Med. Oncol. & Tumor Pharmacother. Vol. 2, No. 3, pp. 137-142, 1985 Printed in Great Britain

11736--{1118/85$3.1~) + .(HI Pergamon Press LTd.

F A T I N T A K E AND C A N C E R O F T H E G A S T R O I N T E S T I N A L T R A C T A N D P R O S T A T E

ROBERT MACLENNAN Queensland Institute of Medical Research, Bramston Terrace, Herston, Brisbane 4006, Australia

(Received 27 March; accepted23 April 1985) The epidemiological evidence for and against the postulated role of dietary fat in the aetiology of human gastrointestinal and prostate cancers is critically reviewed. Evidence for a causal association is inconsistent, and much is of low validity. Several studies of colorectai cancer provide evidence to refute the fat hypothesis, at least in some populations. The expanding interest in precursor adenomas and the possible role of fat in their aetiology has led to case-control studies of colorectal adenomas and clinical preventive trials to see if fat is related to risk, but virtually no results have yet been published. The few studies done in relation to prostate cancer justify further investigation of the role of fat, but there appear to be no putative mechanisms to explain an association. Key words: Dietary fat, Epidemiology, Review, Precursors, International.

INTRODUCTION

migrants to Australia from Southern Europe notably Italy, Greece and Yugoslavia.-" The possibility that migrants are in some way genetically selected samples would appear unlikely since similar changes occur in many different groups of migrants. A large proportion of the international differences in cancer occurrence can thus be attributed to environmental differences between countries. Many changes occur following migration. Hopkins 3 found that 68% of Italian migrants to Australia increased their meat consumption at a time when Australians decreased their intake. Italians also tended to increase their fat consumption and decrease sugar and starch, vegetables, fruit, salt and dairy products. Migration was the most common reason given for dietary change. Although this study did not document it, many migrant groups tend to follow traditional dietary patterns acquired prior to migration but changed by availability of ingredients and affluence. Thus Chinese migrants still follow a Chinese cuisine, and Italian migrants would continue to consume meals within a distinctly Italian repertoire. As shown by Hopkins 3 there appears to be a change in the quantities of many foods, with increases in the 'prestige' foods, notably meats and fats. It thus seems that the dietary changes in first generation migrants are more likely to be quantitative than qualitative. Among Japanese migrants to the U.S.A. colon cancer rates appear to shift towards the higher United States rates to a greater extent in males than in females, possibly reflecting

Human societies vary greatly in diet and range from primitive hunter-gatherers, nomadic pastoralists, simple 'Swidden' horticulturalists, to agriculturalists and modern industrial agriculturalists. Within this range human effort is reduced in more complex societies and is replaced by energy from fossil fuels. In all societies, however, meats and fats are highly valued foods and their low intake in poorer countries is determined, not primarily by choice but through lack of availability. Thus the types of food and quantities eaten vary greatly among different countries as does the incidence of most human cancers.

CHANGES IN MIGRANTS Our belief that much of this variation is attributable to environmental factors comes from studies of migrants where many cancers rapidly change towards the incidence rates of the countries to which they migrate. For large bowel cancer, this increase occurs relatively rapidly within a lifetime. ~ Colorectal cancer rates are very high in the English speaking world - - in Great Britain and her former colonies, notably the United States, Australia and New Zealand. The increases in colorectal cancer among migrants from China, Japan, Norway and Poland to the United States have been known for some 20 years. More recently it has been demonstrated that similar increases in colon cancer occur among 137

