Food Insecurity Predicts Excessive Exercise, Dietary Restriction, Cognitive Restraint, and Purging, but Not Binge Eating, in College Students Across 3 Months

ABSTRACT

Objectives

College students are at elevated risk for both food insecurity and eating disorder (ED) symptoms. Prior literature supports cross‐sectional associations between food insecurity and ED symptoms, including binge eating, purging (e.g., diuretic and laxative misuse, self‐induced vomiting), and dietary restriction. However, less is known about the temporal relation, particularly among college students.

Methods

We tested associations between food insecurity and cognitive restraint, binge eating, dietary restriction, purging, and excessive exercise across one college semester (three months). College students [N = 259; mean (SD) age = 19.22 (1.23)] were recruited to complete the Eating Pathology Symptoms Inventory (EPSI) and the 30‐day version of the United States Department of Agriculture Adult Food Security Survey Module in August (baseline) and November (follow‐up). We conducted five multiple regression models to examine baseline food insecurity as a predictor of each EPSI subscale score of interest, adjusting for baseline EPSI score, sociodemographic characteristics, and body mass index.

Results

Baseline food insecurity significantly predicted greater cognitive restraint (β = 0.12, p < 0.05), dietary restriction (β = 0.18, p < 0.001), excessive exercise (β = 0.15, p < 0.01), and purging (β = 0.14, p < 0.05) at follow‐up, adjusting for baseline levels, sociodemographic characteristics, and body mass index. Baseline food insecurity did not predict binge eating at follow‐up when the baseline level, body mass index, and sociodemographic characteristics were considered.

Discussion

Experiencing food insecurity may contribute to the development or exacerbation of excessive exercise, dietary restriction, cognitive restraint, and purging among college students. Findings highlight the potential need for food insecurity interventions to include support for disordered eating.

Summary.

• Food insecurity is defined as the inability to consistently access affordable, nutritious foods.

• Over the course of one semester (3 months) college students experiencing food insecurity endorsed greater cognitive restraint, dietary restriction, purging, and excessive exercise at follow‐up, regardless of baseline ED symptoms and sociodemographic factors.

• Efforts to screen and mitigate food insecurity on college campuses may reduce or prevent the development of eating disorder symptoms.

1. Introduction

Food insecurity, the inability to consistently access affordable, nutritious foods, is a serious international problem (Jones 2017). In the United States (US), food insecurity affects over 12% of households (Rabbitt et al. 2023). College students are at disproportionate risk: 20%–50% of students report experiencing food insecurity (Freudenberg et al. 2019). Food insecurity among college students is associated with poor academic functioning and mental health concerns, including eating disorder (ED) symptoms (Christensen et al. 2021; Zickgraf et al. 2022). Previous research has been primarily descriptive and conducted using cross‐sectional designs (e.g., Christensen et al. 2021), leaving unanswered whether food insecurity predicts future ED symptoms among college students. In the current study, we investigated the relationship between food insecurity and ED symptoms among college students across one semester.

1.1. Food Insecurity and Disordered Eating Among College Students

College students experiencing food insecurity are significantly more likely to report concurrent ED symptoms (Christensen et al. 2021; Zickgraf et al. 2022) and to screen positive for an ED (Barry et al. 2021) than students without food insecurity. Student food insecurity associates with both binge eating and weight control behaviors, such as dietary restriction not due to food scarcity (Christensen et al. 2021). Among population‐based samples of adults, food insecurity is cross‐sectionally associated with greater purging (e.g., diuretic and laxative misuse, self‐induced vomiting), excessive exercise, and dietary restriction not related to food scarcity (Becker et al. 2017; Becker et al. 2019; Middlemass et al. 2021).

The “feast or famine” theory might explain this association. It posits that binge and loss of control eating occur during food abundance (e.g., immediately following receipt of assistive benefits), while restricted eating occurs during periods of food scarcity (Hazzard et al. 2020). Indeed, adults with food insecurity report binge eating during periods of greater food availability (Hazzard, Loth, Crosby, et al. 2023) and longitudinally across 18 months (Hazzard, Loth, Fertig, et al. 2023). Importantly, restrained eating may present as either cognitive restraint or dietary restriction. Cognitive restraint includes preoccupation with restricting food intake (either by amount or type of food) while dietary restriction includes behavioral food restriction (Forbush et al. 2014). Individuals experiencing food insecurity are likely to experience both dietary restriction and cognitive restraint, due to inconsistent food availability and associated worries (Becker et al. 2017). The experience of restriction for any reason predicts later increases in food consumption and loss of control eating (Keys et al. 1950), which might also help explain why binge eating associates with food insecurity.

Food insecurity is also associated with weight control behaviors. Due to the higher caloric density and lower nutritional value of low‐cost foods, individuals experiencing food insecurity are more likely to consume a less nutritious diet and may experience unintentional excess weight gain (Morales and Berkowitz 2016), potentially prompting engagement in weight control behaviors (Hazzard et al. 2022). Although food insecurity longitudinally predicted self‐induced vomiting and diet pill use among parents of young children (Hazzard, Loth, Fertig, et al. 2023) such associations have not yet been examined in college students.

