Physical activity and obesity prevention


Professor N Wareham, MRC Epidemiology Unit, Elsie Widdowson Laboratories, 120 Fulbourn Road, Cambridge CB1 9NL, UK. E-mail:


Increasing physical activity is a key element in the treatment of individuals who are obese and in the prevention of weight regain in those who have successfully lost weight. This review, however, focuses on the role of physical activity in the primary prevention of weight gain and obesity and outlines key uncertainties. It is based on a recent systematic review of physical activity and obesity prevention, which was reported at the Nutrition Society and subsequently published in the Proceedings of that group (1).

Ecological data

Data suggest that the rising prevalence of obesity has occurred simultaneously with changes in physical activity patterns. This evidence is ecological, and as such does not provide as high a level of causal inference as individual-level data. The evidence is also rather limited. One of the main reasons for this is that physical activity, in contrast to obesity, is not simple to assess as it is a complex, multidimensional behaviour (2). Physical activity takes place in a variety of domains – in transportation, domestic life, occupation and recreation. Each domain probably needs to be assessed separately, not only because this allows the information to be more specific, but also because it is more likely to be valid. In addition to the different domains, physical activity assessment needs to consider intensity, frequency, duration and the type of activity undertaken. Many historical physical activity instruments, however, are rather simple and often reduce this complex behaviour to a global self-report index (3). Even questionnaires that ask about specific activities sometimes group them across domains, creating questions that are difficult to answer. Few people can accurately report how far they walk in total per day. It is more likely that self-reporting will be accurate when questions address activities in the domain in which they take place. Thus, questions might separately address walking to and from work or school, walking at work and walking for pleasure. Even when questionnaires are logically constructed, with attention to the different domains of activity, they are still relatively imprecise as a measure of total energy expenditure (4).

It is likely that the inherent limitations of self-report measures of activity for population surveillance of energy expenditure are so great that alternative strategies, including use of objective monitoring, are required. This is not to suggest that self-report data are not valuable as, as a measure of reported behaviour, it provides information about the context of activity that is not available from objective measures. The issue with the subjective data relates to the limited validity of measures of total energy expenditure calculated from them. In the UK, population-level surveillance data on overall physical activity are extremely sparse. The Health Survey for England (HSE) included questions originally developed for the Allied Dunbar National Fitness Survey, which have been reconfigured to allow comparison to contemporary definitions of desirable activity levels. As yet, insufficient time has elapsed for temporal trends to become apparent, but, if anything, the HSE data currently suggest that between 1997 and 2003 an increasing number of individuals achieved a physical activity target of a minimum of 30 min or more of moderate intensity activity on a minimum of 5 days a week (5), which is perhaps a paradox given the rising prevalence of obesity. There is no published validation study demonstrating the accuracy of these questions in assessing true activity levels. There is also an added problem that the questions have been altered over time, further limiting the ability to track temporal changes in activity. Thus, our knowledge in the UK concerning population levels of physical activity mostly stems from proxy domain-specific measures. For example, since the 1960s, there has been a large increase in second-car ownership (6), an increase in the use of labour-saving devices in the house (e.g. dishwashers, tumble driers) (7), an increase in the hours of television viewing per week (7), a decline in the distance children walk per year (6), and a massive downward change in the proportion of the workforce employed in manufacturing, farming and other physically demanding occupations. Given that the aetiological effects of activity are likely to be related to the totality of physical activity rather than the domain-specific components, it is a major deficiency that we do not have population-level data on temporal trends in total activity. It could be that the apparent temporal changes in activity in domestic life, work and travel are compensated for by an increase in recreational activity (8), but this is uncertain in the UK. In the USA, there is some suggestion from the Behaviour Risk Factor Surveillance System of a secular decrease in the proportion of people reporting total inactivity during recreational time. Whether this can be taken to imply an increase in sufficient physical activity to prevent weight gain is uncertain.

In summary, the data that exist seem to indicate a secular decline in overall physical activity that occurred at the same time or possibly before the temporal increase in obesity. To be able to have a clearer knowledge of temporal trends in physical activity, it is essential that the UK and other countries establish a consistent, securely funded collection of long-term population surveillance data both on subcomponents of physical activity and total physical activity-related energy expenditure. As the latter is difficult to assess by questionnaire, it would be preferable to use objective methods. This proposition requires a long-term commitment to surveillance, which in turn implies investment in the development and use of inexpensive surveillance methods. Population-level surveillance data not only provide important descriptive epidemiological information about temporal trends, but also act as an outcome measure for interventions aimed at changing population activity. The problem in the UK is that no single institution has overall responsibility for the collection of this type of data. This is a soluble issue.