138

Robert MacLennan

differences in biological response to an altered diet, or males consuming more meats and fat. INTERNATIONAL CORRELATIONS The hypothesis that fat is causally related to cancers of the colon and rectum and prostate is based on international correlation studies in which total national food disappearance data are correlated with national cancer incidence or mortality. Information on food is generally available only for countries as a whole and data by sex and age, or on a regional basis are not usually available. Correlation analyses based on heterogenous populations are less reliable than those based on more homogeneous populations. Lack of available data has prevented correlation analysis being done in populations likely to show the greatest contrasts in diet and in cancer. Most analyses have been limited to relatively few countries (generally the same set). Such studies can at best give clues to causation and suggest possible hypotheses for further testing using other study designs. Correa 4 comprehensively reviewed epidemiological correlations between diet and cancer frequency. Prior to 1975 only cancer mortality data were available for correlation studies. Armstrong and Doll s used the newly available cancer incidence data from volumes I and II of Cancer hTcidence in Five Continents ~''7 and limited the ages to 35--64 years, thus reducing the effect of differences in level of diagnosis in different countries which could result from different utilization or availability of medical services for the aged. They noted the high correlation between many dietary variables, e.g. 0.70 between total fat and total protein intake. In general their dietary data were from the early 1960s and cancer data from the late 1960s. The strongest association found for gastric cancer was a negative correlation with total fat consumption. Several dietary variables were highly correlated with cancers of the colon and rectum including fat and meat, but that for meat was highest. The correlation of fat with colorectal cancer was reduced to an insignificant level by controlling for meat intake, but the correlation with meat was reduced only slightly on controlling for fat. There was a negative correlation between cereals and colorectal cancer. High correlations were found between total fat disappearance and prostate cancer mortality, but not with incidence. Armstrong and Doll 5 believe that ascertainment of the true incidence of prostatic cancer is likely to be deficient, varying from country to country according to diagnostic practices. They note that although the incidence of latent cancer of

the prostate is similar in Japan to Hawaiian Japanese, mortality is very much lower in Japan. They postulate that environmental factors may affect the rate of progression of latent to invasive prostatic cancer. In an attempt to take account of the long latency between initiation and cancer diagnosis, Liu et al ~ correlated food disappearance data for 1954--65 and mortality data for 1967-73 from 20 industrialized countries. Their simple correlations were similar to those above. 5 By multivariate analysis they found support for a dietary cholesterol and colon cancer hypothesis. Correa 4 found strong correlations between death rates for cancers of the colon and prostate, and consumption of total fat and of nutrients derived from animal sources especially beef, pork, eggs and milk.

REGIONAL CORRELATIONS Correlations based on a small number of homogeneous populations or population samples have given somewhat different results from tile international correlation analyses. In different regions of Great Britain, Bingham et al 9 found no significant correlation betwen fat intake and mortality from colorectal cancer, although in the United States Blair and Fraumeni ~~found that regions with high prostate cancer mortality among whites had high fat intakes.

ETHNIC CORRELATIONS Kolonel et al ~ used quantitative food consumption histories from 4657 adults 45 years of age and over from the five major ethnic groups and correlated them with corresponding population-based cancer incidence rates. Significant correlations were found between prostate cancer and fat (saturated, animal) and protein (animal, total); stomach with fat (fish only) and protein (fish only). There was no correlation of fat with cancers of the colon and rectum, a finding attributed to the low incidence of colorectal cancer in native Hawaiians of Polynesian origin despite their very high fat diets; and the high colorectal cancer rates in Japanese who had a relatively low fat intake. The Hawaiian findings are consistent with provisional data from New Zealand, where Maori populations (also of Polynesian origin) have low rates of colon cancer compared with high rates in Europeans t2 despite apparently similar high fat low fibre diets (Pomare, personal communication).

Fat a n d cancer

TIME TREND CORRELATIONS Wynder and Shigematsu ~3noted the low incidence of colon cancer and to a lesser extent rectal cancer in Japan, and that the Japanese diet contained about 12% of energy from fat, usually in an unsaturated form, in contrast to 40--44% fat in the American diet, nearly half of which was in saturated form. They further noted that second generation Japanese in Hawaii had comparable intakes to Americans. They postulated that the increase of colorectal cancer among Japanese immigrants in the United States was a most important epidemiological clue that could be accounted for by dietary changes. The data from this early study by Wynder and Shigematsu, when taken alone, "could not establish a major environmental factor or factors relating to the development of cancer of the large bowel". They concluded that the "dietary pattern that may fit the distribution of cancer of the large bowel includes a high intake of fats. The effect of this pattern appears to be more marked for cancer of the colon than cancer of the rectum". There was a three-fold increased per capita fat intake in Japan from 1949 to 1971, with a parallel large increase in mortality from prostate cancer, and colorectal cancer to a lesser extent. 14 Other analyses of the time trend data in the United States, England and Wales, Australia and New Zealand showed no consistent associations between fat and meat and colon cancer. 15

