Can adherence to dietary guidelines address excess caloric intake? An empirical assessment for the UK

The facilitation of healthier dietary choices by consumers is a key element of government strategies to combat the rising incidence of obesity in developed and developing countries. Public health campaigns to promote healthier eating often target compliance with recommended dietary guidelines for consumption of individual nutrients such as fats and added sugars. This paper examines the association between improved compliance with dietary guidelines for individual nutrients and excess calorie intake, the most proximate determinant of obesity risk. We apply quantile regressions and counterfactual decompositions to cross-sectional data from the National Diet and Nutrition Survey (2000-01) to assess how excess calorie consumption patterns in the UK are likely to change with improved compliance with dietary guidelines. We find that the effects of compliance vary significantly across different quantiles of calorie consumption. Our results show that compliance with dietary guidelines for individual nutrients, even if successfully achieved, is likely to be associated with only modest shifts in excess calorie consumption patterns. Consequently, public health campaigns that target compliance with dietary guidelines for specific nutrients in isolation are unlikely to have a significant effect on the obesity risk faced by the population.

m odest changes in patterns of excessive energy intake, and consequently, with only sm all shifts in obesity risk in the population 5 .

. Obe s ity a n d Adh e re n ce to Die tary Gu id e lin e s in th e U K
The Cabinet Office report Food Matters (J arvis, 20 0 8 ) identifies the promotion of healthier dietary choices by consumers as a key element in the UK Government's food strategy for the 21 st century. This follows from the recognition of the enormous health gains that would accrue to the UK if diets matched nutritional guidelines on fruit and vegetable consumption, saturated fats, added sugars and salt intake. The Cabinet Office report estimates that adherence to nutritional guidelines would reduce the risks related to cancer, heart disease and other illnesses leading to 70 ,0 0 0 fewer people dying prem aturely every year. Im proved dietary choices are also crucial for meeting the challenge of obesity, with over a quarter of adults and 10 % of children in the UK already classified as "obese". In addition to the social im pacts, the econom ic burden of diet related ill-health is estimated at almost £ 6 billion a year by way of additional National Health Service costs alone.
The changing Body Mass Index (BMI) profile of the UK adult population based on data from the Diet and Nutritional Survey of British Adults 1986-87 (OPCS, 1991 and the National Diet and Nutrition Surveys of 20 0 0 -0 1 (ONS, 20 0 5) and 20 0 8 -0 9 (Bates, Lennox and Swan, 20 10 ) is presented in Table-1 which summarises the quantiles of the BMI distribution in the three tim e periods. While the average BMI has steadily increased, what is significant and of greater concern is the rightward shift of the BMI distribution. The proportion of adult individuals who are overweight (BMI≥25 and <30 ) or obese (BMI≥ 30 ) has increased from 40 % to 62%, while the proportion of individuals who are obese has increased from 10 % to 27%.

[ Ta ble -1 h e re ]
Obesity is a characteristic with a com plex set of socio-econom ic, dem ographic, environm ental and cultural determ inants (Ulijaszek,20 0 7). A wide array of these factors has been exam ined in the literature, including access to food, lifestyle factors, fast-food consumption (Chou et al., 20 0 4), econom ic well-being and security (Sm ith et al., 20 0 9), food advertising (Andryeva et al.,20 11), physical activity and the built environment (Salois,20 12,Sandy et al.,20 12). The influence of these factors arises through their effect on the two most proxim ate determ inants of body weight -food choice and the resultant nutrient and energy composition of diets and energy expenditure. In this paper, while recognising the wide array of determinants that influence obesity, we focus on the association between the nutrient composition of diets and excessive energy intake and the consequent risk of obesity.
