J Cancer Res Clin Oncol (1991) 117:177-185 017152169100035U

C~6eer ~esearch Clinical 9 a~a

9 Springer-Verlag1991

Guest editorial* Epidemiological approaches to primary and secondary prevention of cancer * * Anthony B. Miller Department of Preventive Medicine and Biostatistics, Universit~r of Toronto, Toronto, Ontario, M5S 1A8, Canada Received 6 December 1990/Accepted 12 December 1990

Summary. Primary prevention of cancer requires control of both involuntary and voluntary exposures. Involuntary exposures include carcinogens in air and water, and various forms of radiation. Often these exposures are difficult to characterise individually and difficult to study epidemiologically. Although it is unlikely that they account for more than a small proportion of cancers, it is important that we refine our techniques of study to facilitate their control. Voluntary (lifestyle) exposures are responsible for the majority of cancers. In many developed countries, tobacco accounts for approximately 30% of cancer deaths, and major public health endeavours are justified to reduce this toll. Dietary factors may be as important, with dietary fat the most important risk factor, vegetables and fruits being protective. In several studies, including a cohort study in Canada, dietary fat increases breast cancer risk, though other studies have been negative. The evidence for fat increasing the risk of colorectal is more consistent. Epidemiology has shown that secondary prevention of cancer is applicable by screening for breast cancer with mammography with or without physical examination in women age 50-69, and screening for cervix cancer in women age 25-60 with cervical cytology. Organised screening programmes are essential to ensure that a high proportion of women are screened, and that the tests are high quality with adequate quality control. Under these circumstances screening every 2 years for breast cancer and every 3 years for cervix cancer is costeffective. Screening for other cancers cannot be recommended currently. There is a time to effect that must be recognised in planning primary or secondary prevention. Full effect of most primary activities will not be achieved for decades, screening may require a decade. Available * The "Journal of Cancer Research and Clinical Oncology" publishes in loose succession "Editorials" and "Guest editorials" on current and/or controversial problems in experimental and clinical oncology. These contributions represent exclusively the personal opinion of the author The Editors ** Based on a Plenary Lecture presented at the 15th International Cancer Congress, Hamburg, 18 August 1990

knowledge must be applied now, however, to ensure the effect will eventually be seen, as is now occurring in some countries with the downturn in lung cancer mortality following smoking reduction in men. Key words: Cancer control - Breast cancer - Colorectal cancer - Primary prevention Screening

Introduction Our main objective for cancer is cancer control, comprising the whole spectrum of activities from primary prevention, through screening and early detection (secondary prevention), treatment, rehabilitation, palliative care and pain relief. In this paper I shall concentrate on the epidemiological approaches that will enable us to approach primary and secondary prevention of cancer; i.e. to try and remove the causes o f cancer, and practice effective screening and early detection. This is not to minimize the other approaches to cancer control, but just to emphasize primary and secondary prevention as those approaches which I believe are now becoming more and more important in leading to the eventual control of cancer. In most countries in recent years there have been increases in both incidence and mortality from cancer. In Canada, taking 1970 as the base year, there has been a gradual rise in cancer mortality in men, to a 12% increase over baseline in 1988 (National Cancer Institute of Canada, 1990). In women there was a fall leading to about a 4% reduction in cancer mortality in 1976, followed by a slow rise, to about 2% above the baseline in 1988. If we exclude lung cancer from the male rates we find that for a period things seemed to be improving to a plateau about 4% below baseline, but more recently there has been an increase to 1% above baseline in 1988. For women if we exclude lung cancer we find that there is a decrease throughout the time period, to about a 12% reduction below baseline in 1988. These differences in the sexes are reflected in differences in trends for specific

178 cancers, particularly colorectal and the lymphomas, and has elicited some comment over trends in other countries (Davis et al. 1990; Doll 1990). For the lymphomas the increase in men may represent a mixture o f increase in chemical exposure and the effect of the AIDS epidemic. However, the decrease in all cancer mortality in women after exclusion of lung cancer is particularly important because it may reflect emerging changes that relate to the subject of this paper. Nevertheless, the fact that lung cancer is increasing in women has to be regarded as one of our major failures in cancer control, and highlights the needs there are for much more important research in terms of changing behaviour, research that will inevitably involve collaboration between epidemiologists, behavioural scientists and others who are expert in working in communities.

Primaryprevention International variation in cancer incidence is well recognised, even the extent of differences that can be seen within large countries. It was these differences that led Oettl~ (1960) and later Higginson (1969) to state that approximately 90% of all tumours are influenced by exogenous factors. It has been the task of epidemiologists interested in the epidemiology of cancer to try and determine what these exogenous factors are. People have often misunderstood the statement, but it can be clarified if you distinguish between involuntary exposures, which are what the general public regard as associated with the environment, and the voluntary exposures associated with our lifestyle.

