Food insecurity may promote opportunistic eating through heightened satiety insecurity: A cross-sectional mega-analysis study

Abstract

Background:

Food insecurity promotes opportunistic eating (overeating when food is available), contributing to obesity. Opportunistic eating may persist after food security is restored; it is unclear why. Food insecurity may produce lasting worries about maintaining fullness between meals, which may motivate opportunistic eating independent of food access.

Objective:

We developed the novel Satiety Insecurity Scale (4-items), reflecting worries about maintaining fullness. We tested whether satiety insecurity contributes to food insecurity’s relationship with opportunistic eating and higher BMI, and whether relationships were moderated by sex.

Methods:

We mega-analyzed six cross-sectional studies (n=1,240 from Singapore, United Kingdom, United States) that included measures of food insecurity, satiety insecurity, and virtual portion size selection. Mega-analyses tested the aggregated indirect effects of satiety insecurity on relationships of food insecurity with virtual portion sizes and BMI. One study included an experiment where participants selected ideal virtual dinner portions while expecting to consume a large or small hypothetical lunch.

Results:

Higher food insecurity was associated with higher satiety insecurity, which was independently associated with selecting larger virtual portion sizes (Indirect effect = 6.82 kilocalories, 95%CI: 4.15, 9.50) and higher BMI (Indirect effect = 0.28, 95%CI: 0.14, 0.42). The association between food insecurity and satiety insecurity was stronger among males. However, the serial indirect effect of food insecurity on BMI through satiety insecurity and portion selection was only observed among females. Satiety insecurity was also associated with selecting larger virtual dinner portions only when expecting to consume a large virtual lunch, suggesting satiety insecurity may motivate eating when hunger is minimal.

Conclusions:

Satiety insecurity is an unexamined driver of overeating and partially explains why food insecurity motivates opportunistic eating and higher BMI (especially among females). Efforts to mitigate the lasting effects of food insecurity on obesogenic eating behaviors should both restore food access and target insecurities about maintaining satiety.

Introduction

In the USA, about 12.8% of households (44 million people) experience food insecurity (1), which is troubling given food insecurity is an important risk factor for many chronic conditions, such as overweight/obesity, diabetes, heart disease, and stroke (2–4). Food insecurity is marked by inconsistent or complete lack of access to adequate food for healthy living. Food insecurity is associated with obesity, yet this relationship is more pronounced among females compared to males (5).

Food insecurity may impact health by promoting opportunistic eating, in which individuals eat in the absence of hunger or eat excessively when food is available to compensate for prior or anticipated bouts of smaller or delayed meals. Opportunistic eating may lead to excess energy intake and weight gain. Indeed, food insecurity is associated with poorer diet quality (10, 11), greater ad-libitum energy intake (12), eating beyond satiation and in the absence of hunger (13), and binge eating (14, 15). Further, longitudinal studies suggest that food insecurity during adolescence predicts binge eating and obesity in young adulthood (16, 17). Animal models of food insecurity have demonstrated that unpredictable food availability, but not lower total energy intake per day, may lead to increased meal size and weight gain when ad-libitum access to palatable high-fat/sugar foods becomes available (18).

This tendency to opportunistically overeat persists even after food security has been restored. Independent of current food (in)security, childhood experiences of food insecurity are associated with adults’ tendencies to engage in greater daily carbohydrate intake (19) and greater energy intake from snacks following a high-energy preload (20). Further, early life socioeconomic disadvantage, which may be associated with food insecurity, continues to predict eating in the absence of hunger and higher adiposity in adulthood (21, 22). Overall, the persistence of these eating behaviors in later stages of life when food security may be restored suggests that food insecurity may program motivations towards food in a lasting way. If so, relieving food insecurity alone may be insufficient to address its detrimental effects on eating behaviors and health.

Yet, mechanisms by which food insecurity may shape or program the tendency to opportunistically overeat remain unclear. Measures of food insecurity typically assess concerns about reliable access to food itself (23). However, the persistence of opportunistic eating behaviors when food security is restored suggests that food insecurity may elicit insecurities about subjective states related to food (i.e., attaining satisfaction, delaying hunger) rather than access to food itself. This is supported by prior findings demonstrating that adult females who experienced childhood food insecurity are more likely to value sensations of fullness (19).

We propose that satiety insecurity - insecurities about one’s ability to reliably maintain sensations of fullness (and suppress hunger) - may be a mechanism by which food insecurity promotes tendencies to eat opportunistically. To our knowledge, there have not been any formal constructs proposed that capture insecurities about one’s ability to maintain fullness or control hunger onset. Thus, satiety insecurity may be an important construct for understanding how food insecurity shapes eating behaviors and as general drivers of overeating. Satiety insecurity may also provide an explanation for the Food-Insecurity Obesity Paradox, in which food insecurity is related to higher BMI, especially among women (5).

