Foundations of A Healthy Diet1

Foundations of A Healthy Diet1

Walter C. Willett

Meir J. Stampfer

Nutritional science has provided a wealth of data ranging from detailed molecular descriptions of nutrients and their actions to epidemiologic findings from large prospective studies and controlled randomized studies on selected population groups. Integrating this vast literature into a description of a healthy diet is a challenging yet essential step to provide the public and those responsible for food programs with the best information regarding their food choices. Efforts to develop descriptions of a healthy diet include the US Dietary Guidelines for Americans (1), a synthesis by the Institute of Medicine (IOM) (2) and the World Health Organization (WHO) (3). Because information on diet and health is accruing rapidly, these syntheses of dietary information require frequent updating, which is recognized by the requirement that the US Dietary Guidelines be reviewed every 5 years. This chapter discusses considerations for developing the definition of a healthy diet and briefly reviews some of the main issues, recognizing that they are addressed in detail elsewhere in this text. Finally, several alternative representations of a healthy diet are described.

Until recently, a primary focus of human nutrition was the prevention of nutrient deficiency, and achieving the recommended dietary allowances (RDAs) (4) for essential nutrients was the central objective. This approach led to the development of the seven food groups during World War II and later the “basic four” (meat, dairy, grains, and fruits and vegetables) as the definition of a healthy diet to be conveyed to the public (5). This effort, together with selective fortification and greater availability of a variety of foods, successfully eliminated clinically evident nutrient deficiencies from the United States and Europe. In the last several decades, the definition of a healthy diet has been expanded to include the optimization of long-term health. An underlying motivation for this expansion in scope has been epidemiologic evidence that coronary heart disease (CHD) and cancer have become the major causes of death in Western countries. Thus, considerations regarding a healthy diet have come to include macronutrient composition, qualitative aspects of macronutrients such as the glycemic index, food constituents not considered to be nutrients such as fiber and carotenoids, and possible benefits of essential nutrients at intakes above those known to prevent overt deficiency.

In describing a healthy diet, an immediate issue is whether this should be expressed as foods or nutrients. Using foods is attractive because this provides an easy form of communication that is recognizable by all. Although this is desirable in principle, those attempting to describe an optimal diet only in terms of foods find this challenging. The main reason is that the same foods can be made in many ways. For example, a cracker can be made with lard, partially hydrogenated vegetable oil, or nonhydrogenated corn oil; and vegetables served at a restaurant can be prepared in butter, margarine of unknown composition, or olive oil. The implications for health vary greatly. This issue is becoming increasingly important as the proportion of the food supply that is already processed or is eaten away from home increases. Most groups that have grappled with these issues have developed guidelines
that are hybrids, using a combination of food and nutrient criteria. For example, many written guidelines include both a quantitative description of fat intake and suggestions about servings of fruits and vegetables. However, when translating dietary guidance into graphic form (e.g., a food guide pyramid), this is often done by only using foods, which may fail to convey essential information.


Excessive body fat caused by an imbalance between energy intake and expenditure is currently the most important nutritional problem in developed countries and is rapidly becoming a global epidemic. A definition of a healthy diet that fails to address this would be deficient. Some well-intended guidelines are highly prescriptive in terms of energy intake or servings per day of each food group. A fundamental problem is that even the healthiest combination of foods consumed in slight excess, by only a few percent, over an extended period of time, will lead to overweight. Even with the best of methods, our assessments of intake are not sufficiently precise to measure these fine differences, and our assessments of energy expenditure are at least as imperfect. This problem is further compounded by the imprecise estimation of intake of quantities of foods by individuals and even by differences in definitions of serving sizes among branches of government (e.g., the US Food and Drug Administration [FDA] and US Department of Agriculture [USDA]). For these reasons, attempts to address overweight by detailed definitions of energy intake in dietary guidelines will not be successful. Weight itself, however, is well measured and represents a sensitive indicator of the long-term balance between energy intake and expenditure. For this reason, a definition of a healthy diet needs to be closely linked with the importance of maintaining a healthy weight and the need to make adjustments in intake or physical activity if an imbalance exists. Whether the qualitative aspects of diet may help facilitate weight control is discussed later in this chapter.


