*Review* **Food Addiction and Psychosocial Adversity: Biological Embedding, Contextual Factors, and Public Health Implications**

#### **David A. Wiss 1, Nicole Avena 2,3 and Mark Gold 4,\***


Received: 18 September 2020; Accepted: 13 November 2020; Published: 16 November 2020

**Abstract:** The role of stress, trauma, and adversity particularly early in life has been identified as a contributing factor in both drug and food addictions. While links between traumatic stress and substance use disorders are well documented, the pathways to food addiction and obesity are less established. This review focuses on psychosocial and neurobiological factors that may increase risk for addiction-like behaviors and ultimately increase BMI over the lifespan. Early childhood and adolescent adversity can induce long-lasting alterations in the glucocorticoid and dopamine systems that lead to increased addiction vulnerability later in life. Allostatic load, the hypothalamic-pituitary-adrenal axis, and emerging data on epigenetics in the context of biological embedding are highlighted. A conceptual model for food addiction is proposed, which integrates data on the biological embedding of adversity as well as upstream psychological, social, and environmental factors. Dietary restraint as a feature of disordered eating is discussed as an important contextual factor related to food addiction. Discussion of various public health and policy considerations are based on the concept that improved knowledge of biopsychosocial mechanisms contributing to food addiction may decrease stigma associated with obesity and disordered eating behavior.

**Keywords:** food addiction; eating disorder; obesity; stress; trauma; early life adversity; adverse childhood experience; dopamine; epigenetics; biopsychosocial

#### **1. Background**

The quest to discover the precise mechanisms of hedonic overeating began decades ago. While many theories have been proposed, none have been widely accepted, and the obesity epidemic continues to grow. The *Nutrition Transition* theory describes a global trend toward consumption of processed foods that are low in fiber and high in added sugars and fats [1]. The changing global food landscape in the past four decades have increased access to convenient "snack" foods and decreased time spent preparing foods at home [2]. Several lines of research have explored the idea that highly palatable foods can alter brain reward pathways. For example, a landmark study showed that dopamine (DA) receptors were significantly lower in individuals with obesity [3]. Soon after, investigators documented overlapping neuroimaging characteristics in humans with obesity and those with substance use disorders (SUDs), showing reductions in DA-D2 receptors [4]. It was then suggested that individuals may overeat to compensate for DA-D2 receptor dysfunction [5]. To date, it is not clear whether these neurochemical associations are a cause of addiction-like overeating or a

consequence [6]. However, similar to other addictions, changes that occur in obesity show that food reinforcement adapts, strongly implicating biological underpinnings. Given the limited success in reversing the obesity trends, a better understanding of the various biopsychosocial mechanisms may help inform public health efforts.

Bart Hoebel pioneered the concept of food addiction (FA) research using animal models, showing evidence of bingeing, withdrawal, craving, and concomitant changes in dopaminergic and opioidergic systems in response to overeating sugar [7–14]. In rodent studies, early life adversity (ELA) has been shown to induce alterations in DA neuronal activity and synaptic function [15], impacting reward-directed behavior and partially accounting for individual variation along the mesolimbic DA projection [16]. More recently it has been shown that chronic stress dysregulates the reward system, promotes addiction-like eating, and contributes to the development of obesity [17]. Furthermore, palatable diets buffer against the negative impact of social stressors in juvenile rats [18]. Interestingly, environmental enrichment (larger space with conspecifics and novel objects) reduced sugar seeking and consumption [19]. Other rodent studies documented early and persistent alterations in amygdala circuitry and function following exposure to ELA, which were not diminished when the stressor was removed [20]. This suggests that ELA is not always redeemable by subsequent experience. At present, there is a gap in our understanding of how various forms of stress, trauma, and adversity link to addiction-like eating in real-world settings, particularly when viewed in social context, as well as over the lifespan.

In humans, various forms of ELA are associated with illicit drug use later in life [21–23]. In addition, there are established links between ELA and obesity [21,24], however, the exact mechanisms are not understood. A recent systematic review on childhood obesity implicated stress as a midstream factor that can lead to "junk food" self-medication and subtle addiction, in order to alleviate uncomfortable emotional states [25]. In a nationally representative sample of young adults (n = 10,813) exposure to multiple types of child maltreatment predicted excessive sugary beverage consumption [26]. In a Brazilian sample (n = 7639) FA was independently associated with early life physical and sexual abuse [27]. A positron emission tomography (PET) study also found that long-term exposure to adversity is associated with reduced striatal DA synthesis capacity [28]. Functional magnetic resonance imaging (fMRI) studies have linked ELA to blunted subjective responses to reward-predicting cues [29], and to altered connectivity in the extended reward network, leading to increased vulnerability to FA and obesity later in life [30]. While there are sufficient data that describes life course associations between ELA and adult weight outcomes [24], the actual biological mechanisms are less understood, which is a primary focus of this review. Another aim is to integrate psychologically relevant contextual factors such as weight stigma and pathological dieting.

The purpose of this review is to focus on literature from FA as well as obesity in the context of exposure to trauma, stress, and adversity, in an effort to answer three questions: (1) is FA a biologically plausible explanation for a life course association between ELA and obesity? (2) how might other relevant psychological, social, and environmental factors contribute to FA and to obesity? (see Figure 1) and finally, (3) what does it mean for public health? For simplicity, we have conceptually merged stress/trauma/adversity (STA) at several points throughout, particularly when reviewed outside of the context of early life, however we acknowledge these are not identical concepts. We also acknowledge that FA does not always lead to obesity, and that obesity can occur in the absence of FA. Additionally, ELA is used synonymously with adverse childhood experience (ACE) to describe exposures in the first 18 years of life. This review draws from literature across multiple disciplines in order to consider both individual and population health perspectives, and to describe contextual factors related to the neurobiology of FA. It is important to translate obesity science into a relevant social context, in order to identify achievable intervention targets which may have a meaningful impact upstream.

**Figure 1.** Food addiction and obesity following exposure to stress, trauma, and adversity: A biopsychosocial perspective of contextual factors.

