Body mass index (BMI) is the cornerstone of the current classification system for obesity and its advantages are widely exploited across disciplines ranging from international surveillance to individual patient assessment. However, like all anthropometric measurements, it is only a surrogate measure of body fatness. Obesity is defined as an excess accumulation of body fat, and it is the amount of this excess fat that correlates with ill-health. We propose therefore that much greater attention should be paid to the development of databases and standards based on the direct measurement of body fat in populations, rather than on surrogate measures. In support of this argument we illustrate a wide range of conditions in which surrogate anthropometric measures (especially BMI) provide misleading information about body fat content. These include: infancy and childhood; ageing; racial differences; athletes; military and civil forces personnel; weight loss with and without exercise; physical training; and special clinical circumstances. We argue that BMI continues to serve well for many purposes, but that the time is now right to initiate a gradual evolution beyond BMI towards standards based on actual measurements of body fat mass.

Fast foods are frequently linked to the epidemic of obesity, but there has been very little scientific appraisal of a possible causal role. Here we review a series of studies demonstrating that the energy density of foods is a key determinant of energy intake. These studies show that humans have a weak innate ability to recognise foods with a high energy density and to appropriately down-regulate the bulk of food eaten in order to maintain energy balance. This induces so called 'passive over-consumption'. Composition data from leading fast food company websites are then used to illustrate that most fast foods have an extremely high energy density. At some typical outlets the average energy density of the entire menus is approximately 1100 kJ 100 g(-1). This is 65% higher than the average British diet (approximately 670 kJ 100 g(-1)) and more than twice the energy density of recommended healthy diets (approximately 525 kJ 100 g(-1)). It is 145% higher than traditional African diets (approximately 450 kJ 100 g(-1)) that probably represent the levels against which human weight regulatory mechanisms have evolved. We conclude that the high energy densities of many fast foods challenge human appetite control systems with conditions for which they were never designed. Among regular consumers they are likely to result in the accidental consumption of excess energy and hence to promote weight gain and obesity.

OBJECTIVE: To investigate the degree of bias in under-reporting of food intake in obese and non-obese subjects, hypothesising that under-reporting may be selective for either macronutrient content (carbohydrate (CHO), fat, protein, alcohol), speci®c food types or eating occasions (meals, snacks). DESIGN: Thirty-three women (18 obese, 15 non-obese) were recruited to a long-stay metabolic facility for 24 h. Ad libitum food intake was covertly measured throughout the study and a reported food intake completed at the end of 24 h. RESULTS: Reported total daily energy intake was signi®cantly lower than measured intake. Whilst meals were accurately reported, energy from snack foods eaten between meals was signi®cantly under-reported. (P`0.001) Reported total carbohydrate and added sugar intakes were signi®cantly lower than measured, whilst reported protein and fat intakes were not signi®cantly different from measured. Reported alcohol intake was also considerably lower than measured, but high variability prevented signi®cance. CONCLUSIONS: In both obese and non-obese women the major cause of under-reporting, as assessed by covert study design in subjects restricted within a metabolic facility, is the failure to report between-meal snack foods. There is some evidence for increased under-reporting in high CHO, but no evidence of a bias in under-reporting towards high fat or high protein foods.

Several epidemiological studies have observed an inverse relationship between people's habitual frequency of eating and body weight, leading to the suggestion that a 'nibbling' meal pattern may help in the avoidance of obesity. A review of all pertinent studies shows that, although many fail to find any significant relationship, the relationship is consistently inverse in those that do observe a relationship. However, this finding is highly vulnerable to the probable confounding effects of post hoe changes in dietary patterns as a consequence of weight gain and to dietary under-reporting which undoubtedly invalidates some of the studies. We conclude that the epidemiological evidence is at best very weak, and almost certainly represents an artefact. A detailed review of the possible mechanistic explanations for a metabolic advantage of nibbling meal patterns failed to reveal significant benefits in respect of energy expenditure. Although some short-term studies suggest that the thermic effect of feeding is higher when an isoenergetic test load is divided into multiple small meals, other studies refute this, and most are neutral. More importantly, studies using whole-body calorimetry and doubly-labelled water to assess total 24h energy expenditure find no difference between nibbling and gorging. Finally, with the exception of a single study, there is no evidence that weight loss on hypoenergetic regimens is altered by meal frequency. We conclude that any effects of meal pattern on the regulation of body weight are liely to be mediated through effects on the food intake side of the energy balance equation. Meal frequency: Nibbling: Gorging: Energy balanceThe present paper reviews the epidemiological studies relating meal frequency to body weight, and attempts to integrate these with the results of physiological investigations on meal frequency and energy metabolism. The paper will focus exclusively on human subjects because, although there is a considerable literature describing the metabolic effects of differing meal patterns in experimental animals (Fabry & Tepperman, 1970;Adams & Morgan, 1981; Bellisle, 1995), there are important ecological differences between the natural feeding patterns of species, and because the human experimental data are adequate to answer the key questions relating to the efficiency of energy utilization in this context. EPIDEMIOLOGICAL ASSOCIATIONS BETWEEN HABITUAL MEAL FREQUENCY AND BODY WEIGHTThe extensive work of Fabry and co-workers (Fabry et al. , 1966Hejda & Fabry, 1964) was the first to demonstrate strong and reproducible inverse relationships between habitual meal frequency and body weight in human subjects. Results from their first large sample study involving 379 Czechoslovakian men aged 60-64 years are illustrated in Fig. 1. Both the proportion of overweight subjects and mean skinfold thicknesses were significantly inversely related to meal frequency. Similar trends were observed in a smaller sample of eighty-nine men aged 30-50 years (Hejda 8z .

In patients with HIV infection, total energy expenditure is reduced during episodes of weight loss. Reduced energy intake, not elevated energy expenditure, is the prime determinant of weight loss in HIV-associated wasting.