NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

LeBlanc EL, Patnode CD, Webber EM, et al. Behavioral and Pharmacotherapy Weight Loss Interventions to Prevent Obesity-Related Morbidity and Mortality in Adults: An Updated Systematic Review for the U.S. Preventive Services Task Force [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2018 Sep. (Evidence Synthesis, No. 168.)

Cover of Behavioral and Pharmacotherapy Weight Loss Interventions to Prevent Obesity-Related Morbidity and Mortality in Adults: An Updated Systematic Review for the U.S. Preventive Services Task Force

Behavioral and Pharmacotherapy Weight Loss Interventions to Prevent Obesity-Related Morbidity and Mortality in Adults: An Updated Systematic Review for the U.S. Preventive Services Task Force [Internet].

Show details

Chapter 3Results

Included Studies

Our literature search resulted in 15,483 unique citations. For these, we provisionally accepted 572 articles for full-text review based on titles and abstracts (Appendix C). Following review of full-text articles and critical appraisal, we included 124 trials of weight loss or weight loss maintenance interventions160, 161, 168, 172, 173, 205, 206, 213-330 reported in 238 publications (Appendix D). Of the included trials, 80 trials examined the effectiveness and/or harms of behavior-based weight loss interventions,205, 206, 214, 215, 217, 219, 221, 224, 225, 228-232, 234, 235, 237, 240, 242, 243, 245, 249-258, 261, 262, 264-267, 269-272, 274-281, 283, 286, 288-291, 293, 295, 300-302, 305, 306, 308, 310, 314-316, 318-330 and 32 examined the effectiveness and/or harms of medication for weight loss160, 161, 168, 172, 173, 213, 216, 218, 220, 222, 226, 227, 236, 238, 239, 241, 244, 246, 248, 259, 260, 263, 268, 273, 285, 292, 297-299, 304, 307, 311 (Table 2). An additional 12 trials (nine behavior-based223, 233, 282, 284, 294, 296, 303, 309, 313, 317 and three medication-based247, 287, 312) evaluated the effectiveness of a weight loss maintenance intervention. We carried forward 41 studies from our prior review and added 83 new studies (Table 2).

Table 2. Characteristics for All Trials, by Intervention Type and Author (k=124).

Table 2

Characteristics for All Trials, by Intervention Type and Author (k=124).

Of the 572 articles that were reviewed, the most common reasons for exclusion were: a lack of relevant outcomes (k=51), less than 12-month followup (for effectiveness studies) (k=59), and a lack of an appropriate comparator (comparative effectiveness, controls told specifically not to lose weight) (k=92). Appendix E contains a list of all excluded studies and their reasons for exclusion.

Given the diversity of interventions included in this review, we organized the results by: 1) behavior-based weight loss interventions (k=80), 2) behavior-based weight loss maintenance interventions (k=9), 3) medication-based weight loss interventions (k=32), and 4) medication-based weight loss maintenance interventions (k=3). Weight loss maintenance trials are those in which participant randomization occurred after weight loss (either as part of or outside of the study).

Study and Population Characteristics

Behavior-Based Weight Loss Interventions

Of the 80 behavior-based weight loss trials, 20 were carried forward from the previous review and 60 new studies were added (Table 2). All of the included studies were RCTs; 11 were cluster RCTs with randomization of health centers or primary care practices, physicians, or families.214, 225, 231, 232, 235, 269, 271, 318, 320, 329, 330 Sample sizes ranged from 30 to 2161, and the median sample size was 240. Followup at 12 months ranged from 57 to 100 percent. The majority of the trials (k=47) took place in the United States, and the remaining trials were conducted in Europe (k=15), the United Kingdom (k=11), Japan (k=3), Australia (k=2), and Canada (k=2). Recruitment varied, with at least some self-selected into the trial based on broad-based recruitment methods (35 trials), or direct recruiting through methods such as targeted mailings or appointments with their primary care providers (PCPs) (40 trials). The remaining trials applied mixed recruitment methods.

Half of the behavior-based weight loss trials represented a general, unselected population of adults who were eligible for participation based on their BMI alone (k=40) (with or without other demographic limitations [e.g., age, race/ethnicity]) (Table 3). Five additional trials specifically enrolled adults at elevated cancer risk (i.e., cancer survivors, those with colorectal adenomas).217, 229, 235, 288, 310 The remaining 35 trials selected participants based on increased subclinical (k=19) (e.g., prediabetes, family history of diabetes, high-normal blood pressure) or clinical (k=16) (e.g., hypertension, dyslipidemia) cardiovascular risk. Across all 80 trials, regardless of participant selection into the trials, cardiovascular risk status of the participants was underreported and variable among those that did report baseline prevalence. Among those that reported risk status, the proportion of affected participants varied broadly: prediabetes (8.5% to 100%; k=16), diabetes (0% to 43%; k=49), hypertension (0% to 100%; k=24), and dyslipidemia (0% to 67.7%; k=12).

Table 3. Population Characteristics for All Included Trials, Sorted by Intervention Type and Author (k=124).

Table 3

Population Characteristics for All Included Trials, Sorted by Intervention Type and Author (k=124).

The majority of trials (k=71) included adults who were overweight or had obesity, with eight trials limited to adults with obesity (i.e., BMI ≥30 kg/m2) and one trial limited to adults who were overweight (BMI 25 to 29.9 kg/m2).251 Very few trials placed an upper bound for eligible BMIs; in those that did, the upper bound ranged from 29.9 kg/m2 (in the only trial limited to adults who were overweight) to 60 kg/m2. The mean baseline BMI ranged from 25.2 kg/m2 (among a sample of Japanese adults ages 50 to 69 years) to 39.2 kg/m2 (among a sample of African American women ages 30 to 65 years) with a median of 33.4 kg/m2. The standard deviations (SDs) were large, indicating that there was a wide range of baseline BMIs and that baseline BMIs overlapped, even among trials with different BMI inclusion criteria (Figure 2). Only five trials225, 265, 270, 286, 291 included eligibility criteria based on central adiposity (i.e., waist circumference).

Figure 2 displays the distribution of mean and standard deviations of baseline BMIs in behavior-based weight loss trials. The standard deviations were large and baseline BMIs overlapped, even among trials with different BMI inclusion criteria.

Figure 2

Distribution of Baseline BMI in Behavior-Based Weight Loss Trials, by BMI Inclusion Criteria*. *Four trials not included because baseline mean BMI was not reported. Abbreviations: BMI = baseline; SD = standard deviation

The mean age of included participants ranged from 22.4 to 66.0 years (median, 50.3 years). While none of the trials restricted participation to older adults, the mean age was older than 60 years in six trials.217, 229, 235, 243, 270, 326 One trial focused on college students ages 18 to 35 years.242 Four trials were limited to men249, 272, 281, 286 and 14 were limited to women.229, 234, 235, 240, 250, 261, 269, 277, 288, 289, 295, 310, 321, 330 Of the trials restricted to women, some were further restricted to specific subgroups of women, including women with a history of breast229, 235, 288 or endometrial cancer,310 postpartum women,250, 277, 330 and African American women.240, 269 Eleven trials focused on specific racial/ethnic groups, including African Americans,240, 269 Asians257, 270, 274, 316 and South Asians,225 American Indian,276 or those of Hispanic ethnicity.278, 290, 321 There was no consistent reporting of socioeconomic status of the participants; however, based on the variables that were reported, most of the sample represented adults with medium to high socioeconomic status based on education, income, and employment.

Behavior-Based Weight Maintenance Interventions

Of the nine behavior-based weight loss maintenance trials, three were carried forward from the previous review (Table 2). The nine studies included eight RCTs and one cluster RCT (randomized based on assignment to a previous weight loss intervention).233 Sample sizes ranged from 92 to 1032 (median, 201), with a followup of 74 to 95 percent at 12 to 18 months. The majority of the trials (k=6) took place in the United States, and the remaining trials were conducted in the United Kingdom, Finland, and Australia. Recruitment procedures varied across trials, but at least some of the participants self-selected into the trials as a result of broad-based recruitment methods such as advertising with the community, health insurance, or PCP.

All but one behavior-based weight loss maintenance trial represented a general, unselected population of adults who were eligible for participation based on BMI alone (k=8), with one trial specifically enrolling adults with cardiovascular risk (i.e., hypertension and/or dyslipidemia) (Table 3).303 Six trials233, 282, 284, 303, 309, 317 conducted weight loss interventions prior to randomizing participants into the maintenance interventions, with three trials randomizing only those with at least 4 kg of weight loss.233, 282, 284 The mean BMI at enrollment in these trials was 34.2 kg/m2. The remaining three trials selected patients based on achieving a weight loss of 5 to 10 percent in the 1 to 2 years before randomization and did not include weight loss as part of the trial (mean BMI, 33.1 kg/m2).294, 296, 313

The mean age of included participants across studies ranged from 46.4 to 61.8 years (median, 49.2 years). One study examined only women,233 and one was limited to men.317 The majority of studies did not report information regarding participant race/ethnicity or socioeconomic status. Based on the limited information available, the study populations appeared to be majority white (percent of nonwhite participants ranged from 5.4% to 41.9%) with medium to high socioeconomic status based on limited data on education, income, and/or employment.

Medication-Based Weight Loss Interventions

Of the 32 medication-based weight loss studies, 16 were carried forward from the previous review (all related to orlistat) and 16 new studies were added (Table 2). Among the 32 studies, 20 RCTs were included in the review of the benefits of weight loss medications (KQ1 and KQ2), and all of the studies (30 RCTs, one retrospective cohort, and one event monitoring study) were included in the review of potential harms of medications (KQ3). Sample sizes ranged from 48 to 3731 in the RCTs (median, 542). About one-half of the trials (k=15) took place solely in the United States. The remaining trials were conducted in Europe (k=14), Australia/New Zealand (k=1), and multiple countries/regions (k=2). Fourteen studies had run-in periods to assess compliance with taking the medication. The trials that examined health outcomes (KQ1 and KQ2) lasted 12 to 48 months, with six trials contributing outcome data at 24 months or longer.160, 161, 172, 220, 241, 246, 285, 292, 297 Of the trials included for the effectiveness of weight loss, followup at 12 to 18 months ranged from 50 to 96 percent. The body of evidence regarding harms (KQ3) also included trials with shorter followup (1 to 6 months). The retrospective cohort and event monitoring study examined harms over a median of 150 days to 3 years.213, 248

The majority of studies recruiting from academic, research, or specialty care settings; recruitment procedures were not well described. Five trials specifically reported conducting at least some communitywide recruitment (via local advertising).161, 220, 239, 259, 297 One trial recruited participants from a primary health care setting,246 and two studies were conducted in a primary care setting.246, 263

Almost two-thirds of the weight loss medication trials (21 of 30 trials) were conducted in generally unselected populations based on their BMI alone. Thirteen of these trials in unselected populations required BMI to be greater than or equal to 30 kg/m2 but allowed those with BMIs greater than or equal to 27 kg/m2 if cardiovascular risk factors were present.160, 161, 168, 172, 173, 216, 218, 220, 239, 244, 246, 285, 311 One trial selected participants based on the presence of prediabetes,259 and eight trials selected those with one or more cardiovascular risk factors (e.g., hypertension, dyslipidemia).222, 226, 227, 236, 241, 263, 273, 304 Across all 30 trials, cardiovascular risk status of the participants was underreported. In those reporting risk status, 1 to 68 percent had prediabetes (k=8), 0 to 27 percent had type 2 diabetes (k=21), 0 to 100 percent had hypertension (k=11), and 21 to 100 percent had dyslipidemia (k=13). Overall, approximately two-thirds of the trials (k=19) included adults who were overweight or had obesity, with 11 trials limited to adults with obesity (BMI ≥30 kg/m2). The mean baseline BMI ranged from 31 to 42 kg/m2 (median, 36.1 kg/m2). The large SDs indicate a wide range of baseline BMIs that overlapped among trials with different BMI inclusion criteria (Figure 3). No trial had eligibility criteria based on central adiposity (i.e., waist circumference).

Figure 3 displays the distribution of mean and standard deviations of baseline BMIs in medication-based weight loss trials. The standard deviations were large and baseline BMIs overlapped, even among trials with different BMI inclusion criteria.