Involuntary exposures The main routes of exposure to the factors in the general environment are by inhalation and ingestion, though under special circumstances involving close contact exposure can also be through the skin. The exposures in the air are those derived from pollution from industry, those increasingly recognized from passive smoking, those derived from what many are now calling the sick building syndrome, to which formaldehyde is a major contributor, and sometimes from toxic waste dumps. Exposures from drinking water include the contamination of aquifers with chemicals from toxic waste dumps and the problem of chlorination by-products derived from the chlorination o f water to ensure its safety, which, when the water has a lot of humus in it, results in various halogenated chemicals many of which have been noted to be carcinogenic. For diet the public is often concerned about the question of additives and various contaminants though, as illustrated by Ames (1983), there is also the problem of natural carcinogens. There are a number of difficulties in studying involuntary exposures. The exposure itself is often ill defined and misclassification of the exposure is probable. To acquire historical data other than current is even more problematic, but clearly in terms of cancer historical data are important. It is likely, particularly if we are concerned

over exposures from toxic chemical dumps, that there will be a mixture of carcinogens. The cohort of people we want to study may be often incompletely ascertained and mobility, sometimes induced by concern over the exposure, may exacerbate those difficulties. This has been particularly true for the exposures that occurred in relation to the Love Canal in New York State (Janerich et al. 1981). In addition the population at risk, at least the population at risk to fairly intense exposures, is usually small, sometimes far too small for investigation by the routine methods of cancer epidemiology. Further, the end-points we may be concerned with following such exposures are several. Birth defects have been studied by many investigators as a possible early indicator of problems, and these are relatively immediate. For cancer of course there is a problem over the latent period, which can be long, but some exposures have been prolonged and ecological studies have suggested excesses of cancer in connection with some dump sites (Griffith et al. 1989), while there are some indications already that for some cancers, induced by unusual exposures, the latent period may not be as long as we have sometimes believed possible. As an example, an incident occurred in a city in Ontario which led to very heavy chemical exposure of some firemen, and there have been some rapidly progressive cancers occurring in that group, which justifies further investigation. Leukemia has often been studied, but of course it is relatively rare, though there are already a number of instances where it seems possible that leukemia has been increased in relation to exposures from a chemical dump (Fagliano et al. 1990; Lagakos et al. 1986). One of the major problems for scientists is the politics that surround these issues. There is a great deal of media attention. There is a great deal o f distrust of government and scientists and an enormous amount of public anxiety. And under these circumstances, even though people want to know the answers, they may not wish to collaborate with those who are trying to find the answers. When we get to the end o f an investigation causality is usually in doubt, and when we are dealing with clusters, chance is extremely difficult to exclude. This is one of the unresolved areas that those concerned with the control of cancer have to bear in mind. Even though the public may be misinformed over some of these exposures, it is beholden on us to try and reassure them and, if there are problems, to identify them and to institute appropriate curative actions.

Voluntary exposures In this section of the paper I concentrate on diet. I do this largely because o f the interest of myself and my group in diet and some recent findings of interest, but also because diet potentially is one of the major factors we have to be concerned about in primary prevention of cancer. This is not to minimize the importance o f tobacco. Clearly the evidence for tobacco carcinogenesis is much stronger than for dietary causation. There is absolutely no excuse for ignoring the evidence that has been accumulated by

179 so many. The action that is necessary seems obvious though unfortunately not often applied. In Canada we seem to come from a privileged society in that the public in general appears to have accepted to a great extent the message that the Canadian Cancer Society and others have been trying to bring of the causality of lung and other cancers and other diseases in relation to tobacco exposure. It comes as some surprise to us when we come to Europe to found ourselves bombarded with tobacco smoke to the extent we are, even in certain public areas during a cancer congress! So it is important to re-emphasize the importance of research in cancer control in terms of the measures that have got to be taken to alleviate the problems of tobacco. This is particularly an example of research that has to be done in conjunction with others, research that will involve epidemiologists and behavioural scientists, people working in the population. The priorities are to identify the means to prevent children and adolescents taking up smoking and to encourage adults to stop smoking. It is remarkable, as we have learned in Canada, how much you can accomplish once you have got the legislative bodies on your side. One of the major contributors to success in Canada was the almost single-minded approach by the Medical Officer of Health in the city of Toronto, who started a movement that led to municipal legislation and is leading towards making tobacco smoking socially unacceptable, a movement that built on the strong foundation laid by R.M. Taylor and others (Delarue 1990). The other voluntary exposures are less important in etiology, nevertheless there have to be major efforts in terms of alcohol, and certainly we have to try and change the attitude of the public towards exposure to sunlight. Table 1 updates an analysis, which I performed very much on the lines of Doll and Peto (1981) for the United States, looking at the reasonable ranges for the causes of cancer deaths in Canada. In a developed country like Canada it appears likely that diet is slightly more important than tobacco, though both together contribute to over half of the causes of cancer deaths. This message that we have been slowly learning in Canada, and I hope will be learned in Europe, has hardly begun to be learned in developing countries and has to be one of our major concerns in the future in relation to tobacco. Turning to diet, Carroll et al. (1968) and others (Armstrong and Doll 1975) pointed out the very strong association between total dietary fat intake, as measured by disappearance of fat in populations, and the age-adjusted death rate from breast cancer. This sort of association does not demonstrate causality but it can be regarded as an indication of the potential importance of dietary variation. More recently there have been further attempts to evaluate these international associations by Prentice et al. Table

