In recent years many countries have issued recommendations in relation to nutrition and health. Many of these encourage the maintenance of a healthy weight, but the specific advice to achieve this goal varies considerably. Most countries recommend changes in both diet and physical activity. Dietary guidelines usually include a reduction in fat intake, but the precise recommendations regarding physical activity range from a reduction in sedentary activities or an increase in habitual activities, to vigorous exercise of a precise duration or frequency. These recommendations are largely based on the results of experimental research studies, and there has been relatively little analysis of the role of each of these factors in free-living subjects.
This review will consider ecological, cross-sectional, and prospective observational studies, which include measurements of weight (or body mass index (BMI)), dietary intake, and physical activity. It has not been possible to include those studies that have collected all these data but where it has been published in independent publications. Furthermore, only those studies in which weight, BMI, or weight change is the outcome variable have been included. In spite of these stringent criteria, there is enormous heterogeneity between studies that limits the interpretation of the data. There are also a number of methodological issues, particularly in relation to the methods used to measure the key exposures—diet and physical activity. Carefully controlled experimental investigations that include independent measurements of dietary intake and physical activity have illustrated the tendency to underreport dietary intake and overreport physical activity (14,20). Gross misreporting of dietary data can be identified from basic physiological principles, and many of the studies in this review are suspect (10). At a population level, larger and heavier subjects would be expected to have higher habitual energy needs than their smaller counterparts, but many studies show no such association. More detailed interpretations of the macronutrient composition of the diet are therefore vulnerable to errors caused by biased reporting of dietary intake. There is no reliable method to validate information on physical activity, but it would be naive to imagine this data is entirely accurate. In view of the small short-term energy imbalance, which in the long-term leads to obesity, these methodological limitations thwart any robust conclusions from the following studies.
It is particularly difficult to precisely identify the contribution of a sedentary lifestyle per se to the etiology of obesity because few studies have quantified any sedentary activities and there has been no attempt to describe any form of global “sedentary lifestyle index.” Instead, sedentariness is often inferred from the absence of active pursuits. However, this indirect approach may limit the validity of the data. Indeed, it is possible that the quantification of specific sedentary activities, e.g., TV viewing, may prove to be more robust than the measurement of physical activity. The potential for important interactions between inactivity and diet has also largely been ignored, yet it is plausible that sedentary lifestyles are associated with specific dietary habits, with regard to the macronutrient composition of the diet or eating frequency.
Understanding the differences in the prevalence of obesity between countries, particularly in the developing versus developed world, is beset by methodological difficulties. No single study has attempted such an analysis, although the ecological data relating the prevalence of obesity to dietary intake; specifically, fat (2,15), sugar (12), and physical activity (7) have recently been independently reviewed.
Some studies have reported secular decreases in energy intake concurrent with increases in weight and or fatness, in both children and adults, and it has been inferred that this corresponds to a decrease in physical activity (e.g., 6,26,30,31). Only one study has attempted to simultaneously compare data on energy intake and physical activity with trends in obesity (22). In the United Kingdom, there has been an increase in the prevalence of obesity from 6 and 8% in men and women in 1980 to 17 and 20% in 1997. However, the National Food Survey (NFS) shows a decrease in household food intake, corrected for changes in confectionery, soft drinks, and alcohol consumption, of over 20%. It is probable that there have been increases in energy consumed outside the home over this period, but because this represents such a small proportion of total energy intake (11% at present), it is unlikely to account for the entire decrease in energy intake over this period. Moreover, the NFS data is supported by the accumulated results of cross-sectional dietary surveys which together also show a fall in reported energy intake (21). This implies that there have been even greater decreases in physical activity. Unfortunately, there are no direct measurements of secular trends in physical activity, but good data exist to show significant increases in the time spent watching TV (a proxy measure of leisure-time sedentary activities) and increased car ownership (a proxy measure for the decline in the personal energy cost of transport).
