Affective responses to overeating episodes in women participating in a behavioral weight loss program

Abstract

Background:

Much research has focused on precursors to dietary lapses in weight loss programs, but less is known about how individual responses to lapses may influence future non-adherence and program success. The current study examined affective responses to overeating lapses and their influence on subsequent overeating and overall weight loss.

Methods:

Women (n = 60) with overweight or obesity (BMI (mean ± SD): 34.3 ± 3.9 kg/m²; age: 48.1 ± 10.1 years) participated in a 3-month group behavioral weight loss intervention (BWLI). At baseline and 3 months, participants completed anthropometric assessments and a 10-day ecological momentary assessment protocol sent 5 times per day reporting on overeating and affect (stress, shame, anxiety, and feeling good about oneself). Across time points, multilevel models were used to examine affective responses to overeating and to predict likelihood of subsequent overeating. Linear regression models were used to examine the effect of affective responses to overeating (at baseline and collapsed across time points) on weight loss.

Results:

Following self-reported overeating episodes, compared to non-overeating episodes, feeling good about oneself decreased. These decreases lessened with time from overeating. Overeating predicted subsequent overeating episodes, with decreases in feeling good about oneself following overeating marginally predicting increased likelihood (p = 0.065). Neither overeating frequency at baseline nor change in overeating frequency predicted weight loss; however, greater decreases in anxiety following overeating were associated with less weight loss.

Conclusions:

Self-reported overeating during a BWLI was associated with negative affective responses and may have increased the likelihood of subsequent overeating, but did not affect overall weight loss in this sample.

1. Introduction

Overeating, which can be measured either objectively or subjectively but generally involves eating more than one’s current energy needs, is a barrier to healthy weight management and is associated with distress and negative affect among individuals with obesity (Berg et al., 2015; Goldschmidt et al., 2012; Goldschmidt et al., 2018). Within behavioral weight loss interventions (BWLIs), overeating is a specific type of dietary lapse, which is an umbrella term that encompasses many types of discrete deviations from the calorie-restricted diet prescribed in BWLIs. Dietary lapses are common in BWLIs and greater frequency of lapses has been shown to interfere with overall weight loss (Forman et al., 2017). The abstinence violation effect theorizes that negative emotional responses to engaging in an unwanted behavior may promote dejection and lead to further unwanted behaviors (Collins & Witkiewitz, 2013). This has been found following binge eating episodes in individuals with disordered eating and in those with substance use disorders following a lapse in abstinence (Grilo & Shiffman, 1994; Ward et al., 1993). However, little empirical evidence exists regarding affective responses to lapses, specifically overeating episodes (OEs), among adults with over-weight/obesity in a BWLI.

Initial work in samples of primarily female adults enrolled in BWLIs has demonstrated that lapses have been followed by feelings of guilt and failure as well as decreasing self-efficacy to lose weight and resist future temptations (Carels et al., 2004; McKee et al., 2014). The only study to date to examine momentary psychological responses to a dietary lapse as a risk factor for subsequent lapses in BWLI participants showed lower momentary self-criticism was linked to increased likelihood of having another lapse later in the day (Schumacher et al., 2018). This finding contradicts the abstinence violation effect and encourages further exploration of the role of affect in predicting future eating behavior.

The current study, which involves secondary analyses from a BWLI parent trial (R03 DK115978), expands on prior work by examining changes in additional affective variables (stress, anxiety, shame, and feeling good about oneself) following OEs. Additionally, it tests these changes in affect as predictors of subsequent OEs as well as overall weight loss. We hypothesized that 1) the occurrence of self-reported overeating would be associated with increased subsequent levels of negative affect (e.g., greater shame) and 2) greater negative affect following OEs would be associated with greater risk for subsequent OEs as well as less weight loss overall.

2. Methods

2.1. Participants

Women (n = 60) ages 18–60 with overweight or obesity (BMI of 25.0–40.0 kg/m²) were primarily recruited in Rhode Island and Massachusetts via Facebook and enrolled in 2 cohorts. Exclusion criteria included recent weight loss of ≥10 pounds, participation in another BWLI, pregnancy or plans to become pregnant, recent participation in a stress-management or mindfulness-based treatment program, and physical or mental conditions that would impair participation. Participants were required to own a smart phone, though no one was excluded based on this criteria.