For many years nutritionists have recognized that individuals differ in their response to nutrient intakes (e.g., in the response of serum cholesterol to dietary cholesterol [6], or the response of blood pressure to sodium intake [7]). In extreme cases of inborn genetic defects (e.g., phenylketonuria [PKU]), standard diets can be lethal. The elucidation of the human genome and rapid identification of polymorphisms in almost all genes is creating new opportunities to individualize dietary guidance. For example, a homozygous polymorphism in the methylenetetra-hydrofolate reductase (MTHFR) gene, present in about 10% of the population, increases the amount of dietary folic acid needed to minimize blood concentrations of homocysteine (8). Does this mean that special dietary guidance needs to be given to those persons? Although we could now easily screen for MTHFR polymorphisms and individualize dietary advice, this is still probably not a logical strategy, and having different dietary advice for folic acid for different persons would create considerable complexity within populations and even within families. Because these variations probably exist for almost every nutrient, the possible combinations are almost infinite and would mean that each person would have a unique dietary recommendation. An alternative is to define healthy diets that would be sufficiently high in folic acid to meet the needs of this subset of the population. This has been the general approach in setting RDAs, whereby a margin of error has been added above average requirements to include individual variations in nutrient needs. This is an appropriate approach when variation in requirements is known to exist and we have no practical way of identifying individuals with different requirements or the reason for these differences, and often it is still a reasonable strategy even though we have the potential to identify individual differences in requirements.

The ability to identify individuals with different requirements allows more detailed studies to be sure that their needs are being met. Also, for some carefully selected genetic variants, different dietary approaches may be appropriate (e.g., PKU, as mentioned). Also, it may well be that diets to address specific abnormalities, such as elevated cholesterol, will be prescribed based on genetic information.

Genetic variation is only one of several factors that can influence nutritional requirements, and may not even be the most important. Age, body size, activity level, and pregnancy have long been recognized as factors to be considered, and requirements often are specific for these groups. However, as Hegsted pointed out (9), if requirements are expressed in terms of dietary quality (e.g., as nutrient densities), many of the differences in requirements diminish. One fundamentally important influence on the response to diet is the underlying degree of insulin resistance. This was described by Jeppesen et al (10, 11), who noted that the adverse effects of high carbohydrate intake on metabolic markers of the insulin resistance syndrome were strongly correlated with baseline insulin resistance. This relation has been confirmed in population studies showing a much stronger relation between the dietary glycemic load and blood triglyceride concentrations (12) and risk of CHD (13) among persons with a greater body mass index, which is a major determinant of insulin resistance. The implication is that a person who is lean and active can better tolerate a high-carbohydrate diet than someone who is less active and overweight. This also has important implications on a population basis because of strong evidence that most Asian groups have a higher prevalence of insulin resistance, possibly because
of genetic reasons, compared with European populations (14). Neel (15) described this as the “thrifty gene.” Until recently, these populations generally were highly active and lean, and thus protected from the adverse effects of this genetic predisposition. However, with the reductions in activity and gains in body weight that typically accompany a modern lifestyle, the ability to tolerate a diet high in refined carbohydrates diminishes. Nevertheless, this does not necessarily require different dietary recommendations if diets with low amounts of refined carbohydrate, even if not as critical, would be desirable for other populations as well.


Traditionally, animal experiments and small human metabolic studies provided the data underlying the basis of dietary recommendations, based on short-term or extreme effects, such as overt deficiency. Inevitably, the study of chronic disease in humans has required epidemiologic approaches. Until recently, these largely consisted of international comparisons and case-control studies, which examined dietary factors retrospectively in relation to cancer and other diseases. Now, large prospective studies of many thousands of persons are providing data, based on both biochemical indicators of diet and dietary questionnaires, which have been rigorously validated (16). Ideally, each potential relationship between diet and a health outcome would be evaluated in a randomized trial (17), but this is often not feasible because of practical constraints. The best available evidence usually is based on a synthesis of epidemiologic, metabolic, animal, and mechanistic studies. Major aspects of diet are discussed briefly here.

