Food Addiction Is Strongly Associated With Psychopathology and Reduced Psychological Well-Being Among Adults Irrespective of BMI
Handling Editor: Hubertus Himmerich
Funding: This work is supported by grants from the Beckett Foundation (17-0-0822), the A.P Møller Foundation of Medical Science (17-L-0013), and the Lundbeck Foundation (grant number: R381-2021-1315) (all to C.H.). J.M.B. is employed at Steno Diabetes Centre Aarhus, Aarhus University Hospital, Denmark, which is partially funded by an unrestricted donation from the Novo Nordisk Foundation, Denmark. Outside this study, S.D.Ø. reports funding from the Lundbeck Foundation (grants R358-2020-2341 and R344-2020-1073), the Novo Nordisk Foundation (grant NNF20SA0062874), the Danish Cancer Society (grant R283-A16461), the Central Denmark Region Fund for Strengthening of Health Science (grant 1-36-72-4-20), the Danish Agency for Digitisation Investment Fund for New Technologies (Grant 2020-6720), and Independent Research Fund Denmark (Grant 7016-00048B and 2096-00055A). The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
ABSTRACT
Background and Aims
Food addiction has been linked to psychopathology and reduced psychological well-being. Here, we investigated whether these associations are mainly driven by food addiction itself or mediated via an increase in BMI.
Methods
Data stem from a nationwide survey from Denmark (n = 1474 participants). The survey questionnaire included the Yale Food Addiction Scale 2.0 (YFAS 2.0) measuring food addiction, questions on height and weight (to compute BMI), and a range of self-reported measures of psychopathology and psychological well-being. The association between food addiction and psychopathology/psychological well-being, stratified by weight category (normal weight (BMI 18.5-24.9), overweight (BMI 25–29.9) and obesity (BMI ≥ 30)), was assessed via multivariable regression analyses, adjusted for sex, age, socioeconomic status and BMI.
Results
Across all BMI categories, having food addiction was strongly positively associated with psychopathology (depression, anxiety, and interpersonal sensitivity) and strongly negatively associated with psychological well-being (all p-values < 0.001), despite adjustment for BMI. These associations remained following exclusion of participants either having received a diagnosis of mental disorder or having redeemed a prescription for psychopharmacological treatment.
Conclusion
The findings from this study are compatible with food addiction itself, and not increased BMI likely arising from it, being associated with psychopathology and reduced psychological well-being.
Summary
-
Across BMI categories, having food addiction was strongly positively associated with psychopathology and strongly negatively associated with psychological well-being, despite adjustment for BMI.
-
These findings are compatible with the hypothesis that food addiction itself, and not the increased BMI associated with it, may lead to psychopathology and reduced psychological well-being.
-
Longitudinal- and mendelian randomisation studies are needed to examine this hypothesis more closely.
1 Introduction
Food addiction is characterised by an addiction-like attraction to foods with high levels of refined carbohydrates and added fats, which, in some individuals, seem to induce neural and behavioural changes (Gearhardt and Schulte 2021; Gearhardt and DiFeliceantonio 2023; Volkow et al. 2017; Lindgren et al. 2018; LaFata and Gearhardt, 2022). Food addiction is operationalised according to the DSM-5 diagnostic criteria for substance use disorder and symptoms include compulsive consumption, craving, tolerance, and continued use despite negative consequences (American Psychiatric Association 2013; Gearhardt et al. 2016).
Prior research has identified a strong association between food addiction and both increased body mass index (BMI: kg/m2) (Gearhardt and Schulte 2021) and psychopathology (Camacho-Barcia et al. 2021; Nunes-Neto et al. 2018; Lacroix and von Ranson 2021), including clinically diagnosed mental disorders, and lower psychological well-being (Nunes-Neto et al. 2018; Lin et al. 2020; Borisenkov et al. 2018; Skinner et al. 2021; Hong et al. 2020; Brytek-Matera et al. 2021; Burrows et al. 2018; Zhao et al. 2018). Furthermore, it is well-established that high BMI is linked to depression (Luppino et al. 2010; Mannan et al. 2016; Milaneschi et al. 2019) and that this link is likely causal (Speed et al. 2019). However, it remains unclear if the association between food addiction and psychopathology/reduced psychological well-being is mainly driven by food addiction itself, or whether it is mediated via increased BMI.
Additionally, low socioeconomic status is associated with both high BMI (McLaren 2007) and psychopathology/reduced psychological well-being (Dohrenwend et al. 1992; Fryers et al. 2003). It follows that the association between the two latter could be confounded by socioeconomic status, which does not seem counterintuitive from a bio-psycho-social perspective.
Unfortunately, there is a relative paucity of studies examining the aspects outlined above. A large community sample survey from Brazil, examined psychopathological correlates of food addiction, while accounting for self-reported sociodemographic aspects, but not BMI. They found that, despite adjustment for sociodemographic status, food addiction was associated with increased self-reported depressive symptoms and interpersonal sensitivity (feelings of inferiority, self-consciousness, and discomfort in interpersonal interactions) (Nunes-Neto et al. 2018). Similarly, a randomised controlled trial from the US investigating a weight reduction intervention in individuals with obesity found that food addiction was positively associated with depressive symptoms, despite adjustment for BMI and medical comorbidities.