We developed and validated a measure of satiety insecurity (Satiety Insecurity Scale). We tested whether satiety insecurity contributes to the relationship of food insecurity with intentions to consume more energy from a meal (portion size selection). Traits that promote opportunistic eating behaviors, such as eating in response to food-related cues and in the absence of hunger, have been associated with selecting larger portion sizes on virtual portion selection tasks (24–26). We hypothesized a mediating role of satiety insecurity, such that greater food insecurity would be associated with higher satiety insecurity, and subsequently with selection of larger virtual portions and higher BMI (Figure 1; Hypothesis 1). We also tested an exploratory moderated mediation by sex, to examine whether this hypothesized mediation by satiety insecurity is stronger for females than males. We also hypothesized that those reporting greater satiety insecurity would be more likely to engage in opportunistic eating by selecting larger virtual dinner portions, even when expecting to consume a large virtual lunch (Hypothesis 2). This could reflect motivations to extend sensations of satiety by intending to consume a large portion at dinner even when hunger may be minimal.

Figure 1.

Mediation models depicting the indirect effect of higher satiety insecurity on the relationships between: A) higher food insecurity and greater selected mean portion sizes, and B) higher food insecurity and higher BMI (total n = 1,240). The coefficients for the paths, direct, total and indirect effects are based on aggregating individual participant data across six studies into a mega-analysis. This analysis was conducted and pooled across 25 multiply imputed datasets. Coefficients are unstandardized. Parentheses include 95% confidence intervals. BMI = Body mass index.

Methods

Studies

Secondary analyses were conducted on six independent cross-sectional studies that were conducted in the USA, UK and Singapore (Table 1). Although the measures and procedures varied for each study, each study included questions assessing insecurity about maintaining satiety and satiation, food insecurity, and a virtual portion size selection task. Analyses for Hypothesis 1 were first conducted on three studies (Studies 1–3). Hypotheses 1 and 2 and analyses were then pre-registered for the remaining three studies (Studies 4–6) along with the plan to aggregate the results of the six studies through meta-analysis. Although a meta-analysis of these six studies was originally pre-registered, we instead conducted a mega-analysis (individual participant database analysis) (27) that pooled data across the six studies to test Hypothesis 1. The indirect effect of satiety insecurity on the relationship between food insecurity and portion selection was tested on this pooled data set using mega-analysis. Participants in all studies provided informed consent prior to taking part in the study, and all studies were approved by the university’s Institutional Review Board. Present analyses were pre-registered¹ on the Open Science Framework (https://osf.io/j93g8).

Table 1.

Descriptive statistics of analytic variables and additional contextual variables for the six studies.

	Study 1	Study 2	Study 3	Study 4	Study 5	Study 6
N	283	261	312	152	160	72
Location	USA	Singapore	United Kingdom	United Kingdom	Singapore	Singapore
Recruitment Dates	Oct-Nov 2020	Sept 2020	Sept-Oct 2020	March-April 2021	August-Sept 2020	March-May 2021
Format	Online	Online	Online	Online	Lab	Lab
Portion selection task (PST)	Multiple choice	Multiple choice	Multiple choice	Slider	Slider	Incremental change
Number of foods in PST	12	10	10	10	25	13
Age	39.26 (8.27)	37.38 (9.77)	43.55 (12.72)	31.13 (9.22)	21.40 (2.25)	22.53 (2.74)
% Female	53%	44%	57%	59%	59%	75%
Race/Ethnicity frequency: n (%)	48 Asian (17.0%) 13 Black (4.6%) 5 Multiracial/Mixed (1.8%) 189 White (66.8%) 6 Hispanic/Latino (2.1%) 22 Other/Unsure (7.8%)	203 Chinese (77.8%) 20 Malay (7.7%) 16 Indian (6.1%) 2 Eurasian (0.8%) 18 Other (6.9%)	14 Asian (4.5%) 5 Black (1.6%) 8 Mixed (2.6%) 266 White (85.8%) 9 Other (2.9%)	19 Asian (12.5%) 2 Black (1.3%) 124 White (81.6%) 5 Biracial (3.3%) 2 Other (1.3%)	138 Chinese (86.3%) 3 Malay (1.9%) 13 Indian (8.1%) 1 Eurasian (0.6%) 4 Other (2.5%)	59 Chinese (1.9%) 3 Malay (4.2%) 6 Indian (8.3%) 4 Other (5.6%)
Household income	US$41892.96 (40770.53) yearly per member of household	S$26635.80 (50438.85) monthly	£36205.81 (35284.58) Yearly	£60305.21 (88060.00) yearly	S$11323.80 (29949.59) monthly	Not collected
Education Frequency: n (%)	Not collected	16 Vocational (6.1%) 17 GCE O Level (6.5%) 3 N Level (1.1%) 21 GCE A Level (8.0%) 161 Bachelor (61.7%) 26 Master(10.0%) 4 Doctoral (1.5%) 11 Other (4.2%)	78 GCSE/CSE/O-Level/National 5 (25.2%) 54 A-Level/GCE/BTEC (17.4%) 9 BTEC (2.9%) 18 Higher National Certificate (5.8%) 22 Higher National Diploma (7.1%) 86 Bachelor’s Degree (27.7%) 23 Master’s Degree (7.4%) 7 Doctorate Degree (2.3%) 5 Other (1.6%)	12 GCSE/CSE/O-Level/National 5 (7.9%) 37 A-Level/GCE/BTEC (24.3%) 5 BTEC (3.3%) 2 Higher National Certificate (1.3%) 10 Higher National Diploma (6.6%) 61 Bachelor’s Degree (40.1%) 19 Master’s Degree (12.5%) 4 Doctorate Degree (2.6%) 2 Other (1.3%)	12 Vocational (7.5%) 11 GCE O Level (6.9%) 109 GCE A Level (68.1%) 23 Bachelor (14.4%) 3 Master (1.9%) 1 Doctoral (0.6%)	Not collected
Subjective socioeconomic status	6.04 (2.03)	5.72 (1.88)	5.06 (2.01)	Growing Up: 5.45 (1.81) Currently: 5.74 (1.59)	Not collected	Not collected
Satiety insecurity	2.63 (0.87)	2.61 (0.80)	2.53 (0.94)	2.76 (0.85)	3.11 (0.69)	2.93 (0.74)
Food insecurity	2.88 (1.80)	2.12 (1.04)	1.96 (1.05)	1.37 (0.64)	1.69 (0.75)	1.56 (0.72)
Portion selection	331.47 (121.90)	272.32 (109.27)	314.69 (118.08)	358.43 (92.32)	382.46 (129.48)	454.86 (117.31)
BMI	26.95 (7.08)	22.97 (4.60)	26.03 (6.08)	24.15 (4.14)	22.00 (3.95)	22.26 (3.83)
Current hunger	45.05	53.52 (25.79)	47.71 (28.24)	48.00 (24.41)	51.25 (25.01)	25.32 (18.03)
TFEQ: Hunger	N/A	N/A	N/A	6.47 (3.62)	N/A	6.43 (3.85)
TFEQ: UE	N/A	2.21 (0.61)	2.21 (0.69)	N/A	2.39 (0.56)	N/A
TFEQ: EE	N/A	2.13 (0.76)	2.23 (0.81)	N/A	2.19 (0.82)	N/A
TFEQ: CR	2.40 (0.63)	2.37 (0.56)	2.33 (0.62)	N/A	2.36 (0.60)	2.31 (0.65)
Satiety goal: Stop hunger	59.51 (21.78)	65.98 (19.41)	62.38 (22.64)	57.99 (21.22)	Not collected	55.58 (15.18)
Satiety goal: Comfortably full	66.63 (17.40)	70.61 (16.72)	66.66 (20.62)	68.95 (14.78)	Not collected	67.37 (11.80)
Satiety goal: Completely full	74.11 (20.11)	78.62 (17.61)	80.24 (21.03)	89.15 (12.64)	Not collected	83.33 (12.49)
Satiety goal: Stop hunger or feel full?	57.36 (28.53)	60.92 (31.38)	56.00 (32.66)	Not collected	Not collected	56.48 (29.77)

