Binge Eating Disorder and the Outcome of Bariatric Surgery in a Prospective, Observational Study: Two Year Results

Abstract

Objective

We previously reported that preoperative binge eating disorder (BED) did not attenuate weight loss 12 months after bariatric surgery. This report extends our prior study by examining weight loss at 24 months.

Methods

We used a modified intention-to-treat (ITT) population to compare 24-month changes in weight among 59 participants treated with bariatric surgery, determined preoperatively to be free of a current eating disorder, with changes in 33 surgically-treated participants with BED. Changes were also compared with 49 individuals with obesity and BED who sought lifestyle modification for weight loss. Analyses included all available data points and were adjusted for covariates.

Results

At month 24, surgically-treated patients with BED preoperatively lost 18.6% of initial weight, compared with 23.9% for those without BED (p=0.049). (Mean losses at month 12 had been 21.5% and 24.2%, respectively; p=0.23.) Participants with BED who received lifestyle modification lost 5.6% at 24 months, significantly less than both groups of surgically treated patients (p<0.001).

Conclusions

These results suggest that preoperative BED attenuates long-term weight loss after bariatric surgery. We recommend that patients with this condition, as well as other eating disturbances, receive adjunctive behavioral support, the timing of which remains to be determined.

Introduction

Binge eating disorder (BED) is characterized by the consumption of an objectively large amount of food in a discrete period of time (i.e., 2 hours) with an accompanying loss of control over eating (1). Using criteria from the Diagnostic and Statistical Manual, fifth edition (DSM-5), the prevalence of BED has been estimated at 15.7% to 26.6% in individuals who are obese and seek bariatric surgery (2,3). Prior studies have yielded conflicting findings concerning whether the preoperative presence of BED is associated with lesser weight loss following surgery (4–9). Most studies have been of short duration (i.e., ≤ 12 months) and, thus, have not assessed patients at 18 months or longer post-surgery, when weight loss typically plateaus and dietary adherence declines (10,11).

We previously reported that there were no significant differences in weight loss 12 months after bariatric surgery in participants who preoperatively were determined (by the Eating Disorder Examination (12) to meet DSM-5 criteria for BED (N=33) or to be completely free of an eating disorder (N=59) (13). The two groups lost a mean of 21.5% and 24.2% of initial weight, respectively (p=0.23) (14). The present follow-up study examined weight losses of these groups an average of 24 months post-surgery. In addition, we assessed changes in the occurrence of binge eating in both groups to determine the effects of bariatric surgery on the remission (or precipitation) of this behavior.

The study also enrolled a non-randomly selected group of individuals with BED who met criteria for bariatric surgery but sought treatment with comprehensive lifestyle modification (13). The lifestyle group was included to compare outcomes of this approach with those produced by bariatric surgery, in the event that patients with BED were initially encouraged to pursue behavioral rather than surgical treatment. Participants who received this intervention lost less than half as much weight at 12 months as those with BED who underwent surgery (8.3% vs 21.5%, respectively) (14). The present study also reports 24-month changes in weight and binge eating in these lifestyle-modification participants.

Methods

Details of the study’s procedures have been reported previously (13). We recruited three groups: bariatric surgery candidates with and without BED, and a control group of patients with obesity and BED who sought weight loss with lifestyle modification. The study was approved by the University of Pennsylvania’s Institutional Review Board, and informed consent was obtained from all participants.

Bariatric Surgery Participants

Surgery participants were recruited from the Bariatric Surgery Program at the Hospital of the University of Pennsylvania (from 09/2005 to 12/2008). All candidates completed several pre-operative evaluations, including a behavioral assessment to determine that they were free of contraindications to surgery. The behavioral assessment included the Weight and Lifestyle Inventory (WALI) (15) and a clinical interview conducted by a psychologist. The clinicians informed all participants who appeared to meet criteria for BED, as assessed by the Questionnaire on Eating and Weight Patterns (16) (contained in the WALI), that we were conducting a research study to compare changes in weight, eating behavior, and other outcomes in surgical candidates with and without BED. All individuals who appeared to be free of a current eating disorder were identified in a similar manner, and both groups of candidates were directed to the study’s research coordinator to obtain additional information.