Secular trends in obesity can also be observed in the comparison of two cross-sectional studies in Finland (8). In a subsample of participants in the Finnmonica survey, detailed measurements of energy intake are available from 3-d food diaries and questionnaire-derived estimates of occupational activity, transport to and from work, leisure activity, and time spent sleeping. Over a 10-yr period the prevalence of overweight (BMI > 27 kg·m−2) increased from 39 to 43% in men but was almost unchanged in women, increasing from 33 to only 34%. Energy intake declined by approximately 10% in both sexes, although when the data were edited to remove those clearly underreporting their habitual energy intake, the decrease was only 4%. There was a decrease in the energy cost of transport to work and other work-related activity. This was almost counterbalanced by an increase in the energy cost of leisure activity in women but less adequately in men. Thus, it could be argued either that the greater increase in leisure-time activity in women, or greater decrease in energy cost of work in men, has led to the disparity in the rates of increase in the prevalence of obesity, rather than differences in energy intake.
The analysis of secular trends in obesity is complicated by a multitude of other concurrent changes in lifestyle. However, an analysis of cross-sectional trends from large nationally representative surveys in the United Kingdom provides further supporting evidence of the importance of physical activity over and above energy intake (Figs. 1 and 2). There is a marked social class gradient in the prevalence of overweight and obesity, especially in women where it ranges from 11.1% in professional women to 23.6% in those who are partly skilled or unskilled (23). Despite differences in the types of food consumed there are remarkably small differences in energy intake between groups (11). However, there are significant differences in reported physical activity, with 18% men and 21% women in social classes IV and V reporting no activity in the preceding 4 wk, compared with only 14% and 13% in social classes I and II (27). A similar analysis across the age range shows that the prevalence of obesity in men and women increases from 5.8% and 8%, respectively, in 16- to 24-yr-olds to 21.5% and 23.2% in 55- to 64-yr-olds (23). Between 16 and 44 yr, there is little or no relationship between the increasing prevalence of obesity and differences in either energy intake or levels of inactivity, but from 44 to 64 yr, there is a strong association between rising levels of inactivity and obesity in both men and women.
These ecological analyses, albeit using data from diverse sources and collected at different points in time, provide indirect evidence that sedentary lifestyles play an important and possibly dominant role in the etiology of overweight and obesity. However, these association studies, using data predominately collected in the United Kingdom, have not yet been replicated elsewhere. These data fall into Evidence Category C, being from observational studies only.
There are a number of relevant cross-sectional studies that include data on weight, BMI, or fatness and measurements of diet and physical activity. Studies that have considered the difference in diet and physical activity between groups of lean and obese subjects have been excluded from this analysis because of the high probability of post hoc changes in lifestyle as a consequence of obesity. However, this problem cannot be entirely eliminated from studies in which relative weight is used as a continuous variable. Cross-sectional studies identify associations rather than etiological agents.
Only two cross-sectional studies have been identified in children that report measurements of diet, physical activity, and weight (or fatness) (25,26). Each included approximately 700 children, but although there was a trend toward an association between low activity (or increased sedentary activities) and fatness, these relationships were not significant after adjustment for confounding variables. Sunnegardh et al. (26) reported that children, especially girls, whose parents had low levels of education had more body fat than those of better educated parents, whereas Shannon et al. (25) observed that the weak relationship between TV viewing and obesity was only apparent in the less affluent school districts.
Other cross-sectional analyses can be drawn from the baseline data collection for two of the prospective studies described below (9,13). In a small study of United States Caucasian subjects (142 men and 152 women) Klesges et al. (13) measured dietary intake using a food frequency questionnaire and physical activity by using the Baecke scale comprising 16 items covering work, sport, and nonsport leisure activities. They observed a strong positive effect of the proportion of dietary fat on BMI but no relationship with physical activity. For women, parental obesity was the strongest predictor of BMI, although this was not significant for men.
French et al. (9) studied 1913 women and 1639 men as part of the Health Worker Project conducted in 32 companies in Minnesota. Energy intake was measured using an 18-item food frequency questionnaire, based on food groups and physical activity using a 13-item exercise frequency questionnaire that included high intensity activities, moderate intensity activities, group and racquet sports, and occupational activity. For women, high intensity activities and walking were both significantly inversely related to body weight, as was alcohol intake. Consumption of soft drinks was positively associated with weight. For men, only high intensity activities were significantly inversely related to body weight. The consumption of sweets was also inversely correlated with weight, whereas dairy products, alcohol, and meat were positively related to body weight. In addition, previous dieting, participation in a weight loss program, or currently dieting were all associated with higher body weight. Thus, although offering some support to the hypotheses that the absence of aerobic activities and consumption of high-fat foods may contribute to weight gain, this study shows the importance of previous dieting history as a determinant of subsequent weight change. This is consistent with other studies (3,4) and suggests that the mechanism of this weight rebound (increased intake versus decreased expenditure) has not been adequately described by the available measures of diet and physical activity.