#### *1.1. The Biopsychosocial Model & Other Foundational Theories*

Social and biological processes overlap and are inextricably linked. However, research methods are often incapable of capturing all features of an observed phenomenon, such as the various drivers of obesity (see Figure 1). Another example is how addiction neuroscience overlooks key social factors such as exclusion and marginalization which would make these findings more clinically relevant [31]. Biopsychosocial models were originally proposed by Engel as a new way to understand health and disease, by considering influence from various domains [32]. Biopsychosocial obesity research has found that lower educational attainment is associated with higher BMI, after adjusting for biological (energy intake and expenditure), psychological (decisional balance) and social (support) factors [33]. A biopsychosocial approach to childhood obesity should consider the (1) biology of the child (2) family environment and immediate psychosocial influences and (3) wider environmental, social, and cultural influences [34]. This creates opportunity for collaboration across multiple academic and clinically-focused disciplines. The current review employs a biopsychosocial perspective on FA, considering obesity as one possible outcome. This manuscript also incorporates Krieger's *Ecosocial Theory* which emphasizes the social production of disease over biomedical individualism, describing "embodiment" as the biological incorporation of social and ecological circumstances into everyday life [35]. A *Life Course Perspective* is used to link ELA to adult health [36–38]. Finally, a *Developmental Psychology* perspective views human development as relational, pertaining to dynamics (e.g., community features) which require individuals to be contextually situated into multidirectional and reciprocating ecological systems [39–41].

#### *1.2. Food Addiction & Eating Disorders*

With the validation of the *Yale Food Addiction Scale* (YFAS) in 2009 [42] and the updated YFAS 2.0 in 2016 [43], FA in humans has been operationalized across hundreds of studies. At the present time, FA has not been recognized as an official eating disorder (ED) in the Diagnostic and Statistical Manual (DSM) of Mental Disorders. Unique aspects of addictions include the importance of the substance, withdrawal, and tolerance, whereas unique aspects of EDs include restraint/rules, and shape/weight concerns [44]. It is well-established that dietary restraint/restriction can lead to rebound bingeing [45] yet it remains unclear if this is a cause or consequence of FA symptoms (discussed in Section 5.1). Thus, disordered eating characterized by dietary restraint provides important context for FA data. It has been recently suggested that the presence of dieting behavior must be carefully evaluated in order to separate the FA "signal" from the "noise" [46]. For example, ED research has identified significant overlap between FA and bulimia nervosa (BN), with FA symptoms improving when BN symptoms remit [47]. FA prevalence is the highest in BN [48] compared with other EDs, suggesting that FA treatment models should consider symptom contribution from dietary restraint and other compensatory behaviors. It has been proposed that FA is a transdiagnostic disorder associated with neurobiological vulnerability in certain people, who are more susceptible to using food as a coping mechanism [49]. It has also been shown that FA predicts a worse treatment outcome in patients with binge eating disorder (BED) [50].

Among those with an ED diagnosis, Brewerton (2017) proposed that the presence of FA be conceptualized as a meaningful correlate of post-traumatic stress disorder (PTSD) severity and symptoms [51]. For example, data from the Nurses' Health Study II has shown that severe physical and sexual abuse are associated with a 90% increase in FA risk [52]. The same dataset also suggested that symptoms of PTSD are associated with an increased prevalence of FA [53]. In a sample of 301 overweight and obese women, the association between FA and childhood trauma remained significant after adjusting for potential confounders such as socioeconomic status (SES) [54]. In a sample of bariatric surgery seeking patients (n = 1586), elevated ACE scores correlated with an increased likelihood of screening positive for FA [55]. A recent meta-analysis showed that multiple ACEs increased the odds of adult obesity by 46% (95% CI: 28–64%) [24] but several unmeasured confounders likely influence this estimate, such as the presence of EDs and SUDs, which frequently cluster, co-occur, and lead to weight fluctuations [56,57]. Therefore, risk estimates between childhood adversity and adult obesity would likely be higher after adjusting for these diagnoses often associated with dietary restriction and weight control, however this has not been formally tested. While data linking ELA to EDs are robust, only recently has it been shown that FA symptoms can mediate this pathway, as well as exacerbate ED symptoms significantly across all forms of childhood maltreatment [58]. Although EDs are not directly featured in Figure 1, the constructs of dietary restraint and weight stigma are used to contextualize important associations between EDs and FA. Notwithstanding, there are likely paths from ELA to obesity that are better captured by more classic ED pathology (e.g., BED) rather than FA, which are not directly featured by the model.

#### **2. Food Addiction Neuroscience & Social Context**

A frequent criticism of FA data in clinical settings is that the measure itself does not account for restrained eating [46] (discussed in Section 5.1). Another criticism is that it remains unclear how to intervene once FA has been detected. A recent systematic review of mostly pilot and feasibility studies concluded that currently there are no empirically supported psychosocial interventions for FA [59]. The authors recommend that clinicians assess for comorbid ED, and if present, provide evidence-based treatments (e.g., cognitive behavioral therapy) for those conditions. There is growing support for the FA construct in studies using the YFAS as well as neuroimaging studies on obesity that do not use "food addiction" terminology, with several examples provided below. It is worth noting that many authors reject the FA term in favor of other language such as eating addiction [60], or with additional qualifiers such as refined or processed food addiction [61,62], and even food use disorder [63]. Figure 1 proposes

that FA is one driver of obesity, although there are several others, including some not captured by the model (discussed further in Section 2.1).

