Epidemiology of Cancer


The epidemiology of cancer is the investigation of the incidence and causes of the disease in people under different conditions of life. Such investigations have generally been the way in which reliable evidence about causal agents for cancer, and the magnitude of the risks from these agents, have been found. They have shown that any type of cancer that is common in one population is rare in some other, and that the differences between populations are mostly not genetic, but rather the consequences of behaviours and circumstances of life. In principle, cancers are therefore largely preventable.

The range of incidence rates between geographical and ethnic groups is more than 10-fold for each of the common cancers, and for some cancers is more than 100-fold. Large changes in rates of many tumours can occur in migrants compared with rates in their homeland, and large changes have occurred in rates within populations over time, indicating the scope for prevention.

The causes of cancer

These can be divided into nature (biological factors), nurture (environment and behaviours), and chance.

Biological factors—important biological factors are genetic susceptibility, age, and sex.

Tobacco smoking—this is the most important extrinsic factor causing cancer in developed countries, and is a major cause of cancers of the mouth, pharynx (other than nasopharynx), oesophagus, larynx, lung, pancreas, renal pelvis, and bladder (and it also causes a proportion of several other types of cancer). In 2005, smoking is estimated to have caused 28% of all fatal cancers in the United Kingdom.

Other extrinsic causes—there are many, including (1) alcohol—the cause of at least six types of cancer, including liver, various upper aerodigestive sites, and breast; (2) ionizing radiation—can cause cancer in most tissues; in the United Kingdom the main sources of exposure are natural sources including radon, and medical uses; (3) ultraviolet radiation—causes skin cancer; (4) infection, principally viral, but also bacterial and parasitic—a major cause of cancer of several sites, especially in developing countries; (5) immunosuppression—patients with persistent immunosuppression from therapeutic, infective, or genetic causes have raised risks of certain cancers, notably non-Hodgkin’s lymphoma; (6) chemotherapeutic agents—about 20 of these, used for treatment of specific diseases and including several that are used to treat cancer, have been shown themselves to cause cancer, of different anatomical sites according to the agent; (7) other drugs—hormone replacement therapy and oral contraceptives, both widely used in the general population, affect the risk of certain female reproductive-related malignancies, increasing risk for some cancers but (for combined steroid contraceptives) decreasing it for others; (8) occupation—numerous occupational groups have been found to be at raised risk of cancer, mainly of the respiratory tract, especially the lung; (9) air, water, and food pollution—these are probably responsible for a small percentage of cancers in Western countries; (10) diet—this may well have an effect on the aetiology of a substantial proportion of cancers, but there is considerable uncertainty on the figure and specific dietary associations are largely unknown.

Other factors of particular note are (1) menstrual and reproductive history, also certain hormonal drugs—these affect the risks of breast, endometrial, and ovarian cancers in women; (2) obesity—relates to increased risks of breast, endometrial, colonic, kidney, and possibly other cancers; (3) physical inactivity—relates to increased risk of breast, endometrial, colonic, and possibly other cancers.

Epidemiology and aetiology of particular cancers

The most common cancers worldwide are those of the lung, breast, and colorectum, and the most common causes of cancer death are lung, stomach, and liver cancers. Descriptive and aetiological epidemiological information is given about 33 types of cancer in this chapter.

Lung cancer—the major cause is smoking tobacco, particularly cigarettes. Lung cancer became epidemic in men in Western countries during the mid 20th century, with rates rising later in women, and in Western countries there have been considerable decreases in men in recent years. In developing countries, however, the epidemic has arrived later, with rising rates to be expected in future years as a consequence of current smoking levels. Occupational causes of lung cancer include exposures to asbestos, polycyclic hydrocarbons, and radon. Air pollution in towns may have been a factor, largely in smokers, and radon in indoor air contributes to a small percentage of cases.

Breast cancer—the most common cancer worldwide in women. It has greatest incidence in Western countries, where rates have tended to increase slowly over decades; rates have generally been much lower in Asia and Africa. Hormonal and reproductive factors are important to risk: early menopause, late menarche, nulliparity, and older age at first full-term pregnancy all increase risk, as do postmenopausal hormone replacement therapy and combined oral contraceptives, while tamoxifen treatment of unilateral breast cancer decreases risk in the unaffected breast. There is also raised risk of breast cancer in relation to a history of benign breast disease, alcohol consumption, lack of physical exercise, postmenopausal obesity, taller height, and ionizing radiation exposure at young ages, as well as genetic predisposition.


All cancers have certain pathological and clinical characteristics in common, but those arising in different organs often have very different causes. The epidemiology of cancer, by which is meant the study of the incidence and causes of the disease in people under different conditions of life, is, therefore, the epidemiology of specific types of cancer, usually, but not always, defined as cancers of specific organs. In this sense, the subject has a history dating back nearly 300 years to Ramazzini’s observation that cancer of the breast occurred more often in nuns than in other women of similar age and to Pott’s observation, over 200 years ago, that scrotal cancer in young men occurred characteristically in chimney sweeps. The high risk in nuns (which largely reflected the protective effect of multiple pregnancies in the general population) helped the realization that hormonal factors can substantially affect the incidence of several types of cancer, while the latter led to the recognition that the combustion products of coal to which sweeps had been exposed could cause cancer on any part of the skin with which they came into repeated contact and to the isolation of the first specific chemical carcinogen. Many other similar observations were made over the next 150 years, mostly as a result of the acumen of individual doctors who noticed clusters of cases of a particular type of cancer occurring in patients with a similar occupational or cultural background. Lip and tongue cancers were found in pipe smokers, bladder cancer in certain aniline dye workers, buccal cancer in those who habitually chewed mixtures of tobacco and betel in India, lung cancer in miners of particular ores (who, it was subsequently realized, were heavily exposed to radon and its daughter products), and skin cancer in the early radiologists and radiographers who were heavily exposed to X-rays and in farmers and seamen heavily exposed to sunlight. Gradually, however, clinical anecdotes were replaced by statistics as the epidemiological methods that are described below began to be applied to the study of cancer and other noninfectious diseases. As a result, many other causes were identified with sufficient certainty to justify preventive action and data were obtained to suggest hypotheses that could be tested in the laboratory.

Preventability of cancer

Perhaps the most important result of such observations has been the realization that any type of cancer that is common in one population is rare in some other, and that the differences between populations are mostly not genetic. Hence, where they are common these cancers occur, in large part, as a result of the way people behave and the circumstances in which they live and they are, therefore, at least in principle, preventable. This does not mean that we can at present envisage a society in which any of the common cancers are completely eliminated (although this may prove to be possible when we understand more clearly the mechanisms by which they are produced). What it does mean is that we can envisage a society in which the age-specific risk of developing any type of cancer is low.

Differences in incidence between communities

Variation in incidence of cancer between different ethnic and geographical groups around the world can be ascertained from data provided by population-based cancer registries. Table 1 below shows, for selected types of cancer, the range of variation recorded by cancer registries that have produced data sufficiently reliable for the purpose of international comparison, or, in one instance, the range determined by special surveys. The age ranges are constrained to exclude the oldest ages, at which the records of the incidence of the disease are least reliable. Types of cancer have been included if they are common enough somewhere to have a cumulative incidence among men or women of at least 2% by 75 years of age. The range of variation is never less than 13-fold and is sometimes more than 100-fold. Despite the selection of reasonably reliable registries, some of this tabulated variation may still be an artefact, due to different standards of medical service, case registration, and population enumeration. In many cases, however, the true ranges will be greater, because much of the world is not covered by reliable cancer registries, and because the data generally refer to cancers of whole organs and do not distinguish between different histological types or different locations within an organ, for which greater variation may apply.

The variation in incidence is not limited to the common cancers. Burkitt’s lymphoma, for example, never affects more than 1 in 1000 of the population, but it is at least 100 times as common among children in parts of Uganda as it is in Europe and North America; while Kaposi’s sarcoma, which was extremely rare in most of the world until the advent of AIDS, was so common in children and young adults in parts of Central Africa, even before 1970, that it accounted for 10% of all tumours seen in one of the African hospitals surveyed by Cook and Burkitt. Some few cancers occur with approximately the same frequency in all communities, but all are relatively uncommon. Acute myeloid leukaemia at 15 to 25 years of age is an example; nephroblastoma is another, except that it appears to be only half as common in Japan as elsewhere.

The figures above refer to cancer incidence in communities defined by geography, but substantial differences are found between communities defined in other ways such as by ethnic origin, religion, or socioeconomic status. Jewish women, for example, have a low incidence of cervical cancer irrespective of the country in which they live, and Mormons and Seventh Day Adventists living in the United States of America have low incidence of cancers of the respiratory, gastrointestinal, and genital systems.

Table 1 Range of incidence rates of common cancers (men, unless specified otherwise)
Site of origin of cancera High-incidence areab,c Cumulative incidence (%) in high-incidence aread Low-incidence areac Ratio of cumulative rates in high- and low-incidence arease
Nonmelanoma skin Australiaf (Queensland) >20 Several nonwhite populations >200
Prostate US (Detroit, black) 28 China (Tianjin) 240
Oesophagus China (Cixian) 23 Singapore, Malay 550
Stomach China (Changle) 20 Thailand (Songkhla) 100
Lung US (Detroit, black) 12 Mali (Bamako) 40
Breastg Uruguay (Montevideo) 13 Gambia 18
Liver Thailand (Khon Kaen) 11 Algeria (Algiers) 190
Colon Japan (Hiroshima) 7 India (Karunagappally) 70
Uterine cervixg Zimbabwe (Harare, black) 7 China (Tianjin) 40
Melanoma of skin Australia (Queensland) 5 Mali (Bamako) >300
Bladder Italy (Genoa) 5 Uganda (Kyandondo) 24
Kaposi’s sarcoma Zimbabwe (Harare, black) 5 Several >500
Rectum Japan (Hiroshima) 3 Algeria (Algiers) 36
Corpus uterig US (Connecticut, white) 3 India (Karunagappally) 24
Non-Hodgkin’s lymphoma US (San Francisco, non-Hispanic white) 3 Mali (Bamako) 13
Kidney Czech Republic 2 Vietnam (Hanoi) 40
Nasopharynx Hong Kong 2 Mali (Bamako) >200
Larynx Spain (Zaragoza) 2 China (Qidong) 40

a Sites of cancer are shown if somewhere they reach a cumulative incidence by age 75 of at least 2% in either sex.

b The geographic area of highest recorded incidence by age 75.

c Excluding very small cancer registry populations, with unstable numbers.

d By age 75 years, in the absence of other causes of death.

e By age 65 years, in the absence of other causes of death.

f Special survey.

g Women.

Few of the large differences between communities can be explained by genetic factors, apart from some of the differences in the incidence of cancer of the skin, the risk of which is much greater for whites than for blacks, and possibly also for some of those in the incidence of testis cancer, which rarely affects black populations, and in chronic lymphocytic leukaemia, which rarely affects people of Chinese or Japanese descent. Genetic factors cannot explain the differences observed on migration or with the passage of time, which are discussed below, nor can they explain the correlations observed between the national rates for particular types of cancer and aspects of the lifestyle in different countries.

Changes in incidence in migrant groups

That changes in the incidence of cancer occur on migration is certain. Numerous groups have been studied, particularly migrants from many countries to Australia, Israel, and the United States. These show, for example, that Afro-Americans experience incidence rates for internal cancers that are generally much more like those of white Americans than those of the black populations in West Africa from which most of their ancestors came, while Japanese in Hawaii have experienced rates that are much more like those of the white residents of Hawaii than those of Japanese living in Japan (Table 1 above). The ancestors of black Americans and Hawaiian Japanese will have come from many different parts of West Africa and Japan, some of which are likely to have cancer rates somewhat different from those cited in Table 2 below. Nevertheless, the contrasts are so great that there can be no serious doubt that new factors were introduced with migration.

Changes in incidence over time

Within one population there may be substantial changes in the incidence of a particular type of cancer over a period of a few decades that provide conclusive evidence of the existence of preventable factors. Changes in incidence over time may, however, be difficult to assess reliably, chiefly because it is difficult to compare the thoroughness of the selection and registration of particular types of cancer at different periods and partly because few incidence data have been collected for long enough, so we often have to fall back on changes in mortality rates even though these may be influenced by changes in treatment as well as by changes in incidence.

There are no simple rules for deciding which of the many changes in recorded cancer incidence and mortality rates are reliable indicators of real changes in incidence. Each set of data has to be assessed individually. It is relatively easy to be sure about changes in the incidence of cancer of the oesophagus, as the disease can be diagnosed without complex investigations and its occurrence is nearly always recorded, at least in middle age, because it is nearly always fatal. By contrast, it is much more difficult to be sure about changes in the incidence of basal-cell carcinomas of the skin, which—although easy to diagnose—seldom cause death and can be treated effectively outside hospital, and so often escape registration. What appears to be a change in incidence may therefore be a change only in the completeness of registration. Cancers of the pancreas, liver, and brain, and myelomatosis, in contrast, usually cause death, but even when they do they may be misdiagnosed as another disease (e.g. brain tumours in old people could frequently in the past be misdiagnosed as other neurological conditions), so that an increased incidence or mortality rate may be wholly or partly due to improvements in diagnosis or in the availability of medical services. Such changes are particularly likely to affect the rates recorded for people over 65 years of age, as many old people who were terminally ill used not to be intensively investigated.

Table 2 Comparisons of cancer incidence rates in migrantsa and residents in homelands and adopted countries (men, unless otherwise specified), mid 1990s. Cumulative rate to age 65, per 1000 persons
    Hawaii West Africab USA
Japanb Japanese Whites Blacks Whites
Oesophagus 6.5 3.0 1.7 1.9 7.4 2.6
Stomach 32.9 7.4 3.6 7.3 6.6 3.1
Colon 14.8 17.2 14.0 0.8 16.3 11.0
Rectum 10.0 11.6 6.6 0.9 6.8 6.4
Liver 17.3 3.5 2.6 32.1 4.9 1.9
Pancreas 4.4 4.0 3.9 1.3 7.0 3.4
Larynx 1.7 1.8 5.2 0.4 6.6 3.2
Lung 15.2 13.4 21.7 1.9 49.5 25.8
Breastc 24.2 61.7 72.2 10.5 59.8 65.1
Uterine cervixc 4.7 3.4 5.7 27.5 7.5 5.3
Corpus uteric 3.0 13.1 9.9 1.9 7.6 12.4
Ovaryc 4.7 7.4 10.2 1.4 5.5 9.1
Prostate 2.0 17.6 40.6 2.2 91.2 47.4
Testis 1.0 1.7 5.1 0.2 0.8 4.1
Non-Hodgkin’s lymphoma 3.4 4.3 10.4 1.4 11.0 10.1

a ‘Migrant’ rates are based on ethnicity, and hence include all generations of migrant.

b Average of rates in two regions.

c Women.

Despite these difficulties, some of the decreases and increases in the recorded rates of particular types of cancer have been so gross that there must have been real changes in their incidence. Examples include the increase in lung cancer throughout most of the world (and its recent large decrease in men in the United Kingdom), the increase in mesothelioma of the pleura in men in industrialized countries, the decrease in cancer of the tongue in the United Kingdom, and the decrease in cancers of the uterine cervix and stomach throughout western Europe, North America, and Australasia.

Identification of causes

More-specific evidence of the preventability of cancer, and of measures to enable this, has come from identification of agents and circumstances that cause the disease. In general, reliable evidence of causality (and particularly of the magnitude of any risks) has come from epidemiology and not from laboratory experiments, although the latter can often provide reinforcement of epidemiological findings and understanding of the mechanisms of cancer causation. Reliable epidemiological evidence does not require randomized trials within particular populations, but it does require the study of different individuals within populations and not just the comparison of incidence rates between populations. Nonrandomized epidemiological studies of individuals have often yielded proof of causation beyond reasonable doubt (like that required to convict in a court of law), and has been the decisive evidence of aetiology for almost all proven carcinogens. Action based on such evidence has, moreover, often been followed by the desired result—e.g. a reduction in the incidence of bladder cancer in the chemical industry on stopping the manufacture and use of 2-naphthylamine and, on a national scale, the reduction in the incidence of lung cancer in men in the United Kingdom following the decrease in smoking since the mid 20th century.

The causes of cancer can, briefly, be divided into nature (biological factors), nurture (behaviour and environment), and chance.

Biological factors

Genetic susceptibility

Various genetic factors are known that affect cancer risks to different extents. Very large risks are seen in patients with certain rare cancer-associated genetic syndromes in which bearers of one gene (if the condition is dominant) or two (if recessive) almost invariably develop a particular type of cancer. Examples include the dominant genes for polyposis coli that lead to cancer of the large bowel, and the recessive genes for retinoblastoma and xeroderma pigmentosum that lead (in the latter case) to squamous carcinoma and (less commonly) melanoma of the skin. Similar evidence has shown that other genetic syndromes frequently, but not invariably, lead to cancer, such as von Recklinghausen’s neurofibromatosis leading to fibrosarcoma, the Peutz–Jeghers syndrome leading to carcinoma of the small bowel, the Wiskott–Aldrich syndrome leading to non-Hodgkin’s lymphoma, and ataxia telangiectasia, Bloom’s syndrome, and Fanconi’s anaemia leading to leukaemia. Very high cancer risks are also present in individuals with various cancer susceptibility genes—e.g. raised risks of breast and ovarian cancers in patients with mutations in BRCA1 and BRCA2 genes. The recognition of these genes is important to the individual, as it may provide an opportunity for prophylactic surgery, or enable the diagnosis of malignancy to be made at an early stage when treatment is more likely to be effective, or (rarely) enable precautions to be taken to prevent exposure to the relevant carcinogens, as in the case of sufferers from xeroderma pigmentosum or albinism, who can be protected against sunshine. The proportion of all cancers that occur in people who are highly susceptible to cancer in this way is, however, very small, although substantially greater at young than at older ages.

