Health risk is better established for obese persons than for overweight persons. However, overweight status also carries risk (11). Even mild to moderate overweight in young adults predicts subsequent obesity (12), and weight gain is associated with adverse outcomes (13). Visceral fat versus subcutaneous fat is particularly linked with adverse cardiovascular profiles in diverse ethnic and racial groups (14-20). Body composition varies with race and ethnicity. For example, Asian persons may be more likely (21) and African-American persons may be less likely to accumulate visceral fat than white persons (15, 22, 23). Health implications may also vary (14-20).

Estimated direct obesity costs are 5.7% of total U.S. health expenditures (24). Expected lifetime costs for cardiovascular disease and its risk factors increase by 20% with mild obesity, by 50% with moderate obesity, and by nearly 200% with severe obesity (25).

We reviewed the medical literature to determine the effectiveness of adult obesity screening—the conscious measurement of weight status to clinically address body weight—and treatment. Although obesity may seem to be an obvious condition, only 42% of obese U.S. adults report that health care professionals have advised them to lose weight (26). In 1996, the U.S. Preventive Services Task Force (USPSTF) recommended periodic height and weight measurement (7). Because of increased obesity prevalence, therapeutic changes, and accumulating evidence of associated health risk, this recommendation needed to be updated. The Research Triangle Institute–University of North Carolina Evidence-based Practice Center developed a systematic review of evidence to assist the USPSTF in this process.

Methods

Top Methods Results Discussion Author & Article Info References

We developed an analytic framework of obesity screening components with key questions and eligibility criteria (Appendix Table 1). Randomized, controlled trials (RCTs) or systematic reviews of RCTs were preferred evidence. When these were lacking, we evaluated cohort and nonrandomized controlled studies. Because long-term data were limited, we accepted pharmacotherapy efficacy trials with a minimum of 6 months' follow-up; otherwise, we required at least 12 months' follow-up. Study quality was rated by using USPSTF criteria (Appendix Table 2) (27).

We then reviewed primary literature not covered by previous reviews. At least 2 authors independently reviewed abstracts and articles, excluded those that did not meet eligibility criteria, and abstracted eligible articles. We abstracted or calculated 95% CIs for treatment efficacy from available data whenever possible. When sample size was not reported with variance (32, 33), the baseline sample was used.

The U.S. Agency for Healthcare Research and Quality funded this research. Agency staff and USPSTF members participated in the initial study design and reviewed interim analyses and the final manuscript.

Results

Top Methods Results Discussion Author & Article Info References

Although no RCTs evaluated the efficacy of obesity screening, we found studies that addressed the health risks of obesity, treatment efficacy, and the health implications of weight loss.

Weight and Health Risk

Longitudinal data showed J-shaped or U-shaped relationships between absolute mortality and BMI (34-45). Elevated risk at low BMI may partly reflect smoking (35, 37, 42) or the limitations of BMI in approximating fat mass (46). The BMI that carried the lowest mortality risk varied but was generally within the normal range for men and the normal-to-overweight range for women (34-45). Morbidity risk increased fairly linearly with BMI. Risk was strongest for cardiovascular disorders (37, 43, 47). Breast, colon, uterine, and ovarian cancer incidence increased with BMI (44, 48).

In the United States, the association between excess body weight and mortality may be weaker for African-American persons than for white persons (41, 42, 49). However, race-specific data are rare, and concerns about sample size limit conclusions. Mortality risk from excess weight may lessen with age; health risks from obesity are unclear beyond age 74 years (50).

Approaches to Screening

Body mass index, the most common screening test for obesity, is easy to measure, highly reliable, and closely correlated (r = 0.7 to 0.8) with adult body fat (7, 51, 52). Validity may vary by demographic characteristics, including ethnicity (53-55) and possibly age (51, 56). Clinical relevance is established by prospective links with diverse health outcomes (37, 40-43, 47, 57).