Unless otherwise noted, means are presented with standard deviations in parentheses. Household income in Study 1 refers to household income per member of the household, rather than total household income. Gray cells indicate that the measure was not administered in the study.

Abbreviations for the Three Factor Eating Questionnaire (TFEQ) subscales: dispositional hunger subscale (hunger), uncontrolled eating (UE), emotional eating (EE), and cognitive restraint (CR).

Measures that were used for the current hypothesis tests and analyses across the six studies are described in the Measures section below, while descriptions of other measures in each of studies are provided in the Supplementary Materials.

Participants

Across the six studies, 1,532 participants were initially recruited. Participants were excluded from analyses for failing/not completing attention checks (Studies 2–4; n =268), and reporting outside the plausible ranges of height (1.2–2.2m), weight (30–400kg), and BMI (14–70kg/m²; n =24) (28). The final sample size across the six studies included 1,240 participants (see Table 1 for sample sizes for each study). Four studies were conducted entirely online (Studies 1–4) and two were conducted in laboratory environments (Studies 5–6). Compensation varied across studies (e.g., research credit for laboratory studies conducted at a university, monetary compensation). Characteristics of the studies and samples are summarized in Table 1.

Measures

Current Hunger.

Participants responded to four questions regarding their levels of hunger and desire to eat at the time of the survey (e.g., How hungry are you right now?; How strong is your desire to eat right now?; How much could you eat right now?). Responses were recorded on a slider scale (0 = not at all, 100 = very much). One item (How full are you right now?) was reverse scored. An average score was generated across the items (α = .76-.92), with higher scores indicating higher levels of hunger.

Food Insecurity.

An adapted version of the U.S. Department of Agriculture Food Security survey was used to assess food insecurity (29). Response options were modified to indicate how often in the past 12 months the respondents experienced the following situations (i.e., How much have you been worried that the food you bought wouldn’t last and you would not be able to get more?) on a 7-point Likert scale (i.e., 1 = almost never, 7 = almost always). Responses were averaged to generate a composite score (α = .87-.96), with higher scores indicating higher food insecurity.

Satiety and Satiation Insecurity.

Novel measures of satiety and satiation insecurity were developed. Eight items were developed that either reflected insecurities about maintaining fullness (or avoiding hunger; satiety insecurity) or motivations/behaviors to maximize fullness and satisfaction from meals (satiation insecurity). The response format for each item was a 5-point Likert scale of frequency (i.e., 1 = almost never, 5 = almost always).