Genes conveying less-raised risks of cancer are far more widespread in the population, and likely to be involved in the causation of a much larger proportion of cancers, but their discovery is only just beginning. It has been shown that many of the common types of cancer tend to cluster in families to some extent. Differences of this sort do not necessarily imply that the familial clusters are genetic in origin; they could be due to familial similarities of behaviour or environment. That socially important genetic variants exist is demonstrated by the greatly increased risk of developing basal-cell and squamous carcinomas of the sun-exposed skin in fair-skinned populations compared with dark-skinned, and there may be other genes associated with localized populations, which, for example, diminish the risk of chronic lymphatic leukaemia and myelomatosis in Chinese, Japanese, and Indians.

Discovery of genetic factors that affect particular types of cancer is unlikely to explain much of the social and geographical differences in the distribution of cancer other than skin cancer, but it should help to elucidate mechanisms and may help to focus health education and costly methods of early diagnosis on the sections of the populations that are most at risk.


Some risk of cancer occurs at every age, but the risk of developing any particular type varies with age. The most common relationship with age is a progressive increase in incidence from near zero in childhood and adolescence to a high rate in old age. This type of relationship is shown by carcinomas of the skin, lung, and gastrointestinal and urinary tracts, and by myelomatosis and chronic lymphatic leukaemia. The rate of increase is rapid, being typically proportional to the fourth, fifth, or sixth power of age in years, so that the annual incidence may be 100 or 1000 times greater above age 75 than before age 25. It is probable that this reflects the cumulative effect of processes that operate steadily throughout life, starting at around the time of birth or at young ages (e.g. for lung cancer, in adolescence). With most of these cancers, the recorded incidence may stabilize, or even decrease, in the oldest age groups, but this is partly or wholly an artefact due to incomplete investigation of the terminal illnesses of old people.

A less common pattern is a peak incidence early in life, which may be followed either by a decline virtually to zero or by a slow rise in middle and old age. Retinoblastomas and nephroblastomas occur only in childhood, with peak incidences (respectively) in the first and second years of life. Teratomas and seminomas of the testis have peak incidence rates at about 25 and 35 years of age, respectively, and later almost cease to occur, while osteogenic sarcomas have a peak incidence in adolescence and then show a slow increase with age from a lower rate in young adult life.

The remaining cancers show a variety of patterns. Carcinomas of the breast and uterine cervix of women, for example, begin to appear in young adulthood and become rapidly more common up to the menopause. After the menopause the incidence of carcinoma of the breast may remain approximately constant, or may even become slightly reduced for a few years, before increasing again with age, though at a slower rate. Carcinoma of the cervix continues to increase fairly steeply for a few years after the menopause, before showing a stable or declining rate. Hodgkin’s disease, on the other hand, appears in childhood but thereafter continues relatively evenly throughout life with only modest peaks in young adult life and at older ages, while connective tissue sarcomas become progressively more common from childhood on, but with a much slower rate of increase than is shown by the common carcinomas.

Some of these relationships with age, like that for retinoblastoma in early childhood, seem to be invariant everywhere and, as far as is known, at all times. Others vary from community to community, or from time to time. In postmenopausal women, for example, cancer of the breast becomes progressively less common with increasing age in parts of Asia, but more common in Europe, while carcinoma of the lung used to show a peak incidence at about 60 years of age in the United Kingdom, which gradually moved to older ages, as a generation that had not smoked substantial numbers of cigarettes throughout adult life was replaced by one that had, and the same process is now being repeated in many developing countries.

These various patterns provide information, either about the period of activity of the stem cells from which the cancers derive, or about the period when the main exposure to causative agents occurs and the duration of that exposure. Some of this variation has already helped to explain some of the causes of cancer, as was the case with the shift in the peak incidence of bronchial carcinoma, but much of it still awaits elucidation.


Cancer used to be more common in women than in men in many countries due to the great frequency of carcinoma of the breast and of the uterine cervix and to the rarity of bronchial carcinoma, and this is still the case in populations for which similar conditions persist, as in parts of Latin America. Elsewhere, cancer is now more common in men, among whom lung cancer often predominates. This overall male preponderance hides, however, a wide range of sex ratios for cancer of different organs. If the sites of cancer that are peculiar (or almost peculiar) to one sex are ignored, the sex ratio varies (in Britain) from a male excess of about 6 to 1 for pleural mesothelioma and carcinoma of the larynx, through many types of cancer with only a small male preponderance, to carcinoma of the thyroid, which is about twice as common in women.

For many types of cancer the sex ratio is much the same in different countries and at different times. For some, however, and particularly for cancers of the mouth, oesophagus, larynx, and bronchus, the sex ratio is extremely variable—not only between countries and at different times, but sometimes also between different ages at the same time and in the same country. The most marked variation is shown by cancer of the oesophagus, which may affect both sexes equally or be 20 times more common in men than in women. As with the various patterns of incidence with age, these different sex ratios and their variation can provide useful clues to the causation of the particular type of cancer, not all of which have yet been successfully followed up.

Delay between cause and effect

One reason why it has been difficult to recognize causes of cancer in humans is the long delay that characteristically occurs between the start of exposure to a carcinogen and the appearance of the clinical disease. This ‘latent period’, as it is commonly, but rather misleadingly, called is often several decades, although it may be as short as 1 year or as long as 60. The exact relation between the date of exposure and the date of the appearance of different cancers is still uncertain, partly because the interval is subject to random factors, partly because few cancers are induced by a single, brief exposure, and partly because there are still very few sets of quantitative data with detailed information about the dates when exposure began and ended.

When cancer is induced by short but intensive exposure to ionizing radiation, as following the explosions of the atomic bombs in Hiroshima and Nagasaki or in patients treated by radiotherapy, the excess incidence of solid tumours rises for 15 to 20 years and then may continue to rise, level off, or decline. In the case of acute leukaemia, however, a peak incidence occurs much earlier (c.5 years after irradiation) and relatively few cases appear after more than 30 years.

Short, intensive exposure to a carcinogen is, however, exceptional. The more usual situation is for sporadic or continuous exposure to a carcinogen to be prolonged for years—a decade or two in the case of occupational exposure, several decades in the case of tobacco smoking, and a lifetime in the case of ultraviolet radiation. In this situation the incidence of cancer increases progressively with the length of exposure. In the last two cases cited, the incidence appears to increase approximately in proportion to the fourth power of the duration of exposure, so that the effect after (say) 40 years is more than 10 times as great as that after 20 years, and more than 100 times as great as that after 10 years. Whether the same holds for occupational exposure is not known, but it has been shown to hold in some experiments in which chemicals were repeatedly applied to the skin of genetically similar mice and it may prove to be a general biological rule for many types of carcinoma and many carcinogens.

There is still less quantitative information about what usually happens when exposure ceases, but in the case of cigarette smoking the rapidly rising annual risk among those who continue to smoke stabilizes for one or two decades after smoking ceases before increasing again slowly. The ex-smoker consequently avoids the enormous progressive increase in risk suffered by the continuing smoker.

These delayed effects accord with the idea that the appearance of clinical cancer is the end result of a multistage process in which several mutations have to be produced in a single stem cell to turn it into the seed of a growing cancer. From the practical point of view, the important conclusions are that cancer may be very much more likely to occur after prolonged exposure to a carcinogen than after short exposure; that it is seldom likely to appear within a decade after first exposure (except in the case of leukaemia, certain hormone-related cancers, and the specific cancers of childhood); that it commonly occurs several decades after first exposure; and that some excess risk may continue to occur for decades after exposure has ceased. The exact relationship may, however, differ for different carcinogens and different types of tumour. Bladder tumours, for example, began to appear within 5 years of intensive exposure to 2-naphthylamine in the dye industry, while mesotheliomas of the pleura have seldom, if ever, appeared within 10 years of exposure to asbestos, but they continue to increase in incidence for up to 50 years after first exposure, even if the exposure was relatively brief.


There remains the influence of chance, which is commonly ignored; yet it is important for the individual as it is the reason why two animals of identical genetic constitution that have been treated in the same way do not, in general, develop cancer in the same place at precisely the same age. It reflects the element of chance that determines whether a particular series of events all occur in one particular stem cell out of the many thousands of stem cells that exist that do not give rise to a malignant clone. For any one individual the role of good or bad luck in determining the occurrence of cancer may be large (just as luck plays a substantial part in whether or not an individual driver has a traffic accident), but in a large population luck has little net effect on the incidence of cancer and only nature and nurture are important.

Avoidable factors


Tobacco is by far the most important single cause of cancer in developed countries. Chewed, it can cause cancers of the mouth and oesophagus; smoked, it is a major cause of cancers of the mouth, pharynx (other than nasopharynx), oesophagus, larynx, lung, pancreas, renal pelvis, and bladder. For these eight cancers, epidemiological evidence indicates that prolonged smoking of average numbers of cigarettes per day increases the risk 3 to 20 times. It is, however, now clear that cigarette smoking also causes a proportion of several other types of cancer, increasing the incidence up to twice that in nonsmokers: namely, cancers of the lip, nose, nasopharynx, stomach, liver, and renal body, and also myeloid leukaemia. Although the proportional increases are not large, the consistency of the findings in different countries, the evidence of dose–response relationships, the lower mortality in ex-smokers than in continuing smokers, the lack of evidence for important confounding, and the presence in the smoke of many different carcinogens provide strong grounds for believing that most or all of these observed associations are causal.

In sum, smoking is estimated to have caused 28% of all fatal cancers in the United Kingdom in 2005, down from 34% 30 years earlier. The reduction was substantial in men (down from 52% in 1975 to 39% in 1995 and 33% in 2005) but it was largely counteracted by the increase in women (from 12% in 1975 to 20% in 1995 and 22% in 2005). Comparable figures from the United States and from some other developed countries are shown in Table 3. In men, there have in the past decade been decreases in many developed countries, but in each the proportions remain substantial. In women, the proportion of cancer deaths attributed to smoking was generally low in 1975, but has subsequently increased in all developed countries and must be expected to increase further. It was, however, still small in countries such as France or Spain, where many young women now smoke but few middle-aged or elderly women have been smoking for long enough for any material effect to be produced.

Table 3  Per cent of cancer deaths attributed to smoking, 1955, 1975, 1995, and 2005, by sex: various countries


Country Male Female
1955 1975 1995 2005a 1955 1975 1995 2005a
Australia 20 40 33 29a 0 4 14 15a
Finland 38 46 37 31 1 1 5 8
France 17 34 37 35 0 0 3 7
Hungary 21 36 51 50 2 5 13 18
Spain 13 28 39 37 0 0 0 2
UK 41 52 39 33 3 12 20 22
USA 23 42 42 39 0 10 25 27

a 2005, except Australia for which the most recent year of data available from WHO was 2004.

In developing countries, the effects of smoking have only recently begun to be studied systematically and much remains unclear. Large, nationally representative studies of smoking and death have, however, been conducted in China and India. In general, women in developing countries do not smoke (although there are particular regions in China and India where they do so). In men, however, there has been a very large increase in worldwide cigarette consumption over the past few decades, the full effects of which have yet to materialize. China, with 20% of the world’s population, smokes 30% of the world’s cigarettes and by 1987 smoking was already responsible for about 20% of male cancer deaths (and 12% of all male deaths in China at ages 35–69), and this proportion is likely to at least double between the 1980s and the 2020s. In India, where many men have smoked ‘bidis’ (small, home-manufactured cigarettes) for decades, smoking now causes 32% of all male cancer deaths (and 20% of all male deaths in India at ages 35–69), partly because smoking can act as a cofactor for the production of cancers of the mouth, oesophagus, or stomach in those who habitually chew quids containing betel and tobacco. In some parts of South America and China the male lung cancer rates from smoking are already as high as in developed countries. Overall, tobacco may be causing about as many cancer deaths in developing as in developed countries, in which case it would be responsible for about 20% of cancer deaths throughout the world.


At least six types of cancer are caused in part by the consumption of alcohol. One, liver cancer, is produced mainly by the production of liver cirrhosis and is, consequently, caused mainly by heavy and prolonged consumption. Four are causally related to smoking as well as to alcohol: namely, cancers of the mouth, pharynx (other than nasopharynx), oesophagus, and larynx. The two agents act synergistically, increasing each other’s effect, so that the risk from alcohol in nonsmokers or long-term ex-smokers is very small, while that in heavy smokers is disproportionately large. The remaining type, cancer of the breast, has been shown to be related to alcohol more recently. Epidemiological cohort studies show that the risk increases progressively with the amount drunk (at least up to moderately high levels) and laboratory studies that show that alcohol increases the level of oestrogen in the blood suggest a plausible mechanism.

Cancers of the large bowel have also been associated with alcohol in many studies, but the relationship is weak and its nature uncertain: it could be due to confounding by diet.

Ionizing radiation

Ionizing radiations, of whatever type, and whether from external sources or from inhaled or ingested radionuclides, share the characteristic of having sufficient energy to damage DNA through ionization when they pass through the tissues of the body. It is not surprising, therefore, that they have been found to increase the incidence of cancer in practically every organ. The radiosensitivity of different organs varies greatly, however, and particularly large risks, relative to background rates, occur for thyroid cancer in people exposed as children, for myeloid leukaemia, and for cancers of the breast and bladder. In contrast, there is no good evidence that exposure to ionizing radiation can increase the risk of chronic lymphocytic leukaemia, Hodgkin’s disease, or testis cancer. Many exposures to ionizing radiation are specific to certain parts of the body or to certain organs or tissues, and this determines the site of the induced cancer. For example, the sites of cancers after radiotherapy depend on the sites exposed to radiation as a consequence of the treatment; Thorotrast (a radioactive contrast agent used in the mid 20th century) tends to be incorporated in the liver and bone marrow and hence to cause liver cancer and leukaemia; inhalation and ingestion of iodine-131 was the principal exposure from the Chernobyl accident to the general population in the surrounding areas, which led to an increase in thyroid cancer in children; and inhalation of the natural radioactive gas radon and its progeny gives rise to lung cancer.

Estimates of the carcinogenic effect of X-rays have been derived by following groups of people with unusual but well-documented exposures, including patients given radiotherapy, or repeatedly screened radiologically, and the survivors of the atomic bombings of Hiroshima and Nagasaki (in whom exposure was principally to γ-rays, which are high-energy X-rays). Estimates of the risk of lung cancer from the inhalation of radon and its progeny have been derived both from studies of uranium miners and from studies of indoor radon exposure carried out in the general population.

At low doses (<c.20 mGy) it seems probable that the carcinogenic effect of ionizing radiation is linearly proportional to the dose, while at higher doses the same is true for most cancers other than leukaemia, for which the risk is approximately proportional to the square of the dose. For most sites of cancer the risk is higher in people exposed in childhood than those exposed as adults, and it starts within the first 5 years and lasts for several decades. The International Commission on Radiological Protection (2007) has concluded that the lifetime risk of developing a fatal cancer is approximately 10%/Gy for X-rays (or per Sievert for other types of radiation) to the whole body if the radiation dose is moderate and given acutely, and about half that if the dose is low and spread out over time (that is 5 per 100 000 per mGy (or mSv)), with corresponding reductions if only part of the body is exposed.

It has not been possible to detect by direct observation the effect of very small exposures, including, for example, the effect of a single chest radiograph given to an adult. However, theoretical considerations and the dose–response relationship observed at larger doses both indicate that there is unlikely to be any threshold below which no effect is produced. This conclusion is reinforced by the observation that children who received doses of 10 to 20 mGy in utero (because their mothers were irradiated for diagnostic purposes while pregnant) were subject to an added risk of developing cancer in childhood of approximately 1 in 2000.

People are exposed to different amounts of radiation in different countries, depending principally on the concentration of radon in indoor air and the medical use of radiation for diagnosis. In the United Kingdom, the average indoor radon concentration is 20Bq/m3 and this is estimated to give rise to about 1000 lung cancer deaths each year. In the United States the average radon concentration is about twice that for the United Kingdom and, within the United Kingdom, it varies from one part of the country to another. Most notably, Devon and Cornwall have average indoor radon concentrations three or four times greater than the national average and there are a few houses with concentrations that are 10 or even 100 times greater. Other sources of radiation are estimated to give rise to an average annual dose of about 1.4 mSv in the United Kingdom, which, on the basis of the risk estimate recommended by the International Commission on Radiological Protection, would lead to about 4200 deaths per year in the national population of about 60 million. This gives an estimated total of 5200 deaths per year from radiation-induced cancer, or just over 3% of total cancer deaths. The major contributors are radon (20%), other natural sources (55%), and medical uses (24%). In addition, there is a contribution from radiotherapy for cancer, but less than might at first sight be implied from the collective dose, because a substantial proportion is received by people who will not survive long enough for a radiation-induced cancer to appear. Less than 1% of all radiation-induced deaths in the United Kingdom can be attributed to occupational exposure, fallout from past nuclear weapons tests, manufactured products, or radioactive waste.

