Chapter 11 : Cancer and dietary fat

Contents - Previous - Next

Intervention studies
Dietary fat and cancer of the breast and colon
Animal experiments
Cancer at other sites

Early studies clearly demonstrated that mice and rats which were fed high-fat diets were far more susceptible to skin and mammary cancer than animals fed diets low in fat (Tannenbaum and Silverstone, 1953; Tannenbaum, 1959). These studies attracted little attention until epidemiologic data demonstrated that the prevalence of cancer at several sites was much higher in countries with high-fat diets than in those with low-fat diets (Carroll, 1975; Carroll and Khor, 1975). Since that time, abundant experimental data have shown that feeding rats or mice a variety of high-fat diets increases susceptibility to cancer of the breast, skin, colon, pancreas and prostate. Such studies have used a variety of carcinogens to induce cancer or, in some cases, spontaneous tumours. Much of the evidence relating diet to cancer has been presented and discussed in recent reviews (U.S. Department of Health and Human Services, 1988; National Research Council, 1982, 1989).


Several types of epidemiologic data have been reported. The first can be categorized as inter- and intra-country correlations.

Data collected by WHO on cancer incidence or mortality have been adjusted for age and standardized (Kurihara and Aoki, 1984; Parken et al., 1992) and then compared to dietary fat disappearance data collected by FAO (FAO, 1980b). Fat disappearance data do not take wastage into account and the early data, expressed as grams consumed per person per day, grossly overestimated actual intake. However, when fat is expressed as a percentage of total energy, the values approximate usual consumption and show a strong positive correlation with cancer mortality.

Because the populations compared are large in size, variations in dietary habits and genetic susceptibility are taken into account in these studies. Therefore, it is likely that the observed differences can be safely attributed to environmental factors. On the other hand, the accuracy of mortality data differs for each country and type of cancer. Similarly, the reliability of the fat disappearance data varies and the crude estimates of fat intake give no indication of the dietary practices of those individuals who do or do not develop cancer.

Comparisons of groups within countries, such as studies in the USA comparing the Seventh Day Adventists with the population as a whole (Phillips et al., 1980) and ethnic groups in Hawaii (Kolonel et al., 1982) have been presented. These involve smaller groups than the inter-country comparisons and the correlations with fat intake are generally weaker.

Time trends and migrant studies provide a second type of epidemiological data. Changing patterns of diet and cancer within countries or in groups who have migrated provide some of the strongest evidence that environmental factors, rather than heredity, are responsible for geographic differences in cancer incidence and mortality. Such trends in Japan (Hirayama, 1979) and among migrants from Japan (Prentice and Sheppard, 1990), China and various European countries (Gori, 1978; McMichael and Giles, 1988; Prentice and Sheppard, 1990) generally indicate that the trend from low-fat to high-fat diets is accompanied by increases in cancer at sites, such as breast and colon, that are positively correlated with fat intakes in the intercountry data. There are no reports on migrants changing from high- to low-fat diets; if such data exist, it would be of considerable interest.

Case-control studies give a third type of evidence. The diets of groups of people with cancer have been compared with those of individuals of similar age without cancer. Generally, these studies have failed to indicate a connection between diet and cancer. Such comparisons involve individuals who are clearly susceptible to cancer or those who have been exposed to adequate initiating stimuli; these factors may not be true of the people in the control group. Such studies depend upon the ability of individuals to accurately recall their dietary habits over the years which is probably unrealistic. Also, their current dietary habits may be influenced by the presence of disease. Substantial data indicate that dietary fat may act primarily during the promotional stage of carcinogenesis. Hence, it is not surprising that these kinds of data do not provide strong evidence of an association between diet and cancer (Carroll, 1994).