Principal components² factor analysis with varimax rotation was conducted within each study on the items of the satiety and satiation insecurity items. Four items consistently loaded onto one factor representing satiety insecurity across the studies (How often do you desire more food after you finish a meal or snack? How often do you worry that typical portion sizes will not keep you feeling full? How often do you worry that you will start feeling hungry between meals? How often do you feel like the meals you eat just aren’t keeping you full long enough?). Three of these items were developed a priori to represent concerns related to satiety (fullness experienced between meals). While one item (How often do you desire more food after you finish a meal or snack?) had originally been developed to represent concerns about satiation (fullness experienced within a meal), it was averaged with the other three items to create a composite measure of satiety insecurity across the studies, given its consistent loading with the other satiety-related items (α = 0.73 to 0.88; see Supplementary Table 1 for factor loadings). However, there were less consistent factor loadings for the remaining four items representing satiation insecurity across the studies (How often do you select foods that will keep you full? How often are you satisfied after eating a meal or sack? How often do you consume a meal and feel full after? How often do you eat until you feel completely full?; Supplementary Table 1). These remaining satiation insecurity items generally exhibited low reliability across the studies (α = .34-.71) and were removed from analyses. Thus, our analyses focused on the four items pertaining to satiety insecurity (Satiety Insecurity Scale). Detailed results for the factor analysis for each study are reported in Supplementary Materials (Supplementary Table 1).

Portion Selection Task.

All studies contained variations of a virtual portion selection task (PST), where participants chose their ideal portion sizes of snack and meal foods common in their country (i.e., chicken rice, Singapore; penne pasta, UK, mac n cheese, USA). A virtual portion selection task (PST) was used as one of our primary outcome measures, since these tasks are reliable proxies of portion sizes that people select and consume of real food (30, 31). While ad-libitum portion selection and intake of real food can only be practically tested for a single meal or food type per session, the PST represents an estimation of each participant’s desired portion sizes across multiple possible meals and food types. Portions throughout the tasks ranged from a minimum of 20 kcal to a maximum of 740 or 1,000 kcal (depending on the study). The number of foods on the task ranged from 10 to 25. Response formats included using left/right arrows to decrease/increase portions depicted on the screen in 20 kcal increments (Study 6; 31, 32), or versions of the portion selection task that involved multiple choice between seven portions sizes (Studies 1–3) or choosing an ideal portion on a slider (Study 4, Study 5; 33). Selected portion sizes for each food were averaged for each participant to generate mean selected portion sizes (in kcal; α = .85-.93). A summary of the characteristics of the PSTs across the studies is displayed in Table 1.

Three Factor Eating Questionnaire-R21 (TFEQ).

The 21-question version of the TFEQ was administered in Studies 2, 3, and 5 to assess individual differences in cognitive restraint (6 items; α = .75-.81), uncontrolled eating (9 items; α = .83-.89), and emotional eating (6 items; α = .91-.93; 34). The cognitive restraint subscale was administered independently in Studies 1 and 6. Participants responded using a 4-point scale (1 = definitely false, 4 = definitely true). Studies 4 and 6 also included the dispositional hunger subscale of Stunkard and Messick’s 1985 version of the TFEQ (35). Participants responded to 15 items, 12 with true-false responses and three with 4-point rating scale responses (i.e., 1 = almost never, 4 = almost every day; α = .81, Study 4; α = .85 Study 6). Responses for each subscale were averaged to compute a composite score for each subscale.

Satiety goals.

Satiety goal items measured the amount of fullness required to attain three different states of satiety (36, 37). Participants viewed an outline of a person with a stomach that was filled to three different levels of fullness: empty, half full, and totally full, which were respectively positioned over the left end, midpoint, and right end of a 100-point visual analogue (VAS) scale. Participants selected the appropriate point on separate scales to indicate how full their stomach would have to be to stop hunger, feel comfortably full, and feel completely full. On a separate item, participants were asked to indicate whether they typically eat to stop hunger or to feel full on using a 100-point VAS anchored with “stop hunger” and “feel full” on the two ends.

Large/Small Lunch Manipulation PSTs.

In addition to a standard PST, two additional versions of the PST were administered in Study 6. The measure was identical to the standard PST, except participants were presented images of two foods together on the screen (Figure 2). The food on the left side of the screen was always a fixed portion of fried rice that represented a hypothetical lunch portion of fried rice one would consume. The food on the right side of the screen was the target food that had adjustable portion sizes (from 20 to 1000 kcal in 20 kcal increments), which represented the intended dinner portion one would consume, assuming they had consumed the accompanying portion of fried rice for lunch. In the small lunch version of this PST, the fried rice portion for lunch was relatively small (260 kcals). In the large lunch version of this PST, the fried rice portion for lunch was relatively large (500 kcals). The target foods presented on the right side of the screen were identical between the small and large lunch PSTs. The three PSTs (standard, large lunch, small lunch) were completed by each participant in a randomized order. Within each of the three PSTs, portion sizes selected for target foods were averaged to generate a composite portion size selected for each PST (responses in small lunch condition, α = .90; responses in large lunch condition, α = .91). While most people would compensate for a small portion size for lunch by selecting larger desired portion sizes for dinner, those who are worried about maintaining satiety and motivated to extend sensations of satiety may be more likely to select larger desired portion sizes for dinner, despite expecting to consume a large portion for lunch.

Figure 2.

Sample stimuli from the large lunch (A) and small lunch (B) Portion Selection Tasks administered in Study 6. For each trial, the food on the left side of the screen was always a large (500 kcal) or small (260 kcal) portion of fried rice, which the participant assumed they would be consuming for lunch. The food on the right side of the screen varied within each trial, and participants could adjust the portion size (from 20 to 1000 kcal) to the desired amount.

