In a session moderated by Loretta DiPietro, Ulf Ekelund and Robert Ross discussed evidence on the impact of physical activity on the prevention of obesity in adults. This chapter summarizes their presentations and the discussion that followed.
Ekelund described evidence showing only a very weak association between physical activity and weight gain and other evidence suggesting that physical activity reduces the risk of obesity, but only in people with normal weight status at baseline. In contrast, Ekelund continued, yet other evidence indicates a strong relationship between physical activity and other health outcomes, including all-cause mortality. Increasing physical activity by simply adding 20 minutes of brisk walking per day has been shown to reduce the risk of mortality by 24 percent in people of normal weight and 16 percent in those who are obese. Ekelund called for a greater focus on promoting physical activity for health rather than for weight.
Ross differentiated between efficacy studies on the association between physical activity and weight gain and effectiveness studies on the association between encouragement to increase physical activity and weight gain. Those are two different questions, he said, and they require different study designs. Based on a review of evidence from randomized controlled efficacy trials, Ross concluded that unless individuals eat more, exercising under supervised conditions increases energy expenditure and leads to both weight loss and reduced waist circumference. Echoing keynote speaker James Hill’s sentiment, he said, “I just don’t think there is any ambiguity there. I think
the evidence from rigorously controlled studies is very clear.” Turning to evidence from randomized controlled effectiveness trials, Ross said this evidence indicates that lifestyle interventions designed to prevent weight gain are generally effective, although it is unclear which component of the interventions—diet or exercise—is responsible. He suggested that treatment of obesity (e.g., sustained weight loss) may be too great a challenge and that a more desirable outcome may be prevention of weight gain.
Cross-sectional studies have revealed a strong inverse relationship between higher levels of physical activity and body weight and obesity, Ulf Ekelund began. Based on a systematic review of evidence from observational studies, however, Summerbell and colleagues (2009) concluded that overall, physical activity is not associated with subsequent excess weight gain and obesity. When studies do report associations, the associations are small and negative. To focus on the association between physical activity and body weight and obesity in more detail, Ekelund addressed five questions:
- Do higher levels of physical activity prevent weight gain over time?
- Do higher levels of physical activity prevent the development of obesity over time?
- Does the association between activity and weight gain differ depending on baseline weight status?
- Is change in activity associated with change in body weight?
- Does weight status predict physical inactivity over time?
Does Physical Activity Prevent Weight Gain?
To measure the effect of physical activity on weight gain, epidemiologists measure physical activity at baseline, measure body weight at followup, and adjust for confounding factors, Ekelund explained. Confounding factors are associated with both the exposure (in this case, physical activity) and the outcome (in this case, body weight at follow-up) and include, for example, diet, alcohol consumption, socioeconomic status, age, and sex. Importantly, to study temporal associations, baseline body weight needs to be included in the model, in Ekelund’s opinion, given that it is probably the most important predictor of follow-up body weight. He noted
1 This section summarizes information and opinions presented by Ulf Ekelund, Ph.D., FACSM, University of Cambridge, United Kingdom.
that there has been some debate around whether baseline body weight is a confounding factor, with some experts considering adjustment for it to be overadjustment.
In a prospective cohort study on physical activity and abdominal adiposity and body weight gain among 288,498 men and women, Ekelund and colleagues (2011) followed individuals for 5 years. They assessed physical activity using a validated questionnaire and categorized individuals into four groups: inactive, moderately inactive, moderately active, and active. At follow-up, they measured weight and waist circumference. The researchers adjusted for a number of confounding factors, including baseline body weight or waist circumference, depending on the outcome being considered. They concluded that physical activity at baseline did not predict weight gain at follow-up, but it did predict waist circumference in both men and women. However, the magnitude of the association with waist circumference was minor, Ekelund said, with the difference in gain over 5 years between the inactive and moderately inactive groups being only 0.05 cm. While statistically significant in such a large sample, such a small difference, in Ekelund’s opinion, is not clinically significant.
Does Physical Activity Prevent the Development of Obesity?
To measure the effect of physical activity on obesity at follow-up, Ekelund continued, the same confounding factors need to be controlled. In the same study discussed in the previous section (Ekelund et al., 2011), he and colleagues observed that baseline physical activity was associated with the risk of developing obesity at follow-up, with a one-category increase in physical activity index (e.g., moving from inactive to moderately inactive) reducing the risk of obesity in women by 10 percent and in men by 7 percent.