College students face unique challenges accessing food on campus. Students who pre‐purchase a meal plan have a set amount of funds to use throughout the semester. Depending on the price of on‐campus foods or the meal plan amount, funds may run out sooner than anticipated, resulting in food insecurity, which may prompt or exacerbate cognitive restraint, or increase vulnerability to binge eating once meal plan funds replenish the following semester. Additionally, because food insecurity is associated with weight gain (Carvajal‐Aldaz et al. 2022), students may become more vulnerable to weight control behaviors such as purging, dietary restriction not due to food scarcity, or excessive exercise, following and during periods of food insecurity.

Given (1) cross‐sectional associations between food insecurity and ED symptoms among college students and (2) food insecurity as a prospective risk factor for ED symptoms among non‐college adults, longitudinal examination of the relation between food insecurity and ED symptoms among college students is needed. Understanding associations could inform intervention efforts tailored to this group.

1.2. The Current Study

We examined the prospective relation between food insecurity and ED symptoms among a sample of college students across 3 months (one fall semester). This period was specified because most college students spend a full semester on campus before returning home or going elsewhere for winter breaks. Given the high prevalence of food insecurity and EDs on college campuses (Christensen et al. 2021; Zickgraf et al. 2022), we sought to understand whether food insecurity at the semester's start increased the risk for ED symptoms at the semester's end, before a student left campus and potentially had greater access to food. We hypothesized that food insecurity would longitudinally predict higher levels of binge eating, excessive exercise, dietary restriction, purging, and cognitive restraint, adjusting for baseline ED symptoms and relevant sociodemographic and personal characteristics.

2. Methods

2.1. Participants

Data come from 259 undergraduate college students from a large southeastern university in the United States. Criteria for inclusion were current enrollment at the university and age 18–25 years. Eligibility was restricted to this age range because it is the average age of college students, who are at elevated risk for both food insecurity and EDs (Barry et al. 2021; Christensen et al. 2021; Royer et al. 2021). Two participants were excluded because they did not provide baseline data on food security. Participants provided data in August 2022 (baseline) and November 2022 (follow‐up). At follow‐up, 212 participants provided data (retention rate = 81.9%). The mean (SD) age of participants was 19.22 (1.23) years. Participants' self‐identified gender breakdown was 78.0% female, 19.7% male, 1.5% nonbinary, and 0.8% not reported. Participants' self‐reported racial breakdown was 71.0% White, 4.2% Black, 16.2% Asian/Pacific Islander, 5.0% Biracial, 1.2% Middle Eastern, 0.4% American Indian, and 1.9% not reported. Participants' self‐identified ethnic breakdown included 9.7% Hispanic/Latinx, 88.4% Non‐Hispanic/Latinx, and 1.9% not reported. Most (67.2%) of participants reported receiving some form of financial aid, 35.9% reported having a job, and 20.8% were first‐generation college students. This demographic breakdown is similar to the university at large (Virginia Tech 2023). Participants' year in school was as follows: first (44.8%), second (18.1%), third (22%), fourth (13.5%), and fifth (1.5%) year.

2.2. Procedure

The study was advertised on the SONA System at Virginia Tech, an online database where students enrolled in psychology courses can receive course credit for research participation. Students were told the study was an investigation of food access and eating behaviors open to all students age 18–25 years. Interested students completed an eligibility screen; eligible students were invited to participate. Participants first provided electronic informed consent, then completed the survey online at baseline and 3 months later at follow‐up. For completion of each survey, participants received either $15 or 1 credit toward their psychology research participation requirement (depending on their preference). This study was reviewed and approved by the Virginia Tech Institutional Review Board.

2.3. Measures

2.3.1. United States Department of Agriculture (USDA) Adult Food Security Survey Module (AFSSM)

The USDA‐AFSSM is a 10‐item self‐report measure validated by the USDA Food and Nutrition Service that assesses household food security over the past 30 days (Bickel et al. 2000). Responses are summed to compute an individual's household food security score (range = 0–10). Lower scores reflect greater food security; higher scores indicate more severe food insecurity. Food security status is assigned according to the total score: 0 = High food security, 1–2 = Marginal food security, 3–5 = Low food security, 6–10 = Very low food security. For this study, participants in the marginal, low, and very low categories were coded as having food insecurity. Marginal food security was considered food insecure, given recent research that college students with marginal food security are more demographically and academically similar to students with low or very low food security than to students with high food security, suggesting they may not be food secure (Brescia and Cuite 2022). The scale had good internal consistency in the current sample (α = 0.84).