The rising trend in the incidence of obesity has been associated with significant deviations from the recom m ended dietary guidelines suggested by the World Health Organization (20 0 3) and national public health authorities. These dietary guidelines are chiefly related to the share of energy derived from macronutrients (fats and sub-com ponents, sugars and protein) and the absolute intake levels of fruits and vegetables, fibre, salt and cholesterol. In the US, which reports the highest incidence of obesity in the developed world (OECD, 20 12), Volpe and Okrent (20 12) report that consumer diets are a long distance away from conform ity with the Dietary Guidelines for Am ericans (US Departm ent of Health and Hum an Services and the US Departm ent of Agriculture, 20 0 5). Using the Healthy Eating Index (HEI) (US Department of Agriculture, 20 10 ), an overall measure of diet quality based on adherence to the Dietary Guidelines, Volpe and Okrent (20 12) find that the average HEI score for food-at-home purchases of US consumers was 56.4, far below the maximum score  Table-2. Inform ation on trends in adherence to key dietary guidelines in the UK over last the two decades is sum m arised in Table-3. For 20 0 8 -0 9, the average contribution to energy intake from all fats (34.8 %), saturated fats (13.2%) and non-milk extrinsic sugars (13.5%) was clearly in excess of the prescribed guidelines, while the contribution of polyunsaturated fatty acids (PUFA) to energy (5.2%) and average cholesterol consumption were within the acceptable range.
Average fruit and vegetable consumption fell well short of the recom m ended intake (40 0 grams per day), as did the intake of dietary fibre. The average consumption of salt remained considerably higher than the recom m ended guideline (less than 2.36 grams of sodium per day). The average figures, however, mask the true extent and severity of non-com pliance with the recomm ended guidelines. From an obesity risk perspective, it is the large proportion of non-com pliers in the population whose dietary choices are a matter of concern.
For instance, the average intake of energy from saturated fats (13.5%) m ay not appear to be a large violation of the recom m ended guideline for saturated fats. However, Table-3 shows that 86.6% of the adult population consum e saturated fats in excess of the recomm ended am ount, with 25% obtaining m ore than 15% of energy from saturated fats.

[ Ta ble -2 h e re ] [ Ta ble -3 h e re ]
The dietary guidelines suggested by the Department of Health (DoH) and the World Health Organization (WH O) are based on an extensive review of m edical and nutrition literature exam ining the relationship between dietary choices/ nutrient intakes and the prevalence of chronic disease and ill-health. They are intended to reduce the incidence of a range of chronic diseases such as diabetes mellitus, cardiovascular disease, hypertension, stroke, certain types of cancer and also combat obesity. It is no doubt true that not all of the dietary guidelines are aimed at combating obesity. However, the promotion of diets derived from these guidelines (such as the US Department of Agriculture's "MyPlate" and the UK Food Standard Agency's "Eatwell Plate") is a crucial element of the strategy to improve the dietary patterns of the population. The reduction in the risk of obesity that can be achieved through adherence to the recomm ended dietary guidelines remains an empirical question. It is this question, which could be of considerable interest to those designing public health campaigns and other measures to promote healthy eating, that we address in this paper. We develop a framework to assess the potential changes in the distribution of excessive energy intake in the population if adherence to prescribed dietary guidelines for nutrient intakes is successfully achieved.
We note that dietary guidelines in the UK also include Dietary Reference Values (DRVs) for overall energy intake. The Scientific Advisory Committee on Nutrition (20 11) provides DRVs based on estim ated average requirements (EARs) of infants, children, adolescents and adults. Clearly, adherence to recom m ended energy intake guidelines would have a significant effect on the obesity risk of the population. However, our framework considers only the potential effects associated with adherence to dietary guidelines for individual nutrients/ foods. This is because public health campaigns and other measures to promote healthier diets focus predominantly on adherence to individual nutrient intake guidelines and influence calorie consumption only indirectly. Our interest is in the patterns of calorie intake that are likely to em erge as a consequence of com pliance with nutrient intake guidelines.
In assessing the potential effects of adherence to dietary guidelines on obesity, it is tempting to model BMI (or other obesity indicators) as a function of intakes of nutrients or adherence to guidelines. However, using such an approach with cross-sectional data sets (which comprise the majority of nutrition data sets) raises a number of problems and issues which have been well recognised in the literature. J ebb (20 0 7) notes that "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" (p. 93). Variyam et al. (20 0 2) observe that nutrition studies using self-reported data intakes fail to find a strong positive association with between energy intake and obesity. This is attributed to the under-reporting of intakes by overweight persons and also to the possibility that at any point in time, a proportion of overweight persons m ay be on weight-loss diets. Rosenheck (20 0 8 ) finds "discrepant associations between frequent fast food consumption, increased energy or overweight in terms of BMI" (p. 535), with a number of studies finding a negative correlation between BMI and fast food consum ption. Data in cross-sectional surveys do not contain all the information for analysing the determinants of BMI, as BMI observed at any point in tim e is the result of a lifetime of consumption and habits (Komlos and Brabec,20 11). Therefore, relating BMI directly to dietary intakes in a cross-sectional setting m ay not be a useful approach for assessing the obesity implications of adherence to dietary guidelines 6 . In a cross-sectional setting, observed dietary intakes can signal only the potential risk of weight gain/ obesity faced by different segments of the population.
