extent, of the colon: these patients carry an increased risk of adenocarcinoma of the small intestine. Similarly, patients with Crohn’s disease have a 10-fold increased risk of small intestine adenocarcinoma.
Malignant lymphomas represent about one-quarter of neoplasms of the small intestine: they are mainly of diffuse histiocytic type. Patients with AIDS and coeliac sprue are at increased risk of small cell lymphomas. Carcinoid tumours, which originate from the enteroendocrine (argentaffin) cells, are another important histological type. Limited data are available on the risk factors for this type of neoplasm.
The evidence of a role for environmental factors, such as tobacco smoking and diet, in the genesis of small intestinal neoplasms is at present inconclusive.
Colon (Schottenfeld and Winawer 1996)
The highest rates of colon cancer (around or above 30 per 100 000 in men and 25 per 100 000 in women) are recorded in Oceania (the islands of the Pacific Ocean and neighbouring seas), the United States (in particular among black people), and western Europe. Rates in developing countries are lower (5 to 15 per 100 000). The disease is rare before the age of 45 years, after which the incidence increases exponentially with age. In most populations, rates are higher in men than in women, with a ratio of the order of 1.5; however, given the predominance of women at older ages, the number of cases is similar in the two genders. A small increase in the incidence of colon cancer has been observed during the last decades in most populations, but not in North America, where rates have been stable.
Studies of migrant populations have repeatedly shown that the risk of colon cancer approaches that of the country of adoption within 20 years of residence; the incidence is higher in urban than in rural populations.
The predominant histological type of malignant neoplasms of the colon is the adenocarcinoma. This neoplasm is usually preceded by a polyp, or adenoma, less frequently by a small area of flat mucosa exhibiting various grades of dysplasia. The malignant potential of an adenoma is increased by a surface diameter greater than 1 cm, by villous (rather than tubular) organization, and by severe cellular dysplasia. Carriers of one adenoma larger than 1 cm have a two to four times increased risk of developing colon cancer; this risk is further doubled in carriers of multiple adenomas. On a geographical basis, the prevalence of adenomas detected during colonoscopy closely parallels the incidence of colon cancer.
Many investigations have addressed the possible role of nutrition in colon carcinogenesis (WCRF 1997). Different aspects of diet have been subject to special scrutiny as risk factors for colon cancer: (a) increased total energy intake; (b) a diet rich in animal fat, in particular in meat; (c) high intake of heterocyclic amines and other mutagens formed during the cooking of meat and fish; (d) low intake of vegetable fibres; (e) a diet poor in fresh fruits and vegetables; (f) low intake of antioxidant micronutrients, such as vitamin E, vitamin C, carotenoids, selenium, and zinc. While there is some evidence in favour of a role of each of these aspects of diet in increasing the risk of colon cancer, the interpretation of the data is complicated by their correlation and by the limitations of assessment of dietary factors in observational epidemiology. Two mechanisms have been postulated for the possible carcinogenic role of a diet high in fat and low in fibre: enhancement of conversion of bile acids into mutagens and tumour-promoting agents by the colonic bacterial flora, and formation of oxygen radicals and other lipid peroxidation products (WCRF 1997). While the mechanisms are unknown at present, the strongest evidence indicates a protective effect of a diet low in calories and animal foods, and rich in fruit and vegetables. Following the results of experimental studies, calcium and vitamin D have received attention as potential chemopreventive agents: the available evidence, however, does not consistently suggest a protective effect of dietary or supplemental intake of either agent.
There is growing evidence of an effect of heavy physical activity, both at the workplace and during leisure time, in protecting from colon cancer. This effect seems to be independent of that of diet, although the exact role of different components of energy balance (i.e. excessive intake from diet and insufficient expenditure through physical activity), as well as of adiposity and body composition, is unknown at present.
Several studies have associated tobacco smoking with an increased risk of colonic adenoma. While the studies of colon cancer have not consistently demonstrated a positive association, a modest increased risk following prolonged heavy smoking has been shown in some of the largest prospective studies.
Several large observational studies have provided evidence that regular aspirin intake reduces by 40 to 60 per cent the risk of colorectal cancer as compared with non-use. However, bias and confounding cannot be completely excluded from these studies (IARC 1997b).
Patients with ulcerative colitis and Crohn’s disease are at increased risk of colon cancer. The overall relative risk has been estimated in the range of 4 to 20, and it is higher for young age at diagnosis, severity of the disease, and presence of dysplasia. The contribution of shared genetic and environmental factors in the genesis of the two inflammatory conditions and of colon cancer is not known. Cholecystectomy has been associated with a moderate (1.5- to twofold) increased risk of right-sided colon cancer, possibly due to continuous secretion of bile. Patients with one cancer of the colon have a double risk to develop a second primary tumour in the colon or rectum, and the risk is greater for early age at first diagnosis. In women, an association has been shown also with cancers of the endometrium, ovary, and breast, possibly due to shared hormonal or dietary factors.
There are several rare hereditary conditions that are characterized by a very high incidence of colon cancer. Familial adenomatous polyposis, due to inherited or de novo mutation in the adenomatous polyposis colon gene on chromosome 5, is characterized by a very high number of colonic adenomas and a cumulative incidence of colon or rectal cancer close to 100 per cent by the age of 55 years. Other, more rare, diseases characterized by colonic polyposis, among other features, are Gardner’s syndrome, Turcot’s syndrome, and juvenile polyposis. All these hereditary conditions, although very serious for the affected patients, account for no more than 1 per cent of colon cancers in the general population.
Two syndromes characterized by hereditary non-polyposis colon cancer, i.e. with increased familial risk of colon cancer in the absence of adenomas, have been described. Lynch syndrome type I is characterized by increased risk of cancer of the proximal (right) colon, and is due to inherited mutation in one of two genes involved in DNA repair. Patients of Lynch syndrome type II also have an increased risk of extracolonic neoplasms, mainly of the endometrium and breast. As a whole, hereditary non-polyposis colon cancer may account for a sizeable proportion of cases of colon cancer in Western populations. In addition to these hereditary conditions, first-degree relatives of colon cancer patients have a two- to threefold increased risk of developing a cancer of the colon or rectum. It is unclear whether the underlying mechanisms are unknown genes at high or moderate penetrance or shared environmental factors.
Rectum (Schottenfeld and Winawer 1996)
The distribution of cancer of the rectum, including the rectosigmoid junction and the anus, parallels the distribution of colon cancer: the highest rates are recorded in Oceania, North America, and central Europe and are of the order of 20 per 100 000 in men and 10 per 100 000 in women. In most populations, incidence rates have been stable in recent decades. The male to female ratio is close to 2.
Most biological and epidemiological features of rectal cancer resemble those described for colon cancer, including the preneoplastic role of adenomas and non-polypoid dysplastic mucosa, the presence of familial syndromes, the increased risk among patients with chronic inflammatory bowel diseases, and the likely protective role of dietary factors and physical activity. In addition, several studies have provided evidence, although not fully consistent, of an association between elevated intake of alcohol, of beer in particular, and increased risk of colorectal adenoma and adenocarcinoma.
Anus (Schottenfeld and Winawer 1996)
The incidence of cancer of the anus is, in most populations, between 0.5 and 1 per 100 000. This neoplasm may occur either in the anal canal or in the perianal region; the predominant histological types are squamous cell carcinoma and transitional cell carcinoma. Women have a higher incidence of cancer of the anus, in particular of the canal, than men.
The epidemiological features of the disease resemble those of sexually transmitted diseases, with an increased incidence among never-married people, men with homosexual preference, and people with an increased number of sexual partners. The disease has also been associated with previous or subsequent history of cancers of the uterine cervix, vagina, vulva, and penis.
Chronic infection with human papilloma virus, in particular types 16 and 18, is the main known risk factor for anal squamous cell carcinoma. Human papilloma virus DNA has been detected both in a high proportion of carcinomas and anal intraepithelial neoplasias, the precursor lesions of invasive cancer.
Other probable risk factors that are likely to interact with human papilloma virus infection are infection with herpes simplex virus, history of anogenital condylomata genital warts, chronic inflammatory conditions such as fistulas, and immunosuppression, including infection with HIV.
Increased intake of vegetables and reduced intake of total calories, fat, and meat high in heterocyclic amines, together with increased physical activity, are reasonable suggestions for the primary prevention of colorectal cancer. Control of human papilloma virus infection, possibly via vaccination, would represent the main preventive measure for anal cancer. Chemopreventive strategies cannot be recommended at present.
Surveillance via flexible sigmoidoscopy, involving removal of adenomas, is a recommended secondary preventive measure for people at increased familial risk. An additional approach consists in the detection of occult blood in the faeces. The method suffers from low specificity and, to a lesser extent, low sensitivity, in particular in the ability to detect adenomas. However, trials have shown a reduced mortality from colorectal cancer after annual testing, although this is achieved at a high cost due to an elevated number of false-positive cases. Current recommendations for individuals aged 50 years and over include either annual faecal occult blood testing or flexible sigmoidoscopy every 5 years (Winawer et al. 1997).
Cancer of the liver and biliary tract (London and McGlynn 1996)
The descriptive epidemiology of liver cancer is complicated by the large number of secondary tumours of the liver, which are difficult to separate from primary cancers without histological verification. The most common histological type of liver malignant neoplasm is hepatocellular carcinoma; other forms includ childhood tumour hepatoblastoma, and adult tumour cholangiocarcinoma (originating from the intrahepatic biliary ducts) and angiosarcoma (from the intrahepatic blood vessels). Cancers of the extrahepatic biliary ducts are of the adenocarcinoma type. Most liver cancers originate from cirrhotic tissue.
The incidence of liver cancer is high in all developing regions of the world, with the exception of northern Africa and western Asia (Fig. 7). The highest rates (above 40 per 100 000 in men and above 10 per 100 000 in women) are recorded in Thailand, Japan, and certain parts of China. In most industrialized countries, age-standardized rates are below 5 per 100 000 in men and 2.5 per 100 000 in women. Intermediate rates (5 to 10 per 100 000 in men) are observed in areas of southern and central Europe. Rates are two- to threefold higher in men than women, and the difference is stronger in high-incidence than in low-incidence areas. While in developed countries the rates increase linearly with age, in high-risk areas of developing countries a plateau is reached during the fourth decade of age, and rates decrease after the age of 65 years (it is unclear whether this pattern is due to underdiagnosis at older ages). The estimated worldwide number of new cases of liver cancer in 2000 is 560 000, of which more than 80 per cent are from developing countries (54 per cent from China alone) (Ferlay et al. 2001). Given the poor survival from this disease, the estimated number of deaths is similar to that of new cases (550 000); liver cancer is the third most frequent cause of neoplastic death in developing countries.
Fig. 7 Incidence of liver cancer: age-standardized rates (worldwide) for men (all ages). (Source: Ferlay et al. 2001.)
Incidence rates of biliary tract cancer are high (above 3 per 100 000 in men and above 5 per 100 000 in women) in central Europe, South America, Japan, and western Asia. In the United States, rates are higher among people of American-Indian, Hispanic, and Japanese origin than in other groups. Most of the geographical variation is accounted for by cancer of the gallbladder, which represents the majority of biliary tract cancers. Rates of gallbladder cancer in women are generally higher than in men, while other biliary tract cancers are slightly more frequent in men: the overall pattern is characterized by female predominance in the occurrence of this neoplasm.
Chronic infections with hepatitis B and hepatitis C virus are the main causes of hepatocellular carcinoma. The risk increases with early age at infection (in high-risk countries, most hepatitis B virus infections occur perinatally or in early childhood) and the presence of liver cirrhosis. Hepatitis B virus is the main agent in China, Southeast Asia, and Africa, while hepatitis C virus is the predominant virus in Japan and southern Europe. The most frequent routes of hepatitis C virus transmission are parenteral and sexual, perinatal infection being rare. The estimated risk of developing hepatocellular carcinoma among infected subjects, relative to uninfected, ranged between 5 and 50 in different studies. On a global scale, the fraction of liver cancer cases attributable to hepatitis B virus is 60 per cent, and attributable to hepatitis C virus is 24 per cent (Pisani et al. 1997).
Ecological studies have shown that the incidence of hepatocellular carcinoma correlates not only with hepatitis B and hepatitis C virus infection, but also with contamination of foodstuff with aflatoxins, a group of mycotoxins produced by the fungi Aspergillus flavus and Aspergillus parasiticus, which cause liver cancer in many species of experimental animals. Contamination originates mainly from improper storage of cereals, peanuts, and other vegetables and is prevalent in particular in Africa, Southeast Asia, and China. The investigation of the carcinogenic role of aflatoxins in humans has been complicated by the inadequacy of traditional methods of exposure assessment (e.g. questionnaires). During the last decade, however, prospective studies have shown a strong association between biological markers of aflatoxin exposure in serum or urine and risk of subsequent liver cancer. A carcinogenic role of aflatoxins, in particular of aflatoxin B1, has therefore been confirmed and shown to be independent of that exerted by hepatitis B virus infection.
Alcohol intake increases the risk of hepatocellular carcinoma. The most likely mechanism is through development of liver cirrhosis, although alternative mechanisms such as alteration in activation and detoxification of carcinogens may also play a role. Alcoholic liver cirrhosis is probably the most important risk factor for hepatocellular carcinoma in populations with low prevalence of hepatitis B and hepatitis C virus infection and low exposure to aflatoxins, such as North America and northern Europe. The association between tobacco smoking and hepatocellular carcinoma is not fully consistent, and a possible confounding effect of excessive alcohol intake or infection with hepatitis B and hepatitis C virus has not been completely ruled out. Despite these limitations, the evidence indicates a possible limited effect of tobacco smoking on liver carcinogenesis.
Use of oral contraceptives greatly increases the risk of liver adenomas, and an effect on the risk of hepatocellular carcinoma is plausible, although its magnitude is likely to be small. Case reports have associated the use of anabolic steroids with the development of liver cancer, but the evidence is not conclusive at present. An increase in iron storage in the body is a likely cause of hepatocellular carcinoma: the evidence comes from studies of patients with haemochromatosis or other disorders of iron metabolism. The effect of iron overload seems to be independent of development of cirrhosis and may interact with hepatitis B virus infection.
Other types of primary liver cancer
Infestation with the liver flukes, Opisthorchis viverrini and Clonorchis sinensis, is the main known cause of this form of cholangiocarcinoma that is rare in most populations but relatively frequent in infested areas in Southeast Asia. Infection occurs via consumption of improperly cooked fish. Exposure to thorotrast, a contrast medium containing radioactive thorium used for angiography in Europe and Japan between 1930 and 1955, resulted in an increase of cholangiocarcinoma and of liver angiosarcoma. Workers exposed to vinyl chloride, a monomer used in the chemical industry for production of the plastic polymer, polyvinyl chloride, experience an increased risk of angiosarcoma. The identification of clusters of cases of liver angiosarcoma in these workers has led to a drastic reduction in exposure to vinyl chloride.
Cancer of the extrahepatic biliary ducts
The main known risk factor for cancer of the gallbladder is the presence of gallstones. The relative risk is of the order of 3, and it is higher in patients with large (more than 3 cm in diameter) rather than small (less than 1 cm) stones. In Western populations, most gallstones are formed by cholesterol, and their formation is associated with hypersecretion and saturation of cholesterol in the bile. The possible causes of cholesterol saturation (obesity, multiple pregnancies, and other hormonal factors) are also associated with increased risk of gallbladder cancer. An additional role of gallbladder hypomotility in stone formation is likely. In Asia, the main types of gallstone are formed by bilirubin salts and have as a risk factor bacterial infection of the biliary system: their association with gallbladder cancer, however, is not clear.
Other suspected risk factors for gallbladder cancer include chronic inflammation; biliary stasis and infection, in particular status of chronic typhoid and paratyphoid carrier; history of gastric resection; reproductive history resulting in increased exposure to endogenous oestrogens and progesterone (early age at menarche, late age at menopause, high parity); obesity; and, possibly, increased energy intake. It is likely that these factors act through gallstone formation, although the available data do not allow a conclusion with respect to their possible role in gallbladder carcinogenesis.
Fewer data are available on risk factors for cancer of extrahepatic biliary ducts. Infestation with the liver flukes causing intrahepatic cholangiocarcinoma, and history of ulcerative colitis are established risk factors but explain only a small proportion of these cancers. Tobacco smoking has been suggested as an additional cause.
Genetic factors are likely to play a role in biliary tract cancer. Patients with hereditary cancer-prone conditions, such as ataxia telangiectasia and hereditary non-polyposis colon cancer, have an increased risk of biliary tract cancer. Furthermore, a familial aggregation has been shown in population-based studies, which might be due to familial predisposition to gallstone formation.
The strong role in liver carcinogenesis of infection with hepatitis B virus, a virus for which effective and relatively cheap vaccines are available, suggests that prevention of liver cancer is achievable. In high-prevalence areas, hepatitis B virus vaccination has to be introduced in the perinatal period. In the last decades, many countries from Asia, southern Europe, and, to a lesser extent, Africa have expanded the national childhood vaccination programme to include hepatitis B virus. Large-scale studies are on-going to assess the efficacy of such programmes in reducing the incidence of liver cancer. A similar primary preventive approach is not available for hepatitis C virus; medical treatment of carriers with interferon might represent an alternative approach (which is also available for hepatitis B virus carriers).
Control of aflatoxin contamination of foodstuffs (below 20 ppb) represents another important preventive measure. While this is easily achieved in industrialized countries, its implementation is limited by economic and logistic factors in many high-prevalence regions. Control of tobacco smoking and alcohol drinking might represent additional primary preventive measures, although their impact on liver cancer is likely to be of less importance than for other cancers.