CASE-CONTROL STUDIES Wynder and Shigematsu ~3 state that dietary information on specific food items as obtained in retrospective histories is of little value in determining the influence of diet on colorectal cancer, and reviewed the apparent problems in obtaining such information. Despite this warning there have since been many case-control studies of diet and colorectal cancer with generally inconclusive results. They have been reviewed by Zaridze.16 Dietary methodology is not precise and associations are likely to be detected only where there are large constrasts in diet as in groups in transition such as migrants in Hawaii or rural to urban migrants in Greece, or where there are large risks from diet. Manousos et a117 in Athens found significantly more frequent consumption of meat among cases and reduced intake of vegetables, particularly beet, spinach, lettuce and cabbage. Their analysis was based on food items and not nutrients so that it is not possible to sharply distinguish between animal protein and fat. However, they conclude that their results seem to indicate that animal protein and/or fat and lack of

139

vegetables may represent independent, though converging, aetiological factors for cancer of the large bowel. Compared with colorectal cancer, there have been only few case-control studies of diet and carcinoma of the prostate gland. Graham et al h~ note the previous findings by Schuman et aP 9 and Rotkin 2~ that prostate cancer patients ate margarine and other fatty foods more frequently than controls, and report an analysis of data collected from 1957-1965 at the Roswell Park Memorial Institute. The risk of prostate cancer increased with the frequency of ingestion of a number of food items, the highest risks being found with animal fat (RR 3.0). There was also an increased risk with frequency of intake of foods rich in Vitamin A (in contrast to Schuman et a119) and of those rich in Vitamin C. Although these data were collected using a brief interview with only a few questions on diet, the results are supported by preliminary results from a case-control study of prostate cancer in Hawaii. Kolonel et a121"22 found an increased relative risk for cholesterol intake and for total fat, and a similar positive association between Vitamin A intake and prostate cancer in men over 69 years of age.

RELIGIOUS GROUPS Religious groups with prescribed diets often have different cancer patterns from the general population and may provide unusual opportunities for studies of diet and cancer. The possibility that differences in mortality in such religious groups are due to self-selection of factors other than diet has been raised in studies of Californian Seventh Day Adventists. Adventists in California have around 70% of the rates of colorectal cancer among nonAdventists, but among Adventist physicians (mainly graduates of Loma Linda University, California), the cancer pattern was very similar to that in physicians graduating from the University of Southern California. 2s The diet of Adventist physicians was not documented and although they follow certain dietary prohibitions such as non-use of pork, the proportion following Adventist avoidance of meat was unknown. Phillips '-3 gives results from a small case-control study showing relative risk for colon cancer of 2.8 for past use of meat and 2.1 for a combined group of highly saturated fat foods. He suggested that the lacto-ovo-vegetarian diet might protect against colon cancer. If this were so, it could be due to protective factors in the diet other than reduced fat intake. Kinlen 24 investigated cancer mortality betwen 1911 and 1978 in two groups of

140

Robert MacLennan

enclosed religious orders for women; one of 1769 nuns who have no meat and one of 1044 nuns who ate little meat. Mortality from colorectal cancer was not significantly lower in either group than in the general population. Kinlen concluded that there are aetiologically important factors in diet other than fat or meat intake.

COHORT STUDIES Several sources of error in case-control studies, such as poor recall of past diet, and diseaseproduced changes in diet, are avoided in cohort studies in which the diet of well individuals is first documented and their subsequent cancer incidence determined. Unfortunately a large number of person-years of follow-up is required. Large cohort studies have been organized in Japan, Norway and Hawaii. All have documented diet initially but generally not during subsequent follow-up. Cancer occurrence has been determined in Norway and Hawaii by means of population-based cancer registries, and in Japan through mortality records. In Hawaii the initial examination and documentation of diet was done on 7554 from an eligible 11,148 men of Japanese ancestry born in the years 1900 through 1919 and living on the island of Oahu in 1965. After 15 years follow-up, Stemmermann et al 2s found a statistically significant negative association betwen colon cancer and the intake of saturated fat, whether assessed on the basis of grams per day or as a percentage of the caloric intake. There was a similar association with total fat intake when expressed in terms of percentage of total calories. The strongest negative relationship was found in cancer of the right colon. In contrast rectal cancer showed a weakly positive relationship to the intake of saturated fat when assessed on the basis of percentage of caloric intake. Previous studies among Hawaiian Japanese have shown that Japanese who subsequently develop myocardial infarction have previously reported a significantly higher fat intake as percentage of total energy. In 1960, 6763 Californian white male Adventists completed a dietary questionnaire. 2r Overweight men had a significantly higher risk of fatal prostate cancer (RR 2.5), and there was an increased risk for consumption of four animal products - - milk, cheese, eggs and meat. Relative risks for individual products were small (1.3-1.5), but combined high consumption of all four gave a predicted relative risk of 3.5. Another cohort analysis of Californian Adventists showed no relation of meat consumption to colon cancer. 27