Neurobiological overlaps between obesity and addiction have been described within the mesolimbic pathway between the ventral tegmental area (VTA) and the ventral striatum, with further projection to limbic (amygdala and hippocampus) and cortical regions (prefrontal cortex [PFC] and cingulate gyrus) [64]. Recent data suggests that among 110 healthy lean adults, exposure to a Western-style diet for one week led to rapid declines in hippocampal-dependent learning and memory, as well as appetitive control [65]. Research has shown that obesity (similar to SUD) is associated with deficits in executive functioning, an umbrella term encompassing the higher-order cognitive processes that help people take goal-directed action [66]. In a sample of women with obesity (n = 36), FA severity has been associated with impaired decision-making, compared to controls [67]. Resting-state fMRI data has shown decreased functional connectivity in the frontal gyrus in adults with obesity (n = 20) compared to controls [68]. A large cross-sectional study of children ages 9–11 (n = 2700) showed that increased BMI is associated with a reduced mean cortical thickness as well as lower executive functioning [69]. A follow-up report from the same study (n = 3190) suggested that BMI is associated with PFC development as well as diminished working memory [70]. Interestingly, a nationally representative sample of US adults (n = 4769 mean age 29) found that obesity is associated with poor working memory in women, but not men [71]. While tempting to consider that biological sex differences explain these findings, social context would suggest that the experience of weight stigma (discussed in Section 5.2), which is higher in women than men [72] may be a contributing factor. Recent data on school-age children (n = 176) suggests that weight-related stereotype threat (fear of confirming a negative stereotype) may explain working memory deficits more so than excess body weight [73].

Among patients with obesity (n = 224), FA is more closely correlated with psychological factors (depressive symptoms, quality of life) than with metabolic parameters (BMI, fat percentage, waist circumference) [74]. In a small sample of adult community members (n = 52), individuals with FA had significantly higher scores on depressive symptoms, emotion dysregulation, emotional eating, demand characteristics, motives, impulsivity, and family history of mental health problems and addiction [75]. Impulsivity can be defined as decision-making with limited forethought (rash-spontaneous behavior), having strong associations with FA [76–79]. Impulsivity hinders inhibitory control and is associated with increased intake of food [80] and drugs [81], often heightened in response to novel stimuli [82]. Delay discounting (preference for "smaller sooner" rather than "larger later" rewards) is closely associated with impulsivity and has been correlated with YFAS scores [83]. These authors believe it to be a predisposing factor rather than a consequence, although bidirectionality is likely. It has been suggested that impulsivity-related domains such as lower self-control, higher reward sensitivity, and negative affect help explain some similarities between addiction and obesity [84]. While impulsivity has heritable components linked to the mesocorticolimbic system [85] as well as serotonin-related candidate genes (e.g., HTR2A) [86], the potential for these traits to be influenced by epigenetic modification following psychosocial adversity will be explored in Section 3.1.

#### *2.1. Food Environment as a Driver of Food Addiction & Obesity*

The proposed pathways in Figure 1 suggest that FA may partially mediate the relationship between the food environment and obesity. Other pathways also exist. For example, census tract data have been used to show that wealthier neighborhoods (using median income) have better access (physical availability) to markets with healthier foods compared to poor neighborhoods [87]. Considering the importance of the built environment, neighborhood features (e.g., crime) that discourage outdoor physical activity are consistently linked to higher BMIs [88]. Recognizing that built, socioeconomic, and social characteristics co-occur [89], several investigators have advocated for a better understanding of theory-driven mediators and moderators in the relationship between neighborhood context and obesity [90,91]. Given the link between STA and FA, unsafe environments associated with lower SES neighborhoods are likely to impact BMI through increases in reward-based eating. It has been

established that diet quality tends to follow a SES gradient [92]. It has also been shown that parental fruit/vegetable consumption is linked to adolescent fruit/vegetable consumption [93], suggesting that the home food environment is important. Not eating dinner as a family has also been linked with increased BMI in kindergarten age children, regardless of SES [94]. It has been proposed that the maltreatment-obesity association is spurious, driven by confounding through the home food environment. However, after testing, researchers have found limited confounding influence [95] which supports arguments in favor of biological mechanisms.

Innovative methods that assess food environments include examining the (1) ratio of fast-food to full-service restaurants (2) ratio of bars/pubs to liquor stores and (3) presence of markets [96]. Multilevel models typically adjust for individual factors (education, hours of walking per week) as well as neighborhood factors (deprivation, walkability score). Perhaps the combination of food environment features matters more than individual components [96]. While several studies have described neighborhood "food swamps" (high density of high-calorie junk food) as predictive of obesity [97], few studies have looked at the potential roles of psychosocial pathways (mental health and wellbeing) [98]. A recent systematic review found that overall psychological resources (i.e., stress) had more consistent evidence of mediation than external neighborhood in the relationship between SES and BMI [91]. One study (n = 1112 adults) showed that paths from neighborhood characteristics to BMI could be partially explained by psychological distress and measures of inflammation [99] (discussed in Section 3). Taken together, FA is one potential pathway linking the food environment to obesity, however there is a "backdoor path" [100] through SES to STA, as well as a pathway that may not include FA (i.e., through the built environment). Comprehensive biopsychosocial frameworks cannot be tested or explained by any single study. There is conceptual support for the theory that the external environment (quick, cheap, highly palatable foods) is an upstream driver of EDs however this not been thoroughly investigated (discussed in Section 6.1). Figure 1 represents a synthesis of literature reviewed so far, as well as a roadmap for subsequent sections.

#### *2.2. Socioeconomic Status*

Given the inverse relationship between SES and BMI [101,102], obesity can also be viewed as a social phenomenon. This negative relationship has been shown in numerous countries outside of the US (e.g., Netherlands, Turkey, Morocco, South Asia) [103]. Hot-spot analysis in the US shows that higher BMI clusters are more likely in socioeconomically disadvantaged minority neighborhoods [104]. Meanwhile, large datasets (n = 43,864) have shown that obesity risk is decreased when positive contextual factors (maternal mental health, school safety, and child resilience) are present [105]. A large population-based cohort from the UK (n = 18,733) found that home-based deprivation was more closely associated with changes in child BMI than school-based deprivation [106]. On the other hand, some authors believe that the root cause of ACEs are largely based in the community, originating from an accumulation of contextual risk factors beyond a child's control, including family history, failed attachment, safety/security, and neighborhood risks [107]. While models of addiction and obesity are incomplete without psychosocial context, current research methods cannot adequately contextualize risk. Hence, literature from multiple disciplines was used to synthesize our conceptual model, which could be expanded with additional constructs such as resilience.