Thus, although the literature points toward an association between food addiction and psychopathology, beyond the influence of obesity and socioeconomics, they are marred by limitations. First, the study samples were either highly selected (Chao et al. 2017), or based on self-inclusion and recruitment through public advertising, increasing the risk of selection bias (Nunes-Neto et al. 2018; Minhas et al. 2021). Furthermore, the sociodemographic information was obtained by self-report (Nunes-Neto et al. 2018; Minhas et al. 2021), and the studies only investigated the association between food addiction and quite selected psychopathological aspects (depressive symptoms (Nunes-Neto et al. 2018; Chao et al. 2017) and interpersonal sensitivity (Nunes-Neto et al. 2018)).
To overcome the limitations of prior studies, we examined the association between food addiction and the severity of self-reported psychopathology and well-being in a large random sample of individuals from the general Danish population with fine-grained data on socioeconomics available from national registers. Specifically, the study aimed at examining (i) whether the positive associations between food addiction and psychopathology are reduced or maintained following adjustment for BMI, and socioeconomic status, and (ii) whether these associations are also consistent across weight categories.
2 Material and Methods
2.1 Data Source
This study is based on data from the Food Addiction Denmark (FADK) project, a combined survey- and register-based research project conducted in Denmark in 2018, which is described in detail elsewhere (Horsager et al. 2019). In brief, a random sample of 5000 adults aged 18–62 years were drawn from the Danish Civil Registration System (Schmidt et al. 2014), where all Danish citizens are registered with a unique numerical identifier. To be eligible for inclusion individuals had to be Danish born and have Danish born parents. This criterion ensured that invites were able to understand and read Danish. Individuals with an unregistered address were not eligible for inclusion.
2.2 Survey Procedure
The randomly drawn individuals were invited to participate in a web-based survey. The invitations were sent via the electronic mail system (eBoks (The Agency for Digitisation M of FD 2020)) used by Danish public authorities, and included information on the study purpose and a link to the survey questionnaire. If the invitees did not respond within 6–8 weeks, a reminder was forwarded via surface mail. Invitees who did not have an eBoks account were invited via surface mail and given the possibility to fill out a paper version of the questionnaire. No compensation was provided, but all respondents entered a lottery for two iPads.
2.3 Survey Measures
The survey consisted of validated rating scales on food addiction, subjective psychological well-being, and different measures of psychopathology extracted from the Hopkins Symptom Checklist (Carrozzino et al. 2016; Bech et al. 2014).
2.3.1 The Yale Food Addiction 2.0 (YFAS 2.0)
The YFAS 2.0 is based on the criteria for substance dependence in the Diagnostic and Statistical Manual of Mental Disorders Fifth Edition (DSM-5) (American Psychiatric Association 2013), and includes 35 items that evaluates food addiction symptomatology (Gearhardt et al. 2016). Symptoms are reported for the past 12 months. The YFAS 2.0 has a dual scoring option allowing for both a ‘diagnostic’ categorical scoring option with four severity levels: no food addiction, mild food addiction, moderate food addiction and severe food addiction, and a total symptom score ranging from 0 to 11. We used the categorical scoring to compare those with food addiction (at least mild food addiction) to those with no food addiction. The cut-off for food addiction is as follows: at least two of the 11 symptoms are endorsed along with ‘significant impairment’ and/or ‘significant distress’ (Gearhardt et al. 2016). The total symptom score was used for the association analysis. The total score is calculated by adding up the symptoms that are endorsed, yielding a total score between zero and 11 (Gearhardt et al. 2016).
Prior studies have demonstrated that the YFAS 2.0 has sound psychometric properties across settings and languages (Gearhardt and Schulte 2021; Oliveira et al. 2021; Penzenstadler et al. 2018; Praxedes et al. 2022; Meule and Gearhardt 2019). This includes our Danish validation (based on data from the present sample (Horsager et al. 2020) where the confirmatory factor analysis (single factor model) found item factor loadings ranging from 0.43 to 0.77 (all with p-values < 0.001), confirmatory fit index of 0.91, Tucker Lewis Index of 0.89, root-mean-square error of approximation of 0.09, and Chi2 with F = 549.59, df = 44, p < 0.001). The internal consistency measured by Kuder-Richardson alpha (KR-20) was 0.87.
2.3.2 The Five-Item WHO Well-Being Index (WHO-5)
The WHO-5 is a widely used questionnaire that examines subjective psychological well-being over the past 2 weeks with. Its psychometric validity across different age groups, cultures and study settings is well-established (Topp et al. 2015).
2.3.3 The 92-Item Hopkins Symptom Checklist (SCL-92)
The SCL-92 questionnaire examines the occurrence of a broad range of psychopathology within the past week. The full SCL-92 consists of several subscales, which have undergone psychometric validation (Carrozzino et al. 2016; Bech et al. 2014). In this study, we included the following subscales: Interpersonal sensitivity (SCL-IPS5), major depression (SCL-MDI9 excl. the SCL-92 item regarding suicidality, which was removed for ethical reasons), the Brief Depression Rating Scale (SCL-D6), and the anxiety symptom scale (SCL-ASS8) (Bech et al. 2014). The subscale score is defined as the sum of the individual item ratings.
2.3.4 Body Mass Index (BMI)
The BMI was computed from self-reported weight and height as kg/m2. The text in the questionnaire encourages participants to be as precise as possible when reporting these data. The computed BMI was used to define the following weight categories: underweight (BMI < 18.5 kg/m2), normal weight (BMI: 18.5–24.9 kg/m2), overweight (BMI: 25–29.9 kg/m2), and obesity (BMI: ≥ 30 kg/m2).
