3.1 INTRODUCTION
The simplest objection to a link between chronic disease and diet comes from those who argue that humans have always had to eat, that diets have always contained the same nutrients—protein, fat, carbohydrates, vitamins, minerals, and that, therefore, the present pattern of diseases and changes in that pattern cannot be causally linked to dietary intake. This argument, despite its simplicity, raises some important issues for nutritional epidemiology. The contemporary epidemiologic endeavour that explores the relation between diet and specific diseases is based, obviously, on the premise that this argument is false. This chapter is an attempt to explore the evidence for a significant and causal association between eating patterns and cancer. Its second role is to show that, far from being an implausible link, much of the explanation for the relationship between dietary patterns and cancer comes from the dependence of humans on their food supply; this dependence is not merely the issue of providing the fuel for the organism but relates, rather, to long-standing adaptive patterns of food intake and aberrations in those patterns.
This chapter takes, as its starting point, a diet to which humans are adapted, noting, especially, intakes of substances for which we are dependent on the environment and intakes of substances to which we have low or infrequent exposure. It is a diet with seasonal variability in the availability of total food intake and of specific foods.
There are four broad ways in which aberrations in dietary patterns could produce disease and perhaps cancer. The first is an imbalance in energy intake and output. Secondly, there may be an alteration in the pattern of both macro- and micro-nutrients. Thirdly, there may be cases of specific deficiencies.
Fourthly, there may be present, in the food supply, substances to which the organism is almost never exposed and therefore, for which there may not be the relevant metabolic responses.
3.2
IF THERE IS A DIET TO WHICH HUMANS ARE ADAPTED, WHAT MIGHT IT LOOK LIKE?
There is a long history of evoking a lost Golden Age to compare with current miseries or explain current misfortunes: from Eden (where fruit eating habits were a part of the problem) through Rousseau’s noble savage to the recent discussion of the nature of paleolithic diets (Eaton and Konner, 1985). The fact is that we cannot know exactly to what kind of diets humans are well adapted biologically (although dentition, metabolic enzyme patterns, and the length and morphology of the GI tract provide some clues). Nonetheless, it is appropriate to ask if there is some probable picture we can draw of our early diet. If we can, we should also acknowledge the extensive variability in details that must have existed in the same way that we can describe extensive geographic variability in contemporary diets. Some of the common features of that early diet must have been: a high intake of a wide variety of foods—roots, leaves, nuts, seeds, fruit (grains become a staple only in the last 10 to 15 thousand years but were probably gathered regularly in season); sporadic intake of lean and/or saturated- fat meat (a more secure and regular supply of fish and seafood for coastal dwellers); intake would also have included insects, grubs (high in protein) and bone marrow and organ meats; very low intake of alcohol; little refining or fractionation of food into parts; low and irregular intake of eggs, milk and milk products; variability, by season, either of total amount of food available (and therefore body weight), of kinds of foods available, or both. Therefore, there would also have been variability in the availability of particular nutrients. Some of the variations in this overall intake pattern would have been the result of climate (as it varied over time and from place to place including the consumption of high-fat diets in extreme northern populations) but in general, until very recently, saturated fat and alcohol intake would have been low, vegetable food (but not grain) intake high, and the kinds of food eaten highly varied.
One of the significant arguments in favour of human adaptation to certain kinds of foods and eating habits, is that there are some substances for which we are dependent upon the environment. This well-established concept in nutrition has, as will be argued below, some important implications for cancer aetiology.
The short-term and long-term consequences of a variety of nutrient deficiencies are a spectrum of well-described disorders. But deficiency disorders can arise only if the organism is incapable of synthesis. This argues that there has been no selection pressure to develop such a capacity in the population and, therefore, that the essential substances are widely available in the environment.
Paradoxically, then, because essential nutrients are widely available in naturally- occurring human food, deficiencies are possible. Essential amino acids, essential fatty acids, micro-elements and vitamins are examples. The fact that what is essential varies among mammalian species underlines the importance of the adaptation process. To question whether we have defined all of the essential
64 DIET AND CANCER [CH. 3
nutrients and whether we have exhaustively characterized all the consequences of low/absent intake (whether that intake is in the normal range or not) is one function of this chapter.
