Condition

One of the Healthy People 2020 national objectives is to increase the prevalence of a healthy weight among adults to 34 percent and to reduce the prevalence of obesity among adults to less than 30 percent.¹ From 2005 to 2008, only 31 percent of adults were a healthy weight.² Obesity was estimated to cost $79 billion in the United States during 1995. By 2008, health care costs associated with obesity were thought to have risen to $147 billion. The Federal Government pays about one half of these costs through Medicaid and Medicare spending.³

Body mass index (BMI)—expressed as weight in kilograms divided by height in meters squared (kg/m²)—is commonly used to classify underweight (BMI <18.5 kg/m²), healthy or normal weight (BMI 18.5–24.9 kg/m²), overweight (BMI 25.0–29.9 kg/m²), obesity (BMI ≥30.0 kg/m²), and extreme obesity (BMI ≥40.0 kg/m²).

Adults tend to gain weight progressively through middle age. Although the average weight gained per year is 0.5 to 1 kg, the modest accumulation of weight over time can lead to obesity.⁴ The estimated age-adjusted prevalence of overweight and obesity (BMI ≥25.0 kg/m²) was 68 percent in the United States during 2007 and 2008. Despite the doubling in the prevalence of obesity between 1976 and 1980 and 2007 to 2008 (13 to 34 percent), the prevalence of overweight has remained stable between the same time periods (32 to 34 percent).

Obesity is a risk factor for chronic conditions including cardiovascular disease, type 2 diabetes, arthritis, certain types of cancer, and cancer recurrence.^5-12 Weight is associated with an increased risk of some forms of cancer and cancer recurrence. There is growing evidence that breast cancer survivors or women with breast cancer have better outcomes if they lose or maintain their weight. Obesity can also be caused by medications used to treat chronic disease, as is the case for antipsychotic treatments,¹³ some treatments for type 2 diabetes,^14,15 and tamoxifen and aromatase inhibitors for treatment or prevention of breast cancer or cancer recurrence.¹⁶ Higher grades of obesity are associated with excess mortality, primarily from cardiovascular disease, type 2 diabetes, and certain types of cancer.^16,17

We aimed to review studies of strategies to prevent weight gain among adults. The strategies of interest were self-management techniques, diet, physical activity, use of the dietary fat absorption inhibitor orlistat, or combinations of these strategies applied at the individual, community, or environment level. These strategies could have been implemented in any setting, including clinical care sites, community settings, higher education institutions, and workplaces. Strategies could have targeted individuals at high risk of gaining weight because of a family history of obesity or diabetes mellitus, personal risk factors for diabetes mellitus and cardiovascular disease (such as borderline values of laboratory measures), use of medication associated with weight gain,¹⁸ or have had more inclusive enrollment criteria.^19,20

Literature Search Strategy

We searched the following databases for primary studies: MEDLINE^®, Embase^®, the Cochrane Central Register of Controlled Trials, CINAHL^®, and PsycINFO^® through June 2012. We developed a search strategy for MEDLINE, accessed via PubMed^®, and developed comparable searches using the other databases. We also reviewed the reference lists of each included article, relevant review articles, and relevant studies identified in ClinicalTrials.gov.

Title, abstract, and full article reviews were performed by two independent reviewers to identify relevant publications. Only one reviewer had to identify the publication as relevant to be included at title review. At abstract review, both reviewers had to agree that the study did not include any exclusion criteria (Table A). At full article review, both reviewers had to agree that the article met the inclusion criteria. Conflicts were resolved by consensus adjudication.

Table A

Study inclusion and exclusion criteria.