A total of 1506 individuals responded to the survey corresponding to a response rate of 30.1%. For the present study, we only included data from survey participants who had filled in the YFAS 2.0 and reported both height and weight. Furthermore, participants with underweight (n = 32) were excluded as they likely represent a subgroup, where a positive food addiction screen is rather an expression of a subjective experience of overconsumption and loss of control than actual food addiction. Thus, the final sample size for the present study was n = 1474.
2.4 Register Data
Data from Danish registers (Thygesen et al. 2011) were available for all respondents included in the present study. Accurate linkage was ensured via the 10-digit unique identifier assigned to all people at birth or when obtaining legal residency.
2.4.1 Socioeconomic Status
Data on educational level (Lower secondary school; High school; Vocational or short-cycle higher education; Medium-cycle higher education including bachelor degrees; and Long-cycle higher education), occupational status (In the labour force; Unemployment, sick pay, leave of absence; Disability pension, social security benefit; and Enroled in education), and personal income (categorised in quintiles: < 21,906 euro; 21,906 euro—38,145 euro; 38,146 euro—48,914 euro; 48,915 euro—63,329 euro; and > 63,329 euro) were available from the Register on Personal Level of Education, the Registers on Personal Labour Market Affiliation, and from the Income Statistics Register (Jensen and Rasmussen 2011; Petersson et al. 2011; Baadsgaard and Quitzau 2011).
2.4.2 Prior Diagnosis of Mental Disorder
Data on diagnoses assigned following and inpatient and outpatient contacts with a Danish psychiatric hospital in the period from 1969 to 2017 were available from the Danish Psychiatric Central Research Register (Mors et al. 2011). This register contains diagnoses from inpatient contacts from 1968-1994, and for diagnoses from inpatient, and outpatient contacts since 1995. The ICD-8 was used as diagnostic reference from 1969-1993 and was replaced by the ICD-10 in 1994. For the present study, we used these data to assign a binary value of any prior diagnosis of mental disorder (yes/no) to each participant.
2.4.3 Prior Psychopharmacological Treatment
Data on prescriptions redeemed at pharmacies in Denmark since 1995 were available from the Danish National Prescription Register (Pottegård et al. 2017). We used these data to assign a binary value of prior psychopharmacological treatment (yes/no) to each participant. The following groups of medications (Anatomical Therapeutic Chemical classification code (ATC-code) were included in this composite variable; antipsychotics (ATC-code: N05A—N05AX17 excl. N05AN); Lithium (ATC-code: N05AN); Anxiolytics (ATC-code: N05BA, N05CD02, N05CD05, N05CD06, N03AX16); Antidepressants (ATC-code: N06A—N06AX26); Medication for attention-deficit hyperactivity disorder (ATC-codes: N06BA09, N06BA04, N06BA12, N06BA02, N06BA07, C02AC02); Medication for addiction disorders (ATC-codes: N07BB, N07BC).
2.5 Statistical Analyses
All analyses were conducted using STATA version 18.0. Prior to all analyses, their assumptions were tested, and if assumptions were not met, sensitivity analyses using bootstrap with 1000 iteration were conducted.
Descriptive statistics were used to characterise individuals with and without food addiction, stratified by weight category. Comparisons between individuals with and without food addiction were carried out using Chi2-test, Fischers exact test and students t-test, as appropriate.
The association between food addiction (total symptom score on the YFAS 2.0) and self-reported psychopathology/psychological well-being (total scale/subscale scores on the individual measures) was examined using linear regression for the 1474 participants as a whole and for those with normal weight, overweight and obesity, respectively. Both crude and adjusted analyses were conducted. The covariates for adjustment were chosen a priory and a 3-step approach was applied as follows: Model i) adjusted for sex and age, Model ii) adjusted for sex, age and self-reported BMI (calculated based on self-reported weight and height available from the survey data: weight/height2 (kg/m2)), respectively), and Model iii) adjusted for sex, age, BMI and socioeconomic factors (educational level, disposal income, and affiliation with the labour market). We chose to conduct the association analysis both on the group as a whole and within the three weight categories separately as the relationship between food addiction and psychopathology/psychological well-being, and the role of BMI in this context, may not necessarily be uniform across the weight categories. In robustness analyses, as an attempt to narrow in on the effect of food addiction upon psychopathology and psychological well-being and not vice versa, the linear regressions described above were repeated following exclusion of participants with a prior diagnosis of mental disorder or having received psychopharmacological treatment.
2.6 Ethics
The respondents provided indirect informed consent as it was explicitly stated in the survey invitation that filling in the questionnaire was a consent for research use. Furthermore, the FADK project was approved by Statistics Denmark and the Danish Health Data Authority. The study complies with the General Data Protection Regulation and is registered on the North Denmark Region internal list of research projects.
