The relationship between food insecurity and binge and night eating symptoms in prebariatric surgery patients is mediated by depressive symptoms

Abstract

Background:

Eleven percent of households in the United States experience food insecurity, which is a lack of access to adequate, desirable food for a healthy lifestyle. Although food insecurity is associated with increased risk of obesity and nonadherence to dietary management of chronic diseases such as diabetes, the correlates of food insecurity have not yet been studied in a bariatric surgery population.

Objectives:

To replicate, in a bariatric sample, previous findings that food insecurity is related to eating pathology and to test the hypothesis that this relationship is mediated by depressive symptoms.

Setting:

University hospital, United States.

Methods:

Two hundred forty bariatric surgery candidates responded to self-report measures of food insecurity and mood, night-eating, and binge-eating symptoms. The sample was 74% female and 71% white, with a mean age of 41.09 (11.84) years. Based on responses to the United States Department of Agriculture Adult Food Security Survey Model, 15.8% were categorized as food insecure and 25.8% as marginally food secure. Multiple regression models with bootstrapping for confidence interval estimates were used to explore mediation hypotheses.

Results:

Food insecurity was positively associated with symptoms of night eating and binge eating, and these relationships were cross-sectionally mediated by depressive symptoms.

Conclusions:

Food insecure bariatric candidates may be at increased risk of poorer postoperative outcomes because of lack of access to needed food and the detrimental mental health impact of this lack of access.

In 2017, 11.8% of U.S. households experienced food insecurity (FI), defined as a lack of reliable access to sufficient food for active healthy living [1,2]. A further 10.5% of households with children and 6.6% without experienced marginal food security (i.e., anxiety about continued access to food without evidence of disrupted food intake) [3]. People with FI experience higher rates of obesity and increased risk for chronic diseases such as hypertension and diabetes when controlling for demographic variables related to FI, including nonwhite race, Hispanic ethnicity, poverty, and lower educational attainment [3]. FI in adults has also been linked to mental health disorders such as depression and anxiety [4-6] and eating pathology, including binge eating, night eating, and preoccupation with shape and weight [7-11]. On physical and psychological morbidity outcomes, people with marginal food security most often resemble those with FI more than those with food security or report outcomes intermediate between the two other categories [4,10]. Two published studies have explored trends in disordered eating across levels of FI, suggesting worsening disordered eating beginning with marginal FI [10,11].

Existing hypotheses suggest a direct relationship between FI and disordered eating, due to the cognitive effects of food restriction and/or the food choices and eating behaviors associated with FI [9,12]. Food restriction, even when not initially driven by weight and shape concerns, can be a risk factor for the development and maintenance of disordered eating symptoms [9]. FI might also increase the risk of dysregulated eating because of factors in the home food environment. People with FI often rely on inexpensive, energy-dense, processed foods that are highly palatable [13], which can itself be a trigger for food cravings and disrupt the hormonal control of appetite regulation [14,15]. Some people with FI report misjudging the amount of food needed early in the month and experiencing scarcity while waiting for their next benefit period [13]. Inconsistent access to food, especially highly palatable foods, triggers binge-like behavior in animal models [16].

Alternately, the relationship between FI and disordered eating might be mediated by a third variable, such as symptoms of depression. This mediation hypothesis implies that FI causes or exacerbates depressive symptoms, which in turn cause or exacerbate disordered eating symptoms. Depression symptoms are cross-sectionally associated with FI [5,6,17,18], and prospective and naturalistic experimental studies suggest a causal relationship, with changes in FI preceding subsequent changes in mood symptoms [5,19,20]. Longitudinal evidence also supports a causal relationship between depression and dysregulated eating behaviors [21,22].

While there is growing interest in the FI/disordered eating relationship, to date there are no published studies investigating this relationship in a prebariatric population. Adherence to pre- and postoperative dietary requirements is crucial for healthy weight loss and maintenance [23]. After bariatric surgery, FI may cause poor or inconsistent dietary adherence. Postoperative depression and disordered eating symptoms are predictive of poorer postoperative weight loss [24] and greater risk of complications and rehospitalization [25]. There is evidence that that presurgical disordered eating can persist after surgery, negatively affecting postsurgical outcomes [26]. If FI is related to increased risk of mood and disordered eating symptoms in this population, patients with FI might face even more barriers to achieving optimal surgical outcomes.