METABOLIC STUDIES The chance observation in toxicity studies that cycasin produces intestinal tumours in conventional but not in germ-free rats stimulated the concept that intestinal bacteria might act on exogenous or endogenous substrates reaching the large intestine to form carcinogens. The first major study in relation to this concept was that of Hill et al 2s who demonstrated an association between colon cancer mortality and the faecal concentration of bile acids and the faecal degradation of cholesterol and its metabolites. The metabolites of bile acids have certain structural similarity to the carcinogen 3methyl cholanthrene. In further studies in Hong Kong, Denmark, Finland and New York 29-31 it appears that the total amount of neutral and acid faecal steroids is determined by fat intake, and that the concentrations in the faeces are related to faecal bulk which in turn is determined by other aspects of diet, notably dietary fibre. The search for mutagens is faeces has been disappointing and no consistent result has been found in studies of population samples in Africa, Denmark and F i n l a n d . Newmarket a132 have proposed that fatty acids from cellular and microbial debris in the colon and free bile acids in the ionized form are highly irritating to the colon epithelium. This results in cell loss and a compensatory increase in the rate of cellular proliferation for regeneration and repair, thus forming the basis of the promoting effect of fat. They have proposed the use of dietary calcium to increase calcium ions in the colon and thus form insoluble soaps of fatty acids and free bile acids. This hypothesis is perhaps supported to some extent by the comparisons of populations in Denmark and Finland where Copenhagen and rural Kuopio had similar total fat intakes, but the low risk Finnish population had much higher intake of milk and dietary fibre. 33 High rates for colon cancer in Chinese and Japanese in the United States are difficult to reconcile with dietary hypotheses. 3z The interesting question is why Japanese and Chinese in Hawaii and San Francisco whos~ diet is in transition have rates the same as or a little higher than those of U.S. whites. At issue is whether there are other dietary factors to explain this, or whether they have a greater genetic susceptibility. 34 Although one can attribute a large proportion of the international variation in cancer to differences in environmental factors, it is likely that much of the variation within a population is attributable to genetic differences. Although first degree relatives of colon cancer cases have a 2-3-fold increase risk of colon cancer, there

Fat and cancer

appear to be no increased risk in their spouses who presumably share a similar diet after marriage 35 (MacLennan et al, in preparation).

EXPERIMENTAL Animal experimental data were recently reviewed. 36 It was concluded that numerous experiments have shown that dietary lipid influences tumourigenesis in the colon and that dietary fat appears to affect tumour promotion rather than tumour initiation, although the latter cannot be excluded. The specific mechanism involved in promotion is unknown.

DISCUSSION Although there is a large literature suggesting that dietary fat might be causally related to colorectal cancer and possibly to cancer of the prostate, the evidence is of low validity with many of the positive associations being found in international correlation studies, Well designed cohort studies with distinct variation in fat intake within the cohort could contribute further to our understanding, although the Japanese data in Hawaii are against the fat hypothesis. The high correlations between fat and other nutrients suggests that only an experimental approach in humans is likely to resolve the issue. Due to the low incidence and the consequent very large numbers required, controlled clinical trials of fat reduction to reduce mortality from colorectal cancer or prostate cancer would be impracticable. Hence, for colorectal cancer, the study of precursor' adenomas may be rewarding, Adenomas have a relatively high prevalence in the population and are now routinely removed from the entire large bowel through flexible fibre-optic colonoscopy. Follow-up colonoscopy is necessary to remove new adenomas which appear following the complete removal of all visible lesions, and the cumulative incidence over 2 years is of the order of 40%, Several case-control studies are currently being done in several centres incuding Brisbane and New York and clinical trials, first done in Toronto, have also been started in the United States and Australia. A major aim of the Australian study is to see whether or not dietary fat is related to the growth Of colorectal adenomas. In our Brisbane case-control study we distinguish between patients coming for the first time to colonscopy and those seen at follow-up - - adenomas in the latter have oceurred during the preceding 2-3 years, and hence are mare likely to be related to promoting effects of recent diet.