A second issue to be addressed is that the effect of adherence to dietary guidelines m ay vary along the distribution of obesity risk. For instance, the effect of a unit increase in the share of energy derived from saturated fats may be very different for individuals facing different levels of obesity risk. A number of studies in the literature have used multiple linear or logistic regressions to analyse the effects of dietary intakes on BMI. These approaches assess the m ean response of the outcome variable to changes in explanatory variables. In these approaches, the effect of the covariates is same along the whole range of outcomes -for instance, the effect of a unit increase in fat consum ption on BMI would be the sam e for an underweight or obese individual. In designing public health campaigns, we are m ore interested in the effect of adherence to guidelines in the upper tails of the outcome variable (denoting higher obesity risk). The heterogeneity of response to adherence to different dietary guidelines at different levels of overweight/ obesity risk is a key elem ent of interest.
We would like to explore the hypothesis that the effects of adherence to dietary guidelines could vary along the whole distribution of obesity risk and could be significantly different from the mean response values in the upper tail of the obesity risk distribution. The effects of compliance could be very different for potentially obese and non-obese individuals.
A third issue is that in assessing the potential effects of adherence to dietary guidelines, we are m ore interested in how the population distribution of obesity risk would change as a result of compliance, rather than in the average effects of compliance with dietary guidelines.
Even if the entire population were to conform to the recom m ended dietary guidelines, there would still be substantial variation across individuals in the degree of conformity. We are interested in the potential distribution of obesity risk that can be expected in a population of compliers.

. Me th o d s
Given the poor correlation between BMI and dietary intakes in a cross-sectional context, we develop a proxy indicator for obesity risk relying on one of the few facts about obesity that is not controversial -"that weight is only gained when energy intake exceeds energy needs over a prolonged period"(J ebb,20 0 7, p.98). Large scale nutrition data sets, however, do not record food consumption over an extended period and we have to rely on indicators derived from cross sectional data. We build a measure of "excess calorie consumption" (ECC) and m odel it as a function of the nutrient com position of diet. ECC is computed as (Total calories consum ed/ Total calorie need)*10 0 . If total calories consum ed equal total calorie need, then ECC equals 10 0 . ECC <10 0 denotes deficient calorie consum ption. Adherence to dietary guidelines influences obesity risk via its association with ECC. We use ECC in our analysis because it clearly signals obesity risk and at the same time appears to have a systematic association with the nutrient composition of diets available from cross-sectional data. To assess calorie need, we rely on the predictive equations for resting energy expenditure Committee on Nutrition (20 11) in the UK for deriving the dietary reference values for energy for different segments of the population and in other studies on energy requirem ent in hum ans (e.g., Pellet, 1990 ).
The ECC measure derived from observation of food consumption over a short period of time in cross-sectional data may not reflect the cumulative consumption patterns of the respondents. Some individuals may be on weight loss diets and their observed nutrient intake profile may not reflect their consumption history 9 . The observed pattern of ECC will translate into the risk of obesity only if it is m aintained over a period of tim e. However, the NDNS is a nationally representative sample, and the ECC distribution derived from it will reliably reflect the more stable consumption patterns for the population as a w hole. Our approach to assessing the effects of conformity to dietary guidelines requires only a robust approxim ation of the ECC distribution of the population, which the NDNS can provide. The QR results are used to carry out a "counterfactual decomposition" to assess the shifts in the distribution of excess calorie intake associated with adherence to dietary guidelines. We use the technique used by Machado and Mata (20 0 5) which allows changes in the distribution of the outcome variable to be decomposed into "co-efficient" and "covariate" effects. For instance, if ECC is related to fat consum ption, then a change in ECC over two tim e periods could be due to an increase (or decrease) in fat consumption (covariate effect) and/ or due to a change in the effect of fat consumption on ECC (co-efficient effect). The same technique can also be used to assess how the distribution of the outcome variable would change if the distribution of one the covariates were to change, other covariates remaining the same. We use this counterfactual decomposition technique to assess how the

[ Ta ble -4 h e re ]
The variables used in this paper are summarised in Table- to give replies that are seen as being socially appropriate (Warnecke et al, 1997). The underreporting of nutrient intakes and over-reporting of physical activity m ay imply that ECC m ay not be estimated precisely and may be subject to a significant downward bias. This is an im portant caveat to be kept in m ind when interpreting the data on ECC. The counterfactual decomposition method used in this paper assesses the likely change in the ECC distribution associated with adherence to dietary guidelines (relative to existing patterns). The results provide useful insights into the direction and m agnitude of changes in ECC that can be expected from improved conform ity to dietary guidelines, notwithstanding the potential downward bias in ECC estim ated from NDNS data.