Figure 3

Distribution of Baseline BMI in Medication-Based Weight Loss Trials, by BMI Inclusion Criteria. Abbreviations: BMI = baseline; Nal/Bup = Naltrexone HCL and bupropion HCL; Phen/Top = Phentermine-topiramate extended release; SD = standard deviation

The mean age of included participants ranged from 41 to 58 years (median, 45 years). All studies comprised both men and women (25% to 90% female), with all but one study299 including more females than males. Of the 18 trials that reported race/ethnicity of the sample, the percent of nonwhite participants ranged from 5 to 37 percent. No trials focused on specific racial/ethnic groups and there was no reporting of socioeconomic status. Baseline characteristics were similar in the two non-RCT studies.213, 248

Medication-Based Weight Maintenance Interventions

Two medication-based weight loss maintenance trials were carried forward from the previous review,247, 287 and one new trial was added (Table 2).312 All were RCTs conducted in research clinics and were set in the United States, Canada, and Scandinavia. Sample sizes ranged from 309 to 542, with a followup of 65 to 74 percent at 12 to 36 months. All three trials began with an active weight loss phase, which lasted for 4 to 24 weeks, during which all participants were prescribed hypocaloric diets and exercise with no pharmacologic intervention. Participants were required to lose 5 to 8 percent of their baseline weight prior to randomization to the maintenance intervention. One trial was limited to adults with at least one cardiovascular risk factor. The mean baseline BMI at enrollment into the maintenance phase ranged from 32.8 to 37.5 kg/m2.287 The mean age of included participants was 46 to 47 years, with the majority of the participants being female (51% to 84%) and white (12% to 16% nonwhite). There was no reporting of socioeconomic status.

Intervention Characteristics

Behavior-Based Weight Loss Interventions

Within the 80 weight loss trials, 105 unique weight loss interventions were evaluated against control conditions (Table 4). The interventions were highly variable across the included trials in terms of the modes of delivery, number of sessions and contacts, and interventionists. However, specific weight loss messages and behavior change techniques were consistent across the trials (Table 5; Appendix F Table 1). Duration of interventions ranged from 3 months (in six trials) to 5 years (in one trial), with the majority taking place for a minimum of 1 year. One-third of interventions provided a “core” intervention period (described as “core,” “active,” or “intense” phases) generally for 3 months to 1 year and then followed up with a support phase (also described as “maintenance” in some trials), generally for 9 to 12 months. The remaining interventions did not distinguish between “core” and “support” phases.

Table 4. Behavior-Based Weight Loss Intervention Characteristics (k=80, 105 intervention arms).

Table 4

Behavior-Based Weight Loss Intervention Characteristics (k=80, 105 intervention arms).

Table 5. Behavior-Based Weight Loss Intervention Components.

Table 5

Behavior-Based Weight Loss Intervention Components.

To better summarize the interventions, we categorized each intervention arm according to the main mode of intervention delivery into the following groups: 1) group, 2) individual, 3) mixed, 4) technology-based, and 5) print-based (Table 4). Across the 105 intervention arms, one-third (35 arms in 25 trials) were primarily group-based counseling interventions.214, 215, 221, 230, 243, 249, 253-255, 261, 262, 266, 267, 270, 276, 280, 295, 300, 301, 314, 316, 321, 323, 327, 329 Group-based interventions ranged from eight group sessions over 2.5 months to 52 weekly group sessions over 1 year (median, 23 total sessions in the first year). Twelve interventions (in seven trials254, 261, 300, 301, 314, 327, 329) provided group sessions beyond 1 year (1.5 to 3 years total intervention time). Groups typically consisted of classroom-style sessions with 8 to 12 participants per group, and each session lasted 1 to 2 hours. Within the group-based interventions, five trials (nine arms) provided referral and free access to commercially available group-based weight loss programs including Weight Watchers,253, 255, 267, 323 Slimming World221, 255 and the Size Down program255 (both provided by the U.K. National Health Service), and Rosemary Conley (U.K.-based weight loss program).255 Six of the group-based interventions offered minimal supplemental support, with one brief individual counseling session.214, 221, 300, 301, 327, 329

In 30 trials (with 33 arms), the main mode of intervention delivery was individual-based support.205, 206, 217, 219, 224, 225, 229, 232, 234, 237, 250, 252, 255, 257, 265, 269, 271, 275, 283, 286, 289, 291, 305, 306, 308, 318, 320, 324, 325, 328 In most of these (24 arms), counseling was provided through face-to-face intervention sessions with or without ongoing telephone support. The remaining nine individual-based interventions were provided remotely through telephone counseling calls (average 15 to 30 minutes) and Web-based self-monitoring and support. One trial evaluated three individual-based strategies that included telephone support only, a mailed food basket only, and telephone support plus the mailed food basket.320 In general, the individual-based counseling interventions had fewer sessions or contacts than the group-based interventions; the median number of sessions in the first year for individual-based interventions was 12 compared with 23 in group-based interventions. DPP was the most intense individual-based intervention, offering participants weekly and then bimonthly individual counseling sessions with case managers over 3 years.205 Another example of an intense individual intervention was one that offered free access to weekly individual counseling sessions through Jenny Craig.289

We categorized 17 interventions (within 15 trials219, 228, 240, 251, 256, 258, 274, 278, 288, 290, 302, 310, 315, 326, 330) as “mixed” interventions as they included relatively equal numbers of group- and individual-based counseling sessions with or without other forms of support (telephone, print, Web-based). All but four of these interventions took place for more than 1 year, and most had more than 12 sessions in the first year (median number of sessions [23] was same as median for group-based interventions).

In another 18 interventions (16 trials), the main component of the intervention was technology-based, including computer- or Web-based intervention modules,231, 264, 272, 277, 279, 281, 308, 315 Web-based self-monitoring, mobile phone-based text messages, smartphone applications, or social networking platforms,242, 245, 265, 293, 302, 319, 322 or DVD learning.266 In all but two of these interventions,231, 264 there was no face-to-face interaction with an interventionist. There was only one trial (two arms) that delivered its intervention entirely through print-based tailored materials.235

Across all intervention types, 19 interventions included interaction with a PCP, although the level of interaction with the provider was variable across the interventions.206, 219, 221, 224, 231, 232, 237, 255, 257, 269, 271, 305, 310, 318, 324, 327, 328 In three of these interventions, PCP involvement was limited to encouragement to take part in and/or referral to interventions conducted by other providers (i.e., group-based interventions conducted by lifestyle coaches or registered dietitians) or in other settings (i.e., commercial weight loss program).219, 221, 224 In seven trials, PCPs reinforced intervention messages through brief counseling sessions.206, 231, 237, 310, 324, 327, 328 A PCP was the primary interventionist in only six interventions, providing 3 to 12 months of individual counseling.232, 255, 257, 269, 305, 318 The intervention providers were highly diverse in the remaining trials not involving PCPs and included behavioral therapists, psychologists, registered dietitians, exercise physiologists, lifestyle coaches, and other study-hired medical or public health staff. Most trials included interventionist training prior to the start of the intervention; in those that gave specific details (k=43), training was fairly intense, ranging from 2 hours to 4 days as well as regular check-ins or supervised sessions to ensure fidelity to the intervention protocol. In one trial,271 the focus of the intervention was to educate PCPs on the benefits of weight loss and effective treatment options through small group meetings. Each practice was then asked to devise an individual weight management protocol for their patients who had obesity to help them achieve 10 percent loss of their body weight.

The trials had very similar messages in terms of specific weight loss and behavioral goals. Most of the interventions were designed to help participants achieve a 5 percent or greater weight loss through a combination of dietary changes (including specific caloric goals) and a gradual increase in physical activity (generally promoting at least 150 minutes of moderate-intensity activity per week). A few trials mentioned promoting specific dietary approaches including the DASH diet (Dietary Approaches to Stop Hypertension),228, 302, 320 a Mediterranean food pattern,291 or the Magedeburg Dual Diet (500 kcal/day reduction and consumption of low-glycemic index foods).265 Only one trial314 encouraged a very low-calorie diet (800 to 1000 kcal/day). In two trials,254, 289 prepackaged meals were provided directly to participants. In one trial,256 the intervention was exclusively focused on dietary changes and participants were specifically told not to exercise.

In addition to group, individual, and technology-based education and counseling, most interventions provided additional tools to assist with weight loss (e.g., pedometers, food scales, exercise videos). One intervention provided monetary incentives for weight loss.254 Most of the trials targeted individual participants, but a few encouraged participants to invite family members to join intervention activities228, 243, 300, 301, 306, 316, 320 and two specifically targeted family pairs or units (i.e., mother-daughter pairs235 or adult relatives225).

Twelve trials (14 intervention arms) provided interventions modeled closely after the DPP lifestyle intervention for application in the community214, 215, 230, 258, 277, 278, 290, 316, 319, 321 or primary care.266, 328 A number of these trials tailored the DPP intervention for a specific population (e.g., Latinos, postpartum women with recent gestational diabetes) and included additional intervention components such as individual counseling sessions with community health workers. All but one328 of these trials were among adults at increased diabetes (9 trials) or cardiovascular risk (2 trials). Three of the trials adapted the DPP core curriculum to be provided strictly by DVD,266 text messages,319 or a Web site.277 Three additional trials described using or adapting DPP materials as part of their interventions but did not closely follow the DPP framework and were conducted among unselected adults or those at low cardiovascular and diabetes risk.237, 261, 305

In general, rates of participation or participant adherence were relatively high (Appendix F Table 1). Most of the studies reported that more than two-thirds of the intervention participants completed the full intervention, or alternatively, that all participants completed more than two-thirds of the intervention. However, participation rates appeared to decline over time, especially as intervention intensity lessened. This pattern held true even among interventions that were primarily technology-based.

The majority of trials employed a minimal weight loss intervention (k=41) or usual care (k=23) arm for the control group (Appendix F Table 1). Most of the minimal intervention and usual care groups consisted of generic self-help print or Web-based materials focused on weight loss, diet, and physical activity changes, and diabetes prevention (e.g., the NHBLI’s “Aim for a Healthy Weight” brochure). A handful, however, were more intense and included 30 minutes to 2 hours of nontailored group weight-loss counseling sessions, brief (2 to 3 minutes) quarterly counseling sessions with a PCP, or more intense individualized counseling two to four times per year.205, 225, 237, 265, 267, 272, 288, 289, 305, 306, 324, 326, 328, 329

Behavior-Based Weight Maintenance Interventions

Within the nine behavior-based maintenance trials,233, 282, 284, 294, 296, 303, 309, 313, 317 there were 15 unique weight loss interventions evaluated against control conditions; four trials had more than one active intervention arm compared with a control condition (Table 6; Appendix F Table 2). The maintenance interventions included group interventions (six arms),282, 284, 313 technology-based (four arms),233, 303, 313, 317 individual counseling sessions conducted in person or by phone (four arms),294, 296, 303 or a combination of individual and group counseling (one arm)309 (Table 7). Duration of the maintenance interventions ranged from 6 months to 5 years, with the majority taking place for 12 to 18 months. The number of sessions within the first year ranged from 0 (Web-based self-monitoring only) to 26, with the majority of the interventions having greater than 12 sessions in the first year. Only one study, which included four intervention arms, specifically reported that the intervention included interaction with a PCP (physician or nurse) paired with a clinical psychologist.284

Table 6. Behavior-Based Weight Loss Maintenance Intervention Characteristics (k=9, 15 intervention arms).

Table 6

Behavior-Based Weight Loss Maintenance Intervention Characteristics (k=9, 15 intervention arms).

Table 7. Behavior-Based Weight Loss Maintenance Intervention Components.

Table 7

Behavior-Based Weight Loss Maintenance Intervention Components.

The interventions were designed to help participants maintain weight loss by continuing dietary changes and physical activity. There was a focus on reviewing nutrition, exercise, and behavioral topics as well as self-monitoring, identifying barriers, problem-solving, peer support, and relapse prevention. Programs also provided participants with tools to assist in weight loss maintenance (e.g., food diaries, pedometers) and one intervention included monetary incentives for program adherence.284

In most trials, the majority of the sessions were attended or contacts completed during the first 6 to 12 months (Appendix F Table 2). However, similar to the trials of behavior-based weight loss interventions, participation began to drop off, especially beyond 12 months.

Following an administered weight loss intervention233, 282, 284, 303, 309, 317 or after study enrollment,294, 296, 313 the control groups received either no intervention (k=4),233, 282, 284, 317 minimal intervention (e.g., generic self-help print or Web-based materials or minimal phone contact) (k=3),294, 303, 313 or usual care (e.g., care offered as part of health plan enrollment) (k=2).296, 309

Medication-Based Weight Loss Interventions

All of the medication-based weight loss studies examined FDA-approved dosages of medications (Table 8): liraglutide at 1.8 mg QD or 3.0 mg QD, lorcaserin at 20 mg (10 mg BID), naltrexone and bupropion at 32/360 mg (16/180 mg TID), orlistat at the prescription strength dosage of 360 mg daily (120 mg TID) and over-the-counter dosage of 180 mg (60 mg TID), and phentermine-topiramate at 15/92 mg and 7.5/46 mg. We did not abstract data on nonapproved dosages.

Table 8. Medication-Based Intervention Characteristics (k= 35, 41 arms).

Table 8

Medication-Based Intervention Characteristics (k= 35, 41 arms).

Within all trials, both groups received identical behavioral interventions. Participants were told to follow energy-restricted diets (generally with a 500- to 800-kcal/day deficit) and increase physical activity in addition to taking the medication. The extent of the behavior-based component of the intervention varied widely among studies—from a single visit with a study physician246 to weekly, 90-minute group sessions.311 The most common behavior-based intervention was to require participants to complete food records that were discussed with nutritionists at study visits (which ranged from monthly to quarterly).