Relevant data were extracted from eligible trials of interventions and observational studies of approaches with a focus on items related to the population, interventions and approaches, comparisons, outcomes, timing, and setting. Each article was serially abstracted first by a first reviewer and then by a senior reviewer. Serial data abstraction involved a senior reviewer (faculty-level project investigator) abstracting data from articles while having access to the first reviewer's data. Differences in opinion were resolved through consensus adjudication and, for difficult cases, during team meetings. The timepoints of interest for data abstraction of weight outcomes were at 1 year, 2 years, 5 years, and the last reported timepoint after 5 years of followup. For the intermediate outcomes, safety, clinical, and quality of life outcomes, we only abstracted data for the last reported timepoint on or after 1 year.

Quality Assessment of Individual Studies

Study quality was assessed using the Downs and Black methodologic quality assessment checklist (Appendix F in the full report).²¹ This checklist was developed to assess the quality of reporting, internal validity, and external validity of randomized and observational studies. We used information on study quality to assess the risk of bias (using the internal validity items) and directness (using the external validity items) of the studies. Two reviewers independently completed the checklist for each article and came to consensus for each item.

Data Synthesis

When there were three or more studies with comparable interventions and comparable outcome measures, we considered quantitative pooling of the results. We examined the studies' designs for qualitative similarities. Because we found that no groups of studies were amenable to pooling with meta-analyses, we calculated and displayed the mean differences, risk differences or relative risks with 95% confidence intervals (CI) for the individual studies grouped by study population and comparable interventions. Observational studies did not report categories of approaches consistently, prohibiting the use of summary figures.

We selected a meaningful between-group-difference threshold in addition to a statistically significant threshold (p<0.05) for reporting on the outcomes. A meaningful difference threshold was defined as 0.5 kg of weight,^4,22 0.2 units of BMI (based on a 0.5 kg change for an individual with a BMI of 27), or 1 cm of waist circumference relative to the comparison group. The meaningful threshold was annualized to account for the different durations of the studies. For example, the weight difference threshold was 1 kg for outcomes reported at 2 years and 2.5 kg at 5 years.

Rating the Body of Evidence

At the completion of our review, we graded the quantity, quality, and consistency of the evidence addressing Key Questions 1 through 6 by adapting an evidence grading scheme recommended by AHRQ's “Methods Guide for Effectiveness and Comparative Effectiveness Reviews.”²³ We created evidence grades for each comparison and outcome by population or setting. We used four domains to yield a final evidence grade: Risk of Bias, Consistency, Directness, and Precision.

We classified evidence pertaining to Key Questions 1 through 6 into four categories: (1) “high” grade (indicating high confidence that the evidence reflects the true effect and further research is very unlikely to change our confidence in the estimate of the effect); (2) “moderate” grade (indicating moderate confidence that the evidence reflects the true effect and further research may change our confidence in the estimate of the effect and may change the estimate); (3) “low” grade (indicating low confidence that the evidence reflects the true effect and further research is likely to change our confidence in the estimate of the effect and is likely to change the estimate); and (4) “insufficient” grade (no evidence identified). A comparison-outcome pair with high strength of evidence was one with low risk of bias, consistency (or not applicable if only one study contributed), directness, and precision. Moderate strength of evidence indicated that one of the following was observed: a moderate risk of bias, inconsistency, indirectness, or imprecision. Low strength of evidence indicated a high risk of bias or two or more of the following: a moderate risk of bias, inconsistency, indirectness, and imprecision. Details on how the risk of bias, consistency, directness, and precision were identified are provided in the body of the report. For consistency with the reporting of the results, we graded the strength of evidence using the above process for each population or setting.

The team members discussed the process they used to grade the evidence throughout the report writing process. When a team member felt the evidence grade was questionable, this comparison-outcome evidence grade was discussed at a team meeting.

Applicability

We describe the applicability of studies in terms of the degree to which the study population, interventions or approaches, outcomes, and settings were relevant to individuals at risk of weight gain and features that may affect the effectiveness of the strategy.²⁴

The populations included in the studies affect the generalizability of the results. For this reason, we report the results ordered by the studied population. The most inclusive population (adults from the general population not selected based on underlying comorbidity or setting) is reported first, followed by strategies that were evaluated in individuals in a specific setting (workplace based and college based) and finally by groups of individuals with a disease or at risk of a disease (cardiovascular disease, cancer, and mental health).