Ultraviolet radiation

Photon energies in the ultraviolet (UV) range are sufficient to damage DNA and hence to cause cancer mutations. UV does not penetrate much below the skin, so that it is chiefly within the skin that it is directly carcinogenic. Within the skin, however, it is the principal cause of all types of cancer, other than Kaposi’s sarcoma. Whether it has any indirect carcinogenic effect on other tissues (notably the lymphopoietic tissue) remains uncertain. The main source of human exposure to UV radiation is sunshine.

Table 4  Viral causes of cancer


Virus Cancer
Hepatitis B Cancer of liver
Hepatitis C Cancer of liver
HPV types 16, 18, and others Cancers of cervix, vulva, vagina, penis, anus; some skin cancers
HHV type 4 (EBV) Burkitt’s lymphoma
Post transplant lymphoproliferative disease
Nasal T-cell lymphoma
Other non-Hodgkin’s lymphoma
Hodgkin’s diseaseb
Nasopharyngeal cancer
HHV type 8 (Kaposi-associated herpesvirus)
  • Kaposi’s sarcoma
  • Primary effusion lymphoma
Human T-cell leukaemia type 1 Adult T-cell leukaemia/lymphoma
HIVa Kaposi’s sarcoma
Non-Hodgkin’s lymphoma
Hodgkin’s disease
Conjunctival carcinoma

EBV, Epstein–Barr virus; HHV, human herpesvirus; HPV, human papillomavirus.

a In most cases, if not in all, by facilitating the effect of other viruses, probably via immunosuppression.

b Causal nature of observed association unproven.

Simian virus 40 (SV-40) has been suspected in the aetiology of mesothelioma and several other types of cancer, without clear proof.


Infection, principally viral, but also in some cases bacterial and parasitic, is a major cause of avoidable cancer, especially in developing countries: in sub-Saharan Africa about 40% of cancers in women and 30% in men, compared with about 10% in developed countries, are attributable to infections.

Viral infection

Viruses that are known to cause human cancers, or suspected of doing so, are listed in Table 4 above, along with the types of cancer with which they are associated. Not all infected people develop the disease. In some cases the proportion doing so is quite small, unless other factors are also present. Such cofactors include endemic malaria for Burkitt’s lymphoma, the consumption of a type of salted fish for nasopharyngeal cancer, and the consumption of aflatoxin, a metabolic product of fungal infection with Aspergillus flavus, for liver cancer. What they are for the cancers produced by the human papillomavirus is not known.

Quantitatively, chronic infection with hepatitis B virus is one of the most important causes of cancer in many parts of the world. In China, for example, liver cancer accounts for about 18% of all cancer deaths, the large majority of which are due to chronic lifelong infection with the virus. Infant vaccination against the virus is now being introduced and will protect those born in the present century, but will not provide retrospective protection for those born in the 20th century.

Bacterial infection

Only one specific bacterial infection has been closely linked with the development of cancer: Helicobacter pylori. Persistent H. pylori infection acquired early in life leads to chronic gastritis in the antrum of the stomach and increases the risk of gastric cancer two- to three-fold. Nonspecific chronic infection in the bladder may increase the risk of bladder cancer.

Parasitic infection

In parts of Africa and Asia, parasitic infection is a major cause of cancer. Infection with Schistosoma haematobium, which excretes its eggs through the bladder wall, causes a high incidence of bladder cancer in Egypt and East Africa while infection with S. japonicum, which excretes its eggs through the wall of the large bowel, is responsible for a high incidence of intestinal cancer in parts of China. Liver flukes (Clonorchis sinensis and Opisthorcis viverrini) are similarly responsible for the high incidence of cholangiosarcoma of the bile ducts in parts of South East Asia. The parasites may not cause cancer directly, but chronic infection may start a chain of events that leads to cancer in other ways, such as chronic bacterial infection and the local formation of nitrites and nitrosamines.


Patients with persistent immunosuppression, either therapeutic (notably immunosuppressive drug treatment for organ transplant patients) or as a consequence of infection (e.g. HIV) or genetic (e.g. in ataxia telangiectasia and Wiskott–Aldrich syndrome) have greatly raised risks of non-Hodgkin’s lymphoma, and often of other viral infection-related cancers including Kaposi’s sarcoma, although the pattern of malignancy and scale of risk varies according to the type of immunosuppression—Kaposi’s sarcoma, for instance, predominates after HIV, whereas non-Hodgkin’s lymphoma is the most common consequence of genetic immunodeficiency and of transplantation.

Medical drugs

Apart from ionizing radiation, some 20 agents have been used therapeutically that are known to cause cancer in humans. These are listed in Table 5 below. That so many carcinogenic agents should have been prescribed medically is not surprising when it is borne in mind that treatment often requires modification of cellular metabolism and is sometimes intended to interfere with DNA. The hazard of cancer, however, need not necessarily be a bar to the use of a drug if this risk is outweighed by the therapeutic benefits, as is commonly the case with antineoplastic agents, immunosuppressive drugs, and radiotherapy.

Some of the chemotherapeutic agents listed in Table 5 were soon abandoned, while others have continued to be used for the treatment of uncommon conditions, and the sum of the cancers that these now produce cannot amount to more than 100 or so a year in the United Kingdom.

Two of the listed drugs are, however, used extensively in the general population: hormonal replacement therapy (HRT) for postmenopausal women, and selected steroids for contraception, which increase the risk of breast cancer. Both also increase the risk of endometrial cancer, but HRT does so substantially only when given in the form of oestrogen alone and steroid contraceptives do so only in the form (now abandoned) in which oestrogen and progestogen are given sequentially. The combined steroid contraceptives currently in use can also rarely cause liver cancer and they may possibly increase the risk of cervix cancer. Combined steroid contraceptives, however, also reduce the incidence of endometrial cancer and halve the risk of ovarian cancer for many years after they have been used, and HRT and combined steroid contraceptives are associated with a reduction of some 20% in the risk of colorectal cancer, although whether this is causally related to their use remains unknown.

Table 5  Carcinogenic agents used in medical practice (other than ionizing radiations)


Agent Type of cancer
Antineoplastic agents including:  
 Busulphan Leukaemiaa
 Carmustine (BCNU) Leukaemia
 Chlorambucil Leukaemiaa
 Chlornaphazine Bladder
 Cyclophosphamide Bladder, leukaemiaa
 Lomustine (CCNU) Leukaemia
 Epipodophyllotoxins Leukaemia
 Melphalan Leukaemiaa
 MOPPb Leukaemiaa, probably lungc
 Thiotepa Leukaemia
 Treosulfan Leukaemiaa
Arsenic Skin, liver (angiosarcoma), lung
Immunosuppressive drugs:  
 Azathioprine Non-Hodgkin’s lymphoma, Kaposi’s sarcomad
 Ciclosporin Non-Hodgkin’s lymphoma, Kaposi’s sarcomad
Methoxypsoralen (plus UV radiation) Skin
Phenacetin Renal pelvis, bladder
Polycyclic hydrocarbons (coal-tar ointment) Skin
Sex hormones:  
 Unopposed oestrogens Endometrium, breast
 Transplacental diethylstilbestrol Vagina and cervix (adenocarcinoma)
 Oxymetholone (an anabolic steroid) Liver (hepatoma)
 Oral contraceptives (combined)e Breast, liver (hepatoma)
 Tamoxifene Endometrium

a Acute or nonlymphocytic.

b Combination of nitrogen mustard, vincristine, procarbazine, and prednisone.

c Lung cancer might also be caused by certain other alkylating agents or regimens.

d There have also been excesses of several other cancers in transplant patients treated with immunosuppressive drugs.

e Oral contraceptives also reduce the risk of ovarian and endometrial cancers and tamoxifen reduces the risk of contralateral breast cancer.

Other drugs that may inhibit cancer rather than cause it are the nonsteroidal analgesics, most notably aspirin, the prolonged use of which may somewhat reduce the risk of colorectal cancer and perhaps cancers of certain other sites.

Taken altogether, it seems unlikely that medically prescribed drugs can be responsible for more than 1% of all today’s fatal cancers and may, in total, reduce the risk by somewhat more.


In the years that followed Pott’s observation that chimney sweeps tended to develop cancer of the scrotum, many other groups of workers were found to suffer from specific hazards of cancer; indeed, more substances that are known to be carcinogenic to humans have been unearthed by the search for occupational hazards than by any other means. Most of these occupational cancers are in the respiratory tract, especially the lung. The hazards are listed in Table 6 below. Many of the hazards that have been recognized caused large, or at least relatively large, risks, albeit for limited populations, and it may well be that other occupational causes exist that have not yet been detected, either because the added risk is small in comparison with that due to other causes, or because only a few workers have been persistently exposed, or simply because the hazards have not been suspected and so not looked for. It must also be borne in mind that cancer in humans seldom develops until one or more decades after exposure to the carcinogen first occurs and it is, therefore, too soon to be sure whether agents that have been introduced into industry only during the last 20 years are carcinogenic or not.

Many groups of workers not listed in Table 6 have been suspected of having a special risk, but it has not been possible to decide whether the risk is real and attributable to their work. Some of these excesses may have arisen by chance alone, especially if the excess has not been confirmed in other studies, and others may be due to confounding; i.e. they may have been produced by social factors or behaviours that are associated with the occupation in question rather than by the occupation itself.

Given sufficient details and the ability to repeat the observations, it is usually possible to obtain a fairly clear idea of whether an excess incidence in an occupational group reflects an occupational hazard e.g. by seeing whether the effect is related to the length of employment, the time after first exposure, and a specific type of work within the industry. Unfortunately these details are not always available and the reasons for many of the moderate excesses of cancer that have been reported in certain industries are still uncertain.

In addition to the directly occupational cancers discussed above, workers may be indirectly exposed to carcinogens while at work—e.g. tobacco smoke from clients or colleagues.

At present it seems likely that occupational hazards account for only a few per cent of all fatal cancers in developed countries such as the United Kingdom. The three principal causes are probably asbestos dust (lung and pleural cancer), the combustion products of fossil fuels (skin and lung cancer), and ionizing radiation (a wide range of cancers).


The idea that pollution might be an important cause of cancer has been in the forefront of the minds of cancer research workers since it was realized that the incidence of lung cancer tended to be higher in towns than in the countryside and that the combustion products of coal, which used to produce a pall of smoke over all large cities in Britain, contained carcinogenic hydrocarbons. Subsequently, with the rapid expansion of the chemical industry and the discovery that some of its products are mutagenic in vitro and carcinogenic in laboratory animals, anxiety increased about the possible effects of distributing such products ubiquitously in the air we breathe, the water we drink, and the food we eat.

The effects of pollution of this sort are, however, peculiarly difficult to assess directly by epidemiological methods, as pollutants are likely to be present in most areas, the absolute risk from each is likely to be small, and there may be little difference in the extent to which individuals are exposed over a wide area. Reliance is, therefore, often placed mainly on two indirect methods: extrapolation from the effects of chronic exposure to much larger amounts in an occupational setting, and prediction of the effects on humans from laboratory tests. Both, however, (but particularly the latter) involve substantial uncertainties.

Table 6  Occupational causes of cancer


Agent Type of cancer Occupationa
Aromatic amines: Bladder Dye manufacturers
 4-Aminodiphenyl   Rubber workers
 Benzidine   Coal-gas manufacturers
 2-Naphthylamine   Some chemical workers
Arsenic Skin, lung Copper and cobalt smelters
Pesticide manufacturers
Some gold miners
Asbestos (all forms) Lung, pleura, peritoneum Asbestos miners
Asbestos textile manufacturers
Carpenters and general builders
Insulation workers
Shipyard workers
Benzene Leukaemia Workers with glues and varnishes
Berylliumb Lung Beryllium refiners and machiners
Bis-chloromethyl ether and technical-grade chloromethyl methyl ether Lung Makers of ion-exchange resins
Cadmiumb Lung Cadmium refiners
Chromiumb Lungc Manufacturers of chromates from chrome ore; pigment manufacturers
Ionizing radiations Lung Uranium and some other miners
Bone Luminizers
Leukaemia, skin Radiologists, radiographers
Mustard gas Larynx, lung Poison-gas manufacturers
Nickelb Nasal sinuses, lung Nickel refiners
Polycyclic hydrocarbons in soot, tar, oil Skin, scrotum, lung, and sometimes bladder Coal-gas manufacturers, roofers, asphalters, aluminium refiners, and many groups exposed to tars and selected oils
Silica, when crystalline as quartz or cristobalite Lung Miners, stone workers, refractory brick workers
Sulphuric acid mists (strong acid) Nasal sinuses, larynx Many industries, isopropanol manufacture, ‘steel pickling’
Ultraviolet radiation Skin (melanoma and nonmelanoma) Farmers, seamen
Vinyl chloride Liver (angiosarcoma) PVC manufacturers
? Nasal sinuses Hardwood furniture manufacturers
? Nasal sinuses Leather workers

a Typical occupations with proven hazards.

b Certain compounds or oxidation states.

c And possibly nasal sinuses.

So far as atmospheric pollution is concerned, the epidemiological picture is complicated by the personal pollution produced by tobacco smoke and the social distribution of smoking habits. Despite this complication, however, the various methods that have been discussed under lung cancer all lead to the conclusion that the pollution of the past may have contributed to the production of a few per cent of all lung cancers in Western countries, but that the levels over the last three decades (principally from the combustion of fossil fuels, but also from asbestos, dioxins, and various other materials) are unlikely to be responsible for more than a fraction of 1% of future cancers—although there may be exceptions awaiting discovery in the neighbourhood of particular factories. The greater effect of the modern type of pollution with ultra fine particles and of the intense indoor pollution with smoke that occurs in parts of China is examined later under lung cancer and of erionite in certain Turkish villages under pleural cancer.

The effect of polluted drinking water and food is more obscure. Modern analytical techniques permit the detection of chemicals at concentrations of less than 1 part per billion in both food and water and, in consequence, many have been detected that might arguably be carcinogenic, including pesticide residues and a variety of halogenated organic materials produced by the chlorination of water supplies. Relationships have been reported between the concentrations of some of these compounds in water and the mortality from cancers of the bladder and, possibly, the large intestine, in different localities, but it is extremely difficult to know what these relationships mean as there are many potentially confounding factors.

Mortality rates from cancers of the gastrointestinal and urinary tracts are, for the most part, stable or decreasing in early middle age, when the effects of new agents might be expected to show themselves first, and, in the absence of more specific evidence, it seems unlikely that chemical pollution of water and food could have a greater effect than the small effect already estimated for pollution of the air.


For many years there has been suggestive evidence that most of the cancers that are currently common could be made less so by modification of the diet, but, with few exceptions, there is still little reliable evidence as to the modifications that would be of major importance. If we define diet to include all materials that occur in natural foods, are produced during the processes of storage, cooking, and digestion, or are added as preservatives or to give food colour, flavour, and consistency, the ways in which diet could influence the development of cancer are legion.

Ingestion of preformed carcinogens

The most obvious is the ingestion of small amounts of powerful carcinogens or precarcinogens. Several have been identified in foodstuffs but only two have been related at all clearly to the production of cancer in humans. One is aflatoxin, a metabolic product of Aspergillus flavus, which contaminates stored or oily foods such as grains and peanuts in many countries, and is a major cause of liver cancer in the tropics among those individuals who are also chronic carriers of the hepatitis B (or less commonly hepatitis C) virus. Likewise, the salted fish eaten extensively in South China probably acts synergistically with Epstein–Barr virus to cause nasopharyngeal cancer. A third possible source is bracken fern, an extract of which is carcinogenic in animals. It is eaten extensively in Japan and has been tentatively linked with the development of oesophageal cancer. The polycyclic hydrocarbons and other mutagens that are produced in food by grilling or smoking have often been suspected of playing a role, but intensive investigation has failed to detect one.

It seems, therefore, that if diet does affect the incidence of cancer in the Western world in any material way, it is likely to do so by more indirect means, such as affecting the formation, transport, activation or deactivation of carcinogens in the body or affecting the secretion of hormones.


That overnutrition could affect the incidence of cancer was first suggested by Tannenbaum’s experiments on mice during the Second World War. These showed that the incidence of various spontaneous tumours and tumours produced experimentally could be halved by moderately restricting the intake of food without modifying the proportions of the individual constituents. This protective effect has subsequently been demonstrated repeatedly, but has attracted little attention. It is now clear, however, that what is considered normal nutrition in developed countries increases the risk of breast cancer (by bringing forward menarche and increasing body size). With greater consumption obesity (i.e. a body mass index >25 kg/m2) has been estimated to be responsible for 5% of all incident cancers in Europe and 10% of all cancer deaths in nonsmokers in the United States: most notably, cancer of the breast in women after the menopause and cancers of the endometrium, large bowel, and kidney, probably cancers of the oesophagus and gallbladder, and perhaps cancers of the prostate and thyroid. The increases in the two female cancers in postmenopausal women are probably attributable to the formation of oestrogen from androstenedione in adipose tissue, but for others the explanation is unclear.

Meat and fat

Figures for food consumption and cancer incidence and mortality rates in different countries show fairly close correlations between the consumption of fat, and to a lesser extent the consumption of meat, and the incidence of several types of cancer. The correlations are closest for breast cancer and cancer of the large bowel and are less strong for cancers of the endometrium, pancreas, and prostate. When, however, attempts are made to associate the consumption of either type of food with the disease in individuals within a country, the evidence is commonly conflicting. This could be because the international correlations are misleading, indicating only that the risks are correlated with something that is itself correlated with fat and meat consumption (e.g. some other aspect of a high gross national product), but it could be partly because of the inaccuracy of dietary histories and partly because people within developed countries eat such similar diets. Overviews of the published data, however, do suggest that a high consumption of fat is associated with a high risk of colorectal cancer, but the claim that a high consumption of fat (or of particular types of fat) is associated with high risks of breast and endometrial cancer after the menopause, other than by providing a high-calorie diet leading to obesity, is controversial.

