Eating Expectancies Before Bariatric Surgery: Assessment and Associations with Weight Loss Trajectories

Abstract

Background:

While presurgical eating behaviors have demonstrated limited prognostic value, cognitions regarding the effects of eating may serve as important predictors of weight loss outcomes following bariatric surgery. The Eating Expectancies Inventory (EEI) is a commonly used self-report measure of expected consequences of eating; however, its psychometric and predictive properties have not yet been evaluated among bariatric surgery patients.

Objectives:

This study sought to examine the factor structure and internal consistency of the EEI among bariatric surgery candidates, to examine relationships between EEI factors and measures of eating psychopathology, and to explore the effects of eating expectancies on postsurgical weight loss.

Setting:

Data originated from an interdisciplinary bariatric surgery center in the Midwest United States.

Methods:

262 women completed self-report questionnaires prior to bariatric surgery. Presurgical data and available postsurgical weights (at 6, 12, and 18 months) were obtained from medical records.

Results:

Analyses indicated that the original five-factor model was a good-to-excellent fit for the EEI data. All EEI factors demonstrated good reliability and were significantly associated with eating disorder symptoms and behaviors at baseline. Higher scores on EEI Factor 1 (negative affect) and Factor 5 (alleviates boredom) predicted poorer weight loss at 18-months post-surgery (n = 132).

Conclusions:

Findings support the reliability and validity of the EEI among female bariatric candidates. Presurgical eating expectancies were linked to pathological eating patterns and also predicted postsurgical weight loss trajectories, suggesting that eating expectancies may have prognostic value as predictors of bariatric surgery outcomes.

Bariatric surgery continues to grow as a method for durable weight loss. However, a large number of patients experience insufficient initial weight loss or significant weight regain following the procedure ⁽¹⁾, prompting efforts to identify factors that might inhibit successful outcomes. While postoperative maladaptive eating behaviors (e.g., binge eating, grazing) predict weight loss outcomes ^(2–4), evidence is inconsistent regarding the prognostic power of presurgical maladaptive eating behaviors in predicting postsurgical weight loss ^(5,6). Therefore, continued efforts to illuminate presurgical predictors of outcome remains an important area of research, with valuable clinical implications.

While presurgical eating behaviors have demonstrated limited prognostic success, patients’ cognitions or expectancies about the positive or negative effects of eating may hold promise as a predictor of subsequent eating behaviors and weight loss trajectories. Eating expectancies refer to the beliefs that one holds regarding the consequences of their eating behavior ⁽⁷⁾. Originally derived from outcome expectancy theory ^(8,9), eating expectancies are theorized to evolve out of an individual’s prior reinforcement and learning history, and to shape his or her subsequent eating behaviors ⁽¹⁰⁾. Specifically, individuals who learn to anticipate greater positive (e.g., eating is pleasurable) or negative (e.g., eating alleviates negative mood) reinforcement from eating episodes are hypothesized to be at greater risk for problematic eating behaviors such as overeating and binge eating ⁽⁷⁾.

Consistent with theory, empirical evidence indicates that eating expectancies are a robust predictor of problematic eating, suggesting that expectancy may be an important mechanism contributing to the onset and maintenance of maladaptive eating behaviors. For example, eating expectancies positively relate to concurrent disinhibited eating among overweight/obese samples ⁽¹¹⁾, and predict increases in binge eating in community samples ⁽¹²⁾. While no study to date has examined prospective associations between eating expectancies and weight loss, some work suggests an association between related constructs (e.g., eating in response to internal thoughts and feelings, equating eating with reductions in negative affect) and weight trajectories among individuals seeking weight loss ^(13,14). As bariatric candidates frequently report using food to regulate emotions or experience pleasure ^(15,16), it is possible that patients’ learned expectancies regarding the positive consequences of food intake may contribute to difficulty reducing maladaptive eating behaviors following surgery and ultimately hinder weight loss outcomes.

