Predictors of Obesity, Weight Gain, Diet, and Physical Activity Workshop

Bethesda, MD

August 4-5, 2004


Agenda and Abstracts
Speaker Roster

Rationale and Objectives for the Workshop

National data have shown continuing increases in overweight among adults and children over the past 30 years (Flegal et al., 2002, Ogden et al., 2002). Obesity among adults 20 years and older has nearly doubled during that time. By 1999-2002, approximately two-thirds (65%) of adults were overweight or obese while nearly one-third (31%) of children ages 6-19 were overweight or at risk of being overweight (Hedley et al., 2004). These unabating increases in overweight and obesity clearly indicate the need for more effective interventions that reach all segments of the population. Efforts are needed to augment those currently in place, and will need to include multi-factorial approaches at the individual, community, and national level.

The main purpose of the workshop was to identify predictors of obesity, weight gain, and related lifestyle behaviors--diet and physical activity--that may help guide the design of intervention studies aimed at weight loss or prevention of weight gain. An underlying premise of the workshop was that a logical sequence of studies on obesity-related behaviors can lead to evidence-based public health interventions (Sallis et al., 2000). The steps involved in developing such interventions include (1) establishing links between obesity and specific behaviors, (2) developing and improving methods for accurately measuring these behaviors, (3) identifying factors that explain or influence these behaviors, (4) evaluating interventions to modify these identified behaviors, and (5) translating research into practice, i.e., disseminating effective interventions.

The workshop focused on existing data from longitudinal cohort studies that were ongoing during the time when the obesity epidemic developed. Most data presented were from studies of adults, although some were from studies on children and adolescents. Presentations about intervention studies, including some in managed care settings, as well as one on statistical methods were included. Investigators involved in intervention studies aimed at weight loss or prevention of weight gain were invited to discuss findings and develop recommendations for further research.

The specific objectives of the workshop were to:

  • Stimulate new data analyses of existing data
  • Identify key data and research gaps
  • Discuss implications for designing intervention studies
  • Facilitate discussion between cohort and intervention study investigators

A framework used to focus presentations and discussions was an ecological model of factors possibly influencing weight gain or the adoption of relevant lifestyle behaviors (Figure 1- graphic , Figure 1 - descriptive text for screen readers ). The model, developed for the workshop, captures multiple levels of influence on behaviors, including a wide range of factors: demographic, intra-personal, socio-cultural, organizational, physical environment (both natural and built), and policies. The workshop focused on modifiable factors, primarily intra-personal, socio-cultural, and physical environment factors.

This report provides a summary of findings, recommendations, and abstracts (to be posted) for each presentation.

Summary of findings

Longitudinal cohort studies have documented increases in weight and obesity prevalence, declines in physical activity, and increases in caloric intake over the past twenty or more years. These trends were found in all segments of the populations examined although some segments were identified as at higher risk than others, particularly African-American women, American Indians, and Mexican-American women. Such data provide an empirical rationale for targeting weight control efforts in specific subgroups.

Diet, Physical Activity, and Sedentary Behaviors as Predictors of Weight Outcomes

Most of the prospective studies measured diet at least once and physical activity multiple times and thus were able to examine relationships of diet and physical activity to weight outcomes. Prospective analyses clearly established a link between these behaviors and adverse weight outcomes (e.g., weight gain, incident overweight or obesity). In addition, they generally identified more consistent and stronger predictors of weight outcomes than cross-sectional analyses.

Several dietary components and behaviors consistently predicted adverse weight outcomes including greater fat intake (total or saturated), higher consumption of fast foods, and lower intake of fiber and/or whole grain foods. Other dietary components found to be predictive of weight gain but not examined in multiple studies included greater trans-fatty acid intake, higher consumption of sodas, a dietary pattern characterized as “empty calorie”, and snacking between meals.

Weight gain was consistently associated with declines in physical activity and fitness. Obesity was lowest in participants who maintained their physical activity levels over time. Although the type of activity was not routinely examined, one study found that weight training was associated with less subsequent gain in waist circumference among adult men.

Television watching was the most often measured sedentary behavior and was directly related to weight gain in most studies. Television watching was inversely associated with parental income and education in white but not black girls. In young adults, television watching was positively associated with smoking, alcohol consumption, low physical activity, hostility, and depression but inversely associated with age, education, income, and employment in cross-sectional analyses. One study that examined other sedentary behaviors found that adult women with more frequent television watching had the highest risk of developing obesity.