1. Causes of cancer deaths in Canada (%)

Diet Tobacco Occupation Family history Alcohol

32 29 9 8 6

Parity Sexual activity Sunlight Drugs Radiation

4 3 1 1 1

(1988), who have shown that the nutrient that contributes most to the variation in breast cancer incidence in women age 45-69 is calories derived from fat, a highly significant association; whereas calories derived from protein, alcohol and carbohydrates do not seem to contribute to the association. There have been a number of case-control studies, which show that various approaches to evaluating dietary fat show a positive relationship with breast cancer (Miller 1990a). An example is a study in Israel where those who had the highest risk of breast cancer were those who had a dietary pattern involving high consumption of fat and protein and the lowest consumption of fibre, with relative risks of the order of 1.6 to 2.0, compared to those with low consumption of fat and high consumption of fibre (Lubin et al. 1986). So it was the dietary pattern that appeared to be important. Because cohort studies avoid the problem of recall bias, many felt that the potential asse~ iation of diet and breast cancer had largely been dismissed by the study of Willett et al. (1987). This reported the results at 4 years of a follow-up of 90 000 nurses, who were given self-administered dietary questionnaires in 1980 and for whom it was possible to characterise the diet in terms of the percentage of calories from total fat, ranging from a low of 32% of calories from fat to a high of around 44% o f calories from fat. If dietary fat had been responsible for the risk of breast cancer in this 4-year period, one would have expected to see not a reduction in risk with increasing consumption of dietary fat, a reduction which is not significant, but an increase in relative risk at the highest fat consumption levels of about 1.5 or higher. The fact that this was not seen was regarded as extremely important negative evidence against the association. However, more recently a combined analysis of 12 case-control studies from many parts of the world N o r t h America, Europe, A s i a - by Howe et al. (1990a), showed a different finding. Using modern analytical methods and analysing the individual data from these 12 studies with more individuals in the combined analysis than had ever been studied before, there was a highly significant association for total fat in increasing the risk for breast cancer by about 50% when you compare the high consumption levels to the low. There was also increased risk of calories themselves, but calories contributed by total fat seemed to show the strongest association. This was not found for protein or carbohydrates. Although the findings for premenopausal women as distinct from post-menopausal women moved in the same direction, they were not significant. Recently we have had the first results of an analysis o f what we call the NBSS Diet Cohort (Howe et al. 1991). The NBSS is a randomized trial of screening for breast cancer involving just under 90 000 women, who were enrolled between 1980 and 1985, in 15 screening centres in Canada. We received funding in 1983 to offer self-administered dietary questionnaires to these women and nearly 57000 of them agreed to co-operate. In the course of a follow-up to the end of 1987 over 1000 breast cancers have been diagnosed in the total group, and o f these 519 were newly diagnosed breast cancers, which occurred

180 Table 2. Relative risks for incidence of breast cancer for total fat, protein and carbohydrate intake Dietary factor

Relative risk"

95% Confidence interval

Total fat Protein Carbohydrate

1.35 0.86 0.72

1.00-1.82 0.35-2.16 0.52-0.99

Per 693 calories/day. All models include other sources of calories

after the dietary questionnaire had been completed. These 519 and 2 controls for each of the 519 cases have been the subject of the analysis. Table 2 summarises the findings for calories contributed by total fat, protein and carbohydrate. Overall the relative risk for total fat was almost 1.4, a 35% increase in risk for the comparison between those whose diets differed by 693 calories a day. After taking into consideration the calories contributed by other dietary factors, we find this is a significant association. When you look at calories contributed by protein there is no relationship; when you look at calories contributed by carbohydrate there appears to be a protective effect. So what appears to be happening is that fat calories are increasing the risk of breast cancer, and carbohydrate calories or other things associated with carbohydrates appear to be reducing the risk of breast cancer; and when you put all this together in a model you end up with these two significant associations. It is not clear whether the positive association we have found in our study as distinct from the negative association found in Willett's study is due to a slightly different way of collecting information, a different follow-up or different types of people completing the interviews, but nevertheless this is the first cohort study to show a positive association. It largely confirms the findings from the case-control studies. Other analyses show that one of the concerns over case-control studies, namely that there is recall bias (those who have a diagnosis of cancer being more likely to give a positive effect), does not occur in a secondary administration of a questionnaire in the women contributing to these data (Freidenreich 1990), so that we have some confidence that this is a real finding and that possibly some o f the negative findings in the past were due to misclassification rather than a true negative biological relationship. So in my view, the descriptive studies, the positive case-control studies, the combined analysis and now this positive cohort study combine to show that dietary fat is related in a causal way to breast cancer; and some other calculations suggest that it might be responsible for at least 25%, maybe nearer 50%, of the differential in breast cancer risk between countries (Miller 1978). The other main diet-associated cancer is colorectal cancer. For men and women in a case-control study in Canada that we reported on some years ago we found a significant association for increasing risk with saturated fat, and a possible trend in men for fibre as a protective factor, although not in women (Miller et al. 1983). Another case-control study to show a similar association was that reported by Graham et al (1988) on colon