Waist circumference and the waist-to-hip ratio may capture increased cardiovascular risk for central adiposity, even among nonobese persons (44, 58-61). Waist circumference more closely approximates visceral adiposity, particularly in African-American persons (15, 20). Skinfold thickness measurement requires training for accuracy and so was judged undesirable [7]. We focused on BMI because 1) it is linked with the broadest range of health outcomes, 2) entry criteria for most treatment studies are BMI-based, and 3) such trials typically report weight or BMI change.

Effect of Counseling and Behavioral Interventions on Body Weight

Counseling aims to promote change in diet, exercise, or both. Behavioral interventions are strategies to help patients acquire the skills, motivations, and support to change diet and exercise patterns. For comparison with other treatments, we considered counseling for diet, exercise, or some combination, potentially with behavioral theory, in aggregate. Of importance, each counseling component included diverse options, possibly in combination. Also, although primary care–based physical activity counseling has uncertain efficacy (62), physical activity has diverse health benefits (63) and fitness may reduce obesity's cardiovascular risk (64). Previous systematic reviews found modest effects of counseling and behavioral interventions, while more recent RCTs showed consistent findings (Table 1).

In 29 trials with at least 1 year of follow-up, the U.S. NIH review found that average weight change in diet or physical activity groups (some including behavioral therapy) was 1.9 to –8.8 kg (mean, –3.3 kg), corrected for change in controls (Table 1) (11). Counseling for low-calorie diets (1000 to 1200 kcal per day) reduced body weight by an average of 8% over 3 to 12 months and decreased abdominal fat. Although very-low-calorie diets produced greater initial weight loss than low-calorie diets, results were similar beyond 1 year. Counseling for physical activity in 24 RCTs led to weight loss of 2% to 3% and reduced abdominal fat. A combination of diet and physical activity counseling produced greater reduction of weight and abdominal fat than either approach alone. Behavior therapy was a useful adjunct to diet or physical activity counseling. Longer-term efficacy depended on continued intervention.

The U.K. NHS review found that behavioral interventions, combined with diet or exercise, appeared effective, and long-term maintenance strategies were useful (29). In 24 studies, mean net weight change (intervention groups corrected for controls) was –3 kg over 12 to 60 months (Table 1). The CTFPHC review found that weight reduction was most effective during supervised dietary treatment and that patients then gradually regained weight (28). In 6 trials, net weight change was –0.2 to –4.5 kg after 24 to 84 months.

We identified 17 additional RCTs of counseling (65-82). We examined weight loss and weight loss maintenance trials separately (68, 73). Limitations included loss to follow-up (rates varied from 5% to 38%) and differential attrition between treatments. External validity concerns included volunteer enrollment versus random community sampling and poor gender and ethnic diversity.

To compare diverse programs (Appendix Table 3), we assessed intervention mode (group or individual), components (diet, exercise, behavior), and intensity (low, moderate, high). Intensity was rated by using frequency of person-to-person contact in the first 3 months. Moderate intensity was defined as monthly contact, high intensity was defined as more frequent contact, and low intensity was defined as less frequent contact.

View larger version (29K): [in this window] [in a new window]   Figure 1. Differences in mean weight loss between intervention and control groups for counseling and behavioral interventions. Only studies for which the difference in mean weight loss could be calculated are included. Error bars represent 95% CIs and are presented for studies in which those data were available. Data presented are as close as possible to 1-year follow-up. An asterisk indicates that the difference was statistically significant (P < 0.05) but there were insufficient data to calculate CIs. B = behavioral therapy; D = diet; E = exercise; EP = exercise program; EQ = exercise equipment; L = lottery entry; MR = meal replacement; SES = socioeconomic status. +++ = high intensity; ++ = moderate intensity; + = low intensity.