Observational studies of physical activity and change in weight

A higher level of causal inference comes from studies where activity and weight change are measured in individuals rather than populations. A systematic review in [2000] describing data from observational cohort studies on physical activity and weight gain in adults concluded that there was inconsistent evidence of the predictive effect of baseline physical activity on subsequent weight gain (9). However, they observed that the association between weight gain and change in activity or activity at follow-up was stronger, although still modest. These results may be interpreted in three different ways:

  • • physical activity is an important factor in preventing weight gain but the true association is not detectable because of measurement error;
  • • less weight gain leads to better exercise adherence – a reverse causality argument;
  • • the self-reported physical activity may be a proxy for a general healthy lifestyle – a confounding argument.

In our updated review, we identified an additional 14 observational studies on physical activity and weight gain in adults, of which only two included an objective measure to assess physical activity (10,11), whereas the other 12 studies assessed physical activity by means of self-report (12–23). Comparing these results to the previous review, the more recent studies more often reported associations in the expected direction. Overall, the magnitude of the effect was small in the studies reporting the expected inverse association between physical activity and subsequent weight gain, even in those studies using objective measurement.

As with the adult studies, a review in 2000 (24) suggested that, of seven studies in children, four reported that physical activity was associated with less weight gain in children, whereas the other studies did not observe an association. We identified a further 16 articles and, again, the results were mixed (25–40). Five of the studies using a self-report measure of activity did not find an association of physical activity or sedentary behaviour with weight gain (25–29). Six studies showed an inverse association between higher levels of self-reported physical activity and weight gain, or a positive association with time spent on sedentary activities (30–35). However, as in the studies in adults, the measures of association tended to be small (24). Five studies in children have reported the longitudinal association between objectively measured physical activity and weight gain (36–40). In three of these studies, the results were inconclusive, and in two, the association between activity and weight gain was weak but in the expected direction.

Trials to prevent weight gain

The observational studies of activity and weight gain in adults and children are affected by issues of measurement error, residual and unmeasured confounding, and reverse causality. Although improved study design, with greater emphasis on objective measures and with a known degree of measurement error, can deal with a number of these issues, it cannot resolve problems of the direction of causality nor can it deal with unmeasured confounding, for which randomized clinical trials are required. Although there are a large number of clinical trials on the treatment of individuals with obesity or the prevention of weight regain among weight losers, a recent systematic review of trials to prevent weight gain de novo only revealed a total of nine trials (41). Not only are these trials relatively few in number but also, for various methodological reasons, they are uncertain in their conclusions about whether increasing activity will be effective in preventing obesity. Our updated review (1) found a further six trials aimed at increasing physical activity and preventing weight gain in adults (42–47) and these interventions were mostly aimed at populations that were either defined on the basis of their risk for weight gain or because they represented a specific group on whom an intervention might be targeted, e.g. couples in their first 2 years of living together, the working population, pregnant or middle-aged premenopausal women, or patients taking drugs that induce weight gain as a side effect. In general, the interventions were of a high intensity and were spread over a relatively long period of time, with face-to-face counselling on behaviour change in either group or individual settings. Two studies had follow-up measurements more than 3 months after the intervention had ended. In general, the description of the underlying theories supporting the interventions was limited.

Our review also showed that the picture was similar for trials of physical activity and obesity prevention in children. We identified a further 11 trials aimed at preventing unhealthy weight gain by increasing physical activity or reducing sedentary behaviour in children (48–58). Nine trials studied the effectiveness of school-based interventions, whereas the others studied home-based or family-orientated interventions (50,54). Effectiveness was mostly assessed directly after the intervention and only one study included a follow-up measurement at more than 3 months after the end of the intervention (57). Three of the 11 trials were able to detect a small intervention effect on body composition at follow-up (49,55,58), with two of them reporting significant effects for boys only (55,58). Although several of the non-effective trials reported on positive changes in physical activity levels or on dietary behaviour, they did not show statistically significant differences in body weight or body composition at follow-up between the intervention and control group. When studying the intensities and settings of the interventions, it seems that comprehensive school-based interventions aiming at increasing physical activity levels through physical education classes and behaviour change are most likely to be effective in preventing weight gain in children, whereas interventions aimed at reducing sedentary behaviour and family based interventions seem to be less effective.