Whether meat increases the risk of any type of cancer, apart from the contribution it makes through its calorie content, is also uncertain. The low incidence of several types of cancer that is commonly observed in vegetarian communities is not necessarily due to the absence of meat from the diet, as it can generally be explained by the increased consumption of protective foods (vegetables and fruits) and commonly by associated behavioural characteristics (below average use of tobacco and alcohol). Some studies that make allowance for these confounding factors have claimed that meat specifically increases the risk of large-bowel cancer, but the evidence is weak and the increase in risk, if any, is small.


That fibre may play a part was suggested by Burkitt’s observation that several intestinal diseases, including cancer of the colon, were common in countries in which cereals were processed to remove the fibre and rare in rural Africa and Asia where they were not. The idea was attractive, as ‘fibre’ passes through the small bowel unchanged and serves as pabulum for the colonic bacteria, thus increasing faecal bulk and possibly protecting mechanically against the development of cancer by diluting any carcinogens present and hastening their transit through the bowel. The idea was too simple, however, and has not been confirmed (using the original definition of fibre) by either epidemiological studies on individuals in developed countries or by experiments aimed at reducing the recurrence of colorectal adenomas. In fact, fibre is difficult to define and the term is better replaced by ‘nonstarch polysaccharides’ as there are many that share the characteristics of passing through the small bowel unchanged and being, for the most part, partially or wholly degraded by bacteria in the large bowel. Some starch, moreover, known as ‘resistant starch’ and found in green bananas and cold potatoes, has similar physiological characteristics. Further studies that take these complexities into account are, therefore, needed before ‘fibre’ in any of its manifestations can be considered as having any place in protecting against the development of cancer.

Retinoids and carotenoids

Experiments on animals and on cell cultures in vitro have suggested that vitamin A (retinol) and its esters and analogues (retinoids) may, in appropriate circumstances, reduce the risk of cancer by reducing the probability that partially transformed cells become fully transformed and proliferate into clinically detectable tumours, although in other circumstances they appear to have opposite effects. Human studies, however, did not support the idea that serum levels of retinol are related to the risk of any type of cancer, at least in countries in which clinical symptoms of vitamin A deficiency seldom or never occur. Such studies suggested that the risks were inversely related to the serum level of β-carotene, which acts as an antioxidant and is broken down to produce retinol. When β-carotene was put to the test of clinical trials, however, it provided no benefit and the inverse relationship commonly observed in epidemiological studies is presumably due to confounding with some other protective factors in vegetables.

Other components

Many other components of the diet, including lycopene in tomatoes, indoles in brassicas (e.g. cabbages and sprouts), phyto-oestrogens (plant chemicals structurally similar to oestradiol), fresh fruit and vegetables, vitamins C, D, and E, and calcium and selenium have also been proposed as protective agents. Conversely, nitrates, nitrites, secondary amines, and the preservation of food by salting, have been thought to increase the risk of cancer. For some the evidence is strongly suggestive: notably for vitamin C as protective against gastric cancer and for salt-preserved foods predisposing to it. In general, however, the evidence of benefit or harm is too weak to justify any firm conclusion.


Some of the uncertainties about the effect of diet could be resolved only by means of controlled trials in which volunteers are allocated at random to micronutrient supplements (such as vitamin C or lycopene) or a dietary schedule that requires a substantial reduction in the consumption of fat. Several such studies are under way, which may give answers within a few years. Practicable modifications of the diet may well provide the means for reducing cancer deaths in developed countries by one-third, but the range of uncertainty about this figure is large. Meanwhile the only dietary changes that can be recommended with confidence in developed countries are a general increase in the use of fresh fruit and vegetables and a sufficient limitation of calories to avoid obesity.

Reproduction, other factors affecting secretion of reproductive hormones, and other hormones

Epidemiological observations have shown clear relationships between a woman’s menstrual and reproductive history and the risk of cancers of the breast, endometrium, and ovary, which are generally thought to reflect changes in hormonal secretions. Which hormones are concerned, however, and the mechanisms by which they act are, for the most part, still uncertain. An exception is endometrial cancer, the risk of which is directly related to the degree of exposure to oestrogen not followed after an appropriate interval by progestogen. Strong evidence that oestrogenic stimulation of the mammary tissues is a cause of most cases of breast cancer in developed countries has been provided by randomized trials of tamoxifen, an antioestrogenic drug that blocks the oestrogen receptors in the cells of the normal breast. The effect is large and rapid: 5 years of tamoxifen approximately halves the incidence of contralateral breast cancer in a woman who has had previous breast cancer, not only while the drug is being taken but also for some years afterwards. Exogenous oestrogen also increases the risk of breast cancer when given as hormonal replacement therapy and endogenous oestrogen accounts for the increased risk associated with adiposity after the menopause, as androstenedione, which continues to be produced by the adrenals, is converted to oestrogen in adipose tissue. It is presumably oestrogens, too, that cause a small increase in risk of breast cancer during and immediately after pregnancy and the oestrogen component of the steroid contraceptives that causes a similar small increase in risk during their use and for a few years after their use is stopped. It is, however, unclear which hormone-related processes are involved in reducing the long-term risk for the rest of a woman’s life that occurs some years after the occurrence of each pregnancy and it is equally unclear why the use of oral contraception and the consequent suppression of ovulation reduces the long-term risk of ovarian cancer.

Sex hormones, it is thought, may also be involved in producing cancers of the testis and prostate in men. For testis cancer, the strongest evidence has been for an effect of maternal oestrogen levels on the developing testis in utero, whereas for prostate cancer the evidence relates to adult androgen levels, but for neither has causation been established. Randomized trials of the effects of physical or medical castration in men who already have prostate cancer have shown that progression of the disease can be slowed substantially, presumably by the reduction of androgenic stimulation.

In recent years it has been found that prior raised endogenous levels of insulin-like growth factor-1 (IGF-1) are associated with raised risk of cancers of the breast, colon, and prostate, and there has also been evidence that endogenous and exogenous growth hormone affects the risk of colorectal cancer.

Physical inactivity

Physical inactivity contributes to the risk of cancer indirectly by increasing the risk of obesity but it may also contribute directly. Associations with colon, breast, and endometrial cancers have fairly consistently been reported, and an association with prostate cancer has been found in most studies. The mechanism of effect is uncertain, but possibilities include effects of exercise on hormone levels, on immune function, and on intestinal transit time and hence the duration of exposure of the colonic mucosa to faecal carcinogens.

Interaction of agents

Attribution of the risk of cancer to different causes is complicated by the fact that some agents interact with others to produce effects that are much greater than the sum of the separate effects of each on its own. An example is provided by smoking and asbestos, which multiply each other’s effects so that, compared with nonsmokers in general, the incidence of cancer of the lung was increased sixfold among a group of asbestos insulation workers in the United States of America who did not smoke, but were heavily exposed to asbestos dust in the 1940s, 10- to 20-fold among cigarette smokers in general who did not work with asbestos, and nearly 90-fold among the asbestos workers who also smoked cigarettes regularly. Other examples are provided by smoking and radon (which interact similarly, though somewhat less than multiplicatively, to produce cancer of the lung), by smoking and alcohol (which interact to produce cancers of the mouth, pharynx, larynx, and oesophagus), and by infection with the hepatitis B virus and aflatoxin (which interact to produce cancer of the liver).

Such interactions complicate the attribution of risk, as we may find ourselves appearing to claim that more cancer can be prevented than actually occurs by attributing, say, 80% of lung cancers in men heavily exposed to asbestos to their occupational exposure and 90% of the same cancers to cigarette smoking. Each, separately, is correct, but they cannot be added to show the combined effect.


Estimates of the proportions of fatal cancers that can be attributed to environmental and behavioural factors, grouped into 11 main categories, are given in Table 7 below. The evidence on which these estimates are based is summarized in this article and in greater detail by Doll and Peto (1981), Tomatis et al. (1990), Stewart and Kleihues (2003), and Schottenfeld and Fraumeni (2006).

The sum of the best estimates in Table 7 amounts to less than 100%, despite the fact that some of the listed agents interact with one another to augment each other’s effect and that some fatal cancers are consequently counted twice. The total would be somewhat more than 100%, however, if the true proportions attributable to some of the categories turn out to be nearer the upper end of the acceptable estimates.

The estimates in the second and third columns of Table 7 do not distinguish between factors (such as tobacco) that are sufficiently understood to enable specific action to be taken with a guarantee of success and those (such as diet) that are not. They should not, therefore, be taken as guides to the proportion of cancer deaths that can now be prevented by practicable means. This is illustrated by the fourth column in Table 7, which shows the proportions of United Kingdom cancer deaths in 2005 that are reliably known to be avoidable by practicable means. The percentage attributed to tobacco is more than the sum of the percentages reliably attributable to other specific factors for which practicable preventive measures are available: and tobacco causes about twice as many deaths from other diseases as it does from cancer. The position is different in countries such as China, where hepatitis B virus causes about as many cancer deaths as tobacco and the hazard for future generations can be avoided in a cost-effective way by infant vaccination.

Table 7  Estimated proportion of United Kingdom cancer deaths in the year 2005 attributed to previous exposure to different environmental and behavioural factors and proportion of future United Kingdom cancer deaths avoidable by known effects of practicable changes in current exposure levels


Factor or class of factors Percentage of UK cancer deaths in 2005 attributed to previous exposure levels Estimated percentage of future UK cancer deaths avoidable by known effects of practicable changes in current exposure levels
Best estimate Range of acceptable estimates
Tobacco 28 25–30 <28a
Alcohol 6 4–8 <6b
Ionizing radiation 3c 2–4 <1
Ultraviolet radiation 1 1 <1
Infection (virus 3%, bacteria 2%) 5 4–15 <1d
Medical drugs <1 0–1 <1
Occupation 4 3–5 <1e
Pollution 2 1–5 <1
Diet 25 15–35 <8f
Reproductive and hormonalg 15 10–20 <1
Other and unknown ? ? ?

a The proportion of UK cancer deaths attributed to smoking fell from 34% in 1985 to 28% in 2005 and is still decreasing (as current levels of smoking would cause less than 28% of future UK cancer deaths).

b Mostly cancer of the upper aerodigestive tract that could have been avoided by not smoking.

c 0.8% diagnostic x-rays, <0.1% all other manmade, 0.7% natural radon in houses, 1.8% other cosmic or terrestrial natural sources.

d Cervical cancer currently causes 1.4% of UK female cancer deaths. Without screening it would cause several per cent; with even better screening it would cause a fraction of 1%.

e Although asbestos exposure is now strictly controlled, the delayed effects of past exposure probably account for about 3% of current cancer mortality (including more than 1% from mesothelioma), and diesel, coal and other smoke must also have appreciable effects, especially if they potentiate tobacco smoke.

f In later middle age in the UK almost half are overweight (BMI 25–30 kg/m2) and a quarter are obese (>30 kg/m2); minimal cancer mortality is at c.25 kg/m2, and about 8% of current cancer mortality would be avoided if the overweight and obese had a BMI of 25 kg/m2. (Instead, BMI is increasing by about 1 kg/m2 per decade.)

g Includes other factors affecting the secretion of reproductive hormones.

Epidemiology of cancer by site of origin

In the following account of the epidemiology of cancers arising in specific organs, the description of each type is preceded by notes showing its importance in England. One figure gives the proportion of all cancers that arise at the site, from national cancer registrations for England in 2004 and another gives the proportion of all cancer deaths allocated to the site in the national mortality statistics for England and Wales for 2004. A third gives the ratio of the age-standardized incidence rates in England for each sex. The way in which the incidence of the disease varies with age is shown for males and females in a series of graphs, using data for England over a 5-year period (1993–7). Trends in incidence and mortality for each type, along with the trends in possible causative factors, are given by Swerdlow et al. (2001). Trends internationally are commented on in the text, and described more fully in Doll et al. (1994). Comments on the total worldwide frequency of different cancers are based on data from Parkin (2001); they disregard incidence of nonmelanoma skin cancers, for which reliable international statistics are not available.


  • 0.1% of all cancers and 0.01% of cancer deaths
  • Sex ratio of rates 2.0:1; age distribution like oesophageal cancer

Carcinoma of the lip was one of the first types of cancer to be related to an extrinsic cause when, more than 200 years ago, it was noted to occur characteristically in pipe smokers. Many years later it was realized that the disease could also be produced by smoking cigarettes, although much less readily, so that it must be produced by the chemicals in smoke rather than by the nonspecific effect of local heat. It is also much more common in outdoor than in indoor workers and is induced by ultraviolet radiation in the same way as other cancers of the exposed skin. Solar ultraviolet radiation and tobacco account, between them, for the great majority of all cases in the United Kingdom, probably multiplying each other’s effects. The disease is much less common than it used to be, because of the decrease in both pipe smoking and outdoor work.

Oral cavity and pharynx (excluding salivary glands and nasopharynx)

  • 1.2% of all cancers and 1.0% of cancer deaths
  • Sex ratio of rates 2.2:1; age distribution like oesophageal cancer

Cancers of the tongue, mouth, and pharynx (other than nasopharynx) are all related to smoking (of pipes, cigars, and cigarettes) and to the consumption of alcohol. The two factors act synergistically and cancers in these sites are extremely rare in nonsmokers who do not drink alcohol. There is also fairly consistent evidence of an association of risk with low intake of fruit and vegetables.

Cancer of the tongue is much less common in Britain than it was 100 years ago, but the reason for this sharp decline is unknown. One explanation could be the decrease in syphilis, which was commonly believed to be a predisposing factor because of the clinical association with syphilitic leucoplakia. Recent increases in oral and pharyngeal cancer in men are partly due to increased consumption of alcohol and possibly, in the case of pharyngeal cancer, to human papillomavirus infection.

Cancers that occur low in the hypopharynx are distinguished by a tendency to affect women who have suffered from iron-deficiency anaemia and dysphagia.

Cancers of the mouth and pharynx (excluding nasopharynx) are particularly common in south east and central Asia where tobacco smoking is largely replaced by chewing tobacco, betel nut or leaf, and lime (calcium hydroxide). A close association with such chewing habits has been established by studies that have shown that the cancers tend to originate in the part of the mouth in which the quid is usually held—a characteristic that varies both between individuals and between areas. The materials chewed differ in different places and, although the disease is commonly described as ‘betel chewer’s cancer’, betel is not invariably a component of the quid and the most characteristic constituent seems to be a small amount of lime and, in most cases, some form of tobacco. In parts of Asia, the disease is so common that it accounts for 20% of all cancers and in those populations the abandonment of chewing would be the single most effective means of reducing the total incidence of cancer—so long as the habit was not replaced by an increase in tobacco smoking. Among habitual quid chewers, the risks are particularly elevated in those who both chew and smoke—indeed, in parts of India the majority of deaths from betel chewer’s cancer could have been avoided if those affected had not also smoked. The incidence might also be reduced by improved nutrition, as the disease in southern Asia tends to be associated with vitamin A deficiency.

In parts of India where women tend to smoke local cigars and cigarettes with the burning end inside the mouth to prevent them going out, the habit is associated with cancer of the palate.

Salivary glands

  • 0.2% of all cancers and 0.1% of cancer deaths
  • Sex ratio of rates 1.5:1; age distribution, 

The salivary glands are not common sites for cancer anywhere. They are, however, relatively more common in circumpolar Inuits than others. No causative factors are known other than ionizing radiation exposure, and no notable changes in incidence over time have been reported.


  • 0.1% of all cancers and of cancer deaths
  • Sex ratio of rates 2.5:1; age distribution, 

Cancers of the nasopharynx, unlike those in other parts of the pharynx, are not related to alcohol and are only weakly related to tobacco. They are rare in most populations but are common in southern China, especially so in Cantonese originating from parts of Guangdong, where the disease is the most common type of cancer. A weak relationship with HLA type has been reported. Moderately high rates have been observed in Eskimos, American Indians, Malays and Filipinos. Rates decrease over succeeding generations in Chinese migrants to (low-risk) Western countries.

DNA characteristic of the Epstein–Barr virus (EBV) has been detected in the nuclei of nasopharyngeal cancer cells and patients with the disease tend to have unusually high antibodies against EBV-related antigens. Among adults, sudden increases in certain EBV antigens in the blood often precede the appearance of a cancer by a few years. Infection with the EBV is, however, almost universal and can be only one of several agents that act in combination to produce the disease. One such agent in Southern China occurs in the ‘salted fish’ on which children are commonly weaned. This strongly flavoured delicacy bears little relation to the salted fish eaten elsewhere, and might better be described as decomposing fish: it contains various mutagens, and exposure to it in childhood when infection with EBV first occurs may alter the usual lifelong balance between host and virus in some hazardous way.