BED Diagnosis

Trained assessors, masked to the clinicians’ provisional diagnoses, administered an abbreviated version of the Eating Disorder Examination (EDE) (12) to assess the presence of BED (per DSM-5 criteria). Participants who reported, for the prior 3 months, an average of one or more binge episodes per week, and met all associated behavioral criteria, were diagnosed with BED. Those who reported no objective binge eating episodes and not more than one objective overeating episode per week during the prior 3 months -- and were free of symptoms of bulimia nervosa -- were considered free of a current eating disorder.

Surgical intervention

Surgery candidates were provided detailed information about the procedures offered, which from 2005 to 2008 were primarily Roux-en-Y gastric bypass (RYGB) and laparoscopic adjustable gastric banding (LAGB). Patients, in consultation with their surgeon, decided which operation they preferred. Both procedures were conducted laparoscopically, following standard methods described previously (13,17).

Lifestyle Modification Participants

These participants were recruited using advertisements for a behavioral weight loss program for individuals with BED. Similar to the surgery participants, enrollment was open to individuals who were at least 18 years of age and had a BMI ≥ 40 kg/m² (or ≥ 35 kg/m² in the presence of a co-morbid condition). Exclusion criteria have been described previously (13).

Respondents to advertisements completed a telephone screening. Those who appeared to meet study criteria were scheduled for an interview with a psychologist who obtained informed consent and reviewed participants’ responses to the Questionnaire on Eating and Weight Patterns (16) to determine whether they appeared to meet criteria for BED. However, this latter determination was made with the EDE following the same procedures used in selecting surgical participants.

The lifestyle modification intervention provided weekly group treatment sessions from weeks 1–20, every-other-week meetings from weeks 22–40, and monthly sessions through week 52. Sessions lasted 90 minutes, included 7–10 participants, and were led by a psychologist, following a protocol described previously (13,18). During the first 12 weeks, participants were prescribed a 1200–1300 kcal/day diet consisting of four servings daily of a liquid meal replacement (160–170 kcal/serving; HMR 800; HMR, Boston, MA) and a 250–300 kcal frozen food entrée with a cup of salad and a serving of fruit. Participants consumed additional fruits and vegetables to achieve the recommended caloric intake. At week 14, participants began to replace the liquid meal replacements with conventional foods, with a goal of consuming 1400–1600 kcal/day of self-selected conventional foods by week 18. Participants were encouraged to gradually increase their walking (or other aerobic activity) to 180 minutes per week (19).

Outcome Measures

Weight

Weight was measured on a digital scale (Detecto, model 6800A), with participants dressed in light clothing, without shoes. Weight was assessed at baseline, 1–2 weeks before surgery participants had their procedures and 1 week before lifestyle participants began treatment. It was re-measured at 2, 6, 12, 18, and 24 months post-baseline. For individuals who missed the 24-month assessment, we searched their electronic health records to identify the closest weight within 6 months of this time (which yielded weights for three participants).

Binge eating

Binge eating was assessed using the EDE (12) at the screening visit and months 2, 6, 12, 18, and 24. On each occasion, the assessors determined the number of objective binge eating episodes (OBE) and subjective binge eating (SBE) episodes participants experienced in the past 28 days. The latter episodes were characterized by the participant’s report of loss of control (LOC) eating in the absence of consuming an objectively large amount of food. We examined SBEs because, following surgery, most patients are physically unable to consume an objectively large amount of food. Studies have suggested that LOC eating after surgery may be associated with lesser weight loss, regardless of the size of binge episodes (20–23). Assessors also evaluated the number of OBEs (and related behaviors) participants reported in the past 3 months to determine whether they met full DSM-5 criteria for BED.

Statistical Analyses

The primary outcome analysis compared 24-month percentage change in baseline weight in surgically-treated participants diagnosed with and without BED at the screening visit. The main secondary outcome was the 24-month percentage change in baseline weight in participants with BED who received bariatric surgery versus lifestyle modification. A two-tailed p-value of <0.05 was used for both comparisons. Additional secondary outcomes included the percentage of participants in each group that met different categorical weight losses, as well as 24-month changes in the prevalence of BED and the occurrence of OBEs and SBEs. Only participants who completed the 24-month EDE were included in these latter analyses.