Because not all trials used a null control (many compared one counseling intervention with another), our treatment efficacy estimates (intervention effect minus control) may be conservative. Of 11 high-intensity interventions to promote weight loss, 6 used a true control. Four were successful (loss of 2.5 to 5.5 kg beyond controls in 12 to 54 months) (66, 67, 70, 81), and 2 showed borderline (76) or transient (69) weight reduction (Table 2). In 5 trials, 1 high-intensity intervention led to more weight loss than another (65, 72, 74, 78, 82). Moderate-intensity interventions showed mixed results (71, 79), and 2 of the 3 low-intensity weight loss interventions were ineffective (77, 83).

Successful interventions typically included 2 to 3 components (diet, exercise, and behavioral therapy). Only 1 trial (65) examined a combination of counseling and pharmacotherapy. In this trial, adding lifestyle counseling to sibutramine therapy led to a mean weight reduction of 7.3 kg (CI, 1.6 to 13.0 kg), and adding a low-calorie diet to counseling and sibutramine therapy led to a mean weight reduction of 12.8 kg (CI, 8.2 to 17.4 kg) (65).

Twelve- to 18-month and prolonged follow-up was reported in 3 high-intensity weight loss studies (67, 70, 76), 2 of which included long-term maintenance strategies (67, 76). Although participants regained weight, modest net loss ( 2 kg) was maintained for 24 to 36 months in 3 of 4 interventions (67, 70, 76).

Trials designed to maintain weight loss showed some success (68, 73). One promoted an additional 5-kg loss over 1 year (68). In another, weight-focused counseling promoted weight maintenance in 36% more participants than exercise-focused counseling (73). Overall, counseling promoted modest average weight loss (3 to 5 kg). Multicomponent, intensive interventions that included behavioral therapy most often led to weight loss. Maintenance strategies helped sustain loss.

Effect of Pharmacotherapy Interventions on Body Weight

Pharmacologic obesity treatment has changed substantially in the past decade. Safety concerns have eliminated several options. Evidence of the efficacy of sibutramine (a dopamine, norepinephrine, and serotonin reuptake inhibitor) and orlistat (a gastrointestinal lipase inhibitor) has increased. Both of these drugs, in combination with lifestyle change, are approved for people with BMIs of 30 kg/m² or more or people who have BMIs greater than 27 kg/m² along with other risk factors (for example, hypertension, diabetes, or dyslipidemia). Efficacy trials have also examined several drugs developed for non–weight-related purposes.

A recent systematic review of pharmacotherapy for obesity found that in 7 RCTs, sibutramine promoted weight loss of 2.8 to 4.2 kg over 8 to 52 weeks in healthy adults and those with controlled hypertension (31). However, participants regained weight after the treatment was discontinued. Orlistat had similar efficacy (mean loss of 3.5 kg in 10 RCTs of 1 to 2 years' duration). Phentermine (7.4-kg average loss in 1 RCT) and mazindol (3.8-kg average loss in 1 RCT) caused modest weight loss in adults who were more than 15% overweight; however, mazindol is no longer manufactured in the United States. Other small RCTs showed limited and inconsistent efficacy of diethylpropion (2 RCTs) and fluoxetine (2 RCTs).

We identified 18 additional RCTs meeting eligibility criteria (a). Seven evaluated sibutramine (32, 33, 84-88), 8 evaluated orlistat (89-96), 2 evaluated metformin (81, 97), and 1 evaluated several drugs (98). Three trials examined maintenance strategies (84, 92, 93). Attrition (3% to 50%) and poor adherence data were primary quality limitations. Generalizability issues were similar to those in the counseling trials.

In 6 weight loss trials (Figure 2) (32, 33, 85-88), sibutramine-treated participants lost 2.8 kg (CI, 1.6 to 4.0 kg) to 7.8 kg (CI, 5.9 to 9.7 kg) more than patients given a placebo (Appendix Table 4). Frequency of response, when recorded, was high. Twenty-seven percent (CI, 18% to 36%) to 65% (CI, 60% to 70%) of sibutramine-treated patients lost 5% of their body weight and 6% (CI, 1% to 10%) to 34% (CI, 26% to 40%) lost 10% (33, 85-88). Nineteen percent (CI, 9% to 29%) to 53% (CI, 36% to 70%) more drug-treated participants than control participants lost 5% of body weight, and 5% (CI, 1% to 10%) to 27% (CI, 18% to 36%) lost more than 10% of body weight.