cancer; a subsequent report on rectal cancer showed similar findings (Frudenheim et al. 1990). A dose/response relationship showing increasing risk with increasing consumption of dietary fat was found for both cancers. Both of our groups took a lot of care over the dietary instrument; some of the negative studies in the past were probably negative because they did not collect sufficiently detailed dietary data. Recently there has been an intriguing and important report from Whittemore et al. (1990), who studied the dietary factors that might be related to colon cancer simultaneously in North America and the People's Republic of China. The Chinese, like the Japanese in North America, have shown increasing risks of colon cancer with migration, and perhaps it is not surprising that within the group in N o r t h America there is a highly significant relationship between saturated fat intake and risk of colon cancer. There is an association in the People's Republic of China but it is not significant. This differential was to be expected; the people who migrated are in the process of changing their diets to the North American patterns. These are the people who show increasing risks with migration and these are the people who show the association with saturated fat but not with other sources of calories. This is extremely strong evidence compatible with the dietary fat hypothesis. There have been other studies of colorectal cancer (Miller 1990a). One conducted in the Mediterranean region of France around Marseille did not find increased risk for the main fat component of the diet (MacquartMoulin et al. 1986), but it is important to note that the main fat component in the diet in this part of the world is olive oil not saturated fat, and if anything it looks as though this protects rather than increases the risk. What was found was a protective effect for two different types of vegetables, the strongest effect in fact being for vegetables containing low amounts of fibre. In a study we conducted of gastric cancer in Canada in different parts of the country, we found that nitrites, smoked meats and smoked fish increased the risk, and that nitrates and vitamin C reduced the risk (Risch et al. 1985). It is o f some interest that this should be so. Many people think that nitrite and nitrate should work similarly. The point about nitrate ist that the main source in the diet is vegetables and, like vitamin C, nitrate appears to be protective. Perhaps it is vitamin C which is the protective factor. Be that as it may, the protective effect of nitrate is probably a reflection of the main source of nitrate; that is, vegetables. So for gastric cancer, although the factors that increase the risk are different from those leading to colorectal cancer and breast cancer, the factors that are protective begin to look similar. In a study we have just reported on of pancreatic cancer we found slightly different effects (Howe et al. 1990b). We did not find a significant effect of saturated fat or other fat sources. We did find an effect of carbohydrate, largely sugars as contributors to carbohydrates rather than complex carbohydrates, in increasing the risk of pancreatic cancer. There is a similar finding in a combined analysis of case-control studies of pancreatic cancer conducted in Canada and many other parts of the

181 world, coordinated by the IARC, which will be reported shortly. A study of oral and pharyngeal cancer also found a productive effect for increasing consumption of citrus fruits, dark yellow fruits, and cruciferous vegetables (McLaughlin et al. 1988). Indeed, for increasing numbers of cancers fruits and vegetables are found to be protective (Le Marchand 1989; Miller 1990a). It is not certain what the protective factors are, but this is an important thing to emphasize when considering recommendations concerning dietary modification. The Committee on Diet and Health of the US National Research Council (1989) looked not only at cancer but at cardiovascular disease and other chronic diseases. There were two major recommendations that were designed to reduce the risk of all these conditions. The first was "Reduce total fat intake to 30% or less of calories; reduce saturated fatty intake to less than 10% of calories; and reduce intake of cholesterol to less than 300 mg daily." Obviously for cancer the most important are the first two. The cholesterol recommendation is more in relation to cardiovascular disease, though there are some studies that suggest that cholesterol may increase the risk of some cancers (Miller 1990a). The second main recommendation was "Every day eat five or more servings of a combination of vegetables and fruits, especially green and yellow vegetables and citrus fruits. Also increase intake of starches and other complex carbohydrates by eating six or more daily servings of a combination of breads cereals and legumes." These recommendations are made largely because of the work of epidemiologists in trying to determine the factors that can be used in primary prevention. As I have already emphasized, the task is now to learn how we can deliver this to the population by working with others to do the appropriate research on behavioural change.

Secondary prevention In this section I am going to draw very heavily on the deliberations of the UICC Project on Screening for Cancer (Chamberlain et al. 1986; Day et al. 1986; H a k a m a et al. 1985; Miller et al. 1990; Porok et al. 1984). In terms of breast cancer screening, the HIP Study (the study of the Health Insurance Plan of Greater New York) is the longest running and perhaps the most well known. In the initial reports devoted to the first 5 years the major effect was seen in women over the age of 50 (Shapiro et al. 1974). In the more recent reports that cover an 18-year follow-up of a program, which involved four annual screens of mammography and clinical examination, we find very little difference in the apparent effect by age (Shapiro et al. 1988). In fact, the difference has been in the timing of these beneficial effects in that the beneficial effect in the women over the age of 50 appeared to begin around 3 or 4 years after the screening began, whereas the apparent beneficial effect in women under the age of 50 was very much later. There is still some dispute as to whether this is a real finding. Some of the more detailed analyses suggest this difference may not be entirely due to

screening (Miller 1990b), so we depend on other more modern studies to find out whether indeed there is a differential and, if there is a differential, when we might possibly see an effect in younger women. One of the most important of these studies, recently updated to give findings to 9 years after initiation of screening, is a trial of single oblique-view mammography conducted in two rural counties of Sweden (Tabar et al. 1989). The significant findings remain restricted to women aged 50 69, where there has been a 40% or a 35% reduction in breast cancer mortality in this 9-year followup period. There is no benefit seen as yet in women age 40-49 and the benefit in women over the age of 70 is small and not significant. A smaller trial is being conducted in the city of Malta6 in Sweden, in which double-view mammography was used (Andersson et al. 1988). Women in the age range 4564 were recruited, and the investigators reported their results by dividing the age group into two - less than 55 and 55 or more. There was a rather strange non-significant excess of breast cancer mortality in the first 5 years in the younger women, which disappeared in the next 5 years. There was no benefit in the first 5 years in the older women and then a reduction in breast cancer mortality, not significant with these numbers, but nevertheless compatible with the findings in the two-county study in Sweden. In the United Kingdom a non-randomized trial was conducted and so far the results from the overall trial have not been given in terms of age; they included women age 45 or more. Again though, when you look at the comparison of the screening centres that offered mammography every 2 years and physical examination annually, we find no benefit in the first 5 years but then a significant reduction in breast cancer mortality in women in the years 6 and 7 after entry (UK Trial of Early Detection of Breast Cancer Group, 1988). They also had two centres studying the effect of teaching breast self-examination to classes. They did not find a reduction in mortality in the first 5 years, and though there is a possible reduction after that it is certainly not significant. There are actually preliminary results from a study we have been carrying out with Dr. Gastrin in Finland which do suggest, on a long-term follow-up, some benefit of brest self-examination at least in women over the age of 55 (Gastrin et al. in preparation, 1991). These results and the results from earlier studies, were considered by the UICC Project this April (Miller et al. 1990) and we concluded: "Screening for breast cancer by mammography every 1 3 years can reduce breast cancer mortality substantially in women age 50-70. In women under age 50 there is little evidence for a benefit, at least in the first 10 years after screening is initiated." It is quite unclear at present why there is this age difference, but it is there. It was first seen in HIP, the European studies replicated it, the emerging results in Canada are also replicating it. It may be there is a problem over treatment of breast cancer. Even though you can find it earlier in younger women, definitive mortality reduction has not been demonstrated. So that, at present, policies for applying this epidemiological knowledge have to be re-