The primary outcome was examined using a modified intention-to-treat (ITT) population that included all participants who received treatment and provided at least one post-baseline measurement of body weight. (Participants without a post-baseline measurement of weight were excluded because of the lack of adequate estimates of weight change.) Changes in weight were examined using linear-mixed-effects models, which controlled for the covariates initial BMI, type of surgery (i.e., RYGB vs LAGB), age, gender, ethnicity, and presence of type 2 diabetes. Models were fit using maximum likelihood estimation (24). A piecewise model with breakpoints at month 6 and month 12 was selected for the comparisons based on model fit (25).

We used one-way analyses of variance and Fisher’s exact tests to examine 24-month differences between groups in categorical weight loss and the prevalence of BED, as well as the percentage of individuals who endorsed OBEs, SBEs, and LOC eating (i.e., either OBEs or SBEs) in the prior 28 days. We also conducted an exploratory analysis using linear mixed effects models to examine differences in 24-month weight loss between surgery patients with and without any LOC eating episodes in the prior 28 days and a more liberal categorization of any LOC in the prior 3 months. Analyses were adjusted for the same covariates listed above. Only participants who completed the 24-month EDE assessment were included in this analysis.

Results

Participant Enrollment and Retention

Participant enrollment and retention up to month 12 have been reported previously (13). In brief, 48 participants were diagnosed with BED at baseline and proceeded to have bariatric surgery. Of these, 33 (68.8%) provided ≥1 post-operative measure of weight, 27 (56.3%) completed the 12-month assessment, and 27 (56.3%) returned at month 24. Eighty participants were determined to be free of an eating disorder at baseline (i.e., non-BED participants) and had surgery. Of these, 59 (73.8%) provided ≥1 post-operative measure of weight, 51 (63.8%) returned at month 12, and weights were available for 50 (62.5%) participants at month 24 (including 2 measurements obtained by chart review). For the 51 lifestyle modification participants with BED, 49 (96.1%) provided at least 1 post-operative measure of weight, 42 (82.4%) completed the 12-month assessment, and weights were available for 41 (80.4%) participants at month 24 (including 1 measurement obtained by chart review).

Participants’ Baseline Characteristics

Baseline characteristics of participants in the two surgery groups did not differ significantly from each other (Table 1). Corresponding characteristics of the lifestyle modification group were generally similar, except there were significantly (p<0.001) more minorities in the Lifestyle-BED group (63.3%) than in the Surgery-non-BED (28.8%) and Surgery-BED groups (24.2%). Baseline BMI and weight were both significantly lower in the lifestyle modification group than in the Surgery-non-BED (p<.0.001) and Surgery-BED groups (p=0.007) (Table 1). Equal percentages of patients in the BED and non-BED groups had RYGB (57.6% vs 64.4%; p=0.517.).

Table 1.

Baseline characteristics of participants in the modified ITT population.

Characteristic	Surgery-non-BED (N = 59)	Surgery-BED (N = 33)	Lifestyle-BED (N = 49)
Sex, N (%)
Female	49 (83.1%)	24 (72.7%)	39 (79.6%)
Male	10 (16.9%)	9 (27.3%)	10 (20.4%)
Race/Ethnicity, N (%)
Black/African American	14 (23.7%)	7 (21.2%)	26 (53.1%)
Non-Hispanic White	42 (71.2%)	25 (75.8%)	18 (36.7%)
Other	3 (5.1%)	1 (3.0%)	5 (10.2%)
Education N (%)
< 12th grade	2 (3.4%)	1 (3.0%)	3 (6.1%)
High school/G.E.D.	34 (57.6%)	16 (48.5%)	26 (53.1%)
Bachelors degree	13 (22.0%)	9 (27.3%)	14 (28.6%)
Graduate or professional     degree	10 (16.9%)	7 (21.2%)	6 (12.2%)
Age (year)	43.8 ± 1.3	46.9 ± 1.7	43.8 ± 1.4
Weight (kg)	139.3 ± 3.7	140.2 ± 4.2	125.8 ± 2.9
Height (cm)	167.3 ± 1.1	169.0 ± 1.6	168.2 ± 1.2
BMI (kg/m2)	49.5 ± 1.0	49.3 ± 1.2	44.3 ± 0.7

Note: Values shown are the number of participants (%) or the mean ± standard error. Statistically significant differences are described in text. ITT = intention to treat.