A fourth type of epidemiological data comes from cohort studies. In these studies, a group of people who are free of disease are followed over time after their diet and other characteristics have been assessed. As some individuals develop cancer or other diseases, attempts are made to compare the characteristics of those who have developed disease with those who have not. Such studies are advantageous in comparison to case-control studies since there is no selection bias with regard to susceptibility to disease. Furthermore, the current diet and other characteristics can presumably be evaluated more accurately than past characteristics. However, the group usually has rather homogeneous dietary practices and it is uncertain that the techniques available can correctly define the eating habits of the individuals. Correlations between nutrient intake assessed by different methods are relatively poor (Willett et al., 1985) and the limitations of this approach have frequently been noted (Goodwin and Boyd, 1987; Freudenheim and Marshall, 1988; Hebert and Kabat, 1991; Hegsted, 1989).

Intervention studies

In clinical trials, the participants are randomly assigned to a "treatment" and a "control" group; the "treatment" group is given counselling on the appropriate diet or other treatment and the "control" group is given standard advice. The groups are monitored over time to determine compliance and who develops disease. Many consider this approach to be the proper way to establish proof of the effectiveness of diet or other methods in treating or preventing disease. The problems in dietary intervention trials, however, are formidable. The groups must be large and the studies are labour-intensive, expensive and must extend over long periods of time. A primary problem is the assessment of individual dietary practices within the groups, particularly when in many instances, the general public is receiving the same advice as the treatment group. Since relatively few individuals can be expected to develop disease, the diets of individuals must be identified to ascertain the degree of compliance in those who do or do not develop disease. Current methods of assessing individual intakes show considerable bias in reported energy intakes, apparently regardless of the methodology used (Livingstone et al., 1990; Schoeller, 1990; Black et al., 1993). It is apparent that participants in treatment groups are under considerable pressure to report consumption patterns which approximate the instructions. In large-scale attempts to test lowfat diets, study participants have reported energy intakes that are unreasonably low, thus casting doubt on the reliability of the food consumption data (Henderson et al., 1990; Hunninghake et al., 1993). Such data appear to confirm the conclusion that the more dietary advice groups receive, the less reliable the data on food consumption become (Mertz, 1992). Thus, the utility or practicality of intervention trials to test dietary effects on disease incidence is questionable.

Experimental studies with animals allow absolute dietary control, can be continued over the lifetime of the animal and have been used to test a wide range of carcinogens and types of diet. The question is whether results found for animals can be extrapolated to humans. If the effects of dietary constituents are consistent with the epidemiologic correlations which have been observed, it is plausible that there are causal relationships between diet and cancer.

Dietary fat and cancer of the breast and colon

Epidemiological evidence. Cancer of the breast and colon account for a large proportion of the total cancer in Western populations. Both show a strong positive correlation with fat consumption in inter-country studies. Time trends and migrant studies are consistent with these patterns. However, in Japan, colon cancer has increased more rapidly than breast cancer as fat intake has risen (Figure 11.1). Migrants also show a more rapid increase in colon cancer (Berg, 1975) than breast cancer.

In case-control studies, dietary fat appears to be more consistently associated with colon cancer than breast cancer (Carroll, 1994; Willett, 1989), which would be expected because of the longer latent period for breast cancer. Although some studies show a positive association of breast cancer with dietary fat and others do not, a combined analysis of 12 case-controlled studies indicated a statistically significant, positive association with saturated fat intake and breast cancer in post-menopausal women (Howe, 1990). Case-control studies have generally shown a positive relationship between fat intake and colorectal cancer (Whittemore et al., 1990).

Cohort studies have usually failed to show an association between dietary fat and breast cancer (Howe et al., 1991; Van den Brandt et al., 1993; Willett et al., 1987, 1992). In their study of American nurses, Willett and his colleagues (1990) reported that colon cancer was positively associated with dietary animal fat, saturated fat and monounsaturated fat but the associations were weak and non-significant when adjusted for total energy intake (Figure11.2). Studies of this kind involve estimates of current diets and do not preclude the possibility that a life-long exposure to a low-fat diet might reduce susceptibility to breast and colon cancer.

FIGURE 11.1 : Relative changes in fat intake and age adjusted mortality for breast and colon cancer in Japan

FIGURE 11.2 : Relative risk of colon and breast cancer by animal fat intake

Animal experiments

Studies with experimental animals have shown with remarkable consistency that a high-fat diet increases the incidence and yield of mammary and colon cancer (National Research Council, 1982, 1989; Welsch, 1992). The effect appears to be primarily at the promotional stage of carcinogenesis (Carroll, 1975) although initiation may be influenced to some degree (Rogers and Lee, 1986).

Animal studies also indicate that non-lipid materials influence cancer incidence. Rats which were fed commercial chow had fewer tumours than those fed purified diets of similar lipid content (Carroll, 1975; Ip, 1987). There is accumulating evidence that diets which are high in dietary fibre, fruits and vegetables reduce cancer risk in humans (National Research Council, 1982, 1989; Ziegler, 1989; Block, Patterson and Subar, 1992; Sandler et al., 1993; Hunter et al., 1993). In animal experiments, a variety of anticancer agents have been identified in natural materials (Birt and Bresnick, 1991; Wattenberg, 1992). A reduction in fat intake often results in increased consumption of non-fat constituents in the diet. Hence, the difference in cancer incidence and mortality in populations consuming high- and low-fat diets may involve a variety of dietary components other than fat. However, studies to observe the effects of dietary fat using purified diets clearly implicate fat.

Experimental studies demonstrate that there is a requirement for n-6 polyunsaturated &t in the case of mammary tumours. Tumour yields increase with additions of linoleic acid up to a threshold of about 4-5 percent of total calories. When this threshold is reached, increasing total fat causes further increases in incidence and yield, apparently independent of the type of fat added (Ip, 1987). Colon cancer also seems to have a requirement for linoleic acid although this appears to be smaller than that required for mammary cancer (Bull, Bronstein and Nigro, 1989). There have been few studies on the effects, if any, of other specific fatty acids.

Fish oils, containing mainly n-3 polyunsaturated fatty acids, do not appear to promote mammary cancer when fed at high levels although they may have a stimulating effect at low levels (Carroll, 1989; Cave, 1991 a,b). Studies with mixtures of n-3 and n-6 fatty acids indicate that the promotional effect of n-6 fatty acids may be neutralized by a high ratio of n-3 to n-6 fatty acids in both mammary and colon cancer (Cave, 1991a; Reddy, 1992).

There has been much speculation about the mechanisms involved in the promotion of mammary cancer by dietary fat. This has included effects on the endocrine and immune systems; changes in the amount or composition of adipose tissue; changes in the fatty acid composition of the membranes of cancer cells; and effects on eicosanoids and lipid peroxidation (Welsch, 1987, 1992).

Many have suggested that energy balance is more important than the level of dietary fat (Boutwell, 1992; Pariza, 1988; Welsch, 1992) since the stimulating effect of high-fat diets can be overcome by caloric restriction. The degree of caloric restriction imposed in most experimental studies, however, is unrealistic for humans. Obesity is often associated with increased risk of breast cancer (Dao and Hilf, 1992), especially when the fat is more localized in the upper body (Ballard-Barbash et al., 1990; Schapira et al., 1990). Lean athletes have been reported to be at reduced risk of breast cancer (Frisch et al., 1992). It has been suggested that this may be related to the ability of adipose tissue to convert androgens to estrogens. Exercise in animals is reported to either inhibit or enhance mammary cancer depending upon the intensity and duration of the exercise (Thompson, 1992).

The promotion of colon cancer by dietary fat could be due to enhanced excretion of bile acids and might be mediated by protein kinase C and/or ornithine decarboxylase. The lack of promotion of colon cancer by fish oils may involve effects on eicosanoid production. Prostaglandin inhibitors have been shown to inhibit colon carcinogenesis (Ready, 1992).

