The incidence of all types of cancer varies, not only with the sex and age of individuals but also at each age in different communities. This variation may be 100-fold or more, as with Burkitt's lymphoma, hepatocellular carcinoma, Kaposi's sarcoma, carcinoma of the renal pelvis, and basal cell carcinomas of the skin, or it may be less than two-fold, as with some childhood cancers and myeloid leukaemia in young adults. However, the incidence of all cancers that are at all common anywhere varies at least five-fold and most 10- or 20-fold. This variation is due to three factors: nature, or hereditary constitution; nurture, or the influence of the external world and of individuals' behaviour; and luck, or the play of chance.

Luck is often overlooked, but it is an important determinant of why one individual develops cancer at (say) 40 years of age while another lives twice as long in apparently similar circumstances without developing the disease. This fortune is partly due to differences in genetic susceptibility, but it is also due to luck. Even in the laboratory, genetically identical animals kept under conditions as uniform as possible develop cancer at different ages. The reason is easy enough to understand; the development of cancer is a process that requires several rare changes to take place in a cell before it is capable of giving rise to a malignant clone. Each change also has to evade the cell's mechanism for repair of the damage. If luck played no part, there would be no possibility of treating a patient with breast cancer by simple excision of the lump: all of the other stem cells in the breast have the same genetic constitution and have been exposed to the same general environment since the individual's birth. If there was a simple one to one relationship between genetic susceptibility and exposure in the development of cancer each stem cell would be expected to give rise to a malignant clone at about the same time, which fortunately they never do. The fewer the events that are required to transform a given cell into a malignant clone, the less opportunity there is for luck to play a part, and the more likely it is that cancers will occur in individuals at about the same age. This situation is seen in families with familial adenomatous polyposis and par excellence in the two eyes of children genetically susceptible to retinoblastoma.

Nature and nurture determine the probability that each individual will develop cancer; luck determines exactly who will do so and when. Although the role of luck in any individual case is large, it has little effect on determining the number of cases that occur each year in a large town and a negligible effect on the number of cases that occur annually in a whole country; the larger the population the more certain it is that good and bad luck will average out. In the rest of this chapter luck is, therefore, ignored, as it plays no part in determining the differences in the incidence of cancer in different communities that have led to the conclusion that most cancers are largely avoidable, in the sense that practicable modifications of behaviour and the environment could reduce age-specific incidence rates nearly everywhere by some 80 to 90 per cent.

The role of nature in determining the probability that an individual develops cancer may be large. Its role in determining the variation in the incidence of cancer in different communities is, however, small. Exceptions occur: skin colour plays a role in determining the risk of some types of skin cancer, and other hereditary factors may be responsible for differences in the incidence of chronic lymphatic leukaemia and (to a lesser extent) multiple myelomatosis in Europeans on the one hand and in Chinese and Japanese on the other. Nature, however, cannot play any part in determining the changes in the incidence of cancer that are occurring nearly everywhere nearly all the time, most markedly in cancers of the lung and stomach and in melanoma, nor can it account for the dramatic changes seen when migrants move from one country to another. Nature cannot be reasonably held to account for the striking correlation between dietary factors, such as fat consumption and the onset rates of cancers of the breast, large bowel, and body of the uterus. These correlations do not of course imply that fat in the food is the cause of these diseases (though it may well contribute to them) but they do imply that a large proportion of all these cancers is due to some variable factor which, if not fat itself, is correlated with it.

Many of the avoidable causes of cancer are now known, and together these account for nearly half of the cancers that occur annually throughout the world.

Some causes act synergistically with others, so that the risks they produce together are much greater than would be anticipated from their risks when acting alone. This has been established with combinations of smoking, alcohol, ionizing radiation, asbestos exposure, and some components of diet, and it probably holds for many other agents. That agents should interact in this way is understandable, when it is borne in mind that the production of cancer is a multistage process and that several different changes have to take place in a cell before a malignant clone is produced. In practice, this means that the risk of a particular type of cancer may be greatly reduced in several ways. For example, the risk of developing cancer of the oesophagus can be reduced to much the same extent by avoiding smoking or by avoiding alcohol, and the avoidance of both has very little more effect than the avoidance of one.

Ionizing radiation causes every type of cancer, except chronic lymphatic leukaemia and possibly Hodgkin's disease, because of its ability to penetrate to the nucleus of a cell and mutating the DNA. It acts synergistically with other agents, so that the effect of a unit dose on a given tissue is approximately proportional to the normal incidence of cancer in that tissue. It follows, therefore that a given dose will cause more cancers of the stomach in (say) Japan, where the disease is normally common, and fewer cases in the United States, where the disease is uncommon; the reverse will be true for cancer of the lung. Similarly, a given dose will produce more cancers in a tissue in which the incidence of cancer is normally high than in a tissue in which it is low. The precise quantitative relationship in each tissue has yet to be established, but it is assumed that most tissues are equally susceptible and that a given dose multiplies the normal incidence of cancer in that tissue by a standard amount. As with other carcinogens, there is a latent period of about 10 years before any effect is observed. The effect then increases for a few years and subsequently persists in a constant proportional relationship with the normal incidence of the disease. The actual numbers of cancers produced will therefore increase progressively with age of the irradiated subjects. On the basis of this simple relationship and the experience of survivors of the Hiroshima and Nagasaki atomic bombs, it has been estimated that the exposure of the whole body to 1 Gy of low linear energy transfer radiation, such as that produced by X-rays, &ggr;-rays, and cosmic rays, causes a life-time risk of death from cancer of approximately 10 per cent in a population with the age distribution of the United Kingdom. Larger and smaller risks occur in direct linear proportion to the size of the dose. At low delivery rates, such as those encountered in the nuclear industry and possibly also with low doses generally, such as those employed in normal medical radiography, some allowance can be made for the efficacy of cellular repair and the risk is estimated to be reduced to about 5 × 10&supminus;&sup4; cGy.

Increasing knowledge will certainly modify this simple rule and two tissues that react differently are already known. The cells that give rise to acute leukaemia and chronic myeloid leukaemia are more susceptible than average to the induction of cancer and the risk produced rises to a maximum earlier (in 5 - 10 years) and then falls practically to zero after 30 years, while the cells that give rise to chronic lymphatic leukaemia do not appear to be susceptible at all. This apparent absence of an effect is probably due to the particular susceptibility of these cells to destruction by irradiation, so that they are killed more easily than they are converted to malignancy. The thyroid also appears to be more susceptible than average, while tumours induced in the bone increase in incidence and then decrease, like those induced in the marrow.