24-Month Weight Loss: Surgery Participants

At month 24, surgery participants diagnosed with BED preoperatively lost a mean (±standard error) of 18.6±2.3% of baseline weight, which was significantly (p=0.049) smaller than the 23.9±1.6% lost by surgery patients without BED (Figure 1). A significantly (p=0.008) greater percentage of the non-BED (59.3%) than BED (30.3%) surgery participants lost ≥20% of initial weight (Table 2). There were no significant differences between groups in the percentage of participants who lost ≥5% (p=0.073) or ≥10% (p=0.114).

Figure 1.

Linear mixed-effects models for 24-month weight losses of participants in the three observational groups (for the modified ITT population). Values shown are mean (±standard error) percentage reduction in initial weight over 24 months for participants who had binge eating disorder (BED) and elected lifestyle modification (Lifestyle-BED; N=49) or bariatric surgery (Surgery-BED; N=33) or were free of BED and had bariatric surgery (Surgery-non-BED; N=59). Weight losses were estimated using linear mixed-effects models controlling for initial BMI, gender, type of surgery, age, ethnicity, and presence of diabetes.

Table 2.

Number of participants (%) in each of the three interventions who lost ≥5%, ≥10%, and ≥20% of initial weight at month 12 and 24.

Weight Loss Criterion	Month 12	Month 24
≥ 5% Loss
Surgery-non-BED	48 (81.4%)	49 (83.1%)
Surgery-BED	25 (75.8%)	22 (66.7%)
Lifestyle-BED	20 (40.8%)	18 (36.7%)
≥ 10% Loss
Surgery-non-BED	44 (74.6%)	45 (76.3%)
Surgery-BED	23 (69.7%)	20 (60.6%)
Lifestyle-BED	13 (26.5%)	11 (22.4%)
≥ 20% Loss
Surgery-non-BED	35 (59.3%)	35 (59.3%)
Surgery-BED	15 (45.5%)	10 (30.3%)
Lifestyle-BED	6 (12.2%)	3 (6.1%)

Note: The categorical data shown are cumulative such that the numbers of participants who lost ≥20% of initial weight is included in the numbers who lost ≥10% and ≥5%; and the number who lost ≥10% is included in those who lost ≥5%.

Factors affecting weight loss in surgery participants

Of the covariates included in the analyses, type of surgery was the only that was significantly associated with weight change. The 57 participants who received RYGB lost 28.0±1.4% of initial weight at month 24, compared to 11.1±1.8% in the 35 participants who underwent LAGB (p<0.001; Figure 2a). Exploratory analyses revealed no significant interactions between surgery type, BED status, and time (all ps >0.45). Post-hoc comparisons revealed that among RYGB-treated patients, those without preoperative BED (n=38) lost 29.6±1.6% at month 24, compared with 24.2±2.3% for those with preoperative BED (n=19) (p=0.05). Among LABG-treated patients, non-BED (n=21) and BED (n=14) patients lost 12.4±1.6% and 9.1±2.9%, respectively (p=0.375; Figure 2b). Thus, preoperative BED was associated with non-significantly smaller weight losses with both surgeries.

Figure 2.

Linear mixed-effects models for 24-month weight losses of participants who underwent laparascopic adjustable gastric banding (LAGB; N=35) vs Roux-en-Y gastric bypass (RYGB; N=57; Figure 2a). Figure 2b includes within surgery comparisons of LAGB-BED (n=14) vs LAGB-non-BED patients (n=21), as well as RYGB-BED (n=19) vs RYGB-non-BED patients (n=38). Values shown are mean (±standard error) percentage reduction in initial weight over 24 months. Weight losses were estimated using linear mixed-effects models controlling for initial BMI, gender, age, ethnicity, and presence of diabetes.

24-Month Weight Loss: Lifestyle Modification and Surgery Participants with BED

At month 24, participants with BED who received lifestyle modification lost 5.6±1.6% of baseline weight, significantly (p<0.001) less than the 18.6±2.3% lost by BED participants who had surgery (Figure 1). At month 24, a significantly greater percentage of BED participants treated by surgery than by lifestyle modification lost ≥5% (p<0.001), ≥10% (p<0.001), and ≥20% (p=0.003) of baseline weight (Table 2).