141

Populations in transition, such as migrants from one country to another as from Italy to Australia and from Japan to Hawaii, or within a country such as rural to urban are worth investigating. There may be opportunities in Italy with immigrants from the south to the more industrialized north. The n o r t h south geographical differences in mortality from cancer of the prostate in Italy 37 suggests analysis of regional data on diet. In studies of diet and cancer we should investigate anomalous populations such as those in Hawaii and New Zealand which do not appear to be consistent with o u r current aetiol0gical hypotheses. Progress in science rarely comes from simply attempting to sustain weak hypotheses, but rather by refutation. This can lead to new hypotheses for further testing in human studies, with the ultimate aim of the dietary prevention of cancer.

Acknowledgements - The assistance of the Sanrocco International Cancer Symposia in attending the Symposium and of the Queensland Cancer Fund with return travel is gratefully acknowledged.

REFERENCES 1. Haenszel W, Kurihara M: Cancer mortality among U.S. Japanese for 1959-62. J. Natn Cancer lnst 40, 43 (1968). 2. McMichael A J. McCall M G, Hartshorne J M, Woodings T L: Patterns of gastrointestinal cancer in European migrants to Australia. The role of dietary change, hlt J Cancer 25, 43l (1980). 3. Hopkins S, Margetts B M. Cohen J, Armstrong B K: Dietary change among Italians and Australians in Perth. Commun Hlth Stud 4, 67 (1980). 4. Correa P: Epidemiological correlations between diet and cancer frequency. Cancer Res 41, 3685 (1981). 5. Armstrong B, Doll R: Environmental factors and cancer incidence and mortality in different countries, with special reference to dietary practices. Int J Cancer 15, 617 (1975). 6. UICC: Cancer Incidence in Five Continents. Springer. Berlin (1966). 7. UICC: Cancer Incidence in Five Continents, Vol. II. Springer, Berlin (1970). 8. Liu K, Moss D, Persky V, Stamler J, Garside D, Soltero I: Dietary cholesterol, fat and fibre, and colon-cancer mortality. Lancet 2, 782 (1979). 9. Bingham S, Williams D R R, Cole T J, James W P T: Dietary fibre and regional large-bowel cancer mortality in Britain. Br J Cancer 40, 456 (1979). 10. Blair A, Fraumeni J F: Geographic patterns of prostate cancer in the U.S. J Nam Cancer Inst 61, 1379 (1978). 11. Kolonel L N, Hankin H. Lee J, Chu S Y. Nomura A M Y, Ward Hinds M: Nutrient intakes in relation to cancer incidence in Hawaii. Br J Cancer 44, 332 (1981).