Demographics.

Participants were asked to provide the following demographic information: age, height (dropdown menu ranging from 4 to 7 feet with equivalent options in meters and centimeters), weight (dropdown menu ranging from 84 to 336 pounds with equivalent options in kilograms and stones), household income (free response; household income per member used in Study 1; not assessed in Study 6), sex (male, female, other, prefer not to say), race/ethnicity, and educational attainment. Response options for race/ethnicity and educational attainment varied by country, representing racial/ethnic groups and structure of education typical of the country (see Table 1). Studies from different countries included inconsistent response options for race/ethnicity (e.g., race/ethnicity options from Singapore did not match the USA). Thus, we created a binary variable representing whether the respondent selected their respective country’s racial/ethnic majority group (i.e., ‘White’ for USA and UK, ‘Chinese’ for Singapore) for use in the mega-analysis.

Analyses

To conduct mega-analyses, the data from the six studies were pooled into a common dataset. To establish construct validity of the satiety insecurity measure, Pearson correlation coefficients were computed between satiety insecurity and other measures related to food- and eating-related dispositions, such as subscales of the TFEQ and implicit satiety goals (Table 2). Correlations were examined on complete cases (not multiply imputed data), since some measures were not administered across all six studies. All analyses were conducted in SPSS version 29 (IBM, 2023).

Table 2.

Aggregated (through mega-analysis) Pearson correlation coefficients for the relationships between satiety insecurity and other relevant study variables.

Measure	Mega-analysis correlation (r) with satiety insecurity	# Effect sizes	Sample Size
Food insecurity	0.38 (0.33, 0.43)	6	1186
Portion selection task	0.27 (0.22, 0.32)	6	1194
BMI	0.05 (−0.01, 0.11)	6	1173
Current hunger	0.25 (0.20, 0.30)	6	1207
TFEQ: Dispositional hunger	0.67 (0.55, 0.73)	2	169
TFEQ: Uncontrolled eating	0.67 (0.62, 0.70)	3	726
TFEQ: Emotional eating	0.44 (0.38, 0.50)	3	727
TFEQ: Cognitive restraint	0.07 (0.02, 0.13)	5	1071
Satiety goal: Stop feeling hunger	0.23 (0.17, 0.29)	5	1042
Satiety goal: Feel comfortably full	0.22 (0.16, 0.28)	5	1041
Satiety goal: Feel completely full	0.15 (0.09, 0.21)	5	1041
Satiety goal: Do you eat to stop hunger or feel full?	0.23 (0.16, 0.29)	4	920
Age	−0.30 (−0.35, −0.25)	6	1197
Household income	−0.05 (−0.11, 0.01)	5	1115
Subjective socioeconomic status	0.01 (−0.05, 0.07)	5	969

95% confidence intervals are presented within parentheses. The number of effect sizes refers to the number of correlation coefficients aggregated (number of studies). BMI: Body mass index, TFEQ: Three Factor Eating Questionnaire.

Given missingness on key variables and covariates to test our hypotheses (food insecurity: 2.1%, satiety insecurity: 2.5%, portion selection task: 1.1%, BMI: 3.1%, Sex: 1.6% missing or preferred not to say), multiple imputations were used to impute missing data (m = 25 imputed datasets) for analyses of hypothesis tests. We tested Hypothesis 1 using mediation to test the indirect effect of satiety insecurity on the relationship between food insecurity and selected portion sizes (Figure 1A). The analysis was run using the Process macro for SPSS (Model #4) (38) with 95% confidence intervals estimated through 5,000 bootstrapped samples. Coefficients for the mega-analysis were unstandardized (b) and pooled together across each of the imputed datasets using Rubin’s rules. Covariates included the study that the variables came from, composite rating of current hunger, age, race/ethnicity majority group status, and BMI (when portion selection was the outcome)³. We used the same mediation analysis and covariates (without BMI) to test the indirect effect of satiety insecurity on the relationship between food insecurity and BMI in the mega-analysis (Figure 1B). For sensitivity analysis, we conducted an alternate version of the mediation analyses testing reverse causation with satiety insecurity as the predictor and food insecurity as the mediator (Supplementary Materials).

Given that the relationship between food insecurity and BMI is more pronounced for women than men (5), we conducted exploratory moderated mediation analysis with sex as the moderator. The same mediation models for Hypothesis 1 were tested with portion selection and BMI as outcomes, while testing whether each of the paths in the model, along with total and indirect effects, differed between men and women using the Process macro for SPSS (Model #59) (38). Evidence of moderation of individual paths was assessed via interaction term coefficients and 95% confidence intervals while evidence of moderation of the entire indirect effect was determined using the index of moderated mediation (presence of a significant difference in the indirect effect between males and females).

Relationships between satiety insecurity and virtual dinner portion sizes selected following a hypothetical small or large lunch in Study 6 were analyzed using multiple linear regression on the multiply imputed datasets (m = 25) generated from Study 6’s data. In two separate models, portion sizes selected for the small lunch and large lunch were regressed on satiety insecurity and covariates (food insecurity, current hunger, age, sex, ethnicity, and BMI). Effect size sensitivity analysis estimated a medium effect size (f² = 0.11) as the minimum detectable population effect size based on power of 0.80, alpha of 0.05, and Study 6’s sample size.

