Ultra-Processed Food Addiction

The prevalence of obesity doubled since the 1980s, resulting in exponentially higher incidence rates of diet-related diseases [1^, 2]. These drastic changes cannot be attributed to rapid increases in the genetic predisposition for obesity only nor sudden reductions in individuals’ personal responsibility in making food decisions [3^, 4]. Decreases in physical activity, due to industrialization onset in the 1910s, predated the uptick in obesity rates [1]. Converging evidence convincingly demonstrates that the timeline of rising obesity rates has mirrored increases in the availability of calorie-dense, nutrient-poor ultra-processed foods (UPFs) in the environment, thus implicating UPFs as a primary causal driver of the obesity pandemic [1].

The term “UPF” originated from the NOVA Food Classification System [5], which is the most prominent approach used across disciplines in the scientific literature for categorizing foods based on their level of processing [6^, 7]. Industrial processing is common in our modern food supply and involves techniques that can be beneficial (e.g., canning vegetables to increase shelf life) or harmful (e.g., creating cheap, nutrient-poor ingredients like corn syrup and refined oils) for public health [8]. Thus, the NOVA system was created to differentiate between forms of food processing and underscore the uniquely problematic natures of UPFs [8^, 9]. While numerous terms have been used to describe rewarding foods [10^, 11] and adoption of a uniform term remains controversial [12^-14], we use the UPF term to align with the widely used NOVA system.

NOVA categorizes UPFs as group 4 foods, reflecting the highest degree of processing and greatest deviation from naturally occurring foods [5]. UPFs have few, if any, whole food ingredients [5]. Instead, UPFs have been industrially formulated to maximize palatability (and therefore profitability) by combining artificially elevated amounts of rewarding ingredients like hydrogenated oils, refined carbohydrates (e.g., white flour, added sugars), sodium, and other additives (e.g., flavor enhancers) [5^, 15]. Categories of UPFs include packaged snacks (e.g., potato chips, cookies), candy, frozen desserts, packaged white flour bread/pasta, sugary breakfast cereals/bars, fast foods (e.g., pizza, French fries), and beverages with added sugars (e.g., soda, sweet tea, sports drinks) (for a comprehensive description and list, see [16]). While homemade versions of some of the aforementioned foods have been enjoyed for generations (e.g., cookies, breads), research has demonstrated that the recent uptick in the availability of industry-created UPFs poses the greatest public health risk, as UPFs are conveniently accessible, affordable, heavily marketed, and dominate the food supply [5^-9].

Notably, emerging evidence suggests that UPFs are so highly reinforcing that they may trigger addictive biological and behavioral responses, thus directly driving forward excessive patterns of food intake and contributing to obesity [13]. While the changing food environment has been considered extensively in the context of obesity [17], less attention has been paid to how potentially addictive UPFs may also contribute to disordered eating, particularly binge-type eating disorders. Thus, the overarching aim of this editorial is to evaluate how food addiction (FA), a clinical presentation consistent with a substance-based addiction to UPFs, encompasses unique theoretical, phenotypic, and treatment perspectives relative to obesity and existing eating disorders. Specifically, we review (1) how the changing food environment contributes to problematic eating behavior, (2) the operationalization of FA, (3) the association between FA and obesity, (4) the association between FA and eating disorders, and (5) the diagnostic and public health implications of FA.

The Changing Food Environment and Eating Behavior

The rise in UPFs in the modern food environment began in the 1980s [18], which coincides with when major tobacco corporations acquired large food companies (e.g., Kraft, General Foods) and became the biggest producers of UPFs [19^, 20]. Tobacco industry documents reveal that tactics used to formulate cigarettes to be as addictive as possible have been applied to create reinforcing UPFs with optimal combinations of rewarding ingredients (e.g., fat, sugar) to maximize palatability and profitability [20^-22]. The food industry has been extremely successful in profiting from UPFs, which now account for over 60% of all calories consumed in the USA [23]. Emerging empirical evidence supports the direct role that UPFs play in motivating overeating behavior. For instance, an inpatient-feeding trial demonstrated that eating a diet of UPFs, compared to a macronutrient-matched diet of minimally processed whole foods, led individuals to consume 500 more calories per day [24].

Research has not only linked UPFs to obesity and all-cause mortality [25^, 26], but UPFs are also the predominant food type consumed during binge episodes [11^, 24^, 27^-32]. A review of food diaries of individuals with eating disorders found that 100% of the foods consumed in binge episodes were UPFs [33]. While anorexia nervosa (AN) (i.e., restrictive eating leading to dangerously low body weight) was included in the first Diagnostic and Statistical Manual of Mental Disorders (DSM) in the 1950s [34], the clinical emergence of binge-type eating disorders occurred between 1980 and 2013 [34]. Thus, when UPFs began to increasingly dominate the food environment, clinical and academic recognition of binge-type eating disorders increased substantially. This suggests that greater exposure to UPFs may be a key factor in the emergence of binge-type eating disorders. However, traditional approaches to conceptualizing eating disorders have focused less on whether UPFs may directly contribute to binge eating.

In contrast, the FA model emphasizes that the attributes of UPFs, the proposed addictive substance, trigger compulsive patterns of excessive intake. Individual differences are important in addiction models, as only a subset of those who use addictive substances exhibit the individual risk factors that enhance susceptibility for becoming addicted (e.g., reward dysfunction, emotion dysregulation, impulsivity) [35]. Although the FA concept first emerged in the scientific literature in the 1950s, there has been a marked increase in this area since the early 2000s [36]. At this time, animal models demonstrated that UPFs (and UPF ingredients like sugar and fat) could cause biological (e.g., dopamine down-regulation) and behavioral (e.g., use despite consequences, withdrawal) responses akin to addictive substances [37^-43]. Neuroimaging technology demonstrated that obesity was associated with neural differences in the reward system (e.g., lower D2 receptor availability, heighten striatal activation to UPF food cues) seen in addictive disorders [44]. This spurred ongoing scientific investigation into whether UPFs directly motivate overeating behavior, akin to how addictive substances perpetuate problematic use, and trigger the FA phenotype in at-risk individuals [13].

Operationalizing UPF Addiction

The empirical literature to date is consistent with conceptualizing FA as a substance-based addiction to UPFs [13]. FA is most commonly operationalized using the Yale Food Addiction Scale (YFAS), a self-report questionnaire that asks individuals the extent to which they experience problematic patterns of UPF consumption that align with the DSM diagnostic criteria for substance-use disorders (SUDs) [45^, 46]. The current version of the YFAS aligns with the 11 DSM-Fifth Edition (DSM-5) symptoms of SUDs [45^, 47]. Examples of YFAS symptoms include continued UPF intake despite negative physical/psychosocial consequences, persistent yet unsuccessful attempts to cut down on UPFs, withdrawal, and tolerance (see Table 1 for a list of all 11 criteria). Consistent with DSM-5 diagnostic criteria for SUDs [47], individuals can meet a “diagnostic” score for FA on the current versions of the YFAS by endorsing at least two of the 11 behavioral indicators of addiction plus clinical impairment/distress [45^, 48]. Further paralleling DSM-5 criteria, severity thresholds for this “diagnostic” score are defined by the number of symptoms endorsed (mild: 2–3 symptoms; moderate: 4–5 symptoms; severe: 6–11 symptoms) [45^, 47]. The YFAS measures have been adapted and validated for the assessment of developmentally appropriate indicators of FA in youths [49^, 50] and in dozens of different languages (e.g., Japanese, Spanish, Farsi) [51]. While not intended to be a diagnostic tool, the YFAS measures have demonstrated excellent clinimetric properties, evidenced by incremental associations between FA symptoms and diagnostic score severity with clinically significant outcomes, including poorer quality of life, more severe psychopathology, and worse treatment outcomes [52^-54]. The YFAS measures also exhibit strong clinimetric sensitivity, given their ability to differentiate between individuals with compulsive patterns of UPF intake versus healthy controls and to detect changes in response to interventions (e.g., decreases in FA following behavioral weight loss treatment and bariatric surgery) [51].

Open multimedia modal

Using the YFAS “diagnostic” score, recent meta-analyses have estimated the prevalence of FA in general population samples to be 12–15% for children and adolescents and 14–20% for adults [55^, 56]. The prevalence of FA is thus similar to rates of SUDs with legal addictive substances (alcohol-use disorder: 19.2%; tobacco-use disorder: 18.2%) [57^, 58]. Biological and behavioral evidence has convincingly highlighted parallels in the mechanisms implicated in both FA and SUDs, such as reward dysfunction, emotion dysregulation, impulsivity, elevated cravings, and psychological comorbidities (e.g., depression, anxiety, childhood trauma) [59^-63]. FA is more likely to occur in individuals with a family history of addiction and has a high co-occurrence rate with other SUDs [64], which supports its conceptualization as an addictive disorder.

The Association between FA and Obesity

Early support for the plausibility of FA in humans resulted from neuroimaging studies showing overlapping responses in reward-related neural regions among individuals with a SUD for relevant drug cues and persons with obesity for UPF cues [44^, 65]. Yet, obesity is a heterogenous condition that may be caused by a caloric imbalance or an array of other factors (e.g., medication side effects) [66^, 67] and is thus an imprecise and insufficient proxy for FA. As such, operationalizing FA using behavioral criteria that align with DSM-5 diagnostic indicators of SUDs is a more theoretically informed approach for evaluating the relevance of FA within obesity.

Notably, only a subset (19–28%) of individuals with obesity meet for FA on the YFAS measures [55^, 56], suggesting that FA may represent a subtype of obesity characterized by greater susceptibility to the reinforcing natures of UPFs. In support, in studies of obesity, persons with versus without FA report elevated intake of UPFs and increased tendencies related to enhanced addiction vulnerability (e.g., impulsivity, emotion dysregulation) [68]. Addictive-like UPF consumption may at least partially explain observations that individuals with obesity and FA, compared to those with only obesity, endorse more frequent and severe maladaptive eating behaviors (e.g., emotional eating, uncontrolled eating) and an array of comorbidities, including poorer cardiometabolic indicators (e.g., higher visceral fat, poorer glycemic control), an increased prevalence and severity of comorbid physical (e.g., hypercholesterolemia, neuropathy) and psychological (e.g., depression, trauma) disorders, and higher psychological distress [52^, 69^-74].

It is important to note that FA is prevalent across weight classes (e.g., 12–17% of youths and adults with normal weight) and exhibits similar associations, regardless of BMI, with addiction risk factors (e.g., impulsivity), poorer quality of life, compulsive eating behaviors, and increased physical and psychological comorbidities [52^, 75^-80]. For all adults, FA may also increase the risk of weight gain over time [81^-83]. Strikingly, adults in the USA with versus without FA reported gaining, on average, nearly six times more weight during the first year of COVID-19 (12.42 vs. 2.14 pounds, respectively), and FA predicted weight gain above and beyond BMI [83]. No studies have yet evaluated the associations between FA and weight gain in youths, but findings in adults suggest the need for future research.

Implications of FA for the Treatment of Obesity

FA has been strongly related to clinical outcomes relevant to obesity like dietary adherence, treatment participation, and weight loss. Individuals with versus without FA report consuming significantly greater quantities of UPFs [32^, 84^-87] and are more likely to exhibit compulsive patterns of consumption with these foods [11], which likely interferes with adherence to the sustained caloric restriction required for weight loss. FA has been associated with poorer outcomes in behavioral weight loss treatments for both adults and youths, including lower attendance, higher attrition, lower weight loss, and higher weight regain [88^, 89]. Particularly strong evidence comes from a randomized clinical trial of 609 adults with overweight or obesity enrolled in a 12-month behavioral weight loss program, which found that FA was the strongest psychosocial predictor of treatment dropout and poorer weight loss [53]. In two studies of adults who had bariatric surgery, FA was associated with less weight loss and greater weight regain (mean follow-up durations were 7.7 and 13.7 years) [90^, 91]. However, some studies, particularly those with smaller sample sizes and/or briefer durations, have not found an association between FA and weight loss (e.g., [92]).

