Review Scope

The current review is an update of the 2011 LeBlanc et al review.¹⁹⁷ Unlike the previous review, populations selected based on the presence of a chronic disease in which weight loss or weight loss maintenance is part of disease management (e.g., arthritis, known CVD, type 2 diabetes) have been excluded. Pharmacological interventions included in this review are limited to those that are approved by the FDA for long-term chronic weight management; therefore, although metformin was reviewed in the 2011 evidence review, it was not included in the current review. We included four new weight loss medications that have been approved since the last review: liraglutide, lorcaserin, naltrexone and bupropion, and phentermine-topiramate, and one medication included in the previous review (orlistat).

Analytic Framework and Key Questions

We developed an Analytic Framework (Figure 1) and three Key Questions (KQs) to guide the literature search, data abstraction, and data synthesis.

Figure 1

Analytic Framework. Abbreviations: HRQoL = health-related quality of life

Data Sources and Searches

In addition to considering all studies from the previous review on this topic¹⁹⁷ for inclusion in the current review, we performed a comprehensive search of MEDLINE, PubMed Publisher-Supplied Records, PsycINFO, and the Cochrane Central Registry of Controlled Trials. We searched between January 1, 2010, and June 6, 2017, building upon the most recent full search for this topic. We worked with a research librarian to develop our search strategy, which was peer-reviewed by a second research librarian (Appendix B). All searches were limited to articles published in English.

In addition to these database searches, we searched ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform (www.who.int/ictrp) for ongoing trials through August 2017. We also examined the reference lists of previously published reviews, meta-analyses, and primary studies to identify any potential studies for inclusion. We examined the FDA review documents for each included medication to identify any additional studies not published in the primary literature. We supplemented our searches with suggestions from experts and articles identified through news and table-of-content alerts such as those produced by the USPSTF Scientific Resource Center LitWatch activity.¹⁹⁸ We managed literature search results using version X7 of Endnote® (Thomson Reuters, New York, NY), a bibliographic management software database.

Study Selection

Two reviewers independently reviewed the title and abstract of all identified articles using DistillerSR (Evidence Partners, Ottawa, Canada) to determine if the study met our a priori inclusion and exclusion criteria for design, population, intervention, and outcomes (Appendix B Table 1). Two reviewers then independently evaluated the full-text article(s) of all potentially relevant studies against the complete inclusion and exclusion criteria. Disagreements in the abstract and/or full-text review were resolved by discussion.

For all KQs we included RCTs, including cluster randomized trials and controlled clinical trials focused on weight loss in individuals who are overweight or have obesity, or maintenance of previous weight loss. In addition, for KQ3 (potential harms of weight loss/maintenance interventions) we included systematic reviews and large cohort, case-control, or event monitoring studies. We excluded studies with a primary aim of the prevention of overweight or obesity. Studies included for KQ1 and KQ2 had to report weight/adiposity change at least 12 months following the start of the intervention to be included. No minimum followup for KQ3 was required.

We included studies among adults age 18 years or older who were candidates for weight loss/maintenance interventions and selected based on an above normal BMI (e.g., ≥25 kg/m²) or other weight-related measure (e.g., waist circumference). In cases where lower BMI thresholds were used for eligibility (e.g., ≥23 kg/m²) or where participants were selected based on other cardiovascular risk factors (e.g., hypertension, impaired fasting glucose) without weight-related eligibility criteria and the focus of the intervention was clearly weight loss, we examined the distribution of the mean BMI at baseline to evaluate potential inclusion. We allowed in studies where 100 percent of the sample had a BMI above 23 kg/m², 95 percent of the sample had a BMI above 24 kg/m², or 90 percent of the sample had a BMI above 25 kg/m². These individuals may have additional risk cardiovascular risk factors (e.g., hypertension); however, we excluded studies in adults with a chronic disease for which weight loss/maintenance is part of disease management (e.g., known CVD, diabetes mellitus). In addition, we excluded studies in adults with known chronic diseases not generalizable to the primary care population (e.g., eating disorders, chronic kidney disease). Studies in adults with secondary causes of obesity, pregnant women, and institutionalized adults were excluded. The evidence related to weight loss in children and adolescents is addressed in a separate review.¹⁹⁹

We included interventions that were conducted in or recruited from primary care or a health care system or that we judged could feasibly be implemented in or referred from primary care. We included studies of commercial weight loss programs that are widely available in the community at a national level. We excluded studies that took place exclusively in or in conjunction with worksites, churches, or other settings that are not generalizable to primary care given pre-existing social ties that are not easily reproducible in primary care.