Re s u lts a n d D is cu s s io n 5 .1 Qu an tile Re gre s s io n Re s u lts
The linear QR m odel was estim ated was: 12 The ethn icity variable in the NDNS 20 0 0 -0 1 refers to self-reported m em bership of one of the following groups -White, Black-Caribbean, Black-African, Black-Other, Indian, Pakistani, Bangladeshi, Chin ese or 'None of These'. These groupings include an elem ent of the region/ country of origin besides race. Therefore, we use the term "race/ ethnicity" to describe these groups -but have grouped these nine categories in to four -White (reference group), Asian, Black an d Others.
ECC = 1 + 1 satfat + 2 pufat + 3 m onofat + 4 sugar + 5 prot + 6 star + 7 frv eg + 8 fibre + 9 sodium + 10 sm oking + 11 alcoh + 12 chol + 13 phyact + 14 age + 15 fem ale + 16 race-ethnic + 17 m anual + where the dependent variable ECC denotes excess calorie consumption, the explanatory variables are as described in Table- It m ay be noted that the share of energy from all m acronutrients will sum to 10 0 %. A unit increase in the share of energy from one m acronutrient im plies a decline in the share of som e other m acronutrient. In the equation above, the share of energy derived from carbohydrates (other than from sugar and starch) has been omitted as it is treated as "residual".
14 The physiological effects of changes in the m acronutrient com position of diet on calorie intake m ay operate through energy den sity effects and through the differing effects of m acronutrients on appetite intake -for fruits and vegetables, fibre, sodium (salt) and cholesterol consum ption. The coefficient for fruit and vegetable intake can be interpreted as the effect of moving to a diet with a unit increase in fruit and vegetable intake. The consumption of fibre, salt and cholesterol can generally increase only as a result of increased consumption of foods in which they are constituents. The effect of increased consumption of these nutrients on calorie consumption arises on account of the increase in consumption of the associated foods 15 . We would, therefore, expect the coefficients of these nutrients to be positive unless there are large offsetting effects. For instance, increased consumption of foods containing fibre would add to calorie consumption-so we would expect the coefficient of fibre to be positive. A negative coefficient would arise only if the move to a diet with a higher level of fibre consumption is associated with a reduction in calorie intake from other foods in the observed diets of the population. The effects of smoking and alcohol consum ption m ay arise from their effects on behavioural, sensory and physiological processes influencing the ingestion of calories from all other foods. Alcohol, however, is itself a source of calories, therefore, its coefficient will reflect the net effect of a unit increase in alcohol consum ption arising from its own calorie content and its effect on consumption of other foods. As noted previously, the WHO and UK dietary guidelines are based on an extensive survey of the literature on how individual nutrient intakes affect obesity and chronic diseases. We do not exam ine the physiological effects of changes in individual nutrient intakes here. Our m ain interest in this paper is to assess (1) how the changes in ECC associated with conform ity to dietary guidelines vary across ECC quantiles, i.e., across segments of the population with low or high risk of obesity and (2) how these effects are likely to play out em pirically in a population of compliers (in the context of an ecological study of free living populations).