In another study, based on data from the Women’s Health Study, which followed women for 11.6 years, Britton and colleagues (2012) showed that vigorous-intensity physical activity was associated with about a 20 percent lower risk of becoming overweight or obese, without adjusting for baseline body mass index (BMI). When the researchers adjusted for baseline BMI, the association disappeared.
In yet another study, using Coronary Artery Risk Development in Young Adults (CARDIA) data, with physical activity and body weight being measured repeatedly over time, Hankinson and colleagues (2010) found that individuals who were consistently active over 20 years showed less increase in BMI; this was the case even for individuals with lower BMIs at baseline. Ekelund noted that despite the smaller BMI gains among active individuals, BMI increased in all individuals over the course of the study. Also of note, only 11-12 percent of study participants remained in the
active group over the course of the study (total n = 3,554 [1,689 men and 1,865 women participated]). The researchers concluded that maintaining high activity levels through young adulthood may lessen weight gain as young adults, particularly women, transition into middle age.
Does Baseline Weight Matter?
The third question addressed by Ekelund was whether associations between baseline physical activity and the development of obesity differ depending on baseline weight status. Adjusting for confounders, Lee and colleagues (2010) used Women’s Health Study data (n = 34,079) collected over the 15-year period from 1992 to 2007 to assess the effect of baseline BMI on the association between physical activity and weight gain at 3-year intervals. The women gained an average of 2.6 kg every 3 years. For the analysis, the researchers stratified the women into three groups based on metabolic equivalent of task (MET) hours per week: less than 7.5 MET hours per week, between 7.5 and 21 MET hours per week, and 21 or more MET hours per week (7.5 MET hours corresponds to about 150 minutes of physical activity). The women also were stratified into three BMI groups: normal weight, overweight, and obese. The only observed association between physical activity and weight gain was among the women with normal BMI (i.e., lower than 25), with greater levels of physical activity being associated with less weight gain. The researchers found that women who were successful in maintaining normal weight and gaining less than 2.3 kg over 13 years averaged approximately 60 minutes per day of moderate-to vigorous-intensity physical activity (MVPA).
As part of the same study discussed in the previous two sections, Ekelund and colleagues (2011) conducted a similar analysis. When they stratified their participants into normal weight, overweight, and obese BMI categories, they observed a significant inverse association between physical activity and weight gain in the normal weight category among both men and women. Ekelund concluded, “There may be differential associations between physical activity and weight gain depending on the initial BMI or body weight status.”
Is Change in Physical Activity Associated with Change in Body Weight?
The simplest model used to determine whether a change in physical activity is associated with a change in body weight entails measuring both exposure (physical activity) and outcome (body weight) at two different points in time and calculating the associations between changes in the two variables. While this type of analysis is very robust, Ekelund said, he cautioned that it cannot be used to determine the direction of an association.
Using data from the Nurses Health Study, Mekary and colleagues (2009) found that, compared with women who were active for less than 30 minutes per day, women who sustained 30 or more minutes of activity per day had a 32 percent reduced risk of gaining more than 5 percent of their baseline body weight between 1989 and 1997. Women who increased their level of physical activity over time (i.e., went from being active less than 30 minutes per day to being active 30 or more minutes per day) had an even greater reduced risk of gaining weight (36 percent). Women who decreased their physical activity over time (i.e., moved from being active 30 or more minutes per day to less than 30 minutes per day) had a 12 percent elevated risk of gaining more than 5 percent of their baseline body weight.
Using a combined dataset from three different American cohorts of men and women (n = 120,877), Mozaffarian and colleagues (2011) stratified participants into five equal-size quintiles based on level of physical activity and examined change in weight at 4-year intervals. They found that as physical activity increased, change in body weight decreased (i.e., individuals gained increasingly less weight over time), with the most active quintile being associated with a –1.86 pound reduction in body weight (i.e., individuals gained 1.76 fewer pounds within each 4-year period). In the results section of their paper, the authors state, “Absolute levels of physical activity, rather than changes in these levels, were not associated with weight change (data not shown).” Ekelund interpreted these findings to mean that physical activity at baseline did not predict weight gain, but change in physical activity did.
Reverse Causality: Does Weight Status Predict Physical Inactivity?
Finally, Ekelund explored reverse causality, as Kathleen Janz had done during her presentation (see Chapter 2): Do higher levels of body weight or adiposity at one point in time predict lower levels of physical activity or higher levels of sedentary behavior at a later point in time? As with the other four questions, answering this question requires measuring both exposure (in this case, baseline body weight) and outcome (in this case, activity levels), and then controlling for confounding factors.
