Objective: We conducted this systematic review to support the U.S. Preventive Services Task Force (USPSTF) in updating its 2012 recommendation on screening for and treatment of adult obesity. Our review addressed three key questions: 1) Do primary care–relevant behavioral and/or pharmacotherapy weight loss and weight loss maintenance interventions lead to improved health outcomes among adults who are overweight or have obesity and are a candidate for weight loss interventions? 2) Do primary care–relevant behavioral and/or pharmacotherapy weight loss and weight loss maintenance interventions lead to weight loss, weight loss maintenance, or a reduction in the incidence or prevalence of obesity-related conditions among adults who are overweight or have obesity and are a candidate for weight loss interventions? 3) What are the adverse effects of primary care–relevant behavioral and/or pharmacotherapy weight loss and weight loss maintenance interventions in adults who are overweight or have obesity and are a candidate for weight loss interventions?
Data Sources: We performed a search of MEDLINE, PubMed Publisher-Supplied, PsycINFO, and the Cochrane Central Registry of Controlled Trials for studies published through June 6, 2017. Studies included in the 2011 USPSTF review were re-evaluated for potential inclusion. We supplemented searches by examining reference lists from related articles and expert recommendations and searched federal and international trial registries for ongoing trials. We conducted ongoing surveillance through March 23, 2018 to identify any major studies published in the interim.
Study Selection: Two researchers reviewed 15,483 titles and abstracts and 572 full-text articles against prespecified inclusion criteria. Eligible studies were those that focused on weight loss in adults who are overweight or have obesity, or maintenance of previous weight loss. Trials among populations selected based on the presence of a chronic disease in which weight loss or maintenance is part of disease management (e.g., known cardiovascular disease, type 2 diabetes) were excluded. Studies included for health and intermediate outcomes (including weight loss) were randomized or clinically controlled trials that report data at least 12 months following the start of the intervention. In addition, for studies of potential harms of interventions we included large cohort, case-control, or event monitoring studies in addition to trials with fewer than 12 months of followup. Included interventions were those conducted in or recruited from primary care or a health care system or were judged to be feasible for implementation or referral from primary care and included behavior-based interventions as well as five U.S. Food and Drug Administration–approved medications for long-term chronic weight management (liraglutide, lorcaserin, naltrexone and bupropion, orlistat, and phentermine-topiramate). Studies of surgical and nonsurgical weight loss devices and procedures were excluded. We conducted dual, independent critical appraisal of all provisionally included studies and abstracted all important study details and results from all studies rated fair or good quality. Data were abstracted by one reviewer and confirmed by another.
Data Analysis: We synthesized data for behavior- and medication-based weight loss and weight loss maintenance interventions separately. Health outcomes and harms were sparsely reported and the specific outcomes measured differed across trials, precluding meta-analysis, so we summarized those data in tables and narratively. For weight loss outcomes related to behavior-based weight loss interventions, we ran random-effects meta-analyses using the DerSimonian and Laird method to calculate the pooled differences in mean changes (for continuous data) and pooled risk ratio (for binary data). 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. Meta-regression was used to explore potential effect modification by various study, population, and intervention characteristics. We generated funnel plots and conducted tests for small-study effects for all pooled analyses. Meta-analysis of the medication trials was not performed due to the small number of included trials and inconsistency in outcome reporting; therefore, results from these trials were summarized narratively and in illustrative forest plots. Using established methods, we assessed the strength of evidence for each question.
Results: We included 124 studies that were reported in 238 publications. We carried forward 41 studies from our previous review and 83 new studies were added. Of the 124 included studies, 89 trials focused on behavior-based weight loss (80 trials) or weight loss maintenance (nine trials) interventions. Thirty-five studies addressed medications for weight loss (32 studies) or weight loss maintenance (three trials). The majority of trials took place in the United States. Over half (73 trials) represented a general, unselected population of adults who were eligible for participation based on being overweight or having obesity; the remaining trials specifically enrolled participants who were also at elevated clinical or subclinical risk of cardiovascular disease or cancer. The mean baseline body mass index ranged from 25 to 42 kg/m2 and mean age ranged from 22 to 66 years. Eleven trials focused on specific racial/ethnic groups (African American, Asians and South Asians, American Indian, or those of Hispanic ethnicity). In the remaining trials, race/ethnicity and socioeconomic status were not well reported and when described, the majority of participants were white, with medium to high socioeconomic status.
The behavior-based interventions were highly variable across the included trials in terms of the modes of delivery, number of sessions and contacts, and interventionists. Across the 120 intervention arms, the primary mode of intervention delivery was: group based (41 arms), individual-based (37 arms), technology-based (22 arms), “mixed” (18 arms), or print only (two arms). Twenty-three interventions included interaction with a primary care provider. The 41 medication-based studies addressed: liraglutide (four trials), lorcaserin (four trials), naltrexone and bupropion (three trials), orlistat (19 trials, two observational studies), and phentermine-topiramate (three trials).
Health Outcomes: Health outcomes were minimally reported in the behavior-based weight loss and maintenance trials (k=20; n=9910). In four weight loss trials (n=4442) reporting mortality, there were no significant differences between groups over 2 to 16 years. Two weight loss trials (n=2666) reported on cardiovascular events, with neither finding differences between groups over 3 and 10 years, respectively. Health-related quality of life (QOL) was evaluated in 17 weight loss and maintenance trials (n=7120), with almost all showing no differences between groups. Trials of medication-based weight loss interventions examined few health outcomes beyond QOL (k=10; n=13,145). Although most studies showed evidence of a greater improvement in obesity-specific QOL among those on medication compared with placebo, the differences were small and of unclear clinical significance. In addition, interpretation of these finding was limited given high study dropout rates (≥35% in half the included trials). Two medication-based trials (n=6210) examined cardiovascular events, finding few events in any group. None of the medication-based maintenance trials reported the effects of the interventions on health outcomes.