This updated review shows that there are still relatively few trials aimed at the primary prevention of weight gain and that there is thus still insufficient evidence on which to base conclusions about which approaches work. It is evident, however, that understanding from such trials would be enhanced if they were based on an explicit causal model with a clear theoretical foundation, so that even if the overall effect was non-significant, it would be possible to disentangle which bits of the intervention were ineffective. At present, most of the interventions are black boxes, and when they do not work, it is difficult to generalize explanations for their ineffectiveness. In the absence of clearly successful interventions, the evidence from randomized trials does not currently contribute solutions to the issues left unresolved by observational cohort data.

How much activity is enough to prevent weight gain?

Given the uncertainty from the observational studies and the scarcity of randomized controlled trials, the short answer to the question, ‘how much activity is enough to prevent weight gain?’ is that we do not know. A report from a consensus conference published in 2003 attempted to provide recommendations, concluding from two prospective, non-randomized, cohort studies among women who had successfully lost weight, that there was ‘compelling evidence that prevention of weight regain in formerly obese individuals requires 60–90 minutes of moderate-intensity activity or lesser amounts of vigorous activity’ (59). Such observational evidence might be considered by some to be less than complete, but when they came to discussing the amount of activity required for the primary prevention of weight gain, the authors acknowledge that ‘definitive data are lacking’. However, on the basis of a selective review, they concluded that ‘moderate-intensity activity of approximately 45–60 minutes per day . . . is required to prevent the transition to overweight or obesity’. Although one could argue about the details, the most salient points on which most would agree are that we still do not have definitive data and that, on the basis of the studies that are currently available, the amount of activity necessary is likely to be substantial.

While further studies are required to address the deficiencies in the evidence base, public health authorities have to make responsible recommendations using whatever evidence is available now. In England, the Chief Medical Officer’s (CMO) recent review on the evidence of the impact of physical activity and its relationship to health very much followed the lines of the Saris consensus report with respect to weight gain (60). While one might question the scientific basis on which the conclusions of the review were based, the overall conclusion is still sensible. The danger of overplaying the strength of evidence underlying these conclusions is that it may hinder efforts to improve the evidence base and may undermine an approach to prevention if interventions are unsuccessful. The CMO’s overall recommendation was for individuals to accumulate at least five episodes per week of moderate activity lasting 30 min. Although the impact of such a recommendation on the likelihood of weight gain cannot be quantified, it is rational, as any increased activity overall is likely to reduce obesity risk. A low proportion of the population currently meet this recommendation, and the recommended levels are not so distant from people’s everyday experience to be unachievable. Whether the recommended activity levels are adopted remains to be seen, as does their impact on the prevalence of obesity if they are adopted.

Analysis of true population-level approaches to increasing physical activity

All of the trials we found were aimed at changing individual behaviour. However, if one takes a broader perspective, it is clear that this approach ignores important collective determinants of physical activity. These determinants, including environmental influences such as transport policy, are much less amenable to the traditional medical reductionist approach to evaluation and it is unlikely that they will ever be subject to assessment by a randomized controlled trial. However, if they are powerful influences on physical activity and therefore strong drivers of the current obesity epidemic, it is important that opportunities are sought to assess the impact of the environmental changes that are brought about by deliberate policy intention, such as the provision of cycle pathways, or those that are the result of policies in relation to an entirely different issue, for example, congestion charging schemes. In either situation, it is also important to assess not only specific behaviours that are likely to be directly affected by the policy change, but also other physical activity behaviours that might be altered as a consequence and, ideally, to assess the totality of activity as well. A congestion charging scheme might impact on car use with a resultant increase in cycling and walking to work. However, its overall impact on total activity might be neutral if the increase in energy expenditure during transportation to work is compensated for by opposite trends in recreational activity. It is unlikely that researchers will be able to accumulate sufficient pre-change data to be able determine whether the policy change has had an effect, as most policy initiatives work to a short timetable. Therefore, a mixed approach to evaluation is likely to be the most successful, with the incorporation of rapid local measures of specific and total activity into areas likely to be affected by policy changes, with longer-term trend data being obtained from ongoing population surveillance studies. As the introduction to this review indicated, such background data are scarce in the UK and are barely sufficient to be able to describe the current trends in physical activity behaviour, let alone allow for analysis of the impact of societal level interventions. Efforts to address these deficiencies are clearly long overdue and require as comprehensive a system as that in place for infectious diseases. As inactivity and overweight are the major public health challenges of the 21st century, it is timely to consider whether our public health surveillance systems are up to the task of tracking our progress in meeting these new challenges.

Conflict of Interest Statement

No conflict of interest was declared.