182 stricted to women over the age of 50; because if there is no benefit, or even if the benefit for younger women is substantially delayed, the cost effectiveness is clearly low and there are many things that have a higher priority in cancer control. In terms of screening for cancer of the cervix there has never been a randomized trial and here the expertise of epidemiologists and observational epidemiology has really been called into play. One of the most dramatic findings was an analysis conducted by my colleague Matti H a k a m a (1982) of the trends of incidence of cancer of the cervix in the Nordic countries. It showed that in Finland, for example, there was stability in the incidence of cancer of the cervix for up to a few years after they initiated their program of screening women every 5 years by invitation, and then a dramatic decline in cervical cancer incidence. Similar effects were seen in Sweden, and also in Iceland an initial rise (presumably cases were being found by screening), and then a dramatic fall; and in Denmark, where there was formerly a high incidence, a substantial reduction was seen to a level approaching that of other Nordic countries, again related to screening. But in the Nordic country that did not introduce organized screening programs, namely Norway, there was no reduction in incidence until towards the end of this time period, when possibly there is the beginning of an effect of what was in fact going on, unorganized screening, largely taken up by health-conscious women in the country. This is indeed one of the most dramatic examples of the application of policies in countries, contrasting findings between Norway and the other Nordic countries showing the effects of well-organized screening policies. In many countries of the world there are changes of cervical cancer incidence. Figure 1 shows the difference between the age-specific incidence for Cali, Columbia, the registry with the highest incidence of cervical cancer in the world in 1962-1964, and more recent data from 1977-1981. Screening has been introduced in that city and indeed much of this reduction is almost certainly due to the effects of screening. It is of interest that in older age groups they are now finding more cases, which were not found in the past. Figure 2 shows a similar comparison for Alberta, Canada, where again screening was introduced and after a period there was a dramatic reduction of incidence in the middle years of age.

400 -

One of the applications of case-control approaches that has been important has been the use of such studies to evaluate screening. Although there are a number of problems with this approach, the studies have all shown a reduction in risk of disease in those who had had a Pap smear in the past 5 years (Moss 1991). If there is no Pap smear history in 5 years we find the risk of cervical cancer approaches three times the risk of those who were screened. This persists when the analysis is standardized for other factors associated with the risk of cervical cancer (Clarke and Anderson 1979). This approach has been replicated in many studies and the body of information obtained from this, and also from some of the cohort studies, including studies we conducted in Canada, were brought together by the Working Group to Evaluate Cervical Cancer Screening of the International Agency for Research on Cancer (1986) to evaluate the expectation of different frequencies of screening from the data. F r o m these studies, in the absence of screening you can expect a cumulative incidence in women of 1575 per 100000, i.e. 1.6% of women in this age range in the absence of screening can be expected to develop cervical cancer. If you screen annually over the age range 20-64 years there is a dramatic reduction in incidence, about 93%, but this requires 45 tests per lifetime. If you screen as infrequently as every 5 years over a restricted age range you still have an important impact, 70%, with only 6 tests per lifetime. The important thing for most technically advanced countries is that you can do almost as well by 3-yearly screening as you do with annual screening but with substantially fewer tests in a lifetime. And even for developing countries you can do remarkably well with a limited number of tests applied at the right age. These data were incorporated in a report by a World Health Organization Meeting (1986), which pointed out that throughout the world cancer of the cervix is a major problem for which the only immediate remedy appears to be screening. An effect could be anticipated with a single screen appropriately timed in relation to a woman's life, probably around the age of 3 5 4 0 . But once a country can afford it they should increase their utilization of screening to every 10 years over a limited age range, then every 5 years, and eventually achieve the developed country level of 3-yearly screening, which is all one really

o

300o

oo

D Alta, 60-62 9 Alta, 78-82

9 Cali, 77-81 o

60-~

80 ~

0 Call, 62-64

-

40-

200-

N

n-" 1 0 0 -

20-

0

"" 20-

30-

40-

50-

60-

70-

80-

Age

Fig. 1. Age-specificincidence of cancer of the cervix in Cali, Columbia, 1962-1964 and 1977-1981

0

1

20-

1

1

1

30-

1

1

40-

1

1

1

50Age

]