View larger version (25K): [in this window] [in a new window]   Figure 2. Differences in mean weight loss between intervention and control groups for pharmacotherapy interventions. Only studies for which the difference in mean weight loss could be calculated are included; each arm is represented by a data point. Error bars represent 95% CIs and are presented for studies in which those data were available. Intensity of co-interventions was not assessed because most trials provided insufficient information for evaluation. An asterisk indicates that the difference was statistically significant (P < 0.05) but there were insufficient data to calculate CIs. B = behavioral therapy; BID = twice daily; D = diet; E = exercise; QD = daily; TID = 3 times daily.

In 1 trial comparing drug and lifestyle interventions, participants treated with metformin lost 2 kg more than those given a placebo but lost less than participants in the lifestyle group (81). Another trial showed no metformin effect (97). A multidrug trial showed that persons treated with sibutramine lost statistically significantly more weight (13.4 kg) than those treated with orlistat (8 kg) or metformin (9 kg) (98).

Maintenance studies showed moderate success. In 1 (84), sibutramine, taken 6 months for weight loss and 18 months for weight maintenance, promoted a net loss of 4 kg (CI, 2.4 to 5.6 kg) versus placebo. A corresponding 44% (CI, 37% to 50%) of sibutramine-treated participants versus 16% (CI, 6% to 25%) of placebo participants maintained 80% of initial weight loss. Likewise, successful dieters treated with orlistat lost more weight and over 1 year were more likely to maintain 75% of the initial amount lost than those treated with placebo (P < 0.05) (92). In a 3rd trial, participants treated with 1 or 2 years of orlistat lost "significantly more" weight over 2 years than placebo participants (93). However, during the second year, continuous orlistat was no more effective than continuous placebo, and discontinuing therapy with the drug led to excess weight gain (for example, during the second year, mean weight gain in those who discontinued orlistat therapy was 6.3 kg compared with 3.1 kg in those who took placebo throughout) (93).

Overall, pharmacotherapy with sibutramine and orlistat promoted modest mean weight loss (3 to 5 kg) beyond that of controls, and prolonged drug courses helped sustain this loss up to 2 years. Phentermine and mazindol had similar short-term efficacy but are not approved for long-term use (31). Metformin, diethylpropion, and fluoxetine showed mixed efficacy.

Surgical Approaches

Surgical obesity treatment is limited to patients with BMIs exceeding 40 kg/m² or patients with BMIs of 35 kg/m² or more who have associated severe health complications and have not responded to other treatment methods (99). Bariatric surgery is restrictive or malabsorptive, and current techniques are primarily restrictive. Gastric bypass involves complete gastric partitioning with anastomosis of the proximal gastric segment to a jejunal loop. Adjustable gastric banding involves placing an inflatable band around the stomach that can be adjusted to different diameters (100). Vertical banded gastroplasty entails partial gastric partitioning at the proximal gastric segment with placement of a gastric outlet stoma of fixed diameter (28). Practice patterns appear to be shifting away from this technique. These procedures can be performed open or laparoscopically. Although the duodenal switch procedure—a relatively new malabsorptive technique—is fairly common in practice, we found no RCTs evaluating its effectiveness.

Because of practical and ethical constraints to a true randomized, blinded, placebo-controlled trial of surgery for obesity, high-quality evidence is limited. The 3 previous systematic reviews of obesity therapy primarily examined randomized unblinded trials comparing surgical techniques (that is, trials that included no nonsurgical controls).

The U.S. NIH reviewed 5 randomized trials and found that patients who received obesity surgery lost 10 to 159 kg over 12 to 48 months (Table 1) (11). Of 7 trials reviewed by the U.K. NHS (29), 6 showed weight loss with both gastric bypass (mean reduction, 45 to 65 kg) and gastroplasty (mean reduction, 30 to 35 kg). The CTFPHC (28) analyzed 4 surgical randomized trials and 1 prospective cohort study and found a mean weight loss of 17 to 46 kg after 2 to 5 years.