To illustrate further, a recent study showed that individuals with higher ACE scores were more likely to report not finishing high school, unemployment, and living below the poverty level [108]. Sustained activation and loss of capacity to respond to chronic stress might lead to a higher risk of illness and disease among people in lower SES categories [109]. It has been recognized that the processes which mediate the relationship between ELA and adult obesity might differ between men and women [110–112]. For example, a recent systematic review found that perceived stress from structural racism and weight stigma among black women creates negative emotions which predict emotional eating [113]. These stressors may then increase metabolic disturbance. Obesity itself can be a stressful state due to high prevalence of weight stigma [114] (discussed in Section 5.2), as highlighted

by the feedback loop in Figure 1. Obesity may be driving changes in stress biology rather than stress biology driving obesity [115]. Next, we consider the impact of ELA (as well as STA more broadly) from a life course perspective, describing precise (as well as candidate) mechanisms by which social factors impact health, seen primarily through recent SUD and obesity research.

#### **3. Biological Embedding of Stress, Trauma, & Adversity**

The fetal and infant origins of adult disease was proposed by Barker in 1990 [116]. This focus on the biological basis of disease gave rise to concepts such as allostatic load (AL), defined as the "cost of chronic exposure to fluctuating or heightened neural or neuroendocrine response resulting from repeated or chronic environmental challenge" [117]. This can be operationalized using a range of biomarkers that indicate inflammation and long-term "weathering." A simpler definition of AL is the price of adaption that leads to disease states over time. It has been suggested that frequent activation of the stress response and the failure to shut off allostatic activity creates "wear and tear" [118]. A landmark study showed that higher AL scores were associated with poorer cognitive and physical functioning, increasing the risk of cardiovascular diseases independent of sociodemographic risk factors [119]. Higher levels of AL (indexed by measures of blood pressure, C-reactive protein, fibrinogen, cholesterol ratio, triglycerides, and cortisol) have been observed in higher weight individuals [120]. This research suggested that these cumulatively elevated biomarkers link to decreased inhibitory control, highlighting the potential for disordered eating to become very difficult to overcome, similar to drug addiction.

It has been suggested that inflammatory mediators act on cortico-amygdala (threat) and cortico-basal ganglia (reward) circuits in a manner which predisposes individuals to "self-medicating" behaviors such as drug use, smoking, and the excess consumption of highly palatable foods [121]. Such behaviors further propagate inflammation and create a self-sustaining feedback loop. The "neuroimmune network hypothesis" proposes that ELA amplifies the communication between the brain and the immune system, promoting low grade peripheral inflammation [122]. The "glucocorticoid cascade hypothesis" posits that stress hormones impair brain function which further increases cortisol levels [123]. Adolescents exposed to childhood adversity have larger pituitary gland volume, associated with lower cortisol awakening response [124]. These authors propose that attenuation of hypothalamic-pituitary-adrenal (HPA) axis function may derive from stress-induced chronic hyperactivation during childhood. Heightened susceptibility may be due to differences in corticotrophin-releasing hormone (CRH) within the HPA axis, responsible for the output of cortisol [125]. Individual differences in inflammatory reactivity might explain why people have differing susceptibility to the consequences of stress, which may include neuroinflammation from stress-induced pro-inflammatory cytokines [126]. A recent cohort study of nine- and ten-year-old children showed that pro-inflammatory diets (i.e., high in saturated fats) increase neuroinflammation in reward-related brain regions, which in turn lead to further unhealthy eating and obesity [127].

A meta-analysis of 1781 people documented significantly decreased hippocampal volumes following ELA, with weaker evidence of increased amygdala volumes [128]. Alterations in corticolimbic circuitry following exposure to trauma make adolescents (n = 64) less able to relax and more vulnerable to risky behavior [129]. Other observable neural changes following ELA include (1) structural variation in gray and white matter (2) functional variation in brain activity and functional connectivity and (3) altered neurotransmitter metabolism [130]. In line with what has been observed in rodent studies [131], these effects might not be restricted to one's own lifespan but may also be transmitted to offspring [132]. It is well established that paternal drug exposure has long-lasting consequences including altered drug sensitivity in subsequent generations [133] but only recently has intergenerational transmission of trauma consistent with epigenetic explanations been described [134]. In animal models, the effects of maternal care on developing DA pathways and reward-directed behavior may account for individual differences in the mesolimbic DA system [16]. In social context, addictions may promote compromised parenting increasing the possibility that suboptimal care may be provided to the next generation. There

is a timely need for longitudinal studies that capture the precise biological mechanisms that link ELA to addictions over time, as well as their consequences.

#### *3.1. Epigenetic Mechanisms of Biological Embedding*

Epigenetics is the study of how the environment regulates the genome, best described as changes in gene function without changes in gene sequence. DNA methylation is an enzymatically-catalyzed modification of DNA and is one plausible mechanism through which early life exposures (low SES, nutritional patterns) become biologically embedded [135]. Methylation changes are apparent even years after the exposure but can be reversible in some cases. Epigenetic processes may be a key mediator between social environments during childhood and disease risk throughout life. Both DNA methylation and demethylation mechanisms are likely recruited during early life unfavorable experiences [136]. Other forms of epigenetic modification include histone modification and noncoding ribonucleic acids [137].

A milestone study confirmed what had been previously shown in animal models: human parental care impacts epigenetic regulation of hippocampal glucocorticoid receptor gene (NR3C1) expression [138]. Such epigenetic marks that persist into adulthood may influence vulnerability for psychopathology through its impact on HPA axis function. However, it is difficult to determine if DNA methylation changes are the immediate results of ELA or a consequence of the phenotypes associated with such adversity [139]. Notwithstanding, NR3C1 has been linked to prenatal stress [140] and is the most studied gene to date related to abuse and neglect [141]. Other genes involved in HPA axis regulation such as corticotrophin-releasing factor (CRF) have been investigated [142]. It is not implausible that developmental programming of the HPA axis and subsequent regulation of the stress response might impact addiction susceptibility, thereby increasing intake of substances known to activate reward pathways, including highly palatable food.

Epigenetic control of the expression of opioid receptor genes (mu-, delta-, and kappa-) has been reviewed in the context of SUDs [143]. While methylation at the mu-opioid receptor (MOR) gene is most strongly associated with drug addiction [144] as well as incentive motivation for processed food [145,146], decreased methylation at the kappa-opioid receptor (KOR) in the anterior insula has been shown in child abuse [147] (discussed further in Section 4.1). In this study of postmortem brain structures, the investigators were unable to detect a change in MOR expression, suggesting different epigenetic signatures associated with addictions and ELA. It is worth noting that different drugs have impacts at different brain regions and many include histone modifications in the nucleus accumbens (NAc) [148]. Other potentially relevant epigenetic modifications include the serotonin transporters [149–151] and proopiomelanocortin (POMC) [152]. Additional research is needed to determine how various epigenetic modifications associate with various forms of disordered eating, including FA.

In a sample of 206 women with bulimic symptomatology, there was evidence of increased methylation of the DA-D2 gene promoter, compared to controls [153]. Taq1A polymorphisms at the D2 receptor has been well-studied and known to influence impulsive behavior [154]. Recent data shows that DNA methylation in obesity-related genes may relate to obesity risk in adolescents [155]. Increased obesity susceptibility genes (e.g., FTO) have been found in the insula and substantia nigra (brain regions involved in addiction and reward) [156]. It has been shown that the methylation status on DA signaling genes (SLC18A1 and SLC6A3) might underlie epigenetic mechanisms contributing to carbohydrate and calorie consumption, as well as fat deposition [157]. Recently authors have linked specific dietary components with the gut microbiome in an effort to determine epigenetic factors on offspring susceptibility to obesity [158]. Expression levels of candidate genes implicated in glucose and energy homeostasis (e.g., HDAC7 and IGF2BP2) could be epigenetically regulated by gut bacterial populations [159]. The link between epigenetic marks and gut microbes appear to be mediated by host-microbial metabolites acting as substrates and cofactors for key epigenetic enzymes in the

host [160]. More research linking epigenetics to the microbiome is timely and warranted, particularly in the context of dysfunctional eating behavior (including both under- and overeating).

#### **4. Stress & Obesity**

While a PTSD diagnosis is associated with an altered stress response, chronic stress can exist in the absence of PTSD, and has been the focus of several investigations related to eating behavior. Multiple pathways have been described which link stress to obesity, including (1) interference with cognitive processes (executive function, self-regulation) (2) behavior (eating, physical activity, sleep) (3) physiological changes (HPA axis, reward processing, gut microbiome) and (4) production of biochemical hormones and peptides (leptin, ghrelin, neuropeptide Y) [114]. At a basic level, stress may lead to food consumption in the absence of hunger. It is established that poor executive functioning is associated with consumption of palatable food, leading to inflammation and metabolic changes promoting weight gain [161–163]. Other pathways which have been identified include the autonomic nervous system (cardiovascular functioning), the epigenome (intergenerational transmission), and the metabolome (profile of metabolites in body) [164]. The vagus nerve (part of the autonomic nervous system) has been identified as an important physiological stress pathway linked to gut microbiota [165]. With rising interest in the gut-brain axis, novel pathways which include FA are being explored [166]. FA can be considered as a partial mediator in the stress-obesity pathway, likely resulting from one or many of the biologically embedded pathways described herein.

To illustrate further, individual differences in neural response to food cues under stress have been observed in human neuroimaging studies [167], lending support to differential susceptibility. It is well established that amygdala function is moderated by stress-induced glucocorticoid (GC) release [20], and a less efficient HPA axis negative feedback loop may represent a deficiency in emotion and stress regulation [168]. Highly palatable foods stimulate stress hormones that alter the limbic system (emotions) and striatal (motivational) pathways, promoting further food craving and excessive intake [169]. Rewarding foods upregulate CRF in the amygdala and related limbic striatal pathways. The most direct physiological pathway is dominated by cortisol, which stimulates fat storage and changes dietary behavior through increased reward sensitivity (DA and opioid systems) and increased appetite (arcuate nucleus in the hypothalamus) [165]. Future research should attempt to clarify the biological embedding of chronic stress both in the absence and presence of diagnosed PTSD, specifically impacting reward-related pathways associated with consumption behavior. Additionally, more research is needed on biopsychosocial factors of resilience in the context of both FA and obesity.

#### *4.1. Stress & Addictions*

Given the established links between stress and obesity, these links can be used to conceptualize relationships between stress and FA. The phenomenon of stress-induced reinstatement of drug-seeking is generalizable to other substances, including food [170]. To illustrate, the DA and GC systems are both highly involved in substance addictions, and ELA may induce long-lasting alterations in these systems. One of the most profound effects of stress is the activation of the HPA axis with release of CRF from the paraventricular nucleus. Human studies have shown stress exposure increases alcohol craving [171]. Both chronic stress and long-standing alcohol use promote PFC dysfunction [172]. Changes in CRF activity that result from chronic alcohol exposure within the extended amygdala network is thought to be key factor in withdrawal symptoms [173]. It has been proposed that repeated altered activity in the DA system and sustained activation of the CRF system leads to AL and negative emotional states [174]. The central thesis in Koob's allostatic view of stress and addiction is that stress leads to changes in brain CRF that have a direct impact on addiction [175]. Withdrawal can produce elevated levels of GCs and increase release of CRF in the central nucleus of the amygdala [175].

It has also been suggested that increased CRF alters serotonin release in the brain which facilitates DA in the accumbens [176]. Prolonged exposure to stress can lead to irregular changes in GC receptor density (epigenetics) which may increase the reinforcing effects of alcohol and drugs [177,178]. Interestingly, a higher salivary cortisol level in response to stress has been associated with higher drop-out rates in treatment [179]. It has also been suggested that variability in stress-related genes may contribute to the ability of certain individuals to remain abstinent from heroin, possibly due to higher stress resilience [180]. Importantly, not only does STA increase addiction behaviors, some authors have suggested this association also exists in the opposite direction [181,182]. With illicit drugs, their procurement and use can predispose individuals to traumatic stress [183,184]. In animal models, chronic opioid pretreatment is able to robustly augment associative fear learning [185]. These changes were not observed when opioids were given after the traumatic event, and potentiation lasted beyond discontinuation of drug exposure. This concept has been thoroughly described as part of the withdrawal process in widely accepted addiction models [186–188]. However, more research is needed to understand how long-term exposure to highly palatable foods may alter one's long-term response to stressful life experiences, and how this dynamic can play out in reciprocal and bidirectional ways, for example in the presence of weight stigma and dietary restraint (and cumulatively over time).