Results

Correlations

Overall, satiety insecurity was positively correlated with other eating behavior phenotypes that would be expected from an individual who experiences insecurities about hunger or maintaining adequate fullness (Table 2). Aggregated coefficients (using mega-analysis) revealed that higher satiety insecurity was positively associated with dimensions of the Three Factor Eating Questionnaire (TFEQ), such as dispositional hunger (persistent readiness to eat; measured in 2 studies), uncontrolled eating (responsiveness to cues of food and eating; measured in 3 studies), emotional eating (eating in response to emotions; measured in 3 studies), but minimally correlated with cognitive restraint (exerting mental effort to limit food intake; measured in 5 studies). Higher satiety insecurity was also associated with greater self-reported fullness required to stop the sensation of hunger (measured in 5 studies), to feel comfortably full (measured in 5 studies), to feel completely full (measured in 5 studies), and the tendency to eat to feel full over stopping hunger (measured in 4 studies). Satiety insecurity was also negatively correlated with age (measured in 6 studies). Overall, those with higher satiety insecurity report eating behaviors and experiences that may drive opportunistic eating, such as a persistent readiness to eat, responsiveness to cues of food and eating, and valuing sensations of fullness.

Hypothesis 1: Portion Selection

The mega-analysis for Hypothesis 1 (6 studies) revealed that greater experience of food insecurity was associated with higher satiety insecurity (b = 0.28, 95%CI: 0.24, 0.31), such that a one unit increase in food insecurity was associated with a 0.283 unit increase in satiety insecurity. Higher satiety insecurity was associated with larger mean portion sizes selected (b = 24.42, 95%CI: 15.58, 33.35), such that a one unit increase in satiety insecurity was associated with a 24.42 kcal increase in mean portion size. We observed an indirect effect of satiety insecurity on the relationship between food insecurity and selected portion sizes (b = 6.82, 95%CI: 4.15, 9.50), such that a one unit increase in food insecurity was associated with approximately 6.82 kcal increase in portion size through satiety insecurity (Figure 1A).

Exploratory analysis of the moderation of the paths of this mediation model by sex showed that only the relationship between food insecurity and satiety insecurity (predictor to mediator) was moderated by sex (interaction term b = −0.09, 95%CI: −0.15, −0.03), such that this relationship was stronger for males (b = 0.33, 95%CI: 0.29, 0.38) than females (b = 0.24, 95%CI: 0.20, 0.29). There was no moderation of the indirect effect of satiety insecurity by sex (index of moderated mediation = −0.64, 95%CI: −6.33, 5.06), showing that the indirect effect of satiety insecurity on the relationship between food insecurity and portion size was observed for both males (b = 7.64, 95%CI: 2.60, 12.67) and females (b = 7.00, 95% CI: 4.13, 9.87).

Hypothesis 1: BMI

A similar mediation model using mega-analysis with BMI as an outcome showed that greater food insecurity was associated with higher satiety insecurity (b = 0.28, 95%CI: 0.25, 0.32), and that higher satiety insecurity was associated with higher BMI (b = 0.98, 95%CI: 0.54, 1.42). There was a significant indirect effect of satiety insecurity on the relationship between food insecurity and BMI (b = 0.28, 95%CI: 0.14, 0.41) such that a one unit increase in food insecurity was associated with approximately 0.28-unit increase in BMI through satiety insecurity (Figure 1B).

Exploratory analysis of the moderation of the paths of this mediation model by sex showed that the relationship between food insecurity and BMI (total effect) was moderated by sex (interaction term b = 0.73, 95%CI: 0.20, 1.27), such that this relationship was only observed for females (b = 0.59, 95%CI: 0.24, 0.95), but not for males (b = −0.02, 95% CI: −0.40, 0.35). There was no moderation of the indirect effect of satiety insecurity by sex (index of moderated mediation = −0.12, 95%CI: −0.38, 0.14), showing that the indirect effect of satiety insecurity on the relationship between food insecurity and BMI was observed for both males (b = 0.36, 95%CI: 0.13, 0.60) and females (b = 0.24, 95%CI: 0.10, 0.39.

Serial mediation

Given that we observed an indirect effect of satiety insecurity on the relationship of food insecurity with both portion selection and BMI using mega-analysis, we conducted an exploratory moderated serial mediation analysis using mega-analysis and the Process SPSS macro (model #92) (Figure 3). The serial mediation model tested whether higher food insecurity is associated with higher satiety insecurity (mediator 1), which is then associated with larger portion sizes (mediator 2), which is ultimately associated with higher BMI. Males exhibited a stronger relationship between food insecurity and satiety insecurity than females as reported in the results for portion selection as an outcome above. However, females exhibited slightly stronger relationships between satiety insecurity and portion selection (b = 28.85, 95%CI: 17.61, 40.09), and between portion selection and BMI (b = 0.007, 95%CI: 0.003, 0.01) than males (satiety insecurity to portion selection: b = 22.91, 95%CI: 10.02, 35.80; portion selection to BMI: b = 0.003, 95%CI: −0.001, 0.007). Although these pathways were more pronounced for females, sex did not significantly moderate these relationships (satiety insecurity x sex interaction term b = 5.94, 95%CI: −10.24, 22.12; portion selection x sex interaction term b = 0.004, −0.001, 0.10). There was no moderation of the indirect effect of satiety insecurity by sex (index of moderated mediation = 0.03, 95%CI: −0.02, 0.07). Yet, we observed a serial indirect effect through both mediators only among females (b = 0.05, 95%CI: 0.01, 0.09), but not males (b = 0.02, 95%CI: −0.01, 0.05). This model suggests that for females, food insecurity promotes increased satiety insecurity, which in turn motivates behaviors that increase energy intake (selecting larger portions), which then contributes to higher BMI. Results for the serial mediation model that does not include sex as a moderator are reported in Supplemental Materials (Supplementary Figure 1).