Since about half of hepatocellular carcinomas, but not normal adult liver, secrete the fetal antigen a-fetoprotein, the detection of this marker has been proposed as a screening method. However, no population-based studies are currently available showing a decreased mortality from liver cancer in screened populations.
Cholecystectomy is an obvious means of preventing gallbladder cancer. The removal of the gallbladder in asymptomatic patients, however, is not justified, with the possible exception of high-risk circumstances such as large stones and calcified gallbladder. The increased rate of cholecystectomy in many industrialized countries is probably responsible for the temporal decreasing trend of gallbladder cancer.
Cancer of the pancreas (Anderson et al. 1996)
The descriptive epidemiology of pancreatic cancer suffers from geographical and temporal variation in the sensitivity and specificity of clinical diagnosis and in the proportion of histological verification of cases. Even when comparing populations living in the same place at the same time (e.g. different social classes or age groups), differential access to health care might bias the incidence or mortality data.
The great majority of malignant neoplasms of the pancreas originate from the exocrine portion of the organ, which mainly secretes digestive enzymes. They are of the adenocarcinoma type. Rare pancreatic neoplasms include tumours (of uncertain clinical behaviour) of the endocrine portion, which secretes insulin and glucagon, as well as lymphomas and sarcomas.
The highest rates are recorded among black Americans (about 12 per 100 000 in men and 10 per 100 000 in women) and among indigenous populations in Oceania. The lowest rates, which may suffer from underdiagnosis, are recorded in India, Southeast Asia, and northern and central Africa (below 2 per 100 000 in men and 1 per 100 000 in women). In the United States, rates are about 50 to 100 per cent higher in black people as compared with white people living in the same areas. The disease accounts for an estimated 216 000 new cases in 2000, 60 per cent of which occurred in developed countries (Ferlay et al. 2001). Given the very poor survival from this disease, mortality rates closely parallel incidence rates. Rates are about 50 per cent higher in men than in women. An increase in incidence and mortality has taken place since the 1970s, in particular in Europe, that can be partly attributed to diagnostic improvements.
The incidence of pancreatic cancer is very low before the age of 40 years and increases logarithmically, without a clear suggestion of a plateau at old age. Urban populations have been reported to have higher rates than rural populations, but this may reflect differences in quality of diagnosis. Migrant studies suggest that first-generation migrants from low- to high-risk areas experience, after 15 or 20 years, rates that are even higher than those of the country of migration, suggesting an important role of environmental exposures occurring late in life.
The best known risk factor for pancreatic cancer is tobacco smoking. The risk in smokers is two- to fivefold higher than that in non-smokers, and a dose–response relationship and a protective effect of quitting smoking have been shown in many populations. The proportion of cases of pancreatic cancer attributable to tobacco smoking has been estimated to be 27 per cent in men and 11 per cent in women (Parkin et al. 1994). It is noteworthy that some of the features of the descriptive epidemiology of pancreatic cancer (i.e. a high incidence among black Americans as compared with a low incidence in Africa, and a higher risk among men and urban residents) can simply be explained by differences in smoking habits.
An increased risk of pancreatic cancer has been suggested for low vegetable and fruit intake (WCRF 1997). For other dietary components, namely cereals and sugar, the evidence is not consistent. Early reports of an association between coffee consumption and pancreatic cancer risk have not been confirmed by larger, more recent investigations. A positive association between alcohol drinking and pancreatic cancer has been reported in some, but not all, studies that have addressed this question. The current evidence is consistent with a possible weak effect of heavy alcohol drinking, in particular among smokers.
Several medical conditions have been studied with respect to their association with subsequent risk of pancreatic cancer. History of pancreatitis increases the risk more than 10-fold, with little difference between the alcoholic and non-alcoholic forms of the disease. An increased risk has also been shown in several studies of diabetic patients; the relative risk is likely to fall in the range of 1.5 to 2. Gastrectomy patients are at three- to fivefold increased risk of pancreatic cancer; the association does not appear to be confounded by tobacco smoking.
A familial history of cancer of the pancreas is present is 8 to 10 per cent of patients, suggesting a possible role for genetic factors. Specific hereditary conditions carrying an increased risk of pancreatic cancer include the Li–Fraumeni syndrome, hereditary non-polyposis colon cancer, and a group of rare hereditary pancreatites that involve deficiency in enzyme metabolism: these conditions are likely to explain only a small proportion of cases in the general population.
There is no effective cure for pancreatic cancer, with the exception of surgery for a small number of patients. Screening methods are not available. Primary prevention is the only available tool for this disease: avoidance of smoking would have a very important effect in reducing the number of cases; an increase in dietary consumption of fruits and vegetables is the other recommended preventive measure.
Cancer of the respiratory tract
Malignant neoplasms are important in all organs of the respiratory tract: the nasal cavity and paranasal sinuses, nasopharynx, larynx, lung (including the trachea and the bronchi), and pleura. Cancer of the nasopharynx is discussed above.
Nasal cavity and paranasal sinuses (Roush 1996)
The incidence of cancer of the nasal cavity and paranasal sinuses (sinonasal cancer) is low in most populations (less than 1.5 per 100 000 in men and less than 1.0 per 100 000 in women). Higher rates are recorded in Japan and certain parts of China and India. In Western populations and Japan, squamous cell carcinoma is the main histological type, followed by adenocarcinoma, and 30 to 80 per cent of the cancers originate from the maxillary sinus. In other countries, other forms of sinonasal cancer are important, such as Burkitt’s lymphoma of the maxillary or ethmoid sinus in eastern Africa. Time trends have shown in most populations a stable incidence or a small decline in recent decades.
Occupational exposure to wood dust, in particular to dust of hard woods such as beech and oak, is the main known risk factor for sinonasal cancer: the increase in risk (of the order of five- to 50-fold) is strongest for adenocarcinoma and for cancers originating from the sinuses. It is unclear whether dust from soft woods such as pine and spruce increases the risk. The effect is present after 40 or more years since first exposure and persists after cessation of exposure. An increased risk of sinonasal cancer has been shown among workers in nickel refining and chromate pigment manufacture, but not among workers exposed to these metals in other processes, such as plating and welding. Among other suspected occupational carcinogens are formaldehyde, di-isopropyl sulphate, and dichloroethyl sulphide.
A relatively weak association (relative risks in the range of 2 to 5) has been shown between tobacco smoking and sinonasal cancer, in particular squamous cell carcinoma. Presence of recurrent polyps represents a risk factor, and a possible role of chronic sinusitis has been suggested. Exposure to thorotrast, a radioactive contrast agent, represents an additional risk factor. Infection with Epstein–Barr virus is implicated in the origin of Burkitt’s lymphoma of the sinuses and possibly in the origin of undifferentiated carcinoma.
Control of occupational exposure to hard wood dust and nickel and chromium compounds, as well as avoidance of tobacco smoking, are the two main preventive measures for sinonasal cancer.
Larynx (Austin and Reynolds 1996)
More than 90 per cent of cancers of the larynx are squamous cell carcinomas, and the majority originate from the supraglottic and glottic regions of the organs. The incidence in men is high (10 per 100 000 or more) in southern and central Europe, southern Brazil, Uruguay, and Argentina and among black Americans, while the lowest rates (less than 1 per 100 000) are recorded in Southeast Asia and central Africa. The incidence in women in below 1 per 100 000 in most populations. Rates have not changed markedly during the last two decades. An estimated 140 000 new cases occurred worldwide in 2000, of which 140 000 were among men (Ferlay et al. 2001).
Most cases of laryngeal cancer in Western countries (up to 80 per cent in high-risk areas) are attributable to tobacco smoking, alcohol drinking, and the interaction between these two factors. The effect of tobacco, with risks in smokers of the order of 10 relative to non-smokers, seems to be stronger for glottic than supraglottic neoplasms. Studies in several populations have shown a dose–response relationship and a beneficial effect of quitting smoking. Smoking black tobacco cigarettes entails a stronger risk than smoking blond tobacco cigarettes. Studies from India have also reported an effect of chewing tobacco-containing products. The effect of alcohol is stronger for supraglottic tumours than for tumours at other sites; however, it is not clear whether different alcoholic beverages exert a different carcinogenic effect.
A protective effect is probably exerted by high intake of fruits and vegetables, although the evidence regarding specific micronutrients, such as carotenoids and vitamin C, is inadequate to draw a conclusion. It has been suggested in studies from Brazil and Uruguay that maté drinking is a risk factor. Data concerning a possible effect of other foods are not consistent.
Occupational exposure to mists of strong inorganic acids, in particular of sulphuric acid, is an established risk factor for laryngeal cancer. A possible effect has been suggested for other occupational exposures, including nickel, asbestos, and ionizing radiation, but the evidence is not conclusive at present.
Laryngeal papillomatosis is a condition characterized by multiple benign tumours, called papillomas, that are caused by infection with human papilloma virus types 6 and 11, the same types causing genital condylomata acuminata. Infection in children occurs in both genders from vertical infection during delivery; infection in adults is prevalent in men and may occur via orogenital sexual contact. Papillomatosis patients have an increased risk of laryngeal cancer; however, studies aimed at assessing the presence of human papilloma virus DNA have not yet provided conclusive evidence of a higher prevalence of infection in cases of laryngeal cancer than in controls. Another virus for which a causal role has been postulated in laryngeal cancer is herpes simplex virus type 1.
There is no evidence of strong genetic factors in laryngeal carcinogenesis; however, polymorphism for enzymes implicated in the metabolism of alcohol and tobacco, such as glutathione-5-transferase and alcohol dehydrogenase, are likely to represent weak susceptibility factors, with relative risks of the order of 1.5 to 2.
Survival from laryngeal cancer is relatively common (5-year survival rates are of the order of 65 per cent in developed countries and 40 per cent in developing countries) (Sankaranarayanan et al. 1998; Berrino et al. 1999). These patients are at very high risk of developing a second primary tumour in the oral cavity, pharynx, and lung. While shared risk factors are likely to play an important role, it is plausible that host susceptibility factors are also partially responsible. For example, an increased sensitivity to mutagens has been shown in lymphocytes of laryngeal cancer patients, in particular those with multiple tumours, as compared with controls (Spitz et al. 1997).
Control of tobacco smoking and excessive alcohol drinking, possibly together with an increased intake of fruits and vegetables, would prevent the majority of cases of laryngeal cancer (as well as of a number of other cancers) in most populations. Control of exposure to known and suspected occupational carcinogens is an important measure for exposed workers. No screening methods are currently available for laryngeal cancer.
Lung (Blot and Fraumeni 1996)
Lung cancer was a rare disease until the beginning of the twentieth century. Since then, its occurrence has increased rapidly and this neoplasm has become the most frequent malignant neoplasm among men in most countries, and represents the most important cause of neoplastic death worldwide. It accounts for an estimated 901 000 new cases each year among men and 337 000 among women (Ferlay et al. 2001). Survival from lung cancer is poor (5 to 10 per cent at 5 years).
In both men and women, the incidence of lung cancer is low before the age of 40 years, and increases up to the age of 70 or 75. The geographical and temporal patterns of lung cancer incidence are to a large extent determined by consumption of tobacco. An increase in tobacco consumption is paralleled some 20 years later by an increase in the incidence of lung cancer, and a decrease in consumption is followed by a decrease in incidence.
The highest incidence rates (more than 100 per 100 000) are recorded among black people from New Orleans, the United States, and Maori from New Zealand (Parkin et al. 1997). In some central and eastern European countries, very high mortality rates, with further increasing temporal trends, are reported. On the contrary, mortality rates have started to decline among men from the United States and northern Europe. The lowest incidence rates are reported from Africa and southern Asia (Parkin et al. 1997).
Rates in women are high in the United States, Canada, Denmark, and the United Kingdom, and low in countries such as France, Japan, and Spain, in which the prevalence of smoking in women increased only recently. The lowest rates (less than 3 cases per 100 000 people) are recorded in Africa and India. China is a notable exception, with relatively high rates recorded among women (e.g. 37 per 100 000 in Tianjin during 1988 to 1992; Parkin et al. 1997), despite a low prevalence of smoking.
The main histological types of lung cancer are squamous cell carcinoma, small cell carcinoma, adenocarcinoma, and large cell carcinoma. Over the last 20 years, the proportion of squamous cell carcinomas, which used to be the predominant type, has decreased and an increase of adenocarcinomas has taken place in both genders. Despite some minor differences, the main risk factors for lung cancer affect all histological types.
A carcinogenic effect of tobacco smoke on the lung has been demonstrated in the 1950s and has been recognized by public health and regulatory authorities since the mid-1960s (USDHHS 1982). The risk of lung cancer among smokers relative to the risk among never-smokers is of the order of 8 to 15 in men and 2 to 10 in women. This overall risk reflects the contribution of the different aspects of tobacco smoking: average consumption, duration of smoking, time since quitting, age at start, type of tobacco product, and inhalation pattern, with duration being the dominant factor. As compared with continuous smokers, the excess risk decreases in ex-smokers after approximately 5 years since quitting, but a small excess risk is likely to persist in long-term quitters throughout life. The risk of lung cancer is lower among smokers of low-tar and low-nicotine cigarettes than among other smokers, and in non-inhalers as compared with inhalers. A protective effect has also been observed among long-term smokers of filtered cigarettes compared with smokers of unfiltered cigarettes. Smokers of black (air-cured) tobacco cigarettes are at a two- to threefold higher risk of lung cancer than smokers of blond (flue-cured) tobacco cigarettes. A causal association with lung cancer has been shown also for consumption of cigars, cigarillos, pipe, bidis, and water pipe.
An association has been shown in many studies between exposure to passive smoke and lung cancer risk in non-smokers. The magnitude of the risk among non-smokers exposed to passive smoke is of the order 15 to 20 per cent (Hackshaw et al. 1997).
There is convincing evidence that a diet rich in vegetables and fruits exerts a protective effect against lung cancer (WCRF 1997). Subjects in the categories at highest consumption experience 10 to 50 per cent of the risk of lung cancer of subjects in the categories at lowest consumption. Despite the many risk estimates for intake of other foods, such as cereals, pulses, meat, eggs, milk, and dairy products, the evidence is inadequate to allow a judgement regarding the evidence of a carcinogenic or a protective effect (WCRF 1997).
Several studies have suggested positive associations between total and saturated fat intake and lung cancer risk, and the effect seems to be independent of that of tobacco consumption (WCRF 1997). A large number of studies have reported a reduced risk of lung cancer for high intake of b-carotene (IARC 1998a). Similar results have been obtained in studies based on measurement of b-carotene in prospectively collected sera (WCRF 1997). This evidence of a protective effect has been challenged by the results of intervention trials of b-carotene supplementation (IARC 1998a). In two of the studies, which included smokers and workers exposed to asbestos, an increase in the incidence of lung cancer was observed in the treated groups: in the other studies, no difference was found between the treated and the control groups. The difference in the results of observational studies and preventive trials can be explained either by a confounding effect by other dietary components in observational studies, or by a paradox effect of b-carotene at very high, non-physiological doses, in particular among smokers.
There is a suggestion from observational studies that high intake of vitamin C, vitamin E, and selenium protects against lung cancer. The evidence regarding other micronutrients is inconclusive at present (WCRF 1997). The available data suggest a small increase in the risk of lung cancer from alcohol drinking, which does not appear to be fully explained by tobacco smoking (WCRF 1997). Furthermore, there is quite consistent evidence that reduced body mass index (usually below 22) is associated with an increased risk of lung cancer; and a certain degree of protection from regular physical activity has been shown in most relevant studies, after adjustment for tobacco smoking (WCRF 1997).
A positive familial history of lung cancer has been found to be a risk factor in several studies. Segregation analyses suggest that inheritance of a major gene, in conjunction with tobacco smoking, might account for 50 to 80 per cent of cases diagnosed below the age of 60 years (Yang et al. 1999). In addition, low-penetrance genes involved in the metabolism of tobacco carcinogens might influence individual susceptibility to lung cancer.
Studies of atomic bomb survivors and patients treated with radiotherapy indicate that they are at increased risk of lung cancer (National Research Council 1988). Although the magnitude of the increased risk is moderate (relative risk of the order of 1.5 to 2 for cumulative exposure in excess of 1 Gy), the number of extra cases of lung cancer exceeds that of other neoplasms. Underground miners exposed to radio-active radon and its decay products have been found to be at increased risk of lung cancer. The main concern for lung cancer risk from radon and its decay products, however, comes from residential rather than occupational exposure. A meta-analysis of the eight most informative studies of lung cancer risk from residential exposure resulted in a pooled relative risk of 1.14 (95 per cent confidence interval 1.0 to 1.3) at 150 Bq/m3 of exposure (Lubin and Boice 1997).
The risk of lung cancer is increased among workers employed in several industries and occupations. For several of these high-risk workplaces, the agent (or agents) responsible for the increased risk have been identified. In other cases, the identification of the relevant carcinogens has not yet taken place (see section below on occupational and environmental agents for details). Occupational agents are responsible for an estimated 5 to 10 per cent of lung cancers in industrialized countries.
Patients with pulmonary tuberculosis are at increased risk of lung cancer; it is not clear whether the excess risk is due to the chronic inflammatory status of the lung parenchyma or to the specific action of the Mycobacterium. Chronic exposure to high levels of fibres and dusts might result in lung fibrosis (e.g. silicosis and asbestosis), a condition which entails a great increase in the risk of lung cancer. Chronic respiratory diseases (chronic bronchitis, emphysema, and asthma) have also been associated with lung cancer risk.