  • 2.1% of all cancers and 4.7% of cancer deaths.
  • Sex ratio of rates 2.6:1; age distribution, 

Cancer of the oesophagus is the eighth most common cancer in the world. It is exceptional among malignancies in the extent of geographical variation in incidence, both internationally and often also over relatively small distances. Like other cancers of the upper respiratory and digestive tracts, cancer of the oesophagus is closely related to prolonged smoking and the consumption of alcohol. All types of smoking have comparable effects and, so it appears, do all alcoholic drinks, although spirits may be slightly more effective per gram of ethyl alcohol than other alcoholic drinks. Alcohol and tobacco act synergistically and, in the absence of either, the incidence of the disease in Western countries would be greatly reduced. Smoking raises the risk of both squamous cell and adenocarcinoma of the oesophagus whereas alcohol affects largely or solely the former. A few cases originate from the scars produced by poisoning with corrosive substances and a very few in conjunction with a particular hereditary form of tylosis (presenting with keratoses of the palms and soles). The relatively small excess in men probably reflects the existence of other unknown causes in women, possibly nutritional in origin and similar to those responsible for cancers of the hypopharynx. Mortality (which, because of the high fatality rate, approaches incidence) fell progressively in men in Britain from the 1920s to the 1960s, in line with the fall in the consumption of alcohol, and rose again after 1960 when the trend in the consumption of alcohol reversed. Since pipe smoking affects oesophageal cancer risks at least as strongly as cigarette smoking, no large effects on male oesophageal cancer trends could be predicted from the male switch from pipes to cigarettes, although the switch by females from nonsmoking to cigarettes should, other things being equal, produce a large upward trend. It appears, however, that other things were not equal and some other, possibly nutritional, cause of oesophageal cancer seems to have decreased, for the upward trend in oesophageal cancer in women has been moderate. Oesophageal cancer is associated in several studies with low fruit and vegetable consumption. In men, in contrast, the rates have increased when based on smoking they might have been expected to decrease. To some extent this can be accounted for by the increased consumption of alcohol and possibly by an increase in the nitrosamine content of tobacco smoke, which has resulted from changes in the method of curing tobacco and which could have a specific effect on the oesophagus. A part of the increase is due to an increased risk of adenocarcinoma at the lower end of the oesophagus, which may be associated with a decreased prevalence of Helicobacter pylori and gastritis, and an increase in gastro-oesophageal reflux and Barrett’s oesophagus, which is a common precursor of the tumour. Obesity is associated with risk of adenocarcinoma of the oesophagus, perhaps via an effect on reflux. The balance of adenocarcinoma and squamous cell cancer of the oesophagus has altered greatly over time in Western countries, such that the former, which has been increasing, now generally predominates.

In Africa and Asia, the epidemiological features are quite different and present some of the most striking unsolved problems in the field of cancer epidemiology. In parts of China (particularly in north Henan but also elsewhere) and on the east coast of the Caspian Sea in Turkmenistan and Iran, oesophageal cancer is the most common type of cancer, with incidence rates in both sexes that are equal to the highest rates observed for lung cancer in men in European cities. Within China, the disease varies more than 10-fold from one county to another. In parts of Africa, particularly in the Transkei region of South Africa and on the east coast of Lake Victoria in Kenya, extremely high rates are also observed, sometimes equally in both sexes and sometimes only in men. In these and several other areas, the high incidence zones are strictly localized and the incidence falls off rapidly over distances of 200 or 300 miles (c.300–500 km).

When tobacco and alcohol are used they increase the hazard, but they are not the principal agents in these high-incidence areas. Many dietary causes have been proposed, including micronutrient deficiencies, contamination of food and pickled vegetables by fungi (particularly by species of fusaria) with the production of carcinogenic metabolites, an agent associated with the production of beer from maize, drinking very hot beverages, and the residues left behind in pipes from smoking opium (which are commonly swallowed). None, however, is supported by any impressive, consistent epidemiological data. The high-incidence area in Iran, which has been intensively investigated, is characterized by extreme poverty and a restricted diet consisting chiefly of home-made bread and tea, with some sheep’s milk and milk products, and very little meat, vegetables, or fruit. In this area the disease has been common for centuries. In southern Africa, however, it seems to have become common only since the First World War. In China, where cancer of the oesophagus was the second most important neoplastic cause of death in the 1970s, the high incidence has persisted.


  • 2.3% of all cancers and 3.8% of cancer deaths
  • Sex ratio of rates 2.4:1; age distribution like oesophageal cancer

Until about 1980, gastric cancer was responsible for more deaths from malignant disease worldwide than any other; it is now second to lung cancer, with over 600 000 deaths per year, mainly in developing countries. Over the last 50 years, the incidence has declined in Western countries, and recently it has begun to do so in South America and Japan.

The highest rates now are in parts of Japan and China, with high rates also in other parts of Eastern Asia, and countries in the ex-Soviet Union and Eastern Europe, while low rates are found developed parts of Africa. This contrasts with the strong socioeconomic gradient in incidence of the tumour seen within Western countries. Irrespective of whether the incidence in a country is high or low, the sex ratio is generally between 1.5 and 3 to 1.

In migrants from high-risk to low-risk countries, for instance from Japan to the United States of America, risk decreases with longer time since migration, but can take two or more generations to reach local levels. Risk of gastric cancer is raised in relation to gastritis associated with chronic infection by H. pylori (sometimes leading to atrophic gastritis), a diet deficient in fruit and green and yellow vegetables, and a poor diet with large amounts of salt and salt-preserved food. Chronic infection with H. pylori, which is very common, is a major cause of peptic ulcer, a finding that is of considerable practical value in patients with ulcers, because the infection can generally be eliminated from the stomach by a short course of appropriate antibiotic therapy and this provides long-term protection against recurrence. Whether such treatments will have any material effect on the incidence of stomach cancer remains, however, to be shown. How these various factors influence the production of the disease is unclear. One possibility is that they encourage or discourage the formation of carcinogens in vivo, particularly perhaps the production of nitrosamines; but if they do, the intake of nitrates (which can be converted into nitrites by bacterial enzymes) is not a rate-limiting factor. Changes in the prevalence of the three factors above could have contributed to the decline in the incidence of the disease, but they could not have brought about such a large and widespread reduction in risk, and it seems probable that the better preservation of food, resulting from the extensive use of refrigeration, has played the major part.

No risk has been detected from the consumption of mutagens produced by the different methods of cooking meat and fish, nor from food additives or pesticide residues. Some food additives may, on the contrary, have served to reduce risk (by avoiding food spoilage and hence improving nutrition, by avoiding contamination by carcinogen-producing microorganisms, or by some antioxidant or other protective effect on the gastric epithelium).

Risk of gastric cancer is also raised, moderately, by smoking, and raised by exposure to ionizing radiation.

Large bowel

  • 10.1% of all cancers and 10.5% of cancer deaths
  • Sex ratio of rates 1.6:1; age distribution like oesophageal cancer

Cancers of the colon and rectum ought to be considered separately, as their causes are not identical. Cancer of the colon, for example, tends to occur more often in women than in men, particularly when it occurs on the right side, while cancer of the rectum is nearly twice as common in men. The geographical distribution also differs slightly, colonic cancer varying in incidence more than rectal cancer. Separate consideration may, however, sometimes be misleading as cancers commonly occur at the rectosigmoid junction and the site of origin of these cases is not recorded consistently. Moreover, there is a growing tendency to describe both diseases merely as ‘cancers of the large bowel’, which, according to the internationally agreed coding rules, are classed as cancers of the colon. The two diseases will, therefore, be considered together.

Cancers of the colon and rectum are the third most commonly incident cancer, and the fourth most common cause of cancer death, in the world. Almost a million cases occur per year worldwide, mainly in developed countries. The disease is most common in Western countries, but incidence has generally stabilized or decreased in these countries in recent years, especially at younger ages. In Japan, where incidence used to be very low, rates have risen to be similar to those in the United States of America and western Europe. Rates in migrants from low-risk to high-risk countries, for instance in previous times from Japan to the United States, tend to gain much or all of the host population risks within the first generation.

In most parts of Asia, and in Africa and eastern Europe, large-bowel cancer has been relatively uncommon (except in areas where chronic schistosomal infestation of the large intestine is prevalent; for example, high rectal cancer rates are found in Chinese counties in which Schistosoma japonicum was, until recently, a major cause of death). Rates tend to rise markedly, however, with the introduction of a Western lifestyle.

Incidence rates in different countries correlate closely with the per caput consumption of fat and meat and crudely with the consumption of processed foods from which the natural fibre has been removed. Ways in which these and other dietary constituents might influence the development of the disease have been discussed under diet. Other factors associated with increased risk are obesity and physical inactivity. A weak association with smoking has been observed in several cohort studies, which may be the result of confounding with the consumption of alcohol (which is associated with colorectal cancer risk in most studies) and a high-fat diet. It is possible, however, that smoking may cause a few cases indirectly by causing the diet to be modified in the direction of a higher fat content.

Cases in childhood or early adult life occur as a complication of familial adenomatous polyposis and of hereditary nonpolyposis colorectal cancer (HNPCC) syndrome. These conditions are determined by dominant genes, which increase the susceptibility to the disease so much that, unless prophylactic measures such as colectomy are undertaken, cancer is highly likely to develop at or before middle age. Many other cases develop from adenomatous polyps and a few occur as a complication of long-standing ulcerative colitis and Crohn’s disease. There is substantial evidence of reduced colorectal cancer risk among long-term users of nonsteroidal anti-inflamatory drugs. There is also increasing evidence for an association of colorectal cancer risk with hormones of the growth hormone/IGF-1 axis: colorectal cancer risks are increased in patients with acromegaly; raised risks have been found in the general population in relation to prior greater levels of IGF-1; and raised risks have been found in patients treated with growth hormone.

Anal intercourse causing infection with types 16, 18 or some other specific types of the human papillomavirus is a probable cause of some anal carcinomas in both sexes, but patients who have sexually transmitted anal warts that are due to other types of human papillomavirus are not for this reason at special risk of anal cancer.


  • 0.8% of all cancers and 1.7% of cancer deaths
  • Sex ratio of rates 2.0:1; age distribution, 

Liver cancer is the third most common cause of cancer death in the world, with four-fifths of cases occurring in developing countries, and one-half occurring in China alone. Incidence is considerably greater in men than women almost everywhere. Incidence rates have tended to be overestimated in developed countries because the primary condition is often confused with metastases to the liver from cancer in various other organs, particularly at older ages. Recently, however, there has been an increase in the United Kingdom and the United States from the very low level that existed previously, perhaps due in part to an increased prevalence of infection with hepatitis C.

The disease is common in South East Asia (the highest recorded rates are in parts of China and Thailand) and tropical Africa. In China it accounts for about 18% of all cancer deaths and in parts of Africa it is the most common cancer in men. Most cases derive from the main cells of the organ (hepatocellular carcinomas) and are attributable primarily to chronic active infection, established early in life, with the hepatitis B virus, exacerbated by consumption of some specific metabolite (e.g. aflatoxins) of particular types of fungi that contaminate stored foods. Neonatal vaccination against the virus produces a marked decrease in the proportion of children who become chronically infected. This has begun in many countries, now including the whole of China and parts of tropical Africa, and has already produced a decreased risk of hepatocarcinoma at young ages. Some cases, however, are caused by chronic infection with hepatitis C (a blood-borne RNA virus that cannot be avoided by immunization; see Chapter 7.5.22).

In developed countries, although some cases are also due to infection with hepatitis B and C viruses, more arise as complications of cirrhosis of the liver attributable to heavy and prolonged consumption of alcohol or, rarely, to haemochromatosis, certain types of porphyria, α1-antitrypsin deficiency, and hereditary tyrosinaemia type 1. Occasionally, liver cancer is produced by drugs. A few cases have occurred in young men who have taken androgenic anabolic steroids to increase their muscular strength and a few from the use of steroid contraceptives, either arising de novo or from benign adenomas, which are themselves rare complications of the use of steroid contraceptives. Some can be attributed to smoking, for an association has been observed in parts of China where little alcohol is drunk and case–control studies in Europe have shown an association after alcohol consumption has been taken into account.

A second histological type (cholangiosarcoma) arises from the intrahepatic bile ducts, tends to occur at a somewhat later age than hepatocellular carcinoma, and, although generally less common than hepatocellular carcinoma, nevertheless accounts for an appreciable proportion of cases. In parts of China, Thailand, and elsewhere in south or east Asia it can be produced by chronic infection with liver flukes (Clonorchis sinensis or Opisthorchis viverrini). In north-eastern Thailand the latter fluke causes one of the highest rates of liver cancer in the world. In developed countries, primary sclerosing cholangitis is the main known risk factor for cholangiosarcoma.

A third histological type that is extremely uncommon everywhere has been variously described as reticuloendothelioma or angiosarcoma. It was first recognized as a complication of the use of Thorotrast as a contrast agent in neuroradiology, a long-abandoned practice that led to chronic retention of insoluble thorium radionuclides in the marrow, spleen, and liver. In 1973, the disease was found to be an occupational hazard for men exposed to vinyl chloride monomer. A few hundred cases have occurred throughout the world in men who were heavily exposed in the manufacture of vinyl chloride polymer, and linear extrapolation suggests that the minute amounts that have leached out of plastic consumer products might have caused only a dozen or so cases altogether in the general public, if indeed they have produced any. A third, and even rarer, cause is prolonged exposure to inorganic arsenic, such as used to result from the medical prescription of Fowler’s solution. Despite these multiple causes only one case of hepatic angiosarcoma normally occurs annually per 10 million people, which is why the recognition of new causes has been easy.

The relative rarity of cancer of the liver in most developed countries is intriguing, since most of the carcinogens thus far discovered in experimental animals induce, perhaps with other cancers, tumours of the liver.

Gallbladder and extrahepatic bile ducts

  • 0.4% of all cancers and 0.4% of cancer deaths
  • Sex ratio of rates 0.9:1; age distribution like oesophageal cancer.

Cancers of the gallbladder and extrahepatic bile ducts are nearly always classed together, which is unfortunate as the causes differ. The former is more than twice as common in women as in men, is probably associated with obesity, and is usually preceded by (and probably caused by) cholelithiasis. The latter is slightly more common in men and is increased in incidence by liver fluke infection, primary sclerosing cholangitis and long-standing ulcerative colitis. Both types are uncommon, and their aggregate varies only moderately from one population to another. The highest rates are recorded among Japanese, Koreans, American Indians, and in women in Delhi (India) and parts of South America.

The incidence of cancer of the gallbladder has fallen sharply in the United States in the last 25 years, which may be partly due to the decreased consumption of animal fat and, perhaps more importantly, to an increase in the rate of cholecystectomy in people who, having gallstones, are at greatest risk of cancer of the gallbladder.


  • 2.1% of all cancers and 4.7% of cancer deaths
  • Sex ratio of rates 1.3:1; age distribution like oesophageal cancer

Cancer of the pancreas is two to three times more common in regular cigarette smokers than in lifelong nonsmokers. The chemicals in cigarette smoke that specifically cause pancreatic cancer have not been identified, but the volatile nitrosamines in smoke that are absorbed from the alveoli and carried to the pancreas in the bloodstream are likely candidates. The disease is twice as common in diabetics as in the population as a whole and risk is raised in patients with chronic pancreatitis.

Cancer of the pancreas is generally regarded as a disease of the developed world, but the diagnosis is difficult in the absence of a well-developed medical service and some of the relatively small geographical and temporal variations may be due to variation in diagnostic standards. The greatest reported rates are in African Americans. Mortality rates in Britain and the United States have begun to decrease under 65 years of age, and this is more likely to reflect a reduction in incidence from reduction in smoking than to any improvement in treatment, as the 5-year survival rate remains less than 5%.

Nose and nasal sinuses

  • 0.1% of all cancers and of cancer deaths
  • Sex ratio of rates 1.5:1; age distribution like oesophageal cancer

Surprisingly, in view of the widespread exposure of the human nose to tobacco smoke and other airborne toxins, cancers of the nasal cavity itself are extremely rare. Most arise from the paranasal sinuses. Several occupational hazards have been recognized, including the refining of nickel, processes giving rise to exposure to strong sulphuric acid mists, and the manufacture of hardwood furniture and leather goods. It would be wrong, however, to conclude that all contact with nickel, hardwood dust, and leather creates a hazard. The hazards have been observed in special occupational situations in which exposure has been intensive and prolonged. The nickel-refining hazard was first observed in South Wales where the nickel carbonyl process was used, but similar hazards were subsequently observed with other refining processes in Canada, Norway, and the Soviet Union. In the Welsh refinery the workplace exposures were much heavier before the Second World War, and (despite the continued use of the nickel carbonyl process in Wales) no hazard of nasal sinus cancer has been observed among men first employed there since 1950. The hazard in furniture workers was first observed in High Wycombe (southern England) and appears to have followed the introduction of high-speed woodworking machinery early in the 20th century. A hazard certainly affects some other groups of woodworkers, but should not be assumed to affect furniture workers in general.

Most nasal and nasal sinus cancers are squamous carcinomas, but the hazard from hardwood dust characteristically produced adenocarcinomas. In some of the groups exposed to this hazard, as many as 5% of the men developed the disease. This meant that the risk of adenocarcinoma was increased 1500 times (as this histological type of the disease is normally very rare) and the hazard was, in consequence, easy to confirm once suspicion had been aroused.

Chromate workers are sometimes said to experience a hazard of nasal cancer, but this may be an error due to confusion with the characteristic ‘chrome ulcer’ of the nasal septum. Such ulcers have not generally been found to become malignant. A causal excess of nasal sinus cancer has been seen, however, in women employed in the United States in the early 20th century to apply radium-containing luminescent paint to dials and clocks, who ingested the radium when they licked the brushes to shape their tips. Risk of nasal cancer is also modestly related to smoking.