3 Results
The characteristics of the 1474 survey participants included in this study, stratified on weight category (normal weight, overweight and obesity) and food addiction status (yes/no), are listed in Table 1. The mean age of the participants was 44.1 years and 59.0% were female. A total of 9.3% met the criteria for food addiction.
| Characteristics | All participants (normal weight, overweight, and obesity) (n = 1474) BMI ≥ 18.5 | Participants with normal weight (n = 737) BMI = 18.5-24.9 | Participants with overweight (n = 498) BMI = 25-29.9 | Participants with obesity (n = 239) BMI ≥ 30 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Food addiction (n = 137) | No food addiction (n = 1337) | p-valuea | Food addiction (n = 30) | No food addiction (n = 707) | p-valuea | Food addiction (n = 41) | No food addiction (n = 457) | p-valuea | Food addiction (n = 66) | No food addiction (n = 173) | p-valuea | |
| Age in years (SE) | 39.8 (1.10) | 44.8 (0.34) | < 0.001 | 35.4 (2.60) | 42.8 (0.48) | < 0.001 | 43.0 (1.89) | 47.0 (0.54) | 0.034 | 39.8 (1.53) | 46.7 (0.81) | < 0.001 |
| Female sex (%) | 108 (78.8) | 753 (56.3) | 28 (93.3) | 461 (65.2) | 33 (80.5) | 205 (44.9) | 47 (71.2) | 87 (50.3) | ||||
| Male sex (%) | 29 (21.2) | 584 (43.7) | N/A | 246 (34.8) | 8 (19.5) | 252 (55.1) | 19 (28.8) | 86 (49.7) | ||||
| < 0.001 | < 0.001 | < 0.001 | 0.004 | |||||||||
| Civil status (%) | ||||||||||||
| Married or cohabiting | 88 (64.2) | 972 (72.7) | 17 (56.7) | 507 (71.7) | 28 (68.3) | 348 (76.2) | 43 (65.2) | 117 (67.6) | ||||
| Single | 49 (35.8) | 365 (27.3) | 13 (43.3) | 200 (28.3) | 13 (31.7) | 109 (23.9) | 23 (34.9) | 56 (32.4) | ||||
| 0.036 | 0.075 | 0.263 | 0.716 | |||||||||
| Educational level (%) | ||||||||||||
| Lower secondary school | 23 (16.8) | 210 (15.7) | 5 (16.7) | 98 (13.9) | 6 (14.6) | 74 (16.2) | 12 (18.2) | 38 (22.0) | ||||
| High school | 22 (16.1.0) | 144 (10.8) | 9 (30.0) | 91 (12.9) | 5 (12.2) | 39 (8.5) | 8 (12.1) | 14 (8.1) | ||||
| Vocational or short-cycle higher education | 48 (35.0) | 500 (37.4) | 7 (23.3) | 216 (30.6) | 16 (39.0) | 208 (45.5) | 25 (37.9) | 76 (43.9) | ||||
| Medium-cycle higher education including bachelor | 31 (22.6) | 298 (22.3) | 7 (23.3) | 174 (24.6) | 7 (17.1) | 89 (19.5) | 17 (25.8) | 35 (20.2) | ||||
| Long-cycle higher education | 13 (9.5) | 184 (13.8) | N/A | 127 (18.0) | 7 (17.1) | 47 (10.3) | 4 (6.1) | 10 (5.8) | ||||
| Missing n < 5 | 0.395 | 0.107 | 0.611 | 0.688 | ||||||||
| Occupation status (%) | ||||||||||||
| In the labour force | 95 (69.3) | 1051 (78.6) | 17 (56.7) | 531 (75.1) | 31 (75.6) | 380 (83.2) | 47 (71.2) | 140 (80.9) | ||||
| Unemployment, sick pay, leave of absence | 6 (4.4) | 34 (2.5) | N/A | 12 (1.7) | N/A | 14 (3.1) | 5 (7.6) | 8 (4.6) | ||||
| Disability pension, social security benefit | 7 (5.1) | 89 (6.7) | N/A | 46 (6.5) | N/A | 28 (6.1) | N/A | 15 (8.7) | ||||
| Enroled in education | 25 (18.3) | 136 (10.2) | 12 (40.0) | 98 (13.9) | 5 (12.2) | 30 (6.6) | 8 (12.1) | 8 (4.6) | ||||
| Missing n = 31b | 0.024 | 0.003 | 0.204 | 0.149 | ||||||||
| Personal income (%)Quintiles | ||||||||||||
| < 21,906 euro | 31 (22.6) | 204 (15.3) | 11 (36.7) | 131 (18.5) | 7 (17.1) | 54 (11.8) | 13 (19.7) | 19 (11.0) | ||||
| 21,906 euro—38,145 euro | 29 (21.2) | 210 (15.7) | 5 (16.7) | 106 (15.0) | 5 (12.2) | 63 (13.8) | 19 (28.8) | 41 (23.7) | ||||
| 38,146 euro—48,914 euro | 27 (19.7) | 267 (20.0) | 5 (16.7) | 123 (17.4) | 10 (24.4) | 102 (22.3) | 12 (18.2) | 42 (24.3) | ||||
| 48,915 euro—63,329 euro | 26 (19.0) | 314 (23.5) | 6 (20.0) | 174 (24.6) | 8 (19.5) | 105 (23.0) | 12 (18.2) | 35 (20.2) | ||||
| > 63,329 euro | 24 (17.5) | 340 (25.4) | N/A | 173 (24.5) | 11 (26.8) | 131 (28.7) | 10 (15.2) | 36 (20.8) | ||||
| Missing n < 5b | 0.052 | 0.101 | 0.923 | 0.299 | ||||||||
| Prior diagnosis of mental disorder (%) | 27 (19.7) | 91 (6.8) | < 0.001 | N/A | 57 (8.1) | 0.305 | 10 (24.4) | 20 (4.4) | < 0.001 | 13 (19.7) | 14 (8.1) | 0.011 |
