Mediation of Emotional Eating in the Effects of Self-Regulating Eating on Short- And Long-Term Weight Loss: Additional Impacts from Baseline Mood in Women with Obesity

James J. Annesi (1,2), Sara M. Powell (1)

About the authors:

1 California State University, Monterey Bay, Kinesiology Department, CA

2 Central Coast YMCA, Monterey, CA

Abstract

Objective An improved understanding of psychosocial factors related to weight loss is required to improve the continually poor outcomes of behavioral obesity treatments beyond their first several weeks or months. The aim of this investigation was to inform future research and improve obesity-treatment foci via an increased understanding of interrelations of treatment-associated changes in self-regulation, emotional eating, and weight.

Methods Stratified randomization yielded matched groups of United States-based White and Black women with obesity within self-regulation-focused or knowledge-focused treatment content lasting 1 year (both n = 47; overall M_age = 48 years). Changes in self-regulation and emotional eating over 3 months, weight over 6, 12, and 24 months, and baseline mood were assessed, and contrasted by group. Regression analyses assessed the relationship of changes in weight by self-regulation change, with mediation effects of emotional eating change also accounted.

Result Effect sizes for improvements in self-regulation, emotional eating, and weight were greater in the self-regulation-focused group on all variables. Incorporating data aggregated across both groups, change in emotional eating significantly mediated the prediction of weight changes by change in self-regulation. Baseline negative mood significantly moderated the emotional eating → weight change relationships within those mediation models.

Conclusion Findings suggest value in a self-regulation-centered obesity treatment approach in women. It also indicated that emotional eating and initial mood of participants are factors to carefully address in future behavioral treatments seeking increased short- and long-term weight losses.

doi.org/10.29102/clinhp.23003

Introduction

An improved understanding of psycho-social factors related to weight loss is needed to improve the incessantly poor outcomes of behavioral (non-surgical/ non-pharmacological) obesity treatments beyond the first several weeks/months (1). Although most such treatments are focused on educating participants in healthier eating and exercise behaviors, others are driven by established behavioral paradigms such as social cognitive theory (2) and self-regulation theory (3). In such interventions, curricula emphasize obtaining and utilizing self-regulatory skills (4) to address common lifestyle and environmental barriers to controlled eating such as time demands, slow progress, and social pressures (5). Some research suggests that emotional eating is the most important psychosocial correlate of weight change (6), especially in women (7). Impacts on weight loss based on the interaction of self-regulation and emotional eating changes are unclear, as are effects associated with initial mood.

Self-regulation-focused and knowledge-focused protocols were incorporated within a field setting to clarify both treatment and psychosocial impacts on weight in women with obesity over both the short and long term.

Hypotheses were as follows:

1. Treatment group membership will significantly predict improvements in eating-related self-regulation, emotional eating, and weight in favor of the self-regulation skill-focused (SRF) vs. the knowledge-focused (KF) group; and baseline mood will signifi-cantly improve accounting of those effects.
2. Reduction in emotional eating will significantly mediate the prediction of weight loss by increased self-regulation; and baseline mood will significantly moderate path(s) associated with effects derived from hypothesis 1-related analyses.

It was hoped that findings related to (1) the impacts of obesity treatment type (i.e., SRF vs. KF) on self-regulation, emotional eating, and weight loss; and (2) whether the relationship of increased self-regulation on short- and long-term losses in weight are mediated by change in emotional eating; would advance the aims of this study which were to inform future research and large-scale behavioral obesity-treatment targets and foci.

Method

Participants

This investigation was a re-analysis of research on community-based treatments for obesity that had different aims than the present study (8). Through stratified randomization, race was balanced across the SRF and KF groups (41 White, 6 Black, each). This was because racial differences in variables associated with emotional eating were found in women in the United States (9). The 4% of women of other racial/ethnic groups were excluded. There was also no significant group difference in age (M_overall = 47.8 years, SD = 7.8), body mass index (M_overall = 35.5 kg/m², SD = 3.1), or reported family income (Median_overall = US$ 64,500 / year [middle class]). After the protocol received institutional review board (IRB) approval, IRB-endorsed written consent was obtained from participants prior to study start. Principles of the World Medical Association Declaration of Helsinki were maintained.