  • 0.6% of all cancers and 0.5% of cancer deaths
  • Sex ratio of rates 6.2:1; age distribution, 

Cancers of the larynx, like cancers of the oesophagus and buccal cavity, are closely associated with tobacco smoking and with the consumption of alcohol. The two agents act synergistically and in the absence of either the disease is rare. The different parts of this small organ are, however, related to the two agents differently. Cancers of the glottis are strongly related to smoking, particularly to cigarette smoking, and only weakly to alcohol, while cancers of the epilarynx resemble cancers of the neighbouring hypopharynx and are strongly related to both agents and to pipe and cigar smoking equally with cigarette smoking.The highest reported incidence of laryngeal cancer in men is in African Americans in New Orleans (southern United States) and in parts of Spain. Rates in women are relatively low everywhere and generally a small proportion of those in men in the same place. Trends with time vary considerably between countries and between the sexes, reflecting trends in smoking and alcohol consumption, and probably some other aetiological factor, perhaps nutritional in character. There has been fairly consistent evidence for an inverse association of risk with fruit and vegetable consumption. That there are other causal factors is evident from the relatively high incidence rates in parts of India, Turkey, North Africa, and Brazil, which cannot be accounted for by tobacco and alcohol.

The disease has also occurred as an occupational risk in the manufacture of mustard gas and in processes that cause exposure to strong sulphuric acid mists.


  • 10.4% of all cancers and 21.0% of cancer deaths
  • Sex ratio of rates 1.8:1; age distribution, 

Nearly all lung cancers are bronchial carcinomas and should properly be so described. The term ‘lung cancer’ is, however, in such common use that it is used here as synonymous with bronchial carcinoma, although it actually includes a very small proportion of alveolar cell carcinomas and other rare types of cancer with different characteristics. Lung cancer is the most common cancer in the world, with over 1.2 million new cases per year, and the most common cause of cancer death.

Until the 1920s, lung cancer was uniformly rare (except in the Hartz mountains, see below). In the next two decades, German and then British pathologists began to comment on an apparent increase, but this tended to be dismissed as an artefact of the greatly improving methods of diagnosis and the establishment of special centres for thoracic disease. Gradually, however, the increase became so pronounced and the change in the sex ratio so marked that the increase could no longer be dismissed as wholly artefactual and by the late 1940s, when the age-standardized mortality rate in men in the United Kingdom had increased 20 times, it was clear that the developed world had begun to see an epidemic of lung cancer comparable in severity to the epidemics of infectious disease of the past, though with a longer time scale. Until the 1940s, the increase among British women was largely a diagnostic artefact. Since 1950, however, diagnostic standards in middle age have changed very little, the increase in British men has been replaced by a decrease, while the increase among middle-aged women has continued for longer, before also reversing. As a result, the sex ratio (male rate divided by female rate), at e.g. 50 to 54 years of age, which rose from 1.8 after the First World War to 8.9 after the Second World War, was reduced to 1.3 in 2004. Changes in treatment have had little effect on the fatality rate, which remains extremely high, and real changes in mortality closely reflect real changes in incidence.


These time trends can be explained almost entirely by the effect of smoking tobacco, particularly in the form of cigarettes, which caused more than 90% of all lung cancers in the United Kingdom in the early 1990s. Evidence of this effect was first obtained in the middle of the last century by comparing the smoking histories of patients with different diseases (case–control studies). It was found that the proportion of patients who had never smoked was much smaller if they had lung cancer (the ‘cases’) than if they had some other disease (the ‘controls’), and the proportion who had smoked heavily was correspondingly greater.

Further evidence was obtained by asking large numbers of apparently healthy men and women what they smoked and then following them up to determine the causes of death of those who had died. Cohort studies of this type, in the United States, in doctors in the United Kingdom, and in other groups, have all shown similar results, the risk increasing with the amount smoked, and varying with the length of time cigarettes had been smoked. If attention is restricted to populations in which most cigarette smokers had been smoking cigarettes regularly since early adult life, lung cancer is about 20 times more common in regular cigarette smokers than in lifelong nonsmokers and up to 40 times more common in very heavy smokers. At first the relationship was less marked in women than in men, but this was because female smokers who were old enough to have a high risk of cancer either had not begun smoking cigarettes so early in adult life or had smoked them less intensively when they began, and the sex differences in behaviour and risk have both been progressively eliminated with the passage of time.

Further studies have found that the relative risk of lung cancer has increased with decreasing age of starting to smoke and decreased with the number of years that smoking has been stopped (detectable at 5 years after stopping, but never reaching the risk of a lifelong nonsmoker); that the national increases in incidence have appeared at appropriate times after the increase in cigarette sales (after due allowance is made for a spurious increase due to improved diagnosis and appropriate differences in consumption by men and women); and that there is a general parallelism between the incidence of the disease in different countries and social and religious groups and the prolonged consumption of cigarettes. Finally, and most encouragingly, the trend in mortality has reversed following reduction in smoking. By 2004, the mortality from lung cancer among men in their thirties in Britain was only about one-fifth of that of men of the same ages some 50 years earlier, corresponding to the earlier changes in the prevalence of smoking. The reduction in tar delivery between 1939 and 1965 contributed to the reduction in lung cancer in young men after the war, but the later reduction had little effect because of changes in the way cigarettes were manufactured and in the way they were smoked to ensure an adequate intake of nicotine. At older ages the decreases are less striking, but they are now seen at all ages in British men, and up to 75 in British women.

In recent years, it has been shown that indoor air pollution with tobacco smoke—‘passive smoking’—increases lung cancer risk, by about 20 to 30% from long-term adult exposure.


Several other causes of lung cancer have been discovered as a result of observations in industry. Many thousands of men and women have experienced significant hazards from exposure to asbestos or to polycyclic hydrocarbons (from the combustion of fossil fuel). The former has given rise to hazards in asbestos mines, asbestos textile works, and insulation work in the shipbuilding and construction industries and the latter to specific hazards in the manufacture of coal gas in coking ovens, in steel works, in aluminium foundries, and wherever substantial amounts of incompletely combusted fumes were released into the working environment. Much smaller numbers of men have experienced substantial hazards from radon in the air of mines (not only when mining radioactive materials, but also when mining haematite and fluorspar under conditions in which radon seeped into the mine air from streams and the surrounding rock), from the manufacture of chromates and chrome pigments, from the refining of nickel, from arsenic (in the manufacture of arsenical pesticides and in the refining of copper, which is always contaminated with arsenic), from exposure to bischloromethyl ether in the chemical industry and exposure to vinyl chloride, from the manufacture of mustard gas, and, to a small extent, from exposure to silica if sufficient to cause silicosis. In one extreme situation (in the cobalt mines of the Hartz mountains in central Europe, which were subsequently mined for radium and uranium), the absolute risk of contracting lung cancer due to the occupational hazard of radon was so large that more than half the workers contracted the disease. In several other situations with heavy exposure to asbestos or the early stages of nickel refining, the occupational hazard has affected as many as 20 to 30% of the exposed men.

Atmospheric pollution

Some of the materials responsible for these occupational hazards—particularly the combustion products of fossil fuels—are or have been widely distributed in the air of towns and it is still uncertain how far they have, in this way, contributed to the production of the disease in the general population. That lung cancer was more common in big towns than in small towns and rural areas is certain, but this held as strongly for Oslo and Helsinki, two relatively unpolluted cities, as for more polluted ones. Differences between the largest towns and the least populated areas have seldom been more than threefold and much of the difference can be accounted for by past differences in cigarette smoking, a habit that has tended to spread outwards from the major cities. Attempts to ‘allow for’ cigarette smoking have usually been inadequate, as it is impossible to take full account of such factors as the age of starting to smoke cigarettes, the amount smoked daily at different periods, and the method of smoking (number of puffs, depth of inhaling, etc.). It is clear, however, that in the absence of cigarette smoking any effect of urban pollution in developed countries is relatively small. Estimates, based on extrapolation from the heavy pollution with coal smoke that used to occur in large towns, suggest that in such towns it may have contributed, in synergism with smoking, to as much as 10% of the risk of lung cancer, but would have caused very little risk in nonsmokers. On this basis, the present levels of pollution with benzo[a]pyrene and the other known lung carcinogens in town air can be only very small. Modern pollution with ultrafine particles (<10 µm diameter) may, however, be more hazardous. Study of residents in six contrasting cities in the United States in which information about personal smoking habits had been obtained suggests that the risk in the most polluted city compared with that in the least polluted could be increased by about one-quarter in both smokers and nonsmokers. The position in some developing countries is different: notably in parts of China, where intense indoor pollution with smoke and fumes from heating and cooking more than doubles the risk of lung cancer in nonsmokers.


The effect of another form of pollution—that of indoor air with radon arising from naturally occurring radium in rock and soil—has been estimated by extrapolation from the effects of the much larger doses to which some groups of underground miners have been exposed, and by direct observation in studies of people with and without lung cancer. These studies suggest that indoor radon may contribute to about 3% of lung cancers in the United Kingdom and about twice as much in the United States. The absolute effects are far greater in smokers than in lifelong nonsmokers, so that in the absence of smoking few cases would be produced.

Geographical differences

The development of the male lung cancer epidemic and the early signs of its departure have been most prominent in the United Kingdom and Finland, since the switch of young men to cigarettes was largely complete in these countries by the 1920s. In the United States, where cigarette consumption doubled during the Second World War, the peak of mortality occurred a little later. In some other developed countries, the development of the epidemic is still further behind and it is only just beginning to appear in many developing countries. For example, Chinese males, who now consume about 30% of the world’s cigarettes, experienced a 10-fold increase in cigarette consumption per head between the 1950s and 1990s that may well eventually cause almost a million cancer deaths a year when the young men of today reach middle age.

In women, the development of the epidemic has generally been later than in men. (Only in the Maori population of New Zealand did it occur at the same time.) In the United Kingdom, the United States, and a few other developed countries, the female lung cancer rates from smoking are already substantial, but in others, such as France and Spain, the epidemic in women has scarcely begun. The greatest recorded incidence in men worldwide is in African Americans in New Orleans and Detroit, and the greatest in women is in the Canadian Northwest Territories. A relatively high risk has long been noted in Chinese women who are nonsmokers, irrespective of their country of residence, which is probably due to their exposure to mutagens in the fumes from oils used in cooking with a wok and from the coal smoke with which many Chinese homes have been heavily polluted.

Pleura and peritoneum

  • 0.7% of all cancers and 1.3% of cancer deaths
  • Sex ratio of rates approximately 3:1 (but 6.5:1 for mesothelioma); age distribution like laryngeal cancer.

The existence of a specific type of tumour arising from the pleura, or less commonly the peritoneum, was debated by pathologists until 1960 when Wagner and his colleagues reported that six African patients with a similar type of ‘peripheral lung cancer’ had all lived in villages that were heavily polluted with dust produced by the mining of blue asbestos (i.e. crocidolite). Since then, occupational asbestos exposure (in asbestos mines, shipyards, building construction, asbestos product manufacture, and other work) has been shown to be responsible for the great majority of mesotheliomas, which are the predominant cancers of the pleura and peritoneum. They are much less likely to be produced by white asbestos (chryso-tile) than by brown asbestos (amosite) or blue, as the two last persist for longer in the lungs. A few cases arise from neighbourhood pollution with asbestos or secondary contamination (e.g. from household contact with asbestos workers) and some in Turkish villages are due to the weathering into the general atmosphere of erionite fibres in local rock and houses; these fibres are physically similar to asbestos although chemically different. A few cases have been caused by radiotherapy, and natural ionizing radiations may be responsible for most of those that are not associated with asbestos. An SV-40-like virus has been found in some tumours, but it is uncertain whether it plays a part in causing the disease.

Mesotheliomas seldom occur less than 15 years after first exposure to asbestos, commonly occur 25 to 30 years afterwards, and may be delayed for 50 years or more. Hence cessation of use of asbestos (peak imports to the United Kingdom, and peak production worldwide, were in the 1970s) will only lead to decreasing mesothelioma rates several decades later. In the last few years, the recorded mortality under age 70 in England and Wales has begun to decrease.

Almost all mesotheliomas are fatal. Due to confusion with lung or other types of cancer, it is still uncertain how many cases have occurred each year and some of the large increase in Western countries since 1960 may be artefactual. The highest recorded rates of mesothelioma incidence now are in men in Genoa, Italy and in Western Australia. Rates in women in Western countries tend to be much lower than in men.

Pleural mesothelioma is not related to cigarette smoking and the occupational hazard affects smokers and nonsmokers alike.


  • 0.1% of all cancers and 0.2% of cancer deaths
  • Sex ratio of rates 1.3:1; age distribution, 

Sarcomas can affect any bone, but characteristically affect the long bones in adolescence. After 45 years of age they occur most commonly in bones affected by Paget’s disease (osteitis deformans), which predisposes to sarcoma so strongly that as many as 1% of all people affected by the disease eventually develop a bone tumour.

Many different histological varieties occur, some of which appear to have different causes. Osteogenic sarcomas and chondrosarcomas are the most common, the former accounting for nearly all the adolescent peak. One rare type (Ewing’s tumour) occurs only in children and young adults and is almost unknown in black people, irrespective of the society in which they live.

Ionizing radiation is the main known extrinsic cause. Cases have been produced after high dose radiotherapy, especially such radiotherapy to childhood cancer patients, and after internal radiation from radionuclides including thorium in Thorotrast, an erstwhile contrast medium, radium in ‘luminizers’, once used to paint clocks and dials, and radium-224 therapy. Bone cancer risk has also been found raised after alkylating agent treatment of childhood cancers, and osteosarcoma risk is raised in several rare cancer syndromes, e.g. Li–Fraumeni syndrome (OMIM 151623), and in retinoblastoma patients.

National statistics in Britain record a reduction in mortality over the last 50 years, but are unreliable indicators of incidence as many deaths attributed to tumours of bone are due to cancers that have metastasized from other sites. The recorded decrease in mortality is, therefore, largely an artefact due to improved diagnosis (though it has been contributed to in recent years by higher survival rates in childhood) and the true incidence may have remained roughly constant. Internationally, bone cancer is relatively rare everywhere, with less geographical variation than for most cancers.

Connective and other soft tissues

  • 0.4% of all cancers and of cancer deaths
  • Sex ratio of rates 1.6:1; age distribution, 

Sarcomas of the soft tissues include a variety of different diseases, all of which are rare everywhere. Some occur in genetic syndromes, e.g. Li–Fraumeni syndrome and neurofibromatosis type 1 (OMIM 162200), and others are caused by ionizing radiation. A few might be caused by intensive immunosuppression or exposure to chlorophenols and related compounds, but the evidence is inconclusive.

Melanoma of the skin

  • 2.5% of all cancers and 1.2% of cancer deaths
  • Sex ratio of rates 0.9:1; age distribution,

Melanoma accounts for a small proportion of incident skin cancers but for most skin cancer deaths. Incidence, and to a lesser extent mortality, rates have been increasing in white populations as far back as data are available. In recently born generations in several populations, however, this trend has stabilized or reversed.

The incidence of the disease varies inversely with the amount of skin pigmentation, both comparing whites with nonwhites, and when comparing within whites, in whom skin sensitivity to sunshine (ease of burning and tanning) and fair or redhead complexion predict risk. Risk is also related strongly to numbers of benign moles and atypical moles on the skin, and less strongly to markers of cutaneous ultraviolet damage such as solar keratoses. There is a particularly great risk for patients with giant congenital naevi and those with xeroderma pigmentosum. In white people the tumour occurs most commonly on the legs in women and the trunk in men and is least common on the buttocks and soles of the feet (areas not exposed to the sun). In blacks, in whom melanoma is rare, a high proportion occur on the soles of the feet.

Incidence rates in white people vary roughly in proportion to the flux of sunshine (ultraviolet radiation) in the countries in which they live, although the reverse is true across Europe, probably reflecting darker complexions as one goes south. Risks rise in white migrants from countries with low insolation to those with higher insolation, especially if migration is at a young age. The greatest recorded incidence is in Queensland, Australia, where melanoma is the most common cancer (other than nonmelanoma skin cancer). For all skin sites combined, the incidence is not, however, greater in outdoor than indoor workers (rather the reverse, in fact, perhaps due to the protective effects of a semipermanent suntan). The totality of the evidence suggests that recreational intermittent exposure of untanned skin to solar ultraviolet radiation, such as when sunbathing, is the principal cause of melanoma and the reason for the rising rates. The relationship is not simple, however, and indeed melanomas of the head and neck occur typically in elderly outdoor workers and appear to relate to chronic ultraviolet exposure. There is suggestive, but not decisive, evidence that use of sunbeds increases melanoma risk, and inconsistent evidence that PUVA (methoxypsoralen UVA) treatment can do so.

Skin (nonmelanoma)

  • 20.4% of all cancers and 0.4% of cancer deaths
  • Sex ratio of rates 1.5:1; age distribution like oesophageal cancer

Nonmelanoma skin cancers are the most common cancers in fair-skinned populations, although rarely fatal. The predominant cause is sunshine (ultraviolet) exposure, and correspondingly the highest reported rates of incidence are in Australia. Rates have been rising in white populations across the world for many years, and tend to be greater in men than in women. The tumours are of two main types, basal cell and squamous cell carcinomas. The former, also known as rodent ulcers, have a causation that appears to relate to both cumulative ultraviolet exposure and intermittent intense exposures such as sunbathing. They occur mainly on parts of the body that are regularly exposed to the sun and, in particular, on the face, head, and neck. They are more common in outdoor workers, such as seamen and farmers, than in indoor workers; more common in fair-skinned (and blond and red-haired) than in dark-skinned (and dark-haired) people; and are almost unknown in blacks (except those who suffer from albinism). Some few cases have been produced by exposure to X-rays, but the risk is very small unless the dose is very large and they seldom occur after normal courses of radiotherapy. People who suffer from xeroderma pigmentosum, a hereditary condition in which there is a defect in the enzyme responsible for the repair of the damage done to DNA by ultraviolet radiation, develop large numbers of skin tumours at an early age in response to even quite mild sun exposure.