Alpha- and &bgr;-radiation emitted by radionuclides deposited on the skin or absorbed by ingestion or inhalation penetrate only short distances into tissues; their effects depend on the physiological distribution of the elements throughout the body. The effect of &bgr;-radiation is much the same as that of an equivalent dose of X-rays. Alpha-radiation and neutrons, however, which give up their energy over short distances (high linear energy transformation radiation) have up to 10 or 20 times the effect of an equivalent dose of X- or &ggr;-radiation—something that is allowed for by expressing the radiation dose in Sieverts rather than in Grays.

When all types of natural radiation are taken into account it is estimated that the total dose (in conjunction with other agents) accounts for about 5 per cent of all cancer deaths in the United States and somewhat less in the United Kingdom , where radon levels are, on average, lower.

Electromagnetic radiation of lower frequency than ionizing radiation but with a frequency high enough to be detected as ultraviolet light penetrates only a short distance. This is, however, enough to cause cancers of the lip and skin.

Animal experiments provide no reason to think that still lower frequency radiation causes any risk of cancer, except possibly by some non-specific effect of local heating. No heating is produced by the extremely low frequency radiation that accompanies the passage of alternating electric currents, yet reports have been made that the resultant magnetic fields may cause cancer in children particularly cancers of the brain and leukaemia. The evidence is extremely weak and is unlikely to stand the test of time.

Animal experiments leave no doubt that nutrition is an important factor in determining the incidence of cancer in many, if not in all, organs. Carcinogens that cause only a small number of cancers in animals may cause several times more tumours in animals that are fed ad libitum rather than kept half starved. How far variations in the total intake of calories, within the normal human limits, determines the incidence of cancer in humans is, however, uncertain. Levels of nutrition that lead to obesity certainly increase the risk of some cancers, most notably cancers of the endometrium and gallbladder and, after the menopause, cancer of the breast. With respect to the endometrium and breast cancers, oestrogens are formed from androstenedione in adipose tissue, and this constitutes their main source after the menopause. A high intake of calories in childhood also increases the risk of breast cancer by bringing forward the age of menarche and possibly by increasing the size of the breast.

Whether fat, which in Western diets may account for some 40 per cent of calories, has a specific effect on the incidence of some cancers, apart from its contribution as a major source of calories, is still controversial. Ecological correlations suggest that fat may increase the risk of cancers of the breast, endometrium, ovary, colon, rectum, and prostrate: support for this hypothesis has been obtained in many but by no means all detailed epidemiological studies. The relationship between fat and cancer of the breast is controversial: some investigators believe that this relationship is due solely to confounding with total calories. The least controversial relationship is that between fat and large bowel cancer, where fat may have a specific effect by increasing the secretion of bile acids and consequently the promoting effect of bile acid metabolites in the stools. The type of fat is almost certainly immaterial. Polyunsaturated fats are necessary for the growth of cells, and the addition of small amounts of polyunsaturated fats to a diet from which they are otherwise absent increases the effect of carcinogenic agents in animal experiments, while similar amounts of saturated fat added to a diet deficient in saturated fat do not. Once a small amount of polyunsaturated fats is present, however, the type of fat that is subsequently added is immaterial and there is no reason to suspect that changes in the ratio of the types of fat in the diet, which have such a large effect on the risk of ischaemic heart disease, have any effect on the risk of cancer.

There is some suggestion that consumption of meat may increase the risk of rectal cancer, but the evidence is weak and any effect is probably due to confounding with fat.

The possible roles of starch and fibre are discussed under cancer of the large bowel.

Evidence is steadily accumulating that green and yellow vegetables and fruit contain some chemicals that help to reduce the risk of cancer in a wide variety of sites, if not in all sites. The effect of fruit is most marked on the stomach and green vegetables may have some specific effects on the large bowel. Some component of vegetables, possibly &bgr;-carotene or some other carotenoid, appears to protect against the development of nearly all of the principal cancers, possibly by scavenging free oxygen radicals. Vitamin A (retinol), which has been shown to be protective in many animal experiments, is unrelated to the incidence of cancer under the living conditions in developed countries, but many studies have shown that the risk of cancer increases inversely in proportion to the serum levels of &bgr;-carotene. Whether &bgr;-carotene itself is the protective agent should be shown by the large scale randomized prophylactic trial which Hennekens and his colleagues are carrying out in the United States with 20 000 doctors acting as experimental subjects. It may be, however, that some other carotenoid with which &bgr;-carotene is associated is the active agent. At present the best prophylactic advice is to recommend the regular daily consumption of green and yellow vegetables rather than any specific ingredient.

Apart from vitamin C and the carotenoids referred to above, there is some evidence to suggest that vitamin E, with or without the addition of selenium, also helps prevent cancer, but the evidence is not strong.

The grilling and roasting of meat produces a variety of mutagenic compounds, both polycyclic aromatic hydrocarbons and three-ring N-heterocyclic aromatic compounds, but there is nothing to suggest that these are the cause of any significant level of cancer in man. Likewise there is no evidence to suggest that traces of pesticides that are found in food are harmful. Natural foodstuffs contain many chemicals that are carcinogenic in animals when tested under standard conditions. These chemicals have, in many cases, been developed as natural pesticides and Ames has calculated that an average Western diet contains more than 1000 times as much of these ‘natural pesticides’ as it does contaminating man-made pesticides.

In the rest of this chapter the environmental and behavioural factors that are known or suspected to cause or prevent cancer are considered separately in relation to each of the principal organs in which cancer occurs. The description of each group of cancers is preceded by a Table giving the incidence of the main types in England and Wales and in Scotland and corresponding figures for some of the high and low incidence areas throughout the world.