The purpose of this study was to explore relationships between FI and symptoms of depression, binge eating, and night eating in a bariatric surgery-seeking population. Binge- and night-eating behaviors are common in prebariatric patients and may persist after surgery [26,27]. We predicted a positive linear relationship between disordered eating and FI [10] and that depressive symptoms would mediate this relationship, replicating an unpublished study in a community sample showing mediation of the relationship between FI and weight/shape concerns by depressive symptoms [7]. These analyses, while cross-sectional and exploratory, are a first step in generating hypotheses about mechanisms through which FI is related to negative physical and mental health outcomes. This will have implications for both preoperative treatment and postoperative follow-up care of bariatric surgical patients.

Methods

Participants and procedures

Patients seeking bariatric surgery (N = 337) at an academic medical center in [Region and US state Central Pennsylvania] completed study measures at the beginning of the required multidisciplinary presurgical program. Two hundred forty provided data on all study variables (71.2%). This was the final number for the mediation analyses. Because data on weight and height were collected at the first session of the presurgical weight management education program, which took place after the administration of self-report measures, body mass index (BMI) data were only available for patients who began the presurgical program (n = 228). Because BMI was not related to FI, it was not a covariate in mediation analyses. Most participants in the final sample of 240 completed the prebariatic program and were approved for surgery (189, 78.8%). The institutional review board of [Penn State Milton S. Hershey Medical Center] provided approval for the measures and procedures employed in this study. Participants provided informed consent for their data to be used in research.

Measures

FI status was measured using a self-administered version of the 10-item United States Department of Agriculture Adult Food Security Survey Module [1]. Responses were scored according to the United States Department of Agriculture manual, such that any positive response to an item was scored as 1, and items were summed for a raw score ranging from 0 to 10 [1]. The raw FI score (FIS) can be explored continuously or used to categorize patients according to levels of food security, with a score of 0 indicating high food security, scores of 1 to 2 indicating marginal food security, and scores of 3 to 10 reflecting low or very low food security [1].

Demographic characteristics including age, sex, education attainment, and race/ethnicity were collected from the Weight and Lifestyle Inventory [28]. In the current analyses, ethnicity, education attainment, and sex were dichotomized as follows: white versus minority racial/ethnic group, college versus less than college, and female versus male (no participants reported a nonbinary identity). BMI was computed from weight and height measured at program entry.

The Beck Depression Inventory-II (BDI) is a 21-item measure of depression symptoms over the past 2 weeks (range, 0–68) [29]. The Binge Eating Scale (BES) is a 16-item measure of behaviors and cognitions associated with binge eating (range, 0–48, cutoff = 27) [30]. The Night Eating Questionnaire (NEQ) is a 13-item measure corresponding to the Fifth edition of the Diagnostic and Statistical Manual (DSM-5) night-eating symptoms including morning anorexia, ≥25% of daily food intake occurring in the evening, eating during nighttime awakenings, degree of control over evening and nocturnal eating, and mood fluctuation during the day (range, 0–52, cutoff = 25) [31]. All 3 measures have been validated in presurgical bariatric populations and showed acceptable internal consistency in this sample using ordinal alpha coefficients appropriate for Likert-type scales with ≤5 scale points [32]: BDI: α = .94, BES: α = .90, NEQ: α = 71.

Data analyses

Regression analyses were conducted using the PROCESS macro for version 25.0 of SPSS (IBM Corp., Armonk, NY, USA), with bootstrapping used to generate 95% confidence intervals for all estimates. First, we explored the relationship between continuous FIS and BES and NEQ, while controlling for sex, ethnicity, education attainment, and age, which are associated with FI in the general population [1]. We also explored the overlap between FI categories and clinically significant BES and NEQ scores using χ² and Cramer’s V (Table 1). Finally, we conducted mediation analyses with depressive symptoms (BDI) as a mediator in the relationships between FIS and BES/NEQ.

Table 1.

Descriptive statistics: food security score, demographic variables, and depressive and eating disordered symptom scores across 3 categories of food security