1

60-

1

1

~

70-

80-

~g. 2. Age-specificincidence of cancer of the cervix in Alberta, Canada, 1960-1962 and 1978-1982

183 needs to do over the age range 25-60. This policy was largely endorsed by the U I C C Screening Project (Miller et al. 1990) and, at least in terms of this sort of approach, recently endorsed by the W o r k s h o p G r o u p (1990) in Canada. I emphasize that these policies are only going to work, however, if you ensure that you bring the women at risk in the population into the screening program. In our U I C C group in 1985 we pointed out what the elements of an organized screening p r o g r a m should be ( H a k a m a et al. 1985): that you identify the target population and the individual women within it; that measures are available to ensure high coverage and attendance. Y o u do not want repetitively to re-screen the same health-conscious women. It is the women who do not come in immediately who are the ones we are trying to get hold of. Y o u have to have adequate facilities, you have to have organized quality control, you have to have a carefully designed referral system for treatment, adequate facilities for diagnosis and treatment, and then you have to evaluate and monitor the p r o g r a m to put in corrective factors. The Nordic countries have been so successful because this is what they have been able to introduce, but in the rest of Europe and in N o r t h America we have not been able to do this. We are trying to begin the process now in Canada. N o n e of the potential approaches to screening for other cancers, colorectal, stomach, ovary, etc., has yet been shown to reduce mortality from the cancer and such screening cannot be recommended as public health policy (Miller et al. 1990). F o r colorectal cancer the necessary studies are in place, and m a y produce the data needed for policy decisions within the next 5 years. But for m a n y other sites, e.g. ovary and prostate, the necessary studies have hardly begun. Screening is an expensive use of health care resources, and should only be introduced if data that confirm that the p r o g r a m will be effective are available.

Table 3. Estimates of potential effects of cancer control measures on incidence of various cancers Site

Oro-pharynx, larynx and oesophagus Stomach

Colon and rectum Breast

Cervix Endometrium Ovary Prostate Kidney Lung Bladder Pancreas Liver

Action

Eliminate smoking and reduce alcohol, increase fruit and vegetable consumption Reduce nitrite, cured meats and salt-preserved foods, increase fruit and vegetable consumption Reduce fat and increase vegetables Adjuvant chemotherapy Reduce fat and increase vegetables Reduce obesity (postmenopausal women) Screening (age 50-69) Adjuvant chemo- and hormone therapy Screening Reduce obesity Reduce fat Reduce fat Eliminate smoking, reduce fat Eliminate smoking, reduce fat and increase vegetables Eliminate smoking, reduce dietary cholesterol Eliminate smoking, reduce calories (? sugar) and increase vegetables Reduce alcohol

PAR a (%)

Potentially preventable (%)

90

86

68

74

50

79

10b 27

75

12 25b 15b 60 30 ?c ?c 30 80 80 60

82 66 81 98 76 76 73

30

70

30

?c

a Population-attributable risk. Estimates for men except for breast, cervix, endometrium and ovary cancer b Reduction in mortality, not incidence No estimate of effect is available

Time to effect

There is a time scale for impact of these approaches, both to primary and secondary prevention, which we have to bear in mind. F o r all these factors, although we can expect a fairly immediate partial impact-except possibly for HBV vaccination, which we believe will result in the reduction o f liver cancer after about 30 years, a long time is going to be required to see the complete effect of our interventions. This is particularly true for changing diet and nutrition in terms of the changes one might hope to see in breast cancer incidence, because the m a j o r effect will only occur if you concentrate your initial changes on women in their teens and 20s. For colon cancer you m a y expect a more immediate effect; colon cancer would seem to be the index cancer to evaluate, as it changes in incidence relatively rapidly following migration. So we have a long time scale to bear in mind but that is no excuse for not introducing the measures which we believe are effective now in the population, because these changes introduced now can result in important dividends in the next century. Table 3 speculates on the possible effect of the

interventions that can be contemplated now. It is based on the published data from a n u m b e r of studies that enable estimates o f the population-attributable risk for a factor to be computed (the P A R column) and also comparison of the incidence of the relevant cancer between the registry with the highest rate and the registry with the lowest rate (the "Potentially preventable" column). It would seem that full application of available knowledge could m a k e an important impact on cancer incidence, the m a j o r contribution being f r o m primary prevention. W h a t can we expect from epidemiology in the future? I believe we are going to see a continuation of much of the present methodology but it will be m o r e sophisticated, particularly in terms o f exposure measurement. w e need to follow-up the existing cohorts - there are m a n y cohorts now in existence - groups of people with previous exposure identified, who when followed-up will continue to provide knowledge on cancer incidence and mortality related to these exposures. We hope that in the new cohort studies that will be initiated better measures