Defining Effectiveness

For an intervention or approach to be considered effective, it had to meet the meaningful between-group difference and statistical thresholds. We selected a meaningful between-group-difference threshold in addition to a statistically significant threshold (p<0.05) for reporting on the outcomes. A meaningful difference threshold was defined as 0.5 kg of weight,^4,22 0.2 units of BMI (based on a 0.5 kg change for an individual with a BMI of 27), or 1 cm of waist circumference relative to the comparison group. The meaningful threshold was annualized to account for the different durations of the studies. For example, the weight difference threshold was 1 kg for outcomes reported at 2 years and 2.5 kg at 5 years.

Results of Literature Searches

From the 24,870 unique articles identified from electronic resources, 58 publications were included describing 51 studies. Thirty-eight trials included 150,081 participants at baseline. The majority (55 percent) of the trials were randomized trials that were not explicitly designed to prevent weight gain. Thirteen observational studies included 420,986 participants at baseline. Most of the observational studies were subanalyses of existing cohorts or randomized trials. Only one of the observational studies came from a cohort that was explicitly designed to measure weight change over time.²⁵

Results by Population or Setting of the Intervention or Approach

The strength of evidence is not high for any of the tested interventions or the approaches described in observational studies to prevent weight gain as measured by changes in BMI, weight, or waist circumference. Other than workplace-based strategies, which have moderate strength of evidence of effectiveness, most evidence was low or insufficient.

When adherence was reported, it tended to be poor, with less than 80 percent adherence to interventions. The one study that assessed awareness of an environmental intervention found inconsistent awareness of all of the components implemented in the workplace.

Very few studies reported on obesity-related clinical outcomes (mortality, quality of life, or cancer recurrence) or adverse effects. All evidence for these outcomes was graded as low or insufficient. No comparative study of orlistat for weight gain prevention was identified, so Key Question 4 has no evidence.

The interventions and approaches that prevented weight gain are described by population and by setting below. In a population, if none of the strategies prevented weight gain, we describe all of the strategies that we identified. The strength of evidence for the body of evidence is provided in Table B, which also includes the evidence about secondary outcomes.

Table B

Summary of the strength of evidence.

Key Findings and Strength of Evidence

We did not find strong evidence that any strategy prevents weight gain. This conclusion is similar to a previous systematic review on prevention of weight gain.²⁶

Interventions that were potentially effective included a clinic-based program to teach heart rate monitoring, a lifestyle intervention targeted at mothers of young children, workplace interventions with individual and environmental components, small group sessions to educate college women about healthy lifestyles, exercise for individuals at risk of cardiovascular disease and diabetes, and exercise performed at home among women with cancer. Potentially effective approaches, identified from observational studies, included eating meals prepared at home among college graduates and less television viewing among individuals with colorectal cancer.

No strategy was graded as having a high strength of evidence for its effectiveness. Workplace-based interventions and physical activity for women with cancer were graded as having a moderate strength of evidence. All other comparisons had low or insufficient evidence. The strength of evidence was low for many comparisons because the studies were not designed to measure weight maintenance or prevent weight gain and the study staff that measured weight in the intervention studies may have been aware of the participants' exposure groups. For observational studies, only one study mentioned that the original cohort was designed to measure weight longitudinally and qualified as direct evidence.²⁵

Despite the attention on primary prevention of obesity,^1,27-32 there is little evidence to recommend specific strategies. Existing recommendations are based on intermediate measures of changes in diet or physical activity or cross-sectional measures of weight. For example, the recommendations by the Centers for Disease Control and Prevention for community interventions to prevent obesity acknowledge that the evidence to support the recommendations were not based on sufficiently long studies that measured weight as an outcome, but on short-term changes in food choices or use of environmental modifications to facilitate physical activity.²⁷

