Examining Impairment and Distress from Food Addiction across Demographic and Weight groups

Abstract

Rates of food addiction (FA) vary across weight and demographic groups. Factors influencing discrepant prevalence rates are largely unknown. Rates of clinically significant distress or impairment also vary across demographic groups, yet prior studies have overlooked the diagnostic significance of distress/impairment in heterogenous groups. We tested if weight and demographic groups differed in their likelihood of endorsing distress/impairment from FA. Participants (N=1,832) recruited from Amazon Mechanical Turk completed the modified Yale Food Addiction Scale 2.0 (mYFAS). The mYFAS includes 11 dichotomous symptom indicators and one dichotomous distress/impairment indicator. Differences in distress/impairment were tested across weight, sex, race/ethnic, and educational groups using logistic regression. FA severity was controlled for using FA symptom count. There were no differences among racial/ethnic and education groups (p>.05). Compared to men, women were more likely to report distress/impairment (aOR=1.96, 95% CI=1.28-3.03). People with obesity were more likely to report distress/impairment compared to people with overweight (aOR = 2.20, 95% CI = 1.39-3.49) or normal weight (aOR=1.99, 95% CI = 1.26-3.13). Individual characteristics (i.e., sex, weight) may influence reporting of distress/impairment from FA. Further inquiry may be appropriate for men and people with normal weight or overweight presenting with FA symptoms who otherwise deny distress/impairment.

1. Introduction

Food addiction is a construct characterized by the compulsive consumption of highly palatable foods (e.g. sugary, refined foods), which in some people can markedly innervate the neurobiological reward system and result in addiction symptomatology. The definition, measurement, and clinical relevance of the food addiction construct are the subject of scientific debate, sparking controversy regarding its validity¹. However, a systematic review synthesized human and animal studies and concluded there is significantly greater evidence supporting rather than disputing the validity of food addiction².

Food addiction is not formally recognized in the major diagnostic classification systems, but efforts have been made to standardize food addiction assessment. The Yale Food Addiction Scale (YFAS)³ was modeled on the seven substance dependence symptoms published in the Diagnostic and Statistical Manual of Mental Disorder 4^th Edition (DSM-IV-TR)³. The YFAS 2.0 was updated in 2016 to reflect DSM-5 diagnostic changes within the substance-related and other addictive disorders category^4,5. Like other substance use disorders, a food addiction “diagnosis” requires endorsement of at least two symptoms as well as clinically significant distress or impairment.

The clinical significance criterion was originally added to approximately half of the diagnoses in the DSM-IV in an effort to reduce false positives and better discriminate normal and sub-threshold symptom variations⁶. Rates of food addiction differ dramatically when including the clinically significant distress or impairment criterion, with rates of food addiction increasing from 19% to 26% when distress or impairment is not required^7,8. These discrepancies highlight that a sizeable subset of people endorse food addiction symptoms without reporting clinically significant distress or impairment, leading to questions about group differences and clinical correlates.

Meta-analytic research indicate differences in food addiction severity and prevalence based on demographic features (e.g., women and people with higher BMI have greater severity/prevalence), but the role of distress/impairment has been rarely investigated. A single, recent empirical study found that among people seeking weight loss treatment, weight-related features (e.g., BMI, weight related impairment) were not different between those who did and did not endorse FA distress/impairment⁹. Some evidence suggests that women may experience greater distress/impairment from food addiction,^10,11 We are unaware of any food addiction studies that specifically examined distress/impairment differences across racial groups and education groups.

While the distress/impairment and food addiction literature is sparse, evidence from related conditions (e.g., binge-eating disorder and substance use disorders) aids in hypothesizing potential associations. Binge-eating disorder (BED) has phenotypic similarities to and is comorbid with food addiction, and requires endorsement of clinically significant distress/impairment¹². Among BED patients, limited evidence suggests lower levels of distress (measured by self-reported depression) among Black as compared non-Hispanic White or Hispanic people¹³. In the substance use disorder literature, evidence suggests greater functional impairment for People of Color^14-16. The association between education and distress/impairment from psychiatric disorders is also largely unexplored.