2.3.2. Eating Pathology Symptoms Inventory (EPSI)

The EPSI is a self‐report measure that evaluates ED symptoms over the past four weeks using 45 items (Forbush et al. 2013). Items are rated on a 5‐point Likert scale: “Never” to “Very Often.” Subscale scores are generated by summing corresponding individual items. Higher scores indicate more severe eating pathology. This study used the following subscales: cognitive restraint, binge eating, excessive exercise, restricting, and purging. The EPSI has demonstrated excellent validity and internal consistency (Forbush et al. 2014, 2013). In the current study, internal consistency was good to excellent (cognitive restraint baseline α = 0.81, follow‐up α = 0.75; binge eating baseline α = 0.90, follow‐up α = 0.90; dietary restriction baseline α = 0.88, follow‐up α = 0.87; excessive exercise baseline α = 0.90, follow‐up α = 0.89; purging baseline α = 0.89, follow‐up α = 0.88).

2.3.3. Body Mass Index (BMI)

Participants self‐reported height and weight, which were used to calculate their BMI (dividing kilograms by height measured in meters and squared). Self‐reported BMI has been previously shown to have good agreement with anthropometric measurements within young adults (Davies et al. 2020).

2.4. Data Analysis

Statistical analyses were performed using SPSS version 29. Before any hypothesis testing, we screened data for any violations of multivariate non‐normality, homogeneity of variance, and linearity. EPSI Purging and BMI were log transformed due to kurtosis. We then calculated descriptive statistics for key study variables. To allow for interpretation, descriptive statistics for purging and BMI are presented using non‐transformed values, but log‐transformed values are used in primary analyses testing study hypotheses. Differences in demographic characteristics and study variables by food security status (present or absent) were examined using chi‐square tests for categorical variables (e.g., gender, race, ethnicity, year in school, and status of financial aid, employment, housing, first‐generation college student) and independent samples t‐tests for continuous variables (e.g., age, current BMI, cognitive restraint, binge eating, dietary restriction, excessive exercise, purging).

Five hierarchical linear regression models were conducted to test the association of baseline food insecurity with follow‐up levels of cognitive restraint, binge eating, dietary restriction, excessive exercise, and purging. Predictor variables were entered as steps, with baseline food insecurity at Step 1, to allow for crude associations with ED outcomes. We then entered the baseline level of the dependent variable at Step 2 to adjust for prior level in assessing the contribution of food insecurity to the dependent variable. We then entered all potential confounders at Step 3 to allow for interpretation of the effect of baseline food insecurity on ED symptoms when possible demographic variables related to both food insecurity and ED symptoms are included. We included the following variables in Step 3 of the model given theoretical and empirical evidence suggesting that they may be confounders: (1) gender, given evidence that EDs (Grammer et al. 2021) and food insecurity (Ukegbu et al. 2019) are more prevalent among college students who identify as female compared with male; (2) financial aid, employment, and first‐generation college student status, as a collective correction for SES, as theory and empirical evidence suggest that low SES contributes to ED symptoms, likely through stress (Burke et al. 2023; Huryk et al. 2021), and SES is strongly associated with food insecurity (Walker et al. 2021). Age and year in school were also included because theoretically food insecurity differs by food environments and available resources, which differ based on age and year in school (i.e., living on‐campus with access to cafeteria versus living off‐campus in an apartment). Additionally, younger students are at higher risk for ED symptoms relative to older students, and living off campus is associated with an elevated risk of some ED symptoms (Lipson and Sonneville 2017); (3) race and ethnicity, as proxies for exposure to racism (Williams 1999). Racism is a causal factor in both the development of ED symptoms and for differences in food insecurity prevalence (Kwan et al. 2018; Rodgers et al. 2018). At Step 4, we included BMI to allow for the interpretation of food insecurity's prediction of ED symptoms when BMI is considered. We did not consider it appropriate to include BMI with confounder variables in Step 3 because there is evidence that weight gain occurs as a result of food insecurity and is cited as a reason for engaging in weight control behaviors among people experiencing food insecurity (Dhurandhar 2016; Taylor et al. 2020). Therefore, BMI is likely a mediator rather than a confounder of the relation. Regarding gender, 4 participants identified as nonbinary, and 2 did not report their gender. We therefore did not have the power to examine differences between nonbinary students compared with cisgender participants. To be inclusive of participants of diverse gender identities, given the importance of gender in associations with both food insecurity and EDs (Laska et al. 2021; Simone et al. 2020), we also conducted all analyses without gender as a covariate. Variance inflation factor (VIF) values were calculated to check collinearity in the regression models. Because all VIF values were under 10, we assumed there were no problems with collinearity (Fox 2019). A priori power analyses determined that for a model with 12 total predictors (including food insecurity, sociodemographic variables, and BMI) and α = 0.05, a sample of N = 213 was needed to detect a small effect (f ² = 0.05) with 90% power. As such, our sample of N = 259 was sufficient to detect the effects of our models. We report partial r as a measure of effect size [0.10 = small, 0.20 = medium, 0.30 = large; Funder and Ozer 2019)].