The Eating Expectancies Inventory (EEI) ⁽⁷⁾ is the only validated self-report measure of eating expectancies, containing five subscales to assess expectations of both positive and negative reinforcement from eating. While the EEI has demonstrated strong reliability and convergent validity in community samples ⁽⁷⁾, overweight/obese samples ⁽¹¹⁾, and those with diagnosed eating disorders ⁽⁷⁾, some research suggests variation in the factor structure across different populations ⁽¹⁷⁾. As the psychometric properties of the EEI have not yet been stringently evaluated among bariatric surgery candidates, the first goal of the current study was to test the hypothesized five-factor structure of the EEI and to explore the concurrent relationships between the EEI factors and measures of eating psychopathology among a sample of bariatric surgery candidates. Based on previous research ^(18–20), it was hypothesized that the EEI factors would demonstrate significant, moderate positive associations with general measures of eating pathology and binge eating, and smaller associations with compensatory behaviors (e.g., self-induced vomiting). The second goal of the study was to explore the impact of presurgical EEI scores on weight loss trajectories through 18 months post-surgery to determine whether eating expectancies significantly predict weight loss outcomes.

Method

Participants

Participants were 262 women seeking Roux-en-Y gastric bypass or vertical sleeve gastrectomy at an interdisciplinary bariatric surgery center at a large hospital in the Midwest United States. Average age for the sample was 45.30 years (SD = 12.80) and average pre-surgical body mass index (BMI) was 49.48 kg/m² (SD = 8.30). The sample was 80.2% White. 16.8% African American, 1.5% American Indian, 0.8% Hispanic, 0.4% Asian American, and 0.4% “Other”.

Measures

Demographic characteristics.

Demographic characteristics, including height and weight, were collected from participants’ medical records. BMI was calculated using the formula [(weight in pounds) / (height in inches)²] x 703. Weights were collected prior to surgery and at six, 12, and 18 months post-surgery.

Eating expectancies.

The Eating Expectancies Inventory ⁽⁷⁾, a 34-item self-report questionnaire intended to measure learned expectations for eating, is hypothesized to contain five factors: 1) Eating Helps Manage Negative Affect, 2) Eating is Pleasurable and Useful as a Reward, 3) Eating Leads to Feeling out of Control, 4) Eating Enhances Cognitive Competence, and 5) Eating Alleviates Boredom. Items are rated on a 7-point Likert scale, ranging from 1 “Completely Disagree” to 7 “Completely Agree.” EEI subscale scores have demonstrated good convergent validity and internal consistency in female community samples ⁽⁷⁾.

Eating disorder symptomatology.

The Eating Disorder Diagnostic Scale (EDDS) ⁽²¹⁾ is a 22-item self-report measure of disordered eating symptoms and behaviors. Items include Likert-type symptom ratings as well as weekly frequency counts for disordered eating behaviors (i.e., binge eating and compensatory behaviors) occurring over the previous three months. The scale also yields a symptom composite score representing overall severity of eating disorder symptoms ⁽²¹⁾. Previous research supports the reliability and validity of the EDDS among bariatric surgery candidates ⁽²²⁾. Internal consistency for the symptom composite score in the current sample was acceptable (α = .76).

Binge eating symptoms.

The Binge Eating Scale (BES) ⁽²³⁾ is a 16-item self-report measure of binge-eating symptoms, with higher scores reflecting more severe binge-eating-pathology ⁽²⁴⁾. The reliability and validity of BES scores has been demonstrated among bariatric surgery candidates ⁽²⁵⁾. Internal consistency for the current sample was good (α= .89).

Procedure

This research was approved by the hospital’s Institutional Review Board, and all participants provided informed consent. Participants completed questionnaires as part of their presurgical psychological evaluation. All data, including postsurgical weight, was collected, de-identified, and recorded through medical chart review for patients seen in the bariatric program between the years 2012 and 2016. The resulting sample size of 262 female patients provided adequate power for analyses in the current study ^(26,27).

Data Analysis

To test the factor structure of the EEI, a confirmatory factor analysis (CFA) was conducted using Mplus version 8.0. Procedures outlined by Hohlstein and colleagues ⁽⁷⁾ were followed when conducting the CFA, including the use of parceled items to estimate Factors 1 and 2. Missing data on the EEI (2.50%) was handled using full information maximum likelihood estimation within Mplus.