control and satiety (Stubbs et al.,20 0 0 ). The effect on calorie intake m odelled here arises as result of m ovin g from a diet with a lower share of energy from a m acronutrient to a diet with a higher share of energy from that m acronutrient. The graphs in Figure At the same time, other studies (Willett, 1998a;1998b; 20 0 2) have reported that reduction in dietary fat intake in obese individuals is not associated with significant weight loss. Our results suggest that the findings from different studies need not be viewed as being contradictory. While the m ean response to an increase in the share of saturated fat in energy intake is an increase in excess calorie consum ption (as shown by the OLS results in Table- Cigarette smoking does not appear to have a significant assoication with ECC in any of the quantiles 16 . However, increased alcohol consumption is clearly associated with reduced ECCan effect which is consistent and significant across all quantiles and is m ore pronounced in the higher quantiles of ECC. This suggests that the increase in calories from a unit of alcohol is more than offset by reduced intake of calories from other foods 17 .
ECC increases with the age of the respondents. This effect is small but significant across quantiles and increases as we m ove up the ECC quantiles. The gender of the respondent has a much larger effect on ECC. The OLS results suggest that ECC of fem ales is lower by 4.83 percentage points in relation to that of m ales. The effect of gender declines (from 7.21 percentage points in the 5 th quantile) as we move to the higher quantiles but is still substantial (3.36 percentage points) and significant in the 90 th quantile. Asian race/ ethnicity is associated with substantially higher ECC (2.8 9 to 12.34 percentage points higher) compared to Whites (the base group). This effect is much more pronounced in the lower quantiles than in the higher quantiles of ECC. A sim ilar association is seen in the case of the Black racial/ ethnic group although the effects are sm aller and not significant in m any of the quantiles.
As expected, increased levels of physical activity are associated with reduced ECC. The effect of physical activity is largest in the top quantiles of ECC. Thus, individuals who are most prone to obesity through excess calorie intake will benefit the most from increased physical 16 Although a number of studies have found an in verse relationship between sm okin g behaviour and body weight and a rebound weight gain on cessation of sm oking, the em pirical evidence on the effect of smoking on body weight is mixed. Smoking may affect body weight not only through its effect on appetite and energy intake, but also through its effect on m etabolism (Wack and Rodin, 1982), fat accumulation and physical activity. Using genetic instruments to model the effects of sm oking, Wehby et al (20 12) find considerable heterogeneity of effects across BMI quantiles -with sm oking increasing BMI at low/ m oderate levels of BMI but decreasin g BMI at high levels of BMI. The coefficient of smoking in our model gives us only the effect on ECC.
17 Excessive alcohol consum ption has been found to be associated with reduced intake of energy from fats and carbohydrates (Windham and Hansen, 1983). However, excessive alcohol consumption causes liver and health disorders and reduced energy intake may be attributable to the effects of the health condition rather than to a diminution of appetite caused by alcohol. Moderate alcohol consumption has been foun d to provide short term stim ulation of en ergy intake in a num ber of studies (Poppitt et al, 1996; Yeom ans, 20 0 4). However, this short term stim ulation of energy intakes can be offset by lower energy intakes in later m eals or on alcohol free days. NDNS data provide the average consum ption over a 7 day period based on respondent diaries. An overall negative effect of alcohol consumption on ECC can be consistent with the short term stimulation of energy intake provided by moderate alcohol consum ption.
activity. However, it should be noted that the effect of physical activity is much smaller compared to the effects of age, gender and race/ ethnicity. The occupational category (m anual or non-m anual) has an interesting association with ECC. In the lower quantiles, a m anual occupation is associated with reduced ECC. But in the higher quantiles a m anual occupation is associated with sharp increases in excess calorie intake in relation to those in non-m anual occupations.
The QR results are useful in identifying segments of the population that are likely to be prone to excessive energy intakes (and hence at the risk of obesity). Public health cam paigns to promote healthier eating could target or prioritise these groups. For instance, our results suggest that older males in manual occupations are prone to excessive calorie consum ption.
Sim ilarly, the large and significant QR coefficients associated with Asian race/ ethnicity (across ECC quantiles) suggest that this group is prone to excessive calorie consumptionand hence m ay need to be prioritised even if their average calorie consumption patterns are not very different from those of the reference (White) group 18 .