2.2. Intervention

This study used a group-based (2 groups per cohort), in-person, 3-month intervention modeled after the LookAHEAD Trial (Look AHEAD Research Group, 2006). Participants received calorie goals (1200–1800 kcal/day) and physical activity goals (progressing to 200 min/week of moderate-to-vigorous physical activity) and were taught evidence-based behavioral strategies (e.g., stimulus control, problem solving) at weekly sessions. The ninth session focused on strategies for recovering from “slips”, or deviations from planned treatment behaviors, which included addressing negative thoughts that may contribute to future slips. Participants self-monitored weight, food intake, and physical activity daily and submitted self-monitoring records weekly. Interventionists provided written feedback related to weight, nutrition, and exercise on self-monitoring diaries weekly in the first month and monthly thereafter.

2.3. Anthropometric measures

Height was measured at baseline and weight was measured at baseline and 3 months using a digital scale.

2.4. Ecological momentary assessment procedures and questionnaires

Participants completed five semi-random prompts distributed throughout the day for 10 days prior to initiation of the BWLI (baseline) and 10 days post-BWLI (3 months). Participants were sent a text message survey link and had 45 min to respond. A reminder text was sent if the link was not clicked within 10 min of receipt.

2.4.1. Eating episodes

Participants were asked if they ate since the previous survey. If they responded yes, they were asked about the most proximal episode “While you were eating, to what extent did you feel that you overate?”, a commonly used measure in EMA research (Berg et al., 2015; Goldschmidt et al., 2018; Mason et al., 2018; Panza et al., 2021). The response scale was 1 (not at all) to 7 (extremely). OEs were defined as a rating of ≥5 (Goldschmidt et al., 2018). Participants also reported how many minutes ago they had eaten (eating latency).

2.4.2. Affective responses

Participants reported on stress and shame using a scale ranging from 1 = not at all to 7 = very much so following the sentence stem “Right now I feel…”. Anxiety was assessed using the 6-item state version of the State-Trait Anxiety Inventory (Marteau & Bekker, 1992). Responses were summed for a total anxiety score (24 = highest). Participants also rated the extent to which they agreed with the statement “I feel good about myself,” which was taken from the State Self-Esteem Scale (Heatherton & Polivy, 1991), from strongly disagree (1) to strongly agree (7).

2.5. Statistical analyses

Data analysis was conducted using R 3.6.3 (R Core Team, 2013). Using all available participant data (n = 60), affective responses to OEs were compared to eating episodes that did not qualify as OEs (i.e., non-overeating episodes; non-OEs) using multilevel models nested within participant. Episode type (OE vs. non-OE) was the exposure and affective response, the outcome variable. The affective state from the previous survey within the same day was included as a covariate to control for affect prior to eating episodes. Additional covariates included cohort and day (1–10) of EMA data collection. Assessment time point and latency of survey report from eating episode were tested as moderators. A zero-inflated Poisson (ZIP) model, which utilizes both a Poisson model for count data and a logit model to predict excess zeros, was used to model shame, as 86% of all observations indicated no shame.

Multilevel models with a logit link function (and binomial distribution) were used to examine the relationship between affective response to OEs and the likelihood of a subsequent OE. A difference score nested with day (i.e., no night to morning differences) was created that represented acute change in affect pre-post eating episode reported at time t. The model included an interaction between t and acute change in affect pre-to-post eating episode. The outcome was a binary indicator of an OE at time t + 1.

Finally, a non-multilevel linear regression model was used to examine whether (1) the frequency of OEs (both baseline and change in frequency of OEs) and (2) average baseline affective responses to OEs were associated with 3-month weight loss in program completers (n = 51). An OE frequency score was calculated as the number of OEs out of total eating episodes documented. Average affective response for OEs and non-OEs by each participant for each affect variable was calculated (change in affect pre-to-post eating episode averaged across baseline surveys within participant). Baseline BMI, cohort, EMA compliance, and average levels of the affect variables were included as covariates.