We identified 3 additional randomized trials that evaluated gastric banding over 1 to 2 years (Appendix Table 5) (100-102). In addition to lack of nonsurgical controls, quality concerns included lack of co-interventions and comorbidity information. None of the trials showed statistically significantly different weight loss between groups, but all treatments promoted considerable loss (17 to >40 kg). In addition, we identified a large, controlled cohort study evaluating surgery efficacy: the Swedish Obese Subjects (SOS) study (103, 104). This study was a multicenter trial of surgical patients (equally divided among gastric banding, vertical banded gastroplasty, and gastric bypass) and nonrandomized, matched, nonsurgical controls (104). At 2 years, surgical patients had lost 28 kg (CI, 26.9 to 29.1 kg) and controls had lost 0.5 kg (CI, –0.2 to 1.2 kg). Mean weight reduction (±SD) after gastric banding, vertical banded gastroplasty, and gastric bypass was 21% ± 12%, 23% ± 10%, and 33% ± 10%, respectively. After 8 years, subset analysis showed an average weight loss of 20 kg (CI, 18.0 to 22.0 kg) in 251 surgical patients and 0.7 kg (CI, –0.8 to 2.2 kg) in 232 controls (104). Overall, surgery promoted substantial, prolonged weight loss (10 to 159 kg over 1 to 5 years) in patients with extreme obesity.

Intermediate Health Outcomes and Sustained Weight Loss

The U.S. NIH systematic review established that counseling-based weight loss (generally approximately 5 to 10 kg) can improve intermediate health outcomes such as blood pressure, glycemic control, and serum lipid levels (11). We assessed the effect of pharmacotherapy-associated weight loss on serum lipids and glucose. Since the previous drug review did not cover these outcomes, we abstracted these data from the primary literature covered by the review, in addition to more recent articles. We found mixed evidence for improved glucose tolerance with sibutramine-induced weight loss (32, 33, 84, 86, 87, 105). Orlistat generally (90, 96, 106-109) but not always (110) improved glucose levels. This inconsistency may be due in part to medication alterations accompanying weight loss. In 1 trial (90), orlistat-treated patients with diabetes were more likely to decrease or discontinue diabetes medications than controls (17% vs. 8%; P < 0.05), and glycosylated hemoglobin level decreased only when adjusted for these alterations.

Seven trials and 1 review linked orlistat with total cholesterol reduction (90, 92, 106-111). Sibutramine showed less consistent total cholesterol findings. No statistically significant drug versus placebo effect was found in 6 trials (33, 84, 86, 87, 112, 113), and improvement was found in 3 others (32, 114, 115). Orlistat was frequently but not always (116) associated with reduced low-density lipoprotein cholesterol level (90, 92-94, 96, 106-108, 110, 115), and sibutramine had inconsistent effects (32, 84-86, 90, 96, 113, 114). Neither drug consistently affected high-density lipoprotein cholesterol level (32, 33, 90, 96, 105, 113, 114, 116, 117) or triglyceride level (33, 84-87, 90, 94, 96, 105, 107, 110, 112-114).

Surgical cohort studies suggest that extensive weight loss may lead to dramatic improvements in glucose metabolism (118), lipid profiles (119, 120), and blood pressure. Of note, hypertension tended to recur within 3 to 10 years in the SOS study (121). Although weight regain accompanied this recurrence, all surgical groups had maintained at least a 20-kg average loss.

Ultimate Health Outcomes and Sustained Weight Loss

We found less evidence for effects of weight loss on ultimate (generally symptomatic) health outcomes. Limited observational data suggest that intentional weight loss in obese persons (particularly those with comorbid conditions) can reduce mortality (122, 123). Two large RCTs showed that behaviorally mediated weight loss can prevent diabetes among those with glucose intolerance (58% reduction; P < 0.05) (67, 81). A smaller reduction in diabetes incidence (31% [CI, 17% to 43%]) was seen among similar metformin-treated patients (81).