142

Robert MacLennan

12 Waterhouse J, Muir C, Shanmugaratnam K, Powell J, Peacham D, Whelan S: Cancer Incidence in Five Continents, vol IV, IARC Scientific Publications No. 42. Lyon, International Agency for Research on Cancer (1982). 13. Wynder E L, Shigematsu T: Environmental factors in cancer of the colon and rectum. Cancer 20, 1520 (1967). 14. Hirayama T: Changing patterns of cancer in Japan with special reference to the decrease in stomach cancer mortality, in Hiatt H H, Watson J D, Winsten J A (ed): Origins of Human Cancer: Cold Spring Harbor Conf on Cell Proliferation, vol 4, Cold Spring Harbor Laboratory, pp. 55-75 (1977). 15. McMichael A J, Potter J D, Hetzel B S: Time trends in colo-rectal cancer mortality in relation to food and alcohol consumption; U.S., U.K., Aust and NZ. lntJ Epidemiol 8, 295 (1979). 16. Zaridze D G: Environmental etiology of large-bowel cancer. J Nam Cancer lnst 70, 389 (1983). 17. Manousos O, Day N E, Trichopoulos D, Gerovassilis F, Tzonou A, Polychronopoulou A: Diet and colorectal cancer. A case-control study in Greece. Int J Cancer 32, 1 (1983). 18. Graham S, Haughey B, Marshall J, Priore R, Byers T, Rzepka T, Mettlin C, Edson Pontes J: Diet in the epidemiology of carcinoma of the prostate gland. J Nam Cancer lnst 70, 687 (1983). 19. Schuman L M, Mandel J S, Radke A, Seal U, Halbert F: Some selected features of the epidemiology of prostatic cancer. Minneapolis-St Paul, Minnesota case-control study, 1976-1979, in Magnus K (ed): Trends in Cancer Incidence. Causes and Practical hnplications, pp. 345-354. New York, Hemisphere Publishing (1982). 20. Rotkin I D: Studies in the epidemiology of prostatic cancer. Expanded sampling. Cancer Treat Rep. 2, 173 (1977). 21. Kolonel L N, Nomura A M Y, Ward Hinds M, Hirohata T, Hankin J H, Lee J: Role of diet in cancer incidence in Hawaii. Cancer Res (Suppl) 43, 2397s (1983). 22. Kolonel L, Hankin J, Lee J: Diet and prostate cancer. Am J Epidemiol 118, 454 (1983). 23. Phillips R L: Role of life-style and dietary habits in risk of cancer among Seventh-Day Adventists. Cancer Res 35, 3513 (1975). 24. Kinlen L J: Meat and fat consumption and cancer mortality: a study of strict religious orders in Britain. Lancet 1, 946 (1982).

25. Stemmermann G N, Nomura A M Y, Heilbrun L K: Dietary fat and the risk of colorectal cancer. Cancer Res 44. 4633 (1984). 26. Snowdon D A, Phillips R L, Choi W: Diet, obesity, and risk of fatal prostate cancer. Am J Epidemiol 120, 244 (1984). 27. Phillips R, Snowdon D: Fatal colorectal cancer among Seventh-Day Adventists with differing dietary habits (abstract). Am J Epidemiol 116, 565 (1982). 28. Hill M J, Drasar B S, Aries V, Crowther J S, Hawksworth G M, Williams R E O: Bacteria and etiology of cancer of large bowel. Lancet 1, 95 (1971). 29. Crowther J S, Drasar B S, Hill M J, MacLennan R, Magnin D, Peach S, Teoh-Chan C H: Faecal steroids and bacteria and large bowel cancer in Hong Kong by socio-economic groups. Br J Cancer 34, 191 (1976). 30. Jensen O M, MacLennan R, Wahrendorf J: Diet, bowel function, fecal characteristics, and large bowel cancer in Denmark and Finland. Nutr Cancer 4, 5 (1982). 31. Reddy B S, Hedges A R, Laakso K, Wynder E L: Metabolic epidemiology of large bowel cancer. Fecal bulk and constituents of high-risk North American and low-risk Finnish population. Cancer 42, 2832 (1978). 32. Newmark H L, Wargovich M J, Bruce W R: Colon cancer and dietary fat, phosphate and calcium: a hypothesis. J Natn Cancer lnst 72, 1323 (1984). 33. MacLennan R, Jensen O M, Mosbech J, Vuori H: Diet, transit-time, stool weight and colon cancer in two Scandinavian populations. Am J Clin Nutr 31, 239s (1978). 34. Henderson B: Third Symp on Epidemiology and Cancer Registries in the Pacific Basin. Natn Cancer hzst Monogr 62, p. 148 (1982). 35. Jensen O M, Bolander A M, Sigtryggsson P. Vercelli M, Nguyen-Dinh X, MacLennan R: Large-bowel cancer in married couples in Sweden: a follow-up study. Lancet 1, 1161 (1980). 36. Committee on Diet, Nutrition and Cancer, Assembly of Life Sciences, National Research Council: Diet, Nutrition and Cancer, Chapter 5. National Academy Press, Washington DC, (1982). 37. Di Silverio F, La Pera G, Tenaglia R: Age-adjusted mortality rate and regional distribution for prostatic carcinoma in Italy between 1969 and 1978. Prostate 3, 631 (1982).