While existing data indicate the potential need to screen for FA at the onset of weight control interventions, it is important to acknowledge that there are no existing evidence-based treatments for FA. Many gold-standard behavioral weight loss treatments focus primarily on calorie intake and the approach of consuming all foods in moderation. From an addiction standpoint, intermittent substance use can enhance reinforcement and trigger problematic patterns of subsequent use [93]. A crucial next step is to adapt and evaluate whether empirically supported treatments for reducing the use of addictive substances (e.g., alcohol) may be similarly efficacious for reducing UPF intake among persons with FA. Behavioral and pharmacological treatments for FA informed by the addiction literature would ideally target the roles of mechanisms that perpetuate UPF reinforcement and addictive-like eating behavior, such as heightened UPF reward, UPF cravings, and impulsivity, as well as core features unique to addictive disorders like withdrawal and tolerance [94]. These novel treatments may have the potentials to improve behavioral weight loss intervention outcomes among those with both FA and obesity.

The Association between FA and Eating Disorders

A common critique of why FA does not warrant consideration as a distinct clinical presentation stems from the elevated prevalence rates of FA observed among individuals with clinical eating disorders. A recent meta-analysis estimated the prevalence of FA among youths and adults as 44% for persons with AN, 48% in those with bulimia nervosa (BN), and 55% among individuals with binge eating disorder (BED) [55], though individual studies have reported higher rates (e.g., 83.6% FA prevalence across eating disorder diagnoses [95]). While persons with eating disorders may be more likely to exhibit FA, this co-occurrence represents just a fraction of the estimated FA rates in the general population.

Lifetime prevalence rates of eating disorders have been estimated at 0.6% for AN, 1.0% for BN, and 2.8% for BED [96]. Given that 12–20% of individuals in general population samples endorse FA [55^, 56], FA appears to capture a clinical phenotype that extends far beyond existing eating disorder diagnoses, with a similar scope as rates of alcohol- (19.2%) and tobacco-use disorders (18.2%) [57^, 58]. In the absence of eating disorders, individuals with FA report clinically significant impairment and distress and elevated depressive symptoms, impulsivity, and negative affect [97^, 98]. Thus, it is necessary to maintain the perspective that individuals with FA and a clinical eating disorder represent a small proportion of those who exhibit FA in the general population. Nevertheless, a growing body of evidence suggests that FA has clinical relevance in the context of existing eating disorders.

Implications of FA for the Treatment of Binge-Type Eating Disorders

The common features between BN/BED and SUDs translate to the numerous similarities in evidence-based treatments for these disorders, such as restructuring automatic negative thoughts that promote cravings, developing alternative emotional coping strategies, and identifying emotional and situational triggers for problematic patterns of intake [106]. There are several key distinctions between SUDs and BN/BED that speak to the potential novelty and clinical utility of FA. Perspectives of binge-type eating disorders do not address core symptoms unique to SUDs, like withdrawal or tolerance, although evidence is growing that these mechanisms may be contributing to eating pathology [107]. Phenotypically, while BN/BED are diagnosed by the presence of distinct, 2-h binge eating episodes, addictive patterns of substance use can present as binge consumption (e.g., exceeding a threshold of alcoholic beverages on one occasion) or grazing (e.g., consuming alcohol throughout the day). Thus, addressing compulsive patterns of food intake that occur outside of binges is an important area of clinical focus.

Treatments for BN/BED do not consider the possibility that the reinforcing properties of UPFs may directly motivate and maintain binge eating episodes akin to an addictive substance. If the addictive potential of UPFs is contributing to binge eating, this does not mean that individuals need to completely abstain from all UPFs (which may have unintended consequences of increasing unhealthy patterns of dietary restraint for some individuals). Harm reduction approaches are empirically supported treatments for addictive disorders that recognize the addictive potential of the substance and implore strategies that help individuals engage in moderate, less harmful patterns of intake [108]. A similar approach could be adapted for FA that considers both the variability in the addictive profile of specific UPFs and situational factors that moderate risk. There is also emerging evidence that existing pharmacological (e.g., naltrexone and bupropion) and psychosocial interventions designed to address addictive mechanisms are beneficial in treating BN/BED [109^, 110], but larger randomized control trials are needed. The inclusion of other intervention targets associated with disordered eating, such as challenging thin body ideals and shape/weight concerns, would not be contraindicated in addiction-focused treatments and could continue to be important aspects of treatment for individuals with BN/BED and FA. In sum, it will be important to use an individualized and flexible approach to integrate addiction tenants into interventions for disordered eating.