Figure 3.

Moderated serial mediation models depicting the serial indirect effects of satiety insecurity and portion selection on the relationship between higher food insecurity and greater selected mean portion sizes. Separate models are presented for males (A) and females (B). A significant serial indirect effect was only observed for females, not males. Coefficients are unstandardized and parentheses include 95% confidence intervals. BMI = Body mass index

Hypothesis 2: Portion selection after large or small lunch

When testing Hypothesis 2 with data from Study 6, higher satiety insecurity was also associated with selection of larger mean virtual dinner portion sizes when assuming consumption of a large virtual lunch (Study 6; b = 62.25, 95% CI: 13.91, 110.58; Table 3), such that a one unit increase in satiety insecurity was associated with approximately 62 added kcals to one’s selected dinner portion size, independent of food insecurity. There was no relationship observed between satiety insecurity and dinner portion sizes selected when assuming consumption of a smaller lunch (b = 27.17, 95% CI: −23.68, 78.02; unadjusted b = 35.99, 95% CI: −10.59, 85.57).

Table 3.

Results of linear regression models on the relationships between satiety insecurity and mean portion sizes selected on the Large and Small Lunch Portion Selection Tasks in Study 6.

Large Lunch PST (n = 72)
Overall Model	F(9, 51 = 2.44, R2 = 0.30, p = .02
Satiety Insecurity	62.25 (13.91, 110.58)
Food Insecurity	−11.53 (−59.80, 36.73)
Current Hunger	2.18 (0.38, 3.98)
Age	4,58 (−7.97, 17.13)
Sex	53.09 (−47.80, 153.98)
Malay Ethnicity	−6.09 (−128.65, 116.46)
Indian Ethnicity	−122.21 (−300.24, 55.81)
Other Ethnicity	−93.49 (−243.39, 56.394)
BMI	−1.91 (−12.36, 8.54)
Small Lunch PST (n = 72)
Overall Model	F(9, 50) = 1.62, R2 = 0.23, p = .14
Satiety Insecurity	27.17 (−23.68, 78.02)
Food Insecurity	−38.29 (−90.41, 13.83)
Current Hunger	0.97 (−1.00, 2.93)
Age	−0.47 (−14.07, 13.14)
Sex	135.48 (30.60, 240.36)
Malay Ethnicity	2.86 (−125.61, 131.33)
Indian Ethnicity	−80.46 (−268.81, 107.89)
Other Ethnicity	−80.62 (−236.19, 74.96)
BMI	−9.46 (−21.15, 2.23)

Desired virtual portion sizes (in kcal) were selected across multiple foods that would be hypothetically consumed for dinner following an experimental manipulation of large (Large Lunch PST) or small (Small Lunch PST) hypothetical lunch portion size. Coefficients are pooled from 25 multiply imputed datasets, but overall model statistics are based on original non-imputed data. Other predictors in the model represent covariates. PST: Portion Selection Task. BMI: Body Mass Index. Chinese was the reference group for other ethnic groups.

Discussion

Here, we demonstrate that satiety insecurity, or worries about one’s ability to maintain adequate fullness, may be a mechanism by which experiences of food insecurity may promote opportunistic eating behaviors. Using data aggregated from six studies, we present a novel measure of satiety insecurity and demonstrate its contribution to the relationship between experiences of food insecurity and selection of larger portion sizes among males and females. In addition to predicting general intentions to consume more energy, satiety insecurity predicted intentions for greater energy intake in an experimental manipulation of opportunistic eating. Satiety insecurity was uniquely associated with selecting larger portion sizes for a hypothetical dinner when expecting to consume a large (but not a small) portion size for lunch. While most people may expect to be hungry and compensate with a larger dinner after consuming a small lunch, the large lunch condition presents a situation that may be sensitive to motivations to consume large dinner portions even when hunger may be minimal. Even after consuming a large lunch, those experiencing higher satiety insecurity may still capitalize on the opportunity to continue to consume larger portions again at dinner to extend sensations of satiety.

We observed similar indirect effects of satiety insecurity on the relationship between food insecurity and BMI using mega-analysis for both males and females. Satiety insecurity may be a motivational mechanism for how food insecurity increases opportunistic eating and the risk of obesity. This is supported by our exploratory serial mediation analyses showing that food insecurity’s relationship with BMI may be mediated by higher satiety insecurity’s association with the tendency to select larger food portion sizes (Figure 3). Yet, this serial mediation pathway was only supported among females. Males showed no relationship between portion selection and BMI in the serial mediation model when adjusting for satiety insecurity, while females did. One interpretation is that increased adiposity is more likely to be a downstream consequence of satiety insecurity’s influence on eating behaviors of women than men. This finding is consistent with the sex-specific nature of the food insecurity-obesity paradox (5). Our serial mediation results suggest that satiety insecurity may be a partial explanation for why obesity is more likely to emerge from food insecurity experienced by women. Future studies should consider the role of satiety insecurity when examining adiposity and cardiometabolic disorders of previously food insecure females.