There is abundant evidence that lung cancer rates are higher in cities than in rural settings (Speizer and Samet 1994). Although this pattern might result from confounding by other factors, notably tobacco smoking, diet, and occupational exposures, the combined evidence from analytical studies suggests that urban air pollution might be a true risk factor for lung cancer; although the excess risk is unlikely to be larger than 50 per cent.
Two important sources of indoor exposure to potential lung carcinogens are the use of coal-burning heating devices without proper exhaust emission (e.g. the use of kang in northeastern China) and high-temperature cooking, in particular when using unrefined vegetable oils such as rapeseed oil (common in several parts of China). Indoor levels of benzo-(a)-pyrene have been reported to be very high in such exposure circumstances (Smith and Liu 1994). Indoor air pollution is probably a major cause of lung cancer in Chinese women, who experience very high lung cancer rates despite a low prevalence of smoking (Smith and Liu 1994).
Control of tobacco smoking remains the key strategy for the prevention of lung cancer. Reduction in exposure to occupational and environmental carcinogens (in particular indoor pollution and radon), as well as increase in consumption of fruits and vegetables are additional preventive opportunities. No screening approaches are effective to reduce lung cancer mortality.
Pleural mesothelioma (Blot and Fraumeni 1996)
Mesothelioma is the most important primary tumour of the pleura. It can also originate from the peritoneum and the pericardium. Mesotheliomas were considered very rare tumours, until large series of cases were reported in the 1960s among workers exposed to asbestos mining and manufacturing. The descriptive epidemiology of pleural tumours, and mesothelioma in particular, is complicated by geographical and temporal differences in diagnostic accuracy: although underdiagnosis is likely to occur when no information is available on exposure to asbestos, the opposite phenomenon may take place when the doctor or pathologist is aware of past asbestos exposure. In most industrialized countries, the incidence of pleural mesothelioma is of the order of 1 to 1.5 per 100 000 in men and around 0.5 per 100 000 in women. Lower rates are reported from developing countries, where underdiagnosis might be a particularly serious problem. In areas with a high prevalence of occupational exposure to asbestos such as shipbuilding and mining centres, the rates might be as high as 5 per 100 000 in men and 4 per 100 000 in women.
A steep increase in the mortality or incidence rates of mesothelioma has taken place in most industrialized countries since the 1950s. This trend is mainly due to a cohort effect, since individuals heavily exposed to asbestos (those born since 1900) have reached the age of development of mesothelioma. The highest exposures were experienced in the United States by those born around 1910, while in Europe exposure was high also in later cohorts. As a consequence, the increasing trend in mesothelioma occurrence is likely to disappear in the United States, while it will continue until about 2020 in most European countries.
Asbestos is the main known risk factor for mesothelioma. All types of asbestos cause mesothelioma, although the potency of chrysotile appears to be less than that of amphiboles. A distinct feature of the carcinogenic potency of asbestos on the pleura is the importance of time since first exposure: the rate appears to increase with the third power of time since first exposure, with few cases developing within the first 20 years and no apparent decline even after 45 or more years. Circumstances of high non-occupational exposure to asbestos (e.g. spouses of asbestos workers laundering work clothes) have been linked to an increased risk of mesothelioma. The data are inadequate to assess the risk of low-level environmental exposure, although the presence of a small excess risk is plausible.
Environmental exposure to erionite, a naturally occurring mineral fibre, has been linked to high mesothelioma risk in rural areas of Cappadocia, Turkey. The evidence of an effect of other mineral fibres (e.g. fibre glass, ceramic fibres) on the pleura is inadequate. Exposure to asbestos or other mineral fibres is not obvious in a proportion of cases of mesothelioma (40 to 50 per cent in men and 60 to 80 per cent in women): although some of these cases might be due to unrecognized occupational or environmental exposure to asbestos, it is likely that a fraction of cases not related to fibre exposure exists. A role of diet (increased risk among subjects with low fruit and vegetable intake) has been suggested in two case–control studies. Simian virus 40 has been found in a high proportion of tumour samples in various series of cases of mesothelioma, but its aetiological role is still unclear.
Control of occupational exposure to asbestos, in particular to amphiboles, is the main preventive action to reduce the number of mesothelioma cases. This has taken place in the mining and manufacturing industries in many countries: workers at highest risk are today those in the construction industry, mainly in demolition and maintenance. Control of asbestos exposure is deficient in many developing countries. Control of tobacco smoking is not effective to reduce mesothelioma risk; however, given the carcinogenic effect of asbestos also on the lung, and its interaction with tobacco smoking, smoking avoidance and cessation programmes should be incorporated in the surveillance of workers with past asbestos exposure.
Neoplasms of the bone and soft tissue (Miller et al. 1996)
Three main histological types represent the majority of cases of bone cancer: osteosarcoma (originating from the bone tissue and representing 30 to 50 per cent of all bone neoplasms), chondrosarcoma (from the cartilage, 20 to 30 per cent), and Ewing’s sarcoma (possibly from primitive nervous tissue, 10 to 20 per cent). Several other rare types (e.g. chordoma, fibrosarcoma, giant cell tumours) comprise some 10 to 20 per cent of all bone cancers. Ewing’s sarcoma occurs during the second and third decades of life; the incidence of osteosarcoma is bimodal with peaks between the age of 10 and 30 years, and after age 60, while the incidence of chondrosarcoma resembles that of many cancers with a steady increase throughout life. Rates of all major histological types are higher in men than in women, but in most populations the ratio is of the order of 1.5. There are limited geographical differences in bone cancer incidence, with rates ranging between 1 and 3 per 100 000 in men and between 0.5 and 2 per 100 000 in women.
Ionizing radiation is the best known risk factor for bone cancer. The risk is increased two- to threefold in groups of adult patients undergoing radiotherapy for cancer, as compared with other patients; the relative risk is high among radiotherapy-treated children, in particular among those with inherited mutation in the retinoblastoma gene, suggesting an interaction with genetic susceptibility. The excess risk, however, is apparent only among patients receiving high radiation doses, and studies of populations exposed to lower doses, such as patients treated for benign conditions or atomic bomb survivors, did not show an increased risk. Studies of workers and patients exposed to high doses of radium isotopes also confirmed the role of ionizing radiation in bone carcinogenesis.
Despite the strong relative risks, the number of cases of bone cancer due to ionizing radiation in the general population is small, but few data are available on other risk factors. A role of alkylating agents used in cancer chemotherapy seems likely, while the evidence implying viral infections, traumas, and medical implants is inadequate to reach a conclusion.
Patients with Paget’s disease (osteitis deformans, a disease characterized by multiple localized areas of destruction of bone followed by repair) have a high incidence of osteosarcoma and chondrosarcoma; this disease is likely to be responsible for a substantial proportion of cases above the age of 60 years. An hereditary component of Paget’s disease is suggested, and unknown environmental factors are likely to play a role in its aetiology. The risk of bone cancer is also increased in other rare syndromes of bone malformation. Bone cancer is one of the neoplasms appearing in the Li–Fraumeni syndrome, caused by an inherited mutation in the p53 gene. Familial aggregation of bone cancer occurs also outside well-defined inherited conditions, suggesting an important role of still unknown genetic factors.
The causes of Ewing’s sarcoma are largely unknown. It is not associated with ionizing radiation and it does not appear to aggregate in families.
Primary prevention of bone cancer is hampered by the limited knowledge about its causes and mechanisms. No secondary prevention strategies are available.
Soft-tissue sarcomas (Zahm et al. 1996)
Sarcomas originate from the mesenchyma in the space between the organs and within any organ. Sarcomas from the organs are classified within the neoplasms of that organ (e.g. liver angiosarcoma), while those originating from subcutaneous and intervisceral connective tissue are classified in the heterogeneous group of soft-tissue sarcomas. Fibrosarcoma (originating from fibrocytes of the connective tissue), leiomyosarcoma (from smooth muscle cells), and liposarcoma (from fat cells) are common types of soft-tissue sarcomas in most populations. The recent epidemic of AIDS has produced in several populations a sharp increase in the occurrence of Kaposi’s sarcoma, a sarcoma from blood vessel cells. Without considering Kaposi’s sarcoma (that is discussed below), the incidence of soft-tissue sarcoma presents limited geographical variations, with age-adjusted incidence rates ranging from 1 to 3 per 100 000 in men and from 0.5 to 2 to 100 000 in women. Rates are slightly lower in Asia, and higher in western Europe and North America. An increasing trend in incidence is suggested in industrialized countries, which is not explained by the increasing incidence of Kaposi’s sarcoma.
Ionizing radiation is a known risk factor for soft-tissue sarcomas: all types of sarcomas have been reported among patients treated with radiotherapy, the most frequent being malignant fibrous hystiocytoma. In addition, cancer chemotherapy seems to exert an effect independent of that of radiotherapy. A viral agent (e.g. cytomegalovirus, Epstein–Barr virus) has been hypothesized for non-AIDS related Kaposi’s sarcoma and for other soft-tissue sarcomas, but no conclusions can be reached at present.
Several occupational exposures have been linked to an increased risk of soft-tissue sarcoma, including in particular phenoxy herbicides and chlorinated hydrocarbon insecticides. For none of these chemicals the evidence of a carcinogenic effect is consistent, with the possible exception of 2,3,7,8-tetrachlorodibenzo-p-dioxin, for which a two- to threefold relative risk has been reported in studies of manufacturers and sprayers of ontaminated herbicides and of populations exposed during industrial accidents, but not among Vietnam veterans who were also exposed to contaminated herbicides. There is no consistent evidence of an association with dietary factors, traumas, tobacco smoking, and use of smokeless tobacco products.
Transplant patients receiving immunosuppressive therapy as well as patients with primary immunodeficiency syndromes have an increased risk of soft-tissue sarcomas. The occurrence of these neoplasms is also increased in patients with the Li–Fraumeni syndrome, neurofibromatosis type 1, and other rare familial cancer syndromes. Benign tumours in the connective tissue (lipomas, fibromas, and so on) are common and usually do not transform into their malignant counterparts.
Cancer of the skin
There are four main types of skin cancer: squamous cell carcinoma, arising from the epidermal cells; basal cell carcinoma, from basal cells forming sebaceous glands; melanoma, arising from melanocytes; and Kaposi’s sarcoma, arising from endothelial cells. Squamous cell carcinoma and basal cell carcinoma share pathological, clinical, and aetiological features, and are often combined under the definition of non-melanocytic skin cancer.
Non-melanocytic skin cancer (Scotto et al. 1996)
Given the simplified diagnostic and therapeutic procedures (often treated in outpatient clinics and general practitioner offices) of most non-melanocytic skin cancers, reporting of cases to registries is frequently incomplete, and many cancer registries do not attempt to provide incidence figures. The very good prognosis (a more than 95 per cent survival rate in most populations) makes mortality figures useless to estimate incidence. Population-based data incidence derive therefore from ad hoc surveys. A survey conducted in the United States in the late 1970s estimated an age-adjusted incidence rate of squamous cell carcinoma of 68 per 100 000 in white men and 24 per 100 000 in white women; corresponding figures for basal cell carcinoma were 258 and 155 per 100 000. Rates in black people were about 100 times lower than in white people, and squamous cell carcinoma predominates. The comparison with a similar survey conducted in the early 1970s revealed a 4 to 5 per cent increase in the incidence of basal cell carcinoma per year, which can be attributed, at least in part, to improved diagnostic and surveillance procedures. Squamous cell carcinoma rates increased little during the same period. Rates in white people approximate those of all other malignant neoplasms combined. Even higher rates have been recorded in Ireland and among white people living in countries with high solar exposure, such as Australia and South Africa, while black populations have consistently low rates.
The risk of non-melanocytic skin cancer is very low before the age of 20 years, and then increases logarithmically up to about 60 years of age, after which it decelerates. Between 75 and 90 per cent of both squamous and basal cell carcinomas in white people are localized on the face, head, and neck. In black people, the lower extremities are the most frequent location of squamous cell carcinoma.
Solar radiation is the main known risk factor for non-melanocytic skin cancer. For squamous cell carcinoma, the cumulative dose of ultraviolet radiation, disregarding dose rate, appears to be the predominant risk factor, while for basal cell carcinoma sun exposure and sunburning during childhood are the main determinants of subsequent risk. The effect of solar radiation has been shown following occupational, recreational, and involuntary exposure. The strongest association concerns ultraviolet radiation B, with a wavelength in the range 290 to 320 nm. A strong excess of skin cancer has also been shown in psoriasis patients treated with psoralen in combination with ultraviolet radiation A. Solar keratosis is a precursor lesion of squamous cell carcinoma of the skin (not of basal cell carcinoma); it occurs in those areas of the skin exposed to solar radiation. The cumulative progression rate of keratosis to carcinoma (usually through a phase of carcinoma in situ, or Bowen’s disease) is of the order of 5 per cent. Skin pigmentation is a modifying factor of the carcinogenic effect of ultraviolet radiation, with people with light pigmentation having the greatest risk.
An excess risk of non-melanocytic skin cancer has been shown following exposure to ionizing radiation (in studies of medical personnel, uranium miners, radiotherapy patients, and atomic bomb survivors): the shape of the dose–response curve appears to be linear without threshold. Exposure to arsenic and its inorganic compounds has been linked to an excess of skin cancer in people exposed occupationally, from drinking water or from drugs used in the past. Mixtures of polycyclic aromatic hydrocarbons (coal tar, tar pitch, soot, creosote, lubricating and cutting oils) are also carcinogenic to the skin: an excess of non-melanocytic cancer has been shown in classical occupational epidemiological studies among workers such as chimney sweeps, machine operators, and roofers.
Skin cancer occurs in Asian countries as a consequence of burn scars produced by traditional heating devices kept in close contact to the skin: kangri in Kashmir, India, kairo in some areas of Japan, and kang in northern China. It is possible that polycyclic aromatic hydrocarbons released by the burning material interact with heat in causing the cancer. Other types of ulcer and scar, not related to heat, also predispose to skin cancer: they include for example chronic skin inflammation and infection by biological agents such as tuberculosis and leprosy.
Immunodeficiency increases the risk of squamous cell carcinoma of the skin, as it has been shown in patients treated with immunosuppressive drugs following renal transplant or other conditions. Xeroderma pigmentosum and the naevoid basal cell carcinoma syndrome are rare hereditary conditions characterized by a very high incidence of skin cancer. In the former syndrome, the mechanism is a reduced capacity to repair damage to DNA. The action of immunodeficiency and genetic predisposition may be via an enhancement of the carcinogenic effect of ultraviolet and ionizing radiation, since the neoplasms occur on parts of the body exposed to the sun.
Avoidance of sun exposure, in particular during the middle of the day, is the primary preventive measure to reduce the incidence of skin cancer. There is no adequate evidence of a protective effect of sunscreens, possibly because use of sunscreens is associated with increased exposure to the sun. The possible benefit in reducing skin cancer risk, however, should be balanced against possible beneficial effects of ultraviolet radiation in promoting vitamin D metabolism. Control of occupational skin carcinogens has taken place in many industries, although high exposure circumstances may still take place in developing countries. Avoidance of drinking water with a high arsenic level should be a priority in contaminated areas. Secondary prevention can be achieved by regular skin examination, in particular for high-risk individuals: however, there is a lack of controlled trials on skin cancer screening.
Malignant melanoma (Armstrong and English 1996)
Malignant melanomas occur most frequently on the trunk in men and on the lower limbs in women. While pathologists distinguish several histological types of melanoma, these are likely to represent different stages of the same condition. A special type of melanoma, however, is the rare lentigo malignant melanoma, which occurs on the head and neck, in areas with sun damage.
An estimated 133 000 new cases of malignant melanoma occurred worldwide in 2000 (Ferlay et al. 2001). The incidence is highest (of the order of 25 per 100 000) in Australia; it ranges between 5 and 10 per 100 000 in other parts of Oceania, North America, and northern and western Europe, and is below 5 per 1 000 000 in the other regions of the world. In general, the incidence is low in dark-skinned populations. In many white populations, there has been a sharp increase in incidence (3 to 7 per cent per year) during the last decades; this pattern was not observed in non-white populations. Although part of this trend might be explained by increasing awareness and improved diagnosis, it is likely that it largely reflects a true phenomenon. Studies of migrants have shown an increased risk following migration to a country with a sunny climate, such as Israel, South Africa, and Australia. Age at migration is an important determinant of risk, since people who migrated before the age of 10 years retain a risk similar to the risk of the country of origin, while other migrants approach the risk of the host country. The incidence of melanoma is very low before the age of 20 years, and increases linearly with age thereafter, thus showing a higher incidence in young adulthood than many other tumours. The increase with age is steeper for melanomas arising on the face, while the incidence of those on other parts of the body declines after middle age. Overall, there is little difference in risk between men and women, but during reproductive ages the incidence is higher among women.
There is strong evidence of a carcinogenic role of ultraviolet radiation in determining malignant melanoma. Intermittent exposure to the sun seems to play a more important role than total cumulative exposure. However, genetic or other factors, including ‘ability to tan’ may modify the dose–response relationship between sun exposure and melanoma risk. According to this model, the risk in poor tanners would be mainly determined by cumulative exposure while intermittency would be the main factor for good tanners. In most available studies, the relative risk among subjects in the highest category of usual or recreational exposure to the sun or history of sunburns was 2 to 6, as compared with subjects in the lowest category.
The evidence of an increased risk of melanoma following exposure to fluorescent lamps and other artificial sources of ultraviolet radiation is inadequate.