.2 Co u n te rfactu al D e c o m p o s itio n s
The   Adherence to the guidelines for fruit and vegetable, fibre, sodium and cholesterol intake is associated with increased ECC, with the magnitude of the effect increasing as we move up the ECC quantiles. The shifts in ECC associated with adherence to these guidelines need to be interpreted somewhat differently. As already noted, these guidelines are specified in terms of absolute levels of intake rather than in terms of their contribution to food energy.
Table-6 shows that adherence to the fibre guideline would be associated with increased ECC and would substantially increase the share of the population consuming excess calories. As m ay be seen from Table-3, average fibre consumption in UK adults is much lower than the recomm ended guideline with 84.3% of the population not conform ing to the guideline in 20 0 0 -0 1. Adherence to the guideline will call for substantial increases in fibre consumption and consequently for increased consumption of fibre containing foods. The QR m odel is anchored in the consumption patterns of the respondents in the NDNS 20 0 0 -0 1. The effect of adherence to the fibre consumption guideline in Table-6  process is more appropriate for modelling the dietary changes of free living populations in 19 We have referred to som e the studies in the discussion of the QR results (Section 5.1) response to public health campaigns rather than the (controlled) dietary changes imposed in intervention studies. The main reason why adherence to dietary guidelines produces such lim ited shifts in the ECC distribution is because the compliant segments of the population also consume excessive energy. Given the nature of the dietary adjustment process assumed in the paper, it is important that our QR and CFD results are not interpreted as the physiological effects of changes in nutrient composition of diets (e.g., a unit increase in the share of energy derived from fats). The "effects" of compliance with dietary guidelines on ECC that we report from our empirical model arise as a result of non-compliant diets being replaced with compliant diets observed in the population.

.3 Im p licatio n s fo r Co m batin g Obe s ity
We have m odelled ECC as a function of nutrient com position of diets, lifestyle factors, physical activity and demographic variables. Persistent excessive energy intake in relation to need unambiguously signals obesity risk. Our results, therefore, provide a number of insights into how com pliance with dietary guidelines is likely to influence the distribution of obesity risk in the population. Effects of adherence to individual dietary guidelines vary considerably and the potential reduction in obesity risk (or "returns") from adherence to specific guidelines can vary across ECC quantiles (i.e., across groups with high or low risk of obesity).
In general, adherence to individual m acronutrient dietary guidelines in isolation (i.e., independently of other guidelines) is likely to produce relatively sm all shifts in the distribution of obesity risk. This is because com pliance with individual guidelines is not necessarily associated with significant reductions in excess calorie intakes. The high shares of fats and sugars in food energy intakes have been implicated as major factors responsible for the obesity "epidemic". However, our results suggest that compliance with fat and sugar guidelines m ay not lead to a large m ovem ent in the ECC distribution. The counterfactual decomposition exercise demonstrates that a large proportion of the population that complies with the guidelines continues to consum e excessive calories. Compliance with the guidelines will not result in large changes in obesity risk, if as our analysis suggests, consumers are able to comply without significantly reducing calorie intake. In the case of proteins, the average share of energy derived from proteins is currently in excess of the guideline and is inversely related to ECC. Compliance with the guideline would call for a reduction in the share of energy derived from proteins and lead to an increase in ECC, increasing obesity risk. We recognise that com pliance with individual m acronutrient guidelines m ay have substantial health benefits related to the prevention of chronic disease -but compliance with individual guidelines in isolation may do little to address the problem of excess energy intake that drives obesity risk.
Adherence to guidelines for fruit and vegetable and fibre consumption (where the current average consum ption falls below the recom m ended am ount) m ay be associated with increased ECC, worsening obesity risk. Campaigns targeting fruit and vegetable consum ption (e.g., provision of free fruit to school children to meet the five-a-day guideline) and fibre intake need to take into account these unintended consequences. The choice of the basket of foods through which compliance is achieved (their calorie content and the presence of other "unwanted" nutrients) is an im portant factor influencing ECC patterns. In the case of salt consumption, achievement of the guideline can have a large effect on ECC and consequently on obesity risk. It should be noted that the ECC reducing effects of compliance with the salt guideline m ay be (ironically) dam pened by the efforts of the food industry to develop and market healthier product options (with lower salt content). This is because healthier option products make it easier for consumers to comply with the guideline without substantially reducing calorie consumption.