3. Results

3.1. Missing Data

Tables 1 and 2 report the percentage of missing data for each study variable. At baseline, 0%–2.3% of data were missing. At follow‐up, 22.4%–23.6% were missing. Little's missing completely at random (MCAR) test indicated that at baseline and follow‐up, data were missing completely at random, χ2(507) = 542.84, df = 507, p = 0.13. Missing values were estimated using expectation maximization (EM), a likelihood‐based missing data method. EM has been shown to approximate population data more accurately than other methods such as mean substitution or deleting incomplete cases (Enders 2006; Little and Rubin 1989). Using EM allowed us to fully utilize the entire sample (N = 259).

TABLE 1.

Demographic characteristics by food security status.

	% data missing	Food security (n = 178)	Food insecurity (n = 81)	df	t	p	d
Variable	Mean (SD)
Age	0	19.04 (1.20)	19.60 (1.21)	257	−3.51	< 0.001	0.46
		n (%) a			X 2	p	Phi
Race	1.9			5	9.60	0.09	0.19
White		134 (75.3)	50 (57.5)
Black		5 (2.8)	6 (7.4)
Asian or Pacific Islander		28 (15.7)	14 (17.3)
Middle Eastern		0 (0)	1 (1.2)
American Indian		1 (0.6)	2 (2.5)
Two or more races		7 (3.9)	6 (7.4)
Ethnicity	1.9			1	0.31	0.65	0.04
Hispanic/Latinx		16 (9.0)	9 (11.1)
Non‐Hispanic/Latinx		159 (89.3)	70 (86.4)
Gender	0.8			2	5.82	0.05	0.15
Male		40 (22.3)	11 (13.6)
Female		137 (77.0)	65 (80.2)
Nonbinary		1 (0.6)	3 (3.7)
Year in school	0			4	10.26	0.04	0.20
First year		89 (50.0)	27 (33.3)
Second year		31 (17.4)	16 (19.8)
Third year		35 (19.7)	22 (27.2)
Fourth year		19 (10.7)	16 (19.8)
Fifth year		4 (2.2)	0 (0)
Financial aid	0			1	9.13	0.003	0.19
Receiving financial aid		109 (61.2)	65 (76.5)
Not receiving financial aid		69 (38.8)	16 (19.8)
Employment	1.5			1	11.76	< 0.001	0.22
Employed		52 (29.2)	41 (50.6)
Not employed		124 (69.7)	38 (46.9)
Housing	2.7			1	3.61	0.07	−0.12
Off‐campus		103 (57.9)	55 (67.9)
On‐campus		72 (40.4)	22 (27.2)
First‐generation college student	1.5			1	14.59	< 0.001	0.24
No		151 (84.8)	50 (61.7)
Yes		26 (14.6)	28 (34.6)

Note: N = 259. All data were reported at baseline.

^a Categories do not add up to 100% because, for each variable, some participants did not provide information.

TABLE 2.

Scores on study variables by food security status.

	Percentage of data missing	Food security (n = 178)	Food insecurity (n = 81)	t	p	d
Variable	Mean (SD)
Baseline body mass index	1.9	22.93 (3.40)	24.13 (5.08)	−2.22	0.03	0.28
Baseline EPSI cognitive restraint	0.8	5.31 (3.29)	5.88 (3.26)	−1.30	0.19	0.17
Baseline EPSI binge eating	0.4	9.45 (6.71)	11.00 (7.62)	−1.65	0.10	0.22
Baseline EPSI restricting	2.3	6.92 (5.72)	8.31 (6.24)	−1.76	0.08	0.23
Baseline EPSI excessive exercise	1.9	8.30 (5.58)	8.48 (6.16)	−0.24	0.81	0.03
Baseline EPSI purging	0.4	1.30 (2.90)	2.79 (5.16)	−2.97	0.003	0.36
Follow‐up EPSI cognitive restraint	22.8	4.59 (2.62)	5.42 (2.75)	−2.32	0.02	0.31
Follow‐up EPSI binge eating	22.8	8.59 (6.26)	10.04 (5.43)	−1.81	0.07	0.25
Follow‐up EPSI restricting	23.6	6.17 (4.68)	8.32 (5.14)	−3.32	0.001	0.44
Follow‐up EPSI excessive exercise	22.8	6.36 (4.49)	7.88 (5.51)	−2.35	0.02	0.30
Follow‐up EPSI purging	22.4	1.03 (2.07)	2.13 (4.27)	−2.79	0.01	0.33

Note: N = 259. For all tests, df = 257.

Abbreviation: EPSI, eating pathology symptoms inventory.

3.2. Descriptive Statistics

Nearly one‐third (N = 81; 31.3%) of the sample reported experiencing food insecurity at baseline; of those, 41 (15.8%) experienced marginal food security, 26 (10%) experienced low food security, and 14 (5.4%) experienced very low food security. Full sample demographic characteristics are presented in Table 1. Compared with food secure participants, participants experiencing food insecurity were more likely to identify as female, receive financial aid, be employed, live off campus, and a first‐generation college student.