| Prior psychopharmacological treatment (%) | 67 (48.9) | 395 (29.5) | < 0.001 | 13 (43.3) | 199 (28.2) | 0.072 | 20 (48.8) | 130 (28.6) | 0.007 | 34 (51.5) | 66 (38.2) | 0.061 |
|
BMI median (IQR) (n = 1474) |
29.5 (9.7) | 24.7 (4.8) | < 0.001 | 22.3 (2.6) | 22.4 (2.6) | 0.637 | 27.1 (2.8) | 26.8 (1.9) | 0.082 | 35.5 (6.60) | 33.1 (5.15) | 0.170 |
|
WHO-5 mean (SE) (n = 1543) |
49.1 (1.93) | 69.9 (0.46) | < 0.001 | 52.8 (4.12) | 70.4 (0.64) | < 0.001 |
N = 493 48.2 (3.16) |
70.6 (0.77) | < 0.001 |
N = 237 48.0 (3.0) |
66.1 (1.38) | < 0.001 |
|
Depression (SCL-D6) Mean (SE) (n = 1441) |
12.1 (0.51) | 4.4 (0.13) | < 0.001 |
N = 728 10.5 (1.17) |
4.3 (0.17) | < 0.001 |
N = 479 12.3 (0.83) |
4.1 (0.20) | < 0.001 |
N = 234 12.7 (0.76) |
5.5 (0.39) | < 0.001 |
|
Depression (SCL-MDI9)b Mean (SE) (n = 1376) |
2.0 (0.09) | 0.7 (0.02) | < 0.001 |
N = 701 1.7 (0.18) |
0.7 (0.03) | < 0.001 |
N = 460 2.1 (0.13) |
0.7 (0.03) | < 0.001 |
N = 215 2.1 (0.13) |
0.8 (0.06) | < 0.001 |
|
Anxiety (SCL-ASS8) Mean (SE) (n = 1438) |
9.4 (0.59) | 3.1 (0.11) | < 0.001 |
N = 726 7.9 (1.16) |
3.1 (0.15) | < 0.001 |
N = 481 9.1 (1.0) |
2.8 (0.17) | < 0.001 |
N = 231 10.3 (0.92) |
3.8 (0.35) | < 0.001 |
|
Interpersonal sensitivity (SCL-IPS5) Mean (SE) (n = 1429) |
13.7 (0.76) | 4.2 (0.14) | < 0.001 |
N = 725 12.1 (1.72) |
4.0 (0.20) | < 0.001 |
N = 479 14.2 (1.22) |
4.1 (0.23) | < 0.001 |
N = 225 14.2 (1.15) |
5.3 (0.47) | < 0.001 |
- Abbreviations: ASS8, the anxiety symptom scale; D6, 6-item Hamilton Depression Rating Scale; MDI9, Major depression inventory; SCL, The Hopkins Symptom Checklist; WHO-5, The WHO-5 Well-being Index.
- a Comparing respondents with food addition to respondents without food addiction. All tests are performed as Chi2 tests except for the comparison of age, BMI, and self-reported psychiatric measures between groups, where the two-sample t-test was used.
- b Major depression inventory based on nine items as the item regarding suicidality was taken out—the value therefore refers to the mean score pr. item.
Participants with food addiction were generally younger (p < 0.001), and more often of female sex (p < 0.001) compared to individuals without food addiction. With regard to socioeconomics, there was generally no or very little difference in educational level and personal income between those with and without food addiction. Compared to the participants without food addiction, a substantially larger proportion of those with food addiction had a prior diagnosis of mental disorder (6.8% vs. 19.7%, p < 0.001) or had received psychopharmacological treatment (29.5% vs. 48.9%, p < 0.001).
Irrespective of weight category (normal weight, overweight and obesity compiled or apart), those with food addiction reported significantly poorer subjective psychological well-being (WHO-5) with an around 20-point difference in score compared to the participants in the same weight category without food addiction (49.1 vs. 69.9, p < 0.001 in the compiled weight category group; Table 1). Furthermore, participants with food addiction were more likely to report depressive symptoms, anxiety symptoms, and interpersonal sensitivity.
The results of the analyses examining the association between food addiction (total symptom score on the YFAS 2.0) and psychopathology/psychological well-being are listed in Table 2 (all respondents), Table 3 (normal weight), Table 4 (overweight), and Table 5 (obesity). Across these four BMI categories, and across the four levels of adjustment, there was a strong and statistically significant positive association between food addiction and self-reported psychopathology and a strong and statistically significant negative association between food addiction and psychological well-being. Notably, the strength of the associations varied little across the BMI categories, and across the levels of adjustment.