Squamous cell carcinoma is also produced by ultraviolet radiation, risk being proportional to cumulative sun exposure, and in PUVA-treated patients proportional to cumulative PUVA dose. It accounts for about 20% of cancers on ultraviolet-exposed skin. It is, however, the principal type of skin cancer produced by various carcinogenic chemicals, and particularly by polycyclic hydrocarbons in the combustion products of coal. These chemicals have been responsible for the scrotal cancers of chimney sweeps, who accumulated soot in the folds of the scrotum; of mule spinners, whose clothes were saturated with carcinogenic oils; and of various other groups of workers whose clothes were contaminated with tar. They have caused (and still do cause) cancers of the forearm in industrial workers whose arms are regularly splashed with tar or carcinogenic oils, cancers of the groin in India, localized by the continued friction of the dhoti cloth, and cancers of the abdomen in Kashmir associated with the habit of carrying a kangri, or small stove, inside the clothes in winter to keep warm.

Squamous cell carcinoma has also been due to prolonged exposure to arsenic, which is excreted by the skin and in the hair, when it may be accompanied by arsenical pigmentation and keratoses. All these conditions have been produced by prolonged medical treatment with inorganic arsenic, which used to be prescribed for a variety of chronic conditions, by the consumption of well water from arsenic-rich soils, and by occupational exposure in the smelting of copper and cobalt (the ores of which often contain arsenic) and in the manufacture of arsenical pesticides.

How large a part human papillomaviruses play in the development of squamous carcinoma of the skin is unclear. The type 5 virus is responsible for the warty lesions of epidermodysplasia verruciformis, some of which progress to cancer, and other types of the virus may contribute to the greatly increased risk that follows the intensive immunosuppression given to permit the survival of organ transplants.

A third type is Kaposi’s sarcoma, which is now classed as a skin cancer. It is associated with AIDS when AIDS results from homosexual intercourse, but probably only when this is accompanied by orofaecal contact. Frequent at first, particularly in the United States, the association has become progressively less common. Before the advent of AIDS, Kaposi’s sarcoma was common in some parts of central Africa, where it occasionally affected children, progressed rapidly, and could account for as many as 10% of all hospital patients with cancer. Elsewhere it was rare, but indolent cases occurred occasionally in developed countries, principally on the legs of middle-aged and elderly men. The disease is initiated by infection with the human herpesvirus type 8, but cofactors are required for tumour development.


  • 12.7% of all cancers and 8.2% of cancer deaths
  • Sex ratio of rates 0.01:1; age distribution, 

Cancer of the breast is the second most common cancer in the world and the most common in women, with a million cases occurring per year. Incidence rates are greatest in Western countries, somewhat lower in eastern Europe, and much lower in Asia and Africa. The geographical differences are unlikely to be chiefly due to genetic factors, as rates in migrants from low- to high-incidence countries rise considerably, to levels intermediate between the two, and there is a further rise in succeeding generations. In many countries incidence rates have tended to rise slowly over several decades, but mortality rates have started to decrease in recent years in Western countries because of more effective treatments and perhaps the effect of screening.

Hormonal factors, particularly oestrogens, are important in the production of the disease. The duration of ovarian activity is relevant, as the disease is particularly common in women who have an early menarche and a late menopause (the former being more important than the latter). Pregnancy produces a short-term increase in risk, followed after a few years by a lifelong decrease, particularly after teenage or early adult pregnancies. The incidence in later life increases progressively with a woman’s age at the time of her first full-term pregnancy, being about three times greater when the first birth occurs after 35 years of age than when it occurs before 18 years. Full-term pregnancies after the first have an additional protective effect. Pregnancies that end in abortion have little or no effect, however, suggesting that the effects of pregnancy depend on the induction of lactation. The duration of lactation has an additional protective effect but is not marked unless it continues for a year or more.

Risk of breast cancer is raised in women with benign breast disease, the degree of risk varying according to the type of disease. Risk is also raised by alcohol consumption, by lack of physical exercise, and by ionizing radiation exposure at young ages, with particularly high risks in women given high dose mantle radiation for Hodgkin’s disease. There is not, however, good evidence for causation by any form of environmental pollution.

Parity and menstrual differences are insufficient to account for the large variations in the incidence of the disease between different countries, which seem to be correlated with a ‘high’ standard of living: i.e. with life in a developed country. Diet might play a part, but the evidence is complex and inconclusive. Obesity is associated with a reduced risk before the menopause, as it tends to be associated with ovarian dysfunction. After the menopause, obesity increases the incidence and probably the fatality of the disease. Height is associated with increased incidence both for pre- and postmenopausal women. Oestrogens prescribed medically, as hormone replacement therapy (HRT) after the menopause, increase the risk by about 2% for each year of use; combined with progestogens in the contraceptive pill they increase it by about 25% during use, but the increased risk gradually disappears over 10 years, when use is stopped, as it does after HRT is stopped. Tamoxifen, an antioestrogen prescribed for the treatment of breast cancer, reduces the subsequent incidence of the disease in the unaffected breast. As well as the relation to sex hormone levels, there is growing evidence that risk of premenopausal breast cancer relates to prior endogenous levels of IGF-1.

Breast cancer has been a particularly fertile area of genetic epidemiology in recent years, with the identification of several high risk genes, some related to clinical syndromes (e.g. Cowden’s disease, and carriage of an ataxia telangiectasia mutation), but most related only to cancer risk.

Uterine cervix

  • 0.8% of all cancers and 0.7% of cancer deaths
  • Confined to women; age distribution, 

Carcinoma of the cervix is the second most common cancer in women worldwide, and the most common in parts of Africa and Asia; it used also to be common in Europe and North America. It has always been rare in Jewish women and has tended to be less common in Muslim women than in women of other faiths living in the same country (e.g. Hindus in India).

Changes in incidence over time have been difficult to assess, partly because mortality data have not always distinguished between deaths due to cancer of the cervix and those due to cancer of the corpus (or endometrium), partly because the introduction of screening programmes has made it possible to diagnose and treat premalignant lesions (see below), and partly because hysterectomy for benign conditions has become progressively more common, with a corresponding reduction in the number of uteri in which the disease could occur. Despite these complications there can be no doubt that the disease has become substantially less common in Europe and North America than it was before the Second World War.

The rarity of the disease in Jewish women and its relative rarity in Muslim women suggest that male circumcision may reduce the risk of its development, but this is unlikely as the state of circumcision of her husband has no substantial effect on a woman’s risk of developing the disease in communities in which only some men are circumcised. Cleanliness is likely to be protective, as the disease is relatively uncommon in communities that practise ritual ablution before and after intercourse and, within each community, it becomes less common with rising socioeconomic status.

Squamous carcinoma, which constitutes the vast majority of all cases, is intimately connected with sexual activity. It almost never occurs in virgins and increases in frequency with the number of sexual partners that a woman or her partner has had and with younger age at first sexual intercourse. Almost all cases are attributable in part to infection with certain types of the human papillomavirus, most notably types 16 and 18. A vaccine effective against these two types has now been developed, and future vaccination should greatly reduce incidence.

The development of squamous carcinoma is preceded by pathological changes limited to the epithelium, known as cervical intraepithelial neoplasia (CIN) types I, II, and III. CIN III is associated with the same types of virus as squamous carcinoma, but CIN I and CIN II generally are not. The changes may progress from one to another, finally leading to carcinoma, but the early lesions (CIN I and II) commonly regress and even CIN III (previously known as carcinoma in situ) may do so occasionally. The lesions can be recognized in cervical smears and destroyed by lasers or extensive biopsy and the occurrence of clinical disease can be greatly reduced by the examination of all sexually active women every 2 or 3 years and the treatment of advanced CIN lesions.

Other factors associated with the production of the disease are high parity, the use of oral contraceptives, and cigarette smoking. Both of the latter tend to be associated with behaviour conducive to venereal infection, but it is uncertain whether this tendency can wholly account for their association with the disease. That smoking may be responsible for some cases is suggested by the presence of mutagens in the cervical mucus of smokers that are not present in the secretions of nonsmokers.

Adenocarcinoma of the uterine cervix is generally uncommon, but has become somewhat more common recently in several countries. It is related to human papillomavirus infection, but also appears to relate to factors similar to those for endometrial adenocarcinoma.

Endometrium (corpus uteri)

  • 1.8% of all cancers and 1.1% of cancer deaths
  • Confined to women; age distribution like cancer of ovary

The epidemiological features of endometrial cancer are in many respects the opposite of those of cervical cancer. Histologically, it is nearly always an adenocarcinoma. It is common in developed countries, especially parts of the United States of America, and rare in poor populations. It is inversely related to parity, but not otherwise related to coitus, and is unaffected by the number of sexual partners. Like cancer of the breast, it is positively associated with late menopause, and perhaps with early menarche. Incidence in most Western countries has been fairly stable in recent decades, but in the United States there was a peak in the 1970s followed by a decline, as discussed below.

The one factor known to produce the disease is regular exposure to oestrogens, unopposed by progestogens. This leads to endometrial hyperplasia and eventually, in some cases, to cancer. Known causes include oestrogen-secreting tumours of the ovary, the use of oral contraceptives in which oestrogens and progestogens are prescribed sequentially (types that have now been abandoned), the use of unopposed oestrogens to relieve menopausal and postmenopausal symptoms, and adiposity. The last causes the disease because oestrogens are produced in the body after the menopause in adipose tissue from the adrenal hormone, androstenedione. Tamoxifen, an analogue of natural oestrogens, which blocks oestrogen receptors in the breast and hence acts as an antioestrogen, can, due to differences between the hormone receptors in different tissues, have a pro-oestrogenic effect in some other organs, and increases the incidence of endometrial cancer in proportion to the length of treatment. Endometrial cancer risks are reduced in users of combined (concurrent oestrogen and progestogen) oral contraceptives.

It is improbable that oestrogens are initiating agents. They are not mutagens in vitro and the changes that took place in the incidence of the disease in the United States following the increase and subsequent reduction in the use of unopposed premarin (a conjugated oestrogen) for the treatment of menopausal symptoms occurred so quickly that they make sense only if oestrogens act on some late stage(s) of the carcinogenic process. Endometrial cancer risk is raised in women who are physically inactive and in women with diabetes, in each instance with evidence that the relation may be more than just a consequence of obesity. Endometrial cancer risk, at least postmenopausally, appears to be reduced among smokers.


  • 1.8% of all cancers and 2.8% of cancer deaths
  • Confined to women; age distribution, 

About 90% of ovarian cancers are of the surface epithelium, and the causes of the tumour that have been recognized may refer only to these. Ovarian cancer incidence is greatest in countries with a high standard of living; the highest recorded rates are in part of Switzerland, in Iceland, and in Israeli Jews born in Europe or North America. Incidence decreases progressively with increasing number of children. There is no strong relation to age at menarche or age at menopause, however. Risk of the disease is reduced by the use of oral contraceptives, more greatly with longer use, and seems to depend on the lifetime number of ovulations. Risk is decreased by tubal ligation and perhaps by hysterectomy, and increased by mutations in the BRCA1 and BRCA2 genes.


  • 10.0% of all cancers and 6.8% of cancer deaths
  • Confined to men; age distribution, 

Cancer of the prostate is the third most common cancer in men worldwide and is found mainly in Western countries. Rates are particularly low in Asia and some parts of Africa. It is more characteristically a disease of old age than any other cancer, so that it comes to play a much larger part in clinical experience as the proportion of old people in the population increases. It is unusual in that foci of cells resembling cancer can be found in a high proportion of clinically normal prostates, so that the recorded incidence is drastically increased by increasing the number of prostatic biopsies. Increases in incidence have been recorded in many Western countries. The introduction of prostate specific antigen (PSA) testing has given rise to considerable artefacts in recorded rates in the United States of America. Some increase in mortality had been recorded in Britain and the United States, but in the last 10 years there has been a modest decrease, and the weight of evidence suggests that the disease is principally due to factors that have affected society for many years. What these factors are remains obscure. Associations have been reported with both increased and decreased sexual activity and there is some evidence, not conclusive, for a reduced risk in men who eat more tomatoes and tomato products, the main source of lycopene. There is also some evidence for reduced risk in men who take considerable physical exercise. On general grounds it seems likely that the disease is dependent on sex hormone imbalance (particularly as castration or oestrogen administration slows the progression of clinical disease) but the nature of the imbalance is unknown. Decreased risks of prostate cancer have been found in patients with Klinefelter’s syndrome and those with diabetes; the former, and there is some evidence that also the latter, have reduced androgen levels. An association of prostate cancer risk has been found with prior raised circulating levels of IGF-1. Vasectomy was thought to increase the incidence of the disease, but probably does not.

Two epidemiological observations stand out: the exceptionally high incidence in African Americans (much the highest recorded), and the low incidence in Japan in contrast with other developed countries. Both may be partly due to genetic factors, but they are not wholly so, as Japanese and blacks have much higher rates in the United States than they have in Japan and Africa respectively.


  • 0.6% of all cancers and 0.1% of cancer deaths
  • Confined to men; age distribution, 

Testicular cancers are of two main types. Seminomas, which are the more common, have a peak incidence at about 35 years of age and teratomas, commonly called embryonal carcinomas in the United States, have a peak incidence about 10 years younger. Testicular cancer is the most common cancer in young white men in many countries, but much less common in nonwhite groups living in the same areas, except Polynesians. Tumours after 50 years of age are mostly lymphomas and are now classed as such. Both genetic and environmental factors are important. On the one hand, the disease is uniformly rare in black populations, whether in Africa or in the United States. On the other, it has increased in incidence over many decades in white populations around the world. In Britain, the increase began in the 1920s and affected first the higher socioeconomic groups. The increase trebled the mortality at 15 to 34 years of age and produced a sharp peak in young adult life that had not previously been present. Mortality has greatly decreased in recent decades, however, as treatment has improved. The disease is more common in more prosperous populations, greatest in parts of Switzerland. Testicular cancer risk is greatly raised in men with XY gonadal dysgenesis, and in brothers, and to a lesser extent fathers, of cases. The cancer is much more likely to occur in an undescended than in a normal testis (c.10% of cases in whites are in men who have had maldescent), and in a testis opposite one that has been cancerous, and is also associated with prior inguinal hernia, but otherwise its causes are unknown. The leading hypothesis, in part because of the age distribution and association with cryptorchidism, has been that the aetiology is prenatal, due to exposure in utero to raised maternal oestrogen levels during the first trimester of pregnancy. Potential prenatal factors have been extensively investigated, with the strongest evidence for reduced risk with late birth order and raised risk for dizygous twins and boys born prematurely, but none are established.


  • 0.1% of all cancers and of cancer deaths
  • Confined to men; age distribution like cancer of the salivary glands

Carcinoma of the penis is at all common only in some parts of tropical Africa and Brazil, where it has accounted for 10% of all cancers in men. It is avoided almost entirely by circumcision at birth and is very rare if circumcision is carried out in boyhood. Phimosis is a risk factor. In developed countries penile cancer is rare even in the absence of circumcision if the glans, coronary sulcus, and foreskin are kept clean.

The oncogenic types of the human papillomavirus (principally types 16 and 18) can usually be identified in the malignant cells and are important causes of the disease.


  • 1.8% of all cancers and 2.2% of cancer deaths
  • Sex ratio of rates 1.8:1; age distribution like liver cancer

Cancers of the kidney are of three main types: nephroblastomas (or Wilms’ tumours), adenocarcinomas (or hypernephromas) of the renal parenchyma, and transitional- and squamous-cell carcinomas of the renal pelvis. The first are limited to childhood, occur with almost equal frequency everywhere, and apart from a few of genetic origin, are of unknown aetiology. The second constitute by far the majority of all cases, are more common in Europe (greatest in the Czech Republic) and North America than in Africa and Asia, and have been increasing in incidence in many Western countries. Cigarette smoking is one cause, but the association is weak and it does not account for more than about one-quarter of the cases. Obesity is also a risk factor, and phenacetin-containing analgesics may have been but the evidence is less clear than it is for renal pelvis cancers.

The third type of renal cancer (carcinoma of the pelvis) constitutes some 10% of all cases. Three established causes are occupational exposure to the chemicals that cause cancer of the bladder, cigarette smoking, and the consumption of phenacetin in large enough amounts to produce analgesic nephropathy. In all three cases the hazards are relatively small (two to three-fold). A fourth cause, Balkan nephropathy, increases the risk several hundred-fold.