The rates given under these latter heads have been recorded by Cancer Registries and are extracted from the publications of the International Agency for Research on Cancer (1987, 1990). They are not necessarily the most extreme levels which have been recorded, but have been selected as typical. For some cancers much more extreme rates have been recorded in special surveys of populations where no regular Cancer Registry is maintained: for example, in Iran, China, and most of Africa. Rates have, moreover, been cited only when the Registry has recorded more than 1000 cases of all cancers in each sex so as to ensure that the extreme rates are not greatly influenced by random variation. Despite the rigorous standards of registration that the International Agency requires before it accepts data for publication, some of the very low rates are likely to be due to incomplete registration. Incomplete registration probably also accounts for the fact that most of the rates for England and Wales are slightly lower than those for Scotland. More detailed information can be found in publications by the International Agency for Research on Cancer (1990), Doll and Peto (1981), and Schottenfeld and Fraumeni (1982).

Carcinoma of the lip was the first type of cancer to be related to an avoidable cause when, more than 200 years ago, it was attributed to smoking pipes. Subsequently it has been found to be produced, though less easily, by smoking in other ways and by exposure to ultraviolet light. In combination, tobacco and ultraviolet light account for the great majority of all cases and the reductions in pipe smoking and in outdoor work account for the current rarity of the disease in the United Kingdom compared with the past.

Carcinomas of the tongue, mouth, and pharynx (other than the nasopharynx) are all produced almost equally by the consumption of alcohol and by all methods of smoking. Alcohol and tobacco act synergistically, so that the avoidance of either has almost as much effect as the avoidance of both. The risks of all these cancers are greatly increased by chewing quids composed of mixtures of tobacco, lime, and betel nut or betel leaf. In some parts of India that such cancers account for 20 per cent of cancers of all types. Tumours occur particularly in parts of the mouth in which the quid is held for prolonged periods: this varies between individuals and between communities. Though commonly known as ‘betel chewers’ cancers', the risks are not eliminated by the omission of betel and the cancers are also common where chewing mixtures contain only tobacco and lime. These cancers have also been caused by occupational exposure in the production of mustard gas.

Carcinoma of the tongue used to be associated with syphilitic leukoplakia and the major reduction in incidence that has occurred in the United Kingdom over the past 60 years can be attributed in part to the virtual disappearance of tertiary syphilis. In the United States a recent increase in incidence is attributable to the increase in the use of oral snuff.

The nasopharynx is conventionally regarded as part of the aero-digestive tract, but it is not exposed to food and drink and would be better regarded as an extension of the nose. Cancer in the nasopharynx is common in south China and in Chinese who have emigrated from the area. It is moderately common in Malaysians, in Kenya, in parts of Northern Africa, and in the Alaskan Inuit and American Indians. In all other populations, it is rare. In populations with a high incidence of this tumour Epstein-Barr virus is integrated into the DNA of the tumours and high levels of antibodies against the virus are found in the patients' serum. Infection with Epstein-Barr virus alone is not sufficient to cause the disease and its development in the high risk Chinese populations is associated with the consumption of a special kind of salted fish early in life. Nitrosamines extracted from the fish have caused nasopharyngeal cancers in rats, but this observation has to be confirmed.

Cancers of the salivary glands are uncommon in all populations except Eskimos. No special causes are known, apart from a genetic factor that increases the risk of both cancer of the salivary glands and breast.

Carcinoma of the oesophagus, like cancers of the mouth and pharynx, is closely related to both smoking (of cigarettes, cigars, and pipes) and to the consumption of alcohol. In the absence of either the risk in the United Kingdom is reduced by about three-quarters; the risk is reduced only very little more by the absence of both, as the two factors interact synergistically. Nutritional deficiencies have probably also played a part in the past, particularly in women. A few cases arise in scars caused by accidental poisoning with corrosive chemicals and a very few cases as a regular complication of a rare hereditary type of tylosis with palmar keratoses.

Extremely high rates of oesophageal cancer occur in parts of Shansi and Henan provinces of China, on the east coast of the Caspian Sea, on the east coast of Lake Victoria, and in the Transkei region of South Africa, where the incidence in men and occasionally also in women, may equal the highest rates recorded for cancer of the lung in European and North American cities. In these areas, tobacco and alcohol are of minor importance, although when used they increase the risk; the principal causes are unknown. In China, there is evidence to suggest that mycotoxins produced by Fusaria fungi may play a part, but the only established factor, common to all the extremely high incidence areas, is a restricted diet, particularly poor in animal protein and green vegetables. Despite intensive research it has not been possible to implicate any of the nitrosamines that are such a prolific cause of the disease in animal experiments.

In Japan, a few cases have been related to the use of bracken fern as a vegetable.

Although cancer of the stomach used to be the most common type of cancer in the world, apart from the non-melanomatous skin cancers that are poorly enumerated; it was displaced by cancer of the lung in the mid-1970s. High rates persist in Central and South America, the former USSR, China, and Japan but elsewhere it has become much less common, most notably in the United States.

Only in tropical Africa has the incidence been generally low, but in some of these areas it is now beginning to become more common.

In the developed world, the disease has been associated with poverty, the incidence being five times greater in unskilled labourers than in members of the professions. No specific cause has been identified, but the disease is closely associated with consumption of food preserved in salt and, to a much lesser extent, with smoked food. Nitrites and nitrates, which can be converted to nitrates in vivo, have been thought to cause it because they can interact with secondary amines in the stomach and produce locally acting nitrosamines. Attempts to demonstrate such a relationship have, however, generally failed. What is certain is that the risk of developing the disease is reduced when the diet contains substantial amounts of fruit and fresh vegetables. These are the principal sources of vitamin C and vitamin C inhibits the formation of nitrosamines in the stomach.

Pathologically, carcinoma of the stomach is preceded by intestinal metaplasia of the mucosa or by atrophic gastritis, both of which may result from early infection with Helicobacter pylori. The risk is increased by gastroenterostomy but not by the presence of a chronic gastric ulcer. An early carcinoma may cause an ulcer, but chronic gastric ulcers do not become malignant and the two diseases are associated only to the extent that would be expected from their common association with low socioeconomic status.

Classed together as cancers of the large bowel, cancers of the colon and rectum are the second most common type of cancer in most developed countries. Both are less common in Eastern than in Western Europe and much less common throughout Africa and Asia. The incidence in Japanese who migrated to Hawaii soon came to equal that in the caucasian population and in the last two decades it has increased rapidly in Japan itself.