Dietary Fat and Specific Fatty Acids

Until recently, reviews on diet and health consistently recommended reducing total fat intake, usually to 30% of energy or less (17, 18, 19), to decrease CHD and cancer. The classical diet-heart hypothesis has rested heavily on observations that total serum cholesterol concentrations predict CHD risk; serum cholesterol has thus functioned as a surrogate marker of risk in hundreds of metabolic studies. These studies, summarized as equations by Keys (20) and Hegsted (21), indicated that, compared with carbohydrates, saturated fats and dietary cholesterol increase and polyunsaturated fat decreases serum cholesterol, whereas monounsaturated fat has no influence. These widely used equations, although valid for total cholesterol, have become less relevant with the recognition that the high-density lipoprotein (HDL) cholesterol fraction is strongly and inversely related to CHD risk, and the ratio of total cholesterol to HDL is a better predictor (22, 23, 24, 25). Substitution of carbohydrate for saturated fat (the basis of most dietary recommendations until recently) tends to reduce HDL as well as total and low-density lipoprotein (LDL) cholesterol; thus, the ratio does not change appreciably (26). In contrast, substituting monounsaturated fat for saturated fat reduces LDL without affecting HDL, thus providing an improved ratio (26). In addition, monounsaturated fats, compared with carbohydrate, reduce blood sugar and triglycerides in those with type 2 diabetes (27).

Although different saturated fats vary in their influence on LDL concentration (28), this finding probably has limited practical importance because intakes of the various saturated fats are strongly correlated with each other in usual diets, and there is no direct evidence that stearic acid is a lesser risk factor for CHD than other saturated fatty acids (29).

Uncertainty about Optimal Polyunsaturated Fat Intake

The metabolic studies predicting total serum cholesterol (20, 21) suggested that polyunsaturated fat intake should be maximized, and the American Heart Association has recommended intakes of up to 10% of energy (compared with US averages of approximately 3% in the 1950s and 6% at present). Concerns have arisen from animal studies in which omega-6 polyunsaturated fat (typically as corn oil) has promoted tumor growth (30), and the possibility that high intakes of omega-6 relative to omega-3 fatty acids might be proinflammatory and promote coronary thrombosis. However, as described in the following, available evidence from human studies has not supported these concerns at levels of omega-6 fatty acid intakes at least up to 10% of calories.

Dietary Fat and Incidence of CHD

In Keys’ pioneering ecologic study of diets and CHD in seven countries (31, 32), total fat intake had little association with population rates of CHD; indeed, the lowest rate was in Crete, which had the highest fat intake because of the large consumption of olive oil. Saturated fat intake, however, was positively related to CHD. In contrast to international comparisons, which are potentially confounded by many different factors, little relationship has been seen with saturated fat intake in prospective studies of individuals when compared with the same percentage of energy from carbohydrate (33, 34). However, polyunsaturated fat, mainly linoleic acid, has been inversely associated with risk of CHD in these prospective studies, especially when compared with saturated fat intake. Similarly, dietary intervention trials generally have shown little effect on CHD incidence when carbohydrate replaces saturated fat, but replacing saturated fat with polyunsaturated fat has reduced incidence of CHD (35). At intakes within the dietary range, the benefits of omega-3 fatty acids appear to be primarily in prevention of fatal arrhythmias that can complicate CHD rather than in prevention of infarction (36, 37). The amount of long-chain (marine) omega-3 fatty acids needed to prevent arrhythmia is remarkably small—on the order of 250 mg/day—and fish consumption twice a week appears to provide most of the potential reduction
of sudden death (38). The 18-carbon omega-3 fatty acid, α-linolenic acid, also appears to reduce risk of coronary heart disease (39, 40) and may be particularly important when fish intake is low. Based on theoretic concerns about competition in the elongation and desaturation pathways, some have hypothesized that dietary omega-6 fatty acids, mainly linoleic acid, are proinflammatory and counteract the benefits of omega-3 fatty acids; thus some have proposed that the ratio of omega-6 to omega-3 fatty acids is especially important in the prediction of heart disease (41). However, linoleic acid appears to have anti-inflammatory effects mediated by other pathways and does not increase inflammatory markers in humans (42). Also, because both omega-3 and omega-6 fatty acids are essential and reduce the risk of heart disease, their ratio has been unrelated to risk (43); consuming adequate amounts of both are important (44). The optimal amount of linoleic acid in the diet is not yet clear; in the US population, the benefits for heart disease appear to increase monotonically with greater intakes, but few individuals consume more than 10% of energy.