Using a small dataset (n = 390), Ekelund and colleagues (2008) objectively measured both physical activity (using individually calibrated heart rate monitoring) and adiposity (using bioimpedance). They found that while higher levels of sedentary time at baseline did not predict gains in adiposity at follow-up, higher levels of adiposity at baseline predicted higher levels of sedentary time at follow-up. Additionally, they found at follow-up that, compared with individuals who lost fat mass over time, those who gained fat mass over time had spent more sedentary time. This same trend has been observed in other studies as well, according to Ekelund. For example,
Golubic and colleagues (2013) found that weight gain over 10 years was a significant determinant of physical inactivity and that great weight gain over just 3 years (more than 2 kg per year) was a significant determinant of physical inactivity. Ekelund interpreted these findings to mean that there may be some sort of reverse causality or bidirectional association between body weight or adiposity and physical inactivity.
In summary, first, the prospective association between physical activity and gain in body weight and BMI is weak, in Ekelund’s opinion. It has been observed in some studies, but not others. Ekelund suspects that some of the variability may be due to error in measurement of the exposure variable, in this case physical activity. Second, maintaining a high level of physical activity appears to reduce the risk of becoming obese over time; however, this association may be limited to those who are of normal weight at baseline. Third, the association between physical activity and obesity is likely bidirectional, suggesting reverse causation. Fourth, the amount and intensity of activity needed to maintain a healthy body weight throughout adulthood are unknown, but likely substantial.
One of the most important changes that could be made for public health, in Ekelund’s opinion, is to increase population levels of physical activity, with small shifts at the population level likely having significant effects on public health outcomes. To demonstrate, using data from the European Prospective Investigation into Cancer and Nutrition (EPIC) Study and stratifying the data into three different BMI groups, Ekelund and colleagues (2015) found that people of normal body weight had a 24 percent reduced risk of all-cause mortality if they were moderately inactive rather than inactive. People who were obese had a 16 percent reduced risk of all-cause mortality if they were moderately inactive rather than inactive. When people with an unhealthy and a healthy waist circumference were compared, mortality risk reductions observed in the moderately inactive versus inactive groups were similarly substantial. The difference between being moderately inactive and inactive, Ekelund explained, was about 20 minutes of brisk walking per day, or 100 kilocalories expended in physical activity.
Ekelund’s take-home message was that physical activity prevents weight gain in a small segment of the population who are of normal weight and who are highly physically active, but the health benefits of physical activity are well established and indisputable. Therefore, he believes a stronger emphasis should be placed on physical activity for health rather than for weight. The challenge, he suggested, is to shift the focus from losing body weight to promoting lifestyle behavior change across the entire population.
Asking whether exercise is an effective strategy for preventing weight gain in adults is really two questions, Robert Ross began. First, what happens with respect to weight gain when adults exercise or increase their physical activity? Second, what happens with respect to weight gain when adults are encouraged to exercise or increase their physical activity? Those are two very different questions that require different types of studies and yield different results, Ross suggested. Unfortunately, in his opinion, the differences often are misunderstood in the literature. Both questions call for randomized controlled trials, but different types of trials. The first question calls for an efficacy trial, which is characterized by strong internal validity and is concerned primarily with physiological response. That is, if people increase their physical activity, what happens physiologically? The second question, in contrast, calls for an effectiveness trial, which is concerned primarily with changing behavior. That is, if people are encouraged to increase their physical activity, does their behavior change? Ross discussed evidence in the literature derived from both types of trials.
Efficacy Studies: What Happens When Adults Exercise or Increase Their Physical Activity?
Jean Mayer’s initial observations many years ago showed a mismatch between body weight and caloric intake among people who were more physically active (Mayer et al., 1956) (see Figure 3-1). “It was a seminal observation,” Ross said. While Mayer’s observations were limited by their cross-sectional nature, they nonetheless suggested that physical activity can prevent weight gain despite increasing energy intake. As James Hill had elaborated during his keynote presentation, Ross reminded the workshop audience how the mismatch between body weight and energy intake initially observed by Mayer and colleagues (1956) led to the notion of a theoretical threshold for optimal weight regulation (see Figure 1-3 in Chapter 1) (Hill et al., 2012). On one side of the threshold is the “regulated” zone, characterized by the mismatch, while on the other side is the “unregulated” zone.
