Dietary determinants of obesity
Dr S Jebb, MRC Human Nutrition Research Centre, Elsie Widdowsdon Laboratory, Fulbourn Road, Cambridge CB1 9AL, UK. E-mail: Susan.Jebb@mrc-hnr.cam.ac.uk
One of the few incontrovertible facts about obesity is that weight is only gained when energy intake exceeds energy needs for a prolonged period. This is clear from studies of basic physiology under standardized conditions and controlled intervention studies involving manipulations of the components of energy balance (1). However, energy intake must be considered in the context of an individual’s energy needs. Obesity results not from a high absolute energy intake but from intake which exceeds energy needs, even as a small fraction of energy flux. It is thus the coupling between intake and expenditure that lies at the heart of the problem (2).
The search for specific dietary factors that increase the risk of obesity is therefore a quest for factors that undermine the innate regulatory control of body weight. There are multiple mechanisms by which this can occur which are explored elsewhere in this series of reviews, including satiety, palatability, food availability or low-energy needs as a consequence of physical inactivity. This review will focus on the evidence for specific dietary determinants of obesity largely from observational and intervention studies.
Research into the dietary determinants of obesity has largely been based on observational studies of intake and weight or of body mass index (BMI). However, the evidence is weak and inconsistent. In part, this may be a true reflection of the multifactorial nature of the problem, but it also relates to methodological difficulties inherent in this approach that are not easily overcome.
Body weight is the integrated product of a lifetime’s diet and exercise habits, and so nutrients, foods or broader dietary habits measured on a small number of occasions may not be related to the longer-term development of obesity. Many dietary factors are highly correlated, and physical activity or other lifestyle traits are other important covariates. Cross-sectional studies are confounded by post hoc effects, in which dietary differences between individuals arise as a consequence of obesity rather than as a causal factor. Prospective studies can overcome the potential for reverse causality but many rely on basic questionnaires to determine dietary intake. Indeed, the greatest limitation in most studies is the reliability of the data on dietary exposures. In most dietary surveys, energy intake is under-reported, relative to estimated energy needs, by an average of 25% (3). This error is not random and is significantly greater among obese people. The nutritional composition of the ‘energy gap’ is unclear and statistical adjustments tend to be made based on energy alone.
Despite the limitations of observational studies, there is a paucity of controlled dietary intervention studies to test aetiological hypotheses. In addition, a flawed logic often leads to clinical trials to test whether reductions in specific foods or nutrients leads to weight loss, rather than the prevention of weight gain. In either case, dietary intervention studies are challenging. First, in situations in which subjects must prepare their own food, it is rarely possible to design a double-blind trial. Second, the process of informed consent and baseline screening may act as an intervention in itself by raising awareness of the diet or health issue and prompting behaviour change. There is a risk that the control group may more readily adopt generic public health messages, while the intervention group are unlikely to adhere to the intervention as precisely as requested, thus reducing the planned difference between groups. Dietary change is difficult to sustain as traditional dietary practices are well established and strongly habituated. Numbers of dropouts from trials are frequently high and the extent of compliance is often difficult to measure. In terms of prevention in particular, the mean weight change is small (especially superimposed on the background population weight increase) and there is large inter-individual variability which has implications for sample size and power. Accordingly, there are few large-scale, long-term, well-controlled dietary intervention studies.
What do we know?
The literature pertaining to dietary determinants of obesity has been recently reviewed by the World Health Organization (4). While basic physiological principles dictate that habitual energy intake must exceed expenditure for weight gain to occur, relatively few cross-sectional or prospective studies have demonstrated this effect, presumably as a consequence of the limitations in the methodology for collecting data in free-living populations (5). Indeed, research has shown very few clear nutrient- or food-based determinants of obesity. The strongest evidence for an increased risk of obesity relates to diets that are high in dietary fat or low in fibre.
Dietary fat is readily stored as body fat, with minimal energy costs of conversion relative to protein or carbohydrate (6). Fat is less satiating than iso-energetic quantities of other nutrients and habitual consumption of a high-fat diet may down-regulate some elements of the appetite control system. Evidence from observational studies of a specific role for dietary fat is inconsistent (7,8). Intervention studies show modest but significant spontaneous weight loss in people who reduce the fat content of their diet (9). In the recent Women’s Health Initiative, there was a clear dose–response relationship (10). These effects are believed to be mediated primarily through energy density (11).
