Increased food energy supply as a major driver of the obesity epidemic: a global analysis

1Overweight and obesity have become major global public health problems. Worldwide, the proportion of adults with a body mass index (BMI) of 25 kg/m2 or greater increased from 28.8% to 36.9% in men, and from 29.8% to 38.0% in women between 1980 and 2013. Urgent action from governments and the food industry is needed to curb the epidemic. Action needs to be directed at the main drivers of the epidemic to meet the global target of halting the rise in obesity by 2025.

2The drivers of the obesity epidemic have been much debated.– An increased food energy supply and the globalization of the food supply, increasing the availability of obesogenic ultra-processed foods, are arguments for a predominant food system driver of population weight gain. Increasing motorization and mechanization, time spent in front of small screens and a decrease in transport and occupational physical activity, point to reducing physical activity as a predominant driver, of the obesity epidemic.

3A model used to predict body-weight gain, assuming no change in physical activity, follows the simple rule that a sustained increase in energy intake of 100 kJ per day leads to a predicted increase of 1 kg body weight on average, with half of the weight gain being achieved in about one year and 95% in about three years. According to this model, the oversupply of food energy is sufficient to drive the increase in energy intake and increases in body weight observed in the United Kingdom of Great Britain and Northern Ireland and the United States of America.– This is despite the fact that, in the United States, food waste has increased by approximately 50% since 1974, reaching about 5800 kJ per person per day in 2003. Here we test the hypothesis that an increase in food energy supply is sufficient to explain increasing population body weight, using data from 24 high-income, 27 middle-income and 18 low-income countries.

4Food balance sheets of the Food and Agriculture Organization of the United Nations (FAO) estimate the food supply of countries, by balancing local production, country-wide stocks and imports with exports, agricultural use for livestock, seed and some components of waste. Waste on the farm, during distribution and processing, as well as technical losses due to transformation of primary commodities into processed products are usually taken into account. However, losses of edible food, e.g. during storage, preparation and cooking, as plate-waste or domestic animal feed, or thrown away, are not considered. The data are expressed as the annual per capita supply of each food item available for human consumption. The FAO’s database contains national level data from 1961 to 2010 for 183 countries. For each country, data on food energy supply were extracted to match the time periods of data on adult body weight.

5Three major strategies were used to collect data on measured average adult body weight. First, an electronic search of major databases on obesity prevalence and BMI was performed, including the World Health Organization’s (WHO) global infobase, WHO’s global database on BMI, the International Association for the Study of Obesity (now World Obesity Federation) database and the Organisation for Economic Co-operation and Development’s health data. As these databases only included data on obesity rates or mean BMI, the original sources of the data were searched. Second, data on average measured body weight were gathered from reports of national health and nutrition surveys in various countries. The WHO MONICA project and WHO STEPwise approach to surveillance (STEPS) country reports included anthropometric measures for male and female adult samples. We also calculated body weight for women of child-bearing age using mean BMI and height data from Demographic and Health Surveys. Third, an electronic search of Medline was conducted. For each country, a separate search was performed using the following keywords: “obesity”, “weight”, “anthropometric”, “BMI”, “health survey” and “national survey” (using the Boolean operator OR). Finally, specific national health and/or nutrition surveys identified by some of the above sources were electronically searched.

6Studies fulfilling the following criteria were extracted: (i) weight was measured after 1961 and again before 2010 (to match the FAO food balance sheet data); (ii) the study samples were representative of a national or subnational adolescent or adult population; (iii) the survey method was comparable with previous or future surveys conducted in the country; (iv) the year in which each survey was conducted could be identified; at least four years elapsed between the two surveys; and (v) FAO food supply data were available for the relevant period.

7If there were more than two eligible studies from a country, the surveys which we judged to be the best quality were included. Criteria for estimating study quality included national representativeness, sample size and length of time between surveys.

8Demographic data (total population, by age and sex) were retrieved from the United Nations Department of Economic and Social Affairs. Average female and male height at survey time points were derived from http://www.averageheight.co/. For 13 countries, data were not available and average height data from a neighbouring country were used for calculating energy requirements.

9Three types of analysis were performed. First, we compared the changes in food energy supply with changes in average body weight over time for each country. Second, estimates of population energy requirements at survey time points were performed for each country using Institute of Medicine equations. Low active physical activity levels (1.4 ≤ PAL <1.6) were assumed for high- and upper-middle-income countries. Active physical activity levels (1.6 ≤ PAL <1.9) were used for all other countries. Finally, we used a physiologically-based, experimentally-validated predictive energy intake body-weight model, to estimate the change in average population energy intake that would be required to account for the observed change in average body weight.

10In total, 83 countries had at least two surveys with data on measured body weight; 24 countries had more than two surveys at different time points. We excluded countries where the period between surveys was less than four years (eight countries), survey populations were not comparable in terms of area representativeness (eight countries) or FAO food supply data for the country were not available (three countries). Survey pairs from 69 countries were included. Of those, 36 survey pairs included data for women of childbearing age only. One survey pair (Saudi Arabia) included data for men only. Data from 24 high-income, 27 middle-income and 18 low-income countries were included. The average period between the surveys was 12 years (range 4–37 years; Table 1). At the time of the initial survey, food energy supply was greater than the average energy requirements in 52 countries. For 37 of these countries, this excess food energy supply was more than 2000 kJ/day (Table 1).