Measures

Eating-related self-regulation was measured by 10 items derived from a taxonomy of theory-driven self-regulatory methods/skills (4). Response options to sentences such as “I make formal agreements with myself regarding my eating” and “I set eating goals” ranged from 1 = never to 4 = often, which were summed. Internal consistency was reported as Cronbach’s α = 0.81, with 2-week test-retest reliability at 0.74 (10). In the present sample, α = 0.77.

Emotional eating was measured by 15 items of the Emotional Eating Scale (11). Response options regarding the extent feelings such as “on edge,” “angry,” and “bored” lead to an urge to eat ranged from 0 = no desire to eat to 4 = an overwhelming urge to eat, which were summed. In women with obesity, internal consistency was reported as Cronbach’s α = 0.78, with 2-week test-retest reliability at 0.79 (11). In the present sample, α = 0.76.

Negative mood was measured by the brief form of the Profile of Mood States (12). Its 30 items equally represented dimensions of anxiety, depression, anger, fatigue, confusion, and vigor through 1–3-word items such as “anxious,” “sad,” and “vigorous.” Responses ranging from 0 = not at all to 4 = extremely were summed, with vigor-related items negatively keyed. In women, Cronbach’s α = 0.90, with 3-week test-retest reliability averaging 0.70 (12). In the present sample, α = 0.85.

Body weight was measured to the nearest 0.1 kg using a recently calibrated digital scale. Shoes and outerclothing were first removed by the participant.

Procedure

In the SRF group, processes focused primarily on the development of participants’ self-regulatory skills (e.g., goal setting/recording incremental goal progress, cognitive restructuring, relapse prevention). There were 6 30-minute (1-on-1) sessions supporting exercise starting at baseline (through month 6), and 13 50-minute small group sessions on eating behavior-change beginning at week 8 (through month 12). The SRF treatment was based on tenets of social cognitive (2) and self-regulation (3) theories.

The KF treatment focused primarily on educating parti-cipants on generally accepted principles of healthy eating and exercise. After an initial 1-on-1 meeting of 20 minutes, 16 sets of reading materials (13) were provided, spaced by 2–3 weeks (through month 12). Each was estimated to take 30 minutes to complete and followed by a 20-minute interaction with an instructor to reinforce its content. The KF treatment was based on the health belief model (14).

Both treatments consisted of approximately 13.5 participant/hours. Non-instructor staff administered structured fidelity checks on 10% of treatment sessions (indicating strong protocol compliance) and participant measurements at baseline and months 3, 6, 12, and 24.

Data analyses

Suggested criteria (15) indicated no systematic bias in the 17% of missing cases which facilitated use of the expectation-maximization algorithm for imputation and the desired intention-to-treat format. Based on the planned regression analyses, an overall sample size of 91 was required to detect a moderate effect of Cohen’s f² = 0.15 at the statistical power of 0.80 (16). Variance inflation factor scores of < 2.0 indicated acceptable multicollinea-rity. Statistical significance was set at α < 0.05 (2-tailed). SPSS version 28 incorporated the Process 4.1 macro-instructional models 4 and 14, with 20,000 percentile-method bootstrap resamples (17).

Significance of gain scores on eating-related self-regu-lation, emotional eating (baseline–month 3), and weight (baseline–months 6, 12, and 24) were first calculated. Using aggregated data, bivariate regression analyses assessed associations of group (coded: SRF = 1, KF = 0) with those score changes. Baseline negative mood score was then entered in their step 2 to determine significance of associated increases in R² values. Next, incorporating a lagged variable analysis format to enhance confidence in directionality of relationships (18), weight changes were predicted by self-regulation change, with change in emotional eating entered as a mediator. Baseline mood was then entered as a moderator of the above path(s) wherever the R² change-value(s) described above was significant.