Trans-Fatty Acids

Trans-fatty acids are formed by the partial hydrogenation of liquid vegetable oils in the production of margarine and vegetable shortening and can account for as much as 40% of these products. Trans-fatty acids increase LDL and decrease HDL (45); raise the proportion of small, dense, and atherogenic LDL particles (46); raise lipoprotein(a) (47, 48) and increase inflammatory markers that have been related to CHD risk (49, 50). In the most detailed prospective study to date, trans-fatty acid intake was strongly associated with risk of CHD (33) and, as predicted by metabolic studies, this association was stronger than for saturated fat. The association between trans-fatty acid intake and risk of CHD has been confirmed in other prospective studies (51). Since 2006, the FDA has required food labels to include the trans-fat content of foods, which has resulted in major decreases in the amounts used (52); and banning of trans-fat in restaurants in New York and elsewhere has led most national restaurant chains to eliminate them from their products.

Relation between Dietary Fat and Risk of Type 2 Diabetes.

The relation between dietary fat and risk of type 2 diabetes appears to be similar to that for CHD (53). The overall percentage of fat does not appear to be related to risk. However, polyunsaturated fat is inversely associated with risk, consistent with its effect on insulin resistance, and trans-fat has been positively associated with risk (53, 54), consistent with evidence of its effects on inflammatory markers noted in the preceding discussion. Consumption of red meat, particularly processed red meat, has been associated with greater risk (55, 56).

Dietary Fat and Cancer.

One justification for low-fat diets has been the belief that these would reduce the incidence of cancers of the breast, colon and rectum, and prostate (18, 57). The primary evidence has been that countries with low-fat intake (also the less affluent areas) have had low rates of these cancers (57, 58). These correlations have been primarily with animal fat and meat intake rather than with vegetable fat consumption.

The hypothesis that fat intake increases breast cancer risk has been supported by most animal models (59, 60), although no association was seen in a large study that did not use an inducing agent (61). Moreover, much of the effect of dietary fat in the animal studies appears to be caused by an increase in total energy intake, and energy restriction profoundly decreases incidence (30, 59, 61). Data from many large prospective studies, including approximately 8000 cases in more than 300,000 women, have been published (62). In none of these studies was the risk of breast cancer significantly elevated among those with the highest fat intake, and the summary relative risk for the highest versus lowest category of dietary fat composition was 1.03 (62). In a pooled analysis, no reduction in risk was seen, even in those with less than 20% of energy from fat (63). In two large randomized trials, interventions to reduce intake of total fat did not significantly influence risk of breast cancer (64, 65). Thus, over the range of fat intake consumed by middle-aged women, total dietary fat does not appear to increase breast cancer risk. During adolescence and early adulthood, higher intake of animal fat, particularly from dairy products and red meat, has been associated with greater risk of premenopausal breast cancer (66, 67, 68). Vegetable fat was not associated with risk of breast cancer in this study, suggesting that some components of animal foods rather than fat itself may increase risk.