Overall, converging evidence from the food addiction, BED, and substance use disorder literature suggests that clinically significant distress/impairment may influence rates of food addiction and there may be differences observed among certain sociodemographic groups, though some comparisons remain untested. Moreover, differences in the ways in which distress/impairment are assessed influence potential associations. Distress/impairment has been assessed using continuous self-report scales^14,16,17; qualitative themes^11,18; self-report functional health measures^15,19; and proximal constructs for distress (e.g., depression or anxiety)^13,20. Given growing efforts to understand the prevalence and treatment needs of people with food addiction²¹, it is critical to know whether characteristics other than the severity of food addiction symptoms influence endorsement of distress/ impairment. This study aimed to assess whether endorsing food addiction symptoms as distressing/impairing differed across weight, sex, racial/ethnic, and educational groups.

2. Methods

2.1. Participants

Participants (N=1,832) were recruited from Amazon’s Mechanical Turk (MTurk) online platform to complete an online survey on eating and healthcare attitudes. Participants were eligible if they were 18 years or older and spoke English. Responses to all questions were required, though participants could choose to discontinue the survey at any time. Robust studies have indicated that the platform generally produces reliable, high-quality data^22,23. However, some data also suggest the potential for “bots” or other threats to data quality²⁴. Among those who completed at least one item beyond consenting (n = 56 did not provide data beyond consent), the following steps were taken to ensure data quality, including 1) removing incorrect responses to quality control items, n = 8; and 2) removing duplicate IP addresses, n = 28. Participants were paid 0.50 cents, consistent with prior investigations using MTurk²⁵. This study received approval from the Yale Human Investigation Committee, University Institutional Review Board, and all participants provided electronic informed consent.

2.2. Assessments

2.2.1. Dependent Variable.

Clinically significant distress or impairment related to food addiction symptoms was the primary dependent variable, and it was assessed using the Modified Yale Food Addiction Scale 2.0 (mYFAS 2.0)²⁶. The mYFAS 2.0 is a 13-item abbreviated version of the YFAS 2.0 ²⁷. The mYFAS 2.0 demonstrates similar psychometric properties to the original YFAS 2.0, including strong internal consistency reliability, convergent validity, and incremental validity. Clinically significant distress or impairment is scored dichotomously based on two items: “My eating causes me a lot of distress” and “I had significant problems in my life because of food and eating. These may have been problems with my daily routine, work, school, friends, family, or health.” Participants reporting either item occurring at least two to three times per week were coded as positive for clinically significant distress or impairment²⁶. Analyses included all responses to the clinically significant distress or impairment criterion, irrespective of food addiction diagnosis. In the present study the Cronbach’s Alpha for the mYFAS 2.0 was .94.

2.2.2. Independent Variables.

BMI was used as an independent variable and calculated using self-reported height and weight (kg/m²). BMI class was based on CDC guidelines²⁸. Sociodemographic variables also served as independent variables, including sex (women, men), race/ethnicity (White Non-Hispanic, Black Non-Hispanic, Asian Non-Hispanic, White-Hispanic, Other-Hispanic, and Other Non-Hispanic), and education (high school diploma or less, some college, or college degree or higher).

2.2.3. Covariates.

The number of food addiction symptoms (range = 0 to 11) served as a covariate in the logistic regression models.

2.3. Statistical Analyses

Data analyses were completed using IBM SPSS Version 26. Prior to analysis, multiple imputation was used to address missing data. However, a number of participants did not provide BMI data (n = 122) and were excluded from analysis. BMI was unable to be estimated in an unbiased manner. Specifically, survey design required responses, and as such participants missing data on BMI were missing on average 90% of their data. Following imputation, data analyses proceeded in four steps and pooled estimates across 5 imputed data sets are reported. First, using chi-square goodness-of-fit or independent t-tests, bivariate differences in the proportion of people reporting distress/impairment across weight and sociodemographic groups were compared, including BMI class, sex, race/ethnicity, and education. Second, collinearity diagnostics were investigated using the steps outlined by Midi, Sarkar, and Rana (2013). Tolerance and Variance inflation factor (VIF) were examined. Tolerance values less than .10 and VIF values greater than 10 were considered indicative of problematic collinearity²⁹. Third, a logistic regression analysis was completed using non-Hispanic White, women, and high school diploma or less as the reference groups; number of food addiction symptoms was entered as a covariate; and endorsing distress/impairment was the dependent variable. Finally, for independent variables showing an omnibus effect based on the Wald’s statistic, the reference group was changed in subsequent models so that all possible odds ratios were calculated.