Changes in Binge Eating and Body Weight

Thirty-six participants from the original Surgery-non-BED group, 24 from the Surgery-BED group, and 29 from the Lifestyle-BED group completed the 24-month EDE assessment (i.e., 63.1% of the 141 participants in the modified-ITT population). (The 89 participants who completed the assessment did not have more binge eating episodes at baseline than the 52 non-completers, p=0.793.) At month 24, three (12.5%) of the Surgery-BED participants still met criteria for BED, as did four (14.8%) of the Lifestyle-BED participants (p=0.999). No participants in the original Surgery-non-BED group met BED criteria at month 24. Although surgery participants who met criteria for a diagnosis of BED at month 24 had lower weight losses, the difference between those with and without BED was not statistically significant (9.1±7.1% vs 20.6±1.7%, respectively, p=0.19).

Among the Surgery-BED group, three (12.5%) participants reported OBEs, and an additional 11 (45.8%) experienced SBEs in the prior 28 days, which was not statistically different from the Lifestyle-BED group in which six participants (20.7%) reported OBEs and 13 (44.8%) acknowledged SBEs (p=0.487 and 0.588, respectively). The number of Surgery-non-BED participants (n=2; 5.6%) who reported OBEs did not differ statistically from the number in the Surgery-BED group (p=0.380). Two (5.6%) Surgery-non-BED participants reported SBEs, which was significantly (p<0.001) smaller than the number in the Surgery-BED group.

The average number of OBE episodes reported in the prior month (rather than the number of individuals reporting OBEs) did not differ significantly among the three groups at month 24 (p=0.122; Figure 3a) but the number of SBEs did (p=0.001; Figure 3b). Participants in the Surgery-non-BED group reported significantly fewer SBE episodes (0.6±0.3) compared to the Surgery-BED (10.7±3.4) and Lifestyle-BED groups (12.0±3.3; p=0.001). Among patients with preoperative BED, equal percentages of patients in the LAGB and RYGB groups had postoperative SBE (50.0% vs 58.2%, p=0.682).

Figure 3.

Number of objective (a) and subjective (b) binge eating episodes in the prior 28 days at screening and 12 and 24 months.

At month 24, there were no significant differences in weight loss between surgically-treated individuals who did and did not report any type of LOC eating episodes within the prior 28 days (p=0.160). When the observation window was increased to the past 3 months, a trend towards smaller weight losses was observed in participants who did versus did not report LOC eating (15.3±2.8% vs 22.4±2.0%; p=0.051).

Discussion

This study’s principal finding was that patients with a preoperative diagnosis of BED lost significantly less weight 2 years after surgery than did individuals who were free of BED prior to surgery. Participants with BED lost 18.6% of baseline weight at year 2, compared to 23.9% for those without this disorder. One-year reductions in initial weight for these two groups were 21.5% and 24.2%, respectively, and did not differ significantly (p=0.23) (13). These results support the notion of a “honeymoon” period in the first postoperative year in which maladaptive eating habits and dietary patterns may be suppressed by the novelty and strength of surgical intervention (26). Numerous studies have shown that the mean maximum weight loss is achieved approximately 18 months post-surgery, at which time dietary adherence may decline and be accompanied by weight loss plateau or weight regain (7,8,26). The present findings suggest that patients with BED preoperatively are among those at risk of suboptimal weight loss following their 1-year postoperative visit.

Two treatment factors in this study appeared to be far more strongly associated with attenuated l long-term weight loss than was BED, although we did not randomly assign patients to interventions and did not directly test either factor. The first was patients’ receiving LAGB rather than RYGB. The mean 2-year reductions in initial weight for these groups were 11.1% and 28.0%, respectively. Several comparative and randomized controlled trials have documented similar long-term differences in weight loss between these procedures, generally leading to the decreased use of LAGB in favor of RYGB or sleeve gastrectomy (27–29).