Descriptive statistics of baseline and follow‐up EPSI scores by food security status are presented in Table 2. Students experiencing food insecurity at baseline reported higher baseline BMI, greater levels of baseline dietary restriction and purging, and higher levels of follow‐up cognitive restraint, dietary restriction, excessive exercise, and purging.

3.3. Longitudinal Regressions

All longitudinal regression results are presented in Table 3. Table S1 includes results without gender included as a covariate. For all models, the pattern of results related to hypotheses was the same when gender was removed as a covariate.

TABLE 3.

Hierarchical linear regression models predicting eating disorder symptoms.

	Outcome
	Cognitive restraint			Binge eating			Dietary restriction			Excessive exercise			Purging a
	B (SE)	β	t	B (SE)	β	t	B (SE)	β	t	B (SE)	β	t	B (SE)	β	t
Step 1	R 2 = 0.02, ΔF = 5.20*, df = 1			R 2 = 0.01, ΔF = 4.15*, df = 1			R 2 = 0.04, ΔF = 10.40**, df = 1			R 2 = 0.02, ΔF = 5.32*, df = 1			R 2 = 0.04, ΔF = 11.00**, df = 1
Food insecurity	0.84 (0.37)	0.15	2.28*	1.65 (0.81)	0.13	2.04*	2.17 (0.67)	0.20	3.22**	1.56 (0.68)	0.15	2.31*	0.13 (0.04)	0.21	3.32**
Step 2	R 2 = 0.38, ΔF = 144.46***, df = 1			R 2 = 0.35, ΔF = 128.51***, df = 1			R 2 = 0.48, ΔF = 201.91***, df = 1			R 2 = 0.49, ΔF = 223.56***, df = 1			R 2 = 0.39, ΔF = 136.95***, df = 1
Food insecurity	0.47 (0.29)	0.08	1.59	0.87 (0.66)	0.07	1.31	1.34 (0.50)	0.13	2.69**	1.36 (0.49)	0.13	2.77**	0.06 (0.03)	0.10	1.97*
Baseline level of outcome	0.50 (0.04)	0.61	12.02***	0.49 (0.04)	0.59	11.34***	0.55 (0.04)	0.67	14.21***	0.58 (0.04)	0.69	14.95***	0.49 (0.04)	0.60	11.70***
Step 3	R 2 = 0.41, ΔF = 2.13*, df = 9			R 2 = 0.38, ΔF = 2.13*, df = 9			R 2 = 0.47, ΔF = 0.87, df = 9			R 2 = 0.49, ΔF = 1.20, df = 9			R 2 = 0.43, ΔF = 2.82**, df = 9
Food insecurity	0.68 (0.32)	0.12	2.14*	0.96 (0.71)	0.08	1.35	1.85 (0.55)	0.17	3.34***	1.44 (0.54)	0.14	2.67**	0.08 (0.03)	0.13	2.39*
Gender d	−0.87 (0.34)	−0.13	−2.55*	−1.26 (0.77)	−0.09	−1.63	−0.07 (0.60)	−0.01	−0.12	−0.79 (0.58)	−0.07	−1.35	−0.04 (0.04)	−0.05	−0.97
Age	−0.25 (0.25)	−0.12	−1.01	−0.13 (0.56)	−0.03	−0.23	−63 (0.43)	−0.16	−1.45	0.27 (0.42)	0.07	0.63	−0.03 (0.03)	−0.14	−1.28
Year in school	0.26 (0.27)	0.11	0.96	−0.49 (0.60)	−0.10	−0.83	0.81 (0.46)	0.19	1.75	−0.26 (0.45)	−0.06	−0.58	0.02 (0.03)	0.08	0.72
Financial aid	−0.41 (0.29)	−0.07	−1.41	−0.53 (0.66)	−0.04	−0.80	−0.34 (0.51)	−0.03	−0.67	−0.67 (0.50)	−0.06	−1.35	−0.03 (0.03)	−0.04	−0.84
Employment	−0.35 (0.31)	−0.06	−1.13	0.37 (0.69)	0.03	0.53	−1.04 (0.53)	−0.10	−1.95	−0.40 (0.52)	−0.04	−0.77	−0.04 (0.03)	−0.07	−1.25
First‐generation	−0.41 (0.34)	−0.06	−1.20	−0.35 (0.77)	−0.03	−0.46	−0.25 (0.60)	−0.02	−0.42	−0.59 (0.58)	−0.05	−1.02	−0.01 (0.04)	−0.01	−0.27
Race, Asian b	−0.27 (0.38)	−0.04	−0.72	−0.38 (0.85)	−0.02	−0.45	−0.52 (0.67)	−0.04	−0.79	−0.29 (0.65)	−0.02	−0.46	0.03 (0.04)	0.04	0.76
Race, all other categories b	0.27 (0.46)	0.03	0.58	−1.01 (1.02)	0.05	1.00	−0.55 (0.79)	−0.03	−0.69	1.30 (0.77)	0.08	1.68	0.08 (0.05)	0.08	1.63
Ethnicity c	0.71 (0.45)	0.08	1.56	2.50 (1.01)	0.13	2.47*	0.20 (0.79)	0.01	0.25	0.80 (0.77)	0.05	1.05	0.18 (0.05)	0.19	3.83***
Step 4	R 2 = 0.41, ΔF = 0.14, df = 1			R 2 = 0.38, ΔF = 1.11, df = 1			R 2 = 0.48, ΔF = 2.33, df = 1			R 2 = 0.51, ΔF = 6.99**, df = 1			R 2 = 0.43, ΔF = 3.39, df = 1
Food insecurity	0.69 (0.32)	0.12	2.16*	0.89 (0.71)	0.07	1.25	1.95 (0.56)	0.18	3.51***	1.57 (0.53)	0.15	2.93**	0.08 (0.03)	0.14	2.52*
Body mass index	−0.77 (2.03)	−0.02	−0.38	4.87 (4.63)	0.06	1.05	−5.43 (3.55)	−0.08	−1.53	−8.93 (3.38)	−0.13	−2.64**	−0.39 (0.21)	−0.10	−1.84