Cancer at other sites

There have been a limited number of investigations of the relationship between dietary fat and cancer at the following sites: pancreas, prostate, skin, lung, ovary, bladder, oral cavity, non-Hodgkin's lymphoma and leukaemia. The inter-country correlations show prostate cancer to be strongly correlated with fat while the relationship is weaker for other sites. In Figure 11.3, the correlation coefficients between the availability of dietary fat and the age-adjusted mortality rates for malignant neoplasms in different countries are illustrated.

In most case-control studies, prostate cancer has been positively associated with dietary fat (Carroll 1994; Rose and Connolly, 1991). The results have been quite variable with regard to other sites but the data available are very limited. A few cohort studies have included pancreatic cancer (Mills et al., 1988) and prostatic cancer (Mills et al., 1989; Severson et al., 1989) and have shown no association with dietary fat. Bosland (1988), however, concluded that a high-fat diet is involved in the etiopathogenesis of prostate cancer.

Experimental studies. Animal models are not available for the study of many cancers. The effects of dietary fat have been studied in rats treated with azaserine and hamsters treated with N-nitrosobis(2-oxypropyl)-nitrosamine. High levels of dietary fat enhance the tumours in the promotional stage (Roebuck, 1992). In the rat model, there appeared to be a rather high requirement for linoleic acid, 4-8 percent in a 20 percent total fat diet. High levels of fish oil reduced the lesions; when mixtures of n-3/n-6 oils were fed, the threshold appeared to be between 3-6 percent linoleic acid. Caloric restriction also reduced focal neoplastic lesions (Roebuck, 1992; Roebuck, Baumgartner and Mac Millan, 1993). Some progress is being made in the development of an animal model for the study of prostate cancer and this provides evidence that high fat diets enhance risk at this site (Pollard and Luckert, 1986).

Many of the earliest studies with animals involved skin cancer and showed a positive association with dietary fat. In one model it was reported that, contrary to the results with breast cancer, there was an inverse correlation between linoleic acid and skin cancer (Fischer et al., 1992). Skin cancer can also be inhibited by caloric restriction (Birt et al., 1993).

FIGURE 11.3 : Dietary fat in relation to cancer at specific sites


There are abundant data showing that animals fed high-fat diets ad libitum develop tumours of the mammary gland, intestine, skin and pancreas more readily than animals fed low-fat diets, although this effect can be overridden by caloric restriction. These data are consistent with the inter-country correlations linking dietary fat with cancer of the breast, colon, pancreas and prostate. Cohort and case-control studies cast some doubt on these associations, but there are serious limitations in case-control and cohort studies.

In animal experiments, the type of dietary fat, particularly the level and type of the polyunsaturated fatty acids, does have an effect. However, often the animal studies have been conducted with a single source of fat, a situation not found in human populations. The inter-country studies indicate that, over a range of n-6 fatty acids of about 4 to 8 percent of energy, there appears to be no correlation with breast cancer (Carroll et al., 1986). Whether such animal data are relevant to humans may be questionable.

Less data are available on the n-3 levels of fatty acids in human diets and their relevance to cancer in humans. However, the total fat content of the diet, rather than the type of fat, appears to have more influence on cancer in human populations.

The association of some tumours with obesity lends some credence to the possibility that the effects of high-fat diets on cancer may be partially explained by changes in energy balance. Lean athletes appear to have less breast cancer. Low-fat diets are reported to be associated with decreased energy intake and weight reduction (Boyd et al., 1988; Henderson et al., 1990) whereas most efforts to control obesity by simple caloric restriction fail. In addition, low-fat diets are inherently higher in crude carbohydrate sources, dietary fibre and fruits and vegetables which appear to provide additional protection against cancer. Appropriate physical exercise is relevant to cancer prevention as it is part of a generally healthy life-style.

Some concern has been expressed about the possibility that low serum cholesterol levels may be associated with increased risk of cancer, particularly cancer of the colon (McMichael, 1991). However, since colon cancer is rare in populations consuming low-fat diets and having low serum cholesterol levels, it is evident that a low serum cholesterol per se does not increase risk of cancer.

Contents - Previous - Next