Medication adherence was rarely reported; however, almost all trials reported the percentage of participants who completed the trial on the drug/placebo. Completion rates ranged from 10 to 93 percent, with most studies having completion rates between 50 and 70 percent. Of note, completion rates tended to be higher among the intervention groups than control groups.

Medication-Based Weight Maintenance Interventions

Three trials examined the effect of medication on weight loss maintenance following a weight loss intervention. Two medication-based weight loss maintenance studies examined orlistat, one at the prescription strength dosage of 360 mg daily (120 mg TID)287 and one at both prescription and over-the-counter strengths (180 mg [60 mg TID]),247 and one study examined liraglutide 3.0 mg QD (Table 8).312 During the maintenance phase, participants were prescribed energy intakes to either maintain weight or result in a 500- to 600-kcal/day energy restriction, were encouraged to exercise regularly, and met with dieticians or behavioral counselors. While no trials reported on pill compliance, the percentage of participants who completed the trial on the drug/placebo ranged from 70 to 77 percent in the two trials that reported these data.

Study Quality

Within the 89 included behavior-based weight loss and weight loss maintenance trials, we rated 26 as good quality and the remaining 63 as fair quality (Table 2). One study included in KQ3 (harms) had intermediate health outcome data, but these data were not evaluated as the study was rated poor quality for KQ1 and KQ2 because of greater than 20 percent differential attrition.320 In general, the 26 good-quality trials were characterized by valid randomization procedures, comparable groups at baseline (or adequate adjustment for known baseline differences in the analysis), high sample retention (i.e., ≥85% retention at 12 months), the use of reliable and valid measurement instruments applied equally across arms, evidence of fidelity to the intervention protocol, no evidence of selective outcome or analysis reporting, and appropriate analyses, including intention-to-treat principles using multiple imputation or other conservative data imputation procedures for missing data (e.g., baseline observation carried forward). Most of the trials rated as good quality included published design papers or protocols with extensive details on their randomization methods, procedures for maintaining fidelity to the intervention, and data analysis plans. Additionally, several of the good-quality trials were multisite trials with data coordinating centers, including the three POWER trials,206, 219, 224, 331 TOHP phases I and II,300, 301 the TONE trial,326 and DPP.205 Some common limitations of the fair-quality studies included lack of reporting details about allocation concealment, relatively higher (i.e., >20%) and differential attrition between groups, and no attempt (or lack of reporting) to account for missing data or only completers-only analyses. The main risks of bias for the 12 behavior-based intervention studies we rated as poor quality included differential attrition between intervention arms (approximately 10% to 30% differential attrition) with completers-only analyses or unclear methods for handling missing data coupled with other issues in trial conduct, analysis, or reporting of results (e.g., intervention fidelity, possible selective reporting, inappropriate exclusion of participants from analyses, questionable validity of randomization and allocation concealment procedures). In addition, three of the studies excluded for poor quality used different procedures for measuring participants’ weight at baseline and followup. In these trials, baseline weight was objectively measured using standard protocols, whereas weight at followup was self-reported by participants for the full or partial sample and the percent of self-reported weights was not reported by treatment group.

All 35 of the medication trials were rated as fair quality; none were rated as good quality (Table 2). One study included in KQ3 (harms) had intermediate health outcome data, but these data were not evaluated as the study was rated poor quality for KQ1 and KQ2 because of greater than 20 percent differential attrition early on in the study with limited data substitution methods.222 In addition, one study only eligible for inclusion for KQ3 (harms) was excluded for poor quality due to incomplete description of the collection and reporting of adverse events.332 The biggest threat to internal validity within this body of evidence is high rates of attrition and missing data, which is a substantial and frequent issue in weight loss medication trials.201 Because most dropouts are due to adverse events or lack of effectiveness of the intervention and not study design flaws, we rated studies with high attrition as fair quality if they used adequate data substitution methods with sensitivity analyses evaluating various substitution methodologies. A study evaluating data substitution methods in obesity medication trials concluded that data substitution methods were generally adequate for protecting against false-positive and false-negative results in the majority of medication weight loss trials.201

KQ1. 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?

Summary of Results

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. Health-related QOL was evaluated in 17 weight loss and maintenance trials (n=7120), with almost all showing no differences between groups.

Trials of medications for weight loss examined few health outcomes beyond QOL (k=10; n=13,145). Although there was evidence of greater improvement on an obesity-specific QOL scale in those randomized to medications for weight loss compared with placebo within most of the trials, the differences were small and of unclear clinical significance, especially given high dropout rates in medication trials. None of the medication-based maintenance trials reported the effects of the interventions on health outcomes.

Detailed Results

Behavior-Based Weight Loss Interventions

Eighteen trials reported the effects of behavior-based weight-loss interventions on at least one health outcome (n=9543);205, 206, 219, 234, 235, 243, 249, 252, 262, 275, 278, 288, 301, 306, 310, 315, 323, 326 we rated half of these trials as good quality. Thirteen of the 18 trials were newly identified as part of our update; the remaining 5 trials–which included the DPP, the Finnish DPS, PREDIAS, TOHP Phase II, and TONE–were included in our previous review.

All-Cause Mortality

Four good-quality trials (n=4442), all included in our previous review, reported the effect of the intervention on all-cause mortality.205, 301, 306, 326, 333-335 Overall, few deaths occurred in the three trials of adults (ages 25 to 65 years). One trial in older hypertensive adults (ages 60 to 80 years) found higher overall rates of death in both arms; however, none of the four trials found significant between-group differences in mortality over 2 to 16 years of followup. After approximately 4.5 years of followup, DPP (n=2161) found that the placebo group had a nonsignificant higher mortality rate per 100 person-years compared with the lifestyle intervention group (0.2 vs. 0.1).205 In Phase II of TOHP, a hypertension prevention trial (n=1191), 5 versus 2 participants in the intervention and control groups died, respectively, over 2 years.301, 333 The Finnish DPS (n=505) found no significant difference in all-cause mortality after 10.2 years of followup with 6 versus 10 deaths in the intervention versus control groups (hazard ratio [HR], 0.57 [95% CI, 0.21 to 1.58]).306, 334 Finally, TONE (n=585), a study in hypertensive adults ages 60 to 80 years, found no significant difference in the all-cause mortality after 16 years of follow up (HR, 0.82 [95% CI, 0.55 to 1.22]).335

CVD

The DPP205, 336 and the Finnish DPS306, 334 trials (n=2666) reported on the incidence of cardiovascular events over the course of the study, including stroke or myocardial infarction. There was no statistically significant difference between groups on the number of participants in the intervention groups of DPP and DPS who experienced cardiovascular events compared with control participants after 3 and 10 years of followup, respectively.205, 306 Within DPP, nonfatal cardiovascular events occurred in 2.2 percent of lifestyle intervention participants (9.7 events/1000 patient-years) (n=1079) compared with 1.7 percent of placebo participants (7.3 events/1000 patient-years) (n=1082), which was not statistically different.336 Cardiovascular-related deaths occurred in only 2 and 4 participants in the lifestyle and placebo groups, respectively.336 In the Finnish DPS trial, after 10.2 years, there were 57 new cardiovascular events (22.9 per 1000 person-years) in the intervention group and 54 events (22.0 per 1000 person-years) in the control group (HR, 1.04 [95% CI, 0.72 to 1.51]).334

Health-Related QOL and Depression

Fifteen trials (n=6893 examined health-related QOL outcomes) (Table 9).205, 206, 219, 234, 235, 243, 249, 252, 262, 275, 278, 288, 310, 315, 323 The data were limited in that only six trials presented absolute changes in QOL scores; the remaining just reported whether or not there were significant differences between groups in QOL outcomes. Three of 14 trials found statistically significantly greater improvement on the physical component summary score (but not the mental component summary score) after 1 to 3 years of followup among intervention participants versus control participants (absolute between-group differences ranging from 1.5 to 2.5 points on a 100-point scale).205, 215, 249 There were no other significant differences between groups on other measures of QOL.

Table 9. Results of Behavior-Based Weight Loss and Weight Loss Maintenance Interventions on Health-Related Quality of Life (k=17) (n=7120).

Table 9

Results of Behavior-Based Weight Loss and Weight Loss Maintenance Interventions on Health-Related Quality of Life (k=17) (n=7120).

None of the included trials reported the effect of the intervention on the incidence or prevalence of depression over the course of the study. Two trials (DPP and the Finnish DPS) reported the prevalence of participants on antidepressant medications after 3 to 4 years of followup and found no significant differences across treatment arms.337, 338

Behavior-Based Weight Loss Maintenance Interventions

The only health outcome reported in behavior-based weight loss maintenance interventions was QOL, which was evaluated in two trials; both found no significant effects after 1 to 2 years of followup (Table 9).282, 296

Medication-Based Weight Loss Interventions

Ten of 32 medication-based weight loss trials reported the effects of the intervention on health outcomes, including QOL (10 trials) and cardiovascular events (two trials).172, 173, 218, 220, 241, 244, 285, 292, 304, 311 In general, findings related to health outcomes were limited by reduced long-term followup, with many trials reporting rates of 35 to 55 percent loss to followup by 12 to 24 months (Table 2, Table 8).

CVD

Liraglutide. Within one trial (n=3723), there were three cardiovascular events in both the liraglutide and placebo arms (0.12% and 0.24%, respectively) after 13 months (statistical testing not reported).285 Participants with prediabetes at baseline (n=2201) were followed for an additional 23 months (total of 36 months), with an additional two cardiovascular events in those randomized to liraglutide and none in the placebo arm (statistical testing not reported).339

Phentermine and topiramate. One trial of phentermine and topiramate (n=2487) reported similarly low rates of cardiovascular events in the intervention and placebo arms (0.4%, 0.6%, and 0.7% in the 15/92 mg, 7.5/46 mg, and placebo arms, respectively) across 13 months followup (statistical testing not reported).241

QOL

Liraglutide. Two liraglutide trials examined changes in QOL (Table 10).220, 285 The smaller trial (n=196) reported QOL improvement in both arms during the first 12 months, without between-group statistical testing.220 The larger trial (n=3662) found significant improvements in QOL in those randomized to liraglutide versus placebo at 13 months (absolute between-group differences ranging from 0.9 to 3.1 points on a 100-point scale). 285 Among a subset of participants with prediabetes, there were mixed results in QOL changes at 36 months.285, 339

Table 10. Results of Medication-Based Weight Loss and Weight Loss Maintenance Interventions on Health-Related Quality of Life (k=10) (n=13145).

Table 10

Results of Medication-Based Weight Loss and Weight Loss Maintenance Interventions on Health-Related Quality of Life (k=10) (n=13145).

Lorcaserin. Two trials of lorcaserin (n=6139) examined changes in QOL at 12 months, both finding172, 173 that greater improvements in QOL were seen in those randomized to lorcaserin compared with those in the placebo arm (absolute between-group differences not reported; p<0.001) (Table 10).172, 173

Naltrexone and bupropion. Three trials (n=2815) of naltrexone and bupropion examined QOL after 12 to 13 months (Table 10).218, 244, 311 All trials reported that QOL improved more in those who received naltrexone and bupropion compared with those who received placebo (absolute between-group differences not reported; p<0.001).

Orlistat. Changes in QOL were evaluated in two orlistat trials (Table 10).292, 304 One (n=333) noted a statistically significant higher score on one QOL subscale in those on orlistat compared with placebo after 12 months; however, there were no other significant differences.304 Another trial (n=481) found those randomized to orlistat for 24 months had statistically significant greater satisfaction with their medication and overall therapy, and less overweight distress.292

Phentermine and topiramate. One trial (n=2487) identified significantly greater improvements in QOL with 15/92 mg phentermine and topiramate compared with placebo (data not reported) (Table 10).241

Medication-Based Weight Loss Maintenance Interventions

None of the three trials examining the effect of medications for weight loss maintenance reported the effects of the intervention on health outcomes.

KQ2. 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?

Summary of Results

Participants who received behavior-based weight loss interventions generally lost more weight and had greater reductions in waist circumference than those in control conditions at up to 24 months followup. Intervention participants had a pooled −2.4 kg (95% CI, −2.85 to −1.92) greater weight loss at 12 to 18 months. 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. In addition, intervention participants had a 1.94 (95% CI, 1.70 to 2.22) times greater chance of losing 5 percent weight, which translated into a number needed to treat (NNT) of 8. Although weight outcomes were less well-reported beyond 12 months, weight loss remained significantly greater in intervention compared with control conditions in interventions lasting up to 36 months. Participants who received behavior-based weight loss maintenance interventions generally maintained more of their weight loss compared with those in control conditions. The heterogeneity in each individual intervention arm and differences in the populations, settings, and trial quality made it difficult to disentangle what variables might be driving larger effects.