Diabetes may resolve in patients treated surgically. For example, in 2 trials (118, 120), 90% follow-up of 300 surgical patients, 50% of whom were initially glucose intolerant and 50% of whom initially had diabetes, showed that 91% had normal fasting glucose and glycosylated hemoglobin levels. However, these data are not from RCTs. Likewise, in the SOS, lower diabetes incidence over 2 years (odds ratio, 0.10 [CI, 0.03 to 0.28]) was seen in surgical patients versus nonsurgical patients (121).

Harms of Screening and Treatment

Difficulty sustaining weight loss has raised concern that cycles of loss followed by regain potentially carry risk. Observational studies examining weight cycling and mortality show mixed results (124-130). Conclusions are primarily limited by failure to distinguish between intentional and unintentional weight loss. Some studies examining weight cycling with intentional weight loss have found unfavorable effects on coronary heart disease and its risk factors (131, 132), but others have not (133, 134). This literature is further limited by joint consideration of participants with diverse baseline age or weight and measurement issues, such as self-recalled weight and problems characterizing cycling (135-137). For example, in studies not restricted to those with excess weight, some data suggest that weight-cycling risk increases inversely with BMI and so is minimized among obese persons. We did not find studies or previous reviews addressing harms of screening or counseling interventions. Some risk is probably present, particularly since stigma associated with obesity is well established (138-140).

Sibutramine and orlistat both have frequent, although not usually serious, adverse effects. Common side effects of sibutramine include insomnia, nausea, hypertension, dry mouth, dizziness, and confusion (31). In the previously reviewed studies, common adverse effects occurred in 10% to 30% of sibutramine-treated patients versus 8% to 19% of controls (31). Among recent RCTs, side effects were common (11% to 79%) (86-88), but incidence was similar across treatments. The most worrisome side effects of sibutramine are cardiovascular, including increased blood pressure (mean increase, 0 to 3.5 mm Hg [31, 86-88] or 5% [84, 88]) and heart rate (mean increase, 4 to 6.8 beats/min) (31-33, 85, 87). In 1 study (33), elevated diastolic blood pressure ( 5 mm Hg) or pulse ( 10 beats/min) occurred in 18% more sibutramine-treated participants than controls. In people with controlled hypertension, clinically significant blood pressure increases were similar across treatment groups (31), but some persons experienced marked increase in blood pressure (31, 86). When reported, dropout due to hypertension was up to 3.9% higher among those treated with sibutramine than among those not treated; overall, dropout rates for adverse events were similar in drug and placebo groups (84, 86-88).

Adverse events were reported in 7.4% to 18% more participants receiving orlistat than participants receiving placebo (31, 89, 91, 94). Most symptoms were gastrointestinal, including oily spotting, flatulence, and fecal urgency, and were reported by 22% to 95% of orlistat users (1% to 37% more often than controls) (89-92, 96). Other problems have included need for vitamin supplementation and reduced absorption of contraceptive pills (31). In recent trials, dropout due to side effects was 0% to 12% more common in orlistat-treated participants (89, 90, 92, 94, 96). The RCTs of metformin that we reviewed did not report dropouts due to drug effects. Gastrointestinal symptoms were noted to be more common (77.8 per 100 person-years vs. 30.7 per 100-person years) in 1 trial (141) and were present but transient in 4% of patients in another (97). In the latter trial, mean lactic acid levels did not rise. Previous review of other weight loss medications found no evidence of serious adverse reactions for phentermine. However, case reports suggested potentially serious side effects of pulmonary hypertension with mazindol and diethylpropion therapy and psychosis with mazindol therapy (142).