There were some limitations of this study. First, virtual portion selection was the only type of eating behavior outcome examined, and the design of the portion selection task differed between some of the studies. Future studies could examine other potential outcomes of satiety insecurity, such as binge eating or eating in the absence of hunger when real food is available. Second, these cross-sectional studies were correlational in nature and cannot confirm causality. Although food insecurity was measured within the past 12 months, it is unclear how long or repeatedly periods of food insecurity must be experienced to influence satiety insecurity. Serial mediation was based on cross-sectional data, so these findings should be interpreted with caution and require further validation. Prospective longitudinal studies, especially those that span developmental periods (e.g., transition from adolescence into adulthood) will be necessary in future research on satiety insecurity. However, our mediation analyses did not find support for reverse-causality, where satiety insecurity predicts variations in food insecurity (Supplemental Materials).

A major strength of the study was internal mega-analysis of six separate studies across three different continents (North America, Europe, Asia). This suggests the potential generalizability of our findings across high-income countries and our novel measure of satiety insecurity. However, samples from other regions should be considered in future studies. Another strength is the inclusion of an experimental manipulation of a hypothetical situation that simulates opportunistic eating (selection of larger dinner portion sizes despite a large lunch). Although portion sizes were measured virtually, there are some strengths afforded by the virtual nature of the portion selection task. They allow measurement of portion sizes across many virtual foods with varying characteristics (up to 25 foods in one of our studies), rather than a single food or meal that would typically be used if real food was presented. Since selected portion sizes on such tasks are associated with ad-libitum consumption of real food (30, 31), virtual portion selection tasks provide estimates of how much energy one intends to consume in general for a meal.

In addition to identifying a psychological mechanism that contributes to the relationship between food insecurity and eating behaviors, our study also introduces a brief and novel measure of satiety insecurity that fills an important gap in the literature of eating behavior phenotypes. Conventional measures of food insecurity are limited to inquiring about uncertain access to food itself, rather than subjective experiences related to food insecurity. Unpredictable or insufficient access to food likely generates increased threat of helplessness to aversive sensations like hunger. As such, insecurities about reliably maintaining satiety and adequate control over hunger may be proximal mechanisms for how food insecurity motivates opportunistic eating. Importantly, no eating behavior constructs to our knowledge have focused on concerns or worries related to inadequate satiety or the anticipated threat of hunger – a parsimonious yet overlooked explanation for overeating. We observed that satiety insecurity is also reliably associated with other phenotypes and behaviors that would be expected from someone with persistent insecurity about maintaining satiety (i.e., valuing sensations of fullness, reactivity to cues of food and eating).

Satiety insecurity may be a useful phenotype for ingestive behavior beyond the domain of food insecurity. Eating in the absence of hunger, poor compensation for large meals, excess energy intake, or choosing energy dense foods perceived as more ‘filling’ over fruits and vegetables could partly be explained by worries related to the onset of hunger and motivation to proactively extend sensations of satiety. However, this explanation is usually not considered as a driver of these eating behaviors. For instance, measures involving reasons for eating in the absence of hunger typically focus on the role of emotions unrelated to hunger/fullness, boredom or external cues to eat (41). Besides food insecurity, other experiences associated with a lack of control over managing satiety and hunger may contribute to satiety insecurity, such as having unpredictable daily schedules or undergoing a restriction-based diet.

In conclusion, our findings suggest that satiety insecurity, or worries about one’s ability to maintain adequate fullness (or to avoid hunger), may be a factor that contributes to the relationship between food insecurity and intentions to consume more energy. Satiety insecurity may also contribute to the sex-specific nature of the food insecurity-obesity paradox, where women who experienced food insecurity are at greater risk of obesity than men. Future research should prospectively test the role of satiety insecurity in the programming of opportunistic eating behaviors following bouts of food insecurity. Early intervention to alleviate food insecurity is essential. But unfortunately, not everyone who experiences food insecurity will experience relief. Satiety insecurity may be a promising target for future interventions that seek to reduce opportunistic eating behaviors and their effects on health, especially among those with a history of food insecurity. Future research could also explore properties of food assistance programs that may reduce satiety insecurity, such as patterns of assistance distribution that produce the most predictability in food access or perceived control over sensations of hunger and fullness.

Abbreviations

BMI

Body Mass Index

BTEC

Business and Technology Education Council

CR

Cognitive Restraint

CSE

Certificate of Secondary Education

GCE

General Certificate of Education

EE

Emotional Eating

Kcal

Kilocalories

PST

Portion Selection Task

TFEQ

Three Factor Eating Questionnaire

UE

Uncontrolled Eating

UK

United Kingdom

USA

United States of America

VAS

Visual Analogue Scale

Data Sharing Statement

Data described in this manuscript, code book, and analytic code will be made available upon request to the corresponding author.