Note: N = 259. Adjusted R ² is reported.

^a The variable was log transformed due to kurtosis.

^b White was considered the reference group.

^c Non‐Hispanic was considered the reference group.

^d Female was considered the reference group.

^* p < 0.05.

^** p < 0.01.

^*** p < 0.001.

3.3.1. Cognitive Restraint

At Step 1, baseline food insecurity significantly predicted follow‐up cognitive restraint (β = 0.15, t = 2.28, p = 0.02, partial r = 0.15). At Step 2, with the addition of only baseline cognitive restraint, food insecurity did not significantly predict follow‐up cognitive restraint (β = 0.08, t = 1.59, p > 0.05, partial r = 0.10). However, at Step 3, food insecurity at baseline was associated with significantly greater cognitive restraint at follow‐up (β = 0.12, t = 2.14, p = 0.03, partial r = 0.14), accounting for sociodemographic variables and cognitive restraint at baseline. At Step 4, with the addition of baseline BMI, baseline food insecurity remained significantly associated with follow‐up cognitive restraint (β = 0.12, t = 2.16, p = 0.03, partial r = 0.14).

3.3.2. Binge Eating

At Step 1, baseline food insecurity predicted follow‐up binge eating (β = 0.13, t = 2.04, p = 0.04, partial r = 0.13). However, at Step 2, with the addition of baseline binge eating, baseline food insecurity did not predict follow‐up binge eating (β = 0.07, t = 1.31, p > 0.05, partial r = 0.09). At Step 3, with the addition of sociodemographic variables, baseline food insecurity again did not predict follow‐up binge eating (β = 0.08, t = 1.35, p > 0.05, partial r = 0.09). At Step 4, with the addition of baseline BMI, baseline food insecurity again did not predict follow‐up binge eating (β = 0.07, t = 1.25, p > 0.05, partial r = 0.08).

3.3.3. Dietary Restriction

At Step 1, baseline food insecurity significantly predicted follow‐up dietary restriction (β = 0.20, t = 3.22, p = 0.001, partial r = 0.20). At Step 2, with the addition of baseline dietary restriction, baseline food insecurity again significantly predicted follow‐up dietary restriction (β = 0.13, t = 2.69, p = 0.008, partial r = 0.17). At Step 3, the presence of food insecurity at baseline remained associated with significantly greater dietary restriction at follow‐up (β = 0.17, t = 3.34, p < 0.001, partial r = 0.22), adjusting for baseline sociodemographic variables and dietary restriction. At Step 4, with the addition of baseline BMI, baseline food insecurity remained associated with significantly greater dietary restriction at follow‐up (β = 0.18, t = 3.51, p < 0.001, partial r = 0.23).

3.3.4. Excessive Exercise

At Step 1, baseline food insecurity significantly predicted follow‐up excessive exercise (β = 0.15, t = 2.31, p = 0.02, partial r = 0.15). At Step 2, by adding baseline excessive exercise, baseline food insecurity significantly predicted follow‐up excessive exercise (β = 0.13, t = 2.77, p = 0.006, partial r = 0.18). At Step 3, the presence of food insecurity at baseline remained associated with significantly greater excessive exercise at follow‐up (β = 0.14, t = 2.67, p = 0.008, partial r = 0.17), adjusting for baseline sociodemographic variables and excessive exercise. At Step 4, by adding baseline BMI, baseline food insecurity again remained associated with significantly greater excessive exercise at follow‐up (β = 0.15, t = 2.93, p = 0.004, partial r = 0.19).