Thus, low levels of physical activity are frequently associated with an increased prevalence of obesity and vigorous activity with a decreased prevalence of obesity. However, it is unclear whether this is a causal association or a post hoc effect. These data fall into Evidence Category C, being from observational studies only.
There are eight prospective studies that include data on changes in weight, or fatness, along with measurements of energy intake and physical activity (3,9,13,16–18,24,28). However, drawing a consensus across studies is limited by their heterogeneity (Table 1). Most studies include adults only, although one refers to children (17) and one has studied the transitional phase from adolescence to adulthood, beginning with age 13, with a 15-yr follow-up (28). This study also has the longest follow-up period, whereas all others range from 2 to 7 yr. Most studies include 100–400 subjects (13,16,17,28), two include 3000–4000 subjects (9,18), and two have over 12,000 participants (3,24). The measurement of the key exposure variables, diet and physical activity, varies considerably. There are also important differences in the adjustment for potential confounding variables. Only two studies have specifically measured any sedentary pursuits, in each case TV viewing (3,17).
The statistical analysis also differs. Half the studies have used measurements made at a single time point to predict subsequent weight changes (13,16,17,24). The remaining studies have used both baseline and follow-up measurements in an analysis that effectively considers the change in weight in relation to the change in diet or physical activity (3,9,18,28). Both these approaches have their own limitations. By using only baseline measurements of exposure, there is a risk that changes in diet and/or physical activity will mask the associations with subsequent weight changes. Yet studies in which changes in diet and/or physical activity are used to explain changes in weight may be confounded by post hoc effects and do not allow any conclusions about cause or consequence. Nonetheless, prospective studies provide the most robust data to assess the relationship between physical activity and weight change.
The overall relationships between diet, physical activity, and other factors with respect to the risk of overweight or obesity are shown in Table 2. Most studies show some evidence of a link between low levels of physical activity and the risk of obesity. Twisk et al. (28) found an inverse relationship between physical activity and fatness (but not BMI). French et al. (9) and Rissanen et al. (24) found an inverse relationship between physical activity and weight change across all subjects, although in other studies the association is limited to subgroups of the population. In the study of United States male health professionals, there was evidence of both an inverse association with vigorous activity and a positive association with TV/VCR viewing, in men aged 45–64 yr but not in those ≥ 65 yr (3). In the studies of Klesges et al. (13) and Paeratakul et al. (18), the association was only observed for women and in the Women’s Gothenburg study low levels of leisure-time physical activity were a risk factor for weight gain only in those women also consuming a high-fat diet (16). Maffeis et al. (17) found no association between physical activity and or TV viewing and the change in relative BMI in children.
The associations between dietary variables and the risk of obesity are more sporadic. Rissanen et al. (24) and Klesges et al. (13) both report a positive association between greater energy intake and weight gain in women, whereas Paeratakul et al. (18) and Coakley et al. (3) each found associations between the proportion of fat in the diet and weight gain in men. French et al. (9) found a relationship between some food groups and the risk of obesity, specifically dairy products, sweets, meat and French fries in women and sweets and eggs in men, but did not calculate specific energy or macronutrient associations. Maffeis et al. (17) found no association between diet and the change in relative BMI in children. Twisk et al. (28) observed that higher energy and macronutrient intakes were associated with greater lean body mass, but not fatness or BMI. This presumably relates to the increased dietary needs for growth in this group, who were followed from adolescence through to adulthood.
Many studies observe that smoking cessation is strongly associated with weight gain (3,13,24), although this was not seen in the study of Paeratukul et al. (18). Previous voluntary weight loss is also a strong predictor of subsequent weight gain in studies in which this has been recorded (3,9). This is consistent with the commonly observed phenomenon of weight rebound. These factors do not explain the primary mechanism of weight change, but they may be important within the context of preventative strategies to allow the development of interventions that are more precisely targeted at high risk groups.