11For 56 countries (81%) both food energy supply and body weight increased between the survey pairs. For 45 of these countries (80%) the increase in food energy supply was more than sufficient to explain the increase in average body weight. This is shown in Fig. 1 with 56/69 countries being in the top right quadrant and 45/56 being to the right of the model-predicted change in energy intake needed to produce the increase in mean body weight for that country. This same pattern was observed for countries of all income levels (Fig. 2, Fig. 3, Fig. 4 and Fig. 5). For 11 countries (Benin, Chile, the Dominican Republic, Gabon, India, Indonesia, Ireland, Italy, Lebanon, Mauritania and New Zealand) in the top right quadrant, the increase in food energy supply was insufficient to account for the observed increase in weight (Fig. 1).

12Five countries (Barbados, Burkina Faso, Kazakhstan, Nigeria and Switzerland) experienced reductions in both food energy supply and average body weight. For Kazakhstan the food energy supply decreased by 3778 kJ/day, from 13 117 kJ/day to 9339 kJ/day over a four year period (Table 1), accompanied by a decrease in average body weight of 0.9 kg. For the four other countries, decreases in food energy supply were much more modest (100–300 kJ/day; Table 1).

13For five other countries (Eritrea, Iceland, Malaysia, Turkey and Uzbekistan), discordant changes were observed with reductions in food energy supply over the same period as increases in average body weight. The decrease in food energy supply was highest for Uzbekistan (2615 kJ/day) and lowest for Eritrea (63 kJ/day; Table 1). Apart from Eritrea, food energy supply at baseline for those five countries was relatively high (ranging from 12 242 to 15 531 kJ/day) and higher than the values of at least half of the other countries included in this study. In addition, excess food energy supply at baseline was high for those five countries (2757–7251 kJ/day; Table 1).

14For three countries (the Islamic Republic of Iran, Rwanda and South Africa) there were discordant changes in the other direction with increases in food energy supply over the same period as reductions in average body weight. However, for two of those countries, the change in average weight was small (a reduction of 5 g for the Islamic Republic of Iran and 100 g for South Africa). In Rwanda, the reduction in weight was 800 g while the food energy supply over the same time period increased by 674 kJ/day (Table 1).

15The correlation between the change in food energy supply and change in average body weight was significant (P = 0.011). When stratifying by type of country, associations were significant for high-income countries (P < 0.001), but not for other country groups.

16For most of the countries included in this study, the change in per capita food energy supply was greater than the change in food energy intake theoretically required to explain the observed change in average body weight. The associations between changes in food energy supply and average population body weight were significant overall and for high-income countries. This suggests that, in high-income countries, a growing and excessive food supply is contributing to higher energy intake, as well as to increasing food waste.

17Other factors, such as a decrease in physical activity, may also lead to an increase in body weight and could occur simultaneously with an increase in food energy supply. It has been shown that among 3.7 million participants in the United States at the county level, increased physical activity has only a very small impact on obesity prevalence. It is likely that in some countries, such as China, the impact of reduced physical activity on obesity is more important., A reduction in physical activity with no compensatory drop in energy intake will cause weight gain until sufficient weight is gained to create energy balance (through both an increased resting metabolic rate and increased energy required to move the larger body).

18Researchers have suggested additional contributing factors for obesity, such as pollutants, infections and changes in the gut microbiota. These factors have an effect on metabolism, body composition and/or energy balance efficiencies. However, more evidence is needed to understand the importance of these factors in weight gain. Ideally, the cause of obesity in humans would be assessed through randomized controlled trials, where food energy availability is increased randomly and average body weight is then measured. However, such an experiment is not practical, since it is difficult to measure food intake over long time periods and it would require that non-obese subjects be randomly assigned to environments with different food energy supplies.

19Our findings suggest that there is an excess of energy available from an increasing national average food energy supply in countries of varying income levels. Therefore, policy efforts need to focus on reducing population energy intake through improving the healthiness of food systems and environments.,, Achieving WHO’s target to halt the rise in obesity by 2025 will require major action by governments and the food industry. A combination of several policy actions will be needed to significantly improve diets and reduce overconsumption. These policies include restriction of unhealthy food marketing to children, front-of-pack supplementary nutrition labelling, food pricing strategies, improving the quality of foods in schools and other public sector settings. The impact of trade and investment agreements and agricultural policies on domestic food environments should be assessed.

20The main strength of this study is the inclusion of nationally representative body weight and food energy supply data for a range of countries and over many years. Weaknesses include the limitations on the measurement of national per capita food energy supply (e.g. losses of edible food during storage, preparation and cooking, as plate-waste or domestic animal feed, and subsistence farming are not taken into account) and the variable quality of energy supply data. In addition, low- and middle-income countries, in different phases of the nutrition transition,, are likely to have poorer data and have higher levels of subsistence farming, which is not included in the FAO food supply data.

21The association between changes in food supply and changes in body weight may be confounded by changes in physical activity levels, changes in food waste or changes in the demographic profile of countries. Demographic changes, particularly size, ageing, and racial/ethnic diversification of populations, may contribute to increasing obesity levels. About half the data sets on weight status used in this study are for women only and thus only represent half of the population. A limitation of the energy-balance model is that it assumes that metabolic physiology and physical activity levels are similar globally. While this is likely to be true for industrialized countries for which accurate data on the relationship between energy expenditure and body weight are available and for which the model has been calibrated, it is not clear how well this assumption applies for developing countries. The model also assumes that population-wide changes in physical activity are negligible over the periods investigated.

22In conclusion, in high-income countries, observed increases in body weight over recent decades are associated with increased food energy supply. In addition, increases in food energy supply are sufficient to explain increases in average population weight. Due to the nutrition transition and a potential decrease in physical activity, the same pattern is expected to occur in low- and middle-income countries in the future. Policy efforts should focus on reducing population energy intake through improving the healthiness of food systems and environments.