	Observed range in sample	Full sample N = 240	Food secure n = 140 58.3%	Marginally food secure n = 62 25.8%	Food insecure n = 38 15.8%	One-way ANOVA F(2,237), η2
Continuous food security score*	0–9	1.12 (1.86)	0(0)	1.37 (.49)	4.82 (1.72)	-
Age, yr	17–70	41.09 (11.84)	43.62 (11.56)	38.39 (10.55)	36.16 (12.57)	8.64, .06, P < .001
BMI†	35.10–70.32	48.33 (7.56)	47.64 (7.72)	49.49 (7.65)	49.05 (6.70)	1.62, .01, P = .20
BES	0–39	13.67 (7.61)	13.98 (7.35)	12.89 (7.37)	16.33 (8.54)	2.86, .02, P = .06
NEQ	2–37	15.22 (5.65)	14.27 (4.91)	15.98 (5.38)	17.50 (7.64)	5.88, .05, P = .003
BDI	0–48	8.99 (8.61)	6.88 (6.83)	11.88 (9.79)	12.02 (10.28)	10.88, .08, P < .001
Clinically significant disordered eating	Clinical cutoff					χ2(2), Cramer’s V
NEQ	25	10 (4.17%)	3 (2.1%)	2 (3.2%)	5 (13.2%)	9.27, .20, P = .01
BES	27	15 (6.25%)	7 (5.0%)	1 (1.6%)	7 (18.4%)	12.26, .26, P = .002
Categorical covariates						χ2(2), Cramer’s V
Sex (% female)	-	180 (75.0)	104 (74.3)	45 (72.6)	31 (81.6)	1.11, .07, P = .57
Education attainment (% with college education)	-	55 (22.9)	42 (30.0)	11 (17.7)	2 (5.3)	11.62, .22, P = .003
Minority/POC ethnicity (% nonwhite race/ethnicity)	-	69 (28.7)	27 (19.3)	27 (43.5)	15 (39.5)	14.88, .25, P = .001

ANOVA = analysis of variance; BMI = body mass index; POC = person of color; BES = Binge Eating Scale; NEQ = Night Eating Scale; BDI = Beck Depression Inventory.

For continuous variables, means and standard deviations are presented. For dichotomous variables, n and percent of the total sample or food security category are presented. ANOVA: means in the same row with different subscripts are significantly different according to Tamhane’s T2.

^*Due to an error in the survey administration, 39 consecutive patients (19.4%) received a version of the survey that was missing 1 item needed to calculate the food insecurity score. For this subset, mean replacement was used for the missing item. Food insecurity scores did not differ between this group and patients who were administered the full survey (t = 1.09, d = .20, P = .25).

^†Participants who dropped out or were disqualified before beginning the presurgical program do not have BMI data; n = 228, food secure n = 134, marginally food secure n = 57, food insecure n = 37.

Results

See Table 1 for sample descriptives of the full sample and by food security category and univariate associations with categorical food security status. Average scores on the BES and NEQ were below suggested clinical cutoffs, but significantly larger proportions of patients with FI scored in the ranges for clinically significant binge and night eating.

The NEQ model found an overall predictive relationship with FIS and evidence for partial mediation of this relationship by depressive symptoms. FIS and demographic covariates predicted 10.13% of the variance in NEQ score (F_{(5, 234)} = 5.28, P < .001; B_FI = .70 [.31, 1.09], t = 3.53, P = .001). When BDI score was included in the model, variance predicted increased to 19.89% (F_{(6, 233)} = 9.77, P < .001). Both the direct and indirect paths for FIS were significant, suggesting that FIS accounted for variance in NEQ both directly and indirectly through its effect on BDI, which in turn had a large effect on NEQ (Table 2).

Table 2.

Results of BES and NEQ mediation models: Parameter estimates for total, direct, and indirect effects and all covariates

	B	95% CI	t	P
Path a: depression ~ food insecurity	1.08	[.48, 1.69]	3.53	.001
Binge eating symptoms (range, 0–48)
Intercept	9.13	[5.29, 12.96]	4.69	< .001
Path b: BES ~ depression	.34	[.24, .45]	6.34	< .001
Path c: food insecurity Total	.82	[.28, 1.36]	2.99	.003
Path c’: food insecurity Direct	.45	[−.06, .96]	1.72	.09
Indirect: food insecurity via depression	.37	[.10, .72]	2.36	.02
BES ~ ethnicity	−2.02	[−4.03, −.02]	− 1.99	.05
BES ~ sex	−.56	[−2.61, 1.49]	−.54	.59
BES ~ education	2.87	[.70, 5.04]	2.61	.01
BES ~ age	.03	[−.05, .10]	.63	.53
Night eating symptoms (range, 0–52)
Intercept	9.60	[6.78, 13.06]	6.24	<.001
Path b: NEQ ~ depression	.22	[.14, .30]	5.39	<.001
Path c: food insecurity total	.70	[.31, 1.09]	3.53	.001
Path c’: food insecurity direct	.47	[.09, .85]	2.43	.02
Indirect: food insecurity via depression	.24	[.06, .48]	2.17	.03
NEQ ~ ethnicity	1.86	[.37, 3.34]	2.46	.02
NEQ ~ sex	− 1.23	[−2.74, .29]	− 1.60	.11
NEQ ~ education	−.16	[−1.76, 1.45]	−.19	84
NEQ ~ age	.07	[.01, .13]	2.45	.02

CI = confidence interval; BES = binge eating scale; NEQ = night eating scale.