Many prospective studies offer some support to that hypothesis that physical activity can attenuate the rate of weight gain. However, it is difficult to precisely quantify the contribution of physical activity within the context of other potentially important factors, especially diet. Again these data fall into Evidence Category C.
THE BALANCE OF EVIDENCE
The evidence reviewed in this paper shows clearly that low levels of activity are associated with overweight and obesity. This is reinforced by data from other studies (reviewed by DiPietro (5)), which have measured physical activity, but not diet, in relation to weight change. There is good reason to believe that this may be a causal relationship, but definitive statements are limited by the methodological flaws of these studies, particularly with respect to the quantification of physical activity and dietary factors. However, a recent analysis of the errors generated by imprecise measurements of physical activity in the study of weight change have shown that there is a substantial underestimate of the relative importance of physical activity levels (32). Thus, the evidence presented here is likely to represent a conservative assessment of the importance of inactivity as a contributor to weight gain.
This analysis of the evidence relating to the role of sedentary lifestyles in the etiology of overweight and obesity has highlighted the inadequacies of data currently available and the limitations of existing studies. This leads onto a number of key research issues.
The studies to date differ in the demographics of the population sample. Of the prospective studies, only that of Paeratakul et al. (18) is truly randomly selected, and it is rare that the sample accurately reflects the composition of the wider population with respect to gender, age, and ethnicity. The recruitment procedures for studies need to be carefully documented and the characteristics of subjects who drop out identified. There may also be a need for studies that specifically address certain groups of the population who may be at a critical period for the development of obesity. By selecting homogeneous samples at high risk of weight change, e.g., ex-smokers, postpartum women, successful weight-losers, it may be possible to identify important etiological factors using a smaller population base.
The diversity of measurements of exposure and outcome variables in different studies limits their comparability and precludes any overall analysis that may have greater statistical power than individual studies. Much greater use could be made of studies working to common measurement and design protocols.
Most studies use weight and height, or the BMI, as a measure of obesity. However, this fails to distinguish between individuals in whom the excess weight is fat and those who are particularly muscular. Studies in which fatness has been measured have sometimes revealed different associations between BMI or fatness and the exposures under study, especially in children, where the relationship between weight and fatness is more variable (28). Most of the reference methods to measure body composition are unsuitable for large epidemiological studies, whereas prediction methods tend to be less accurate. Other measurements such as waist circumference (fatness) or mid-arm muscle mass (leanness) need to be evaluated to potentially discriminate between individuals of similar BMI but very different body composition.
It is necessary to more precisely define the exposure variables of interest. In dietary studies, the emphasis has been upon total energy, or more recently, macronutrient intake. Issues such as the energy density of the diet have not been included in any of the studies in this review, but this may be an important dietary variable (19). Aspects of eating behavior may also be relevant, e.g., eating frequency, eating alone or in company, and eating out of the home. With regard to physical activity, it is still unclear which dimensions of physical activity are the most relevant to the prevention of weight gain, e.g., total energy expenditure, bouts of vigorous activity, frequency of activity, or the proportion of time spent in sedentary activities. Understanding the mechanism by which dietary factors and physical activity, or their interaction, may influence the development of obesity is integral to the future design of methods to test the relationship between physical activity and obesity.
The quantification of diet and physical activity is beset by methodological errors that tend to be biased in favor of underreporting of intake and overreporting of physical activity (14). Dietary records of energy intake must be evaluated against physiologically derived estimates of energy requirements (10), but in considering their contribution to body weight or weight change, there is the added difficulty that the reported errors appear to increase with the level of obesity (1). However, this procedure only considers total energy and biomarkers for other specific components of dietary intake are also required. Estimates of physical activity by many of the questionnaires used in earlier studies have been shown to relate more closely to physical fitness than habitual activity (29). More sensitive questionnaires are required, validated against reference methods to measure physical activity. Alternatively, there is scope for the development of direct measurements of physical activity which remove the element of subjective reporting. Technological developments in heart-rate monitoring and/or accelerometers may shortly offer a viable solution to large-scale measurements of physical activity and combined with appropriate software may yield data on both energy expenditure and patterns of physical activity. In this respect, equal attention should be paid to the quantification of sedentary pursuits, which are not necessarily the reciprocal of measured activities and perhaps require independent measurement procedures.