It is only sim ultaneous com pliance with a set of dietary guidelines for nutrient intake that can potentially produce large shifts in the ECC distribution of the population. For com batin g the obesity risk of the population, public health measures will have to target combinations of dietary guidelines that are associated with the largest shifts in the ECC distribution. It is important to recognise that even when simultaneous compliance with all the recommended dietary guidelines is achieved, the ECC-reducing effects of compliance with certain guidelines m ay be (partially) offset by the ECC-increasing effects of compliance with other guidelines. The need for simultaneous compliance with dietary guidelines to h ave a significant effect on obesity risk may provide clues to resolve what is an apparent paradox in the nutrition data -the rising incidence of obesity over the last two decades even as consum ers have been m aking healthier dietary choices and compliance with m any of the recomm ended guidelines has been im proving. We propose to explore the question of whether "returns" to healthy eating have been declining over tim e separately.

. Lim itatio n s o f th e Stu dy
We have noted earlier the limitations of the ECC measure derived from cross-sectional data in reflecting the cumulative consumption patterns of respondents. Nevertheless, a large nationally representative sample such as the NDNS does provide a reasonably robust approximation of the consumption patterns of the population as a whole for the purpose of assessing the shifts in the ECC distribution associated with conform ity to dietary guidelines.
This study also relies on reasonably accurate estim ates of energy intake and energy expenditure to assess the magnitude of excess calorie consumption for individual respondents. Both self-reported energy intake and expenditure are known to be subject to systematic mis-reporting in large dietary surveys. Energy intake tends to be under-reported while energy expenditure tends to be over-reported, with the extent of m is-reporting varying across the BMI distribution. However, the model used in this study still offers a useful approach for overcoming the poor correlation between BMI and nutrient intakes in crosssectional studies. Dietary surveys such as the NDNS are now incorporating more reliable m ethods for estim ating energy expenditure (e.g., doubly labelled water m ethods, m easuring physical activity using triaxial accelerom eters). As the reliability of data on energy intake and expenditure improves in dietary surveys, the model will produce more accurate results. The model remains useful in assessing the m agnitude and direction of changes likely to result from better com pliance with recom m ended dietary guidelines notwithstanding the difficulties in accurately assessing excess calorie consumption.
We have estimated the magnitude of shifts in the ECC distribution of the population likely to be associated with conformity to only a subset of dietary guidelines, that is, those related to the nutrient composition of diets and the intake of individual nutrients. UK dietary guidelines include DRVs for energy intake as well. Dietary changes that achieve conformity to the DRVs for energy intake along with conformity to guidelines for nutrient intakes will be associated with much larger shifts in the ECC distribution than what we have estimated. Combating the risk of obesity in these segments may call for different public health messages m ore centrally focused on energy balance. More generally, in order to have significant im pact on the risk of obesity, public health cam paigns and other initiatives prom oting healthy eating need to adopt a m ore co-ordinated approach focusing on sim ultaneous compliance with a range of dietary guidelines. Measures promoting compliance with nutrient intake guidelines will not translate into truly healthier dietary choices unless they explicitly consider the unintended consequences of com pliance on excessive en ergy intakes.
Yeom ans, M., 20 0 4. Effects of alcohol on food and energy intake in hum an subjects: evidence for passive and active over-consumption of energy. British J ournal of Nutrition 92, 31-34.  -1986- -87 (OPCS, 1991 b Com puted from the National Diet and Nutrition Survey 20 0 0 -0 1 (ONS, 20 0 5) c Computed from the National Diet and Nutrition Survey 20 0 8-0 9 (Bates, Lennox and Swan, 20 10 ) The percentage of total energy available after taking into account that consumed as protein and fat, hence the wide range. c The term "free sugars" refers to all monosaccharides and disaccharides added to foods by the manufacturer, cook or consumer, plus sugars naturally present in honey syrups and fruit juices. Source: Departm ent of Health (1991) and World Health Organization (20 0 3)

Figu re -1B : Co n d itio n al Qu an tile Effe cts o f Life s tyle , P h ys ic al Activity a n d D e m o grap h ic Facto rs by Exc e s s Calo rie Co n s u m p tio n ( ECC) Qu a n tile s
(Source: