The case group, consisted of 40 individuals (32 women, 8 men) with a BMI over 35.0 kg/m² had a mean age of 36.05 ± 8.47 years. The control group of consecutive three BMI categories (normal, preobese, and obese) included individuals with a mean age of 32.50 ± 6.34, 36.10 ± 7.41, and 35.77 ± 6.39 years, respectively. Within the case group, 65.0% (n = 26) had a job at the time of the study and the majority were married (n = 25, 62.5%). The highest rate of employment (77.5%, n = 31) was among the normal BMI group, yet the difference between the groups did not differ significantly. The case and control groups did not differ significantly in terms of the assessed sociodemographic variables (Table 1).

The distribution of BMIs of participants in four BMI categories were normal. The mean BMIs of each group were as follows: 22.13 ± 1.92 kg/m² (18.80–25.00), 27.22 ± 1.27 kg/m² (25.20–29.80), 32.74 ± 2.59 kg/m² (30.0–38.05), 45.17 ± 4.92 kg/m² (37.80–58.90). Tobacco and alcohol use and family history were similar among the groups (Table 2), however, those with a BMI >30.0 kg/m² (n = 75) had a significantly higher rate (63.5%) of obesity in family history, when compared to the rest (χ² = 23.19, p < 0.01). The morbidly obese group had a significantly higher rate of accompanying chronic medical disorders, namely diabetes, hypertension, and hyperlipidaemia (25, 20, and 10%, respectively).

The different BMI groups differed significantly in terms of FA diagnosis by both instruments (Table 3). Food addiction was found to be more prevalent in the two groups with BMI >30 kg/m² (morbid obese, n = 40 and obese, n = 35) than in the normal (n = 40) and overweight (n = 40) individuals (p < 0.01), as measured by the YFAS (23.8 vs. 0.0%) and DSM-5 clinical interview (57.5 vs. 12.5%). In terms of severe FA as assessed by the DSM-5 (having six or more symptoms), the obesity and morbid obesity group demonstrates 8.88 times higher prevalence than the normal and overweight groups (33.3 vs. 3.7%).

Food addiction diagnosis by both instruments was associated with a higher rate of chronic medical disorders (χ² = 7.0, p < 0.01) and tobacco and alcohol use (χ² = 4.20, p = 0.04; χ² = 5.41, p = 0.02). Dieting and lifetime number of diet attempts were significantly higher in those with FA (11.22 ± 8.23; median 10) than in those without FA (6.89 ± 7.09; median 4) (z = −2.03, p = 0.04).

The most prevalent symptoms as assessed by the DSM-5 adopted clinical interview were (i) consumption of food in larger amounts or over a longer period than intended (71.3%), (ii) persistent desire or unsuccessful efforts to cut down or control (70.5%), and (iii) craving (45.1%); all indicating loss of control over food. Likewise, persistent desire or unsuccessful efforts to cut down or control (93.9%), tolerance (49.0%), and consumption despite persistent physical or psychological problems caused or exacerbated by it (46.9%) were the most frequently met criteria in YFAS assessments.

Food addiction severity, as defined by symptom count in both assessments, showed a significant correlation between YFAS (out of 7 criteria) and DSM-5 (out of 11 criteria). Greater FA severity correlated with increased BMI. Linear regression analysis showed that the severity of FA, measured as the DSM-5 symptom count predicted an increase in BMI [F_(1.153) = 49.095, p < 0.01, R² = 0.243]. The BIS-11 total and sub-scale scores did not significantly differ among BMI categories (z = −1.19, p = 0.24; z = −1.27, p = 0.21; z = −0.76, p = 0.45; z = −0.79, p = 0.43, respectively). Motor and total impulsivity scores showed a positive albeit weak correlation with the severity of FA (assessed by symptom count) but no significant correlation with BMI (Table 4).

The morbid obesity group had significantly higher rates of current BED (25.0%) and night eating syndrome (27.5%) diagnoses, as assessed by the DSM-5 criteria (p < 0.001). The total and sub-scale scores of the EDEQ were not normally distributed. A Kruskal-Wallis assessment revealed that the total and sub-scale scores of different BMI categories differed significantly, showing that higher BMI was associated with higher EDEQ scores. When the FA and non-FA groups were compared, FA was significantly associated with more severe eating symptomatology as assessed by EDEQ (Table 5).

When morbid obesity, FA, and BED diagnoses were examined together, although comorbidities were present, the majority of FA diagnoses (75%) did not meet the diagnostic criteria for BED. In the morbid obesity group, 22.5% had both FA and BED diagnoses. The comorbid group differed from the FA-only group with greater tolerance (χ² = 6.10, p = 0.01), failure to fulfill major role obligations (χ² = 9.93, p < 0.01), and higher attentional impulsivity scores (z = −2.08, p = 0.04). On the other hand, the FA-only group differentiated from the comorbid FA and BED group, as they met the two following BED criteria significantly less frequently: (i) repetitive binge eating episodes and a sense of lack of control over eating during the episode, and (ii) feeling disgusted with oneself, depressed, or very guilty after overeating (p = 0.02, p = 0.06, respectively).

Finally, a logistic regression was performed to ascertain the effects of age, gender, sociodemographic characteristics, eating disorders and impulsivity on the likelihood that participants had FA. The logistic regression model was statistically significant, [χ(11)2 = 33.46, p < 0.05]. The model explained 45.7% (Nagelkerke R²) of the variance in FA. Women were 6.7 times more likely to exhibit FA than men. The presence of BED (OR: 8.33 %95CI [1.96–22.42]; p < 0.05) and higher BIS-11 scores (OR = 1.09 %95CI [1.02–1.23]; p = 0.03) independently predicted the diagnosis of FA.

The prevalence of FA in the morbid obesity group as assessed by YFAS (15.0%) is comparable with that in two studies in the field, in which 15 and 16.9% of bariatric surgery candidates were diagnosed with FA (45, 46). However, there are other studies conducted among individuals undergoing weight-loss surgery in which higher rates such as 21.1% (47), 25.8% (48), 41.7% (49), 53.7% (50), 57.8 (38) were found. This large variation might be due to the self-report nature of YFAS, which is less-objective than a standardized clinical evaluation and that our sample had a lower mean BMI than the aforementioned studies. It is stated that DSM-IV substance dependence diagnosis corresponds to severe substance use disorder of the DSM-5 (51). Given this, the prevalence of FA in the morbid obesity group appears to be 40.0%, implying that DSM-5 criteria might be more permissive in terms of determining FA, whereas YFAS might overlook some cases. Moreover, as YFAS, which is based on the DSM-IV substance dependence criteria does not include craving, which might contribute to lower than actual rates.

Our findings based on the DSM-5 adopted clinical interviews, especially the prevalence of FA in normal-weight and obese individuals, are similar to the findings of the studies conducted using the revised version of YFAS, YFAS 2.0 (52, 53). In accordance with the studies of Meule (54) and Hauck (53), it was found that there was a moderate level correlation between the FA symptom count and BMI (ρ = 0.56) and the FA prevalence was nine times higher in individuals with a BMI >30 kg/m² than in normal-overweight individuals.

It was found that in the FA group, the dieting onset was earlier and diet-weight gain cycles were more frequent. This also supports that there might be an addictive, relapsing process behind in this subgroup of obese patients, as many other addictions also start during adolescence and early adulthood (55, 56). It is also true that eating pathologies tend to arise during this developmentally sensitive period (57).