A recent review of preclinical data suggests three mechanisms by which DA and GCs interact: (1) GCs upregulate tyrosine hydroxylase (rate-limiting enzyme in DA synthesis) (2) GCs down-regulate monoamine oxidase (enzyme responsible for DA removal) and (3) GCs are hypothesized to decrease DA uptake subsequently increasing synaptic DA [189]. Clearly stress enhances substance abuse-related effects at multiple points along the mesolimbic projection. The KOR system plays an important role in behavioral stress responses and has been implicated in stress-induced maladaptive responses [190]. While MOR activation produces euphoria, KOR is generally aversive and may contribute to negative affect states in withdrawal. According to some authors, it is possible that a stress-induced increase in KOR function promotes drug seeking by reducing DA transmission [190]. Meanwhile, reduced MOR has been observed in comorbid binge eating disorder and obesity [191] and across SUDs [144] which strongly suggest neurochemical overlap in these conditions, and which can persist despite weight loss or periods of drug abstinence. Any change in stress neurobiology is likely to influence reward. Based on observed deficits in the ventral striatum, reward responsiveness and processing may be a primary mediator of the effects of ELA [192]. Taken together, FA is a biologically plausible explanation for the life course association between ELA and obesity, however important contextual factors from the psychological domain deserve further consideration.

#### **5. Psychological Correlates of Food Addiction & Obesity**

Thus far we have highlighted several social and environmental factors associated with STA and addiction-like eating. We have reviewed emerging data on the biological embedding of adversity, which may increase an individual's susceptibility to FA, and potentially lead to obesity over time. Based on the overlap between FA and EDs as well as SUDs, we have recommended including these variables into statistical models which investigate weight outcomes. Finally, we have proposed a comprehensive conceptual framework to further contextualize these relationships by including two psychological (as well as socially constructed) correlates of FA, EDs, and obesity: dietary restraint and weight stigma.

#### *5.1. Dietary Restraint*

Restrained eating is generally defined as a cognitive effort to eat less in order to lose weight [193], which has been viewed both as the problem and solution to obesity [194]. More recently it has become clear that theories of weight loss based on low-calorie dieting are failing, likely due to neurochemical, endocrine, and gastrointestinal factors which are not adequately captured by simple models of energy balance. While the concept of dietary restraint has been linked to some positive outcomes (e.g., weight management, prevention efforts) [195], it is included in our model as a risk factor for eating pathology, often associated with EDs (sometimes referred to as restriction). A classic study conducted by Ancel Keys in the 1940s examined the link between starvation and changes in human biology and behavior [196]. The study showed that significant (intentional) weight loss produced the onset of binge

eating in 30% of participants (n = 36). Many of the individuals who were reduced to 50% of their baseline caloric intake for extended periods of time (months) began collecting recipes and cookbooks. The finding that caloric restriction leads to preoccupation with food has been widely cited in the ED literature. Meanwhile, it is less clear if deliberate efforts to eat differently (focusing on dietary quality rather than quantity) should be classified as pathological restraint. Extreme diets intended for health reasons which impair daily function have been described as "orthorexia nervosa" which appears to be growing problem [197]. Research linking FA recovery to orthorexia is timely and warranted.

A dieting intervention on 121 females which included monitoring and restricting showed that monitoring increases perceived stress, while restricting increases the cortisol output [198]. Dieting is stressful, which may explain why engaging in dieting behaviors aimed at losing weight can actually have the opposite effect. Future iterations of Figure 1 may include an arrow directly from dietary restraint to STA, whereas in the current model there is only a backdoor path through weight stigma. A twin study from Finland (n = 4129) showed that dieters are prone to future weight gain independent of genetic factors [199]. A recent fMRI study showed that "successful" restrained eaters had stronger activation in the middle frontal gyrus and cerebellum (associated with executive function and inhibition) suggesting that food temptations may trigger processes of positive inhibition in some, but not others [200]. It is likely that altered neurochemistry from SUD and/or ELA/STA will impact the degree of success with dietary restraint. More research is needed on the impact of trauma on various eating behaviors, including restriction. It will prove important to better define dietary restraint in the context of FA recovery.

There is considerable debate on how to approach FA from a nutritional standpoint, including incorporating FA data into the traditional ED landscape [201]. Meule has stated that "dietary restraint does not have to be dysfunctional as long as flexible elements are added" [202]. Based on available data linking FA and EDs (Section 1.2), it is proposed that restrained eating moderates the link between food environment and FA, as well as the link between FA and obesity (Figure 1). In other words, individuals engaging in dietary restraint are predicted to display higher levels of FA severity. Future research should examine the directionality as well as cumulative interplay of this relationship. Furthermore, it is proposed that individuals who meet criteria for FA and engage in dietary restraint may experience different effects on their weight status, depending on whether or not the restraint is successful, unsuccessful, pathological, or part of a restrictive ED. These theories need to be tested in both observational and experimental studies in an effort to better develop the emerging field of behavioral health nutrition. Recently, an 8-step process has been proposed to help clinicians discern FA from dietary restraint in order to inform inclusive vs. exclusive nutrition strategies [46]. The key discerning factors include the presence of SUD, PTSD, and ELA, which, if all present, can provide more confidence in the strength of an FA signal, particularly in the absence of dieting behaviors.

#### *5.2. Weight Stigma*

Weight stigma has been described as a "vicious cycle" where weight stigma begets weight gain [203]. Similar to dieting, the experience of stigmatization increases cortisol, which may drive food consumption by sensitizing the reward system [203]. In addition to increased cortisol, weight stigma also increases oxidative stress [204] providing further evidence of biological embedding. In a large sample of adolescents (n = 115,180), perceiving one's body as overweight increases risk of suicidality [205]. Conceptually, weight stigma is similar to other forms of STA, which we consider as midstream drivers of eating behavior and subsequent weight outcomes, both through biological and psychosocial pathways. While unproven, it is possible that body dissatisfaction and self-stigma drives avoidance behaviors (e.g., weight loss to avoid adiposity) which is similar to the avoidance experience in PTSD. This may be one reason why efforts to lose weight can be persistent (or even relentless) for so many, despite the fact that weight loss efforts have been unsuccessful (or unsustainable) in the past.