Implication for Diagnostic Conceptualizations and Public Health

The emerging evidence that UPFs may be addictive and that a substantial percentage of the population exhibits clinically significant FA requires us to reconsider the adequacy of our current diagnostic conceptualizations. Although the inclusion of a FA phenotype has been previously discussed surrounding the release of the DSM-5 [111], it was rejected in part due to the idea that compulsive overeating was already accounted for by BED. Since that time, the science on FA has grown substantially (see [13] for a review) and has elucidated the clinical utility of FA as a unique behavioral phenotype and as an indicator of more severe presentations of obesity and eating disorders. Even if we assume that FA is appropriately captured by existent eating disorders (which appears unlikely), the differences in prevalence between eating disorders (0.6–2.8%) [96] and FA (12–20%) [55^, 56] suggest that scores of adults and children who report clinical impairment/distress are undiagnosed and untreated using the current approach. While eating disorders predominantly occur in females, FA occurs at a similar rate across sexes [55]. Thus, males with FA may be particularly likely to be missed based on the current diagnostic conceptualizations. The inclusion of FA as a provisional SUD diagnosis in the DSM-5 and/or the ICD would expedite the scientific evaluation of the validity of FA and the development of novel treatments that adapt evidence-based addiction interventions to help those with FA reduce UPF intake.

There are also important public health lessons to be learned from the field of addiction (see [112^-116]). Addiction epidemics often occur because a novel and potent addictive substance is created, but its addictive potential is undetected or underestimated. The environment changes in a manner that makes the addictive substance more accessible and, in the case of legal substances, heavily marketed and socially acceptable. Given that the individual risk profile for addiction (e.g., genetic profile, trauma exposure) does not quickly change on a population level, the increased prevalence rates of SUDs are primarily attributable to the addictive potency of the substance and accessibility within the surrounding environment. While the addictive substance is most problematic to those who develop a SUD, broader public health risks exist due to the potential for widespread subclinical overconsumption that results in negative health consequences [117]. Addictive substances (particularly legal ones) are highly profitable and the industries (or individuals) that produce and sell these addictive substances benefit.

Educational efforts and treatment advances alone have not been successful at turning back the tide of addiction epidemics. Instead, policy initiatives that alter the environment (e.g., increased taxation, limitations on product availability) and/or reformulation of the product (e.g., removing menthol from cigarettes, developing non-opioid pain medication) have been necessary to reduce the morbidity and mortality associated with addiction epidemics [112^-116]. Of key importance have been efforts to protect children and adolescents (e.g., restricting marketing to youths, age limits on purchase of addictive substances), as prevention is more effective than treatment in reducing harms associated with addictive substances [21]. The ability to institute these effective changes requires addressing opposition from industries that are rich and politically powerful. However, change is possible. The epidemics and public health responses associated with industrial tobacco products and opioid prescription medications are recent examples of meaningful public health wins that reduce the negative impact of addictive substances and save millions of lives [112^-116].

Conclusion

We propose that the rapid dominance of UPFs in our food supply beginning in the 1980s has resulted in another addiction epidemic, wherein FA and subclinical patterns of UPF overconsumption have contributed to increases in binge-type eating disorders, obesity, and diet-related diseases in the past 40 years. Notably, the clinical correlates of FA both within the contexts of obesity and binge-typical eating disorders and as a distinct presentation underscore the need to develop and evaluate evidence-based treatments for FA. Lastly, the lessons learned from past addiction epidemics are relevant for the negative public health consequences of FA and UPFs, particularly the need for policies that reduce the accessibility of UPFs in the modern food environment, reformulate UPFs to reduce their addictive potentials, and minimize risks for children and adolescents to develop FA.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

No funding was received.

Author Contributions

Erica M. LaFata and Ashley N. Gearhardt were each involved in developing, drafting, and finalizing the manuscript.