Colour of hair and eyes, and skin complexion, have been found in many studies to be risk factors for melanoma. Colour of hair seems to be the main predictor of risk, with relative risks in the range of 1.5 to 2 for blond hair and 2 to 4 for red hair as compared with dark or brown hair. Freckling is likely to be an additional risk factor. Skin response to sun exposure and propensity to burn (or poor ability to tan) have also been associated with melanoma risk, with a relative risk mainly in the range of 1.5 to 4. However, pale complexion and propensity to burn are strongly correlated, and the available data are inadequate to separate these two factors completely.
Presence of a high number of naevi is the strongest known risk factor for melanoma. Assessment of the number and type of naevi is not straightforward, and misclassification is likely to affect studies on naevi and melanoma. The relative risk is of the order of 10 for the category at highest number of naevi. Naevi are likely precursors for melanoma. Their number depends on sun exposure, in particular intermittent exposure and sunburns: exposure in childhood is more important than exposure in adulthood. In subjects with familial melanoma, large atypical naevi, referred to as dysplastic naevi, might be found. Individuals with dysplastic naevi and familial melanoma have a very high risk of melanoma. In subjects without familial melanoma, presence of dysplastic naevi seems to be a risk factor independent of number of total naevi.
The number of atypical naevi and the risk of melanoma are increased among immunosuppressed patients. There is no clear evidence of a role of any other risk factor, including dietary factors and exogenous hormones, in the aetiology of melanoma. There is a two- to fivefold increased risk of melanoma in subjects with an affected relative, which seems independent of exposure to solar radiation. Several putative high-risk genes have been proposed to explain the increased familial risk.
Avoidance of solar exposure, especially in childhood, is the only primary preventive measure that can be recommended at present. Use of sunscreen does not protect against melanoma risk. Early diagnosis, in particular of thin lesions, is associated with better survival; screening via medical examination is justified in high-risk individuals, defined according to familial history, type of skin, and reaction to solar radiation.
This form of slowly progressing cutaneous sarcoma was rare before the appearance of the HIV epidemic. The incidence of non-HIV-related Kaposi’s sarcoma was highest in the Mediterranean basin and in Africa. No risk factors have been identified, although the lesion is more frequent in rural than in urban areas. Subjects (in particular of homosexual preference) infected with HIV have a high risk of Kaposi’s sarcoma. A virus, denominated Kaposi’s sarcoma herpes virus or human herpes virus 8, has been identified in HIV-positive cases (IARC 1996). The same virus has been detected in most analysed cases of non-HIV-related sarcoma and is likely to play a causal role in the development of most or all of Kaposi’s sarcoma cases. The transmission of the herpes virus in HIV-infected individuals is likely to be via sexual contact: mode of transmission and cofactors in non-HIV individuals are unknown.
Cancer of the breast (Henderson et al. 1996)
Over 90 per cent of the neoplasms of the breast originate from the ductal epithelium, while a minority originates from the lobular epithelium. Survival from breast cancer has slowly increased in developed countries, where it now approaches 80 per cent at 5 years. It is unclear how much of the improvement is due to screening as compared with improved treatment. Survival in developing countries remains poor, of the order of 40 to 60 per cent (Sankaranarayanan et al. 1998).
Breast cancer is the most common cancer among women worldwide: the estimated number of new cases in 2000 was 579 000 in developed countries and 471 000 in developing countries (Ferlay et al. 2001). It is also the most important cause of neoplastic deaths among women, causing an estimated 373 000 deaths worldwide in 2000 (Ferlay et al. 2001). The incidence of breast cancer is low (less than 20 per 100 000) in most countries from sub-Saharan Africa, in China, and in other countries of east Asia, except Japan. The highest rates (70 to 90 per 100 000) are recorded in North America, Australia, and northern and western Europe, as well as in Brazil and Argentina. It is important to note that the incidence of breast cancer has grown rapidly during the last decades in many developing countries: for example in Arabic countries breast cancer was a rare disease until the 1970s, and it has now become the most frequent female neoplasm, with rates of the order of 30 per 100 000. In developed countries, incidence rates have slowly increased during recent decades, while mortality rates have remained fairly stable. The incidence increases linearly with age up to menopause, after which a further increase is less marked (developed countries) or almost absent (developing countries) (Fig. 8). Women from high social class have consistently higher rates than women from low social class, the difference being of the order of 30 to 50 per cent.
Fig. 8 Age-specific incidence rates of breast cancer in selected populations in the early 1990s.
The cumulative number of regular ovulatory cycles increase the risk of breast cancer. Ecological and analytical studies have shown an increased risk for early age at menarche and late age at menopause. The risk doubles for a difference of 5 years in age at menarche and of 10 years in age at menopause. Similarly, artificial menopause exerts a protective effect.
Pregnancy increases the risk of breast cancer in the short term, probably because of increases in the level of free oestrogens during the first trimester. This effect is particularly strong at first pregnancy. In the long run, however, pregnancy has a beneficial effect, since parous women have a higher level of prolactin and a lower level of sex hormone-binding globulin than nulliparous women. These two effects result in a protective role of early age at first pregnancy (and a small residual protective effect of additional pregnancies) and in an increased risk of women with late first pregnancy. A protective effect of lactation has been shown in several populations, which is probably attributable to the suppression of the ovulatory function caused by nursing. Furthermore, long-term use of hormone replacement therapy carries also an increased risk of breast cancer.
The combined evidence from reproductive factors indicates an important role of oestrogens in breast carcinogenesis. However, a direct assessment of the role of oestrogens can be made in prospective studies containing measurement of hormones made in prospectively collected biological samples and few results are currently available from such studies.
Women suffering from the two most common benign breast diseases, fibrocystic disease and fibroadenoma carry a two- to threefold increased risk of breast cancer. It is likely, however, that the lesions are not preneoplastic conditions, rather that epithelial proliferation, linked to hormonal alterations, is a feature they share with breast cancer.
A familial history of breast cancer in the mother or in a sister is associated with a two- to threefold increased risk of the disease. This role of familial history is likely to result from low-penetrance genes associated with hormonal metabolism and regulation, in particular several of the genes associated with steroid metabolism are polymorphic; and there is some evidence of an increased risk of breast cancer associated with polymorphisms of genes involved in hormone metabolism. In addition, breast cancer risk is greatly increased in carriers of mutations of several high-penetrance genes, in particular BRCA1, BRCA2, and p53. Although the cumulative lifetime risk in carriers of these genes might be as high as 80 per cent, they are rare in most populations and explain only a small fraction (2 to 5 per cent) of total cases. However, there are exceptions such as Ashkenazi Jews, among whom high-risk BRCA1 or BRCA2 mutations are responsible for an estimated 12 per cent of breast cancers.
Although a role of nutrition in breast cancer risk is strongly suggested by international comparisons, the combined evidence from epidemiological studies is only suggestive of a protective role exerted by high intake of fruits and vegetables, while the evidence is inconclusive for other dietary components, including intake of total fat, saturated fat and fibres, and total energy intake. Similarly, results on micronutrients have been elusive, although there is growing evidence of a protective role played by phyto-oestrogens. Hormonal level and nutritional factors during the intrauterine period and childhood are also likely to be important in breast carcinogenesis.
Many lifestyle factors have been investigated as possible causes of breast cancer. An increased risk with increasing weight has been consistently reported among women older than 60 years, but not among younger women. Drinking of three of more alcoholic drinks per day carries an increased risk of the order of 50 to 70 per cent. It is likely that both overweight and heavy alcohol drinking act on breast cancer risk through mechanisms involving hormonal level or metabolism. Tobacco smoking does not carry an increased risk of breast cancer. Studies of occupational factors and of exposure to organochlorine pesticides have failed to provide consistent evidence of an aetiological role.
About 1 per cent of all cases of breast cancer occur in men. The incidence provides limited evidence of geographical variations, with no clear correlation with incidence in women. Conditions involving high oestrogen level, such as gynaecomastia and sex chromatin positivity, are risk factors for breast cancer in men, as is high weight in youth.
Primary prevention of breast cancer has been attempted via nutritional intervention, involving reduction of energy intake, reduction of proportion of calories from fat, and increase in fruit and vegetable consumption. No evidence of efficacy has been produced so far. Tamoxifen, an anti-oestrogen drug used in chemotherapy, has shown a chemopreventive action against breast cancer, although the magnitude of the protection is uncertain. Conversely, tamoxifen increases the risk of endometrial cancer.
The most suitable approach for breast cancer control is secondary prevention through mammography. The effectiveness of screening by mammography in women older than 50 years has been demonstrated, and programmes have been established in various countries. The effectiveness of mammography in women younger than 50 years is not demonstrated. The benefit of other screening approaches, such as physical examination and self-examination, is not known (Moss 1999).
Cancer of the female genital organs
The female genital organs comprise the ovaries and their annexes, the uterus, the vagina, and the external genitals. The uterus is composed of two parts, the cervix and the corpus, which have very distinct physiological and pathological features. Cancers of the cervix and corpus of the uterus are different histologically, clinically, and aetiologically. However, the distinction between cervix and corpus is often neglected in records used for epidemiological purposes, such as death certificates. Today in Europe and North America, most cancers of the uterus without further specification are likely to be cancer of the corpus. This, however, may not have been the case in the past and in other countries, which complicates temporal and geographical comparisons.
Uterine cervix (Schiffman et al. 1996)
Cervical cancer is a major public health problem in many developing countries. Incidence rates are high (20 to 40 per 100 000) in sub-Saharan African and Latin American countries, as well as in India and southern Asia. In China, the Middle East, northern Africa, and developed countries, rates are of the order of 5 to 15 per 100 000. This results in a number of cases each year in excess of 471 000, 80 per cent of which occur in developing countries, where it represents the second most common female neoplasm after breast cancer (Ferlay et al. 2001). The number of estimated cancer deaths in developing countries (194 000 in 2000) exceeds that from breast cancer. Incidence and mortality rates have decreased steadily in developed countries, but an upturn has been observed among young women. Few data on temporal trends are available from developing countries, but incidence has likely decreased during recent decades. In high-risk countries, rates increase up to the age of 60 years, while in developed countries there is little increase above the age of 40. In most countries, cervical cancer hits preferentially women of lower education and social class.
Most cervical cancers originate from the area of squamous metaplasia called the transformation zone, which is adjacent to the junction between the columnar epithelium of endometrial origin and the keratinizing epithelium of vaginal origin. Most invasive cancers are squamous cell carcinomas or mixed adenosquamous tumours. Invasive carcinoma is preceded by inflammatory and condylomatous atypia, mild dysplasia (also called cervical intraepithelial neoplasia of grade 1), moderate dysplasia (cervical intraepithelial neoplasia grade 2), severe dysplasia and carcinoma in situ (which together represent cervical intraepithelial neoplasia grade 3). Atypia and cervical intraepithelial neoplasia grade 1 represent benign alterations due to infection with human papilloma virus, while cervical intraepithelial neoplasia grade 2 and 3 represent steps towards the malignant transformation.
Chronic infection with human papilloma virus is a necessary cause of cervical cancer. Using sensitive molecular techniques, virtually all tumours are positive for the virus, while the prevalence in non-diseased women represents 5 to 40 per cent in the different populations. Different types of human papilloma virus exist, and those associated with cervical cancer are mainly types 16, 18, 31, and 45. In particular, human papilloma virus 16 is a main carcinogen in many populations, while the distribution of other types varies by geographical region (Fig. 9). Differences in prevalence of human papilloma virus infection explain much of the descriptive epidemiology of cervical cancer (geographical patterns, high risk in low social class, and so on). The host response to human papilloma virus infection is important in determining its possible carcinogenic effect; immunosuppression, as present in transplanted patients and HIV-infected individuals, increases the risk of dysplasia and carcinoma in situ.
Fig. 9 Prevalence of human papilloma virus types in cases of cervical cancer, by region. (Source: N. Muñoz, personal communication.)
Sexual characteristics of women (early age at first sexual intercourse and high number of sexual partners, in particular before the age of 20 years) and of their male partners (high number of sexual partners, presence of genital diseases, and contact with prostitutes) have been found to be risk factors for cervical cancer in many populations. They mainly reflect an increased likelihood of human papilloma virus infection, in particular at a young age.
Studies of infection with other agents, in particular Chlamydia and herpes simplex 2, have failed to provide consistent evidence of an effect independent of human papilloma virus. An increased risk, of the order of twofold, has been detected among long-term users of oral contraceptives, which is not completely explained by sexual behaviour or human papilloma virus infection. Similarly, long-term acting hormonal contraceptives might increase the risk of cervical cancer. Conversely, the use of the condom and diaphragm seems to exert a protective effect, possibly via prevention from human papilloma virus infection.
The carcinogenic effect of tobacco smoking that was suggested by early studies has not been confirmed once human papilloma virus infection was taken into account. A possible protective effect of a diet rich in fruits and vegetables has been suggested in several studies, an effect that might be due to increased intake of carotenoids or folate.
Cytological examination of exfoliated cervical cells (the Papanicolaou smear test) is effective in identifying precursor lesions, resulting in a decrease in incidence of and mortality from invasive cancer. The benefit is of the order of a two- to fourfold decreased incidence. There is no conclusive evidence, however, regarding the optimal timing of the test. Cytological smears are not applicable, however, in countries with limited availability of cytologists and pathologists, as in many countries with a high prevalence of human papilloma virus infection and a high incidence of invasive cancer. Alternative approaches for secondary prevention have therefore been proposed, including visual inspection of the cervix with possible enhancement of precursor lesions by acetic acid. Use of human papilloma virus testing as a screening method, either as a first choice for general application or as the triage method of inconclusive cytological diagnoses, is also under trial. The primary method for prevention of cervical cancer, however, would be human papilloma virus vaccination. Several vaccines are currently under development; this work is complicated by the geographical variations in the distribution of human papilloma virus types.
Uterine corpus (Grady and Ernster 1996)
Cancer of the endometrium is the main malignant neoplasm of the uterine corpus, while sarcomas, originating from the muscular tissue, are relatively rare. The descriptive epidemiology of cancer of the uterine corpus is complicated by the large proportion of hysterectomized women in industrialized countries (up to 30 per cent in some birth cohorts of women from the United States). The number of new cases occurring in 2000 worldwide was estimated of the order of 189 000, of which 60 per cent occur in developed countries. The number of cancer deaths is of the order of 45 000 (Ferlay et al. 2001). Rates are relatively high (10 to 15 per 100 000) in Europe, North America, Brazil, and Argentina, as well as in Australia, while they are below 5 per 100 000 in most African countries and in southern and eastern Asia. In high-risk countries, the incidence increases up to 60 years of age, and is stable or declines above that age. In the United States, the incidence is higher in white people as compared with black people, while the opposite applies to mortality. The incidence of endometrial cancer is declining in developed countries; few data on trends are available from other countries.
Nulliparity, infertility, and late age at menopause are associated with a two- to threefold increased risk of endometrial cancer. The evidence regarding other reproductive factors is less consistent. Elevated levels of endogenous oestrogens (e.g. in the case of oestrogen-secreting tumours and polycystic ovarian syndrome) are associated with an increased risk of endometrial cancer. However, studies of blood oestrogen level were inconclusive.
An increased risk of endometrial cancer was reported in the 1970s following the use of sequential oral contraceptives. Conversely, combined contraceptives reduce the risk of endometrial cancer by about 50 per cent. Use of oestrogen replacement therapy is associated with a twofold increase in risk of endometrial cancer. The strength of the association depends on the dose and the duration of use; the relative risk is in excess of 5 for 5 or more years of use of products with 1.25 mg or more of conjugated oestrogen. Addition of progestin to oestrogen replacement therapy may protect from the increased risk of endometrial cancer, but it may also reduce the beneficial effects of oestrogens on cardiovascular disease and osteoporosis. An increased risk of endometrial cancer has also been shown among breast cancer patients treated with tamoxifen: it is unclear at present whether in non-high-risk women the protective effect of tamoxifen on breast cancer outweighs the increased risk of endometrial cancer.
An increased risk of endometrial cancer has been consistently reported among obese and fat women as compared with lean women; the magnitude of the relative risk between the top and the bottom quartile of distribution of weight or body mass is of the order of 2, but it can be as high as 5 for the most obese individuals. An increased risk has also been reported among women with diabetes, which does not seem to be fully explained by increased weight in these patients. A decreased risk of endometrial cancer, of the order of 50 per cent, has been reported among smokers in many populations: this result has been attributed to an anti-oestrogenic activity of smoking. The results of studies of diet and endometrial cancer have been somewhat inconsistent: a possible increased risk may occur following low intake of fruits and vegetables and high intake of saturated fat. Several other potential risk factors have been addressed, including alcohol drinking, hypertension, and history of gallbladder disease, without conclusive evidence of a causal association.
Although changes in the composition of oral contraceptives and oestrogen replacement drugs have probably contributed to the prevention of endometrial cancer in the last decades, no strategy for primary prevention exists, beside the recommendation to reduce weight. Secondary prevention through endometrial cytological sampling has been proposed, but there is no evidence of a benefit of screening asymptomatic women.
Ovary (Weiss et al. 1996)
Most malignant neoplasms of the ovary originate from the coelomic epithelium; less frequent tumours originate from the germ cells (dysgerminomas and teratomas) and the follicular cells (granulosa cell tumours). The estimated number of new cases worldwide in 2000 was 192 000, that of deaths was 114 000 (Ferlay et al. 2001). High incidence rates (of the order of 10 to 12 per 100 000) are found in western and northern Europe and in North America; the lowest rates (below 3 per 100 000) are from China and central Africa. The rates increase up to about the age of 60 years, after which they remain stable. In high-risk countries the rates have remained stable in recent decades.