All coefficients with the exception of Path a are from the full mediation model. Ranges displayed in this table are the full score ranges possible for the measures; observed sample ranges are displayed in Table 1.

The BES model found a weaker, but still significant, correlational relationship between FIS and BES and evidence for full mediation by depressive symptoms. FIS and demographic covariates predicted 6.8% of the variance in BES score (F_{(5, 234)} = 2.87, P = .02; B_FI = .82 [.28, 1.36], t = 2.99, P = .003). With BDI score in the model, 19.6% of the variance was accounted for (F(6, 233) = 9.49, P < .001). In this model, the direct effect of the FIS on the BES score was no longer significant, and there was a significant indirect effect via BDI (Table 2).

Discussion

FI was relatively common in this sample of prebariatric surgery patients, with 41.6% of the sample categorized as either marginally food secure (25.8%) or FI (15.8%). The prevalence of clinically significant binge- and night-eating symptoms was relatively low (6.25% and 4.17%, respectively), but participants with FI were more likely to score above cutoffs for clinically significant disordered eating. In addition, continuous FIS was significantly associated with subclinical night- and binge-eating symptoms, and there was evidence that this relationship was mediated by depressive symptoms. This replicates and extends findings from a previously published study on the relationship between FI and disordered eating in the general population, which found a linear relationship between degree of FI and the prevalence of disordered eating symptoms [9]. This is the first exploration of the psychosocial correlates of FI, including disordered eating symptoms, in a prebariatric population.

The ability to draw causal conclusions about the etiology of disordered eating in our prebariatric population with FI is limited by the low prevalence of clinically significant symptoms in this sample. Future research should address the possibility that dietary restriction driven by FI might be involved in the etiology of dysregulated eating episodes in prebariatric patients [11,12]. Over time, these dysregulated eating behaviors could take on other roles, such as maladaptive coping and emotion regulation, accounting for the mediating role of depressive symptoms in the current models and the maintenance of disordered eating behaviors independent of periodic dietary restriction. Whereas there was no evidence of a direct effect of FI on binge eating when controlling for depressive symptoms, there was a direct effect on night eating. Prebariatric patients with FI may be at greater risk for night-eating behaviors due to a third variable, such as working long hours, shift work, greater likelihood of having a medical illness that affects sleep quality, or the effect of antidepressant medications on hunger and satiety. The present study did not control for these potential confounders; these variables might help to account for some of the direct effect of FIS on night-eating symptomatology that was not mediated by depressive symptoms.

The present study highlights the importance of examining and addressing FI in patients seeking bariatric surgery. However, limitations should be considered when interpreting the results. The cross-sectional design did not allow us to test the causal hypothesis implied by mediation analyses [5,19-22]. Because we did not include a measure of cognitive restraint, we could not test its direct effects and relationship with FI in predicting disordered eating behaviors. This is a direction for future research. In addition, data on household income and composition, important covariates of FI, were not available. Finally, the United States Department of Agriculture Adult Food Security Survey Module was originally developed as an interview measure, and evidence for the validity of the self-report version is limited. Preliminary evidence from our bariatric surgery–seeking population offers support for the convergent validity of the self-administered version of the measure with known sociodemographic correlates of FI including age, ethnicity, education attainment, and living in urban versus rural/suburban areas. The current findings should be replicated in other bariatric samples, including preoperative samples limited to patients who go on to complete surgery and in postoperative samples.

Conclusion

Our findings indicate that FI is common in presurgical bariatric patients and that it is related to psychological symptoms that may interfere with health and weight loss after surgery, including depressive symptoms and binge- and night-eating symptoms. Although testing for mood and eating disorder symptoms is already a standard component of the prebariatric psychological evaluation [28], there is a dearth of research on the prevalence and consequences of FI in this population. FI is of particular relevance to professionals who work with potential bariatric patients to make healthy dietary changes, prepare for specific postsurgical dietary requirements, and address psychological symptoms that contribute to unhealthy eating patterns.

When patients with FI report night eating, binge eating, and/or depression symptoms during the preoperative psychological evaluation, interventions addressing FI should be implemented alongside psychiatric or psychosocial interventions for mood and eating symptoms. At the same time, given the evidence for mediation of the FI/disordered eating symptom relationship by depressive symptoms, psychological interventions for binge- and night-eating behaviors, particularly those that involve healthy emotion regulation and coping strategies, may be most helpful for food insecure patients.