Burkitt's idea that cancers of the large bowel occurred because of a deficiency of dietary fibre, that is of vegetable fibre that is not digested by the alimentary enzymes, has proved too simple; however, the idea that the risk is increased if very little foodstuff reaches the large bowel is probably correct. Fibre, however, is not the only type of polysaccharide to do so. A variable proportion of starch is also not digested, the proportion varying with its origin and the method of serving (being, for example, relatively high if it comes from bananas and from potatoes if they are allowed to cool after cooking). Starch that is resistant to digestion in the small bowel and much of the dietary fibre serve as pabulum for bacteria in the large bowel and consequently increase faecal bulk and, in sufficient quantities, modify the pH. A high fat diet also increases the risk of cancer, possibly by stimulating the output of bile acids, the metabolic products of which act as promoting agents. Two additional minor factors are Brassica vegetables (cabbage and brussel sprouts) and alcohol, particularly in the form of beer. The former reduces risk and the latter may increase it, but the evidence that it does so is inconclusive.

Two chronic diseases of the large bowel which increase the risk of cancer are ulcerative colitis and rectal infection with Schistosoma japonicum. The latter is still very common in the Yangtze valley of central China and is the major cause of what is otherwise a rare disease in the rest of the country.

No occupational causes have been established except in so far as cancer of the large bowel is more common in people who lead a sedentary life. An excess has been recorded in asbestos workers, but this may have been an artefact due to misdiagnosis of peritoneal mesotheliomas and abdominal carcinomatosis secondary to cancer of the lung.

Despite all the similarities listed above, cancers of the colon and rectum are in some ways aetiologically distinct. Colonic cancer is more common in women than in men at young ages, particularly when it occurs on the right side, whereas rectal cancer is nearly twice as common in men. The greater relative incidence in women is reflected in a relationship with pregnancy, which suggests the possibility of a hormonal contribution, the disease being slightly more common in women who have not borne children than in those who have.

Pathologically, neither type of cancer often arises de novo but both arise predominantly in adenomatous polyps whether present singly, in small numbers, or in the large numbers characteristic of familial adenomatous polyposis and Gardner's syndrome.

Anal carcinoma is caused, in part, by infection with the same types of the human papilloma virus that cause carcinoma of the cervix. It is, in consequence, associated with anal intercourse in both sexes.

Hepatocellular carcinoma is the most common liver cancer, and its incidence varies enormously from one country to another, being rare in most developed countries and the most common type of cancer in men in parts of tropical Africa. In areas with a high incidence the disease is due to a combination of infection with the hepatitis B virus and exposure to aflatoxin, a powerful hepatic carcinogen in animal experiments. In these areas infection with the virus commonly occurs in infancy and persists in adult life. Aflatoxin is produced as a metabolite produced by the fungus Aspergillus flavus contaminating peanuts, maize, and other oil-bearing foods stored in hot and humid conditions. Experimental programmes to determine whether immunization against hepatitis B virus within a few days of birth prevents liver cancer are being carried out in Japan, Singapore, and the Gambia.

In the United Kingdom where food contains little or no aflatoxin, most liver cancers are associated with cirrhosis, whether this is caused by hepatitis, alcoholism or haemochromatosis. Rare cases may also occur in non-cirrhotic livers following the use of steroid contraceptives or anabolic steroids.

Cholangiosarcomas are less common and tend to occur in older people than hepatocellular carcinomas. In China, Korea, Japan, and Thailand they occur as a complication of infection with the liver flukes, Clonorchis sinensis and Opisthorchis viverrini.

A third histological variety of liver cancer, variously described as angiosarcoma or reticuloendothelioma, is so rare that no more than two or three cases occur annually in the United Kingdom. This type of cancer may be induced by inorganic arsenic given medicinally and thorium which was given as a radiological contrast medium in the form of Thorotrast. The medical use of these agents has now ceased. Vinyl chloride, used extensively in the manufacture of plastics, may also cause liver cancer but exposure is now so controlled that few if any cases continue to be produced.

Although cancers of the gallbladder and extrahepatic bile ducts are nearly always classed together, they differ aetiologically in many ways. The former is more than twice as common in women as in men, is strongly associated with obesity, and is nearly always preceded by cholelithiasis. The latter is more common in men than women; its incidence is increased by clonorchiasis, and possibly also by chronic ulcerative colitis. Both types of cancer are uncommon in the United Kingdom and not much more common elsewhere. The highest incidence rates in women, about four times the British rate, are recorded in Jewesses in Israel. In the United States the incidence has fallen sharply in the last 25 years due principally, if not entirely, to the high rate of cholecystectomy for gallstone.

A little over half of all cancers of the pancreas in the United Kingdom and the United States are caused, rather surprisingly perhaps, by the smoking of cigarettes. Tobacco smoke contains nitrosamines and many other carcinogens that can be absorbed and reach the pancreas via the bloodstream, but none of these has been shown to cause pancreatic cancer in experimental animals. The incidence of pancreatic cancer is doubled in diabetics and the highest rate (about double that in Britain) is recorded in New Zealand Maoris, who smoke heavily and are prone to obesity, diabetes, hypertension, and myocardial infarction. The disease is generally regarded as characteristic of the developed world, but the diagnosis is difficult in the absence of a well developed medical service and much of the recorded variation between communities and over time is probably due to differences in the availability of medical care.

Cancers of the nasal cavity and nasal sinuses are rare in all countries. An increased risk of such tumours is seen in individuals working in nickel refineries: nickel subsulphide, nickel oxide, and nickel sulphate are probably responsible. The manufacture of hardwood furniture, leather goods, isopropyl alcohol, and mustard gas has also caused increased risks. In some instances the risks have been increased several hundred times, and as many as 5 per cent of heavily exposed workers have contracted the disease. Most cases, whether occupationally induced or not, are squamous carcinomas, but those attributable to exposure to hardwood dust are characteristically adenocarcinomas.

A few cases are caused by tobacco smoke. The increased risk in smokers is, however, small, being of the order of 50 per cent; with such a rare disease, this is difficult to detect.