Pregnancy protects against ovarian cancer, probably through reduction in secretion of gonadotropins. The reduction in risk is of the order of 30 to 60 per cent for three or more pregnancies as compared with none. The same mechanism can be invoked to explain the protective effect exerted by breast feeding which, however, has not been consistently found in all studies. Similarly, the use of oral contraceptives has been consistently shown to protect from ovarian cancer: each year of use carries approximately a 10 per cent decrease in risk. On the contrary, there is no consistent evidence that use of hormonal replacement therapy has an effect on ovarian cancer risk.
The role of hormonal factors in ovarian cancer is unclear; a reduction in risk has been shown following hysterectomy, which might be due to removal of subclinical tumours. A similar reduction in risk, however, has been observed following tubal ligation. The risk of germ cell tumours is increased in girls and young women exposed to exogenous hormones in utero.
An increased risk of ovarian cancer has been reported in relation to high body mass (relative risk of the order of 2 in the highest versus the lowest group), although the evidence is not fully consistent. Similarly, an effect of pregestational body weight of mothers has been shown for germ cell ovarian cancers.
The evidence regarding an effect of diet on ovarian cancer risk is largely inconsistent. A possible protective effect of a diet rich in fruits and vegetables has been suggested. Atomic bomb survivors have an increased risk of ovarian cancer, as have women exposed to irradiation following benign pelvic conditions: an effect of exposure to low-level ionizing radiation, however, has not been shown. Early suggestions of an increased risk of ovarian cancer among women exposed to asbestos are now interpreted as resulting from misclassification of peritoneal mesotheliomas. A possible association has been suggested between perineal use of talc and ovarian cancer risk.
The prevention of ovarian cancer is currently hampered by the limited knowledge of its causes and the lack of availability of early diagnostic techniques.
Vagina and vulva (Daling and Sherman 1996)
The incidence of vaginal and vulvar cancers ranges between 0.5 and 2.5 per 100 000: high rates are reported from Europe and North America, while low rates are recorded in eastern Asia. The incidence of vulvar cancer is about three times that of vaginal cancer. They are mainly squamous cell carcinomas; rare types are melanoma, sarcoma, and clear cell adenocarcinoma. Their incidence increases exponentially through life.
Infection with human papilloma virus is the best known risk factor for these cancers, with a relative risk of 10 or greater. Human papilloma virus 16 is the most commonly reported type. Other sexually transmitted diseases have been reported more frequently among cases of vulvar cancer than controls: they include syphilis, infection with herpes simplex type 2, and lymphogranuloma venereum. The evidence, however, is based on few studies that have not controlled for concomitant human papilloma virus infection. Similarly, the associations reported with early age at first intercourse and number of sexual partners might simply reflect human papilloma virus infection.
Pelvic irradiation for cervical cancer or other diseases is another risk factor for vaginal and vulvar cancers. The occurrence of these neoplasms is associated with the presence of other cancers in the anogenital tract; this phenomenon might be explained by human papilloma virus infection at several sites. An association between tobacco smoking and vulvar and vaginal cancers has been suggested in a few studies, which did not control for human papilloma virus infection.
Clear cell adenocarcinoma of the vagina is a rare tumour occurring in girls and young women, mainly following in utero exposure to diethylstilboestrol, a synthetic oestrogen used between 1940 and 1970 in several countries. An increased risk of the order of twofold has also been reported for squamous cell carcinoma of the vagina and cervix.
Control of human papilloma virus infection is likely to be an important primary preventive measure for these as well as other anogenital cancers. Screening through visual inspection of high-risk women is recommended.
Cancer of the male genital organs
Prostate (Ross and Schottenfeld 1996)
Cancer of the prostate shows considerable geographical variability. It is the most common malignant neoplasm in men from North America, where the incidence is as high as 100 per 100 000. In other developed countries, the incidence is of the order of 20 to 40 per 100 000, and in most developing countries it is below 30 per 100 000, and it can be as low as 5 per 100 000 in southern and eastern Asia. The estimated number of new cases occurring worldwide in 2000 is about 543 000 (Ferlay et al. 2001). Mortality rates show less variability among regions, suggesting that the number of non-fatal cases diagnosed in different countries varies depending on screening and other diagnostic procedures. The estimated number of deaths is 204 000, of which two-thirds occur in developed countries. The incidence of prostate cancer increased slowly during the last decades in most populations; in the United States and Canada, a very rapid increase has been observed since the mid-1980s. Mortality has also been increasing in most countries, although in the last year a decline has been shown in North America. The disease is more common in African-Americans than in European Americans. In most countries, it is more common among affluent groups of the population.
Increased testosterone level in serum has been reported as a risk factor for prostate cancer in several retrospective studies. However, these results might be due to altered hormonal level related to the development of the tumour. The available prospective studies failed to provide convincing evidence of an increased risk linked to increased levels of testosterone or other sexual hormones. Similarly, an increased risk of prostate cancer was reported in retrospective studies following a history of benign prostatic hypertrophy, but no excess risk was found in prospective studies. If an association exists between prostatic hypertrophy and cancer, it could be due to shared aetiological factors or to a common pathological process.
Carriers of BRCA1 and BRCA2 mutations have a four- to fivefold increased risk of prostate cancer. Generally, a history of prostate cancer in first-degree relatives carries a two- to threefold increased risk of developing the same neoplasm. Similar associations, of smaller magnitude, are also suggested for a family history of breast and colon cancers.
It has been shown in several populations that the risk of the disease increases with the number of sexual partners and number of encounters with prostitutes, and with previous history of syphilis and gonorrhoea. Serological studies of human papilloma viruses 16 and 18 have shown an increased risk among positive subjects. It is not clear at present, however, whether syphilis and human papilloma virus are causal factors or markers of infection with other sexually transmitted agents. A possible increased risk following vasectomy has been reported in some studies: other investigations, however, have failed to confirm this association, and no conclusions can be drawn at present.
A possible protective role of high intake of vegetables has been shown in several studies; high intake of meat, dairy products, total fat, and saturated fat might represent a risk factor. The evidence concerning other dietary factors, including fruit intake and intake of specific micronutrients, is inconclusive at present, with the possible exceptions of lycopene, a retinoid present in particular in tomatoes which has been found to be associated with a reduced risk in several studies, and calcium which has been associated with an elevated risk. There is also a suggestion of an increased risk among individuals with elevated serum levels of insulin-like growth factor 1. An increased risk of the disease has been repeatedly reported among subjects with a high weight or body mass.
The wide geographical variability of prostate cancer strongly suggests that environmental factors probably related to diet and other lifestyle factors, such as physical activity, are the main determinants of the disease. Primary prevention, however, is hampered by the fragmentary knowledge of its precise causes. Secondary prevention has been proposed, based on digital rectal examination and measurement of prostate-specific antigen. There is no evidence from controlled trials that either procedure decreases the mortality from prostate cancer. Despite this lack of evidence, these procedures, in particular the prostate-specific antigen testing, have gained popularity in many countries, and are most likely the cause of the steep increase in number of diagnosed cased since the mid-1980s in North America. It is unclear whether the decrease in mortality reported since the mid-1990s in the United States can be attributed to a beneficial effect of unplanned use of prostate-specific antigen testing.
Testis (Schottenfeld 1996)
Some 95 per cent of malignant neoplasms of the testes arise from the germinal tissue. About half of the germinal neoplasms are seminomas, while the remaining comprise teratomas and a variety of rare lesions. Testicular cancer is common at a young age, and its incidence decreases after the age of 30 years (Fig. 10). Teratomas and other non-seminomatous neoplasms predominate before the age of 15 years, after which most tumours are seminomas. Incidence rates are high (3 per 100 000 or more) in the northern part of South America and western Europe, and are low (1 per 100 000 or less) in most of Africa and in eastern and southern Asia. In the United States, rates are higher in European Americans than in African-Americans. The global number of new cases in 2000 has been estimated at 49 000, that of deaths at 8700. The incidence has increased in most countries during the last decades, with evidence of a birth cohort effect. In many countries, the risk is higher in the more affluent groups of the population.
Fig. 10 Age-specific incidence rates of testicular cancer in white Americans, 1988 to 1992. (Source: Parkin et al. 1997.)
Cryptorchism is the best known cause of testicular cancer. The relative risk is of the order of 3 to 10; this risk factor might be responsible for up to 10 per cent of all cases of testicular cancer. The risk is lower when orchiopexy is performed before the age of 10 years than at older ages. This finding suggests that micro-environmental factors might be responsible for the development of cancer in the undescended testis. Several rare diseases in gonadal differentiation, including Klinefelter’s syndrome, increase the risk of non-germinal tumours of the testis. Familial aggregation has also been shown for different types of testicular cancer.
Exposure to elevated oestrogen levels during pregnancy, from either endogenous or exogenous origin, might be a risk factor for testicular cancer, although the evidence of an association is not fully consistent among studies.
The limited knowledge about the causes of testicular cancer makes it difficult to devise effective preventive strategies, with the exception of early surgical treatment of cryptorchism.
Penis (Wideroff and Schottenfeld 1996)
Cancer of the penis comprises mainly squamous cell carcinomas, which are commonly preceded by intraepithelial lesions, including Queyrat’s erythroplasia and Bowen’s disease. The neoplasm in more common (rates of the order of 2 to 4 per 100 000) in Latin America and in south and Southeast Asia, and is rare (0.5 to 1 per 100 000) in North America and Europe. In developed countries, incidence declined during the last decades.
Chronic infection with human papilloma virus is the most important risk factor for penile cancer: the proportion of positive cases is of the order of 90 per cent. Human papilloma virus type 16 is the most frequently found, followed by types 18, 31, 33, and 35. Although the distribution of penile cancer and penile intraepithelial neoplasia parallels the distribution of the corresponding cervical lesions in women, the incidence is generally lower, suggesting differences in either tissue susceptibility or in the role played by cofactors.
A decreased risk of penile cancer has been reported among circumcised men, in particular among those operated on in infancy. The most likely explanation of this effect is a protection against human papilloma virus infection; however, the available studies do not clearly exclude a possible confounding effect of low-risk sexual behaviour. Furthermore, a high risk of penile cancer (relative risk of the order of 10 to 30) has been found among men with phimosis, in particular when it is associated with balanoposthitis (inflammation of the glans penis and prepuce).
Patients treated with oral 8-methoxypsoralen and ultraviolet A radiation for psoriasis or other skin disorders are at increased risk of penile and scrotal cancer, with a relative risk of the order of 50. The carcinogenic effect is attributable to the combination therapy rather than to the skin disease.
Prevention of human papilloma virus infection through vaccination would be the most important preventive measure for penile cancer. Human papilloma virus testing might be used for screening of high-risk individuals. A beneficial effect of circumcision remains unproven. Prevention of exposure to ultraviolet radiation of the genitals might also be beneficial.
Cancer of the urinary organs
The urinary organs comprise the kidneys (renal parenchyma), renal pelvis, ureter, urinary bladder, and urethra. The majority of urinary cancers occur within the bulk of the kidney (renal parenchyma) and bladder. Cancers of the renal pelvis, ureter, and urethra are much rarer and less well studied and are, for statistical purposes, often grouped under kidney cancers. However, these tumours are histologically more similar to those of the bladder than the renal parenchyma, and their aetiology is also likely to be more similar to the aetiology of bladder cancer.
Kidney (McLaughlin et al. 1996)
Worldwide, there were approximately 189 000 kidney cancers in 2000, two-thirds of which occurred in developed countries (Ferlay et al. 2001). The vast majority of cancers which arise in the renal parenchyma are adenocarcinomas, although nephroblastoma (Wilms’ tumour) occurs in children. The highest incidence of renal cell cancers are observed in the Czech Republic, with rates of approximately 15 per 100 000 in men and 7.5 per 100 000 in women. High rates are also observed in parts of Italy and Poland, and among some black populations of the United States. Conversely, rates up to 10 times lower are reported in most Asian and African populations and some, but not all, South American populations. The incidence in males is approximately twice that among females in most populations. The incidence of kidney cancer increases logarithmically from the age of 30 years and plateaus around the age of 60. Strong increasing trends in kidney cancer are occurring among all Caucasian populations and in Japan. Some of the most steep increases are observed among black Americans. It is unlikely that these increasing trends can be explained by increasing detection of presymptomatic tumours and are instead likely to reflect real increases in the number of new cases.
Cigarette smoking has been consistently observed to be a moderate risk factor for kidney cancer, with increased risks compared with never smokers of the order of 1.2 to 2.3. A number of studies have also demonstrated a dose–response relationship with increasing consumption, with risks for heavy smokers ranging from 2.0 to 3.0. The risk appears to decline with increasing years of smoking cessation. Population attributable risk estimates indicate that cigarette smoking, both past and present, is responsible for between 27 and 37 per cent of kidney cancer cases among men, and between 10 and 24 per cent of cases among women. Approximately half of this attributable risk is due to current smoking.
Obesity has been consistently linked with kidney cancer, especially among women, with less consistent and weaker results among men. The mechanism by which obesity causes kidney cancer is unclear, although hormonal changes such as increased levels of endogenous oestrogens may be responsible. Other correlates of obesity, such as hypertension and lack of physical exercise, have not been found to explain this relationship. The population attributable risk of kidney cancer associated with excess weight (defined as being in one of the last three quartiles of body mass index) has been estimated to be over 40 per cent in women and 5 per cent in men.
A history of hypertension has also been linked to kidney cancer, although the strength of this relationship has generally been greatly reduced after adjustment for use of diuretics and other antihypertensive drugs. These findings suggest that use of medications may be the primary risk factor and not hypertension per se. Both diuretic and non-diuretic antihypertensive medications have been linked to kidney cancer, with supportive evidence from animal studies. However, identifying whether the real risk is due to the hypertensive state or due to antihypertensive medication has not so far been possible. Whichever of the two is the real risk factor, it is likely to account for a substantial proportion of cases. The attributable risk of reported hypertension or treatment with antihypertensive drugs has been estimated to be 21 per cent overall, and 39 per cent among women.
High protein consumption from meat and dairy products has been associated with chronic renal conditions that may predispose to kidney cancer, and has been weakly associated with an increased risk of kidney cancer of between 30 and 70 per cent, although the evidence is inconsistent. A possible protective effect has also been identified in a number of studies for high consumption of vegetables.
An excess risk of renal cell cancer has been observed in a variety of occupations with exposure to polycyclic aromatic hydrocarbons such as coke and coal oven workers, fire-fighters, and asphalt and tar workers. Excess risks have also been reported for occupations with exposure to gasoline and other petroleum products such as oil refinery workers and petrol station attendants. Exposure to organic solvents and in particular to chlorinated liphatic hydrocarbons has also been suggested as a risk factor in occupations including dry-cleaning and printing. Overall however, the evidence for associations with specific occupational exposures is still inconclusive.
The main avoidable causes of kidney cancer are cigarette smoking and excess body weight, which together account for up to 50 per cent of all cases. Primary prevention in reducing cigarette smoking and obesity are therefore the clearest strategies for reducing the incidence of the disease. A substantial proportion of cases are also likely to be related to hypertension although further information on whether the true risk factor is the disease or the treatment is required in order to clarify implications for prevention.
Renal pelvis and ureter (McLaughlin et al. 1996)
Renal pelvis and ureter cancers are mainly transitional cell carcinomas as opposed to adenocarcinomas of the kidney. Renal pelvis cancers occur with approximately 10 per cent of the frequency of renal cell cancers, and ureter cancers with a frequency of approximately 5 per cent. For both sites, high rates are observed in parts of Europe including northern Italy, Switzerland, the Czech Republic, and Poland, as well as Australia and the United States. In the United States, rates are higher among white people than black people, in contrast to the incidence of kidney cancer. The male to female ratio is again approximately 2 to 1.
A region of specific interest is the Balkan peninsula because of the high incidence of Balkan endemic nephropathy, a chronic kidney disease associated with progressive renal failure. This condition is mainly restricted to specific rural areas of the Balkans and is also associated with a very high incidence of tumours of the renal pelvis and ureter. The cause of the phenomenon is not known although food-related mycotoxins such as ochratoxin A are suspected.
The main risk factor for renal pelvis and ureter cancers is cigarette smoking, with increased risks for smokers compared with non-smokers in the region of 2.5- to sevenfold. The attributable proportion of renal pelvis and ureter cancers due to cigarette smoking has been estimated to be as high as 70 per cent among men and 37 per cent among women.
A consistent relationship has also been observed between use of phenacytic containing drugs and cancer of the renal pelvis, with relative risks varying from 2.4-fold to over 12-fold (IARC 1987), and also some evidence of a dose–response relationship. Because of concern over links between nephropathy and phenacetin use, it has been removed from analgesics in most industrialized countries, starting in the late 1960s.
Because of the high proportion of renal pelvic and ureter tumours caused by cigarette smoking, especially current smoking, encouraging current smokers to give up will probably result in a substantial reduction in the number of cases.
Bladder (Silverman et al. 1996)
Bladder cancer accounts for approximately two-thirds of all urinary tract cancers with 336 000 cases worldwide in 2000, approximately 50 per cent of which occurred in developed countries (Ferlay et al. 2001). The male to female ratio of 3 to 1 is higher than for kidney cancer. The vast majority of bladder cancers are transitional cell carcinomas, with the exception of bladder cancers related to schistosomiasis infection, which are mainly squamous cell carcimonas. High bladder cancer rates are observed throughout southern and western Europe, Scandinavia, the United States, and Australia, although also in Israel, Egypt, and Uruguay. These high rates generally reflect those countries with high levels of cigarette smoking in the past, with the exception of Egypt where most cases are related to infection with Schistosoma. Bladder cancer incidence is either rising moderately or is steady in most developed countries, probably reflecting past cohort effects in tobacco consumption.