3. Results

3.1. Descriptive data

Participants were 48.1 ± 10.1 (M ± SD) years and 80% were non-Hispanic White. Fifty-one participants completed the program. There were no demographic differences between participants who completed the program and participants who did not. EMA survey response rates were 91.4 ± 6.0% and 90.2 ± 9.3% at baseline and 3 months, respectively. At baseline, an average of 30.5 ± 5.7 eating episodes were documented over 10 days, with 4.7 ± 4.6 OEs. Baseline average levels of stress (2.5 ± 1.1), shame (1.3 ± 0.7), anxiety (10.3 ± 2.8), and feeling good about oneself (5.1 ± 1.4) were calculated. At 3 months, an average of 27.5 ± 6.8 eating episodes were documented over 10 days with 2.0 ± 2.8 OEs. Three-month average levels of stress (2.5 ± 1.3), shame (1.2 ± 0.6), anxiety (10.6 ± 3.2), and feeling good about oneself (5.6 ± 1.2) were calculated. Most participants (80%) experienced at least one OE at baseline (range: 0–16) and 55% experienced at least one at 3 months (range: 0–11). OEs significantly reduced from baseline to 3 months, t (1,51) = 4.42, p < 0.001. Average weight loss in the program was 6.5 ± 3.5%.

3.2. Affective responses to OEs

The intra-class correlation coefficients for stress, anxiety, shame and feeling good about oneself were 0.31, 0.59, 0.34, and 0.64, respectively, justifying the use of within-person models. As seen in Table 1, mixed models indicate that following an OE, compared to a non-OE, feeling good about oneself was lower (p < 0.001), while stress and anxiety showed no differences. ZIP models indicate that OEs were not related to shame levels; however, reports of any shame (versus no shame) were more likely to be seen following OEs than non-OEs (OR = 3.3, 95% CI = 0.14–0.72).

Table 1.

Results from multilevel models of eating episode type (overeating (OE) vs. non-overeating (non-OE)) predicting affective response presented as estimate (SE).

	Stress	Shame (conditional)	Shame (zero-inflation)	Anxiety	Feeling good
Fixed
Intercept	1.33 (0.21)	−4.55 (1.22)	−1.01 (090)	5.78 (0.55)	2.58 (0.19)
Cohort	−0.63 (0.10)	1.23 (0.72)	1.49 (0.47)**	−0.02 (0.24)	0.003 (0.07)
Intervention time point (pre vs. post)	0.01 (0.01)	0.04 (0.04)	0.19 (0.12)	0.06 (0.05)	0.04 (0.01)***
Intervention day	0.20 (0.01)*	−0.02 (0.05)	−0.02 (0.06)	0.06 (0.02)*	−0.02 (0.01)**
Affect at prior survey	0.38 (0.02)***	0.09 (0.04)*	−4.71 (1.43)***	0.40 (0.02)***	0.52 (0.02)***
Eating episode type (OE vs. non-OE)	−0.04 (0.09)	0.12 (0.09)	−1.19 (0.44)**	−0.30 (0.20)	−0.58 (0.05)***
Random
Variance of intercept	0.34	4.72	4.72	2.31	0.34

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

Average latency of reporting from eating episode was 96.9 ± 19.3 min. Latency was a moderator of feeling good about oneself only (p = 0.016). If reporting was more proximal to the eating episode, feeling good about oneself was lower following OEs than non-OEs, whereas if reporting was more distal, there were no differences between types of eating episodes in feeling good about oneself (Fig. 1a). Assessment time point did not moderate affective responses to OEs (i.e., responses did not differ between baseline and 3-month assessments).

Fig. 1.

Model results of (a) latency since eating and reports of feeling good about oneself in overeating and non-overeating episodes and (b) changes in feeling good about oneself following overeating and non-overeating episodes and the probability of subsequent overeating episodes. Feeling good about oneself ranges from 0 to 7.

3.3. Affective responses to OEs predicting subsequent OEs

OEs, versus non-OEs, were more likely to be followed by a subsequent OE (p < 0.001, OR: 2.7, 95% CI: 1.68–4.39). No affective variables moderated this effect; however, feeling good about oneself showed a trend-level interaction (B = −0.45, p = 0.065) suggesting that if feeling good about oneself decreased following an OE, the likelihood of a subsequent OE was higher than if feeling good about oneself increased (Fig. 1b).

3.4. Affective responses predicting weight loss

Neither OE frequency at baseline nor change in OE frequency from baseline to 3 months significantly predicted 3-month weight loss (p’s > 0.45). Using baseline affective responses to OEs to predict weight loss, greater reductions in anxiety following overeating at baseline were associated with less weight loss at 3-months (B = −0.58, p = 0.01).