184 o f exposure, biological m a r k e r s o f exposure, a n d biological m a r k e r s o f genetic susceptibility will be o b t a i n e d , as it is increasingly clear t h a t cancer is a result, n o t j u s t o f external factors, b u t o f the i n t e r a c t i o n between external factors a n d the i n t e r n a l (genetic susceptibility) factors in the host. I believe there will be greater use o f selective int e r v e n t i o n s in r a n d o m i z e d c o n t r o l l e d trials. Such approaches have been extremely i m p o r t a n t in the evaluation o f screening. T h e y will be extremely i m p o r t a n t in e v a l u a t i n g different ways to try a n d induce the necessary changes i n c o m m u n i t i e s . I reemphasise that we need to c o l l a b o r a t e with b e h a v i o u r a l scientists a n d others in cancer c o n t r o l research a n d we need to have greater reco g n i t i o n by the public of o u r needs. It is i m p o r t a n t to p o i n t o u t that the r e g u l a t i o n s t h a t are stifling epidemiology in the F e d e r a l R e p u b l i c o f G e r m a n y did n o t exist in the G e r m a n D e m o c r a t i c Republic. O n e has to look very carefully at the possibility that some o f the epidemiology which has been possible a n d the c o n s e q u e n t advances in the G e r m a n D e m o c r a t i c R e p u b l i c will be p r e v e n t e d if the legal restrictions, which are s t r a n g l i n g epidemiology in the F e d e r a l R e p u b l i c o f G e r m a n y , persist in a u n i t e d G e r m a n y . This is true I m i g h t say also for some other countries in Europe. W e are f o r t u n a t e in C a n a d a that we have been able to avoid some o f these problems.

Acknowledgements. I should like to express my personal indebtedness to my colleagues on the UICC Project on Screening for Cancer (Jocelyn Chamberlain, Nick Day, Matti Hakama and Phil Prorok) for the work they have done over the last 8 years and the work I hope they will continue to do in the next 4. I want to express my personal indebtedness to my colleagues in Canada, particularly Geoffrey Howe, Meera Jain, David Burch, Harvey Risch, Cornelia Baines, Teresa To, and Claus Wall. And finally, I should like to express my personal indebtedness to the man who recruited me to Canada 19 years ago, Robert N. Taylor. Bob Taylor believed as a result of epidemiology, which he felt was important, that cancer was a people's disease, and if we are going to have a major impact it will be because of what we are able to persuade people to do. It is this sort of leadership in the UICC which has been so important in the fight against cancer.

References Ames B (1983) Dietary carcinogens and anticarcinogens. Science 221:1256-1264 Andersson I, Aspergren K, Janzon Let al. (1988) Effect ofmammographic screening on breast cancer mortality in an urban population in Sweden. Results from the randomized Malta6 mammographic screening trial (MMST). Br Med J 297:943 948 Armstrong B, Doll R (1975) Environmentalfactors and cancer incidence and mortality in different countries, with special reference to dietary practices. Int J Cancer 15:617-631 Caroll KK, Gammal EB, Plunkett ER (1968) Dietary fat and mammary cancer. Can Med Assoc J 98:590-594 Chamberlain J, Day NE, Hakama M, Miller AB, Prorok PC (1986) UICC workshop of the project on evaluation of screening programmes for gastrointestinal cancer, lnt J Cancer 37:329-334 Clarke EA, Anderson TW (1970) Does screening by "Pap" smears help prevent cervical cancer? Lancet II:l~4 Committee on Diet and Health (1989) Diet and health. Implications for reducing chronic disease risk. National Research Council, National Academy Press, Washington DC Davis DL, Hoel D, Fox J, Lopez A (1990) International trends in cancer mortality in France, West Germany, Italy, Japan, England and Wales, and the USA. Lancet I:336:474-481

Day NE, Baines CJ, Chamberlain J, Hakama M, Miller AB, Prorok PC (1986) UICC project on screening for cancer: report of the workshop on screening for breast cancer. Int J Cancer 38:303-308 Delarue NC (1990) Smoking or health: an early historical perspective. Annals Royal College of Physicians and Surgeons of Canada 23:431-435 Doll R (1990) Are we winning the fight against cancer? An epidemioligical assessment. Eur J Cancer 26:500-508 Doll R, Peto R (1981) The causes of cancer: quantitative estimates of avoidable risks of cancer in the United States today. J Natl Cancer Inst 66:111%1308 Fagliano J, Berry M, Bove F, Burke T (1990) Drinking water contamination and the incidence of leukemia: an ecologic study. Am J Public Health 80:1209-1212 Freidenreich CM (1990) Recall bias in the association of diet and breast cancer. PhD Thesis, University of Toronto Frudenheim JL, Graham S, Marshall JB, Haughey BP, Wilkinson G (1990) A case-control study of diet and rectal cancer in Western New York. Am J Epidemiol 131:612-624 Graham S, Marshall J, Haughey B, Mittelman A, Swanson M, Zielezny M, Byers T, Wilkinson G, West D (1988) Dietary epidemiology of cancer of the colon in Western New York. Am J Epidemiol 128:490-503 Griffith J, Duncan RC, Riggan WB, Pellom AC (1989) Cancer mortality in U.S. counties with hazardous waste sites and ground water pollution. Arch Environ Health 44:69-74 Hakama M (1982) Trends in the incidence of cervical cancer in the Nordic countries. In: Magnus K (ed) Trends in cancer incidence. Causes and practical implications. Hemisphere, Washington, pp 279-292 Hakama M, Chamberlain J, Day NE, Miller AB, Prorok PC (1985) Evaluation of screening programmes for gynaecological cancer. Br J Cancer 52:66%673 Higginson J (1969) Present trends in cancer epidemiology. In: Morgan JF (ed) Proceedings of the 8th Canadian Cancer Research Conference. Pergamon, Oxford, pp 40-75 Howe GR, Hirohata T, Hislop TG, Iscovitch JM, Yuan JM, Katsouyanni K, Lunbin F, Marubini E, Modan B, Rohan B, Toniolo P, Shunzhang Y (1990a) Dietary factors and risk of breast cancer: combined analysis of 12 case-control studies. J Natl Cancer Inst 82:561 569 Howe GR, Jain M, Miller AB (1990b) Dietary factors and risk of pancreatic cancer: results of a Canadian population-based casecontrol study. Int J Cancer 45:604~608 Howe GR, Freidenreich CM, Jain M, Miller AB (1991) A cohort study of fat intake and breast cancer risk. J Natl Cancer Inst (in press) IARC Working Group on Cervical Cancer Screening (1986) Summary chapter. In: Hakama M, Miller AB, Day NE (eds) Screening for cancer of the uterine cervix. IARC Scientific Publications no. 76. International Agency for Research on Cancer, Lyon, pp 133 142 Janerich DT, Burnett WS, Feck G, Hoff M, Nasca P, Polednak AP, Greenwald P, Vianna N (198l) Cancer incidence in the Love Canal area. Science 212:1404-1407 Lagakos SW, Wessen BJ, Zelen M (1986) An analysis of contaminated well water and health effects in Woburn, Massachusetts. J Am Statist Assoc 81:583-96 Le Marchand L, Yoshizawa CN, Kolonel LN, Hankin JH, Goodman MT (1989) Vegetable consumption and lung cancer risk: a population-based case-control study in Hawaii. J Natl Cancer Inst 81:1158-1164 Lubin F, Wax Y, Modan B (1986) Role of fat, animal protein, and dietary fiber in breast cancer etiology: a case-control study. J Natl Cancer Inst 77:605-612 Macquart-Moulin G, Riboli E, Corn6e J, Charnay B, Berthezene P, Day N (1986) Case-control study on colorectal cancer and diet in Marseilles. Int J Cancer 38:183 191 McLaughlin JK, Gridley G, Block G, Winn DM, Preston-Martin S, Schoenberg JB, Greenberg RS, Stemhagen A, Austin DF, Ershow AG, Blot WJ, Fraumeni JF (1988) Dietary factors in oral and pharyngeal cancer. J Natl Cancer Inst 80:1237-1243

185 Miller AB (1978) An overview of hormone-associated cancers. Cancer Res 38:3985 3990 Miller AB (1990a) Diet and cancer. A review. Acta Oncol Rev Oncol 3:87-95 Miller AB (1990b) Breast cancer screening. Who should be included? J Gen Intern Med 5:$19-$22 Miller AB, Howe GR, Jain M, Craib KJP, Harrison L (1983) Food items and food groups as risk factors in a case-control study of diet and colorectal cancer. Int J Cancer 32:155 161 Miller AB, Chamberlain J, Day NE, Hakama M, Prorok PC (1990) Report on a workshop of the UICC project on evaluation of screening for cancer. Int J Cancer 46:761 769 Moss SM (1991) Case-control studies of screening. In: Miller AB, Chamberlain J, Day NE, Hakama M, Prorok PC (eds) Cancer Screening. Cambridge University Press. Cambrisge (in press) National Cancer Institute of Canada. (1990) Canadian cancer statistics, 1990. Toronto, Canada Oettl6 AG (1960) Cancer in the South African Bantu. In: Raven RW (ed) Cancer progress. Butterworth, London, pp 23~240 Prentice RL, Kakar F, Hursting S, Sheppard L, Klein R, Kushi LH (1988) Aspects of the rationale for the women's health trial. J Natl Cancer Inst 80:802-814 Prorok PC, Chamberlain J, Day NE, Hakama M, Miller AB (1984) UICC workshop on the evaluation of screening programmes for cancer. Int J Cancer 34:1~4 Risch HA, Jain M, Choi NW, Fodor JG, Pfeiffer CJ, Howe GR, Harrison LW, Craib KJP, Miller AB (1985) Dietary factors and

the incidence of cancer of the stomach. Am J Epidemiol 122:947-959 Shapiro S, Strax P, Venet L, Venet W (1974) Changes in 5-year breast-cancer mortality in a breast cancer screening program. In: Seventh National Cancer Conference Proceedings. American Cancer Society Inc, New York, pp 663-678 Shapiro S, Venet W, Strax P, Venet L (1988) Periodic screening for breast cancer. The health insurance plan project and its sequelae, 1963-1986. Johns Hopkins University Press, Baltimore Tabar L, Fagerberg G, Duffy SW, Day NE (1989) The Swedish two county trial of mammographic screening for breast cancer: recent results and calculation of benefit. J Epidemiol Community Health 43:107-114 UK Trial of Early Detection of Breast Cancer Group (1988) First results on mortality reduction in the UK trial of early detection of breast cancer. Lancet II:411~416 Whittemore AS, Wu-Williams AH, Lee M, Shu Z, Gallagher RP, Deng-ao J, Lun Z, Xianghui W, Kun C, Jung D, Teh CZ, Chengde L, Yao XJ, Paffenbarger RS, Henderson BE (1990) Diet, physical activity, and colorectal cancer among Chinese in North America and China. J Natl Cancer Inst 82:915-926 Willett WC, Stampfer MJ, Colditz GA, Rosner BA, Hennekens CH, Speizer FE (1987) Dietary fat and the risk of breast cancer. N Engl J Med 316:22-28 Workshop Group (1991) Report of a National Workshop on screening for cancer of the cervix Can Med Ass J (in press) World Health Organization Meeting (1986) Control of cancer of the cervix uteri. Bull WHO 64:607 618