  • 2.8% of all cancers and 3.2% of cancer deaths
  • Sex ratio of rates 3.4:1; age distribution like oesophageal cancer

Cancer of the bladder is almost universally several times more common in men than women. Greatest recorded incidence in men is in parts of Europe, with low rates in Africa, Asia and much of South America, and in nonwhites compared with whites in the United States of America. The tumour can be produced by cigarette smoking, occupational exposure to a group of chemicals classed together as aromatic amines, infection of the bladder with Schistosoma haematobium, the use of phenacetin-containing analgesics, and the medical prescription of chlornaphthazine (N,N′-bis(2-chloroethyl)-2-naphthylamine) and cyclophosphamide, and ionizing radiation. There has been evidence for a relation to chronic consumption of inorganic arsenic, a contaminant of the water supply in parts of Taiwan and other countries. There has also been some evidence for an association with urinary tract infection, supported by raised risks in paraplegics, who tend to have frequent urinary tract infections. Most bladder cancers are transitional cell carcinomas, but those associated with schistosomiasis are characteristically squamous carcinomas. It is not surprising that the bladder should be affected by many chemicals, as any noxious small molecules in the blood will tend to be found at greatly increased concentration in the urinary tract. Cigarette smoke contains several mutagenic chemicals that enter the bloodstream and thence the bladder, so that when tested in vitro on bacterial DNA the urine of cigarette smokers is found to be mutagenic, while that of nonsmokers is barely active.


An occupational cause was first suspected in 1895 in Germany, when Rehn commented on a cluster of cases in men using aniline for the manufacture of dyes. Aniline, however, is not carcinogenic in experimental animals; more recent studies have failed to incriminate it epidemiologically, and it seems likely that other carcinogenic chemicals were present as impurities. Four aromatic amines that are carcinogenic in experimental animals have been shown to cause bladder cancer in humans: 2-naphthylamine, benzidine, 3,3′-dichlorobenzidine, and 4-aminobiphenyl. The first is one of the most powerful human carcinogens yet known and was responsible for the development of bladder cancer in all the 19 men who were employed in distilling it in a British factory. Its manufacture in Britain was stopped in 1949, but small amounts continued to be imported until the 1960s. Other aromatic amines that may cause bladder cancer include auramine, magenta, and, perhaps, 1-naphthylamine. The last is dubiously carcinogenic in experimental animals and it seems probable that the cases associated with its use have been due to a small proportion of 2-naphthylamine present as an impurity in the commercial material. These chemicals were used in the manufacture of dyes, in the rubber industry as antioxidants (1-naphthylamine and 4-aminobiphenyl) and hardeners (benzidine), and in laboratories as a reagent (benzidine). 2-Naphthylamine is also found in the combustion products of coal and may have been responsible for the hazard of bladder cancer in men who made coal gas. Various other occupational associations, including leatherworking and aluminium work, have been reported, but are less clearly aetiological. As many as 10% of cases were, at one time, attributable to occupational causes in Britain and North America; but the proportion should now be much less.


The most important cause numerically is cigarette smoking, which probably accounts for about half the total number of cases in Britain and North America. 2-Naphthylamine and 4-aminobiphenyl are present in cigarette smoke, but whether the amounts are sufficient to account for the carcinogenic effect is uncertain.


The medicinal causes have, by contrast, been responsible for relatively few cases. Chlornaphthazine was used briefly for the treatment of myelomatosis, until it was found to be metabolized into 2-naphthylamine. Cyclophosphamide is used primarily for the treatment of malignant disease, but it is also used as an immunosuppressant. In large doses it may cause sloughing of the bladder mucosa and, occasionally, cancer. High levels of consumption of phenacetin-containing analgesics led to bladder cancer as well as renal cancer risk, but these drugs have been banned in Western countries for the last 25 years.

Parasitic infection

Heavy infection of the bladder with Schistosoma haematobium has been found to be a cause of the disease, most notably in Egypt and Tanzania.

Artificial sweeteners

Artificial sweeteners came under suspicion as potential bladder carcinogens because of the results of animal experiments in which, first, mixtures of cyclamates and saccharin and then saccharin alone were shown to cause bladder cancer in rats. The human use of cyclamates was banned before saccharin came under suspicion and it now appears that the ‘positive’ results of animal experiments with cyclamates alone were due to impurities. Saccharin has been shown to cause bladder cancer in rats but the quantities that had to be given were large, constituting a few per cent of the feed. The human evidence is extensive and could hardly be more negative, except that it does not cover lifelong use.

Brain and other central nervous system

  • 1.3% of all cancers and 2.3% of cancer deaths
  • Sex ratio of rates 1.6:1; age distribution, 

Tumours of the brain and nervous system are of several different histological types, some of which may not be clearly either benign or malignant. One type occurs characteristically in childhood (medulloblastoma), another in adult life (glioblastoma), and a third (astrocytoma) at all ages. Despite the overall male excess, one type (meningioma) is more common in women. Recorded incidence of brain and other central nervous system tumours tends to be greatest in white populations in Western countries, but geographical variations are less marked than for most tumours.

A moderately large secular increase in incidence in old age has been recorded in many countries, which might be attributable to improved diagnosis with CT scans and MRI. Little or no increase in mortality has been reported in or before middle age and the recorded increases in incidence are certainly largely, and possibly wholly, artefactual. The only established external cause is ionizing radiation. No new environmental cause has been established, but many have been suspected without conclusive evidence, including electromagnetic fields associated with the use of electricity (50–60 Hz) and mobile telephones (cellphones). A small proportion of brain tumours are attributable to high-risk hereditary syndromes, most commonly neurofibromatosis. The presence of allergy is associated with reduced glioma risk, but the reason is unknown.


  • 0.5% of all cancers and 0.2% of cancer deaths
  • Sex ratio of rates 0.4:1; age distribution, 

The thyroid is particularly sensitive to ionizing radiation in childhood, but risks after adult exposures are relatively small. Substantial numbers of cases have occurred among the survivors of the atomic explosions in Hiroshima and Nagasaki, children who were exposed to large amounts of radioactive iodine following the Chernobyl accident, and young people whose necks were irradiated in infancy for the treatment of an enlarged thymus (a condition now considered to be perfectly normal, but at one time thought to be a cause of sudden death). Fortunately, the thyroid tumours produced by ionizing radiation are nearly all of the papillary and follicular types, which respond well to treatment. No external causes are known of the medullary and anaplastic types, which have a high fatality and occur only in adult life. Many medullary thyroid cancers, however, are inherited as an autosomal dominant, alone or as part of the multiple endocrine neoplasia type 2 (MEN2) syndrome. Papillary and follicular thyroid cancers are less often genetic. Associations of thyroid cancer risk have been found with various benign thyroid conditions, but causality remains unclear.

The disease is most common in women in several Pacific islands and in Iceland. Increases in recorded incidence have been seen in recent decades in the United States of America and several other countries; it is unclear how much of this is due to changes in diagnostic completeness and criteria.

Hodgkin’s disease (Hodgkin’s lymphoma)

  • 0.4% of all cancers and 0.2% of cancer deaths
  • Sex ratio of rates 1.3:1; age distribution, 

Hodgkin’s disease is best thought of as at least two diseases, one affecting primarily youths and young adults, the other primarily middle-aged and older people. This division is suggested partly by the existence of two peaks in the age-specific incidence rates, partly by the histological appearances (younger patients tending to have the nodular sclerotic form of the disease and older patients the mixed cellular form), and partly by the clinical distinction that young patients show mediastinal involvement in more than 50% of cases and infradiaphragmatic involvement in less than 5%, while the reverse tends to be true in older people.

There are several reasons for thinking that the characteristic type in young people is infective in origin. In developing countries, Hodgkin’s disease occurs in childhood, but as the standard of living rises, the childhood cases disappear and are replaced by a larger number, and a peak of incidence, in young adults. This is reminiscent of what happened to poliomyelitis in the first half of the 20th century and suggests that the disease may be due to a ubiquitous infective agent that tends to be contracted at older ages as hygiene improves, rarely causes Hodgkin’s disease, but is more likely to do so if infection is at an older age. Risk in young adults decreases with factors likely to facilitate early infection, e.g. late birth order. That the infectious agent is likely to be the Epstein–Barr virus (EBV or human herpesvirus type 4) is suggested by the findings that the incidence is increased 5 to 20 years after a clinical attack of infectious mononucleosis, that abnormal EBV antibody profiles can be present years before the disease, and that the virus is often present in the DNA of the malignant cells characteristic of the tumour (the Reed–Sternberg and tumour reticulum cells). As with other virus-induced cancers, there are likely to be cofactors (at present unknown) that determine whether cellular infection leads to the production of a malignant clone. In recent years a raised risk has been noted in AIDS patients, with EBV genome detected in the tumour in most cases.The principal trend in Hodgkin’s disease in Western countries has been a dramatic decrease in mortality as modern radiotherapy and chemotherapy have transformed it from a disease that was largely incurable before the mid 20th century, to one with high survival.

Non-Hodgkin’s lymphoma

  • 2.8% of all cancers and 2.9% of cancer deaths
  • Sex ratio of rates 1.4:1; age distribution, 

Non-Hodgkin’s lymphoma embraces several diseases with different histological appearances. The histological classification has, however, varied from place to place and from time to time, and it has been difficult to collect epidemiological information about the individual types.

One type that has been clearly distinguished is Burkitt’s lymphoma, derived from B lymphocytes. This affects children everywhere, but is common only in a few areas in which malarial infection is both heavy and widespread. In parts of Uganda, Tanzania, and Nigeria the disease is 100 times more common than in Europe and North America. In high-incidence areas, EBV can nearly always be recovered from the lymphomatous cells and part of its genome is identifiable in the cells’ DNA. Infection with the virus is, however, not necessary for the development of the disease, as some cases occur in its absence; nor is it sufficient, as infection is almost universal and occurs at a very young age in high incidence areas. It seems, therefore, that EBV is a potential cause and that its carcinogenic effect is precipitated by the intense stimulation of the reticuloendothelial system that is characteristic of heavy and chronic malarial infection.

Another type occurs as part of the adult T-cell leukaemia–lymphoma syndrome that follows infection with the human T-cell lymphotropic virus-1 (HTLV-1), especially after infection in childhood. The disease is common in South Japan and the Caribbean, but may occur occasionally anywhere.

A third type, primary upper small-intestinal lymphoma (PUSIL), affects young people in many populations with a low standard of living, not only in North Africa and the Middle East (where its frequency gave it the earlier name of Mediterranean lymphoma) but also in South Africa and Central and South America. Malnutrition is not, however, a sufficient cause as it is uncommon in Bangladesh and several other malnourished populations.

A fourth type, the mucosa-associated lymphoid tissue tumour known as a maltoma, occurs in the stomach as a result of H. pylori infection and can be cured by aggressive treatment of the infection.

The remaining lymphomas, which constitute the majority in developed countries, should probably be divided further. Some in childhood might be better classed with acute lymphatic leukaemia, from which they are distinguished arbitrarily only by the number of lymphocytes in the blood. At present, however, they have to be considered as a group. As such they constitute one of the few types of cancer that have been increasing in incidence at all ages in Western countries for several decades. Internationally, recorded rates are greatest in these countries.

Two factors that have contributed to the increase, but which cannot account for it all, are the use of immunosuppressive drugs and the spread of AIDS. Intense immunosuppression for organ or bone marrow transplantation is followed within 1 or 2 years by an increase in the incidence of the disease of the order of 50- to 100-fold; most of the tumours are EBV-associated and are reversible by cessation of immunosuppression, but there is also an excess of non-EBV-related tumours. Smaller increases follow the less intensive use of immunosuppressive drugs for the medical treatment of patients with rheumatoid arthritis and other similar conditions. Raised risks have been see in patients with several immunological diseases including Sjögren’s syndrome, systemic lupus erythematosus, coeliac disease, and dermatitis herpetiformis. AIDS leads to a 50 fold or greater risk in Western countries, with EBV often present in the tumour; a rare subtype is associated also with human herpesvirus type 8 infection. Greatly increased incidence rates, largely EBV-related, are also seen in a variety of rare hereditary disorders characterized by major immunological impairment, such as the Wiskott–Aldrich syndrome. There is also increased risk of non-Hodgkin’s lymphoma in patients who have had Hodgkin’s disease.

Multiple myeloma

  • 1.0% of all cancers and 1.7% of cancer deaths
  • Sex ratio of rates 1.5:1; age distribution like oesophageal cancer

Myelomatosis has been much easier to diagnose since marrow puncture and then serum electrophoresis became standard diagnostic tools and since the improvement in the management of renal failure, which is often the presenting symptom. As a result it is difficult to be sure whether the increase that was recorded until recently, in both incidence and mortality rates, was due solely to improved diagnosis, or whether it also reflects the introduction of major new causes into Europe and North America between the two World Wars. In southern Sweden, where there has been a long-standing interest in, and search for, cases of myelomatosis, no large increase was seen over the same period; the recorded rates were higher than in other developed populations, but in recent decades those in other populations have caught up.

The disease is uncommon in undeveloped areas, where it is almost certainly underdiagnosed. Genetic factors could be important, as in the United States it is twice as common in African Americans (who have the highest recorded incidence anywhere) as in whites, and is rare in Japanese and Chinese irrespective of where they live. Little is known about the causes of myeloma other than raised risk in patients with monoclonal gammopathy of undetermined significance (MGUS), some raised risk, although not entirely consistent, after ionizing radiation exposure, and probably raised risk in AIDS patients.


  • 2.0% of all cancers and 2.8% of cancer deaths
  • Sex ratio of rates 1.7:1; age distribution, 

Leukaemia may be divided primarily into chronic lymphatic leukaemia (CLL), chronic myeloid leukaemia (CML), acute myeloid leukaemia (AML), and acute lymphatic leukaemia (ALL). CML, AML, and ALL are, in turn, amalgams of two or more different types, with different causes, different age distributions, and different prognoses, but the distinctions between them are still undergoing evolution and, with the exception referred to later, the epidemiological descriptions of each subtype are unclear.

CLL increases progressively with age in the same way as myelomatosis and most of the common epithelial cancers. It is extremely rare in Chinese, Japanese, and Indians, which is presumably due to genetic differences in susceptibility as it continues to be rare in these racial groups even when they migrate to other countries.

AML occurs at all ages. It becomes slowly, but progressively, more common from childhood on and is the most common type in young adult life. In this age group, its incidence is probably less variable throughout the world than that of any other reasonably common type of cancer. CML, by contrast, is very rare in youth, but becomes more common than AML in later middle age. The few cases that occur in childhood should perhaps be regarded as constituting a separate disease, as they lack the Philadelphia chromosome that normally characterizes CML in adult life.

ALL is the most common type of childhood cancer. Three main types can be distinguished. Common (c) ALL arises from B-lymphocyte precursors and is responsible for a peak incidence of the disease at 2 to 3 years of age. Null ALL also arises from B-cell precursors, but lacks the common antigen and accounts for most cases in the first year of life. T-cell ALL occurs more or less equally at all ages in childhood. ALL in adult life can be either B cell or T cell.

Many causes of leukaemia are known. The most important is ionizing radiation, which causes all types except CLL. The leukaemia risk appears, and peaks, sooner after exposure than the risk for solid cancers. The sparing of CLL may be because the relevant stem cells are so radiosensitive that they are killed by small doses that would otherwise be carcinogenic. The other main types are induced by ionizing radiation more easily than most other types of cancer and constitute about 10% of all fatal cancers from exposure of the whole body to moderate doses.

Whether extremely low-frequency nonionizing radiation can cause leukaemia, particularly ALL in childhood, is uncertain. There is evidence to suggest that the risk of ALL is approximately doubled in the small proportion of children who have exposure to power frequency magnetic fields in their homes of an intensity greater than 0.4 µT. The evidence for a causal relationship is, however, inconclusive.

One type of the disease (adult T-cell lymphoma/leukaemia) is caused by a virus (HTLV-1) and has been described under non-Hodgkin’s lymphoma.

Other causes include smoking, which causes a small increase in myeloid leukaemia, several chemicals, and genetically determined diseases. The most important chemical is benzene, which is used widely in industry. Prolonged occupational exposure to large amounts has caused a substantial risk of AML (particularly one of its subtypes, erythroleukaemia) and, less commonly, acute lymphatic leukaemia. Many cases are preceded by periods of aplastic anaemia and there is still some doubt whether leukaemia can be caused by small doses.

Chemotherapy for malignancy can cause leukaemia, mainly AML but also ALL and CML. Several treatment agents are leukaemogenic. Treatment with alkylating agents, such as melphalan, lomustine, and mechlorethamine, greatly increases risk of AML, with a peak risk, as after radiotherapy, at about 5 to 9 years after treatment. Topoisomerase II inhibitors, particularly the epipodophyllotoxins, are leukaemogenic, with leukaemia often occurring within a year or two of treatment. There is evidence of raised risk after treatment with platinum compounds. Patients with chloramphenicol-induced aplastic anaemia may also be at raised leukaemia risk.Of the hereditary causes, Down’s syndrome is the most common and is probably responsible for the greatest number of cases, although the relative risk in some of the other rarer syndromes may be greater than the 20-fold increase in childhood leukaemia that occurs with Down’s syndrome. Ataxia telangiectasia and Bloom’s syndrome predispose to ALL, while Fanconi’s anaemia predisposes to AML.

Further reading


Peto R, et al. (2000). Smoking, smoking cessation, and lung cancer in the UK since 1950: combination of national statistics and two case-control studies. BMJ, 321, 323–9.

Schottenfeld D, Fraumeni JF Jr (eds) (2006). Cancer epidemiology and prevention, 3rd Edition. Oxford University Press, New York.

Stewart BW, Kleihues P (eds) (2003). World cancer report. IARC Press, Lyon.

Swerdlow A, dos Santos Silva I, Doll R (2001). Cancer incidence and mortality in England and Wales: trends and risk factors. Oxford University Press, Oxford.

Tomatis L (ed.) (1990). Cancer causes, occurrence and control. IARC Scientific Publications No. 100. International Agency for Research on Cancer, Lyon.