4. Discussion

The current study assessed associations between emotional reactions following OEs and future overeating and weight loss among adult women with overweight/obesity. Results provide further understanding of how emotional responses to OEs can impact adherence and success in a BWLI. Findings suggest that OEs are associated with transient decreases in feeling good about oneself. These results are in line with many, though not all, prior studies indicating that self-efficacy declines and feelings of guilt and failure increase following dietary lapses among weight loss-seeking individuals (Carels et al., 2004; McKee et al., 2014). However, contributions from the current study indicate these decrements were short-lived and rebounded with increasing latency from the OE, denoting healthy psychological coping. The quick resolution of negative feelings about oneself may help explain why some studies have found increases in negative feelings following lapses while others have found decreases (Berg et al., 2015; Goldschmidt et al., 2018). Moreover, many studies documenting higher negative affect following lapses have focused specifically on weight loss-seeking individuals (Carels et al., 2004; McKee et al., 2014), suggesting responses may be indicative of a misalignment of behavior with weight loss goals within the context of weight loss treatment.

OEs predicted subsequent OEs. This supports the abstinence violation effect and its application to weight management (Herman & Polivy, 1980). Moreover, trend-level results suggested that not feeling badly about oneself following an OE was protective against a subsequent OE. This indicates that preventing negative feelings towards oneself following overeating may be important in inhibiting a downward spiral in behavior. Thus, treatments should consider counseling participants to expect occasional OEs, teaching strategies to counteract negative thoughts, and emphasize the importance of getting “back on track” immediately following an OE. Neither stress, anxiety, nor shame were associated with future overeating, suggesting these emotions may not be tied to subsequent overeating. The shame-related finding was surprising as shame has been identified as a precipitant to binge eating in adults with obesity (Goldschmidt et al., 2018). However, feelings of shame were infrequent in this non-clinical sample of women with obesity, reducing the ability to identify significant effects. Nonetheless, reports of no shame were more common following non-OEs than OEs, providing some support for a relationship. The lack of relationship between both stress and anxiety and overeating may be influenced by a number of factors, including differences in coping skills among participants, as some may be more adaptive than others (Kelly et al., 2012). Relatedly, some individuals eat more in response to stress while others eat less, potentially washing out stress-related effects (Torres & Nowson, 2007).

The finding indicating that greater reduction in anxiety following OEs was linked to less weight loss supports overeating as an unfavorable coping strategy. If greater food consumption is used to counteract negative emotion states and does so effectively, this may be reinforcing and perpetuate additional or larger episodes of increased calorie consumption, leading to lesser weight loss (Polivy et al., 1994; Rosenbaum & White, 2013). Non-food-related coping strategies for anxiety may be beneficial to include in treatment. Frequency of OEs was also not related to weight loss, adding to the equivocal findings in the literature (Forman et al., 2017; Goldstein et al., 2021; Latner et al., 2013). It is possible that participants offset greater calorie intake by consuming fewer calories later in the day or that individuals defined overeating differently, as a subjective measure was used.

Although a strength of this study was good EMA compliance, OEs were not objectively measured or defined. Participant perceptions of overeating are likely more tied to the emotional response that is theorized to influence future behavior in the current study. However, subjective experience does not provide objective measurement of calories which would directly impact weight, potentially explaining the lack of relationship to weight outcomes. Measurement of both constructs in future work could improve understanding of how objective and subjective overeating differ in their effects. Additionally, semi-randomly timed prompts helped reduce reporting bias, but may have led to less detailed data about OEs and, when prompts were unanswered, missed documentation of eating episodes. Finally, due to the use of existing data, variables previously associated with overeating and/or dietary lapses (e.g., sadness, guilt) could not be tested as predictors of subsequent overeating. Moreover, the existing sample was taken from an uncontrolled study, allowing for influence of non-study-related factors, and was limited in diversity of gender, age, race, and socioeconomic status similar to many studies assessing overeating/lapses. Different demographic groups may experience overeating differently (e.g., women [vs. men] more often cite emotions as a cause for weight regain (Sainsbury et al., 2019)). Future studies may include previously identified related variables and a more diverse sample to enhance generalizability (Berg et al., 2015; De Young et al., 2013).