Causes of cancer in different parts of the larynx (the glottis and the supraglottal parts) differ. Glottal tumours are principally due to smoking, while supraglottal tumours, though related to smoking, are equally, if not more closely, related to the consumption of alcohol and in India to betel and tobacco chewing. The high incidence in northern Italy, France, and in the United States black population can probably be attributed to the combination of tobacco and alcohol, but it seems likely that some other factor, possibly nutritional, also contributes to the high incidence seen in S&atilda;o Paulo, Brazil, and to the relatively high rates that have been reported from the Middle East and Northern Africa.

A few cases have been caused by occupational exposure to carcinogens in the manufacture of mustard gas and isopropyl alcohol.

Bronchial carcinoma has been the most common type of cancer in the world since the early 1980s (excluding only non-melanomatous skin cancers which are not adequately recorded). It was a rare disease in all countries until shortly after the First World War, when the death rate attributed to lung cancer began to increase rapidly: some of this increase was an artefact due to improved methods of diagnosis, but much was real. In England and Wales the incidence in males reached a peak in 1973, when the disease accounted for 36 per cent of all cancer deaths in men and 9 per cent of deaths from all causes. Seventeen years later the male rates were falling at all ages and had fallen to less than half the maximum rates recorded at all ages under 50 years. In women the rate has continued to rise, except in the very youngest age group, and lung cancer now challenges breast cancer as the leading cause of cancer death. Similar changes have taken place in all developed countries, although, in most, the onset of the epidemic has been later and the peak has yet to be reached. The rapid increase in incidence, and the subsequent decrease in some countries, are attributable almost entirely to changes in the consumption of tobacco, the switch to cigarettes, and, most recently, to the reduction in the tar content of the smoke. At its peak, smoking was responsible for nearly 95 per cent of all cases of lung cancer in men in the United Kingdom and over 80 per cent of cases in women. The variation in rates throughout the world can be attributed, again almost entirely, to differences in the evolution of smoking: the exceptionally high incidence in Maori women in New Zealand, for example, is due to the adoption of smoking before the end of the 19th century. The incidence of lung cancer in non-smokers is almost identical in all countries, apart from a few places where very low overall rates are almost certainly due to incomplete case-finding. The exception to this is Chinese women, who have a two- to three-fold higher risk of lung cancer than other non-smokers. This appears to be due to the mode of cooking by using a wok and particularly to the use of rapeseed oil as a cooking oil.

Cigarettes have nearly always contributed more to the risk of the disease than pipes or cigars because their smoke, being acid, is less irritating and more easily inhaled. Nicotine consequently reaches the alveoli where it is quickly absorbed, producing a rapid and (to the addict) gratifying increase in its level in the blood. The smoke from pipes and cigars is generally alkaline and more irritating. Low tar cigarettes also contain low levels of nicotine: smoke has to be inhaled more deeply for the same amount of nicotine to be absorbed. Paradoxically, this can spare the bronchi from the deposition of smoke droplets and reduce the risk of bronchial carcinoma; however the same amount of absorbable carcinogens may be delivered to the pulmonary blood, causing the same risk to distant organs such as the pancreas and bladder.

If the smoking habit is stopped an increased risk of cancer persists, due to the damage done by initiating agents; the risk, however, does not continue to rise, with age, showing that the smoke must also act as a promoter. Smoking also acts synergistically with other agents, such as asbestos and ionizing radiation, increasing their carcinogenic effect more than would be expected from the addition of their independent effects.

Many other causes have been discovered as a result of the risks observed in industry but only radon in the concentration that used to occur in the mines of Schneeburg and Jachymov in Central Europe has caused risks comparable to those of tobacco. High risks of lung cancer are also associated with prolonged exposure to asbestos and with the conditions that used to be found in some nickel refineries. Other bronchial carcinogens encountered in industry include the polycyclic aromatic hydrocarbons, produced by the combustion of fossil fuels, particularly in the manufacture of coal gas and coke, but also in iron foundries and aluminium refineries; arsenic trioxide, encountered in the manufacture of arsenical pesticides and the refining of copper; hexavalent chromium compounds, encountered in the manufacture of chromates from chrome ore; and zinc chromate, bischloromethyl ether, and mustard gas. Many other agents have been suspected to be carcinogenic and may have caused relatively small increases in the risk of lung cancer, which may, however, have been substantial in absolute terms as the disease is normally so common. These include man-made mineral fibres, crystalline silica, beryllium and beryllium compounds, diethyl sulphate, and formaldehyde.

Several carcinogenic agents are found in the general environment, either naturally or as a result of human activities. The most important of these is radon gas. The amount of radon in the atmosphere varies from one part of the world to another, depending on the nature of the subsoil, and it may build up in houses to levels comparable to those in some mines in which lung cancer has been observed to be an occupational hazard. Radon and smoking act synergistically, so that the effect in smokers is greater than that in non-smokers. Current estimates suggest that radon exposure may be responsible for 6 per cent of all lung cancers in the United Kingdom and some 10 per cent in the United States. Atmospheric pollution with coal smoke has been important in many countries in the past, and still is in some areas. At its height, atmospheric pollution with coal smoke was estimated to account for about 10 per cent of all bronchial carcinomas in cities. Coal smoke has now been displaced in the United Kingdom by the combustion products of oil and gas, and despite the enormous increase in traffic, the amount of polycyclic aromatic hydrocarbons released is much less and is unlikely to account for more than a small fraction of 1 per cent of all cases. Tobacco smoke in the environment is now an equal or greater hazard and is estimated to be responsible for about a quarter of all cases of lung cancer in non-smokers. In comparison, the hazard from asbestos in the environment, either from its use as a building and insulating material or as a lining for brakes, is trivial.

Not all of the histological types of bronchial carcinoma are produced equally by all known carcinogens. Smoking used principally to increase the risk of squamous carcinomas and to have very little effect on the incidence of adenocarcinomas, but with the introduction of low tar, low nicotine cigarettes this has changed. The low levels of nicotine are compensated for by deeper inhalation with the consequence that adenocarcinomas in the periphery of the lung are becoming relatively more common in smokers. Occupational hazards have generally caused all the main types of the disease, but radon and bischloromethyl ether have tended characteristically to produce small cell carcinomas, while cooking with a wok has principally caused adenocarcinomas.