Efforts by primary care providers to share information from publications on evidence-based research to prevent weight gain may be limited by the lack of reimbursement for the time to provide information and counseling. The World Health Organization European Ministerial Conference on Counteracting Obesity recommended that primary care providers play a more active role in preventing obesity.³⁰ Although the American Medical Association recommends talking with patients about how to prevent inappropriate weight gain,³³ reimbursement for the time required to provide weight maintenance counseling for the nonobese is not supported by Medicare.^34,35 The lack of reimbursement may at least partly explain the low adherence with these recommendations by providers. An analysis of the Behavioral Risk Factor Surveillance Survey data in 2003 indicated that only 2.6 percent of individuals with a BMI between 18 and 25 kg/m² received advice to maintain their current weight by a health care provider.³⁶

Although evidence is limited to support strategies associated with weight gain prevention, the rationale to prevent weight gain is sound given the robust evidence that obesity is associated with poor health outcomes,^5-12 is costly,³ and is difficult to reverse.³⁷

Areas for future research may focus on periods when people are already making other life changes and identifying strategies that people are interested in implementing before using resources to administer an intervention. Three of the interventions targeted populations experiencing life changes such as attending college³⁸ or beginning to cohabitate with a partner.³⁹ During these periods of change, individuals may be more amenable to accept a lifestyle modification or more likely to be adherent to the changes. Although these interventions did not uniformly result in weight maintenance compared with control or result in higher levels of adherence, designing interventions to be implemented during these and other life changes (e.g., postpartum, retirement, relocating to a new region) may be considered in future research.

Identifying an individual's interest level in an intervention prior to recommending a weight maintenance strategy may also be of interest. Many studies randomized participants to an intervention followed by multiple in-person visits, phone calls, and mailings. One trial opted to provide patients with up to three phone invitations to participate in a walking program compared with an information session. Only 33 percent of those invited to walk took part in a walk.⁴⁰ Allowing the option of participating in an intervention (rather than required visits or phone calls) demonstrates that individuals who consent to participate in a weight-related study may not have the motivation to participate in the particular intervention of interest. Combining a time when a person is already in a period of change with an intervention that an individual is motivated to participate in may be an area for future research.

Applicability

These findings apply primarily to overweight individuals. No study included healthy-weight individuals exclusively. The one study of obese individuals included abdominally obese individuals.

Adherence was poor in many trials. The results may have been more useful if they had been reported by adherence status in addition to the intent-to treat analyses. For example, if participants who adhered to an intervention were more likely to maintain weight than the nonadherent participants, this would have been valuable information.

Behavior change is difficult for individuals whose goal is to prevent weight gain, just as behavior change is difficult for those attempting to lose weight. Workplace interventions with environmental-level change may be a way to help those attempting to prevent weight gain and those who aim to lose weight modify their behavior, especially when the workers are made aware of the intervention. For individuals, eating more meals prepared at home and decreasing television viewing are simple, low-cost changes that prevent weight gain. College groups to discuss healthy diets and physical activity also prevented weight gain. The more intensive diet and physical activity interventions reported few adverse events. Although these intensive interventions did not result in strong evidence to promote their adoption, there is no evidence that not adopting a strategy to prevent weight gain is preferable.