Other Predictors of Weight Outcomes

Initial weight at study entry was an important predictor of later weight gain or status in almost all studies--a finding that highlights the importance of preventing weight gain as early in life as possible. Data from white and black girls suggested that weight gain markedly increases during adolescence. In a study of young adults in their twenties, many were already overweight at study entry. Taken together these findings strongly indicate that prevention efforts must begin during childhood.

White girls were less likely to be satisfied with physical appearance and social acceptance with increasing adiposity, whereas black girls were more satisfied and felt socially accepted even with increasing adiposity. Black girls were more likely to engage in eating practices that could lead to overweight, such as eating big helpings and eating in the bedroom.

Among Latino and Asian adolescents, overweight was more prevalent for those who speak English at home. First generation immigrant adolescents consumed more fruit and less fast food than their US‑born counterparts. Among Mexican-American adult men but not women, greater education, income, and functional integration into the broader society (a measure of acculturation and assimilation) were associated with weight gain. These findings suggest the need for interventions among Latinos and other immigrants that are not only culturally-appropriate but also tailored to acculturation status.

Excess weight gain was higher in women following their first pregnancy but there was no association with further pregnancies. Weight gain was greater if new mothers were already overweight at study entry. Additionally, becoming a parent was associated with greater declines in physical activity among women, especially white women. These findings suggest that new mothers might be a possible target for intervention. For example, interventions with counseling and/or materials aimed at preventing weight gain and increasing physical activity could be provided to new mothers attending healthy baby clinics or pediatric visits.

A consistent but paradoxical finding in both cohort and intervention studies was that more frequent attempts to lose weight through dieting (either on their own or through a formal weight loss program) were associated with adverse weight outcomes. Because this finding may be confounded by pre-existing overweight in cohort studies, further research should attempt to clarify the temporal relationship. As a first step, further data analyses should stratify on overweight status at baseline.

Although somewhat inconsistent across studies, participants with better perceived health status were more likely to lose weight than those with poorer perceived health. Among persons who successfully lost weight, those who reported a medical condition were more likely to maintain their weight loss after 1-2 years. Findings related to perceived stress and weight outcomes were also inconsistent but suggest that participants with higher baseline stress levels gained more weight.

Few cohort studies have examined the influence of environmental factors on diet behaviors, physical activity, and weight outcomes. However, plans were presented for research underway in several cohort studies that will link physical environment measures to participant's residential addresses at multiple study visits using Geographic Information Systems (GIS). Such data will allow individual-level, longitudinal analyses that examine the impact of environmental shifts on physical activity, diet behaviors, and weight outcomes. Environmental factors include community design; transportation; availability of restaurants, grocery stores, and physical activity facilities; and socio-economic factors such as crime and community demographics. These studies should improve our understanding of the role of the physical environment in the development of the obesity epidemic as well as yield testable hypotheses for intervention studies.

Predictors of Dietary Behavior and Physical Activity

Higher education, of either the participant or participant's parents, was consistently associated with healthier dietary outcomes. Initial dietary intake was highly correlated with intake 7 years later in young adults. Poorer diets among adults were associated with living in lower income neighborhoods. For African American adults, having at least one large supermarket nearby rather than small grocers or convenience stores was associated with a healthier diet.

Substantial declines in physical activity occurred during adolescence in girls and were greater in black girls than in white girls. Declines in physical activity in this age group were associated with lower levels of parental education, more cigarette smoking, higher BMI, and pregnancy in white and black girls although not consistently in both race groups. Change in fitness in young adults was related to 7-year change in physical activity and BMI but not baseline levels of these variables. Smoking status at baseline and younger age were related to declines in fitness yet education was not.

Higher compared to lower socio-economic neighborhoods had more varied physical activity resources that might influence activity in adolescents, such as number of weekly physical education classes and availability of community recreation centers.

Predictors of Weight Loss in Trials or Weight Regain

Adults who successfully maintained a weight loss of greater than or equal to 30 lbs were followed for 1-2 years. Those who were able to avoid regaining weight were characterized by less depression, and more years at maintaining weight loss. Weight regain was more likely in persons with decreased dietary restraint, increased susceptibility to overeating, small relapses, and inconsistent dietary habits (e.g., weekdays vs. weekends, holidays vs. non-holidays).