Weight Outcomes: Pooled results of 67 behavior-based weight loss trials indicated greater weight loss from interventions compared to control conditions at 12 to 18 months (mean difference in weight change [MD], −2.39 kg [−5.3 lb] [95% CI, −2.86 to −1.93]; k=67; n=22,065; I2=90.0%). Mean absolute changes in weight ranged from −0.5 kg (−1.1 lb) to −9.3 kg (−20.5 lb) among intervention participants and from 1.4 kg (3.0 lb) to −5.6 (−12.3 lb) among control participants. Weight change at followup beyond 12 to 18 months was not as well reported but effects were consistent with short-term weight loss, although generally attenuated, over time. A meta-analysis of 38 trials found that intervention participants had a 1.94 times greater probability of losing 5 percent of their initial weight compared with control groups over 12 to 18 months (risk ratio [RR], 1.94 [95% CI, 1.70 to 2.22]; k=38; n=12,231; I2=67.2%), which translated into a number needed to treat of 8. Among the majority of trials of behavior-based weight loss maintenance interventions, both intervention and control participants regained weight over 12 to 18 months of maintenance; however, the intervention participants experienced less weight regain (pooled MD, −1.59 kg [−3.5 lb] [95% CI, −2.38 to −0.79]; k=8; n=1408; I2=26.8%).
Among 32 medication-based weight loss trials, those randomized to medications experienced greater weight loss compared to those on placebo at 12 to 18 months (mean/least squares mean [LSM] MD ranged from −1.0 kg [−2.2 lb] to −5.8 kg [−12.8 lb]; no meta-analysis conducted). Absolute changes in weight ranged from mean/LSM of −3.3 kg (−7.3 lb) to −10.5 kg (−23.4 lb) among medication participants compared to −0.9 kg (−2.0 lb) to −7.6 kg (−16.8 lb) among placebo participants over 12 to 18 months. Medication participants had a 1.2 to 3.9 times greater probability of losing 5 percent of their initial weight compared with placebo participants over 12 to 18 months. Three medication-based trials indicate greater weight maintenance in medication than placebo participants over 12 to 36 months (MD ranged from −0.6 to −3.5; no meta-analysis conducted).
Intermediate Outcomes: Thirteen trials (n=4095) examined incident diabetes among those in behavior-based interventions compared to control conditions. Absolute cumulative incidence of diabetes at up to 3 years of followup ranged from 0 to 15 percent in the intervention group and 0 to 29 percent in controls. The DPP and Finnish DPS trials found statistically significant lower incidences of developing diabetes at 3 to 9 years; no other trial found differences between groups. However, these trials generally had smaller sample sizes and shorter followup. The pooled relative risk of developing incident diabetes was 0.67 (95% CI, 0.51 to 0.89; k=9; n=3140; I2=49.2%). Four trials of weight loss medications (three weight loss and one maintenance trial) examined incident diabetes. Absolute cumulative incidence of diabetes at up to 4 years of followup ranged from 0 to 6 percent in medication arms and 1 to 11 percent in placebo arms; between-group differences were statistically different in most medication trials. Prevalence of hypertension, metabolic syndrome, use of CVD medications, and estimated 10-year risk of CVD were sparsely reported. There was limited evidence from larger trials that those in behavior-based weight loss arms had reduced prevalence of hypertension and use of CVD medications compared to control conditions; data were limited and mixed for metabolic syndrome and 10-year CVD risk. Four medication trials reported on use of lipid-lowering and antihypertensive medications, prevalence of metabolic syndrome, and 10-year CVD risk score with mixed results.
Adverse Events: There were no serious harms related to the behavior-based interventions and most trials noted no differences between groups in the rates of adverse events, including cardiovascular events. In the three behavior-based trials large enough to examine musculoskeletal issues between groups, results were mixed. Although serious adverse events were relatively uncommon in medication trials and generally similar between groups, adverse event rates were high in both groups by 12 months, with 80 to 96 percent experiencing an adverse event in the medication arms compared with 63 to 94 percent in the placebo arms. The higher rates of adverse events in the medication arms resulted in higher dropout rates than in the placebo arms.
Conclusion: We found that behavior-based weight-loss interventions with or without weight loss medications resulted in more weight loss than usual care conditions. The degree of weight loss we observed with the behavior-based weight loss interventions in the current review is slightly smaller but consistent in magnitude with our 2011 review on this topic. As in the previous review, we noted that weight loss interventions resulted in a decreased risk of developing diabetes, particularly among those with prediabetes, although the prevalence of other intermediate health outcomes was less well reported. Limited evidence exists regarding health outcomes associated with weight loss interventions. Weight loss medications, but not behavior-based interventions, were associated with higher rates of harms compared with control arms. Heterogeneity within each individual intervention arm confounded with differences in the populations, settings, and trial quality, making it difficult to disentangle which variables may be driving larger effects. Long-term weight and health outcomes data, as well as data on important subgroups (e.g. those who are older, nonwhite, or overweight) were lacking and should be a high priority for future study.