In essence, the adaptation argument is as follows: the essential nutrients—
both energy-bearing and micronutrients—are widely available in nature; they have important functions in growth, development and reproduction; the organism is adapted to their ubiquity; deficiencies impair growth, development and reproduction.
This argument regarding nutrients known to be important in growth, development and reproduction has a plausible analogy in relation to the presence of substances necessary for the maintenance of the organism including substances which reduce the risk of carcinogenesis. Here, the argument is that normal function of cells is dependent on the presence of a variety of widespread dietary constituents probably including, but not confined to, those necessary for growth and development. In their absence, cells malfunction. The malfunctioning state may make the cells more susceptible to exposure to carcinogens or may impair some specific protective mechanisms such as enzyme induction. It may even be characterized by an increase in cell replication rates as somatic cells seek to adapt to the new—deprived— conditions. It may be worth noting that maintenance is a continuous function from birth whereas growth, development and reproduction are time-limited.
A converse argument applies to those dietary constituents that are rare in nature. If substances are consumed only occasionally (or not at all), then high ingestion may have untoward consequences. This could apply both to very rare exposures which produce acute toxicity and to unaccustomed levels of intake that overwhelm the cellular and metabolic processes that normally handle the exposure. Bacterial, plant and fungal toxins are all potential members of the first class; a high fat/high calorie intake with consequences for cholesterol metabolism, insulin metabolism, adipose storage, and sex steroid hormone production is an example of the second. A high grain diet (such as that in predominantly agricultural communities) is often associated with a reduced intake of other plant foods: further, such diets contain large amounts of abrasive material that may increase cell replication rates particularly in the upper digestive tract. There may be differing degrees of adaptation in long-exposed vs unexposed populations.
One objection to an adaptation argument of any sort is that natural selection will be an influence only to the age of reproduction and that, as chronic diseases are largely diseases of post-reproductive years, dietary adaptation is an unnecessary postulate. There are four responses to this. The first is to argue that humans have a long period of infant and juvenile dependence and that survival of parents in a healthy state is likely to be selected for.
The second response requires consideration of the unit of selection. If the issue is the survival of tribes or bands, then those bands would have survived better that had sufficient elders who knew how to respond to infrequently met hazards—
food or water shortage, epidemic disease and natural hazards such as fire or extreme weather. The tribal wisdom maintained by the old would have meant survival of the tribe. Tribes without elders and without knowledge would be more likely to perish. Therefore, the tribes in which longevity was selected would, in turn, have survived other threats to pass on their wisdom, their adaptive eating habits, their adapted metabolisms, and their genes.
Thirdly, to argue that chronic diseases are a phenomenon of older age and therefore that resistance to them cannot have been selected for is to forget that these diseases are not a phenomenon of younger ages and that, therefore, some resistance (at least to the point of postponing them to older ages) has been selected for.
Finally, a diet that reduces risk of cancer may also improve reproductive success. There are a wide variety of substances that are both teratogenic and carcinogenic. So selection for improved reproductive success could directly select for reduced risk of cancer.
This chapter will explore the empirical evidence for the existence of unaccustomed exposures and protective dietary constituents, and for risks associated with nutrient and energy imbalance. Some evidence for likely biologic mechanisms and the implications for further research are considered.
3.3
DIETARY EXPOSURE AND CANCER RISK
The available empirical epidemiologic and biologic data on diet and cancer are not readily summarized in a single chapter. It is intended, therefore, to use, as a framework, the adaptation argument outlined above and show that some of the empirical relationships that have been established in the epidemiologic literature are explainable in relation to the four kinds of aberration from the dietary behaviour to which humans are adapted—energy imbalance, nutrient imbalance, specific deficiency, specific exposure. These four relationships will be illustrated in relation to certain cancers, particularly breast, colon and pancreas. Some plausible mechanisms will be outlined.