| Crude | Model 1a | Model 2b | Model 3c | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Coeff. (diff.) | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | |
| WHO-5 (n = 1459) | −20.8 | 1.57 | −23.9;−17.7 | < 0.001 | −19.2 | 1.57 | −22.2; −16.1 | < 0.001 | −17.3 | 1.62 | −20.5; 14.1 | < 0.001 | −17.3 | 1.62 | −20.5;−14.1 | < 0.001 |
| Depression (SCL-D6) (n = 1441) | 7.7 | 0.42 | 6.9; 8.5 | < 0.001 | 7.2 | 0.42 | 6.3; 8.0 | < 0.001 | 6.6 | 0.44 | 5.8:7.5 | < 0.001 | 6.6 | 0.44 | 5.8; 7.5 | < 0.001 |
| Depression (SCL-MDI9) n = 1376) | 1.3 | 0.07 | 1.2; 1.4 | < 0.001 | 1.2 | 0.07 | 1.1; 1.4 | < 0.001 | 1.1 | 0.07 | 1.0; 1.3 | < 0.001 | 1.1 | 0.07 | 1.0; 1.3 | < 0.001 |
| Anxiety (SCL-ASS8) (n = 1438) | 6.3 | 0.39 | 5.5; 7.0 | < 0.001 | 5.8 | 0.39 | 5.1; 6.6 | < 0.001 | 5.4 | 0.40 | 4.7; 6.2 | < 0.001 | 5.4 | 0.40 | 4.6; 6.2 | < 0.001 |
| Interpersonal sensitivity (SCL-IPS5) (n = 1429) | 9.5 | 0.51 | 8.5; 10.5 | < 0.001 | 8.8 | 0.51 | 7.8; 9.8 | < 0.001 | 8.3 | 0.52 | 7.3; 9.3 | < 0.001 | 8.3 | 0.52 | 7.2; 9.3 | < 0.001 |
- Abbreviations: ASS8, the anxiety symptom scale; D6, 6-item Hamilton Depression Rating Scale; MDI9, Major depression inventory; SCL, The Hopkins Symptom Checklist; WHO-5, The WHO-5 Well-being Index.
- a Model 1 adjusted for sex and age.
- b Model 2 adjusted for sex, age and BMI.d
- c Model 3 adjusted for sex, age, BMId and socioeconomic factors (educational level, disposal income, and affiliation with the labour market).
- d Based on self-reported height and weight.
| Crude | Model 1a | Model 2b | Model 3c | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Coeff. (diff.) | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | |
| WHO-5 (n = 729) | −17.6 | 3.18 | −23.8;-11.3 | < 0.001 | −15.0 | 3.17 | −21.3;-8.8 | < 0.001 | −15.0 | 3.18 | −21.2;-8.7 | < 0.001 | −15.0 | 3.17 | −21.2;-8.8 | < 0.001 |
| Depression (SCL-D6) (n = 728) | 6.2 | 0.87 | 4.5; 7.9 | < 0.001 | 5.2 | 0.86 | 3.5; 6.9 | < 0.001 | 5.2 | 0.86 | 3.5:6.9 | < 0.001 | 4.9 | 0.85 | 3.3; 6.6 | < 0.001 |
| Depression (SCL-MDI9) (n = 701) | 1.0 | 0.13 | 0.7; 1.3 | < 0.001 | 0.9 | 0.13 | 0.6; 1.1 | < 0.001 | 0.9 | 0.13 | 0.6; 1.1 | < 0.001 | 0.8 | 0.13 | 0.6; 1.1 | < 0.001 |
| Anxiety (SCL-ASS8) (n = 726) | 4.7 | 0.78 | 3.2; 6.3 | < 0.001 | 4.0 | 0.77 | 2.4; 5.5 | < 0.001 | 3.9 | 0.77 | 2.4; 5.5 | < 0.001 | 3.7 | 0.76 | 2.2; 5.2 | < 0.001 |
| Interpersonal sensitivity (SCL-IPS5) (n = 725) | 8.0 | 1.0 | 6.0; 10.0 | < 0.001 | 6.9 | 0.99 | 5.0; 8.9 | < 0.001 | 6.9 | 0.99 | 4.9; 8.8 | < 0.001 | 6.5 | 0.98 | 4.6; 8.5 | < 0.001 |
- Abbreviations: ASS8, the anxiety symptom scale; D6, 6-item Hamilton Depression Rating Scale; MDI9, Major depression inventory; SCL, The Hopkins Symptom Checklist; WHO-5, The WHO-5 Well-being Index.
- a Model 1 adjusted for sex and age.
- b Model 2 adjusted for sex, age and BMI.d.
- c Model 3 adjusted for sex, age, BMId and socioeconomic factors (educational level, disposal income, and affiliation with the labour market).
- d Based on self-reported height and weight.
|
(n = 498) |
Crude | Model 1a | Model 2b | Model 3c | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Coeff. (diff.) | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | |
| WHO-5 (n = 493) | −22.5 | 2.71 | −27.8;−17.1 | < 0.001 | −21.1 | 2.72 | −26.4;−15.7 | < 0.001 | −20.8 | 2.73 | −26.2;−15.5 | < 0.001 | −20.8 | 2.73 | −26.2;−15.4 | < 0.001 |
| Depression (SCL-D6) (n = 479) | 8.2 | 0.70 | 6.8; 9.6 | < 0.001 | 7.8 | 0.71 | 6.4; 9.2 | < 0.001 | 7.8 | 0.71 | 6.3:9.2 | < 0.001 | 7.8 | 0.71 | 6.4; 9.2 | < 0.001 |
| Depression (SCL-MDI9) (n = 460) | 1.4 | 0.11 | 1.2; 1.6 | < 0.001 | 1.3 | 0.11 | 1.1; 1.5 | < 0.001 | 1.3 | 0.11 | 1.1; 1.5 | < 0.001 | 1.3 | 0.11 | 1.1; 1.5 | < 0.001 |
| Anxiety (SCL-ASS8) (n = 481) | 6.3 | 0.64 | 5.1; 7.6 | < 0.001 | 6.0 | 0.64 | 4.8; 7.3 | < 0.001 | 6.1 | 0.64 | 4.8; 7.3 | < 0.001 | 6.1 | 0.64 | 4.9; 7.4 | < 0.001 |
| Interpersonal sensitivity (SCL-IPS5) (n = 479) | 10.1 | 0.85 | 8.5; 11.8 | < 0.001 | 9.7 | 0.85 | 8.0; 11.4 | < 0.001 | 9.7 | 0.85 | 8.0; 11.3 | < 0.001 | 9.8 | 0.84 | 8.1; 11.4 | < 0.001 |
- Abbreviations: ASS8, the anxiety symptom scale; D6, 6-item Hamilton Depression Rating Scale; MDI9, Major depression inventory; SCL, The Hopkins Symptom Checklist; WHO-5, The WHO-5 Well-being Index.