In trials with weight loss interventions, participants with more confidence in their dietary choices, especially with respect to fat intake, and who showed greater and improved dietary restraint were more likely to lose weight or have greater weight loss. Thus new skills for coping with an environment that has increasingly more convenient ready-to-eat food with larger portions at lower cost may be critical in weight loss and prevention of weight gain. Intervention studies also showed that reduced consumption of high-fat foods and increased consumption of fruits and vegetables were associated with greater weight loss. Increased exercise and support from friends regarding exercise were also predictive of greater weight loss. Improvements in mental and physical health scores were predictive of greater weight loss. Self-monitoring of weight was also associated with greater weight loss. The effects of previous weight loss attempts on current weight loss were counterintuitive. More frequent previous attempts were associated with less weight loss, suggesting that frequent failed attempts may impede future success at weight loss.

Methodological Issues

The strengths and limitations of several statistical approaches to modeling predictors, mediators, and moderators were discussed. One approach, multilevel modeling, has the key advantage of simultaneously examining the effects of variables at both the individual and group level, as well as possible cross-level interactions. Data from longitudinal studies as well as those from trials with weight loss interventions could be used to test specific pathways in behavioral change models using this method.

Translation of Research to Medical Care Settings

Several trials involving weight loss interventions have developed successful strategies that are multifaceted, intensive, and costly. However, concerns have been raised about implementing such programs in medical care settings given the seemingly high costs. One approach presented at the workshop is to use a stepped care approach to behavioral treatment in medical care settings, with screening and advice for all patients, less intensive programs for moderate risk patients, and more intensive treatment programs for high risk patients. Some studies have shown that changes in diet or weight resulted from relatively few intervention contacts; these studies could be used as models for less intensive programs. Electronic medical records could incorporate prompts for physician advice and could also be used to evaluate screening and treatment programs.

Implications for Further Research and Prevention Efforts

Workshop participants noted that the availability of obesity-related predictor data was limited in some cohort studies largely because these studies were designed to examine cardiovascular disease outcomes and a variety of risk factors, not just obesity. If available in cohort studies, psychosocial, neighborhood environment, and social support questionnaires tend to be more general than those used in recent lifestyle intervention studies aimed at weight loss. In the latter studies, such questionnaires specifically target diet, physical activity, and sedentary behaviors as they related to weight outcomes. Thus design differences may account for the relatively few analyses of correlates and predictors of obesity-related behaviors such as diet and physical activity beyond basic demographic factors in most cohort studies. One consequence is that even if data are available in one cohort study, the findings generally can not be confirmed using other cohorts.

On the other hand, some longitudinal studies have begun to retrospectively add data on the physical environment. These data could not only improve our understanding of the influences of the physical environment on diet and physical activity but also may be used to document the development of the obesity epidemic. The addition of weight behavior questionnaires that allow examination of individual and more proximal social influences would greatly increase the value of these data to help us understand how our current environment interacts with individual and social factors to influence weight outcomes.

The longitudinal nature of these studies was recognized as an extremely valuable feature that should be exploited to its fullest potential by utilizing all available data points whenever possible. Data from these studies as well as those from trials with weight loss interventions could be used to test specific pathways in behavioral change models. These analyses could lead to hypotheses to be examined further in behaviorally-focused obesity research, either observational or clinical trials.

Finally, the workshop offered the opportunity for dialogue between investigators involved in longitudinal and intervention studies. Efforts to encourage further opportunities for collaboration and sharing of questionnaires should continue to be facilitated through the NHLBI and other websites (see links below).

Recommendations

Analyses of Existing Data From Longitudinal Studies

  • Examine predictors of long-term weight loss and maintenance taking into account involuntary vs. voluntary weight loss and baseline weight status. appropriate to overcome the small number of participants in these categories.
  • Use detailed dietary data to identify and develop simpler dietary measures, such as key marker foods and behaviors, which strongly predict healthy or unhealthy dietary patterns or weight outcomes.
  • Examine sedentary behaviors and their changes over time as predictors of weight and whether these associations are mediated through physical activity and diet.
  • Analyze changes in diet and physical activity jointly to examine their combined effects on energy balance/imbalance and weight outcomes.
  • Examine the effects on weight outcomes of portion size and their changes over time.
  • Examine predictors of diet patterns and behaviors and physical activity, particularly the relative contributions and potential interactions among personal, social, and physical environmental factors.
  • Exploit the longitudinal study design by utilizing all available time points.
  • Test theory-based behavioral change models, including the use of multi-level analyses that examine predictors, mediators, and moderators of outcomes.
  • Analyze change in predictor variables as they relate to change in weight outcomes and, if possible, model and test specific pathways.