To characterize whether EEI factor means from bariatric surgery candidates significantly differ from controls, a series of one-sample t-tests were conducted using the mean scores from the “normal control” female participants presented in the original EEI validation paper ⁽⁷⁾. Convergent validity of the EEI was assessed via Pearson bivariate correlations with patient BMI and measures of disordered eating symptoms assessed at baseline (see Table S1 for descriptive data).

Weight loss data can be found in Table S1 (see supplementary materials). Percent total body weight loss (%TBWL) was calculated as (weight loss in lbs) / (presurgical weight in lbs) x 100. Percent excess body weight loss (%EBWL) was calculated as (weight loss in lbs) / (presurgical excess weight) x 100, with ideal weight calculated at a BMI of 25 kg/m². The percentage of participants lost to follow-up include 25.57% at six months (n = 195), 34.73% at 12 months (n = 171), and 49.62% at 18 months (n = 132).

A series of mixed linear models with growth curve analyses were calculated in SPSS 25.0 to examine the impact of each EEI factor on trajectories of weight loss (%TBWL) after surgery. This statistical approach was chosen in part because it can include participants with missing data. Growth parameters, including time in months (linear slope) and time quadratic (quadratic slope) after surgery, were entered as main effects into the model¹. Interaction terms between the two growth parameters and individual factors from the EEI were also entered into each model. Due to the number of trajectory analyses conducted, threshold for significance was conservatively adjusted to p < .01. Prediction curves for significant EEI factors were plotted using the following formula:

$$\%{\mathrm{TBWL}}_{\mathrm{ti}}={\mathrm{B}}_{\mathrm{int}}+{\mathrm{B}}_{\mathrm{lin}}(\text{time})+{\mathrm{B}}_{\mathrm{qua}}{(\text{time})}^2+{\mathrm{B}}_{\mathrm{linxEEI}}(\text{time})+{\mathrm{B}}_{\mathrm{quaxEEI}}{(\text{time})}^2$$

where %TBWL_ti is the percentage of excess body weight loss at each assessment time; int is the intercept value; lin is the linear slope parameter, qua is the quadratic slope parameter; and EEI is the value of each individual factor from the EEI.

Results

Confirmatory Factor Analysis and Reliability of the EEI

According to published criteria ^(28–30), results of the CFA indicated that the original five-factor model of the EEI was a good-to-excellent fit for the data: CFI = .957; RMSEA = .066 [90% C.I. = 0.05, 0.08]; SRMR = .044. Internal consistency for the EEI subscale scores was good, with Cronbach’s alphas at .79 or higher (see Table S2). Pearson bivariate correlations between all EEI factors can be found in Table 1. Consistent with previous work among college women ⁽⁷⁾, factors demonstrated medium to large positive associations with one another (r’s from .281 to .610), indicating that the individual EEI factors reflect distinct but related constructs.

Table 1.

Intercorrelations among Eating Expectancy Inventory (EEI) factors and Correlations with Convergent Measures.

	EEI Factors					Convergent Weight and Disordered Eating Variables
	1	2	3	4	5	Pre-surgical BMI	EDDS Symptom Composite	Frequency of Binge Eating	Frequency of Compensatory Behaviors	BES
Factor 1: Eating Helps Manage Negative Affect	--	.571**>	.594**	.595**	.610**	.043	.464**	.435**	−.068	.546**
Factor 2: Eating is Pleasurable and Useful as a Reward		--	.281**	.428**	.564**	.041	.265**	.413**	−.179*	.331**
Factor 3: Eating Leads to Feeling out of Control			--	.355**	.447**	−.025	.628**	.526**	.036	.665**
Factor 4: Eating Enhances Cognitive Competence				--	.396**	.079	.263**	.218*	−.069	.256**
Factor 5: Eating Alleviates Boredom					--	.037	.414**	.373**	−.049	.506**

Note. EEI = Eating Expectancies Inventory; BMI = Body Mass Index; EDDS = Eating Disorder Diagnostic Scale; BES = Binge Eating Scale.