3.3.5. Purging

At Step 1, baseline food insecurity significantly predicted follow‐up purging (β = 0.21, t = 3.32, p = 0.001, partial r = 0.21). At Step 2, adding baseline purging, baseline food insecurity again predicted follow‐up purging (β = 0.10, t = 1.97, p = 0.05, partial r = 0.13). At Step 3, when adding sociodemographic variables, food insecurity at baseline remained associated with follow‐up purging (β = 0.13, t = 2.39, p = 0.02, partial r = 0.16). At Step 4, adding baseline BMI, baseline food insecurity was again associated with follow‐up purging (β = 0.14, t = 2.52, p = 0.01, partial r = 0.16).

4. Discussion

This study is the first to examine food insecurity as a longitudinal predictor of ED symptoms among college students. As hypothesized, food insecurity predicted increased cognitive restraint, dietary restriction, excessive exercise, and purging 3 months later. However, food insecurity did not longitudinally predict binge eating. Effect sizes for all significant relations were of medium magnitude and comparable to or larger than those reported in prior work examining associations between food insecurity and ED symptoms in college students (partial η2 = 0.02–0.04, Christensen et al. 2021; β = 0.17–27, Royer et al. 2021). Therefore, findings align with theoretical and empirical literature suggesting that food insecurity increases vulnerability to ED symptoms among college students and indicate that further study is needed to understand the risk process to inform ED prevention and intervention efforts.

Results related to food insecurity's prediction of dietary restriction align with prior cross‐sectional and longitudinal findings among general population samples (Hazzard, Loth, Fertig, et al. 2023; Middlemass et al. 2021). Importantly, we examined restrictive eating using the EPSI, which delineates cognitive and behavioral food restriction into separate subscales (cognitive restraint and dietary restriction). Other ED symptom measures use a combined scale to simultaneously measure both types of restraint (i.e., Cooper and Fairburn 1987). When combined, it is difficult to draw conclusions regarding potentially distinct relations between cognitive intentions to eat less or behavioral actions of food restriction with food insecurity. Our study indicated a longitudinal relation between food insecurity with both cognitive restraint and dietary restriction. These findings fit with the feast‐or‐famine theoretical model, in which food insecurity is posited to promote preoccupation with, and concurrent behavioral restriction of, food (Hazzard, Loth, Crosby, et al. 2023). Engaging in restrictive behaviors may help to preserve students' limited food supply, especially as meal plan or grocery funds dwindle. Furthermore, because food insecurity is associated with weight gain (Carvajal‐Aldaz et al. 2022), cognitions and behaviors consistent with food restriction may be engaged in with the goal of weight regulation. Notably, the items in the EPSI Dietary Restriction subscale (e.g., “People told me I do not eat very much,” “People encouraged me to eat more”) do not assess motivation for restricting food intake. Therefore, we cannot know if results reflect eating less due to food insecurity or due to ED pathology. Future research to determine if this subscale differentiates restriction for weight loss and restriction due to food insecurity is needed.

Prior studies have not specifically examined excessive exercise as a longitudinal outcome of food insecurity. Future research should investigate the mechanisms underlying this relation. Perhaps college students with food insecurity may feel compelled to excessively exercise to prevent weight gain from relying on lower‐cost, calorie‐dense foods (Morales and Berkowitz 2016). Given qualitative evidence that individuals experiencing food insecurity use weight control behaviors to mitigate weight gain (Taylor et al. 2020), future longitudinal studies should investigate increases in BMI as a mediator of this relation. It could also be that college students with food insecurity are less able to access healthcare services (Kushel et al. 2006), therefore limiting opportunities to receive reliable health information that would improve their eating and exercise habits. There is also literature reporting that food insecurity is associated with higher levels of stress among college students (Coffino et al. 2021), which increases the risk of excessive exercise (Yau and Potenza 2013).

Baseline food insecurity predicted purging three months later, even accounting for associations with baseline purging, sociodemographic factors, and BMI. This finding aligns with work among population‐based adults showing that food insecurity cross‐sectionally associates with self‐induced vomiting, laxative, and diuretic misuse (Becker et al. 2017; Laboe et al. 2023), and longitudinally associates with self‐induced vomiting (Hazzard, Loth, Fertig, et al. 2023). On the EPSI, purging behaviors are collapsed into one subscale, so we cannot know if a specific purging behavior (e.g., self‐induced vomiting) was more likely to occur across time as predicted by food insecurity than another (e.g., diuretic misuse). Future research might examine longitudinal associations between food insecurity and specific purging behaviors among college students to determine if associations mirror those observed in community adults (Hazzard, Loth, Fertig, et al. 2023). Given research suggesting that weight gain occurs because of food insecurity and prompts weight control behaviors, it may be that purging behaviors, like dietary restriction and excessive exercise, are used to regulate weight. As such, future studies should examine whether BMI mediates the relationship between food insecurity and purging.