Nearly all cancers of the pleura are mesotheliomas. About 80 per cent are caused by occupational exposure to asbestos or to similar exposure in unusual domestic or environmental circumstances; some arise from heavy exposure to ionizing radiation, and the remainder may be attributed to the small amount of asbestos that occurs ubiquitously in the general environment and to natural radioactivity. In some Turkish villages such cancers have been produced by a similar fibre, erionite, that occurs in the local rock. Mesotheliomas are produced much more readily by the straight fibres of amosite and crocidolite asbestos which persist in the body for decades, than by the curly and less persistent chrysotile fibres. Similar tumours occur less commonly in the peritoneum, and can be attributed to the same causes except that they may not be produced by chrysotile at all.

The incidence of cancer of the kidney has been increasing slowly in the United Kingdom predominantly in the elderly. About one-quarter of the common adenocarcinomas of the body of the kidney are attributable to smoking, as are two-thirds of the relatively rare squamous carcinomas of the renal pelvis. The increase in incidence may, therefore, be attributable mainly, if not entirely, to smoking.

Renal pelvis cancers are very common in villages along the banks of rivers in the former Yugoslavia, Bulgaria, and Romania where Balkan nephropathy is endemic. The incidence of the disease in these villages is about 100 times that seen in the United Kingdom. The cause is unknown. Some cases worldwide may have been a result of renal nephropathies induced by analgesic mixtures containing phenacetin.

Nephroblastomas are confined to children. The peak incidence occurs between the ages of 1 and 2 years, but some slow growing tumours continue to appear clinically up to 10 years of age.

Cigarette smoking is the most important cause of bladder cancer, accounting for about half of all cases in the United Kingdom and the United States: tobacco smoke contains small amounts of two of the aromatic amines that have caused a high risk of bladder cancer in the chemical industry and the urine of smokers is found to contain increased levels of mutagens.

The hazard in the chemical industry was due to at least three chemicals, the use of which has now been abandoned (2-naphthylamine, 4-aminobiphenyl, and benzidine). In one plant all 15 men who distilled 2-naphthylamine developed bladder cancer, as did the five men who were employed for 15 years or more in the manufacture of benzidine in another plant. Exposure to the dyes made from these chemicals could also cause cancer. Occupational bladder cancer also occurred in the rubber industry, due to the use of an antioxidant containing 2-naphthylamine, and in the retort houses of gasworks where the same chemical was present in the fumes from coal tar.

Bladder cancer has also been caused by two drugs (chlornaphazine, which was used briefly in the treatment of myelomatosis and cyclophosphamide) and occurs as a common complication of infestation with Schistosoma haematobium in Africa. In the latter case, the disease probably results from associated persistent bacterial infection in the bladder with the formation of nitrosamines by bacteria in vivo.

Several dietary causes of bladder cancer have been postulated, partly on the basis of animal experiments (cyclamates and saccharin) and partly on the basis of epidemiological observations (coffee). Intensive investigation has, however, failed to confirm any harmful effect of artificial sweeteners and it is possible that the weak association recorded with the consumption of coffee is due to confounding with smoking.

All these agents cause transitional cell carcinomas, except for schistosomiasis which causes squamous cell carcinomas.

Cancer of the breast is the most common cancer in women in the world (excluding the poorly recorded non-melanomatous cancers of the skin), accounting for nearly 20 per cent of all female cancers. It is more common in the developed countries than in the less developed countries of Africa and Asia and is increasing in incidence nearly everywhere, slowly in the former and rapidly in some of the latter. The occurrence of the disease is determined primarily by hormonal factors and it is more common in women who have an early menarche, a later menopause, or are nulliparous. Amongst parous women, the risk is decreased by an early age at first full-term pregnancy and independently, but less markedly, by increased parity and prolonged lactation. The precise relationship between the disease and hormonal secretion is unclear, but the risk is increased by the long-term use of oestrogens after the menopause and seems to be related to the amount of physiologically available oestradiol in the blood. The use of steroid contraceptives increases the risk in those under about 40 years of age, but not later.

Dietary factors modify the incidence of the disease, partly by determining age at menarche and rate of growth in childhood, partly by determining the degree of obesity, and possibly by some specific mechanisms. A large body size at all ages and obesity after the menopause increase risk, but before the menopause obesity is associated with late menarche, irregular periods, and a decreased risk. Whether dietary fat intake increases risk independently of total calories is still uncertain.

Cancer of the cervix is the second most common cancer in women in the world (excluding the poorly recorded non-melanomatous cancers of the skin). It was much more common in developed countries at the beginning of this century than it is now, and is still very common in Southern and Central America, tropical Africa, and India. In the United Kingdom the risk increased again in women who reached sexual maturity after about 1960.

The disease was long thought to be related to childbearing, as the incidence was much greater in multiparous than in nulliparous women. Childbearing, however, plays little or no part and the risk is determined primarily by the risk of infection with certain types of the human papilloma virus, something that is related to the number of sexual partners that a woman and, independently, her partner, may have. The viral types responsible are not those that cause vulval warts (types 6 and 11) but types 16, 18, and several others that are less common. Infection with these viruses is common and is not enough, in itself, to cause the disease. Two factors that interact with infection to increase the risk are cigarette smoking (by causing the excretion of mutagens in the cervical mucus) and the use of oral contraceptives. These cannot, however, be the most important cofactors as the disease was prevalent before they were introduced. Diet could be one factor, as the disease is associated with relatively low levels of serum carotenene, and some aspect of sexual hygiene could be another, as the disease is associated with poverty and, in some communities, with the absence of male circumcision.

Most cervix cancers are squamous carcinomas and only these have been related to sexual intercourse. The relatively rare adenocarcinomas are, however, equally associated with the use of oral contraceptives.

Endometrial carcinoma is epidemiologically almost the reverse of cervix cancer. It is rare in poor countries and common in the rich; common in nulliparous women and progressively less common as parity increases; unrelated to sexual intercourse and, like cancer of the breast, positively associated with early menarche, late menopause, and postmenopausal obesity. Its incidence is reduced by the use of combined steroid contraceptives and increased by the use of sequential contraceptives and postmenopausal oestrogens and by the presence of oestrogen-secreting ovarian tumours. Lastly, it is one of the few diseases that cigarette smoking helps to prevent. All of these relationships can be simply explained by the extent to which the endometrium is exposed to oestrogens unopposed by progestogens, cigarette smoking exerting its effect partly by bringing forward the menopause and partly by modifying the metabolism of oestrogen.