Measures To Add To Longitudinal Studies

  • Objective measures of physical activity and fitness (e.g., accelerometers, treadmill tests), which use a protocol common to other studies
  • Questionnaires which measure psychosocial, neighborhood environment, social support, and other factors which specifically target diet, physical activity, and sedentary behaviors as they relate to weight outcomes (e.g., social support specific to weight loss attempts, dietary restraint, frequency of self-weighing, and weight loss history)
  • Questionnaires and/or community-level environmental data (e.g., through GIS linkage) to assess the impact of changes in the physical environment on changes in diet, physical activity, and weight
  • Short questionnaires that identify key indicator foods (e.g., soda) and diet behaviors (e.g. meal frequency) at every data collection period
  • Weight and dieting history questionnaires
  • Questions about portion sizes usually consumed
  • Questions about the role of family influences on diet and physical activity

Measures To Add To Intervention Studies

  • Objective measures of physical activity and/or fitness
  • Indicators of the physical environment, either self-perceived or GIS-based with associated data
  • Target the following behaviors in weight loss interventions: reduced consumption of fast foods and sodas, increased consumption of low fat and whole grain foods, decreased television viewing

Methodological Research To Develop New Measures

  • Develop short questionnaires containing key indicator foods and diet behaviors that strongly predict healthy or unhealthy dietary intake or patterns.
  • Develop questions related to portion size as an indicator of dietary behavior.
  • Develop easier methods or more focused questionnaires for self‑monitoring of important diet behaviors, physical activity, and weight.
  • Develop questions assessing the role of media and advertising on dietary choices.
  • Use a qualitative approach to adapt and develop questions that are culturally-appropriate.

Further Studies Needed

  • Longitudinal studies are needed that focus on determinants of weight outcomes and related lifestyle behaviors. These studies should utilize contemporary questionnaires that measure psychosocial, neighborhood environment, social support, and other factors which specifically target diet, physical activity, and sedentary behaviors as they relate to weight outcomes. Study sites should be selected in areas with good, existing environmental databases or, if unavailable, in areas where collection of local physical environment data is feasible.
  • Longitudinal studies in young children are needed to reduce the potential confounding of behavioral influences by pre-existing overweight.
  • Intervention studies are needed in new mothers aimed at returning to a healthy weight after pregnancy.
  • Studies are needed in immigrants to better understand the process and impact of acculturation and assimilation on weight outcomes

Links to Study Resources on the Web

NHLBI Limited Access Datasets

Framingham Heart Study

Jackson Heart Study

CARDIA Study

ARIC Study

Strong Heart Study

PREMIER Study

Active Living Research (Information on funded grants, references, and links to other sites)

Measures and Surveys Available for Download

Citations

Flegal KM, Carroll MD, Ogden CL, Johnson CL. Prevalence and trends in obesity among US adults, 1999-2000. JAMA. 2002 Oct 9;288(14)1723‑7.

Ogden CL, Flegal KM, Carroll MD, Johnson CL. Prevalence and trends in overweight among US children and adolescents, 1999-2000. JAMA. 2002 Oct 9;288(14)1728‑32.

Hedley AA, Ogden CL, Johnson CL, Carroll MD, Curtin LR, Flegal KM.  Prevalence of overweight and obesity among US children, adolescents, and adults, 1999-2002. JAMA. 2004 Jun 16;291(23)2847‑50.

Sallis JF, Owen N, Fotheringham MJ. Behavioral epidemiology: a systematic framework to classify phases of research on health promotion and disease prevention. Ann Behav Med. 2000;22(4)294-8.

Skip footer links and go to content
Twitter iconTwitterExternal link Disclaimer         Facebook iconFacebookimage of external link icon         YouTube iconYouTubeimage of external link icon