In the two largest and longest good-quality trials (n=1818), participants randomized to behavior-based weight loss interventions had a decreased probability of developing type 2 diabetes compared with control conditions over 3 to 9 years. Although 11 smaller and generally shorter-duration weight loss trials did not find significant differences between groups, when pooled with the larger trials, there was a significant 33% reduction in risk of developing diabetes over 1 to 9 years (pooled risk ratio [RR], 0.67 [95% CI, 0.51 to 0.89]; k=9; n=3140; I2=49.2%). Three large trials (n=3916) noted benefits of behavior-based weight loss on hypertension and hyperlipidemia diagnosis and/or medication use; however, effects were not found in five smaller trials. Effects on metabolic syndrome and CVD risk score were mixed.

Participants randomized to weight loss medications had more weight loss and a greater decrease in waist circumference than those on placebo. Participants who received medications to assist with weight loss maintenance generally maintained more of their weight loss and waist circumference decrease compared with those in control conditions. However, the results were limited by high dropout rates and relatively short followup duration in some trials. The most common intermediate outcome reported (k=4; n=9763) was incident diabetes, and there was a decreased risk of developing diabetes over 1 to 4 years in those given medications; however, these trials were similarly limited by high dropout rates. Other intermediate outcomes were sparsely reported with mixed results.

Detailed Results

Behavior-Based Weight Loss Interventions

Weight Loss

All of the included trials reported treatment effects on at least one measure related to weight change (i.e., weight change in kilograms [kg] or pounds [lb], percent weight loss, BMI, waist circumference, or the proportion of participants losing 5%, 7%, 10%, or 15% of their weight from baseline). All weight-related outcomes for all time points and all arms for all trials are reported in Appendix G Table 1 for continuous outcomes and Appendix G Table 2 for dichotomous outcomes. Table 11 summarizes the results for all pooled analyses.

Table 11. Pooled Results of Weight Loss Outcomes for Behavior-Based Weight Loss Interventions.

Table 11

Pooled Results of Weight Loss Outcomes for Behavior-Based Weight Loss Interventions.

A meta-analysis combining the 67 behavior-based weight loss trials that reported kilograms or pounds lost at 12 to 18 months found a pooled mean difference of −2.4 kg (−5.3 lb) more lost in the intervention versus control groups (mean difference [MD], −2.39 kg [95% CI, −2.86 to −1.93]; k=67; n=22,065; I2=90.0%) (Figure 4). Although not all trials found statistically significant results, in all but two cases, intervention participants showed greater reductions in weight than control participants. 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 (3.1 lb) to −5.6 (−12.3 lb) among control participants at 12 to 18 months. Across the trials, however, a wide range of effects was seen within all arms (intervention and control) as demonstrated by large SDs relative to the average change. In other words, some adults showed fairly large reductions in weight, some showed no or modest changes, and some gained weight. All but nine221, 231, 243, 255, 264, 269, 272, 275, 318 of the trials that reported weight change at 12 to 18 months had interventions that spanned at least 12 months. Within the eight trials that had interventions less than 12 months long (i.e., 3 to 9 months), only one reported a statistically significant difference in weight loss at 12 months.221

Figure 4 displays a forest plot of the effect of 67 behavior-based trials (n=22,065). The pooled mean difference of change in weight loss was 2.39 kg (95% CI, -2.86 to -1.93) in favor of the intervention group vs. control group at 12 to 18 months.

Figure 4

Pooled Analysis of Change in Weight (kg) at 12 to 18 Months in Behavior-Based Weight Loss Interventions Compared With Controls. Abbreviations: BL = baseline; Cancer = elevated cancer risk; CG = control group; CI = confidence interval; CV = increased cardiovascular (more...)

Weight change at followup beyond 12 to 18 months was not as well reported. The pooled MD in weight change at 24 months was −1.45 kg (−3.2 lb) in favor of the intervention versus control groups [95% CI, −2.03 to −0.87]; k=21; n= 7268; I2=67.9%) (Figure 5). Absolute changes in weight ranged from a 1.0 kg (2.2 lb) to −5.6 kg (−12.3 lb) among intervention participants and from 0.3 kg (0.7 lb) to −4.0 kg (−8.8 lb) among control participants. Absolute differences between groups ranged from 0.75 kg in favor of the control group to −4.78 kg in favor of the intervention group. Only eight trials reported weight change at greater than 24 months, with most reporting outcomes at 2.5 to 4 years;225, 234, 254, 256, 261, 301, 306, 326 one reported effects of the intervention at both 2.5 and 6.6 years, over 4 years after the intervention ended234).

Figure 5 displays a forest plot of the effect of 21 behavior-based weight loss trials (n=7268). The pooled mean difference of change in weight was 1.45 kg (95% CI, -2.03 to -0.87) in favor of the intervention group vs. control group at 24 months.

Figure 5

Pooled Analysis of Change in Weight at 24 Months in Behavior-Based Weight Loss Interventions Compared With Controls. Abbreviations: BL = baseline; Cancer = elevated cancer risk; CG = control group; CI = confidence interval; CV = increased cardiovascular (more...)

Twenty-eight trials reported effects of the interventions on weight change over time.206, 219, 224, 225, 234, 237, 242, 251, 254, 256, 258, 261, 266, 269, 271, 286, 288, 290, 291, 301, 302, 306, 314, 318, 323-326 Ten trials showed consistent although attenuated, statistically significant benefit of the interventions over time (from 12 to 48 months followup),219, 224, 258, 261, 266, 288, 301, 306, 323, 325 whereas 10 trials showed consistent null effects over time.206, 237, 251, 254, 256, 269, 271, 290, 318 Six trials reported initial statistically significant benefit of the interventions at 12 to 18 months, with attenuation of effects over time such that effects were no longer statistically significant at 18 to 80 months.234, 242, 286, 291, 314, 324 One trial326 reported a consistent (not attenuated) benefit from a 28-month intervention at 12, 18, and 30 months. One trial225 reported no benefit from the 3-year intervention at 12 and 24 months but found a statistically significant greater weight loss at 3 years. A forest plot showing all of the trials that reported weight change over time (i.e., more than one time point), without pooling, is included to visualize the change in effects over time within each trial (Figure 6). Within the 10 trials with a lag time between intervention end and final followup (lag of 2 to 50 months),234, 254, 261, 266, 269, 271, 286, 318, 323, 326 four reported statistically significant differences in weight loss at the final time point.261, 266, 323, 326

Figure 6 displays a forest plot of the effect of all behavior-based weight loss trials that reported weight change over time. Without pooling, this figure visualizes the change in effects over time within each trial.

Figure 6

Change in Weight in Behavior-Based Weight Loss Interventions Among Trials With Multiple Followup Time Points. Abbreviations: BL = baseline; CG = control group; CI = confidence interval; IG = intervention group; Intv = intervention; MD = mean difference; (more...)

Nine trials that reported a weight outcome could not be included in the meta-analyses for weight change at 12 to 18 months or 24 months because of limitations in data reporting (e.g., no measure of dispersion) (Appendix G Table 1).214, 230, 254, 256, 257, 276, 278, 280, 295 Most of these were relatively small trials with sample sizes ranging from 50 to 280. Within all of these trials, intervention group participants experienced greater mean or median weight loss than control group participants, but only three trials reported these differences as statistically significant at 12 to 18 months.278, 280, 295

Separate meta-analyses for between-group mean differences in percent weight change and BMI at 12 to 18 months followup also showed statistically significant associations with weight loss interventions (Table 11).

There was no evidence of small-study effects for weight loss based on the Egger’s test.

Weight Loss of 5 Percent or Greater

Forty-five of the 80 trials reported the proportion of participants losing at least 5 percent of their baseline weight at 12 months or more followup (Appendix G Table 2). A meta-analysis of 38 trials reported 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 (RR, 1.94 [95% CI, 1.70 to 2.22], k=38; n=12,231, I2=67.2%) (Figure 7). Based on an assumed control risk of 14 percent, the NNT to achieve one more adult losing at least 5 percent of their body weight over 12 to 18 months is 8 (NNT, 7.6). At 24 months, the pooled RR was attenuated but still suggested an association between behavior-based weight loss interventions and the proportion of participants losing at least 5 percent of their baseline weight (pooled RR, 1.51 [95% CI, 1.25 to 1.81], k=13; n=4824, I2=63.0%) (Figure 8). There was no evidence of small-study effects for the proportion losing at least 5 percent of their body weight based on the Peters’ test.

Figure 7 displays a forest plot of the effect of 38 behavior-based weight loss trials’ proportion of participants losing at least 5% of their baseline weight (n=12,231). The pooled risk ratio is 1.94 (95% CI, 1.70 to 2.22) times greater probability for intervention participants to lose 5% of their initial weight compared with the control group at 12 to 18 months.

Figure 7

Pooled Analysis of Risk of Losing ≥5% of Body Weight at 12 to 18 Months in Behavior-Based Weight Loss Interventions Compared With Controls. Abbreviations: Cancer = elevated cancer risk; CG = control group; CI = confidence interval; CV = increased (more...)

Figure 8 displays a forest plot of the effect of 13 behavior-based weight loss trials’ proportion of participants losing at least 5% of their baseline weight (n=4824). The pooled risk ratio is 1.51 (95% CI, 1.25 to 1.81) times greater probability for intervention participants to lose 5% of their initial weight compared with the control group at 24 months.

Figure 8

Pooled Analysis of Risk of Losing ≥5% of Body Weight at 24 Months in Behavior-Based Weight Loss Interventions Compared With Controls. Abbreviations: Cancer = elevated cancer risk; CG = control group; CI = confidence interval; CV = increased cardiovascular (more...)

Fewer trials reported the percent of participants losing 10 percent or more of their body weight. Meta-analyses found that intervention participants were 3.1 times more likely to lose 10 percent of their weight compared with controls at 12 to 18 months (RR, 3.06 [95% CI, 2.41 to 3.88]; k=16; n=6975; I2=49.0%) (Figure 9). In the nine trials reporting this effect at 24 months or greater, the effects were attenuated over time; however, six trials still had statistically significant greater probability of 10 percent weight loss in intervention compared with controls, with RRs ranging from 1.6 to 3.8 (Appendix G Table 2).

Figure 9 displays a forest plot of the effect of 16 behavior-based weight loss trials’ proportion of participants losing 10% or more of their baseline weight (n=6975). The pooled risk ratio is 3.06 (95% CI, 2.41 to 3.88) times more likely for the intervention group to lose 10% of their initial weight compared with the control group at 12 to 18 months.

Figure 9

Pooled Analysis of Risk of Losing ≥10% of Body Weight at 12 to 18 Months in Behavior-Based Weight Loss Interventions Compared With Controls. Abbreviations: Cancer = elevated cancer risk; CG = control group; CI = confidence interval; CV = increased (more...)

Waist Circumference

A meta-analysis of 41 trials reported a mean greater reduction of approximately 2.51 cm (1.0 in) in waist circumference among those in behavior-based weight loss interventions compared with control conditions at 12 to 18 months followup (95% CI, −3.15 to −1.87; k=41; n=12,180; I2=94.6%) (Table 11). Absolute changes in waist circumference ranged from 0.1 cm to −11.3 cm among intervention participants and from 1.5 cm to −7.4 cm among control participants. Fewer trials reported the effects of the intervention on other adiposity outcomes such as waist-to-hip ratio and percent body fat; results related to these outcomes are presented in Appendix G Table 1.

Incident Type 2 Diabetes

Thirteen trials (n=4095) reported incident type 2 diabetes associated with behavior-based weight loss interventions (Table 12).205, 215, 225, 258, 265, 266, 277, 280, 283, 288, 306, 314, 321 Twelve of the 13 trials were limited to adults with prediabetes or those who were otherwise at risk for diabetes (family history, history of gestational diabetes, metabolic syndrome); the one remaining trial was conducted among breast cancer survivors.288 Most of the trials reported cases of diabetes over 1 year of followup; only five trials reported type 2 diabetes incidence at 2 or more years. In DPP (n=1295), the estimated cumulative incidence of diabetes at 3 years was 14.4 versus 28.9 percent in the lifestyle-intervention versus placebo groups, respectively (between-group crude incidence difference of −58% [95% CI, 48 to 66]; study-reported NNT, 6.8).205 Similarly, the good-quality Finnish DPS (n=523), a 4-year behavior-based weight loss intervention trial,306, 340 found that after 9 years, intervention group participants were significantly less likely to develop type 2 diabetes compared with the control group (40.0% vs. 54.5%, respectively; HR, 0.4 [95% CI, 0.3 to 0.7]).306, 340 The European Diabetes Prevention Study (EDIPS) (n=102) applied the DPS intervention in the United Kingdom and found a large but nonsignificant reduction in the incidence of diabetes in the intervention group compared with the control group after 5 years (9.8% vs. 21.6%, respectively; RR, 0.45 [95% CI, 0.2 to 1.2]).283 In the remaining 10 trials, progression to diabetes was observed less frequently with absolute cumulative incidence of diabetes at up to 3 years followup ranging from 0 to 15.0 percent in intervention participants and 0 to 28.9 percent among control participants. Although the differences between intervention and control groups were not statistically significant, the studies were generally of shorter duration and smaller than DPP and FDPS.215, 225, 258, 265, 266, 277, 280, 288, 314, 321 When the two larger and seven of the smaller trials that reported rates of incident diabetes were pooled, there was a significant 33 percent reduction in risk of developing diabetes over 1 to 9 years (pooled RR, 0.67 [95% CI, 0.51 to 0.89]; k=9; n=3140; I2=49.2%) (Figure 10).