- a Model 1 adjusted for sex and age.
- b Model 2 adjusted for sex, age and BMI.d.
- c Model 3 adjusted for sex, age, BMId and socioeconomic factors (educational level, disposal income, and affiliation with the labour market).
- d Based on self-reported height and weight.
| N = 239 | Crude | Model 1a | Model 2b | Model 3c | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Coeff. (diff.) | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | Coeff. | SE | 95% CI | p-value | |
| WHO-5 (n = 237) | −18.1 | 2.90 | −23.8;−12.4 | < 0.001 | −15.0 | 2.98 | −20.9;−9.1 | < 0.001 | −14.6 | 2.99 | −20.5;−8.7 | < 0.001 | −14.3 | 2.99 | −20.2;−8.4 | < 0.001 |
| Depression (SCL-D6) (n = 234) | 7.2 | 0.79 | 5.6; 8.8 | < 0.001 | 6.3 | 0.81 | 4.7; 7.9 | < 0.001 | 6.3 | 0.82 | 4.7:7.9 | < 0.001 | 6.3 | 0.82 | 4.7; 7.9 | < 0.001 |
| Depression (SCL-MDI9) (n = 215) | 1.3 | 0.13 | 0.9; 1.5 | < 0.001 | 1.1 | 0.13 | 0.9; 1.4 | < 0.001 | 1.1 | 0.13 | 0.9; 1.4 | < 0.001 | 1.1 | 0.13 | 0.9; 1.4 | < 0.001 |
| Anxiety (SCL-ASS8) (n = 231) | 6.5 | 0.80 | 4.9; 8.0 | < 0.001 | 5.4 | 0.81 | 3.8; 7.0 | < 0.001 | 5.4 | 0.82 | 3.8; 7.0 | < 0.001 | 5.4 | 0.81 | 3.8; 7.0 | < 0.001 |
| Interpersonal sensitivity (SCL-IPS5) (n = 225) | 8.9 | 1.04 | 6.8; 10.9 | < 0.001 | 7.6 | 1.06 | 5.5; 9.6 | < 0.001 | 7.5 | 1.06 | 5.4; 9.6 | < 0.001 | 7.5 | 1.06 | 5.5; 9.6 | < 0.001 |
- Abbreviations: ASS8, the anxiety symptom scale; D6, 6-item Hamilton Depression Rating Scale; MDI9, Major depression inventory; SCL, The Hopkins Symptom Checklist; WHO-5, The WHO-5 Well-being Index.
- a Model 1 adjusted for sex and age.
- b Model 2 adjusted for sex, age and self-reported BMId.
- c Model 3 adjusted for sex, age, BMId and socioeconomic factors (educational level, disposal income, and affiliation with the labour market).
- d Based on self-reported height and weight.
The results of the robustness analyses examining the association between food addiction and psychopathology/psychological well-being following exclusion of the 486 participants with either a prior diagnosis of mental disorder or having received psychopharmacological treatment are listed in Tables S1–S4. While the strength of the associations is generally slightly attenuated compared to the main analyses, there remained a strong and statistically significant positive association between food addiction and self-reported psychopathology and a strong and statistically significant negative association between food addiction and psychological well-being across the four BMI categories, and across the four levels of adjustment.
4 Discussion
This study aimed at examining whether the association between food addiction and psychopathology/reduced psychological well-being is independent of socioeconomic status and, in particular, BMI. This was investigated based on a dataset covering a sample of adults from the general Danish population. Both in the sample as a whole, and across BMI categories (normal weight, overweight and obesity), having food addiction was strongly positively associated with self-reported depressive symptoms, anxiety symptoms, and interpersonal sensitivity, and strongly negatively associated with self-reported psychological well-being, despite adjustment for socioeconomic status and BMI (and sex and age). Notably, the strength of the associations varied little across the BMI categories, and the adjustment for socioeconomic status and BMI only had a minor impact. These associations remained following exclusion of participants with either a prior diagnosis of mental disorder or having received psychopharmacological treatment.
That food addiction was found to be associated with depression, anxiety and interpersonal sensitivity aligns well with findings from studies finding that food addiction is prevalent in individuals with clinically verified depressive disorder and anxiety disorders (Horsager et al. 2021; Lima et al. 2024; Benzerouk et al. 2018; Horsager et al. 2023; Mills et al. 2018; Piccinni et al. 2021). Furthermore, several studies have found that individuals with food addiction/addictive eating patterns tends to report more depressive, anxiety and interpersonal sensitivity symptoms compared to individuals without food addiction (Nunes-Neto et al. 2018; Skinner et al. 2021; Burrows et al. 2018). The current study is, however, to our knowledge, the first study to reproduce this finding in a random sample from a general population, accounting for both socioeconomic status and BMI. These findings are compatible with food addiction itself, and not the increased BMI arising from it, being associated with psychopathology, as appears to be the case for other addiction disorders (Regier et al. 1990; Plana-Ripoll et al. 2019; Toftdahl et al. 2016).