Because of data from RCTs of surgery were limited, we evaluated surgical adverse effects in case series. Adverse effects were both general (for example, need for prolonged follow-up, multivitamin supplementation) and procedure-specific. The RCTs on gastric banding did not report mortality. One showed fewer surgical complications with laparoscopic versus open procedures (100), while the 2 evaluating the site of band placement presented conflicting data about the relative safety of esophagogastric versus gastric placement (101, 102) (Appendix Table 5). Reported symptoms suggest low rates of dysphagia, hunger, vomiting, and esophagitis (101, 102). In the nonrandomized, controlled SOS study, complications were not reported by procedure (104). The postoperative mortality rate was 0.2%, and morbidity included bleeding (0.9%), wound complications (1.8%), abdominal infection (2.1%), thromboembolic events (0.8%), pulmonary symptoms (6.2%), and miscellaneous events (4.8%).

In 38 surgical case series, at least 3 evaluating vertical banded gastroplasty and gastric bypass included patients with substantial comorbid conditions (143-145). Many studies included patients with modest health problems. Generally, mortality rates were low. In 12 cohorts receiving vertical banded gastroplasty (143, 145-155), the perioperative mortality rate ranged from 0% to 1.5% (6 deaths in 1165 patients [pooled data]). Similar rates were seen among patients who underwent gastric bypass (0% to 1.5% per series) (118, 144, 149, 156-161) and those who had adjustable gastric banding (0% to 1.5%) (155, 162-176).

Morbidity was more common. The main complications of vertical banded gastroplasty were reoperation (20% to 25% over 3 to 5 years) (148, 151) and wound infection (8% to 32% of patients) (145, 148, 149). Less frequent events (<6%) included gastric leaks, stomal stenosis, and pouch dilatations. Wound infection was reported in 8% to 20% of patients who underwent gastric bypass (149, 159, 160). Single studies noted staple failure (15%) (118), vitamin B₁₂ deficiency (40%) (118), diarrhea (13%) (160), and gastrointestinal hemorrhage (3%) (149). Among patients who underwent adjustable gastric banding, morbidity often involved reoperation (1% to 20%) (102, 162, 165, 168-170, 175, 177, 178) or band dislocation, leakage, or slippage (0.4% to 8%) (100, 163-165, 167, 168, 170-172, 177, 178).

Discussion

Top Methods Results Discussion Author & Article Info References

Efficacy of Therapeutic Interventions for Obesity

Obesity is common and easy to screen for, poses a substantial health burden in the United States, and has treatment options. Although RCT evidence for long-term improved health with weight loss is limited, weight loss–associated changes in intermediate health variables suggest benefit. In the setting of escalating obesity prevalence, the importance of considering body weight in clinical practice seems clear.

With counseling, obese patients can achieve modest but clinically significant, sustained (1 to 2 years) weight loss (for example, 3 to 5 kg). Because control groups also frequently received some intervention, this estimate may be conservative. More intense programs were generally more successful, as were those incorporating behavioral therapy. Treating patients on an individual rather than a group basis appeared less important.

Sibutramine and orlistat have modest potentially prolonged effects (weight loss of 3 to 5.5 kg). These estimates do not reflect the effects of lifestyle interventions, which should accompany pharmacotherapy. Weight maintenance trials suggest that prolonged therapy with these drugs confers some benefit but that discontinuation may lead to rapid weight regain. Other drugs show inconsistent or short-term benefit. In both counseling and pharmacotherapy trials, a relatively high number of participants have achieved clinically significant (5% to 10%) weight loss.

Surgical options can promote substantial weight loss (10 to 159 kg over 1 to 5 years). Evidence from case series suggests that such loss can be achieved in patients with multiple comorbid conditions and may be prolonged. Although surgical options are appropriate only for the very obese, between 5% and 6% of U.S. adults have a BMI of 35 kg/m² or greater (179), so the number of potentially eligible persons may be substantial.

Limitations of the Literature

Limitations of previous systematic reviews included different eligibility criteria, treatment classifications, and approaches to data synthesis. In addition, aggregate values of their findings do not reflect variations in RCT sample size, length of follow-up, or treatment differences (for example, counseling intensity). There was partial but incomplete overlap in the literature covered by each review. Overall, however, findings were consistent.