Results

Improvements in self-regulation and weight, but not emotional eating (where p = 0.066), were significant at all temporal intervals, overall (Table 1). Effect sizes for improvements were greater in the SRF group on all variables. Using aggregated data, improvement in emotional eating was significantly predicted by group, R² = 0.06, p = 0.014. Subsequent entry of baseline mood (M = 24.76, SD = 13.25) significantly increased the explained variance, R²_change = 0.05, p = 0.024. The explained variance in change in self-regulation, R² = 0.02, p = 0.145; and change in weight over 6 months, R² = 0.29, p < 0.001; 12 months, R² = 0.18, p < 0.001; and 24 months, R² = 0.09, p = 0.003; was not significantly increased by the entry of baseline mood, p-values > 0.40.

Table 1. Descriptive statistics and gain scores in study measures, by group.

Also incorporating aggregated data, change in emotional eating significantly mediated the prediction of weight change over 6 months, 95% confidence interval (CI) = [-0.271, -0.002], 12 months, 95% CI = [-0.345, -0.002], and 24 months, 95% CI = [-0.739, -0.019], by change in self-regulation. The model R²-values were 0.30, 0.26, and 0.27, respectively, p-values < 0.001. Associated path data are given in Figures 1a, 1b, and 1c, respectively. Baseline negative mood significantly moderated the prediction of weight change by emotional eating change (paths b) over the same 3 intervals, 95% CI-values = [-0.018, -0.001], [-0.025, -0.001], and [-0.043, -0.002], respectively. The model R2-values were 0.38, 0.36, and 0.38, respectively, p-values < 0.001.

Data are from both groups aggregated. b–M3, change from baseline to month 3 (or another designated month). Path data are given as unadjusted beta, (its associated standard error), and [95% confidence interval].

Figure 1. Mediation of change in emotional eating in the prediction of weight changes over 6 (a), 12 (b), and 24 (c) months by eating-related self-regulation change (N=94).

Discussion

Based on the more-pronounced improvements in eating-related self-regulation, emotional eating, and weight which supported hypothesis 1, findings suggest value in a treatment approach that is primarily focused on the development of self-regulatory skills. Hypothesis 2 was also supported though the finding that much of the positive effect of increased self-regulation on weight over both the short-term and long-term was through its significant effect on emotional eating in the present sample of women with obesity (see paths a, Figure 1). That corroborates research indicating the importance of addressing emotional eating for weight change (19). The significant moderation of initial mood in the embedded emotional eating → weight change relationships further supported hypothesis 2, and suggests that mood should also be accounted for within behavioral treatments. As previously proposed (20), and substantiated by other research (21), manageable amounts of exercise might be initiated for mood-improvement purposes prior to focusing on eating behavior changes.

Limitations such as a specific sample type, possible expectation effects, self-report biases, and effects of volunteerism should be acknowledged. Nevertheless, the present research indicated importance of (1) a self-regulation-centered treatment approach, (2) an explicit focus on emotional eating, and (3) consideration of initial mood–areas that had only been indirectly addressed in cross-sectional research or studies of a briefer duration (22). Considering the field setting of the present study, its aims of informing large-scale behavioral obesity-treatment targets were met. Hopefully, extensions will account for other, possibly salient, psychosocial variables such as body image, eating disorders, and depressed mood in efforts to better-facilitate weight loss via behavioral means.

Conclusion

Findings suggest value in a self-regulation-centered obesity treatment approach in women. It also indicated that emotional eating and initial mood of participants are factors to carefully address in future behavioral treatments seeking increased short- and long-term weight losses.

Contributors: Conception and design: JJA. Acquisition of data: JJA. Analysis and interpretation of data: JJA, SMP. Drafting, revising and final approving of the article: JJA, SMP.

Competing interests: None declared.

Funding: This research received no external funding.

Patient content: Not applicable.

Ethics approval: Approval from Kennesaw State University institutional review board was obtained for the study protocol (study approval 17173). Written informed consent was obtained from all study participants. Ethical requirements of the World Medical Association Declaration of Helsinki and the American Psychological Association were upheld.

Availability of data and materials: Available from the first author upon reasonable request.

Editorial Office, WHO-CC • Clinical Health Promotion Centre • Health Science, Lund University, Sweden

Copyright © Clinical Health Promotion - Research and Best Practice for patients, staff and community, 2023