^*p < .01;

^**p < .001.

Descriptive Characteristics and Convergent Validity with Baseline Measures

Mean scores for all EEI factors in the current sample are displayed in Table S2. Results of a series of independent sample t-tests revealed that bariatric surgery candidates reported significantly higher scores for eating in response to negative affect (Factor 1), eating when feeling out of control (Factor 3), and eating to alleviating boredom (Factor 5) compared to “normal control” female participants from the original EEI validation paper ⁽⁷⁾ (see Table S2). Alternatively, bariatric surgery candidates reported significantly fewer feelings of cognitive competence when eating (Factor 4). No significant group differences emerged for eating as a reward (Factor 2).

Results of a series of bivariate correlations with other baseline measures failed to find significant correlations between EEI factors and presurgical BMI. However, all EEI factors were positively correlated with the EDDS symptom composite score, BES total score, and binge eating frequency, suggesting that greater eating expectancies were associated with more severe eating disorder symptomology. Compensatory behaviors only exhibited a negative association with Factor 2 (eating as a reward).

Associations between Eating Expectancies and Weight Loss Trajectories

Results of the series of mixed linear models demonstrated the two EEI factors related to negative reinforcement were significantly associated with %TBWL trajectories after surgery (see Table S3). Specifically, a significant interaction emerged between EEI Factor 1 (negative affect) and the quadratic growth factor (B = −0.01, SE = 0.00, p = .008), suggesting that eating to manage negative affect predicted a slight deceleration in weight loss beginning at 12 months post-surgery. Significant interactions also emerged between EEI Factor 5 (alleviates boredom) and the linear (B = 0.17, SE = 0.05, p = .001) and quadratic (B = −0.01, SE = 0.00, p < .001) growth factors, suggesting that eating to alleviate boredom predicted accelerated weight loss through 12 months post-surgery and subsequent deceleration of weight loss through 18 months post-surgery. No other EEI subscale exhibited a relationship with postsurgical weight trajectories (p’s > .01). Interaction plots for the negative affect and alleviates boredom EEI subscales reveal that higher expectancies were associated with the poorest %TBWL outcomes at 18 months post-surgery (Figure 1).

Figure 1.

Trajectories of weight loss (%TBWL) moderated by select factors from the Eating Expectancies Inventory (EEI). Values for “High” and “Low” EEI indicate scores that are one standard deviation above (+1 SD) and one standard deviation below (−1 SD) the mean, respectively. For the expectancy that eating manages negative affect (EEI Factor 1; Figure 1A), rate of weight loss (%TBWL trajectories) began to differ at 12 months post-surgery and patients endorsing the strongest expectancies demonstrated the poorest weight loss at 18 months post-surgery. For the expectancy that eating alleviates boredom (EEI Factor 5; Figure 1B), rate of weight loss (%TBWL trajectories) significantly changed beginning at 12 months-post surgery. Patients with high expectancies that eating alleviates boredom also experienced the poorest weight loss at 18 months post-surgery.

Discussion

The current study sought to examine the psychometric properties of the Eating Expectancies Inventory ⁽⁷⁾ in a sample of women receiving bariatric surgery, and explore the extent to which presurgical eating expectancies predict postsurgical weight loss outcomes. Results replicated the scale’s proposed five-factor structure and supported the reliability and convergent validity of EEI subscale scores in bariatric candidates. Moreover, presurgical EEI subscale scores were predictive of weight loss outcomes. Specifically, individuals who endorsed stronger expectancies that eating helps to manage negative affect and to alleviate boredom at their presurgical evaluation demonstrated poorer postsurgical weight loss.