Choriocarcinomas occur in about 1 in 30000 pregnancies. Nothing is known about their aetiology except that they are somewhat more common in Philippinos, Malays, and Chinese than in other populations.

Adenocarcinoma of the ovary shares many of the epidemiological characteristics of endometrial cancer, in that the risk of developing the disease is greater in rich counties than in poor, increases with the length of time between menarche and the menopause, and decreases with increase in the number of pregnancies and the use of combined steroid contraceptives. It is not, however, affected by medication with oestrogens, and seems, rather, to be related to the total number of ovulations.

The many other histological types of the disease are rare and have not been clearly related to any of the above factors.

Prostate cancer increases in incidence with age more rapidly than any other type of cancer. It has, in consequence, become increasingly prominent as the proportion of old people in the population increases and it now accounts for 9 per cent of all cancers in men. Age-specific incidence rates are also increasing, but the significance of this phenomenon is difficult to interpret as foci of cells indistinguishable from malignant cells are found in a high proportion of apparently normal prostates in elderly men and the recorded incidence rate is largely dependent on the frequency of prostatic biopsy.

Hormonal factors presumably play an important part in the production of the disease, but none has been identified. Clues to aetiology may be found in the high rates recorded in North American black subjects and the low rates in the Chinese and Japanese. There may be genetic factors involved, but the incidence in the Japanese is now increasing and is higher in Japanese migrants to the United States than in Japan. As with breast cancer in women, the incidence of the disease is closely correlated with the per caput consumption of fat.

Testis cancers are either teratomas or seminomas. Both begin to appear in late adolescence, occur most frequently between 20 and 44 years of age (with teratomas peaking before seminomas) and then become progressively less common with increasing age. Both types have become progressively more common over the last 60 years in Europe and North America but have continued to be rare in black people, Chinese, and Japanese, wherever they live. The reason for the increase is unknown. The only known risk factors are incomplete descent of the testis which increases the risk ten-fold, and maternal obesity during pregnancy, which increases it by about 50 per cent. Incomplete descent of the testis has also become more common in the last two or three decades, but it still accounts for only a small proportion of cases.

Carcinoma of the penis, like carcinoma of the cervix, is closely related to infection with certain types of the human papilloma virus; but infection is not alone enough to produce the disease. The disease can be prevented almost completely by circumcision within the first week or two of life and reduced greatly in incidence by circumcision later in childhood. Within populations that do not practice circumcision, the variation in incidence may be related to the degree of personal cleanliness.

Hodgkin's disease consists of a variety of histological and clinical varieties that should properly be divided into at least two main types, one with a peak incidence in young adult life and one which becomes slowly more common with increasing age. In undeveloped countries the peak in early adults is absent, and there is instead a small peak in childhood. The change in age distribution with rising standard of living is reminiscent of poliomyelitis and suggests that the early type may be due to an infectious agent that becomes less widespread as social hygiene improves. A few cases may be due to the Epstein-Barr virus as the incidence is increased 5 to 20 years after an attack of infectious mononucleosis and the virus is sometimes found in the malignant cells. Clusters of cases in schools or young people in close social contact have been reported, but they are not normally characteristic of the disease and may have been due to chance.

One that can be distinguished histologically, clinically, and epidemiologically is the B lymphocyte lymphoma named after Burkitt. It affects children throughout the world, but is rare everywhere except in a few places where malarial infestation is heavy. In these areas, the incidence is 100 times that in Europe and North America. In the high incidence areas the disease is due to a combination of infection with the Epstein-Barr virus, which becomes integrated into the DNA of the prelymphocytes, and heavy malarial infestation which stimulates the multiplication of these cells. Lymphomas due to Epstein-Barr virus infection constitute a proportion of the lymphomas arising following intensive immunosuppression, whether due to drugs or to the development of AIDS.

Another lymphoma constitutes part of the adult T cell leukaemia/lymphoma syndrome that follows infection with the human T cell leukaemia/lymphoma virus. The disease is common in the south of Japan and in the Caribbean, but may occur occasionally anywhere.

Primary upper small intestinal lymphoma 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 its earlier name of ‘Mediterranean lymphoma’) but also in Africa south of the Sahara and in Central and South America. Malnutrition, however, is not enough to cause the disease as it does not occur in many countries where malnutrition is common, such as Bangladesh.

The remaining lymphomas, which constitute the vast majority of those observed in Europe and North America, may need to be further divided, but at present it is better to regard them as a single disease rather than to divide them histologically by criteria that vary from one specialist to another. As a group they are increasing in incidence and there is some evidence to suggest that they may be related to the widespread use of modern pesticides. No occupational risk has, however, been established.

The recorded incidence of myelomatosis has increased greatly in all developed countries over the past 50 years, but it is uncertain how far this is an artefact due to improvements in diagnosis and how far it reflects a real increase in the incidence of the disease. No occupational or other cause for the disease is known.

Leukaemia, like non-Hodgkin's lymphoma, is a constellation of diseases with different clinical and epidemiological characteristics.

Chronic lymphatic leukaemia increases in incidence with age, in the same way as most of the common epithelial cancers. It is common in Europe and North America and rare in India, China, and Japan, due presumably to a difference in genetic susceptibility as the difference persists in immigrants from low to high incidence areas. It is principally a disease of B lymphocytes in the high incidence areas and of T lymphocytes in the low incidence regions. No environmental or behavioural cause is known.

Acute lymphoblastic leukaemia occurs at all ages, but half of all cases occur in childhood with a peak incidence at 2 to 3 years of age. The cases causing the childhood peak derive from B lymphocyte precursors and most possess a distinctive antigen and are described as common acute lymphoblastic leukaemias. The disease is seldom diagnosed in undeveloped countries, but this may be an artefact due to the rapidity with which affected children succumb to infection. Acute lymphoblastic leukaemia in adults is a different disease. One type, which is almost if not entirely limited to migrants in the United Kingdom, is derived from T-lymphocyte precursors and has been described under non-Hodgkin's lymphoma.