Table 12. Results of BehaviorBased Weight Loss Interventions on Incident Diabetes (k=13) (n=4095).

Table 12

Results of BehaviorBased Weight Loss Interventions on Incident Diabetes (k=13) (n=4095).

Figure 10 displays a forest plot of the effect of 9 behavior-based weight loss trials (n=3140) on the incidence of diabetes. The pooled rates showed a significant 33% reduction in risk of developing diabetes over 1 to 9 years (pooled RR, 0.67 [95% CI, 0.51 to 0.89]).

Figure 10

Pooled Analysis of Risk of Developing Diabetes in Behavior-Based Weight Loss Interventions Compared With Controls. Abbreviations: CG = control group; CI = confidence interval; IG = intervention group; RR = risk ratio

Other Intermediate Outcomes

Other intermediate outcomes, including the prevalence of hypertension, use of CVD medications, prevalence of metabolic syndrome, and estimated 10-year risk of CVD were sparsely reported within the trials. Rates of hypertension at 18 months to 3 years of followup were reported for the TOHP Phase I (n=564) and Phase II (n=1191) trials as well as the DPP trial (n=2161) and a smaller study by Nilsen et al (n=213).301, 327, 333, 336 TOHP I and II reported 34 and 22 percent reduced risk of incident hypertension at 18 months among those in the weight loss condition (which also included sodium reduction) compared with the control condition, respectively.333, 336 By 3 years in TOHP II, fewer participants in the intervention group (32%) met criteria for hypertension compared with the control group (39%) (absolute risk difference [RD], 7.3 [NNT=14]). In DPP (n=2161), the prevalence of hypertension remained stable among intervention participants (approximately 30%) but increased among control participants (from approximately 30% to 40%) over 3 years (p<0.001).336 Similarly, use of antihypertensive medications rose from 17 to 23 percent among DPP intervention participants and from 17 to 31 percent among control participants over 3 years (p<0.001). Likewise, fewer weight loss participants (12%) required drug therapy for either elevated triglyceride or low-density lipoprotein cholesterol levels compared with control participants (16%) (p<0.001).205, 336 A smaller study by Nilsen et al (n=213) found no significant difference between groups in the percent of individuals with hypertension by the end of the study; however, there was a high baseline prevalence of hypertension (74%).327 Four smaller trials examining medication changes (n=30 to 772) did not find significant differences in antihypertensive or lipid-lowering medication use between intervention and control arms.232, 253, 288, 291 Five trials (n=3356) reported on incidence of metabolic syndrome in intervention and control arms at 1 to 3 years followup with mixed results.205, 243, 265, 286, 291 Similarly, two trials (n=165) reported mixed findings on the effects of weight loss interventions on estimated 10-year CVD risk at 1 year based on the U.K. Prospective Diabetes Study risk engine or QRISK2.214, 243

Subpopulations

Subpopulation analyses were infrequently reported among included studies and often not prespecified. Even when prespecified analyses were performed, they often lacked interaction testing, limiting the allowable interpretation of treatment effect by subpopulation.

The differential effects of weight loss interventions for individuals with varying baseline BMIs was examined in five trials,234, 266, 288, 301, 302 only two of which266, 301 prespecified such subgroup analyses. No trial found that baseline weight was associated with weight change following interaction testing.

Prespecified analyses of the effect of age were reported in two studies with mixed results. In the CITY trial among young adults (mean age, 29.4 years), a cell-phone based intervention, the oldest tertile of participants (mean age not reported) lost less weight than the youngest tertile of participants (mean age not reported).302 However, in DPP there was a suggestion of a stronger effect of the lifestyle intervention in older individuals (ages 60 to 85 years); however, there was no interaction testing, limiting interpretation of this finding.341

Whether sex influenced the effectiveness of weight loss interventions was reported in eight trials, with six prespecifying interaction testing analyses. Men were generally observed to lose a greater percentage of their baseline weight than women; however, only two301, 342 of six studies found the sex differences to be significant in interaction testing.254, 266, 300, 302 Two exploratory analyses255, 279 also had mixed results, with one trial reporting greater weight loss in men following interaction testing.279

The effect of race/ethnicity on the effectiveness of weight loss interventions was examined in seven trials; six of these analyses were prespecified.215, 300-302, 326, 342 There was a trend toward greater weight loss among white participants than black or Hispanic participants. However, this finding became nonsignificant in three215, 300, 302 of the five trials following interaction testing. The two trials that found a significant racial/ethnic difference were TOHP II and DPP. In TOPH II, white participants lost a net 1.8 kg more than African American participants at 18 months.301 Within DPP, black women exhibited significantly smaller (approximately half) weight losses (p<0.01) with the lifestyle intervention than other race-sex groups.342 One additional trial (TONE) found significantly greater weight loss in white participants than black participants; however, no interaction testing was performed.326, 343 One exploratory analysis among a predominantly non-Hispanic white female population found no difference by race/ethnicity.288

Subgroup analyses for all other outcomes were limited by sparse reporting and limited interaction testing.

Effect Modification

We conducted subgroup analyses and a series of meta-regressions to explore potential effect modification by prespecified study, population, and intervention characteristics (see Appendix B for the full list of variables). We limited these analyses to the main outcome of change in weight at 12 to 18 months followup, and all meta-regressions controlled for risk status of the population.

In terms of intervention characteristics, subgroup analyses according to the number of intervention sessions in the first year (>26 sessions, 12–26 sessions, and <12 sessions) showed slightly higher effect estimates among interventions with more sessions; however, the confidence intervals among all three of the subgroups overlapped (Figure 11). When examined as a continuous measure, a higher number of intervention sessions in the first 12 months was associated with significantly more weight loss (coefficient, −0.03; p=0.023); however, total number of contacts (including text messages, emails, and print materials) was not (coefficient, 0.001; p=0.488). Likewise, the number of sessions in the first 12 months was not associated with greater weight loss after controlling for the presence of any group sessions (coefficient, −0.015; p=0.212). In addition, there was no pattern of effects according to the main mode of intervention delivery (i.e., group vs. individual vs. technology-based vs. mixed) (Figure 11). However, there was evidence of a greater effect among interventions that included any group sessions versus those that did not (coefficient, −1.19; p=0.004). This held true after controlling for the total number of sessions within the first year and the risk status of the population (coefficient, −0.97; p=0.029). Among the subset of trials that included any group sessions (whether or not it was the main mode of delivery), the pooled difference in weight change was −3.03 kg (95% CI, −3.65 to −2.42; k=35; n=15,132; I2=91.3%). Those without any group sessions resulted in a smaller pooled effect estimate (although still statistically significant) and reduced statistical heterogeneity (MD, −1.46 kg [95% CI, −1.84 to −1.09]; k=32; n=6933; I2=49.8%). None of the other intervention characteristics we looked at modified the effect of the intervention, including duration of the intervention, whether there was in-person support, whether individual in-person or telephone sessions were offered, whether the intervention was technology-based, whether self-monitoring of weight or behaviors was encouraged, or whether the intervention was based on the DPP. Descriptions of each intervention, including specific intervention components, are fully described in Table 4, Table 5, and Appendix F Table 1.

Figure 11 displays the pooled effect of 67 behavior-based weight loss trials (n=22,065) by several subgroups based on intervention and population characteristics. The pooled mean difference of change in weight loss is presented for each subgroup at 12 to 18 months.

Figure 11

Pooled Analysis for Prespecified Subgroups of Trials for Change in Weight at 12 to 18 Months in Behavior-Based Weight Loss Interventions Compared With Controls. *k=65. Two trials’ baseline mean BMI were not reported. Abbreviations: BMI = body (more...)

In terms of population characteristics, larger differences in weight change were seen among trials that specifically enrolled adults with increased cardiovascular risk, subclinical risk, and elevated cancer risk versus those who were unselected or generally at low risk (coefficient, −1.15; p=0.004). A meta-analysis of the subset of 33 trials among participants at elevated risk found a pooled MD in change of −2.98 kg (95% CI, −3.58 to −2.39; k=33; n=10,554; I2=87.7%) at 12 to 18 months (Figure 11). A statistically significant association was also found for the subset of trials among low risk or unselected participants but with a significantly smaller effect estimate than that seen for those at risk (MD, −1.82 kg (95% CI, −2.35 to −1.30; k=34; n=11,511; I2=82.8%). Those who self-selected or volunteered to take part in the interventions were also more likely to experience greater weight loss (MD, −2.97 kg (95% CI, −3.87 to −2.07; k=28; n=9626; I2=94.0%) than participants who were recruited directly into the trial (MD, −2.02 kg (95% CI, −2.47 to −1.56; k=39; n=12,439; I2=79.7%) (coefficient, −1.14; p=0.004), after controlling for risk status. Baseline BMI and baseline weight category (i.e., overweight, Class I obesity, and Class II obesity) were not associated with differences in the effects of the intervention on weight change, percent weight change, or the proportion of participants losing at least 5 percent of their baseline weight.

There was no evidence of effect modification by study quality or U.S.- versus non-U.S.-based studies. Sample retention at 12 months was associated with the pooled effect size in that trials with higher retention rates experiencing greater weight loss (coefficient, −0.05; p=0.011).

In summary, a few factors were identified in the subgroup analyses and meta-regressions as potential effect modifiers. However, the heterogeneity in each individual intervention arm, confounded with differences in the populations, settings, and trial quality, make it nearly impossible to disentangle what variables may be driving larger effects. The consistency—yet wide range in effects—seen across specific interventions and across various adult subgroups emphasizes a broad range of benefit that is likely dependent on other individual, social, and environmental factors influencing an individual’s weight loss.

Behavior-Based Weight Loss Maintenance Interventions

Maintenance of Previous Weight Loss

All weight-related outcomes for all time points for the nine behavior-based weight loss maintenance trials are reported in Appendix G Table 3 for continuous outcomes and Appendix G Table 4 for dichotomous outcomes.

Six trials included an initial weight loss intervention (mean weight loss of 5 to 15 kg [11 to 33.1 lb]) for all study participants (Appendix F Table 2).233, 282, 284, 303, 309, 317 Three additional trials did not include a weight loss portion but required that participants have recently lost 5296 or 10294, 313 percent of their body weight. In eight trials, both the intervention and control arms regained weight over a 12- to 18-month followup; however, the intervention arm experienced less weight regain (gain of 0.1 kg [0.2 lb] to 7.5 kg [16.5 lb] in intervention arms and 0.6 kg [1.3 lb] to 8.8 kg [19.4 lb] in control arms), although all participants maintained some of their previous weight loss.233, 282, 284, 294, 303, 309, 313, 317 Only four of the eight trials had statistically significant results. In the ninth trial both the intervention and control arms continued to lose weight (loss of 2.4 kg [5.3 lb] in intervention and 0.6 [1.3 lb] in controls); however these within-group changes were not statistically significant.296 A meta-analysis combining the eight behavior-based weight loss trials that reported kilograms or pounds lost at 12 to 18 months found a pooled mean difference of −1.6 kg (−3.5 lb) in the intervention versus control groups (MD, −1.59 kg [95% CI, −2.38 to −0.79]; k=8; n=1408; I2=26.8%) (Figure 12). The one trial that could not be included in the meta-analysis had similar results.303 Three studies included participant followup beyond 18 months with mixed findings.294, 303, 317

Figure 12 displays a forest plot of the effect of 8 behavior-based weight maintenance trials (n=1408). The pooled mean difference of change in weight was -1.59 kg (95% CI, -2.38 to -0.79) in favor of the intervention group vs. control group at 12 to 18 months.

Figure 12

Pooled Analysis of Change in Weight at 12 to 18 Months in Behavior-Based Weight Maintenance Interventions Compared With Controls. Abbreviations: BL = baseline; Cancer = elevated cancer risk; CG = control group; CI = confidence interval; CV = increased (more...)

Maintenance of 5 Percent or Greater Weight Loss

Only three of the maintenance trials (n=1320) examined maintenance of 5 percent weight loss over 12 to 36 months, finding mixed results. In two small trials (n=92 and 200), those randomized to a maintenance intervention were not more likely to have maintained 5 percent weight loss by 12 to 36 months.282, 317 However, in the larger trial (n=1029), those in the maintenance group were slightly more likely to have maintained 5 percent of their weight loss at 30 months (42% vs. 34%; RR, 1.24 [95% CI, 1.02 to 1.51]) and 60 months (37% vs. 27%; RR, 1.37 [95% CI, 1.03 to 1.82]) compared with the minimal intervention arm.303

Waist Circumference

Three small trials (n=453) reported change in waist circumference after behavior-based maintenance interventions.296, 309, 317 Changes in waist circumference were not significantly different between intervention and control groups at 12 months. Extension of one trial for 36 months did not reveal any significant differences in waist circumference over the longer-term followup.317

Incident Diabetes and Other Intermediate Outcomes

No study reported these outcomes.