Analysis of confounders.
There are many factors that may confound simplistic associations between diet or physical activity and obesity. This is a particular problem in prospective studies, where the relatively precise documentation of factors such as previous weight loss or smoking status may appear to be more strongly associated with weight change than the imprecise estimates of the direct modulators (energy intake and energy expenditure/physical activity). However, the proportion of the variance in weight change that can be accounted for is modest, even in the most comprehensive analyses, suggesting that these studies are either failing to account for important variables or the level of measurement imprecision is too great to reliably assess the true effect of these factors.
Obesity is a chronic condition that develops over many years. During its incubation, there may be many different etiological agents acting sequentially or in concert. Long-term prospective studies that make repeated measurements of both exposure and outcome variables will be particularly powerful. However, careful attention needs to be paid to the design and statistical analysis of such studies to generate the most meaningful data. Additional problems arise in studies that bridge childhood and adulthood because the nature of the outcome variable may need to change. Growth is associated with physiologically orchestrated changes in both fatness and leanness, which lead to variable relationships between BMI and fatness, whereas in adulthood the majority of weight change is due to changes in body fat and thus changes in BMI are a reasonable proxy for changes in fatness at a population level.
Given the measurement imprecision in the assessment of both exposure and outcome variables, the strength of the observed relationship will tend to be diminished. Appropriate statistical procedures to account for this attenuation need to be developed. This will help to refine the association between physical activity and obesity and also to predict the likely outcome of specific intervention targets.
1. Black A. E., A. M. Prentice, G. R. Goldberg, et al. Measurements of total energy expenditure provide insights into the validity of dietary measurements of energy intake. J. Am. Diet. Assoc. 93: 572–579, 1993.
2. Bray G. A., and B. A. Popkin. Dietary fat intake does affect obesity
. Am. J. Clin. Nutr. 68: 1157–1173, 1998.
3. Coakley E. H., E. B. Rimm, G. Colditz, I. Kawachi, and W. Willett. Predictors of weight change in men: Results from The Health Professionals Follow-Up Study. Int. J. Obes. 22: 89–96, 1998.
4. Colditz G, W. Willett, M. Stampfer, et al. Patterns of weight change and their relation to diet in a cohort of healthy women. Am. J. Clin. Nutr. 51: 1100–1105, 1990.
5. DiPietro, L. Physical activity in the prevention of obesity
: current evidence and research issues. Med. Sci. Sports Exerc. 31:(Suppl.) S542–S546, 1999.
6. Durnin J. V. G. A., M. E. Lonergan, J. Good, and A. Ewan. A cross-sectional nutritional and anthropometric study, with an interval of 7 years, on 611 young adolescent schoolchildren. Br. J. Nutr. 32: 169–179, 1974.
7. Ferro-Luzzi, A. Obesity
and physical activity. In: The Origins and Consequences of Obesity
, D. Chadwick and G. Cardew (Eds.). Chichester: John Wiley & Sons, 1996, pp. 207–227.
8. Fogelholm M, S. Mannisto, E. Vartiainen, and P. Pietinen. Determinants of energy balance and overweight in Finland 1982 and 1992. Int. J. Obes. 20: 1097–1104, 1996.
9. French S. A., R. W. Jeffery, J. L. Forster, P. G. McGovern, S. H. Kelder, and J. E. Baxter. Predictors of weight change over two years among a population of working adults: the Healthy Worker Project. Int. J. Obes. 18: 145–154, 1994.
10. Goldberg G. R., A. E. Black, S. A. Jebb, et al. Critical evaluation of energy intake data using fundamental principles of energy physiology. Eur. J. Clin. Nutr. 45: 569–581, 1991.