Although associations between dietary fat and risk of colorectal cancer had been seen in earlier retrospective case-control studies, little relation has been seen in prospective studies (69). However, associations between red meat consumption, particularly processed meats, and colorectal cancer risk have been seen in both case-control and cohort studies (69), suggesting that components other than fat, such as heat-induced carcinogens (70) or the high content of readily available iron, may be responsible (71). Like breast and colon cancer, prostate cancer rates are much higher in affluent compared with poor and Eastern countries (58). More detailed epidemiologic studies are limited, but intake of total fat generally has not been associated with this type of cancer (69). A positive association has been seen with intake of α-linolenic acid in some studies, but the overall evidence is unclear at this time (72).

Overweight is an important cause of morbidity and mortality, and short-term studies have suggested that reducing the fat content of the diet induces weight loss (73). However, in randomized studies lasting a year or longer, reductions in fat to 20% to 25% of energy had minimal effects on overall long-term body weight (74).

In summary, there is little evidence that dietary fat itself is associated with risk of CHD. Metabolic and
epidemiologic data are consistent in suggesting that intake of partially hydrogenated vegetable fats should be minimized. Metabolic studies, epidemiologic studies, and randomized trials indicate that replacement of saturated fat by polyunsaturated fat reduces the risk of coronary heart disease, but the benefits are minimal if carbohydrate rather than unsaturated fat replaces the saturated fat. Although the evidence is more limited, controlled feeding studies of blood lipids as well as the experience of southern European populations suggest that consuming a substantial proportion of energy as monounsaturated fat in the form of olive oil is associated with low rates of coronary heart disease. Available evidence also suggests that total fat reduction has little effect on breast cancer risk, although reducing red meat intake likely decreases the incidence of coronary heart disease, diabetes, colon cancer, and possibly premenopausal breast cancer.


Because protein varies only modestly across a wide range of human diets, higher carbohydrate consumption, in practice, is the reciprocal of a low-fat diet. For reasons discussed under the topic of fat, a high-carbohydrate diet can have adverse metabolic consequences. In particular, such diets are associated with an increase in triglycerides and a reduction in HDL cholesterol (25), and these adverse responses are aggravated in the context of insulin resistance (10, 75, 76).

Complex Carbohydrates

The traditional distinction between simple and complex carbohydrates is not useful in dietary recommendations because some forms of complex carbohydrates, such as starch in potatoes, are very rapidly metabolized to glucose. Instead, emphasis is better placed on whole grain and other less-refined complex carbohydrates as opposed to the highly refined products and sugar generally consumed in the United States. Adverse consequences of highly refined grains appear to result from both the rapid digestion and absorption of these foods, as well as the loss of fiber and micronutrients in the milling process. The glycemic response after carbohydrate intake, which has been characterized by the glycemic index, is greater with highly refined foods as compared with less-refined whole grains (77). The greater glycemic response caused by highly refined carbohydrates is accompanied by increased plasma insulin concentrations and augments the other adverse metabolic changes caused by carbohydrate consumption, noted in the preceding section, to a greater degree than with less-refined foods (12). Higher intakes of refined starches and sugar, particularly when associated with low fiber intake (78), are associated with increased risk of type 2 diabetes (79, 80) and CHD (13, 78). In contrast, higher intake of fiber from grain products has been associated consistently with lower risks of CHD and diabetes (53, 81). Whether these benefits are mediated by only fiber itself, or in part by the accompanying micronutrients is not clear, but for practical reasons this distinction is not essential. Anticipated reductions in colon cancer risk by diets high in grain fiber have not been supported in most prospective studies (82). However, reduced constipation and risk of colonic diverticular disease (83) are clear benefits of such diets.

Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

Jul 27, 2016 | Posted by in PUBLIC HEALTH AND EPIDEMIOLOGY | Comments Off on Foundations of A Healthy Diet1

Full access? Get Clinical Tree

Get Clinical Tree app for offline access