Medication-Based Weight Loss Interventions

All weight-related outcomes for all time points for all medication-based weight loss trials are reported in Appendix G Table 5 for continuous outcomes and Appendix G Table 6 for dichotomous outcomes. Findings were often limited by reduced long-term followup, with the majority of trials reporting 30 percent or greater loss to followup or greater by 12 to 13 months (Figure 13), and limited data reporting (often not reporting statistical significance of findings and lack of description of variance) (Table 2). The study-specific results below and in tables reflect analyses using an mITT analysis (i.e., participants’ last observation postbaseline while still on study drug) as that was the primary analysis reported by studies (mITT is required by the FDA). Results from sensitivity analyses within trials using other data substitution methods (baseline observation carried forward, multiple imputation using mixed effects models) were generally consistent with the mITT results.

Figure 13 displays a forest plot of the effect of medication-based weight loss and weight loss maintenance trials’ proportion of participants losing at least 5% of their baseline weight. This figure visualizes the change in effects over time within each trial by drug name. Due to limitations, the effects were not pooled.

Figure 13

Risk of Losing ≥5% of Body Weight in Medication-Based Weight Loss Interventions Compared With Controls, All Studies and All Time Points. Abbreviations: CG = control group; CI = confidence interval; CV = increased cardiovascular risk; IG = intervention (more...)

Weight Loss

Liraglutide. Two trials reported on degree of weight loss in liraglutide versus placebo arms (n=3853).220, 285 Those in the liraglutide groups lost statistically more weight (−7.8 to −8.4 kg [−17.2 to −18.5 lb]) than those in placebo group (−2.0 to −2.8 kg [−4.4 to −6.2 lb]) over 12 to 13 months, a statistically significant difference (p<0.001) (Table 13). One trial extended followup in those with prediabetes to 36 months; mean weight loss was less by 36 months in both groups, but the mean difference in weight loss between arms was still statistically different (liraglutide arm lost 4.6 kg more than placebo at 36 months; p<0.0001).339

Table 13. Results of Medication-Based Weight Loss Interventions on Weight Loss (kg), by Drug (k=18) (n=22,972).

Table 13

Results of Medication-Based Weight Loss Interventions on Weight Loss (kg), by Drug (k=18) (n=22,972).

Lorcaserin. Two trials reported on degree of weight loss in lorcaserin versus placebo arms using mean or least square mean (LSM) (n=6139) (Table 13).172, 173 Those randomized to lorcaserin lost a mean or LSM of 5.8 kg (12.8 lb), while those in placebo lost 2.2 to 2.9 kg (4.9 to 6.4 lb) over 12 months, which was statistically significantly different in both trials (p<0.001).

Naltrexone and bupropion. Three trials reported on degree of weight loss in naltrexone and bupropion compared with placebo arms (Table 13).218, 244, 311 Those randomized to naltrexone and bupropion lost more weight over 13 months compared with those randomized to placebo (LSM, −6.1 to −6.2 kg [−13.4 to −13.7 lb] and −1.3 to −1.4 kg [−2.9 to −3.1 lb] in the intervention and placebo groups, respectively; p<0.001).218, 244 One trial reported only the percent change in weight, with those in the naltrexone and bupropion arm showing greater percent weight loss than the placebo arm (LSM, −9.1% vs. −5.1%, respectively; p<0.001) (Appendix G Table 5).311

Orlistat. Eleven trials reported on degree of weight loss in orlistat versus placebo arms.160, 161, 227, 236, 239, 246, 260, 263, 292, 297, 304 In every trial, those randomized to orlistat lost statistically significantly more weight loss (mean of 1.0 to 4.4 kg [2.2 to 9.7 lb] more) than those on placebo over 12 months. (Table 13).161, 260, 297 In the two trials that compared the 60 mg TID dosage to 120 mg TID, weight loss was about 0.8 to 0.9 kg less with 60 mg TID compared with 120 mg TID over 12 months.246, 292 Two studies examined weight loss at later time points (18 to 48 months). Mean weight loss was less in both arms at the later time points (1.2 to 2 kg had been regained); however, those in the orlistat arm had still lost more weight since randomization compared with the placebo arm (mean difference, −3.1 to −3.37 kg in 120 mg TID of orlistat and −2.3 to −2.81 kg in 60 mg TID of orlistat vs. placebo arms, respectively; p<0.01).246, 292 Following 4 years of treatment, participants had regained approximately half of their weight loss since randomization; however, the 120 mg orlistat arm still had lost significantly more weight than the placebo arm (p<0.001).161

Phentermine and topiramate. Two trials reported on degree of weight loss for those randomized to phentermine and topiramate versus placebo arms.216, 241 In one trial, participants randomized to phentermine and topiramate lost statistically significantly more weight than those on placebo (LSM, −8.1 kg [−17.8 lb] with 15/92 mg, −10.2 kg [−22.5 lb] with 7.5/46 mg, and −1.4 kg [−3.1 lb] with placebo; p<0.0001) (Table 13).241 The second trial only reported the percentage of weight loss in the two arms. Those randomized to phentermine and topiramate lost a greater percentage of weight compared with the placebo arm at 12 months (LSM, 10.9% vs. 1.5%, respectively; p<0.0001) (Appendix G Table 5).216

Weight Loss of 5 Percent or Greater

Liraglutide. At 12 to 13 months, participants randomized to liraglutide were 2.8 to 4.8 times more likely to lose 5 percent of their body weight compared with those in the placebo arm (63% to 79% compared with 27% to 29%, respectively) and 3.9 to 4.3 times more likely to lose 10 percent of their weight (33% to 40% compared with 10% to 11%, respectively) (p<0.001) (Figure 13, Figure 14).220, 285 In additional followup of a subgroup with prediabetes at baseline, those randomized to orlistat were over 3 times more likely to have achieved 5 and 10 percent weight loss after 36 months compared with those on placebo (p-values<0.001), although absolute percentages who reached this milestone was smaller in both arms.339

Figure 14 displays a forest plot of the effect of medication-based weight loss and weight loss maintenance trials’ proportion of participants losing 10% or more of their baseline weight. This figure visualizes the change in effects over time within each trial by drug name. Due to limitations, the effects were not pooled.

Figure 14

Proportion of Participants Losing ≥10% of Body Weight in Medication-Based Weight Loss Interventions Compared With Controls at Multiple Followup Time Points. Abbreviations: CG = control group; CI = confidence interval; CV = increased cardiovascular (more...)

Lorcaserin. Compared with placebo, those randomized to lorcaserin were 1.9 to 2.3 times more likely to lose 5 percent of their body weight (47% vs. 20%–25% in lorcaserin and placebo arms respectively; p<0.001) (Figure 13) and 2.3 to 2.9 times more likely to lose 10 percent of their weight by 12 months (23% vs. 8%–10% in in lorcaserin and placebo arms, respectively; p<0.001) (Figure 14).

Naltrexone and bupropion. Those randomized to naltrexone and bupropion were 1.6 to 3.0 times more likely to lose 5 percent of their weight (48%–66% in naltrexone and bupropion arm vs. 16%–42% in placebo arm; p<0.01)218, 244, 311 (Figure 13) and 2.0 to 5.0 times more likely to lose 10 percent of their weight (25%–42% in naltrexone and bupropion arm compared with 6%–20% in placebo arm; p<0.001) (Figure 14).

Orlistat. Ten trials reported the percentage of participants who lost at least 5 and 10 percent of their baseline weight.160, 161, 227, 239, 246, 260, 263, 287, 292, 297 Participants randomized to orlistat were 1.3 to 2.3 times more likely to lose 5 percent of their weight at 12 months compared with those given placebo (35%–73% vs. 21%–49%, respectively; p<0.05) (Figure 13).160, 161, 227, 239, 246, 260, 263, 287, 292, 297 In the two trials that examined both orlistat dosages (60 TID and 120 TID), there was little evidence of a dosage effect.246, 292 In the four trials that extended followup beyond 12 months,161, 246, 260, 292 those on either dose of orlistat were still significantly more likely to be 5 percent below their starting weight at 24 to 48 months (RR, 1.41 to 1.74; p<0.05). Nine of these 10 trials also reported 10 percent weight loss;161, 227, 239, 246, 260, 263, 287, 292, 297 the results were similar with those on either dosage of orlistat being more likely to have 10 percent weight loss at 12 to 48 months (RR, 1.31 to 2.95; p<0.05 in all but one study263) (Figure 14); however, the absolute percentage of participants who reached this milestone was smaller in both arms, with rates decreasing as followup time increased.161, 227, 239, 246, 260, 263, 292, 297

Phentermine and topiramate. Those randomized to 15/92 mg or 7.5/46 mg phentermine and topiramate were 3.0 to 3.9 times more likely to lose 5 percent of their weight, respectively, by 12 to 13 months (67% to 70% of 15/92 mg, 62% of 7.5/46 mg, and 17% to 21% of placebo; p<0.0001) (Figure 13). They were 5.1 to 6.4 times more likely to lose 10 percent of their body weight (47% to 48% of 15/92 mg, 37% of 7.5/46 mg, and 7% of placebo); p<0.0001) (Figure 14).216, 241

Waist Circumference

Liraglutide. Participants randomized to liraglutide had significantly greater mean waist circumference decreases than placebo (means, 7.8 to 8.2 cm over 12 to 13 months compared with 3.0 to 3.9 cm in the placebo arm; p<0.001) (Table 14). Among participants with prediabetes at baseline, the change was slightly attenuated by 36 months; however, those randomized to liraglutide still had a statistically significant greater 3.5-cm decrease in waist circumference.339

Table 14. Results of Medication-Based Weight Loss Interventions on Waist Circumference (cm), by Drug (k=14) (n=22,227).

Table 14

Results of Medication-Based Weight Loss Interventions on Waist Circumference (cm), by Drug (k=14) (n=22,227).

Lorcaserin. Waist circumference decreased more in those randomized to lorcaserin compared with those randomized to placebo by 12 months (LSM/means, −6.3 to −6.8 cm vs. −3.9 to −4.1 cm, respectively; p<0.001) (Table 14).172, 173

Naltrexone and bupropion. Waist circumference decreased more in those randomized to naltrexone and bupropion compared with those in placebo over 12 months (LSM/means, −6.2 to −10.2 cm vs. −2.1 to −7.0, respectively; p<0.001) (Table 14).218, 244, 311

Orlistat. There was a greater decrease in waist circumference in the orlistat arms over 12 to 18 months compared with placebo (LSM/means, −3 to −9.6 cm vs. −1.9 to −7.0 cm) (Table 14).161, 227, 236, 260, 263, 292, 304 Statistical significance was reported in only six of the seven trials, with four showing a statistically significance difference between arms. In the one study that examined both 60 mg TID and 120 TID dosages, there was no evidence of a dosage effect.292 By 24 and 48 months, there was regain in waist circumference in both arms, but the statistically significant differences remained except for one 60 mg TID arm.161, 292

Phentermine and topiramate. Over 12 to 13 months, waist circumference decreased significantly more for participants randomized to phentermine-topiramate 15/92 mg compared with placebo (LSM, −9.2 to −10.9 vs. −2.4 to −3.1 cm, respectively; p<0.0001). This effect was also significant in participants randomized to 7.5/46 mg (LSM, −7.6 vs. −2.4 cm, respectively; p<0.0001) (Table 14).216, 241

Incident Diabetes

Liraglutide. In the single liraglutide trial examining incident diabetes (n=3662), fewer participants randomized to liraglutide (n=4 [0.2%]) developed diabetes over 13 months compared with those given placebo (n=14 [1.1%]) (odds ratio [OR], 8.1 [95% CI, 2.6 to 25.3]; p<0.001) (Table 15).285 The trial continued past 13 months among 2210 participants with prediabetes at baseline; participants randomized to liraglutide were less likely to develop type 2 diabetes by 36 months compared with placebo (1.8% vs. 6.2% of participants), with a mean time from randomization to diagnosis of 99 (SD, 47) versus 87 (SD, 47) weeks, respectively (HR, 0.21 [95% CI, 0.13 to 0.34; p<0.0001]). However, these findings are limited by the large number of participants who discontinued medication during the 36-month followup (53% of those on liraglutide and 45% of those on placebo completed the study on medication).339

Table 15. Results of Medication-Based Weight Loss and Weight Loss Maintenance Interventions on Incident Diabetes (k=4) (n=9340).

Table 15

Results of Medication-Based Weight Loss and Weight Loss Maintenance Interventions on Incident Diabetes (k=4) (n=9340).

Lorcaserin. No study reported this outcome.

Naltrexone and bupropion. No study reported this outcome.