The second treatment factor was the modest long-term weight loss achieved by patients with BED who received an intensive 1-year group lifestyle modification program. This intervention produced an 8.3% reduction in initial weight at 1 year (14), which declined to 5.6% at year 2. This 2-year loss compares favorably with that from similar behavioral interventions, such as the Diabetes Prevention Program (30) or the Look AHEAD study (31), which randomly assigned participants to intervention and control conditions. Results of this observational study suggest that bariatric surgery produces substantially greater long-term weight loss than lifestyle modification in patients with BED and is as effective as behavioral interventions in ameliorating binge eating. Multiple randomized controlled trials have demonstrated greater long-term weight loss with bariatric surgery than with lifestyle modification (32–34).

We are not aware of any other observational or controlled trials that have compared changes in BED in patients who received surgical versus behavioral treatment. Both groups in the present study appeared to have very favorable long-term improvements in their binge eating, with only 12.5% of surgery and 14.8% of lifestyle modification participants (who completed the 24-month EDE) continuing to meet full DSM-5 criteria for BED. Of patients who were free of BED prior to surgery (i.e., Surgery-non-BED), none met DSM-5 criteria for this disorder at the 1- or 2-year follow-up assessments. This latter finding provides reassurance that weight loss achieved with surgical intervention does not appear to induce BED. We note, however, that only 63.1% of patients from the modified-ITT analysis completed the EDE at year 2. Patients who did not participate in the long-term follow-up may have had poorer outcomes for both weight loss and eating behavior. In addition, surgically-treated patients typically are not physically able to consume a large amount of food post-operatively, thus, ruling out a diagnosis of BED, but they still may have other types of disordered eating. Use of additional measures, such as the Eating Disorder Examination-Bariatric Surgery Version, may be useful in detecting these complications (35).

LOC eating is a good example of an eating-related disturbance that may confront patients with preoperative BED following surgery. Although only 12.5% of patients in the BED-Surgery group continued to meet full criteria for BED at 2 years, an additional 45.8% reported SBEs in the 28 days prior to the 2-year assessment. (Similar percentages of patients in the lifestyle modification groups reported OBEs and SBEs at this time.) Several studies have reported that post-operative LOC eating (i.e., OBEs and SBEs) is associated with attenuated long-term weight loss (9,21–23). Our results support this hypothesis; surgically-treated patients who did and did not report LOC eating in the 3 months prior to the 2-year EDE assessment lost15.3% and 22.4%, respectively. This difference was significant at a trend level and could be considered clinically meaningful.

Our study had several methodological limitations, including suboptimal retention rates (with greater attrition in the surgery groups relative to lifestyle modification) that resulted in small sample sizes and decreased power to detect differences between groups. We were able to use a modified intention-to-treat analysis for the weight loss analyses but not to estimate post-operative LOC eating, due to the higher percentage of missing data for the 24-month EDE assessment. Our lack of a randomized controlled design prevents us from determining whether the type of surgery used (RYGB vs. LAGB) produced differences in long-term weight loss or whether surgery was superior to lifestyle modification in achieving weight loss in individuals with BED. Patients who sought different treatment modalities may have differed at baseline in personal characteristics that subsequently affected changes in weight and eating. The study also was not designed (or powered) to assess the potential differential effects of BED on the outcome of RYGB vs LAGB.

In summary, 2 years after surgery, patients who had BED preoperatively lost approximately 5% less of baseline weight than individuals who had been free of this disorder preoperatively (18.6% vs 23.9%). Given the generally favorable weight losses and improvements in BED experienced by patients who presented with this disorder preoperatively, we do not believe that BED should be considered a contraindication to surgery. We recommend that treatment staff routinely assess post-operative LOC eating in all patients, particularly at the 1-year follow-up visit and beyond. Bariatric-surgery patients with eating disorders (including LOC eating), determined either pre- or postoperatively, may benefit from additional counseling, such as cognitive behavior therapy. The optimal time to provide such adjunctive care remains to be determined.

What is known about this subject?

• There are conflicting findings concerning whether the preoperative presence of binge eating disorder (BED) attenuates weight loss following bariatric surgery

• Most studies have been of short duration (i.e., ≤ 12 months)

What does this study add?

• At month 24, surgically-treated patients with preoperative BED lost significantly less initial weight (18.6%) than those without preoperative BED (23.9%).

• Participants with BED who received lifestyle modification lost 5.6% at 24 months, significantly less than both groups of surgically treated patients.

• These results suggest that preoperative BED attenuates long-term weight loss after bariatric surgery.