3. Results

A total of 11.5% (n = 211) met food addiction criteria. Table 1 presents demographic characteristics for the total sample and the bivariate differences between endorsing clinically significant distress/impairment and weight/sociodemographic factors. The proportion of people reporting clinically significant distress/impairment differed across BMI, sex, and education groups.

Table 1.

Bivariate Associations between Food Addiction Distress, Impairment, Weight, and Sociodemographic Factors

	Total Sample	No Distress or Impairment	Distressor Impairment	Test statistic†	Effect size‡	Post-Hoca
BMI †				χ2 (3, 1813) = 63.47 p <.001	Cramer’s V = 0.19 (medium effect)	UW > NW OB > NW, OW
Underweight n (%)	52 (2.8)	42 (80.8)	10 (19.2)
Normal Weight n (%)	716 (39.1)	660 (92.2)	56 (7.8)
Overweight n (%)	544 (29.7)	493 (90.6)	51 (9.4)
Obese n (%)	520 (28.4)	405 (77.9)	115 (22.1)
Sex				χ2(1, 1813) = 4.04 p = 0.044	Cramer’s V = 0.05 (< small effect)	Women > Men
Female n (%)	1241 (67.7)	1070 (86.2)	171 (13.8)
Male n (%)	591 (32.3)	530 (89.7)	61 (10.3)
Race/Ethnicity				χ2 (5, 1813) = 3.89 p = 0.565	Cramer’s V = 0.05 (small effect)	–
White Non-Hispanic n (%)	1363 (74.4)	1199 (88.0)	164 (12.0)
Black Non-Hispanic n (%)	120 (6.6)	106 (88.3)	14 (11.7)
Asian Non-Hispanic n (%)	138 (7.5)	119 (86.2)	19 (13.8)
White-Hispanic n (%)	104 (5.7)	86 (82.7)	18 (17.3)
Other-Hispanic n (%)	51 (2.8)	43 (84.3)	8 (15.7)
Other Non-Hispanic n (%)	56 (3.1)	47 (83.9)	9 (16.1)
Education				χ2 (2, 1813) = 9.31 p = 0.010	Cramer’s V = 0.07 (small effect)	College < Some College
High School or Less n (%)	166 (9.1)	144 (86.7)	22 (13.3)
Some College n (%)	645 (35.2)	544 (84.3)	101 (15.7)
College or Greater n (%)	1021 (55.7)	912 (89.3)	109 (10.7)

Notes. NW = normal weight; OB = obesity; OW = overweight; UW = underweight

^†Chi-square statistics are based on original data and n and % is based on imputed data

^‡Effect size interpretation is based on degrees of freedom and guidelines published in Statistical power analysis for the behavioral sciences ³⁰

^a.Post-hoc comparisons are calculated using z-score for column proportions

Collinearity diagnostics did not reveal a problem and all terms were entered into the model (see Table 2). For each additional symptom endorsed, a person was 1.76 times more likely to report distress/impairment. After controlling for the number of food addiction symptoms, people with differing BMI categories and men and women were differentially likely to report distress/impairment. Compared to people with normal weight, people with obesity were significantly more likely to report distress/impairment. When people with overweight were used as the reference category (not included in table 2), people with obesity were significantly more likely to report distress/impairment (OR = 2.22, 95% CI = 1.39-3.54). Compared to women, men were significantly less likely to report clinically significant distress/impairment.

Table 2.

Results of Logistic Regression Model Assessing Impairment and Distress Criterion

	Adj. OR	95% CI
Symptom Count	1.87	1.74-2.00
Sex
Females	–	–
Males	0.49	0.32-0.76
Race/Ethnicity
White Non-Hispanic	–	–
Black Non-Hispanic	0.77	0.36-1.63
Asian Non-Hispanic	1.66	0.85-3.26
White-Hispanic	0.94	0.44-1.98
Other-Hispanic	0.71	0.26-1.93
Other Non-Hispanic	1.44	0.57-3.62
Education
High School or Less	–	–
Some College	0.90	0.46-1.76
College or Greater	0.63	0.32-1.23
BMI
Normal Weight	–	–
Underweight	1.71	0.61-4.81
Overweight	0.90	0.54-1.49
Obese	1.99	1.26-3.13

Notes. Bold values indicate significance at the p ≤ .05 level. BMI = Body mass index

4. Discussion

The current study examined whether the experience of food addiction symptoms as distressing/impairing differed across BMI, sex, racial/ethnic, and educational groups. After controlling for the number of food addiction symptoms, people were more likely to identify the symptoms as distressing/impairing if they had obesity or were women. The current findings raise important questions for assessing and treating food addiction.