Orlistat. A trial of over 3300 persons (21% with prediabetes) found that 6 percent of those randomized to 120 mg TID of orlistat developed type 2 diabetes over 48 months compared with 9 percent of those in the placebo arm (HR, 0.63 [95% CI, 0.46 to 0.87]; p=0.005). However, applicability of these findings is limited by the high discontinuation rate (only 52% and 35% of intervention and placebo participants, respectively, completed a 48-month followup on study medication).161

Phentermine and topiramate. One trial of over 2400 participants with elevated cardiovascular risk (68% with prediabetes) reported that 14 (1.7%) and 12 (2.8%) of those randomized to 15/92 mg and 7.5/46 mg of phentermine and topiramate, respectively, developed type 2 diabetes compared with 30 (3.6%) of those on placebo (RR, 0.47 [95% CI, 0.25 to 0.88] and 0.78 [95% CI, 0.40 to 1.50] for 15/92 mg and 7.5/46 mg arms, respectively) (Table 15).241

Other Intermediate Outcomes

Liraglutide. Compared with placebo at 13 months, those randomized to liraglutide in one trial (n=3662) were less likely to increase use of lipid-lowering medication (2.1% vs. 3.7%) and antihypertensive medication (3.7% vs. 5.7%) and were more likely to decrease use of antihypertensives (6.0% vs. 3.8%), but statistical significance was not reported.285

A second, smaller trial (n=191), in which there was a high prevalence of metabolic syndrome at baseline (42% and 51% in liraglutide and placebo arms, respectively), reported that those randomized to liraglutide had a statistically significant lower prevalence of metabolic syndrome compared with placebo at 12 months of followup (17% vs. 43%; p=0.005).220

Lorcaserin. In one 12-month trial (n=3102),173 4.0 and 5.0 percent of those randomized to lorcaserin and placebo arms, respectively, increased use of lipid-lowering medications and 2.6 and 1.4 percent decreased use. Use of antihypertensive medications decreased in 4.0 and 3.1 percent of participants randomized to lorcaserin and placebo, respectively. However, the number taking these classes of medications at baseline was not given, and statistical significance was not reported.

Naltrexone and bupropion. No study reported other intermediate outcomes.

Orlistat. One trial reported that over 12 months there was no significant difference between change in 10-year CVD risk score and usage of cardiovascular medications in those randomized to 120 mg TID of orlistat compared with placebo.304

Phentermine and topiramate. No study reported other outcomes.

Medication-Based Weight Loss Maintenance Interventions

All weight-related outcomes for all time points for all medication-based weight loss trials are reported in Appendix G Table 7 for continuous outcomes and Appendix G Table 8 for dichotomous outcomes.

Maintenance of Previous Weight Loss

Liraglutide. One trial (n=413) examined weight loss maintenance with liraglutide after a run in weight loss period in which qualifying participants were required to lose 5 percent of their body weight (mean weight loss, 6.3 kg [13.9 lb]) through a hypocaloric diet.312 During the 13-month maintenance phase, those in the placebo arm lost an additional 0.1 kg (0.2 lb), while those randomized to liraglutide lost an additional 6.0 kg (13.3 lb) (p<0.0001).

Orlistat. In the two trials247, 287 examining the use of orlistat for weight loss maintenance, participants had lost an average of 9.9 to 12.0 kg (21.8 to 26.5 lb) through hypocaloric diets prior to being randomized into the weight loss maintenance phase. Those randomized to 120 mg TID of orlistat gained 1.8 to 1.9 kg (4.0 to 4.2 lb) less than those given placebo over 12 to 18 months (2.6–2.8 vs. 4.4–4.7 kg, respectively, a statistically significant difference in the one study reporting statistical results247). In the one trial with both 120 mg TID and 60 mg TID arms, the 60 mg TID arm did not have significantly less weight gain than those in the placebo group over 12 months (3.8 vs. 4.4 kg gain, respectively).247 During longer-term followup in one trial of 120 mg TID, those randomized to orlistat continued to have less weight gain by 36 months compared with placebo (5.1 vs. 7.1 kg; p=0.028).287

Maintenance of Weight Loss of 5 Percent or Greater

Liraglutide. In the one maintenance trial,312 all participants had experienced at least 5 percent weight loss on entry. Compared with placebo participants, those on liraglutide were 2.3 times more likely to have maintained 5 percent weight loss (RR, 2.32 [95% CI, 1.74 to 3.11]) and 4.1 times more likely to have achieved 10 percent weight loss (RR, 4.13 [95% CI, 2.33 to 7.34]) at the end of the 12-month weight loss maintenance period (p<0.0001) (Figure 13).

Orlistat. In the one maintenance trial reporting percent weight loss, all participants had lost at least 5 percent of their weight at study entry (Figure 13, Figure 14).287 After 12 months of maintenance treatment, those randomized to orlistat 120 mg TID were 1.2 times more likely to have maintained their 5 percent weight loss (RR, 1.18 [95% CI, 1.05 to 1.33]; p<0.001), and this difference remained by 36 months (RR, 1.20 [95% CI, 1.00 to 1.43]; p<0.05). However, the percentage with 10 percent weight loss at 36 months was not statistically different between arms (RR, 1.18 [95% CI, 0.85 to 1.64]; p=NS).

Maintenance of Waist Circumference Decrease

Liraglutide. While both arms had continued decreases in waist circumference during the 13-month maintenance trial, the decrease was greater among those randomized to liraglutide (−4.7 vs. −1.2 cm; p<0.0001).312

Orlistat. In the one trial reporting waist circumference, those randomized to 120 TID of orlistat had no increase in waist circumference at 18 months, while the placebo group experienced an average 3 cm increase (p=NR).287 By 36 months, those randomized to orlistat had an average 4.3 cm increase in waist circumference, which was less than the 6.6-cm gain reported in the placebo arm (p=0.032).

Incident Diabetes

Liraglutide. No study reported this outcome.

Orlistat. One trial of 309 participants with at least one cardiovascular risk factor (27% with prediabetes) reported that compared with those randomized to 120 mg of orlistat, almost twice as many persons on placebo developed type 2 diabetes over the 36-month weight loss maintenance intervention (5.2% vs. 10.9%; p=0.041) (Table 15).287

Other Intermediate Outcomes

No study reported other intermediate outcomes.

KQ3. 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?

Summary of Results

Rates of adverse events were sparsely reported in the behavior-based weight loss and weight loss maintenance trials (30 trials; n=12,824). In general, there were no serious harms related to the interventions and most trials noted no differences between groups in the rates of adverse events, including cardiovascular events. In the three trials large enough to examine musculoskeletal issues between groups, results were mixed.

Almost all medication trials reported adverse events. Weight loss medications were associated with more adverse events than placebo, which resulted in higher dropout rates due to adverse events in the medication than placebo arms. However, serious adverse events were not generally more common in those randomized to medications. There are multiple potential harms required by the FDA to be listed on weight loss medication labels, but these harms have not been well evaluated in the trials included in this review.

Detailed Results

Behavior-Based Weight Loss Interventions

Twenty-seven trials reported harms (or lack of harms) associated with a behavior-based weight loss intervention (n=12,235).205, 206, 215, 217, 219, 224, 225, 235, 237, 242, 249, 251, 253, 258, 264, 266, 278, 289-291, 305, 320, 321, 323, 325, 328, 330 We rated 15 of these trials as good quality and 12 as fair quality. Only two of these trials (DPP205 and SLIM325) were included in the previous review; the remaining 25 are new as part of this update. Very few trials reported their methods for capturing adverse events or provided definitions for adverse events or for serious adverse events.

Within the 27 trials, eight stated that no harms or serious adverse events were reported217, 251, 253, 264, 278, 321, 325, 328 and three simply stated that none of the adverse events that were reported were related to the study.290, 305, 323 Within the remaining 16 trials that reported actual data, the rates of any adverse event and serious adverse events were relatively low (ranging from 0.6% to 25% of participants experiencing any adverse event) and often not reported by group. In all but two trials,205, 291 rates of any adverse event, serious adverse events, and specific adverse events did not differ between the intervention versus control participants based on statistical testing or in comparing event rates. Those that did show differences between intervention and control groups are discussed below. Four trials specifically reported that no deaths occurred during the trial period.205, 206, 219, 290

Among four trials reporting specifically on cardiovascular-related adverse events or symptoms (e.g., chest pain, difficulty breathing, and fainting and dizziness),206, 215, 224, 291 three of the four reported low rates of self-reported cardiovascular events (<1%)215, 224 and cardiac disorders (10 cases of angina pectoris, atrial fibrillation, atrial flutter, or syncope among the intervention group and 6 cases of angina pectoris, atrial fibrillation, congestive heart failure, and myocardial infarction among the control group)206 and no differences between groups.215, 224 In the remaining trial among sedentary adults (categorized as low risk, as no data given on baseline comorbidities) (n=490), cardiovascular symptoms (including chest pain, difficulty breathing, and dizziness or loss of consciousness) were less common in the intervention group (115 events among 249 participants [46.2%]) compared with the usual care group (137 events among 241 participants [56.8%]), as were cardiovascular symptoms resulting in physician visits (30.5% vs. 34.4%) or hospitalizations (3.6% vs. 7.5%).291

The adverse events most commonly rated as being related to the intervention were musculoskeletal issues, which varied in severity from soreness to sprains to ruptured tendons.205, 206, 215, 224, 235, 237, 291, 330 Rates of musculoskeletal issues ranged from one musculoskeletal injury (0.006 event rate) to nearly 50 percent of participants experiencing muscle or joint aches. Within the seven trials specifically reporting event rates by group, only three had enough events to make comparisons between groups.205, 215, 291 In the DPP trial (n=2161), a statistically significant higher rate of musculoskeletal symptoms (myalgia, arthritis, arthralgia) was seen among those in the lifestyle intervention group (24.1 events per 100 person-years) compared with those in the control group (21.1 events per 100 person-years) (p<0.0167) over 4 years of followup.205 In the PROACTIVE trial (n=490), rates of musculoskeletal events during or after exercise (pain or cramping in leg, knee, or foot; strained muscle, tendon, or ligament; and broken bone) were slightly higher in the usual care group (311 events among 241 participants [129.0%]) compared with the intervention group (300 events among 249 participants [120.5%]) over the course of the 2-year intervention; musculoskeletal events requiring a physician visit were slightly higher in the intervention (70.3%) versus control (66.4%) participants.291 However, in DEPLOY (n=509),215 intervention participants did not experience more muscle or joint aches (48.6% vs. 50.5%; p=NR) or joint sprains or strains (22.6% vs. 22.9%; p=NR) compared with controls.

Six trials (n=2767) reported on incidence of gallbladder disease, which is more common among those who are overweight and have obesity and is associated with rapid weight loss.206, 224, 242, 249, 289, 320 Across these trials, 6 intervention versus 2 control participants experienced either gallstones or cholecystectomy over 1 to 2 years of followup.

One trial of postpartum women showed higher (38% to 42%) rates of reduced breast milk supply; however, these differences were not significant.330

Behavior-Based Weight Loss Maintenance Interventions

Three of the nine behavior-based weight loss maintenance interventions reported harms (n=589).282, 296, 309 In one trial (n=201), a similar number of participants (31%) in the intervention and control group were treated for adverse events over 24 months.282 In a second trial (n=222), there was one (0.1%) death in the control group and four (3.6%) adverse events (knee pain, low blood pressure, bradycardia, and anxiety) in the intervention group over 56 weeks; of note, the intervention group had more contacts at which to report adverse events compared with controls.309 The third trial (n=166) reported one serious adverse event (group not given) that was felt to be related to a pre-existing condition and not to the intervention.296

Medication-Based Weight Loss Interventions

Liraglutide

Three trials reported harms (or lack of harms) related to liraglutide (n=3990).220, 259, 285 Compared with placebo, those randomized to liraglutide had a higher prevalence of experiencing at least one adverse event (Table 16) (80%–96% vs. 63%–89%) and slightly more serious adverse events (Table 17) (6%–15% vs. 3%–13%) over 12 to 36 months, although statistical testing was not reported.220, 285 As a result, more participants withdrew from the liraglutide (8% to 33%) compared with the placebo arm (0% to 6%) because of adverse events;220, 259, 285 statistical testing was only presented in one study, which noted a statistical difference between groups (p=0.009). One trial reported on total mortality with 1 to 2 deaths per arm.285 Compared with placebo, there was no evidence that more participants on liraglutide developed depression (1%–17% vs. 0%–18%),220, 285 suicidal behavior and/or ideation (0.5% vs. 0.9%),285 or anxiety (2% vs. 1%)220 at 12 to 18 months (p=NR). The most common adverse events were gastrointestinal,285 and 77 to 79 percent of those randomized to liraglutide experienced at least one gastrointestinal event compared with 31 to 46 percent of those on placebo (p=NR).220, 259 Of those on liraglutide, 4 to 8 percent withdrew from the trial because of gastrointestinal adverse events compared with 1 to 2 percent of placebo participants (p=NR).220, 285

Table 16. Percent of Individuals Experiencing at Least One Adverse Event in Studies of Medication-Based Weight Loss and Weight Loss Maintenance (k=16) (n=14,428.

Table 16

Percent of Individuals Experiencing at Least One Adverse Event in Studies of Medication-Based Weight Loss and Weight Loss Maintenance (k=16) (n=14,428.