Ross described evidence from his work showing what happens when people in the regulated zone both increase their physical activity and consume more calories relative to baseline. He and his research team found that when participants, both men and women, exercised an additional 50-60 minutes daily for 4 months and consumed an additional 500-700 calories
2 This section summarizes information and opinions presented by Robert Ross, Ph.D., R.Kin., FACSM, FAHA, Queen’s University, Kingston, Ontario, Canada.
FIGURE 3-1 Body weight and caloric intake as a function of physical activity.
SOURCE: Presented by Robert Ross on April 14, 2015 (reprinted with permission from Mayer et al. American Journal of Clinical Nutrition [1956; 4:169-175]).
every day to offset the exercise-induced energy expenditure, they did not gain weight (Ross et al., 2000, 2004). In fact, Ross said, it was a challenge for them not to lose weight.
In a study on the effects of the amount and intensity of physical activity on blood lipids, Kraus and colleagues (2002) counseled subjects to maintain their weight. If participants started losing weight, they were asked to consume more calories to offset the loss. The researchers found that, regardless of the level of exercise and despite the counseling, participants lost weight when exercise was added.
Ross and his team conducted a similar study on the effects of the amount and intensity of exercise on abdominal obesity and glucose tolerance in 300 obese adults. All participants received personalized diet counseling, but in this case, they were asked to maintain their baseline energy intake, that is, to eat no more (Ross et al., 2015). When exercise was added 5 days per week for 6 months under supervised conditions, the researchers observed a tremendous loss in body weight and decreased waist circumference among all treatment groups (the treatment groups varied in the amount and intensity of exercise). Additionally, the researchers saw no compensation in terms of participants increasing their sedentary time. These findings suggested to Ross that, unless one eats more, if one is exercising under supervised conditions, one will increase one’s energy expenditure and lose weight. “I just don’t think there is any ambiguity there. There is none whatsoever. I think the information is very clear.”
In summary, according to Ross, efficacy randomized controlled trials have demonstrated that exercise without compensation in caloric intake leads to weight loss and that exercise with compensation in caloric intake prevents weight gain.
Effectiveness Trials: Is Exercise an Effective Strategy for Preventing Weight Gain in Adults?
What happens when adults are asked to exercise? In a systematic review of interventions aimed at preventing weight gain in adults, Lombard and colleagues (2009) identified nine randomized controlled trials with intervention lengths varying from a few weeks to 5 years. Overall, the interventions included far more women (1,595) than men (375). All of the interventions included diet and physical activity with behavior change strategies. Five of the nine studies found a significant difference in weight gain (1-3.5 kg) between treatment and control groups, due largely to increased body weight in the control groups. That result was expected, Ross remarked, given that the goal of the interventions was to prevent weight gain. In his opinion, it is impossible to identify which component(s) of the interventions (diet and/or exercise) were responsible for success. Additionally, he noted that
intensive interventions were not always successful, and interventions that included mixed modes of delivery with some personal contact tended to be successful. The authors of the review concluded that more evidence is urgently needed.
In a 24-month intervention trial aimed at preventing weight gain in healthy, nonobese first-year medical students, Hivert and colleagues (2007) used cognitive-behavioral techniques to encourage a healthy lifestyle among the students. They observed what Ross considered a modest difference in weight change (1.3 kg) between the intervention and control groups: the control group gained weight, and the intervention group lost a slight amount of weight.
In another trial, Levine and colleagues (2007) tested three interventions, one behavioral (i.e., clinic-based), one a correspondence intervention, and the third an information-only intervention (the control). Participants were given goals for diet and exercise. The trial was conducted for 2 years, with a 1-year follow-up (i.e., at year 3). Although the researchers observed no significant difference in weight change among groups, they detected a trend toward weight gain in the control and correspondence groups and prevention of weight gain in the clinic group.
In what Ross described as an atypically long 54-month intervention study, Kuller and colleagues (2001) encouraged modest weight loss among 535 peri- to postmenopausal women and then conducted an intervention with the goal of maintaining that weight loss in subsequent years. Reponses were varied, with a mean change in weight of –0.1 kg in the intervention group compared with a gain of 2.4 kg in the control group, representing a modest but significant difference. Additionally, participation in the intervention substantially reduced several cardiometabolic risk factors in the intervention compared with the control group.
As a final example of evidence from effectiveness trials aimed at understanding what happens when people are encouraged to exercise, Bruins and colleagues (2014) conducted a study of individuals with psychotic disorders. They found significant weight gain prevention in the intervention groups.