3.4
ENERGY IMBALANCE
3.4.1
Epidemiologic evidence
The nature of energy imbalance is a major topic; only one or two aspects of it will be considered here. Three measures relevant to energy balance (Pariza and Simopoulos, 1986) have been explored in epidemiologic studies of cancer—total intake (Potter and McMichael, 1986; Willett and Stampfer, 1986; Lyon et al,
66 DIET AND CANCER [CH. 3
1987), energy output (Garabrant et al, 1984) and a variety of measures of growth (Micozzi, 1985) and obesity (Paffenbarger, et al, 1980; Helmrich et al, 1983).
There is no simple established relationship between any of these measures and all cancers, and even for some specific cancers, the empirical data are not clear.
In addition, there are some paradoxes.
The present evidence suggests that higher physical activity is related to a lower risk of colon cancer (Garabrant et al, 1984; Vena et al, 1985; Gerhardsson et al, 1986; Paffenbarger et al, 1987; Wu et al, 1987; Slattery et al, 1988a) but that obesity is probably not a risk factor (Potter and McMichael, 1986; Sidney et al, 1986). For endometrial cancer (Elwood et al, 1977; La Vecchia et al, 1984;
Folsom et al, 1989) and post-menopausal breast cancer (Lew and Garfinkel, 1979; Helmrich et al, 1983) obesity is a risk factor. However, for premenopausal breast cancer, obesity is associated with a reduced risk (Paffenbarger et al, 1980;
Helmrich et al, 1983). Physical activity bears an uncertain but possibly negative relation to risk of breast cancer (Frisch et al, 1985). Total energy intake is an inconsistent risk factor for all three cancers—it is worth noting that dietary instruments vary extensively in their capacity to measure total energy. There are no established relationships with other specific cancers; there is a general association between obesity and overall cancer risk (Lew and Garfinkel, 1979).
Fat distribution, known to be related to diabetes and CHD (Vague, 1956;
Feldman et al, 1969; Larsson et al, 1984; Donahue et al, 1987; Selby et al, 1989) is currently under investigation as a cancer risk factor (Folsom et al, 1989, 1990;
Sellers et al, 1992).
3.4.2
Plausible mechanisms
At least three mechanisms seem to present themselves as explanations—
hormonal, mechanical and total cellular workload. Peripheral adipose tissue is the major source of oestrogens in postmenopausal women (Grodin et al, 1973) via conversion of adrenal androstenedione; this provides a plausible explanation for the association with endometrial and post-menopausal breast cancer, each of which is associated with higher (perhaps cumulative) lifetime oestrogen exposures. Why obesity should be associated with a lower risk of premenopausal breast cancer is unknown—it does not seem to be only a matter of failing to detect cancerous lesions in large breasts (Willett et al, 1985).
The role of physical activity in colon cancer is plausibly a mechanical effect—
higher activity means shorter digestive tract transit time—but this is not an established negative risk factor for colon cancer. That obesity is not a risk factor for colon cancer (while high energy intake and low energy output are) suggests that there may be metabolic differences in those who get colon cancer compared with those who do not. It may also be that the total amount of food passing through the large bowel represents a measure of total cellular work load and therefore likely rates of cell replication (Potter, 1989, 1992).
It is probable that the complex relation between dietary intake, obesity and physical activity, on the one hand, and cancer risk on the other, will only be explicated when we have clarified some of the relevant intermediate metabolic steps including the effects on gut function, on cell turnover, and on hormone production. What remains, however, is the empirical observation that aspects of energy imbalance are related to cancer risk.
3.4.3 Adaptation
The human organism, it was argued above, is adapted to high variability in food intake and able to make rapid use of a sudden increase in food supply in order to survive better through the lean times. This is the original thrifty gene hypothesis proposed as an explanation for the survival advantage of the predisposition to diabetes (Neel, 1969). Clearly, a high intake of food as a regular, rather than sporadic, phenomenon is likely to jam open more than just insulin responses. In provoking obesity, it will also increase peripheral adipose production of oestrogen, a factor perhaps originally associated with reproductive success—
sufficient body fat, plus hormonal support to carry a child to term. Is this related to the fact that obesity is actually protective against premenopausal breast cancer and only a liability later?