Recent primary literature also had deficiencies. Among counseling and pharmacotherapy trials, internal validity was typically fair (with limitations including loss to follow-up and differential attrition between groups), although a few trials were judged to have good validity. Studies tended to report mean weight change but not frequency of response. External validity was an issue: Participants were frequently volunteers, and diversity in sex and ethnicity was limited. No counseling RCT lasted for more than 54 months. Pharmacotherapy trials were accepted with shorter follow-up periods than trials evaluating other treatment methods. Although 6- and 12-month efficacy appeared similar among these trials, their shorter duration could have inflated estimates of sustained weight loss. Surgical data were limited by lack of placebo-controlled RCT evidence; available studies often did not report response frequency, participant comorbid conditions, or co-interventions.

Finally, some studies (particularly those examining pharmacotherapy) used a "last-observation-carried-forward" analytic approach, that is, the final weight outcome available was used as the final weight for participants who dropped out of the study. Because maximal weight loss tends to occur within 6 months of intervention, this technique may overestimate the ability to sustain weight loss. Although this technique is common when a true intention-to-treat analysis is not possible, it should be combined with alternate analyses (180, 181). Many trials showed parallel analyses of trial enrollees and those who completed the trials, but few authors presented parallel "worst-case" analyses.

Harms of Intervention

Treatment appeared reasonably safe. We identified no evidence evaluating harms of counseling. Both sibutramine and orlistat had clinically significant, often mild, adverse effects in trials lasting at most 2 years; long-term adverse effects are less defined. Surgical options clearly have the highest risk. They led to death in less than 1% of patients in pooled samples, but up to 25% of patients may need reoperation over 5 years.

A systematic review of intervention costs was beyond the scope of this project. However, it is important to note that treatment options for obesity may entail considerable cost. Intensive counseling programs require a large amount of time and a substantial staffing commitment. Based on average wholesale price, 1-year supplies of orlistat (120 mg 3 times daily) and sibutramine (15 mg daily) cost $1445.40 and $1464.78 U.S., respectively (182). Surgical costs reflect both the invasive procedure and long-term follow-up. It is possible that long-term health improvements may offset these costs to some extent.

Implications for Clinical Practice and Research

Most efficacy trials reviewed here were not performed in clinical settings. Some interventions, in particular intense counseling, may be difficult to incorporate into medical practice. One option may be referral to programs that offer intense counseling with behavioral therapy. Another may be combining office-based counseling with innovative delivery of behavioral approaches, such as videotapes or Internet-delivered adjuncts. Other topics requiring future research include longer-term follow-up of the efficacy and harms of weight loss strategies (including better characterization of weight-cycling risks), postmarketing safety records of drugs, ability of interventions to alter body fat distribution, race- and ethnicity-specific health effects of purposeful reduction of central adiposity, and efficacy of weight maintenance strategies. In the interest of obesity prevention, treatment efficacy and health effects of lifestyle modification should be clarified for patients who are overweight but not obese. Finally, better estimates of the cost-effectiveness of obesity screening and treatment, including their impact on long-term health outcomes, are needed.

Long-term research on combined treatment methods in more generalized populations is also necessary. We were unable to assess treatment effectiveness by sex or ethnicity. Intervention efficacy trials have focused on white women, and observational evidence for health outcomes is derived mostly from patients of European origin. Treatment efficacy may differ with race (11, 78), and because some ethnic groups have a disproportionate prevalence of obesity, this area needs further attention.

All obesity therapies carry promise and burden, which must be balanced in clinical decision making. Counseling approaches appear the least harmful and produce modest, clinically important weight loss but entail cost in time and resources. Pharmacotherapy promotes modest additional weight loss, but long-term drug use may be needed to sustain this benefit, and long-term adverse events and appreciable cost are unknown. Only surgical options consistently result in substantial long-term weight reduction; however, they carry a low risk for severe complications and are expensive. Body size, health status, and weight loss history all may influence obesity treatment.