Unlike the recent study by Koehler and colleagues⁹, BMI was related to distress/impairment. However, the previous study utilized a sample seeking treatment for weight loss, who had an average BMI in the obese range. Based on the combined findings, it could be that obesity specifically has an effect on FA distress/impairment. The specificity of this effect could be explained by weight bias and weight bias internalization. People with obesity experience higher levels of both weight bias ^31,32 and weight bias internalization ³³ as compared to people with either normal weight or overweight. Future research should investigate the role of weight bias and other potential explanatory constructs for the observed differential endorsement of distress/impairment.

Men and women also differed in their likelihood of identifying their food addiction symptoms as distressing/impairing, with women being more likely. This fits with longstanding sociological theories about emotional expressiveness^34,35. Empirical evidence of the distress/impairment caused by eating pathology between the sexes, however, is sparse. A qualitative study investigating men’s experience of BED found that men were more likely to identify binge-eating as consistent with masculinity and therefore not necessarily distressing³⁶. Overall, two competing explanations may contribute to the observed results for both weight and sex. First, men and people with normal weight or overweight may in fact be less distressed/impaired by food addiction symptoms. Assuming this, the differential prevalence rates observed for women and people with obesity correctly reflect this reality. Alternatively, it may be that nonspecific factors may lead men or people with normal weight or overweight to be less likely to report distress/impairment from food addiction. Nonspecific factors could include measurement bias or broader sociocultural factors. More information is needed to test these competing theories in future research.

There were no differences among racial/ethnic groups. Prior BED literature indicated racial differences in depression, but it may be that distress is distinct from depression. Among people with substance-related or other addictive disorders, there is substantial evidence of increased functional impairment among People of Color. However, confounding variables related to the sociocultural context could contribute to the discrepancy. For example, Black and Hispanic people face disproportionate risk of legal problems as a consequence of their drug use³⁷, which is unlikely to be observed in food addiction. There also were no differences among education groups after accounting for food addiction symptoms.

Study findings should be considered within the context of several limitations. First, this is a nontreatment seeking sample, which limits generalizability to treatment settings. There are also limitations to generalizability based on survey non-completion. It remains unknown if those who do not complete the survey differ systematically from those that do complete the survey. Replication with additional samples can increase confidence in the generalizability of findings. Second, emerging data suggest significant limitations associated with categorical assessment of distress/impairment³⁸. While the current approach is consistent with the validated scoring of the mYFAS2.0, it is likely that dimensional assessment could allow for a more nuanced understanding of distress/impairment. Future research should consider use of dimensional measures, such as the WHODAS^38,39 or the Clinical Impairment Assessment Questionnaire⁴⁰. Third, additional facets of diversity (e.g., sexual orientation) as well as relevant covariates (e.g., acculturation) should be considered for future research. Additionally, some racial/ethnic subgroups were relatively small (e.g., non-White Hispanic subgroup). The small sample subgroups were reflected in the large confidence intervals, indicating a lower precision in the estimate. Future research should evaluate these outcomes in more diverse and balanced samples. Finally, BMI was based on self-report. While self-reported BMI reasonably approximates objective measurements, the standard deviations can be large in some cases⁴¹ and replication in samples with objectively measured BMI is suggested.

4.1. Conclusions

This is one of the first studies to examine how individual characteristics might impact one’s understanding of food addiction as distressing/impairing, independent of the level of food addiction symptoms. The results showed that women and people with obesity were significantly more likely to identify their symptoms as distressing/impairing. These results highlight that further inquiry may be appropriate for men and people with normal weight or overweight presenting with food addictions symptoms who may otherwise deny distress/impairment.

Highlights.

• Distress and impairment are important, understudied features of food addiction (FA)

• After controlling for symptom severity, education and racial/ethnic identity did not appear to influence the experience of FA symptoms as distressing/impairing

• After controlling for symptom severity, women and people with obesity were more likely to experience FA symptoms as distressing/impairing