The incidence of acute myeloid leukaemia increases slowly with age from early childhood on and is the principal type of leukaemia in early adult life. In this age group the incidence of the disease is less variable throughout the world than almost all other types of cancer, other than the very rarest. Chronic myeloid leukaemia, in contrast, is very rare in youth, increases in incidence with age more rapidly, and is more common than the acute form after about 50 years of age. Both types of myeloid leukaemia are more readily induced by ionizing radiation than other cancers and a high proportion can be attributed to natural background radiation. Myeloid leukaemia, and particularly its rare variant, erythroleukaemia, is an occupational hazard of workers exposed to benzene. It is slightly more common in cigarette smokers than in non-smokers, possibly due to the small amounts of benzene and radioactive polonium in tobacco smoke. Myeloid leukaemia can also be produced by several drugs, including chlorambucil, cyclophosphamide, melphalan, methyl-CCNU, myleran, treosulphan, the combination nitrogen mustard, vincristine, procarbazine, and prednisone (MOPP) (probably due to the nitrogen mustard component) and, if given in doses large enough to cause aplastic anaemia, busulphan. The risks are, however, small in comparison with the therapeutic benefits.

Most bone cancers are osteosarcomas and these are responsible for the peak incidence that occurs in adolescence and for the subsequent increase in middle and old age. The only known causes are the genetic abnormality that leads to retinoblastoma, occupational or medicinal exposure to radium, and the agent that causes Paget's osteitis deformans. The last accounts for all the increase in later life, so predisposing to cancer such that osteosarcomas develop in 1 per cent of all affected patients.

Connective tissue sarcomas are rare throughout the world. A relationship with exposure to chlorophenols and phenoxyacetic acid herbicides has been postulated but the evidence is inconclusive.

Basal cell and squamous carcinomas of the skin are common in white populations throughout the world. Nearly all the former and most of the latter occur on the face, head, and neck and are due to exposure to ultraviolet light. The risk decreases with increasing pigmentation and similar cases are almost unknown in black populations unless they suffer from albinism. As a result of a defect in DNA repair, both types of cancer occur in large numbers on the skin of people with xeroderma pigmentosum.

Squamous carcinomas of the skin also have many other causes. Occupational exposure to polycyclic aromatic hydrocarbons in coal tars and lubricating oils have, in the past, caused many cases on the forearms and scrotum, and local customs have caused the ‘kangri cancers’ of the abdomen in Kashmir (due to carrying a charcoal stove inside the clothes for warmth) and the ‘dhoti cancers’ of the groin in India (due to continued friction from the dhoti cloth). In tropical Africa, squamous carcinomas are a common complication of chronic ulceration of the legs. Squamous carcinomas may sometimes be due to infection with the specific types of the human papilloma virus that cause the flat warts of epidermodysplasia verruciformis in people with an inherited defect of cell-mediated immunity, and to infection by other types of the virus in patients treated with immunosuppressive drugs.

Melanomas are less common than basal and squamous cell carcinomas but more important because of their higher fatality rate. In white skinned populations, their incidence is closely correlated with latitude, being higher nearer the equator, and all are related, in one way or another, to exposure to ultraviolet light. Those on the head and neck, like the basal and squamous cell cancers in the same site, are due to chronic exposure. Those elsewhere in the body are more common in office than in outdoor workers, are associated with the frequency of sunburn, and are caused principally by excessive exposure of untanned skin. Changes in customary clothing and the frequency of sunbathing have, in consequence, led to a steady increase in the incidence of the disease in many countries over the last 70 years. In black African populations, who go barefoot, melanomas occur at the junction of the pigmented and unpigmented areas.

Kaposi's sarcoma is now classed as a skin cancer rather than a cancer of the connective tissues. Until the 1980s it had occurred in two forms. One was exceptionally rare, affected elderly men, and occurred everywhere, but particularly in Jews in or from Eastern Europe. The other, which affected principally young men in Central Africa, was so common in some parts that it constituted 8 per cent of all cancer patients in hospital. Since the early 1980s a third type has appeared in association with AIDS, mostly in men who contracted the disease from homosexual intercourse, less commonly in drug addicts, and very rarely in haemophiliacs. The frequency of the association with AIDS has been much greater in the United States of America than elsewhere and has diminished with the passage of time. This last type appears to be spread by orafaecal contact and is presumably due to a virus. If it is, it remains to be seen if the same virus is also responsible for the two types of the disease that occurred before AIDS appeared.

Brain tumours, like so many others, are a mixture of different histological types with different epidemiological features. Medulloblastomas occur only in childhood, glioblastomas only in adult life, while astrocytomas occur at all ages. Most types are slightly more common in males, but meningiomas are more common in females. Several occupational hazards have been suspected, particularly in the chemical industry, but none has been confirmed. The minimal variation in incidence in different communities suggests that environmental factors other than background radiation are unlikely to be of much importance.

The incidence of thyroid cancer is related to the intake of iodine in a complex way. Relatively high rates occur in association with endemic goitre in iodine-deficient areas, such as Switzerland and Columbia, but the highest rates are recorded in areas where the intake is above average, such as Hawaii and Iceland. The histological types differ, however, papillary carcinomas predominating in the deficient areas and follicular carcinomas predominating in the areas with excess. Medullary and anaplastic carcinomas are rare everywhere.

The only other known causes are ionizing radiation, which causes both papillary and follicular carcinomas, and the hereditary factor which causes medullary carcinomas of the thyroid as part of the multiple endocrine syndrome.

Eye cancers in adults are nearly all melanomas. They occur slightly more often than would be expected in rural areas and in people with blue eyes and fair complexions, which suggests an aetiological role for ultraviolet light, despite the filtering effect of the cornea and lens. Retinoblastomas occur in early childhood. The majority are bilateral and due to a parental mutation.

Doll R, Peto R. The causes of cancer: quantitative estimates of avoidable risks of cancer in the United States today. J Natl Cancer Inst 1981; 66: 991 - 1308.

International Agency for Research on Cancer. Muir C, Waterhouse J, Mack T, Powell J, Whelan S, eds. Cancer Incidence in Five Continents, vol. V. IARC Scientific Publication No. 78. Lyon: International Agency for Research on Cancer, 1987.

International Agency for Research on Cancer. Tomatis L, ed. Cancer: causes, occurrence and control. IARC Scientific Publication No. 100. Lyon: International Agency for Research on Cancer, 1987.

Schottenfeld D, Fraumeni JF. Cancer epidemiology and prevention. Philadelphia, PA: Saunders, 1982.