There are animal experimental data to show that on a low intake or in a fasting state, the structure of the gut epithelium is much simpler and cell replication rates much lower (Stragand and Hagemann, 1977). On refeeding, replication rates increase, as does the complexity of the epithelial surface and the total surface area. It seems probable that this is a highly adaptive response to variability in food availability—high cell turnover and large absorptive area during feast; low activity during fast to conserve energy. In the presence of a high-intake diet as a regular phenomenon, however, high cell turnover and maximal epithelial surface area provide a favourable environment for carcinogenesis. It is worth noting that our original observation on the association between increased frequency of food consumption and risk of colon (but not rectal) cancer (Potter and McMichael, 1986) has now been confirmed in several other studies (La Vecchia et al, 1988a;
Young and Wolf, 1988). These data provide additional evidence for the idea that there is a cost for frequent food intake (one concommitment of a high food intake).
Thus, although obesity itself was probably uncommon in our ancestors, the capacity to assimilate food rapidly and store energy when it was available has probably been selected for. The inheritance of this kind of metabolism in a society where food is widely available appears to have consequences for cancer risk at a number of sites. It remains to be seen whether the tendency for different patterns of body fat distribution—now an established risk factor for several diseases—is a marker for the relevant metabolic differences.
68 DIET AND CANCER [CH. 3
3.5
IMBALANCE OF FOOD/NUTRITIONAL INTAKE
3.5.1
Epidemiologic evidence
Much of the current dietary epidemiology of cancer has focused on the role of intakes of specific macronutrients, particularly fat and alcohol. The evidence for a role of fat in the aetiology of breast cancer is rather poor and inconsistent but that for alcohol is surprisingly consistent. A number of ecologic studies of fat and breast cancer have been published. In a regional study in Great Britain, Stocks (1970) noted that dairy products were positively associated with risk but that other sources of fat were inversely associated. Hirayama (1978) noted that pork consumption (but not fat) correlated with risk across 12 regions of Japan.
The study of ethnic groups within Hawaii found what many believe to be an implausibly strong relationship between fat and breast cancer given the data from other studies cited above and below: an approximately 35% difference in fat was associated with about a 200% difference in risk of breast cancer (Kolonel et al, 1981).
Studies have been made of sub-populations where the known intake of animal fat and animal protein is lower than that of the general community. Neither in vegetarian nuns (Kinlen, 1982) nor in Seventh-day Adventists (Phillips et al, 1980), the majority of whom are lacto-ovo-vegetarians, is the risk of breast cancer lower than in the general population. Although it may be argued that, in both these groups, the reproductive rate is lower than comparison populations, these two studies do not provide strong evidence for a role for fat in the genesis of breast cancer.
There have been at least nine case-control studies on the relation between meat or fat consumption and breast cancer (Table 3.1). The study of Lubin et al (1981)—the only study showing a significant increase in risk in association with consumption of fat—was based on frequency of consumption of just eight food items. Two of three studies reporting on dairy product consumption found positive associations with risk.
Of the four cohort studies (Table 3.2) only that of Hirayama (1978) found an association with daily consumption of meat and risk of breast cancer after age 54.
This finding was based on just 14 cases in this category. Combining the risk associated with fat in all of the recent cohort studies results in a risk ratio between the highest and lowest quantiles of 1.01 (personal communication, W Willett and D Hunter).
These findings for breast cancer are in contrast to those for colorectal cancer (Table 3.3). Of 16 studies where saturated fat or total fat were studied, nine found a significantly increased risk. Similarly, meat has been shown to be a significant risk factor in 9 of the 14 studies in which it was examined. Protein was positively associated with risk in 4 of 5 studies.