In a large national sample (n = 5129), weight discrimination was associated with overeating (specifically convenience foods) and less regular meal timing [206]. Individuals who are the target of weight stigma have been shown to decrease self-control and perceived capacity for weight management [207]. In a large sample of adolescents (n = 1497), FA and psychological distress mediated the association between weight-related self-stigma and binge eating [208]. It is worth acknowledging differences between externalized (others) and internalized (self) weight bias. It has also been shown that some people will experience longer term distress from weight stigma than others [209]. Perceptions and/or experiences of weight bias in primary care settings have been shown to negatively influence patient engagement with health care services [210]. In summary, weight stigma has emerged as an important component of obesity context, with strong arguments in favor of adopting weight-inclusive health policy [211]. While the FA explanation for weight control has been shown to decrease weight stigma among groups [212,213] it has been suggested that FA can increase internalized weight stigma among individuals [214]. More research is needed on the role of weight stigma driving dietary restraint, both as a cause and consequence of addiction-like eating.

#### **6. What Does It Mean for Public Health?**

Research on biological programming attempts to identify the most critical and sensitive periods that underlie the developmental origins of later childhood and adult disease [215]. It has been suggested that the timing of adversity explains more variability in DNA methylation than the accumulation or recency of exposure [216]. It has also been observed that different dimensions of adversity have distinct influences on neurodevelopment [217]. The exact mechanisms which link human DNA methylation with psychological disorders have not been elucidated [151]. What we do know is that the cumulative effects of STA can impact neural function with significant implications for substance-seeking behaviors. All addictions share a common neurobiology and have known relationships to STA in both directions. Sugar, salt, and fat added to foods make them more palatable and reinforce "drug-like" behavior with loss of control, continued use despite consequences, binge episodes, and other similarities with traditional drugs of abuse. It is clear that obesity causes changes in opioid and DA signaling which alter reward processing [218]. Given the established links between ELA and the propensity for behavioral health disorders including SUD, ED, and FA, prevention efforts might have a meaningful impact upstream. Addressing clusters of disorders with shared underpinnings jointly may be more fruitful than a one-disorder-at-a-time approach [122].

A recent study of New Orleans children showed that neighborhood stress exerts a direct influence on obesity, after adjusting for diet and activity [219]. Such findings support the need to improve social conditions rather than efforts to address obesity at the individual level. It will be important to identify positive contextual factors such as neighborhood and school safety, as well as resilience [105] and develop community-based programs that promote these protective factors. Resiliency-building programs that reduce delay discounting may decrease addictive behaviors [220]. Meanwhile, socioeconomic differences in the quality of early life create "cumulative disadvantage" that contribute to gradients in health status [37]. SES indicators are upstream determinants of health while biological factors are more proximate determinants [221]. Neighborhood disadvantage creates social context which may become biologically embedded [222]. The impact of low SES can become embedded into inflammatory processes, the HPA axis, and neural function/structure, all of which are epigenetically controlled [135]. It is not unreasonable to assume that normalizing/improving HPA axis function may be beneficial in the treatment and relapse of addiction-related disorders. Given that epigenetic patterns are sculpted during early life [137], reducing stressors appears crucial to the long-term management of FA.

A "systems thinking" multilevel approach will be critical to reverse obesity trends. For example, trauma-informed treatment and stress management curriculum should be made available in underserved communities, starting with schools [107]. Mobilizing cross-sector interdisciplinary partnerships to connect ELA to later life health outcomes will be critical. It has been stated that "fostering increased societal awareness about toxic stress exposures that are often hidden, stigmatized, and attached to shame needs to occur across generations" [223]. Greater awareness of the biological mechanisms discussed herein are likely to reduce weight stigma, which is a known barrier for individual

help-seeking behaviors in those with obesity [224] as well as SUDs [225]. Feelings of rejection associated with weight stigma and disordered eating are additional stressors which may further perpetuate a negative cycle [114,168]. A recent study showed that the FA model explanation for obesity resulted in lower stigma than the traditional "diet and exercise" explanation that attributes obesity to personal responsibility [213]. Given that weight stigma is a psychosocial contributor to maladaptive eating behavior, interventions targeting stigma (at the individual and societal levels) are warranted.

#### *6.1. Food Policy*

This paper has reviewed evidence to suggest that improving the early childhood environment might impact obesity risk and therefore should be a public health priority. Meanwhile, if reward-related neuroadaptations associated with addiction persist over time, addressing only the underlying factors may fail to create lasting changes in eating behavior, suggesting that policies targeting the food environment will also be important. Given that the food environment in the US promotes easy access to foods with addictive potential [226] it is not unreasonable to hypothesize that highly palatable foods leave a biological imprint which may perpetuate FA symptoms across the lifespan and into subsequent generations, as has been shown in animal models [131]. Western-style diets rapidly impair appetitive control, compared to those on their habitual diet [65]. Combined with heightened susceptibility to STA stemming from ELA, efforts to address the obesity epidemic may be futile without strategic multilevel interventions targeting corporate responsibility (i.e., "Big Food") [227].

There is mounting evidence of the harmful effects of processed foods in contemporary diets. A recent trial comparing the caloric intake of those on ultra-processed foods (containing minimal whole foods) compared to unprocessed/whole foods for two weeks found ad libitum intake was increased by approximately 500 kcal/day on the ultra-processed diet [228]. Not surprisingly, people gained weight on the ultra-processed diet and lost weight on the unprocessed. Cross-sectional data (NHANES 2005–2014) has shown that higher consumption of ultra-processed foods is associated with excess weight and is more pronounced in females [229]. A study from Spain showed that four or more servings per day of ultra-processed foods is associated with a 62% increased hazard for all-cause mortality, where each additional serving increased all-cause mortality by 18% [230]. It remains unclear if the negative health effects are due to the direct impact of the processed foods, or the displacement of nutrient-dense high-fiber foods protective against oxidative stress and associated inflammation.