We included interventions focused on weight loss or maintenance of previous weight loss including: behavioral counseling (either alone or part of a multicomponent intervention), training of health care providers, pharmacologic interventions approved by the FDA as first-line long-term weight loss/management medications, and combinations of these interventions. Interventions could be delivered via face-to-face contact, telephone, print materials, or technology (e.g., computer-based, text messages), and by numerous potential interventionists, including but not limited to: physicians, nurses, exercise specialists, dietitians, nutritionists, and behavioral health specialists. Included behavior-based interventions had to focus on healthful diet and nutrition, physical activity, sedentary behavior, or a combination thereof and include behavior change techniques such as: assessment with feedback, advice, collaborative goal-setting, assistance, exercise prescriptions (referral to exercise facility or program) or arranging further contacts. We excluded studies of surgical and nonsurgical devices and procedures, medications not approved by the FDA for long-term weight loss or weight loss maintenance, complementary and alternative treatments, and dietary supplements.

Given the elevated level of lifestyle counseling that now occurs as part of standard care, we allowed more intensive control groups than in the previous review. For studies of behavior-based interventions, we included only studies that had the following controls: no intervention (e.g., wait list, usual care, assessment-only), minimal intervention (e.g., usual care limited to quarterly counseling sessions or generic brochures), or attention controls (e.g., similar format and intensity but different content). We excluded studies that evaluated the comparative effectiveness of two active interventions without the addition of a true control group. For studies of pharmacologic interventions, we included only placebo-controlled studies in which participants all received the same behavior-based interventions. For the greatest applicability to U.S. primary care practice, we included only studies conducted in economically developed countries, defined as member countries of the Organisation for Economic Co-Operation and Development.²⁰⁰ Finally, due to resource constraints, we included only studies for which results were published in English.

Health outcomes included mortality, morbidity, depression, health-related QOL, and disability. Intermediate outcomes included weight measurements, measures of total and central adiposity, incidence or prevalence of obesity-related conditions, and proportion of individuals taking medication for obesity-related conditions. Unlike the 2011 review, the effects of weight loss interventions on intermediate cardiometabolic measures (i.e., continuous measures of blood pressure, cholesterol levels, and glucose levels) was not included; rather, we focused on the incidence or prevalence of specific diseases/risk factors (e.g., diabetes, hypertension). Adverse events included treatment-related harms and discontinuation of medication due to adverse effects at any time point during intervention. We did not include studies that evaluated potential harms of weight loss itself (i.e., harms had to be related to a weight loss or maintenance intervention that met our inclusion criteria, including having an adequate comparison group).

Two reviewers independently assessed the methodological quality of each study using predefined study-design specific criteria developed by the USPSTF.¹⁹⁸ Disagreements in quality were resolved by discussion. Each study was given a final quality rating of good, fair, or poor. Good-quality studies were those that met nearly all of the specified quality criteria (e.g., comparable groups were assembled initially and maintained throughout the study, followup was approximately ≥85%, conservative data substitution methods were used in cases of missing data, no evidence of selective outcome or analysis reporting), whereas fair-quality studies did not meet these criteria but did not have serious threats to their internal validity related to the design or execution of the study. Studies we rated as poor-quality had several important limitations, including at least one of the following risks of bias: very high attrition (generally >40%), differential attrition between intervention arms (generally >20%); lack of baseline comparability between groups without adjustment; methods for ascertainment of weight outcomes were unclear or differed between groups (e.g., self-report or objective measurement and not reported by group), or issues in trial conduct, analysis, or reporting of results (e.g., possible selective reporting, inappropriate exclusion of participants from analyses, and questionable validity of randomization and allocation concealment procedures). Studies rated as poor quality were excluded from the review. In studies of pharmacologic interventions most dropout is due to adverse events or lack of effectiveness and not loss to followup. We allowed studies with more than 40 percent attrition to be rated as fair quality if they used adequate data substitution methods with sensitivity analyses using different methods (e.g., modified intention-to-treat [mITT],¹⁵⁷ baseline observation carried forward, multiple imputation using a mixed effects model).²⁰¹ Because this review was an update of our own work, we did not repeat critical appraisal of the original studies through full dual-quality rating; rather, we confirmed the quality rating during data abstraction. In two cases a study included in the previous report was excluded for poor quality upon rereview due to several methodological issues, including high attrition with lack of adequate data substitution methods, lack of analysis description, and allocation concealment issues.^{202, 203}