Limitations

The strength of evidence is low or insufficient for almost all comparison-outcome relationships. There are several reasons for these low grades based on how we assessed each study's quality and graded the strength of evidence. First, intervention trials were frequently downgraded for lack of blinding, for not reporting the blinding of outcome assessors, or for not accounting for losses to followup. We feel that these quality elements are required to reduce the risk of bias. Although some may argue that an objective measure such as a weight measurement is not subject to bias, we suggest that the role played by a nonblinded assessor still poses a risk of bias in this measurement. Second, we included in this review only observational studies that accounted well for confounding and for losses to followup to ensure that we included only the highest quality observational studies. The inclusion of only high-quality observational studies narrowed the body of evidence but we could not have confidence in outcomes from studies that did not account for confounding or selection bias. Third, very few interventions had a stated goal of weight maintenance or weight gain prevention, a requirement for having direct evidence. We excluded studies that explicitly mentioned that at least some of the patients had a goal of weight loss. The best-known weight gain prevention trial— the Pound of Prevention trial—was excluded for this reason.⁴¹ Only one observational study was nested within a cohort whose original design had a weight-related outcome of interest.²⁵ Fourth, very few studies reported standard errors or confidence intervals for the between-group differences in change in a weight-related outcome over time. When the majority of studies did not report a measure of variability, we graded the body of evidence as imprecise. In some instances, the studies did not report a mean difference or point estimate stating only there was no significant difference in weight change between the groups.

There were also several limitations of the literature base. First, many studies did not report a weight-related goal and yet were included because they did report weight outcomes. We may have inadvertently included some trials that had a goal of weight loss but that did not say so explicitly in the published paper. Studies reported as weight maintenance among overweight and obese individuals may not have been solely targeting weight maintenance, but implicitly implied weight loss. We excluded studies that included specific aims of “weight change” associated with power calculations for an expected decrease in weight among the intervention group. However, some studies did not report power calculations or an expected direction of weight change. These studies were included. We also included observational studies that include participants with unknown weight-related goals. Second, controls had better weight maintenance than expected. In many studies, the weight maintenance in the control groups was better than is expected in a general population. Many control groups had no increase in weight over time. In the general United States population, adults gain about 0.5 kg per year.⁴ Individuals enrolled in intervention studies may be more likely to make behavior changes regardless of the group assignment. It is possible that the knowledge that one will be evaluated on weight regularly may help people to maintain weight without an intensive intervention. This may support the use of simple weight surveillance interventions in a workplace or primary care setting. Third, very few studies reported on obesity-related clinical outcomes or adverse events. Only one study in the general population reported on mortality. The few trials that did report on adverse events stated that none were associated with the intervention. Although, none of these studies stated what adverse events they collected or how they were measured.

Strengths

The greatest strength of the evidence base was the variety of populations included. In addition to including populations with very few exclusion criteria representative of the general population, we also observed a variety of studies targeted at individuals that are known to be more likely to gain weight.

Research Gaps

We suggest that most comparisons and outcomes that have low or insufficient evidence are future research needs. In particular, we recommend future research to examine strategies to prevent weight gain among healthy weight individuals and, separately, overweight and obese individuals. Interventions for individuals initiating antipsychotic medications are also a high priority given that participants of a trial gained 10 kg in the first year of medication use. Diabetes and some cancer medications are also associated with weight gain. Interventions for patients initiating diabetes medications, tamoxifen, and aromatase inhibitors are also a high priority subpopulation, although weight loss goals may be more relevant for some of these patients.

Different degrees of intensity of the strategies should be compared. Less intensive interventions may be possible given that control groups maintained weight comparable with the intervention groups in most of the studies that we included.

There are design and reporting considerations that should be considered for future studies. Observational cohorts should make measuring weight a stated goal in their protocols if that is the intent. Intervention trials should be of sufficient duration to adequately assess the efficacy of interventions to prevent weight gain. We suggest that 1 year should be a minimum duration of followup for these interventions. Longer followup will make it easier to identify true effectiveness if individuals are expected to gain only 0.5 kg per year.

Conclusions

The evidence provides some, although limited, support for strategies to prevent weight gain. Potentially effective strategies included ones that involve minor behavior change (eating more meals prepared at home) or more major changes (endurance exercise training in a gym at least three times per week). Although there is no strong evidence to promote a particular weight gain prevention strategy, there is no evidence that not adopting a strategy to prevent weight gain is preferable.