There is strong experimental support for the ‘energy density’ hypothesis, with plausible mechanisms operating at all levels of the appetite control system, from early gastric distention, the modulation of gut hormones, through to post-ingestive metabolic effects (12). However, supporting data from observational studies of intake among free-living subjects are limited. The theory that energy-dense foods undermine innate appetite control systems is consistent with the concept of ‘fast food’ as a specific risk factor (13), as fast food typically has an energy density very much higher than ‘household’ food (14).
Different types of fat have different metabolic effects and it is possible that this extends to differences in the risk of weight gain (15). However, there is less research on specific fat types, not least because of the difficulties in defining fatty acid intake. Animal studies suggest that saturated fatty acids may be preferentially stored, while more unsaturated fats are more likely to be oxidized. If this has knock-on effects on appetite control, it may provide a plausible mechanism for effects on weight gain and would imply that saturated and trans-fatty acids may be a specific risk factor for obesity.
The proportion of carbohydrate in the diet tends to vary reciprocally with fat and it is difficult to segregate the impact of the total amount of carbohydrate in the diet from total fat (16). While some studies show a protective effect of a high proportion of carbohydrate, many others show no association, particularly for studies in children (17–19). The evidence relating the intake of sugar per se to weight change is inconsistent (20). In part, this may be due to the various sources of sugar in the diet including fruit and milk as well as ‘added’ sugar.
However, a large number of studies have shown an inverse association between fibre intake and weight gain (although the precise definition of dietary fibre is somewhat variable) and intervention studies also show that a high intake of dietary fibre may assist weight loss (21). This may be related to the incomplete digestion and absorption of energy from this type of carbohydrate. In addition, the bulky nature of high-fibre foods, with increased demands on chewing and subsequent gastric distention, may increase satiety and curtail energy intake (22). High-fibre foods may also enhance satiation through delayed gastric emptying and the attenuation of postprandial glucose and insulin responses. It may also have an impact on other gut hormones involved in appetite regulation, such as cholecystokinin. Wholegrain intake has also been negatively associated with BMI and weight gain (23,24), although the effect size is usually attenuated after adjustment for the fibre content of the diet. There is very little data to support an association between glycaemic index and adiposity, although there may be other benefits of a low glycaemic index diet on obesity-related disease.
There has been less research into the effect of protein, perhaps because it makes a smaller contribution to total energy intake than fat or carbohydrate. Evidence from observational studies relating to the proportion of protein and obesity are inconsistent. It is pertinent to note that, in experimental studies, protein preloads are associated with reductions in subsequent intake relative to iso-energetic quantities of other macronutrients, suggesting it may act as a satiety cue (25). The observation that protein is a critical determinant of food intake in insects and prioritized over energy requirements has led to the ‘protein-leverage’ hypothesis (26). The impact of different types of protein is unclear.
There has also been very little consideration of the potential role of micronutrients in the aetiology of obesity, although nutrient imbalances may theoretically impact on the appetite regulation system.
Specific foods: nuts, dairy, sugar-rich drinks, alcohol
A variety of specific foods have been implicated in the aetiology of obesity. Limited data suggest fruit and vegetables may have a modest protective effect (27,28). People who eat the greatest quantities of nuts tend to have a lower body weight (29), but this tends to be associated with many other dietary differences and there is a strong possibility of residual confounding in these analyses. However, intervention studies in which volunteers consumed a small quantity of nuts as a supplement to their usual diet were not associated with weight gain, suggesting this warrants greater consideration (30). Preliminary data have suggested a protective effect of dairy products, but this finding is inconsistent, both in observational and intervention studies (31). Sugar-rich drinks have attracted much attention because their low viscosity is associated with poor satiation (32). Cohort studies show either no relationship or a positive association between sugar-rich drinks and weight gain or obesity in adults or children (33). In observational studies, high-alcohol intake is not associated with increased body weight, although experimental studies suggest that energy from alcohol supplements rather than substitutes for food energy (34). This is a particularly complex area to study given the specific under-reporting of alcohol intake, confounded by other sociodemographic and lifestyle variables and ethical difficulties in conducting intervention studies.
There is growing interest in dietary patterns and eating behaviour rather than specific nutrients or foods. The apparently protective effect of low-fat and/or high-fibre diets is broadly supported by this work, which suggests a protective effect of the Mediterranean or so-called ‘prudent’ dietary pattern characterized by high intakes of vegetables, fruits, legumes, nuts and olive oil, together with a reduced intake of meat/meat products and full-fat dairy products (35–37).