The most important risk factor for bladder cancer is cigarette smoking and is thought to account for approximately 66 per cent of new male cases and 30 per cent of female cases in industrialized populations. It is likely that smokers of black tobacco are at a higher risk than smokers of blond tobacco and this may explain some of the disparity observed in European incidence rates and also the high incidence observed in Uruguay. The risk associated with smoking is likely to be due to aromatic amines present in cigarette smoke, including benzidine, 4-aminobiphenyl, naphthylamine, and 4-chloro-O-toluidine. Bladder cancer risk increases approximately linearly with duration of smoking, reaching a fivefold risk after 5 years. A substantial decrease in risk of bladder cancer is observed within several years for those who gave up smoking, implying a late stage effect in the carcinogenic process.
A high risk of bladder cancer has also been reported among industries which involve exposure to aromatic amines, in particular 2-naphthylamine and benzidine, including the rubber and dyestuff industries. Other occupations which may increase the risk of bladder cancer include leather workers, painters, and drivers, possibly because of exposure to a variety of chemicals including polycyclic aromatic hydrocarbons, polychlorinated biphenyls, formaldehyde, and solvents. The uncertainty surrounding these occupations is again partly due to the difficulty of measuring past exposure to specific chemical agents.
Investigations of diet have provided evidence of decreased risks associated with fruits and vegetable intake, and also some inconsistent evidence of an increased risk associated with coffee consumption. Reasons for the lack of consistency of this latter exposure include the small sample size of many studies and the possibility of residual confounding from other risk factors including cigarette smoking.
The enzyme N-acetyl transferase 2 is involved in the detoxification of various bladder carcinogens including arylamines. Its gene includes a dominant mutation which results in slow metabolization and in higher risk of bladder cancer.
Similar to the renal pelvis, a consistent relationship has been observed between use of phenacytic containing drugs and bladder cancer, with relative risks varying from 2.4-fold to over sixfold (IARC 1987). Cyclophosphamide, an alkylating agent which has been used to treat both malignant and non-malignant diseases has also been linked to bladder cancer. Studies based on cohorts of cancer patients indicate an increase in risk of approximately fivefold associated with cyclophosphamide therapy, with higher risks among heavily exposed subjects.
Schistosoma infection is prevalent throughout Africa and is associated with an increased risk of approximately fivefold. Cases associated with Schistosoma infection are mainly of the squamous cell type. They are responsible for an estimated 10 per cent of bladder cancer cases in the developing world, and about 3 per cent of cases overall. Some studies have reported an increased risk with history of repeated urinary tract infection among both men and women. Bladder infection was more strongly associated with squamous cell bladder cancer. This relationship, however, needs to be confirmed in further studies.
Regarding prevention, avoidance of cigarette smoking is the most effective public health measure against bladder cancer. Approximately 60 per cent of bladder cancer cases are due to smoking, at least half of which could be prevented by smoking cessation among current smokers. Prevention of Schistosoma infection through avoidance of contaminated waters is important in endemic areas. No effective screening approach is available for bladder cancer.
Cancer of the nervous organs
Neoplasms of the eye are rare: the incidence is below 1 per 100 000 in all regions of the world, with the exception of central and southern Africa. The main histological types are squamous cell carcinoma, arising from the conjunctiva; retinoblastoma, which arises in children and is relatively common in Africa; and uveal melanoma, which is the main adult type outside Africa. Solar radiation is a cause of conjunctiva carcinoma and uveal melanoma. About 50 per cent of cases of retinoblastoma are caused by an inherited mutation in the Rb gene.
Nervous system (Preston-Martin and Mack 1996)
Over 90 per cent of nervous system neoplasms arise from the brain and the cranial meninges. Data on the descriptive epidemiology of neoplasms of the nervous system are difficult to interpret because of inconsistent inclusion of benign tumours in different series of cases.
Gliomas arise from the glial cells and are classified pathologically as astrocytomas (low grade) and glioblastomas (high grade). They represent 40 to 60 per cent of primary tumours of the brain and cranial meninges, are predominantly malignant, and are more common in men. Meningiomas arise from the cranial meninges and represent 20 to 35 per cent of brain neoplasms, while schwannomas (or neurilemmomas) arise from the Schwann cells of the nerve sheath (mainly of the eight cranial acoustic nerves) and represent 5 to 10 per cent of all brain neoplasms. These two latter types are mainly benign. Rare types of nervous system neoplasms include pituitary adenomas, childhood primary neuroectodermal tumours (also called medulloblastoma), and tumours of the spine and the peripheral nerves.
The incidence of brain tumours is slightly higher in men than in women; the male to female ratio is approximately 1.5 for gliomas and 0.6 for meningiomas. There is a 10-fold geographical variability in the incidence of brain neoplasms; rates in men are above 6 per 100 000 in most countries from the Americas, Europe, and Oceania, and are below 2 per 100 000 in most African countries and in China. In the United States, rates of gliomas are 30 to 50 per cent higher in white people than in other ethnic groups, while rates of meningiomas are higher in black people. The incidence of gliomas tends to be higher among people from high socio-economic groups.
During the last decades, incidence and mortality from brain tumours have increased in most developed countries. For example, during 1973 to 1987 incidence increased in the United States by 23 per cent and mortality by 9 per cent. However, it is likely that part, if not all, of the increase is due to changes in diagnostic and reporting procedures.
Ionizing radiation is the only established environmental risk factor for brain tumours. It causes all three major types of central nervous system tumours, but the association is stronger for meningioma and schwannoma than for glioma. The evidence comes mainly from studies of atomic bomb survivors and of patients given X-ray therapy in the head and neck region. Head trauma has been suggested as a risk factor for meningioma, and acoustic trauma (as in the case of jobs with exposure to loud noise) as a risk factor for acoustic schwannoma. N-nitroso compounds, in particular nitrosoureas, are potent experimental brain carcinogens, but the evidence of an aetiological role in humans is inconclusive.
Tumours of the central nervous system occur frequently in rare congenital syndromes, such as neurofibromatosis types 1 and 2, von Hippel–Lindau syndrome, and Li–Fraumeni syndrome.
The very limited knowledge about the aetiology of tumours of the central nervous system offers scarce resources for an effective preventive strategy.
Cancer of the endocrine glands
Thyroid (Ron 1996)
An estimated 71 000 new cases of thyroid cancer occurred in 2000 among women, and 33 000 among men (Ferlay et al. 2001). In most areas of the world, the incidence among women is in the range 2 to 5 per 100 000, that in men is between 1 and 2 per 100 000. High-risk areas (incidence over 5 per 100 000 in women) include Central America, Japan, and the Pacific Islands. International comparisons, however, are complicated by possible differences in diagnostic procedures. The most common thyroid neoplasm (50 to 80 per cent of the total) is papillary carcinoma, followed by follicular carcinoma (10 to 40 per cent) and medullary carcinoma (5 to 15 per cent). The incidence of thyroid cancer is very low before the age of 15 years, above which it increases linearly with age, resulting in relatively high rates in young adulthood. In certain populations (e.g. the United States), the incidence in women declines after the age of 50.
Survival from thyroid cancer is very good (a 78 per cent 5-year survival rate in Europe in the early 1990s), resulting in low mortality rates (below 1.2 per 100 000 in women and 0.6 per 100 000 in men in most countries).
In most countries, incidence rates have been stable or have been slowly increasing (less than 1 per cent per year) during the last decades; mortality rates have steadily declined, probably because of improved treatment.
Ionizing radiation during childhood is the main established risk factor for thyroid cancer. The carcinogenic effect seems greater for exposure before the age of 5 years than between 5 and 15 years of age. The pooled analysis of studies of individuals irradiated in childhood for medical conditions and atomic bomb survivors resulted in a summary excess relative risk of 7.7 (95 per cent confidence interval 2.1 to 29) per Gray, and an excess absolute risk of 4.4 (95 per cent confidence interval 1.9 to 10) per 10 000 person years Gray (Ron et al. 1995). An effect of external exposure during adulthood is not established. Several studies have been published on adults exposed to iodine-131 for medical purposes. Although these studies suggest an increased risk, their interpretation is complicated by the fact that these patients were treated because of thyroid diseases. Iodine-131 was the main exposure resulting from the accident of the Chernobyl nuclear reactor in 1986; since then, an increased incidence of thyroid cancer has been reported among children living in the contaminated areas of Belarus and Ukraine. Although part of the excess might be due to increased medical surveillance, it is likely that it also reflects a true phenomenon. Studies of occupational exposure to low-level ionizing radiation, typically in the nuclear industry, have failed to show an increased incidence of thyroid cancer.
The higher incidence of thyroid cancer in women as compared with men and the age-specific rates in women suggest a role of reproductive factors in the development of the disease. A number of studies have shown a moderately increased risk (relative risk of the order of 1.5) in women with more pregnancies and in women with a history of miscarriage. The exact role of hormonal and reproductive factors, however, in unclear. Elevated levels of thyroid-stimulating hormones are associated with thyroid growth and possibly thyroid cancer. The evidence of an association between iodine deficiency (and presence of endemic goitre) and thyroid cancer is equivocal; studies from central and southern Europe support such an association, but is not confirmed in studies from northern Europe and North America. It is possible that iodine deficiency increases the risk of follicular thyroid cancer, while the papillary type is linked to an iodine-rich diet. An increased risk of thyroid cancer has been associated in many studies to a history of thyroid nodules: some of the nodules, however, might represent preneoplastic lesions.
A strong genetic component has been shown for medullary carcinoma: about 20 per cent of these neoplasms are associated with an autosomal dominant gene with penetrance close to 100 per cent. It can also be associated with other endocrine neoplasms within the multiple endocrine neoplasia syndromes. Familial factors also play a role in papillary carcinoma. Among the genes associated with thyroid cancer are the ret proto-oncogene (for papillary and medullary carcinomas) and the APC gene for papillary carcinoma.
A possible role of dietary factors in thyroid carcinogenesis has been addressed in many studies. High intake of vegetables, in particular cruciferous vegetables, might protect against the disease, possibly through interference with iodine metabolism. The evidence regarding other dietary factors is inconclusive.
The prospects for prevention of thyroid cancer are complicated by the limited understanding of its aetiology, with the exception of relatively rare high-risk conditions, such as childhood exposure to ionizing radiation and high-risk families.
Malignant neoplasms of endocrine glands other than the thyroid are rare in most populations. The incidence rates range between 0.1 and 1 per 100 000 people. About two-thirds of these neoplasms arise from the adrenal gland. About one-third of these tumours are carcinomas, and the remaining proportion shows different histological patterns.
Genetic susceptibility plays a relatively important role for this group of neoplasms: in particular, adrenocortical carcinoma is found in cases of Li–Fraumeni syndrome, and malignant phaeochromocytoma is found in multiple endocrine neoplasia type 2 and the von Hippel–Lindau syndromes. Tobacco smoking has been associated with an increased risk of adrenal cancer in a few studies. The evidence for a carcinogenic role of other factors is inconclusive.
Neoplasms of the lymphatic and haematopoietic organs
The term lymphoma encompasses a diverse group of neoplasms which originate from the cells of the lymphopoietic system. Traditionally, two main groups of lymphomas have been distinguished including Hodgkin’s disease, characterized by large polynuclear cells named after Reed and Sternberg, and a diverse group of other lymphomas, defined as non-Hodgkin’s lymphomas. The complexity of lymphomas is reflected by the various classifications that have been used to separate different subtypes. The most recent classification system, the Revised European American Lymphoma classification and its adaptation by the World Health Organization (WHO) represents an effort to reach a wide medical consensus to allocate all lymphoma cases in clear categories. Neoplasms are divided between B- and T-cell lymphocytes, with over 20 different clinicopathological entities. Importantly, this classification incorporates all lymphoproliferative diseases, including multiple myeloma, B-cell acute lymphoblastic leukaemia, Burkitt’s lymphoma, and Hodgkin’s disease. Given that the Revised European American Lymphoma classification has only been in use for a number of years, it is necessary to discuss the characteristics of lymphomas and leukaemias under the traditional groups of Hodgkin’s disease, non-Hodgkin’s lymphoma, multiple myeloma, and leukaemia.
Hodgkin’s disease (Mueller 1996)
The incidence of Hodgkin’s disease varies from low-incidence populations, with rates lower than 1 per 100 000, including areas of southern and eastern Asia and of sub-Saharan Africa, to high-incidence populations, with rates of the order of 3 per 100 000 found in the United States, in certain European regions such as Italy, and in Jews in Israel. The incidence in men is consistently higher than in women with a ratio of between 1.5 and 2. The incidence has been relatively stable over time and may even be declining. The age of onset of Hodgkin’s disease shows a bimodal distribution in developed populations with a first peak between the age of 15 and 35 years and a second peak after the age of 60. In developing countries the first peak tends to be observed during childhood. This bimodal distribution may suggest that the category of Hodgkin’s disease includes at least two different entities.
Viral infections play an important role in the aetiology of Hodgkin’s disease. Its onset may be related to decreased or delayed exposure to infectious agents during childhood, as indicated by its association with having fewer siblings, living in single-family houses, and early birth order.
Epstein–Barr virus infection is associated with a large proportion, and potentially all, of Hodgkin’s disease cases, although its aetiological role remains to be fully established. Epstein–Barr virus is, however, ubiquitous throughout the world with 80 to 100 per cent of individuals being infected by the age of 30 (IARC 1997a). In developing countries infection occurs earlier in life, whereas in developed countries infection is often delayed until adolescence. The Epstein–Barr virus genome is present in about 50 per cent of the Reed–Sternberg cells of cases, and another Epstein–Barr virus-related condition, infectious mononucleosis, is associated with a moderately elevated risk of subsequent development of Hodgkin’s disease. Sero-epidemiological studies indicate that patients with Hodgkin’s disease can be distinguished by an altered antibody profile to Epstein–Barr virus. All of this evidence taken together argues against Epstein–Barr virus simply being a passenger virus, and for a causal association in the development of the disease.
There is only limited evidence of other possible risk factors for Hodgkin’s disease. Employment in the wood and chemical industries may entail an increased risk of Hodgkin’s disease although the responsible carcinogens, if any, have not been identified.
Non-Hodgkin’s lymphoma (Scherr and Mueller 1996)
The incidence of non-Hodgkin’s lymphoma is consistently higher than the incidence of Hodgkin’s disease. High rates of over 10 per 100 000 are reported from the United States, Australia, western Europe, and from Israel and the Middle East, while low rates of less than 5 per 100 000 are reported from southern and eastern Asia and parts of Africa. Men have a 1.5 to twofold higher incidence than women. There is a strong geographical variation of incidence rates for some lymphoma subgroups. For example, Burkitt’s lymphoma is common among children in eastern Africa, and adult T-cell leukaemia/lymphoma is increased in southern Japan and parts of Africa. Conversely, the trend of non-Hodgkin’s lymphoma shows a steady increase with age in most populations. Exceptions are the populations in which a specific type of lymphoma predominates, such as Burkitt’s lymphoma in children.
A striking feature of the epidemiology of non-Hodgkin’s lymphoma is the consistently increasing incidence which has been observed in all developed countries. The rate of increase for non-Hodgkin’s lymphoma has been approximately 4 per cent per year in most populations where accurate data are available, indicating a doubling of incidence every 20 years. The reasons for the increase in non-Hodgkin’s lymphoma incidence have been widely discussed and it is possible that improvement in diagnostic procedures explains part of it, particularly in the elderly. However, it is now accepted that the trend also reflects a real increase in the number of cases, the cause of which is not known.
The current knowledge of potential risk factors for non-Hodgkin’s lymphoma is limited. However, there is strong evidence that altered immunological function, either immunostimulation or immunosuppression, entails an increased risk of non-Hodgkin’s lymphoma. For example, immunosuppressed renal transplant patients have a risk 30 times higher for developing lymphoma compared with the general population. Lymphomas that develop in immunosuppressed patients share common characteristics. They are generally high-grade B-cell lymphomas and are more likely to be extranodal and of worse prognosis. Lymphomas have also been reported for a variety of other conditions which are either autoimmune in nature, or require immunosuppressive treatment, including rheumatoid arthritis and Sjögren’s syndrome.
Infectious agents associated with non-Hodgkin’s lymphoma include HIV, human T-cell lymphotropic virus 1, and Epstein–Barr virus. Hepatitis C virus, human T-cell lymphotropic virus 2, and human herpes viruses 6 and 8 have also been linked to the development of non-Hodgkin’s lymphoma. In addition, infection with H. pylori is a risk factor for gastric lymphoma.
Epstein–Barr virus is particularly prominent in lymphomas developing in immunosuppressed patients, and also in Burkitt’s lymphomas. The relationship with other forms of lymphoma is, however, unclear. Regarding HIV, non-Hodgkin’s lymphoma is 60 times more frequent among patients with AIDS than in the general population (IARC 1996). About 3 per cent of the patients with AIDS develop a non-Hodgkin’s lymphoma, which represents a small contribution to the overall incidence of non-Hodgkin’s lymphoma, except in populations with a high HIV prevalence such as regions of sub-Saharan Africa. The AIDS-related lymphomas tend to be high-grade B-cell lymphomas.
Human T-cell lymphotropic virus 1, and possibly type 2, appear to be associated with the rare adult T-cell leukaemia/lymphoma, a disease entity with strong geographical clustering in Japan, the Caribbean, and parts of Africa. Transmission of the human T-cell lymphotropic virus is similar to that of HIV, involving vertical (mother-to-child) transmission, sexual contact, or blood transfusion.