Most other trials of interventions to prevent weight gain have been successful in achieving that goal, according to Ross. However, all of the trials prescribed both diet and exercise programs, and none of the researchers were able to identify which component of the intervention contributed to weight gain prevention. Additionally, in general, the quality of such trials has been poor, in Ross’s opinion, with none having used intent to treat or being appropriately powered.
The Notion of Small Change: A Focus on the Prevention of Weight Gain
Keynote speaker James Hill was among the first in the field to champion the notion of small change, according to Ross (Hill et al., 2003). Ross echoed the call. Not only is prevention of weight gain important, he said, “but maybe that is what we should be striving for. Maybe treatment of obesity in this environment is just too great a challenge for us.” In a family-based study designed to test a small-change approach, Hill and colleagues asked participants to either increase their steps by 2,000 per day (i.e., about 100 calories spent) or decrease their caloric intake by 100 calories per day (Rodearmel et al., 2006). Those are very small changes, Ross observed. It was a short study, only 13 weeks, so almost a proof-of-concept study, he said. Not only did the intervention successfully prevent weight gain over time, but there was even some modest weight loss. Ross called for larger, longer studies of this type. At the time of this workshop, he and his team had just completed the recruitment phase of a 3-year randomized trial that would be testing a small-change approach.
According to Ross, evidence from efficacy-based randomized controlled trials clearly indicates that if adults increase exercise or physical activity, even if they compensate by eating more, the added exercise or physical activity prevents weight gain. “I don’t think there’s any doubt about that,” he said. Evidence from effectiveness studies suggests that lifestyle interventions designed to prevent weight gain are generally effective, although it is impossible to identify which aspects of such interventions—diet and/or exercise—are contributing to their success. Nor is there sufficient evidence to determine an ideal “dose” of exercise required to prevent weight gain. Many of the trials have been small, short term, and based on weak experimental designs, Ross observed.
Ross concluded by asking two questions. First, is it possible or important to identify the independent contributions of physical activity and diet in preventing weight gain? Second, what are the vital components of the ideal trial for determining the effects of lifestyle on the prevention of weight gain? Given the many ways to conduct trials, Ross encouraged ongoing dialogue.
Following Ross’s talk, he and Ekelund fielded questions from the audience.
Prevention of Weight Gain as a Goal
First, moderator Loretta DiPietro asked whether there is a survival advantage to keeping weight constant through middle age and whether the goal should really be to prevent weight gain, as opposed to attenuating the curvilinear shift that normally occurs.
Ekelund replied that his team’s data, collected over a period of about 5 years, showed a positive association between physical activity and gain in fat-free mass in adults older than 52 years (i.e., with higher levels of physical activity predicting greater gain in fat-free mass). In his opinion, that high levels of physical activity preserve fat-free mass in older adults is tremendously important for a number of different health outcomes.
Advantages of Measuring Body Composition Versus Weight
Jim Sallis observed that in his opinion, the most encouraging outcome described by Ekelund was related to waist circumference, not weight. He asked whether any conclusions presented by either speaker would have been different if the questions they had asked had been phrased in terms of body composition or body fat rather than weight.
Ross responded, “There is absolutely no question.” He and others have demonstrated not just an increase in fat-free mass and a reduction in adiposity with increased physical activity and in the absence of any change in body weight, but reductions in other cardiometabolic risk factors as well. “I have been a big proponent of that,” he said.
Ekelund added that it was important to keep in mind that almost all data obtained from observational research have been on body weight or BMI, and in some cases waist circumference. Very few large-scale epidemiologic studies have measured body composition. The few that have shown a differential effect by age on fat mass versus fat-free mass have demonstrated, as Ross’s data have, that physical activity has a positive effect on maintaining fat-free mass in older adults. Ekelund reiterated that the negative (or inverse) association between physical activity and waist circumference that he described was statistically, but probably not clinically, significant and that the observed change in waist circumference in the intervention group was minor.
Sustained Adoption of Behaviors
There was some discussion around the sustained adoption of behavioral strategies implemented in effectiveness trials—that is, whether indi-
viduals continue, after their trial has ended, to do what they were taught to do during the intervention to prevent weight gain. Ross remarked that participants do very well when they maintain contact with the interventionist. During that time, they sustain and benefit from their activities. “When they start to lose contact,” he said, “we’ve seen an erosion of benefits.”
This page intentionally left blank.