Compared with community controls, bariatric surgery candidates reported stronger beliefs that eating reduces negative emotional states (i.e., negative affect and boredom) and leads to feeling out of control. While stronger endorsement of eating expectancies captured by each of the EEI subscales was related to higher levels of eating disorder symptomatology and greater binge eating severity at the presurgical assessment, those subscales indexing expectancies of relief from negative affect and loss of control experiences demonstrated the strongest associations with disordered eating. These results are consistent with findings from ecological momentary assessment studies among individuals with obesity, which demonstrate that binge-eating episodes are followed by reductions in negative affect ⁽³¹⁾. Interpreted through the lens of expectancy theory, it is possible that these experiences of negative reinforcement lead to the development of associated expectancies, which may then contribute to the continuation of problematic eating behaviors. As binge eating is characterized by experiencing loss of control during eating episodes, it is also likely that the observed relationship between eating pathology and the expectancy that eating leads to feeling out of control may be similarly influenced by prior learning.

Importantly, EEI scores also predicted weight loss outcomes following surgery. Specifically, individuals who endorsed stronger expectancies that eating provides relief from negative affect, as well as expectancies that eating helps to alleviate boredom, experienced less total body weight loss at 18-months post-surgery than those who did not strongly endorse these expectancies. Interestingly, while holding a stronger expectancy that eating leads to feeling out of control was associated with eating pathology at the presurgical assessment, this subscale did not predict weight loss outcomes. This finding is consistent with research indicating that presurgical binge eating, characterized by loss of control eating, is not reliably related to postsurgical weight loss ^(5,6). Finally, positive reinforcement expectancies (e.g., eating is pleasurable/rewarding or increases cognitive competence) demonstrated weaker associations with presurgical eating pathology and were not related to postsurgical weight loss. Ultimately, these results suggest that negative reinforcement and affect regulation processes may be better predictors of negative outcomes within bariatric populations. Further, given previous work suggesting the importance of anticipatory reward processes (i.e., activation in reward regions of the brain when viewing palatable foods) in obese populations ^(32,33), future investigations may seek to clarify the relationships between specific eating expectancies and activation of reward pathways among obese individuals seeking bariatric surgery.

Our findings suggest that a possible mechanism for problematic postsurgical eating behaviors and weight trajectories is that patients with strong expectations that eating alleviates negative affect may continue to use food as an emotion regulation strategy following surgery. In the absence of other effective emotion regulation skills, these patients may experience greater difficulty adjusting to and maintaining postsurgical dietary recommendations. They may also be at increased risk for disordered eating and, consequently, an increased risk for suboptimal weight loss. However, further longitudinal work examining the temporal sequencing of changes in eating expectancies, affect, eating behavior, and weight are needed to test this hypothesis. Findings from the current study have strong clinical relevance, suggesting the utility of evaluating patients’ eating expectancies during presurgical assessment. Patients who demonstrate heighted negative affect eating expectancies may benefit from brief interventions that target these cognitions and support use of alternative affect regulation skills. For example, group therapy incorporating cognitive-behavioral and mindfulness techniques has demonstrated efficacy for reducing binge eating prior to surgery ^(34,35), and it is plausible that addressing eating expectancies within these groups may further improve treatment efficacy, though additional work is needed to explore this possibility.

The present study possesses a number of strengths including a large sample size, use of validated assessments, and longitudinal data-collection. However, the generalizability of results from the current study are limited by the sample’s demographic characteristics (i.e., female and predominantly White). Future work is needed to replicate these findings in male bariatric candidates and those from more diverse backgrounds. In addition, eating expectancies and eating psychopathology were only assessed prior to surgery, and additional work is needed to illuminate the potentially dynamic relationships between eating expectancies, eating behavior, and weight over time.

Conclusions

In sum, the current study supports the reliability and validity of the Eating Expectancy Inventory among female bariatric surgery candidates and indicates that individuals who eat to manage negative affect are at risk for suboptimal outcomes. Findings suggest that the EEI may hold important prognostic value and could help to identify individuals in need of targeted interventions. However, continued work is needed to further evaluate the clinical utility of eating expectancies in bariatric populations.

Highlights.

• Data were a good-to-excellent fit for the original EEI model

• All EEI factors were correlated with baseline eating disorder symptoms

• Greater negative reinforcement expectancies predicted poorer weight loss