We did not find a longitudinal relation between food insecurity and binge eating when baseline binge eating, sociodemographic factors, and BMI were in the model. This is surprising given prior evidence of cross‐sectional associations among college students (Christensen et al. 2021) and longitudinal associations among community adults (Hazzard, Loth, Crosby, et al. 2023; Hazzard, Loth, Fertig, et al. 2023). Potentially, the relation was not observed due to the limited longitudinal period. Specifically, student meal funds usually run on a semesterly basis and thus begin to run out or expire at the end of the semester. Because we assessed binge eating at the end of the college semester, participants may not have had enough access to food to binge eat due to potentially having even less access to food than they had at the beginning of the semester. Future research is needed using (a) longer longitudinal periods or (b) intensive longitudinal methods (e.g., daily diary) to evaluate the relation across shorter time periods.

Findings are clinically relevant for college settings (including student healthcare) and have important practical public health implications. Clinically, providers serving college students should be aware that food insecurity may lead to increases in ED symptoms. Providers treating ED symptoms should screen students for food insecurity and connect them with resources to access low‐cost food on and off campus. Increasing access to campus food pantries increased fruit and vegetable consumption among college students with food insecurity, suggesting that pantries improved both food security and diet quality (Chodur et al. 2024). Importantly, the proliferation of campus food pantries may also allow for better reach of students with EDs. It can be challenging for students to access resources, given the perceived stigma of mental health treatment; providing such resources through food pantries may facilitate the dissemination of ED information and treatment (Graham et al. 2024). Finally, students with food insecurity may benefit from additional support for accessing food outside of the university, such as the US Supplemental Nutrition Assistance Program (SNAP). Many students are unaware of the SNAP eligibility requirements and therefore do not apply (Esaryk et al. 2022). Alternatively, some students do not meet eligibility requirements due to scholarships, work‐study wages, or other financial support that exceeds SNAP eligibility requirements (Esaryk et al. 2022). Policy‐related action is needed to improve SNAP eligibility which may help to mitigate ED‐related disparities among students.

Study strengths include the longitudinal design that allowed for assessment of residual changes in ED symptoms predicted by food insecurity and the use of quantitative measures with strong psychometric properties. There are also limitations. First, our sample size was modest, largely white, and female‐identifying. Participants were psychology students who self‐selected into the study and may not be representative of the wider college population. Given the range of campus food insecurity estimates (20%–50%; Freudenberg et al. 2019) and unique school demographic breakdowns, we cannot know how results may generalize to other settings (i.e., community colleges). Effects should be replicated with larger samples with greater age, racial, ethnic, and gender diversity and from multiple universities to confirm generalizability. Second, we did not assess ED diagnoses, and the self‐report measure we used did not provide a clinical cut score. We cannot know if food insecurity predicted clinical threshold EDs. Additionally, self‐report ED measures may not correlate with symptom frequency as assessed in clinical interviews (Everett et al. 2021). Because ED diagnoses cross‐sectionally associate with food insecurity (Lydecker and Grilo 2019; Rasmusson et al. 2019), future research should use clinical interviews to assess ED diagnosis in relation to food insecurity across longitudinal periods. Third, though food insecurity can be analyzed as an ordinal variable, we analyzed it as a binary variable to maximize our power to detect effects in our modest sample. Thus, we could not examine differences in effects by food insecurity severity. Fourth, participants were assessed only twice across 3 months. We could not investigate mechanisms underlying the longitudinal relation between food insecurity and ED symptoms, and we cannot know if effects persist beyond the 3‐month period. Future studies with additional time points and longer follow‐up periods, perhaps across the college years, may investigate specific factors that explain the relation between food insecurity and ED symptoms such as weight gain, weight stigma, or shame (Davis et al. 2023, 2024; Hazzard et al. 2022). Qualitative studies may reveal additional mechanisms underlying the relation. For example, previous research suggests college students worry about finances and being able to afford food because they are prioritizing other expenses (e.g., rent; Richards et al. 2023; Stebleton et al. 2020). Additionally, we did not assess dining meal plan usage, which may mitigate food insecurity and EDs. Finally, we did not collect data on students' receipt of governmental food assistance; we cannot know if receiving assistance could have impacted results.

The present study expands knowledge of ED symptoms and food insecurity by demonstrating longitudinal associations among college students. Efforts to understand and treat EDs among college students with food insecurity, and to increase access to resources on and off campus, are needed.

Author Contributions

Jacqueline A. Kosmas: conceptualization, methodology, writing – original draft, writing – review and editing. Mariella Garza: conceptualization, writing – original draft, writing – review and editing. Meredith Kells: conceptualization, writing – review and editing. Samantha L. Hahn: conceptualization, methodology, writing – review and editing. Heather A. Davis: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, writing – original draft, writing – review and editing.

Ethics Statement

This study was reviewed and approved by the Virginia Tech Institutional Review Board (IRB #22–257).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1.

Data Availability Statement

Availability of data and material: The data that support the findings of this study are available upon reasonable request from the corresponding author.