Our findings reflect that the impulsivity scores of the case and control groups did not differ significantly from each other. However, significantly higher attentional and motor impulsivity scores were found in the FA group than in the non-FA group, supporting our hypothesis. This might be related to impaired inhibitory control or the so-called negative urgency in these individuals toward specific food products. The findings can be interpreted as impulsivity not playing a direct role in weight gain but rather contributing to obesity in a subgroup of patients who present with FA. A similar mechanism has been implicated in other studies that explore impulsivity and FA. Higher attentional impulsivity (54, 58, 59), negative urgency (60), inability to delay rewards (61), and controlling impulses in negative mood states (62) were found to be related to FA. One study found that impulsivity persists after bariatric surgery (63), which may also be suggestive of impulsivity and FA as the reason behind unsuccessful weight loss outcomes. Neurocognitive studies also suggest deficits in inhibitory functions and dysfunctional frontostriatal circuits leading to eating pathologies and loss of control over food (14).

Tolerance, withdrawal, and craving constitute the most prominent symptoms of FA, whereas in eating disorders, including BED, weight, and body image related concerns are more critical. It was confirmed in several other studies that some FA patients did not meet the BED criteria and vice versa (64, 65). Although loss of control might be the intersecting symptom of the two diagnoses, it was found that BED is related to the binge eating episodes, that is, eating a substantially large amount of food in a very short time. However, in FA, one of the most important aspects of diagnosing substance use disorder the loss of control is observed throughout the day such as grazing, resulting in eating more than intended or planned and higher caloric intake, in a fashion like tobacco consumption. This finding is parallel to the interpretation of Gearhardt and Meule, who evaluated the concept in light of the DSM-5 (66).

As in our study, female gender was repeatedly found to be associated with FA in other studies (10, 60, 67, 68). It is known that women are more likely to self-medicate than men in the acquisition phase of addiction and show a more rapid escalation of use than men, whereas in general substance use disorders are much more common in men (69). The wide availability of hyperpalatable, high-calorie, and inexpensive food and the contextual and social factors that do not necessarily prevent women from consuming these drugs, contrary to that of illicit drugs, might explain why FA is more frequently observed in women than in men.

In our study, risky alcohol use in the morbid obesity group (7.5%) was higher than the general population rates of risky alcohol use reported in other studies from Turkey (70, 71). Higher rates of risky alcohol use might be interpreted as contributing to the development of obesity. This might also mean a predisposition to addiction in this group. Indeed, the higher rate of risky alcohol use in the FA group than in the non-FA group might be related to the cross-addiction phenomenon, which signifies addiction to two or more substances or behaviors, and is frequently observed in substance-related and addictive disorders (72–74).

Our study has some limitations. First, as the study used a cross-sectional method, our findings reflect associations between variables. More prospective and long-term studies, including those which investigate both clinical and neurobiological/hormonal/metabolic correlates of FA are required to better understand the causality. The case and the control groups were not compared for their diet characteristics and food environment which may also play a role in the development of FA. Also, for the classification of body weight, only BMI was used and bioimpedance analysis was not conducted, which would have enabled a more precise comparison, a method that needs to be considered for future studies. The second version of the YFAS, which is prepared in line with the newly established diagnostic criteria for substance use disorder in the DSM-5, was not used, as it had not been validated in Turkish at the time of the study (52). Nevertheless, the contemporary approach to addictive behaviors based on DSM-5 criteria was reflected by means of a clinical interview, which also enabled us to make a complementary evaluation in addition to the self-report assessment by the YFAS. In addition, the inclusion of consecutive BMI groups and comparison of FA-related features provide a more comprehensive understanding of this construct and contribute to the discussions in the field on the validity of FA, which we believe is a strength of this study.

To conclude, our findings support the view that a group of obesity patients demonstrate characteristics of addictive behaviors and FA contributes to the development of obesity, plausibly through impulsive but not-episodic over-eating. However, FA explains only part of the variance in BMI, which is in accordance with the multiple etiologies of obesity. Further neurobiological and clinical studies in different patient groups from different cultural profiles are necessary to address the universal validity of FA and specific treatments targeting this construct, such as brief interventions or other addiction-based treatments such as relapse prevention, which may help improve the treatment outcome in obesity and may facilitate health promotion.