For all of the included studies, one reviewer extracted key elements into standardized abstraction forms in Microsoft Access® 2010 (Microsoft, Redmond, WA). A second reviewer checked the data for accuracy. For each study, we abstracted general characteristics of the study (e.g., author, year, study design), clinical and demographic characteristics of the sample and setting (e.g., age, race/ethnicity, baseline clinical characteristics, setting, country), analytic methods, and results. For intervention characteristics, we abstracted detailed information about specific components: duration, number, and length of sessions; group or individual delivery of counseling; mode of delivery (i.e., in-person, telephone, electronic, or print); providers and provider training; setting; and adherence to the intervention. We abstracted the number of sessions and length of sessions according to what was planned (and not necessarily what was implemented). In order to summarize and compare interventions’ intensity, we abstracted the total number of sessions conducted and the total number of contacts made in the first 12 months for each intervention arm. For this, sessions included any group or individual counseling session, conducted face-to-face or by telephone or any web- or computer-based module or session, whereas contacts included all sessions plus contacts made through mobile phone text messages, emails, or interactions with other web-based or social media platforms. In this case, the number of contacts was always greater than the number of sessions. As described below, both variables were considered when exploring effect modification by intervention intensity.

We categorized each study according to the selection of participants into the study based on their cardiovascular or cancer risk. The four categories of risk were: 1) increased cardiovascular risk (e.g., selection was based on having one or more cardiovascular risk factor such as hypertension, dyslipidemia, metabolic syndrome), 2) subclinical increased cardiovascular risk (e.g., selection was based on having prediabetes, prehypertension, or other clinical risk factor for diabetes such as gestational diabetes), 3) elevated cancer risk (e.g., studies in which participants were cancer survivors or who had a premalignant condition), and 4) low cardiovascular risk or unselected (i.e., studies that did not select participants on the basis of their cardiovascular or cancer risk). Studies categorized as low risk or unselected generally enrolled participants based on overweight and/or obesity status, age, and other demographic characteristics.

Data Synthesis and Analysis

We synthesized data separately for each KQ and according to the focus of the intervention (i.e., behavior-based weight loss interventions, behavior-based weight loss maintenance interventions, medication-based weight loss interventions and medication-based weight loss maintenance interventions). Results for each medication were analyzed and reported separately. The data on health outcomes (KQ1), intermediate outcomes such as incident cases of diabetes or metabolic syndrome (KQ2), and adverse events (KQ3) did not allow for quantitative pooling due to the limited number of contributing studies and the variability in outcomes measured, so we summarized those data in tables and narratively. For the results of medications on weight loss outcomes, there were too few trials (2 to 3) for each drug to be pooled. For orlistat, where there were 11 trials reporting weight loss outcomes, there was inconsistency in the measurements reported for within- and between-group effects (e.g., means, least squares means) and a lack of reporting of variance precluded meta-analyses of continuous outcomes. We chose not to meta-analyze the nine orlistat trials that reported the proportion of participants losing at least 5 or 10 percent of their initial body weight given concerns regarding several of the trials’ high and differential attrition. Instead, we presented a forest plot (without pooling) to illustrate each trial’s results.

For behavior-based interventions, we ran random-effects meta-analyses using the method of DerSimonian and Laird to calculate the pooled differences in mean changes (for continuous data) and a pooled risk ratio (for binary data) for weight outcomes (KQ2).²⁰⁴ Details of our data analysis methods are included in Appendix B. Briefly, we used the between-group differences for each outcome as reported by each respective study and favored adjusted over unadjusted effect estimates. If a between-group effect estimate and variance were not provided, we calculated a crude effect estimate. Within the pooled analyses, we grouped 12- to 18-month followup data together and 24-month data separately. If a trial reported both 12- and 18-month data, we chose 12-month data to pool. If a trial had more than one active intervention arm, we plotted the most intensive arm or the arm that was the most similar with other interventions included in the analysis. Of note, we did not include the Diabetes Prevention Program (DPP) treatment arm randomized to metformin²⁰⁵ or the POWER-UP enhanced brief lifestyle counseling arm, which included the participants’ choice of meal replacements or weight-loss medications (orlistat or sibutramine),²⁰⁶ given our review inclusion criteria. We presented the results of other time points and other intervention arms in tabular format.