11. Gregory, J., K. Foster, H. Tyler, and M. Wiseman. The Dietary and Nutritional Survey of British Adults: HMSO, 1990.
12. Hill J. O., and A. M. Prentice. Sugar and body weight regulation. Am. J. Clin. Nutr. 62:(Suppl.) 264S–274S, 1995.
13. Klesges R, L. Klesges, C. Haddock, and L. Eck. A longitudinal analysis of the impact of dietary intake and physical activity on weight change in adults. Am. J. Clin. Nutr. 55: 818–822, 1992.
14. Lichtman S. W., K. Pisarska, E. Berman, et al. Discrepancy between self-reported and actual caloric intake and exercise in obese subjects. N. Engl. J. Med. 327: 1893–1898, 1993.
15. Lissner L., and B. L. Heitmann. Dietary fat and obesity
: evidence from epidemiology. Eur. J. Clin. Nutr. 49: 79–90, 1995.
16. Lissner L., B. L. Heitmann, and C. Bengtsson. Low-fat diets may prevent weight gain in sedentary women: prospective observations from the Population Study of Women in Gothenburg, Sweden. Obes. Res. 5: 43–48, 1997.
17. Maffeis C., G. Talamini, L. Tato. Influence of diet, physical activity and parents’ obesity
on children’s adiposity: a four-year longitudinal study. Int. J. Obes. 22: 758–764, 1998.
18. Paeratukul S., B. M. Popkin, G. Keyou, L. S. Adair, and J. Stevens. Changes in diet and physical activity affect the body mass index of Chinese adults. Int. J. Obes. 22: 424–431, 1998.
19. Poppitt S. D., and A. M. Prentice. Energy density and its role in the control of food intake: evidence from metabolic and community studies. Appetite 26: 153–174, 1996.
20. Prentice A. M., A. E. Black, W. A. Coward, et al. High levels of energy expenditure in obese women. Br. Med. J. 292: 983–987, 1986.
21. Prentice A. M., and S. A. Jebb. Obesity
in Britain. Br. Med. J. 311: 1568–1569, 1995.
22. Prentice A. M., and S. A. Jebb. Obesity
in Britain: gluttony or sloth? Br. Med. J. 311: 437–439, 1995.
23. Prescott-Clarke P., and P. Primatesta. Health Survey for England 1995: Findings
(Vol. 1). London: HMSO, 1997.
24. Rissanen A. M., M. Heliovaara, P. Knekt, A. Reunanen, and A. Aroma. Determinants of weight gain and overweight in adult Finns. Eur. J. Clin. Nutr. 45: 419–430, 1991.
25. Shannon B., J. Peacock, and M. J. Brown. Body fatness, television viewing and caloric-intake of a sample of Pennsylvania Sixth Grade Children. J. Nutr. Educ. 23: 262–268, 1995.
26. Sunnegardh J., L-E. Bratteby, U. Hagman, G. Samuelson, and S. Sjolin. Physical activity in relation to energy intake and body fat in 8- and 13-year old children in Sweden. Acta Paediatr. Scand. 75: 955–963, 1986.
27. Allied Dunbar Health and Fitness Survey. A report on activity patterns and fitness levels. London: Sports Council and Health Education Authority, 1992.
28. Twisk J. W. R., H. C. G. Kemper, M. van Mechelen, G. B. Post, and F. J. van Lenthe. Body fatness: longitudinal relationship of body mass index and the sum of skinfolds with other risk factors for coronary heart disease. Int. J. Obes. 22: 915–922, 1998.
29. Wareham N. J., S. H. J. Hennings, N. E. Day, and A. M. Prentice. Feasibility of heart-rate monitoring to estimate total level and pattern of energy expenditure in a population-based epidemiological study: the Ely young cohort feasibility study. Br. J. Nutr. 78: 889–900, 1997.
30. Weinsier R. L., G. R. Hunter, A. F. Heini, M. I. Goran, and S. M. Sell. The etiology of obesity
: relative contribution of metabolic factors, diet and physical activity. Am. J. Med. 105: 145–150, 1998.
31. Whitehead R. G., A. A. Paul, and T. J. Cole. Trends in food energy intakes throughout childhood from one to 18 years. Hum. Nutr. Appl. Nutr. 36: 57–62, 1982.
32. Wong M. Y., N. E. Day, and N. J. Wareham. Measurement error in epidemiology: the design of validation studies II: the multivariate situation. Stat. Med.