A familial aggregation is present for both non-Hodgkin’s lymphoma and Hodgkin’s disease: the risk of non-Hodgkin’s lymphoma among first-degree relatives of cases has been reported of the order of 1.5 to 4. However, the risk seems higher for siblings of the same sex, suggesting a role of shared environmental factors rather than genetics. Highly penetrant genetic predisposition to lymphomas is not very common but include ataxia telangiectasia, Wiskott–Aldrich syndrome, and hypogammaglobulinaemia. Approximately 25 per cent of the patients with rare forms of genetic immunodeficiency will develop a lymphoma.
The increasing recreational exposure to ultraviolet radiation in some populations and the decrease in the atmospheric ozone layer have been related to the observed increase in the incidence of non-Hodgkin’s lymphoma. This hypothesis is supported by cancer registry data which show a strong link between non-Hodgkin’s lymphoma and skin cancers, including the frequent occurrence of non-Hodgkin’s lymphoma both before and after skin cancer in the same individual. The hypothesis that exposure to ultraviolet light also causes lymphoma has not been proven but it is biologically plausible as there is experimental evidence showing that ultraviolet light radiation produces systemic suppressive effects on the immune system.
Exposure to pesticides has been associated with non-Hodgkin’s lymphoma risk in studies conducted both on manufacturing workers and applicators in agriculture. The results, however, are not very compelling, with the possible exception of phenoxy herbicides and chlorophenols. This effect might be due to contamination with dioxin. Farming as an occupation has also been weakly associated with non-Hodgkin’s lymphoma risk. It is not clear whether this risk, if present, is accounted for by exposure to ultraviolet radiation, pesticides, or animal viruses. Organic solvents represent another group of chemicals whose association with non-Hodgkin’s lymphoma risk has been widely investigated. Only for tetracholorethylene, however, is the evidence somewhat consistent.
Multiple myeloma is a malignancy of the plasma cells with a variable manifestation. High-incidence areas of around 4 per 100 000 include North America, western Europe, and Oceania. Low rates of around 1 to 2 per 100 000 are reported in most of Asia, although part of this may be due to underdiagnosis. Within the United States, black people have approximately double the incidence rate of white people, with the incidence approaching 10 per 100 000 in some areas. Whether this increase among black Americans is due to genetic or environmental effects is unknown. In most populations incidence rates are higher in men although the ratio is usually less than 2. Multiple myeloma rates are very rare among young adults although the incidence increases exponentially from the age of 30 years and plateaus after 60 years of age. It also appears that the incidence of myeloma is increasing, although these trends are difficult to judge and may be due to diagnostic artefacts.
The only established risk factor for multiple myeloma is monoclonal gammopathy of unknown significance, an asymptomatic non-malignant disorder involving proliferation of plasma cells. The increased risk associated with monoclonal gammopathy of unknown significance appears to be of the order of 10-fold, although the lifetime absolute risk is still relatively low (less than 5 per cent). Other conditions have also been reported to be associated with myeloma including rheumatoid arthritis and allergies, although the evidence is inconclusive.
Ionizing radiation has also been reported to be associated with myeloma although results are again inconsistent. Occupational groups which have been reported to have higher myeloma rates include painters and farmers, which might be due to exposure to benzene, other solvents, and pesticides.
Leukaemias arise in one of the types of white blood cells. They may arise in lymphoblasts, which are lymphoid cells in the early stage of development, resulting in a rapid onset illness termed acute lymphoblastic leukaemia. Alternatively, when the neoplasm involves mature cells, it is termed chronic lymphocytic leukaemia and is usually more sedate. Leukaemias may also be granulocytic in origin, occurring in either young myeloblast cells resulting in acute myeloid leukaemia, or in the mature granulocytes resulting in chronic myeloid leukaemia. There also exist several rarer varieties including monocytic and hairy cell leukaemias.
Acute lymphoblastic leukaemia is the most common childhood cancer, while over 80 per cent of lymphoid leukaemias occurring in adulthood are chronic lymphocytic leukaemia. Incidence rates for chronic lymphocytic leukaemia are difficult to interpret because it is often diagnosed incidentally, or in the course of evaluating other conditions. Differences in medical care may therefore substantially bias incidence data. Bearing this possible ascertainment bias in mind, the highest rates of lymphoid leukaemias are observed in areas of Canada, the United States, western Europe, and Oceania, and are lower in South America the Caribbean, Asia, and Africa. Rates tend to be lower in females although the ratio is usually less than 2. Some increases in leukaemia over time have been reported although the extent to which these represent real increases in incidence is unclear. Some increasing incidence trends have been reported for both chronic myeloid leukaemia and acute myeloid leukaemia, although these are not consistent and may simply reflect changes in diagnostic practices.
Although the cause of most leukaemias is not known, there does exist consistent evidence for two factors, namely ionizing radiation and occupational benzene exposure. Leukaemia was the first cancer to be linked to ionizing radiation after the atomic bombings in Hiroshima and Nagasaki, and clear excesses have been observed for acute lymphoblastic leukaemia, acute myeloid leukaemia, and chronic myeloid leukaemia, but not for chronic lymphocytic leukaemia. Cohorts of patients who have received radiotherapy for both malignant and non-malignant conditions have also been found to be at an increased risk of leukaemia, usually myeloid. Whether there is any increased risk of leukaemia from other sources, including low-level diagnostic radiation, occupational exposure in the nuclear industry for workers and their offspring, or nuclear test explosions, is more contentious. Part of the problem lies in extrapolating from high acute doses experienced in particular circumstances like atomic bomb exposures in Japan, to small or chronic exposures in other instances. There is no consistent evidence that exposure to electromagnetic fields is associated with leukaemia risk.
Some leukaemias are also induced by therapy for a prior malignancy, most notably Hodgkin’s disease. Such patients have a 20- to 40-fold increased risk of leukaemia, most of which are acute myeloid leukaemia. The risk appears to be related to chemotherapy including alkylating agents (in particular mechlorethamine, Oncovin, procarbazine, and prednisone combination therapy). The effect is greater when patients are treated with both chemotherapy and radiotherapy, although whether an independent effect exists for radiotherapy is unclear. Other chemotherapy regimes which appear to be associated with acute myeloid leukaemia are those which contain the epipodophyllotoxin drugs teniposide and etoposide.
Occupational benzene exposure is also a recognized cause of leukaemia, in particular for acute myeloid leukaemia. An increased risk of between three- and fivefold has been observed in several occupational cohorts of benzene-exposed workers, including workers in rubber manufacturing, petrol refinery, and printing. Tobacco smoking is a suspected cause of leukaemia.
The limited knowledge of the causes of lymphatic and haematopoietic neoplasms limits the opportunity for prevention. Avoidance of known risk factors (e.g. unnecessary radiation exposure, benzene) is likely to result in the prevention of a small proportion of these neoplasms in most populations.
Overview of the causes of human cancer
Figure 11 presents the results of a recent review of the contribution of known causes of cancer in the United States (Harvard Center for Cancer Prevention 1996). This picture broadly applies to other developed countries, while no systematic estimate has been proposed for developing countries. These results, however, are subject to many uncertainties and should be interpreted as approximations.
Fig. 11 Proportion of neoplasms attributable to established causes in the United States. (Source: Harvard Center for Cancer Prevention 1996.)
Tobacco smoking is the main single cause of human cancer worldwide (Chapter 10.1). It is a cause of cancers of the lung, larynx, oral cavity, pharynx, oesophagus, pancreas, kidney, and bladder. It is also a suspected cause of cancers of the nasal cavity and stomach, and of leukaemia. Tobacco smoking has been estimated to cause approximately 30 per cent of all human cancers (Doll and Peto 1981; Harvard Center for Cancer Prevention 1996). However, a detailed review of the number of cancers attributable to tobacco smoking in 1985, which was based on very strict criteria for attribution of cases, resulted in a lower estimate, of the order of 15 per cent, corresponding to about 1 100 000 new cases per year (Parkin et al. 1994). This figure should be considered a conservative estimate, and the actual proportion of cancers is likely to be higher. The estimates by Parkin and colleagues were 25 per cent in men and 4 per cent in women and, in both genders, they were 16 per cent in developed countries and 10 per cent in developing countries. The low attributable risk in women (and, to a less extent, in developing countries) is due to the low consumption of tobacco in past decades: the recent upward trend that has taken place among women and in many developing countries will obviously result in a much greater numbers of cancer in the future.
A benefit of quitting tobacco smoking in adulthood has been shown for all major cancers causally associated with the habit. This result emphasizes the need to devise antismoking strategies that address avoidance of the habit among the young as well as reduction of smoking and quitting among adults. In fact, the decline in tobacco consumption that has taken place during the last 20 years among men in North America and several European countries, and which has resulted in decreased incidence of and mortality from lung cancer, has resulted primarily from the increase in quitting at middle age. The great challenge for the control of tobacco-related cancer, however, lies today in developing countries, in particular in China and the other Asian countries: the largest increase in tobacco-related cancers has been forecasted in this region of the world (Peto et al. 1999). Despite growing efforts from medical and public health institutions and the growing involvement of non-governmental organizations, the fight against the spread of tobacco smoking among women and in developing countries remains the biggest and most difficult challenge of cancer prevention in the next decades.
Use of smokeless tobacco products has been associated with increased risk of head and neck cancer. Chewing of tobacco-containing products is particularly prevalent in southern Asia, where it represents a major carcinogenic factor.
Despite considerable research efforts in recent years, the exact role of dietary factors in causing human cancer remains largely obscure. Recently, the World Cancer Research Fund (WCRF 1997) has published a systematic review of the evidence of an association between intake of foods, food groups, and nutrients, and different cancers. Their evaluations are summarized in Table 4. The evidence of a protective role of vegetable intake and, to a less degree, fruit intake has been evaluated as convincing for a number of important human tumours. In the case of most other dietary factors, the evaluation has only been of a possible increase or decrease in risk. This is namely the case for high intakes of total and saturated fat, and of some micronutrients such as carotenoids, vitamin E, and selenium. In addition, IARC has concluded that there is evidence suggesting a lack of cancer-preventive activity for preformed vitamin A (IARC 1998b) and for b-carotene when used at high doses (IARC 1998a). Despite the limitations of the current understanding of the role of diet in human cancer, systematic reviews have been consistent in suggesting that dietary factors are responsible for about one-third of human cancers, at least in developed countries (Doll and Peto 1981; Harvard Center for Cancer Prevention 1996). The only justified dietary recommendation for cancer prevention is to increase the consumption of vegetables and fruits. A reduction in total caloric intake, however, would contribute to a decrease in obesity, which is also an established risk factor for human cancer (see below).
Table 4 Assessment of associations between dietary factors and human cancer (WCRF 1997)
Obesity and physical exercise
Obesity increases the risk of endometrial cancer and probably the risk of postmenopausal breast cancer and kidney cancer (WCRF 1997). An association is possible also for gallbladder and colon cancer. It is likely that obesity exerts a carcinogenic effect in conjunction with other factors such as insulin resistance, low physical activity, and menopausal status. The magnitude of the excess risk is not very high (for most cancers the relative risk ranges between 1.5 and 2 for body weight higher than 35 per cent above the ideal weight); the attributable risk might be large because of the high prevalence of overweight people. Although no detailed assessments of the risk of cancer attributable to obesity worldwide are available, obesity represents an important target for preventive strategies.
Increased workplace or recreational physical activity decreased the risk of colon cancer and possibly the risk of lung and breast cancer (WCRF 1997). The relative risk of colon cancer for regular versus no activity is of the order of 2. Increasing physical activity should be part of any comprehensive cancer prevention strategy.
Alcohol drinking increases the risk of cancers of the oral cavity, pharynx, larynx, oesophagus, and liver. An association is probable in the case of colon, rectal, and breast cancer (IARC 1988; WCRF 1997). For all cancer sites, risk is a function of the amount of alcohol consumed. Alcohol drinking and tobacco smoking show an interactive effect on the risk of cancers of the head and neck. The evidence of a different effect of various alcoholic beverages is inconclusive at present. Although no systematic review of attributable risks to alcohol drinking is available, it has been estimated to account for 4 to 5 per cent of total cancers (Fig. 11).
There is growing evidence that chronic infection with some viruses, bacteria, and parasites represents a major carcinogenic factor for humans, in particular in developing countries. A number of infectious agents have been evaluated within the IARC monograph programme (Table 5), and the evidence of a causal association has been classified as sufficient for several of them. Pisani et al. (1997) provided global estimates of the number of cases of cancer attributable to biological agents. Their estimate for 1990 is 1 454 000 cases, or 15.6 per cent of total cancers. hepatitis B and hepatitis C virus related liver cancer, human papilloma virus related cervical cancer, and Helicobacter-related stomach cancer each provide approximately 30 per cent of the total. Because of the high prevalence of hepatitis B, hepatitis C, and human papilloma virus in developing countries, the estimate of the attributable risk is higher in this part of the world (20.8 per cent of total cancer versus 9.1 per cent in developed countries).
Table 5 Assessment of associations between infections and human cancer
Use of safe, effective, and cheap vaccines represents the best preventive strategy for cancers caused by viruses. Unfortunately, only hepatitis B virus infection can be effectively prevented today. Chronic infection with H. pylori can be prevented by antibiotic treatment and sanitation measures, and changes in dietary practices (e.g. avoidance of raw fish) can prevent infection by carcinogenic parasites.
Occupational and environmental agents
IARC have classified approximately 25 occupational and environmental agents, groups of agents, and mixtures as carcinogenic (IARC 1972–2001). Another 20 compounds are classified as probable carcinogens (Table 6). While some (e.g. mustard gas) represent today a historic curiosity, exposure is still widespread for carcinogens such as asbestos, coal tar, arsenic, and silica. Estimates of the global burden of occupational and environmental cancer result in figures of the order of 4 to 5 per cent (Doll and Peto 1981; Harvard Center for Cancer Prevention 1996). It should be stressed, however, that these cancers concentrate among exposed subjects (mainly male blue-collar workers), among whom they may represent up to 20 per cent of total cancers (Boffetta et al. 1995). Furthermore, unlike lifestyle factors, exposure is involuntary and can largely be avoided. In fact, reduction of exposure to occupational and environmental carcinogens has taken place in industrialized countries during recent decades. Efforts should be made to avoid exposure also in developing countries.
Table 6 Occupational agents, classified by the IARC monographs programme as carcinogenic to humans
The available evidence suggests, in most populations, a small role of purely environmental sources of exposure to carcinogens (air, water, soil pollution). Global estimates are of the order of 1 per cent or less of total cancers. This is in contrast with public perception, which often identifies environmental pollution as a major cause of human cancer. It should be stressed, however, that in selected areas (e.g. residence near asbestos processing plants or in areas with drinking water contaminated by arsenic), environmental exposure to carcinogens may represent an important cancer hazard.
The evidence of a carcinogenic effect of reproductive factors is strongest for breast cancer: early age at menarche, late age at first pregnancy, and late age at menopause are all associated with a 1.5- to twofold increased risk (Harvard Center for Cancer Prevention 1996). In addition, nulliparity increases the risk of ovarian and possibly endometrial cancer. The evidence regarding other types of cancer is inadequate. No detailed estimates are available of the contribution of reproductive factors to the global burden of cancer. Some authors have, however, proposed figures of the order of 3 per cent (Harvard Center for Cancer Prevention 1996).
Perinatal and growth factors
Excess energy intake early in life is possibly associated with breast and colon cancer. The role of attained height, growth factors, and other factors such as insulin resistance or sensitivity in this association is unclear. In addition, high birth weight is possibly associated with an increased risk of breast and prostate cancer. The implications of these findings for preventive strategies will be clarified by a more detailed understanding of the underlying carcinogenic mechanisms (Harvard Center for Cancer Prevention 1996).
Ionizing and non-ionizing radiation
Ionizing radiation causes acute lymphoblastic leukaemia, acute myeloid leukaemia, chronic myeloid leukaemia, and breast, lung, and thyroid cancer. Bone, rectal, and brain cancer may develop following prolonged therapeutic exposure. There is evidence of a linear dose–response relationship between radiation dose and cancer risk. However, levels at which people are commonly exposed to man-made radiation in most countries carry little risk and the main exposure is to natural radiation, including indoor radon. The estimates of the global contribution of ionizing radiation to human cancer are in the range of 1 to 3 per cent (Harvard Center for Cancer Prevention 1996).
Solar (ultraviolet) radiation is carcinogenic to the skin, and it might increase the risk of other neoplasms such as non-Hodgkin’s lymphoma because of immunosuppression. Over 90 per cent of skin neoplasms are attributable to sunlight; because of the low fatality of non-melanocytic skin cancer, solar radiation is responsible for only 1 to 2 per cent of total cancer deaths (Harvard Center for Cancer Prevention 1996). Avoidance of sun exposure, in particular during childhood, is an important cancer preventive behaviour. The evidence of a carcinogenic effect of other types of non-ionizing radiation, in particular electric and magnetic fields, is inconclusive.
Medical procedures and drugs
The drugs that may cause or prevent cancer fall into several groups. Many cancer chemotherapy drugs are active on the DNA, in order to block the replication of cancer cells. This, however, might result in damage to normal cells, including cancer transformation. The main neoplasm associated with hemotherapy treatment is leukaemia, although also the risk of solid tumours might be increased. A second group of carcinogenic drugs includes immunosuppressive agents, notably used in transplanted patients. Non-Hodgkin’s lymphoma is the main neoplasm caused by these drugs. Replacement oestrogen therapy used in the past to treat menopausal symptoms increases the risk of endometrial cancer. Current hormone therapies include a progestin, which inhibits endometrial growth. It is unclear whether combined replacement therapies are associated with an increased risk of cancer. Oral contraceptives reduce the risk of ovarian and endometrial cancer, although it probably increases the risk of liver cancer. Phenacetin-containing analgesics increase the risk of cancer of the renal pelvis.
No precise estimates are available for the global contribution of drug use to human cancer. It is unlikely, however, that they represent more than 1 per cent in developed countries. Furthermore, the benefits of therapies are usually much greater than the potential cancer risk.
Use of ionizing radiation for diagnostic purposes is likely to carry a small risk of cancer, which has been demonstrated only for childhood leukaemia following intrauterine exposure. Radiotherapy increases the risk of cancer in the irradiated organs. There is no clear evidence of an increased cancer risk following other medical procedures, including mammography and surgical implants.
A number of inherited mutations of a high-penetrance cancer gene increases dramatically the risk of some neoplasms (see sections on specific neoplasms). However, these are rare conditions in most populations and the number of cases attributable to them is rather small.
Familial aggregation has been shown for most types of cancers, in non-carriers of known high-penetrance genes. This is notably the case for cancers of the breast, colon, prostate, and lung. The relative risk is of the order of 2 to 4, and is higher for cases diagnosed at a young age. Although some of the aggregation can be explained by shared risk factors among family members, it is plausible that a true genetic component exists for most human cancers. This takes the form of an increased susceptibility to exogenous carcinogens. The knowledge of low-penetrance genes responsible for such susceptibility is still very limited, although research has currently focused on genes encoding for metabolic enzymes, DNA repair, and hormone receptors. Current estimates of the global contribution of genetic factors to human cancer are in the range of 5 to 10 per cent, of which less than 1 per cent is attributable to high-penetrance genes.
Neoplasms are a group of diverse diseases with complex distributions in human populations and with different aetiological factors. Current knowledge of the causes of human neoplasms and the development of control strategies have led to the elaboration of lists of recommendations for their prevention (Table 7). A comprehensive strategy for cancer control might lead to the avoidance of a sizeable proportion, possibly up to one-half of human cancers (Sikora 1999). However, such a strategy would imply major cultural, societal, and economic changes. More modest objectives for cancer prevention should focus on the neoplasms and the exposures that are prevalent in any given population. For example, vaccination of children against hepatitis B virus is likely to be the most cost-effective cancer prevention action in many countries of Africa and Asia.
Table 7 European code against cancer
Neoplasms will continue to be a major source of human disease and death. Considerable efforts are made in the public and private domains to develop effective therapeutic approaches. Even if major discoveries in the clinical management of cancer patients are accomplished in the near future, the changes will mainly affect the affluent parts of the world. Prevention of the known causes of cancer remains the most promising approach in reducing the consequences of cancer, in particular in countries with limited resources. Control of tobacco smoking and of smokeless tobacco products, increased consumption of fruits and vegetables, avoidance of obesity, moderation in alcohol intake, increased physical activity, avoidance of exposure to solar radiation, and control of known and suspected occupational carcinogens are the main approaches to reduce the burden of human neoplasms.
ACS (American Cancer Society) (2000). Cancer facts and figures 2000. American Cancer Society, Atlanta, GA.
Albertini, R.J. and Hayes, R.B. (1997). Somatic cell mutations in cancer epidemiology. In Applicaton of biomarkers in cancer epidemiology. IARC Scientific Publications No. 142 (ed. P. Toniolo, P. Boffetta, D.E.G. Shuker, N. Rothman, B. Hulka, and N. Pearce), pp. 159–84. International Agency for Research on Cancer, Lyon.
Anderson, K.E., Potter, J.D., and Mack, T.M. (1996). Pancreatic cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr), (2nd edn), pp. 725–71. Oxford University Press, New York.
Armstrong, B.K. and English, D.R. (1996). Cutaneous malignant melanoma. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1282–312. Oxford University Press, New York.
Austin, D.F. and Reynolds, P. (1996). Laryngeal cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni, Jr.) (2nd edn), pp. 619–36. Oxford University Press, New York.
Berrino, F., Capocaccia, R., Estève, J., et al. (ed.) (1999). Survival of cancer patients in Europe: the EUROCARE-2 study. IARC Scientific Publications No. 151. International Agency for Research on Cancer, Lyon.
Blot, W.J. (1997). Vitamin/mineral supplementation and cancer risk: international chemoprevention trials. Proceedings of the Society of Experimental Biology and Medicine, 216, 291–6.
Blot, W.J. and Fraumeni, J.F., Jr (1996). Cancers of the lung and pleura. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 637–65. Oxford University Press, New York.
Blot, W.J., McLaughlin, J.K., Winn, D.M., et al. (1988). Smoking and drinking in relation to oral and pharyngeal cancer. Cancer Research, 48, 3282–7.
Blot, W.J., McLaughlin, J.K., Devesa, S.S., and Fraumeni, J.F. Jr (1996). Cancers of the oral cavity and pharynx. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 666–80. Oxford University Press, New York.
Boffetta, P., Kogevinas, M., Simonato, L., Wilbourn, J., and Saracci, R. (1995). Current perspectives on occupational cancer risks. International Journal of Occupational and Environmental Health, 1, 315–25.
Cardis, E., Zeise, L., Schwarz, N., and Moolgavkar, S. (1999). Review of specific examples of QEP. In Quantitative estimation and prediction of human cancer risks (ed. S. Moolgavkar, D. Krewski, L. Zeise, E. Cardis, and H. Møller), pp. 239–304. IARC Scientific Publications No. 131. International Agency for Research on Cancer, Lyon.
Coleman, M., Estève, J., Damiecki, P., Arslan, A., and Renard, H. (1993). Trends in cancer incidence and mortality. IARC Scientific Publications No. 121. International Agency for Research on Cancer, Lyon.
Daling, J.R. and Sherman, K.J. (1996). Cancers of the vulva and vagina. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn) pp. 1117–29. Oxford University Press, New York.
Doll, R. and Peto, R. (1981). The causes of cancer. Oxford University Press.
Ferlay, J., Parkin, D.M., and Pisani, P. (1998). Cancer incidence and mortality worldwide (Globocan I software). International Agency for Research on Cancer, Lyon.
Ferlay, J., Bray, F., Pisani, P., and Parkin, D.M. (2001). Globocan 2000: cancer incidence, mortality and prevalence worldwide. IARC Cancer Base No. 5. International Agency for Research on Cancer, Lyon.
Grady, D. and Ernster, V.L. (1996). Endometrial cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1058–89. Oxford University Press, New York.
Hackshaw, A.K., Law, M.R., and Wald, N.J. (1997). The accumulated evidence on lung cancer and environmental tobacco smoke. British Medical Journal, 315, 980–8.
Hammond, E.C., Selikoff, I.J., and Seidman, H. (1979). Asbestos exposure, cigarette smoking and death rates. Annals of the New York Academy of Science, 330, 473–90.
Harvard Center for Cancer Prevention (1996). Harvard report on cancer prevention. Vol. 1: Causes of human cancer. Cancer Causes Control, 7, S3–58.
Henderson, B.E., Pike, M.C., Bernstein, L., and Ross, R.K. (1996). Breast cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1022–39. Oxford University Press, New York.
IARC (International Agency for Research on Cancer) (1972–2001). IARC monographs on the evaluation of carcinogenic risks to humans, Vols 1–78. IARC, Lyon.
IARC (International Agency for Research on Cancer) (1987). Overall evaluations of carcinogenicity: an updating of IARC Monographs Volumes 1 to 42. IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, Supplement 7. IARC, Lyon.
IARC (International Agency for Research on Cancer) (1988). Alcohol drinking. IARc monographs on the evaluation of carcinogenic risks to humans, Vol. 44. IARC, Lyon.
IARC (International Agency for Research on Cancer) (1994a). Hepatitis viruses. IARC monographs on the evaluation of carcinogenic risks to humans, Vol. 59. IARC, Lyon.
IARC (International Agency for Research on Cancer) (1994b). Schistosomes, liver flukes and Helicobacter pylori. IARC monographs on the evaluation of carcinogenic risks to humans, Vol. 61. IARC, Lyon.
IARC (International Agency for Research on Cancer) (1995). Human papilloma viruses. IARC monographs on the evaluation of carcinogenic risks to humans, Vol. 64. IARC, Lyon.
IARC (International Agency for Research on Cancer) (1996). Human immunodeficiency viruses and human T-cell lymphotropic viruses. IARC monographs on the evaluation of carcinogenic risks to humans, Vol. 67. IARC, Lyon.
IARC (International Agency for Research on Cancer) (1997a). Epstein–Barr virus and Kaposi’s sarcoma herpesvirus/human herpesvirus 8. IARC monographs on the evaluation of carcinogenic risks to humans, Vol. 70. IARC, Lyon.
IARC (International Agency for Research on Cancer) (1997b). Non-steroidal anti-inflammatory drugs. IARC handbooks of cancer prevention, Vol. 1. IARC, Lyon.
IARC (International Agency for Research on Cancer) (1998a). Carotenoids. IARC handbooks of cancer prevention, Vol. 2. IARC, Lyon.
IARC (International Agency for Research on Cancer) (1998b). Vitamin A. IARC handbooks of cancer prevention, Vol. 3. IARC, Lyon.
IARC (International Agency for Research on Cancer) (1999). Surgical implants and other foreign bodies. IARC monographs on the evaluation of carcinogenic risks to humans, Vol. 74. IARC, Lyon.
Joint IARC/IPCS/CEC Working Group (1983). Approaches to classifying chemical carcinogens according to mechanism of action. International Agency for Research on Cancer (IARC) Internal Technical Report No. 83/001. IARC, Lyon.
Kosary, C.L., Ries, L.A.G., Miller, B.A., Hankey, B.F., Harras, A., and Edwards, B.K. (ed.) (1995). SEER cancer statistics review, 1973–1992: tables and graphs. NIH Publication No. 96–2789. National Cancer Institute, Bethesda, MD.
London, W.T. and McGlynn, K.A. (1996). Liver cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 772–93. Oxford University Press, New York.
Lubin, J.H. and Boice, J.D., Jr (1997). Lung cancer risk from residential radon: meta-analysis of eight epidemiologic studies. Journal of the National Cancer Institute, 89, 49–57.
McLaughlin, J.K., Blot, W.J., Devesa, S.S., and Fraumeni, J.F., Jr (1996). Renal cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1142–55. Oxford University Press, New York.
Miller, A.B. (1996). Fundamental issues in screening for cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1433–52. Oxford University Press, New York.
Miller, R.W., Boice, J.D., Jr, and Curtis, R.E. (1996). Bone cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 971–83. Oxford University Press, New York.
Moss, S.M. (1999). Breast cancer. In Cancer screening: theory and practice (ed. B.S. Kramer, J.K. Gohagan, and P.C. Prorok), pp. 143–70. Marcel Dekker, New York.
Mueller, N.E. (1996). Hodgkin’s disease. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 893–919. Oxford University Press, New York.
Muñoz, N. and Day, N.E. (1996). Esophageal cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 681–706. Oxford University Press, New York.
Murray, C.J.L. and Lopez, A.D. (ed.) (1996). The global burden of disease. Harvard School of Public Health, Boston, MA.
National Research Council (1988). Committee on the Biological Effects of Ionizing Radiations. Health risks of radon and other internally deposited alpha-emitters. National Academy of Sciences, Washington, DC.
Parkin, D.M., Pisani, P., Lopez, A.D., and Masuyer, E. (1994). At least one in seven cases of cancer is caused by smoking: global estimates for 1985. International Journal of Cancer, 59, 494–504.
Parkin, D.M., Whelan, S.L., Ferlay, J., Raymond, L., and Young, J. (ed.) (1997). Cancer incidence in five continents, Vol. VII. International Agency for Research on Cancer (IARC) Scientific Publications No. 143. IARC, Lyon.
Peto, R. (1977). Epidemiology, multistage models and short-term mutagenicity tests. In Origins of human cancer (ed. H.H. Hiatt, J.D. Watson, and J.A. Winsten), pp. 1403–28. Cold Spring Harbor Laboratory, Cold Spring Harbor, NY.
Peto, R., Chen, Z.M., and Boreham, J. (1999). Tobacco—the growing epidemic. Nature Medicine, 5, 15–17.
Pisani, P., Parkin, D.M., Muñoz, N., and Ferlay, J. (1997). Cancer and infection: estimates of the attributable fraction in 1990. Cancer Epidemiology, Biomarkers and Prevention, 6, 387–400.
Preston-Martin, S. and Mack, W.J. (1996). Neoplasms of the nervous system. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1231–81. Oxford University Press, New York.
Ron, E. (1996). Thyroid cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1000–21. Oxford University Press, New York.
Ron, E., Lubin, J.H., Shore, R.E., et al. (1995). Thyroid cancer after exposure to external radiation: a pooled analysis of seven studies. Radiation Research, 141, 259–77.
Ross, R.K. and Schottenfeld, D. (1996). Prostate cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1180–206. Oxford University Press, New York.
Roush, G.C. (1996). Cancers of the nasal cavity and paranasal sinuses. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 587–602. Oxford University Press, New York.
Sankaranarayanan, R., Black, R.J., and Parkin, D.M. (ed.) (1998). Cancer survival in developing countries. International Agency for Research on Cancer Scientific Publications No. 145. IARC, Lyon.
Sankaranarayanan, R., Mathew, B., Jacob, B.J., et al. (2000). Early findings from a community-based, cluster-randomized, controlled oral cancer screening trial in Kerala, India. Cancer, 88, 664–73.
Scherr, P.A. and Mueller, N.E. (1996). Non-Hodgkin’s lymphomas. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 920–454. Oxford University Press, New York.
Schiffman, M.H., Brinton, L.A., Devesa, S.S., and Fraumeni, J.F., Jr (1996). Cervical cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1090–116. Oxford University Press, New York.
Schottenfeld, D. (1996). Testicular cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1207–19. Oxford University Press, New York.
Schottenfeld, D. and Fraumeni, J.F., Jr (ed.) (1996). Cancer epidemiology and prevention (2nd edn). Oxford University Press, New York.
Schottenfeld, D. and Islam, S.S. (1996). Cancers of the small intestine. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 806–12. Oxford University Press, New York.
Schottenfeld, D. and Winawer, S.J. (1996). Cancers of the large intestine. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 813–40. Oxford University Press, New York.
Scotto, J., Fears, T.R., Kraemer, K.H., and Fraumeni, J.F., Jr (1996). Nonmelanoma skin cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1313–30. Oxford University Press, New York.
Sikora, K. (1999). Developing a global strategy for cancer. European Journal of Cancer, 35, 1870–7.
Silverman, D.T., Morrison, A.S., and Devesa, S.S. (1996). Bladder cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1156–79. Oxford University Press, New York.
Smith, K.R. and Liu, Y. (1994). Indoor air pollution in developing countries. In Epidemiology of lung cancer. Lung biology in health disease (ed. J.M. Samet), Vol. 74, pp. 151–84. Marcel Dekker, New York.
Speizer, F.E. and Samet, J.M. (1994). Air pollution and lung cancer. In Epidemiology of lung cancer. Lung biology in health disease (ed. J.M. Samet), Vol. 74, pp. 131–50. Marcel Dekker, New York.
Spitz, M.R., McPherson, R.S., Jiang, H., et al. (1997). Correlates of mutagen sensitivity in patients with upper aerodigestive tract cancer. Cancer Epidemiology, Biomarkers and Prevention, 6, 687–92.
Tuyns, A.J., Péquignot, G., and Jensen, O.M. (1977). Le cancer de l’oesophage en Ille-et-Vilaine en fonction des niveaux de consommation d’alcool et de tabac: des risques qui se multiplient. Bulletin de Cancer, 64, 45–60.
USDHHS (United States Department of Health and Human Services) (1982). The health consequences of smoking—cancer. USDHHS, Public Health Service, Office on Smoking and Health, Rockville, MD.
WCRF (World Cancer Research Fund) (1997). Food, nutrition and the prevention of cancer: a global perspective. WCRF and American Institute for Cancer Research, Washington,DC.
Weiss, N.S., Cook, L.S., Farrow, D.C., and Rosenblatt, K.A. (1996). Ovarian cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1040–57. Oxford University Press, New York.
WHO (World Health Organization) (1990). International Classification of Diseases for Oncology (ICD-O) (2nd edn). WHO, Geneva.
Wideroff, L. and Schottenfeld, D. (1996). Penile cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 1220–30. Oxford University Press, New York.
Wild, C.P. and Pisani, P. (1997). Carcinogen-DNA and carcinogen-protein adducts in molecular epidemiology. In Applicaton of biomarkers in cancer epidemiology (ed. P. Toniolo, P. Boffetta, D.E.G. Shuker, N. Rothman, B. Hulka, and N. Pearce), pp. 143–58. International Agency for Research on Cancer (IARC) Scientific Publications No. 142. IARC, Lyon.
Winawer, S.J., Fletcher, R.H., Miller, L., et al. (1997). Colorectal cancer screening: clinical guidelines and rationale. Gastroenterology, 112, 594–642.
Yang, P., Schwartz, A.G., McAllister, A.E., Swanson, G.M., and Aston, C.E. (1999). Lung cancer risk in families of nonsmoking probands: heterogeneity by age at diagnosis. Genetics and Epidemiology, 17, 253–73.
Yu, M.C. and Henderson, B.E. (1996). Nasopharyngeal cancer. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 603–18. Oxford University Press, New York.
Zahm, S.H., Tucker, M.A., and Fraumeni, J.F., Jr (1996). Soft tissue sarcomas. In Cancer epidemiology and prevention (ed. D. Schottenfeld and J.F. Fraumeni Jr) (2nd edn), pp. 984–99. Oxford University Press, New York.