In accordance with the findings on psychopathology reported above, we found that food addiction was strongly associated with reduced psychological well-being. Indeed, the mean WHO-5 total score, which ranges from 0–100, was 21 points lower among those with food addiction (mean: 49) compared to those without food addiction (mean: 70). To put this into context, a WHO-5 total score below 50 is a quite sensitive and specific marker for depression (Topp et al. 2015). This finding is in line with results from prior studies reporting on the association between food addiction and poorer well-being/quality of life (Camacho-Barcia et al. 2021; Nunes-Neto et al. 2018; Lacroix and von Ranson 2021; Zhao et al. 2018; Rose et al. 2017).
Interestingly, we did not find marked differences in socioeconomic status between the group with and without food addiction. Prior studies have found food addiction to be associated with lower income and food insecurity (LaFata et al. 2024; Parnarouskis et al. 2022). These studies, however, stem from the US. Thus, this difference could be due to a combination of higher socioeconomic inequality and increased availability (and lower cost) of ultra-processed foods in the US compared to Denmark (Vandevijvere et al. 2019).
The findings from this study are compatible with the hypothesis that food addiction itself, and not the increased BMI associated with it, may lead to psychopathology and reduced psychological well-being. If this is confirmed by future studies that can leverage causality, it has important clinical implications. Specifically, it would imply that treating/preventing food addiction itself will lead to less psychopathology/increased psychological well-being. Therefore, as further outlined below, the field would benefit tremendously by studies leveraging causality.
While the large random sample and the availability of register-based data on socioeconomics are strengths of the present study, there are also several limitations. First and foremost, the study providing data for these analyses was cross-sectional and does not provide any evidence on the temporality, nor causality, of the reported associations. Specifically, we cannot determine whether food addiction preceded the psychopathology or vice versa. However, the fact that the associations between food addiction and psychopathology/reduced psychological well-being remained strong following exclusion of participants with a prior diagnosis of mental disorder or having received psychopharmacological treatment is compatible with the hypothesis that food addiction may lead to psychopathology/reduced psychological well-being. To test this hypothesis further, we plan to investigate the temporality of the relationship between food addiction and psychopathology/reduced psychological well-being in follow-up studies on this general population sample. Also, if larger genome-wide association studies of food addiction are conducted (Cornelis et al. 2016), it may be possible to use bidirectional mendelian randomisation (Zheng et al. 2017) to test whether the relationship between food addiction and mental illness is causal—including the direction of such causality (Speed et al. 2019; Byg et al. 2023). Such studies will also provide a base for examining the mechanisms driving a causal effect (if verified) of food addiction on psychopathology/psychological well-being. Second, data from the survey were self-reported which introduces a risk of report-bias. Determining the direction and magnitude of this potential bias is not possible based on the data at hand. Third, the response rate of 30.1% comes with a risk of selection bias. However, in a previous study based on the same sample, we did a comprehensive attrition analysis, which indicated that attrition was not a major problem (Horsager et al. 2020). Fourth, our analyses did not factor in the role of perceived weight stigma, which has recently, and after the collection of the data for the present study, been suggested to play a role in the complex relationship between food addiction and psychopathology/psychological well-being (distress) (Huang et al. 2023, 2022, 2024). Fifth, as all participants were living in Denmark and have Danish born parents, the results may not necessarily generalise well to other countries and/or more ethnically diverse populations. Finally, it should be mentioned that there is some controversy with regard to the concept of food addiction. Specifically, it is debated whether food addiction is indeed a valid clinical entity, or if it is in fact better classified as a behavioural addiction, or already covered by the eating disorders that are included in current diagnostic classification systems (Fletcher and Kenny 2018; Gearhardt and Hebebrand 2021; Gordon et al. 2018; Albayrak and Hebebrand 2015). The present study was not designed to inform this debate.
5 Conclusions
In conclusion, the findings from this study is compatible with the hypothesis that food addiction itself, and not the increased BMI associated with it, may lead to psychopathology and reduced psychological well-being. However, longitudinal- and mendelian randomisation studies are needed to examine this hypothesis more closely.
Author Contributions
The study was designed by C.H., E.F. and S.D.Ø. The data were analysed by C.H. and E.F. The results were interpreted by C.H., J.M.B., M.B.L. and S.D.Ø. The manuscript was draughted by C.H. and S.D.Ø., and revised critically for important intellectual content by J.M.B., M.B.L., and E.F. The final version of the manuscript was approved by all authors prior to submission.
Consent
The respondents provided indirect informed consent as it was explicitly stated in the survey invitation that filling in the questionnaire was a consent for research use. Furthermore, they were informed, that their consent to participate could be withdrawn at any time.
Conflicts of Interest
S.D.Ø received the 2020 Lundbeck Foundation Young Investigator Prize. Furthermore, S.D.Ø owns/has owned units of mutual funds with stock tickers DKIGI, IAIMWC, SPIC25KL and WEKAFKI, and owns/has owned units of exchange traded funds with stock tickers BATE, TRET, QDV5, QDVH, QDVE, SADM, IQQH, USPY, EXH2, 2B76, IS4S, EUNL and SXRV. The remaining authors report no conflicts of interest.
Open Research
Data Availability Statement
According to Danish legislation, the original (individual-level) data cannot be shared due to its personal sensitive nature.