Habituation to large portion sizes is likely to be a risk factor, although this has been poorly studied in large cohorts, which frequently rely on food frequency questionnaires to collect dietary data. However, experimental studies suggest that large portions tend to increase energy intake at a meal, with no increase in satiety and little compensation at subsequent eating episodes (38). There is no evidence that eating frequency per se (snacking) is linked to obesity (39), although the qualitative nature of the eating episodes may be important.
Thus, data from diverse sources suggest that foods with a high energy density, sugar-rich drinks and large portion sizes each increase the risk of over-consumption of energy. Traditional dietary patterns such as the ‘Mediterranean’ diet, rich in fruit, vegetables and unrefined carbohydrate, may be associated with a decreased risk, while diets rich in fat and added sugars and low in fibre, energy-dense foods typical of a ‘fast food’ culture, may increase the risk of weight gain.
Research in this area is severely hampered by methodological limitations and there is a concern that current research paradigms may have limited potential to detect what appear to be subtle effects of diet composition on the regulation of body weight. The aetiology of obesity is extremely complex. Food choice and other dietary habits are the product of a wide-ranging of determinants. Additionally, diet must be considered alongside other causal agents, not least physical inactivity.
There is no substitute for time and effort to collect dietary data of the highest possible quality. In the context of large prospective cohorts, resources must be invested in detailed measures of dietary exposures, alongside individual assessments of physical activity and energy needs to address the issue of energy balance rather than energy intake in isolation.
Progress may also be made in understanding broader eating habits, rather than individual nutrients or foods, that are associated with excess energy intake. Improvements in statistical methods to understand the nature and implications of the errors and to refine the calculation of effect sizes in relation to diet and health outcomes are helpful, but are unlikely to overcome the inherent problem in quantifying the nature of the missing energy. More investment is warranted in biomarkers of dietary exposures to test specific dietary hypotheses, although it is unlikely that, given the complexity, it will be possible to define population-attributable risk factors for specific dietary components.
These population studies will increasingly include genetic testing to identify nutrient–gene interactions. Enhanced diligence in dietary assessment will reap benefits in other areas of nutritional epidemiology and contribute to a better understanding of a ‘healthy diet’. Moreover, dietary hypotheses should extend beyond studies of consumption and out into the wider context in which food is selected, including issues of cost, food access and the social norms that drive food choice. These studies also need to encompass physical activity and its determinants to give a more holistic perspective.
Given the limitations of observational data, plausible dietary hypotheses need testing in controlled trials. This may include short-term experimental studies to identify the precise effects of dietary change on food choice, appetite control and ultimately energy intake. Longer-term, community-based studies must clearly separate the efficacy of the intervention from adherence to the treatment. More research is needed now to understand the factors associated with improved adherence to dietary prescriptions. There is also a need to develop new statistical approaches that are less reliant on randomized, controlled trials.
Looking to the future and a time when it is feasible to accurately measure dietary intake (perhaps as a result of technological breakthroughs, improved statistical methods or robust biomarkers), it will undoubtedly be possible to build a more robust evidence base on the dietary determinants of obesity. However, it is important to recognize that this is only the first step in tackling the problem of excess weight gain. Improved knowledge and understanding of effective methods to stimulate changes in dietary habits will be vital. This must consider the role of interventions focused on individuals, together with wider environmental changes, introduced by policymakers or other stakeholders. Research in these areas now will ensure that, in due course, improved knowledge of the specific dietary determinants of obesity can be translated into effective strategies for the prevention of excess weight gain.
In this future scenario, it is also plausible that the evidence will demonstrate that obesity is simply a question of energy balance, which may be achieved with diets of widely different food choices or nutrient composition, rather than a direct consequence of certain nutrients or foods. Accordingly, it may be argued that public policies to redirect food choice are unjustified and, instead, greater efforts should be invested more generically in the attainment of energy balance. But this argument fails to recognize the contribution of dietary composition to the modulation of obesity-related disease. Diet affects so many aspects of physical and mental health, it would be unwise to segregate the prevention of obesity from these wider considerations. It may be that greater emphasis should be placed on linking dietary habits to long-term health outcomes, including the importance of energy balance, rather than considering obesity as an isolated risk factor. This approach provides opportunities for action to improve dietary habits now, while leaving open the possibility of more targeted interventions as the evidence is refined.
Conflict of Interest Statement
No conflict of interest was declared.