Table 17. Percent of Individuals Experiencing at Least One Serious Adverse Event in Studies of Medication-Based Weight Loss and Weight Loss Maintenance (k=25) (n=25,496).

Table 17

Percent of Individuals Experiencing at Least One Serious Adverse Event in Studies of Medication-Based Weight Loss and Weight Loss Maintenance (k=25) (n=25,496).

Other potential harms that are listed in the “warnings and precautions” section of the liraglutide label include malignant thyroid c-cell carcinomas (black box warning), pancreatitis, gallbladder disease, renal impairment, increased heart rate, hypersensitivity/anaphylaxis, hypoglycemia, and cancer.220, 259, 285 One large trial of individuals with prediabetes339 reported 10 cases of confirmed pancreatitis during 3 years among 1505 individuals randomized to liraglutide treatment (0.7%; 0.3 events per 100 person-years) compared with two cases in 747 placebo-group individuals (0.3%; 0.1 events per 100 person-years), with most events (8/12) occurring within the first year of treatment.339 Data on other outcomes were generally sparse among the trials.220, 259, 285

Lorcaserin

Four trials reported harms (or lack of harms) related to lorcaserin (n=6490).172, 173, 238, 268 More participants in the lorcaserin arms experienced at least one adverse event (12% at 1 month, 83% at 1 year) compared with those on placebo (4% at 1 month, 75% at 1 year), although statistical testing was not conducted (Table 16).173, 238 Compared with placebo, the adverse event most commonly associated with lorcaserin was dizziness, with 8 to 10 percent of those randomized to lorcaserin experiencing dizziness compared with 4 percent of controls (p=NR).172, 173, 268 Rates of serious adverse events were low in both groups (0% to 3% in lorcaserin and 0% to 2% in placebo; p=NR) (Table 17).172, 173, 238 In one trial, six serious adverse events were deemed to be potentially related to lorcaserin; three occurred in the placebo group (syncope, ventricular tachycardia, and anaphylactic reaction), and three occurred in the lorcaserin group (syncope, moderate depression, and acute anxiety attack).173 Withdrawals due to adverse events were less than 10 percent in both arms (7% in lorcaserin arm and 5% to 7% in placebo arm; statistical testing was not conducted).172, 173 There were only 2 deaths, both in control groups.172, 173 There was no evidence of increased risk of developing depression or suicidal ideation among those randomized to lorcaserin (2% to 3% and 1%, respectively) compared with placebo (2% and 1%) (p-values=NR).172, 173

Other potential harms that are listed in the “warnings and precautions” section of the lorcaserin label include serotonin syndrome, valvular heart disease, cognitive impairment, and priapism. There were no reports of serotonin syndrome and two studies that conducted echocardiograms during the trial did not report any increased incidence of valvular heart disease in those given lorcaserin.172, 173 There were scarce to no data on cognitive impairment, psychiatric disorders beyond depression, or priapism.

Naltrexone and Bupropion

Three trials reported harms (or lack of harms) related to naltrexone and bupropion (n=3453).218, 244, 311 More participants randomized to naltrexone and bupropion experienced at least one adverse event (83% to 86%) compared with those on placebo (69% to 75%) over 12 to 13 months, although statistical testing was not reported (Table 16).218, 244 The most frequent treatment-related adverse events during the primary treatment period were nausea, constipation, headache, dry mouth, and dizziness.218, 244, 311 Serious adverse events were rare and low in both the naltrexone and bupropion (0.3% to 2%) and placebo arms (0% to 1%) (Table 17). More participants withdrew from the naltrexone and bupropion group for adverse events (20% to 25%) compared with placebo (10% to 14%), although statistical testing was not reported.218, 244, 311 Of the two trials reporting deaths (n=1948), one death occurred in the naltrexone and bupropion group.244, 311 There was no statistically significant association with depression or anxiety; 0 to 5 percent in the naltrexone and bupropion arm developed depression and/or anxiety compared with 1 to 4 percent of the placebo arm.218, 244, 311

In the “warnings and precautions” sections of the naltrexone and bupropion label, additional potential harms include seizures, neuropsychiatric events, increased blood pressure/heart rate, cognitive effects, renal dysfunction (increases in creatinine), liver dysfunction, and glaucoma. Many of these warnings come from trials of naltrexone or bupropion alone. Two of the three trials noted that naltrexone and bupropion may attenuate the positive effects of weight loss on blood pressure as the naltrexone and bupropion arm did not have as much of a decrease in blood pressure (p<0.05),244, 311 and an average 1-bpm greater increase in heart rate (p-values<0.05)218, 244 compared with controls. The other harms were not well reported; however, those with certain conditions such as history of psychiatric illness and seizures were excluded from participating in the trials.

Orlistat

Seventeen trials160, 161, 222, 226, 227, 236, 239, 246, 260, 263, 273, 292, 297-299, 304, 307 (n=10,392) and two observational studies213, 248 (n=209,993) reported harms (or lack of harms) related to orlistat. Sixteen are from the previous review, and three are new as part of this update.248, 298, 299 More participants in the orlistat groups (80% to 96%) experienced at least one adverse event compared with those in the placebo group (67% to 94%) (k=8) over 6 to 18 months; statistical testing was only reported in three studies and between-group differences were significant (Table 16) (p<0.05). The number who withdrew for adverse events was also higher in the orlistat groups (2% to 16%) compared with placebo (0% to 7%), although statistical testing was not presented (k=14). However, the number of participants with serious adverse events was low in both groups and not consistently higher in those randomized to orlistat (0% to 15%) compared with placebo (2% to 26%) (Table 17) (p=NR; k=13). Six trials reported on deaths, but only 0 to 1 deaths were reported in each trial, and none were felt to be related to orlistat. One prescription event monitoring study of 16,021 orlistat users reported 33 deaths (0.21%) in orlistat users over 12 months, but none were felt to be related to orlistat.213

The most commonly reported adverse events were gastrointestinal.213 Sixty-three to 91 percent of participants randomized to orlistat experienced at least one gastrointestinal adverse event (e.g., intestinal borborygmi and cramps, flatus, fecal incontinence, oily spotting, and flatus with discharge) in the first 12 to 24 months of the trials compared with 39 to 65 percent of placebo participants (k=16);160, 161, 213, 222, 226, 227, 239, 246, 260, 263, 273, 297-299, 304, 307 only three trials reported statistical testing and all found statistically significant differences (p<0.05).222, 246, 304 Similarly, compared with placebo participants, more of those randomized to orlistat dropped out because of gastrointestinal adverse events (1%–10% vs. 0%–4%; p=NR; k=12).213, 222, 226, 227, 236, 239, 246, 263, 292, 297, 299, 307 The number of serious gastrointestinal adverse events was much lower in both groups, ranging from 2 to 10 percent in the orlistat group and from 1 to 3 percent in the placebo group (k=3; p=NR).161, 248, 307 There was no clear association with dosage (i.e., 120 mg TID was not associated with more adverse events or gastrointestinal adverse events than 60 mg TID). Gastrointestinal symptoms were described as being of relatively short duration and decreased over time with continued usage. Orlistat was not related to risk of colorectal cancer in a retrospective cohort (n=193,972) designed to examine this association.248

One trial (n=551) found an increased incidence of musculoskeletal problems/injuries in those randomized to orlistat (23%) compared with placebo (16%) (p≤0.05).222 In six trials, more of those randomized to orlistat had episodes of beta carotene or vitamins A, D, or E deficiency (low levels or need for supplementation) compared with those given placebo (0%–12% vs. 0%–8%);161, 239, 246, 292, 297, 307 however, statistical testing was only reported in one trial, which noted differences were significant for beta carotene and vitamin E (p-values<0.01).246

On the orlistat label, there are warnings for specific conditions, including cholelithiasis, liver injury, and impaired renal function. Four trials (n=1230) reported cholelithiasis or cholecystectomy events, with most trials having only 1 or 2 cases.226, 239, 292, 298 One trial (n=222) conducted gallbladder ultrasounds at baseline and at treatment end, reporting that 7 percent of those randomized to orlistat and 11 percent of those on placebo developed gallbladder abnormalities (mostly asymptomatic stones detected by ultrasound; p=NR) over 12 months.239 In the prescription event-monitoring study of 1602 orlistat users, there were 12 reported cases of abnormal liver tests;213 one subject (arm not identified) withdrew in a different trial because of a liver disorder.260 One trial that conducted kidney ultrasounds at baseline and end of treatment noted that renal abnormalities (mainly stones and cysts) developed in 3 percent of those on orlistat and 2 percent of those on placebo over 12 months;239 in a second trial, there was one person in the orlistat arm with a kidney stone exacerbation on orlistat.298

Phentermine and Topiramate

Three trials reported harms (or lack of harms) related to phentermine and topiramate (n=3837).168, 216, 241 In the only trial reporting this outcome, 85 percent of those on 15/92 mg phentermine and topiramate and 73 percent of those on placebo experienced at least one adverse event (p=NR) (Table 16).216 Serious adverse events were rare in all three trials (2%–5%, 1%–3%, and 0%–4% of those randomized to phentermine and topiramate 15/92 mg, 7.5/46 mg, and placebo, respectively); statistical testing was only reported in one study, with no between-group differences noted (Table 17).241 More participants withdrew from phentermine and topiramate (16%–21% on 15/92 mg and 12%–15% on 7.5/46 mg) compared with placebo (7% to 9%) over 6 to 12 months (p=NR).168, 216, 241 In the two trials reporting mortality, only one cardiovascular-related death was reported in the placebo arm of one trial.168, 241 Those randomized to 15/92 mg of phentermine and topiramate had more anxiety (4%) than those in the control arm (1% to 2%) (p≤0.01); the 7.5/46-mg dosage was not associated with increased anxiety compared with control.216, 241 Results on the effects on depression were mixed, with one trial finding a significantly increased incidence of depression in those randomized to 15/92 mg of phentermine and topiramate (5%) compared with placebo (1%) (p=0.0007);216 however, a second trial did not find a significant difference between groups (4% of those on 15/92 mg, 3% of those on 7.5/46 mg, and 3% of those on placebo had incident depression; p=0.90).241 There were no suicide attempts or ideation in either study.

The “warnings and precautions” section of the phentermine and topiramate label list additional potential harms, including fetal toxicity, myopia, mood disorders (irritability), cognitive dysfunction (attention), elevated heart rate, metabolic acidosis, and elevated creatinine. Fetal toxicity risk is based on registries and epidemiologic studies of topiramate alone (with an FDA REMS in place). There was a slightly higher incidence of blurred vision in those on 15/92 mg phentermine and topiramate (5% to 6%) compared with placebo (3% to 5%), but between-group differences were only significant in one of two studies reporting statistical significance; the 7.5/46-mg dose was not associated with an increased risk.168, 216, 241 Irritability and insomnia were higher in the phentermine and topiramate arms (2%–5% and 6%–12%, respectively) compared with placebo (<1%–2% and 5%–6%; p≤0.05).168, 216, 241 Those in phentermine and topiramate arms reported more disturbance in attention (3% to 7%) compared with placebo (<1%) (p<0.001).168, 216, 241 The trials monitored heart rate with conflicting findings.168, 216, 241 Mild decreases in bicarbonate were seen more frequently in those on phentermine and topiramate (p<0.05) but less than 2 percent of all arms experienced substantial reductions.216, 241

Medication-Based Weight Loss Maintenance Interventions

All three medication-based weight loss maintenance trials reported adverse events.247, 287, 312 The adverse events during weight loss maintenance interventions were similar to those seen during weight loss trials.

Orlistat

Two trials reported harms (or lack of harms) during weight loss maintenance with orlistat.247, 287 Adverse events, especially gastrointestinal-related adverse events, were higher in those on orlistat (88% to 95% with gastrointestinal-related adverse events) than placebo (63% to 68% with gastrointestinal-related adverse events) (p≤0.01 and p=NR, respectively). The number who withdrew for adverse events, especially gastrointestinal-related adverse events, was also generally higher in the orlistat groups (5%–15% and 7%–12% for any and gastrointestinal- related adverse events, respectively) compared with placebo (3% to 5% and 0.5%), although statistical testing was not presented. However, the number of participants with serious adverse events was not statistically different in those randomized to orlistat (12%) compared with placebo (18%) in the one study reporting this outcome (Table 16).287 Few participants (<4%) in one orlistat trial required additional vitamin supplementation, and the results are not described by study arm.247

Liraglutide

One trial reported harms (or lack of harms) during weight loss maintenance with liraglutide.312 Persons given liraglutide did not experience more adverse events overall (92%) compared with those on placebo (89%), although statistical testing was not presented (Table 16). However, more participants experienced gastrointestinal adverse events (74% vs. 45% for liraglutide vs. placebo, respectively; p=NR). More participants randomized to liraglutide experienced at least one serious adverse event (4%) than those on placebo (2%), although statistical testing was not conducted (Table 17). Although overall withdrawals for adverse events were similar in the two arms (9%), more participants in the liraglutide arm withdrew for gastrointestinal adverse events (5% vs. 0%; p=NR).312

Views

  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (8.1M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...