We examined statistical heterogeneity among the pooled studies using standard χ² tests and estimated the proportion of total variability in point estimates using the I² statistic.²⁰⁷ We applied the Cochrane’s rules of thumb for interpreting heterogeneity: less than 40 percent likely represents unimportant heterogeneity, 30 to 65 percent, moderate heterogeneity; 50 to 90 percent, substantial heterogeneity; and more than 75 percent, considerable heterogeneity.²⁰⁸ We generated funnel plots to evaluate small-study effects (a possible indication of publication bias) and ran the Egger’s²⁰⁹ or Peters’²¹⁰ test to assess the statistical significance of imbalance in study size as well as findings that suggested a pattern.

We used visual displays and tables grouped and sorted by potentially important characteristics and a series of meta-regressions to investigate whether variability among the results was associated with any prespecified study, population, or intervention characteristics. Specifically, we examined study quality (good vs. fair), percent of participants retained at 12 to 18 months, link to primary care (conducted in or recruited from primary care or not), whether the trial was set in the United States, risk status of the sample (increased cardiovascular, subclinical, or cancer risk vs. low risk or unselected), participant selection approach (self-selected vs. directly recruited), and a number of intervention characteristics (number of sessions and contacts in the first year, intervention duration, the main mode of intervention delivery, the presence of any group, individual, or technology-based components, and the use of self-monitoring).

We used Stata version 13.1 (Stata Corp LP, College Station, TX) for all quantitative analyses. All significance testing was two-sided, and results were considered statistically significant if the p-value was 0.05 or less.

Grading the Strength of Evidence

We graded the strength of the overall body of evidence for each KQ. We adapted the Evidence-based Practice Center approach,²¹¹ which is based on a system developed by the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group.²¹² Our method explicitly addresses four of the five Evidence-based Practice Center-required domains: consistency (similarity of effect direction and size), precision (degree of certainty around an estimate), reporting bias (potential for bias related to publication, selective outcome reporting, or selective analysis reporting), and study quality (i.e., study limitations). We did not address the fifth required domain—directness—as it is implied in the structure of the KQs (i.e., pertains to whether the evidence links the interventions directly to a health outcome).

Consistency was rated as reasonably consistent, inconsistent, or not applicable (e.g., single study). Precision was rated as reasonably precise, imprecise, or not applicable (e.g., no evidence). Reporting bias was rated as suspected, undetected, or not applicable (e.g., when there is insufficient evidence for a particular outcome). Study quality reflects the quality ratings of the individual trials and indicates the degree to which the included studies for a given outcome have a high likelihood of adequate protection against bias. Limitations highlights important restrictions in answering the overall KQ (e.g., lack of replication of interventions, nonreporting of outcomes important to patients).

We graded the overall strength of evidence as high, moderate, or low. “High” indicates high confidence that the evidence reflects the true effect and that further research is very unlikely to change our confidence in the estimate of effects. “Moderate” suggests moderate confidence that the evidence reflects the true effect and that further research may change our confidence in the estimate of effect and may change the estimate. “Low” indicates low confidence that the evidence reflects the true effect and that further research is likely to change our confidence in the estimate of effect and is likely to change the estimate. A grade of “insufficient” indicates that evidence is either unavailable or does not permit estimate of an effect. Two independent reviewers rated each KQ according to consistency, precision, reporting bias, and overall strength of evidence grade. We resolved discrepancies through consensus discussion involving more reviewers.

Expert Review and Public Comment

The draft Research Plan was posted for public comment on the USPSTF Web site from December 10, 2015, to January 13, 2016. Several comments suggested including studies of women during the postpartum period; the USPSTF changed the Research Plan to include postpartum women. A final research plan was posted on the USPSTF’s Web site on March 31, 2016.

Invited content experts and federal partners reviewed a draft of this report. Their comments were presented to the USPSTF during its deliberation of the evidence and were considered in preparing the final evidence review. Additionally, a draft of this report was posted for public comment on the USPSTF Web site from February 20, 2018 through March 19, 2018. A few comments were received during this public comment period. Minor wording changes and clarifications were made based on these comments; however, no significant changes were made to the report. All references suggested for inclusion by commenters were reviewed against full-text criteria; however, none met criteria for inclusion in the report.

USPSTF Involvement

We worked with six USPSTF members at key points throughout this review, particularly when determining the scope and methods for this review and developing the Analytic Framework and KQs. After revisions reflecting the public comment period, the USPSTF members approved the final analytic framework, KQs, and inclusion and exclusion criteria. The Agency for Healthcare Research and Quality funded this review under a contract to support the work of the USPSTF. An agency Medical Officer provided project oversight, reviewed the draft report, and assisted in the external review of the report.