Public health interventions to increase access to healthy foods in lower SES communities have been unsuccessful in reducing obesity, therefore new approaches are needed. Identifying certain foods to be addictive may encourage collective efforts to avoid them [231] and is associated with support for policies to curb their use [232] similar to how public health officials addressed Big Tobacco. Only recently have researchers and policy makers begun to explore targeting the food environment in universal ED prevention efforts [233]. It has been suggested that "processed food addiction is the result of an intentional epidemic of addiction not an incidental by-product of Western environments" [62]. The term "processed food addiction" implicates the food industry rather than the individual. There is a critical need for increased awareness of FA and the role played by multinational food corporations in promoting processed foods with addictive qualities [214]. Evidence suggests that aggressive marketing of these foods to children, adolescents, and young adults disproportionately affects vulnerable groups [234–238]. While it is highly unlikely that food companies will re-formulate their products based on self-regulation, it also unrealistic to expect food-addicted individuals to regularly avoid food-related temptations. Policy support should include warning labels, industry reductions on sugar, and product bans (e.g., energy drinks) [232] while legal tools include advertising restrictions and class-action litigation [239]. Several authors have recommended policies restricting fast food advertising to adolescents [240,241]. Based on growing evidence for FA, this may be indicated.

#### **7. Conclusions**

The biological underpinnings of addictions strongly imply a role for ELA in the development of FA and obesity. Importantly, ELA can alter the physiological response to various forms of psychosocial STA across multiple body systems, which can have a cumulative impact on health behaviors over the lifespan. FA research which began in animal models has since been described in human neuroimaging studies which capture neurobiological and behavioral overlap between FA and SUDs. A biopsychosocial perspective on FA considers biomarkers such as inflammatory markers and other measures of AL, the HPA axis including the output of cortisol, epigenetic mechanisms including those that influence the HPA axis, and various structural, functional, and morphological brain changes, following exposure to ELA and STA. In order to contextualize risk, a biopsychosocial model considers the upstream drivers and fundamental causes of health disparities, such as SES and environmental (e.g., neighborhood) factors that impact food access and food choices. Furthermore, obesity frameworks should incorporate weight stigma as an important cause and consequence of the epidemic, suggested herein as a form of STA that can also become biological embedded. Finally, the role of dietary restraint has been included as an important psychological factor that should be accounted for when conceptualizing FA and obesity, particularly given the strong relationship between ELA and EDs, as recently reviewed elsewhere [46].

Stress proliferates over the life course and across generations, widening health disparities between advantaged and disadvantaged groups [242]. This might explain why public health nutrition interventions in low SES communities have had limited success. Consumption of highly palatable foods to "self-medicate" the long-term biological impact of chronic stress may be a critical factor in understanding the obesity crisis. This is particularly true for marginalized groups with less access (e.g., affordability) to unprocessed foods. Higher SES groups are more likely to have success in reducing addiction-like eating compared to lower SES groups who are constrained by access and resources. Public health interventions should account for the growing inequities in health outcomes. Biopsychosocial approaches that consider the cumulative interplay between social and biological factors are helpful when conceptualizing multiple systems driving substance-related disorders, whether it be alcohol, drugs, nicotine, or food [243]. A biopsychosocial model may contribute to conceptual and methodological advances in our understanding and treatment of obesity. Meanwhile, separating constructs into biological, psychological, and social factors (as in Figure 1) can be contraindicated by ecological models that emphasize the dynamic reciprocity between these levels. However, our conceptual model has discerned between these factors in order to encourage further contextual analysis of FA.

Based on the biological plausibility of FA as a consequence of psychosocial STA, potential solutions to the obesity epidemic may include: (1) improve social conditions in order to reduce exposure to ELA, as well develop community-based programs for early intervention (2) decrease weight stigma based on FA data implying that body weight is not simply a "choice" (3) mind-body approaches (e.g., yoga, meditation) designed to improve the human stress response and (4) policy proposals aimed at the food industry to reduce exposure to highly palatable foods. More information is needed about the role of nutrition in the reversibility of unfavorable gene expression. More research is needed to investigate whether long-term dietary changes such as abstaining from highly palatable foods is even feasible. If so, will this improve the microbiome and stimulate/reverse epigenetic change and/or lead to altered reward pathways in the brain? At a minimum, it is reasonable to predict that reducing exposure to addiction-like eating can improve executive functioning. Since dietary restraint is a known risk factor for the development of EDs, drastic individual nutrition changes should be implemented in consultation with a qualified professional such as a registered dietitian nutritionist, particularly when there is underlying trauma and/or SUD. Treatment models should be trauma-informed and include staff trainings.

FA and SUD share multiple predisposing factors including ELA which can become biologically embedded. These findings may link social determinants to specific health outcomes and elucidate pathway effects of risk across the life course. Epigenetic processes may be a key mediator between social environments during childhood and disease risk in adulthood. Mediating mechanisms such as

AL, the HPA axis, DNA methylation, and altered reward sensitivity (i.e., dopamine systems) have scientific merit, however the fundamental causes of health inequalities present in society should not be overlooked. Low SES and neighborhood disadvantage remain important drivers of ELA, particularly within the context of the obesity epidemic. The cumulative effects of STA that impact neural function and heighten threat vigilance have significant implications for substance-seeking behaviors, including eating. The FA construct has gained credibility from animal and human studies reviewed herein, which may help reduce stigma associated with addiction-like behaviors, including obesity. More research is needed to understand the differential impact of inflammatory signaling markers on the brain, including assessment of blood brain barrier integrity. The study of neuroinflammation is likely to add explanatory power to our conceptual model and guide future research questions.

If the DSM accepts FA, it will lead to better treatment and eventually public health efforts to improve the national food environment and global nutrition landscape. More resources should be allocated for nutrition education during pregnancy and lactation, particularly in underserved communities where stress and adversity are high, and the food environment is suboptimal. Applying the FA framework has the potential to influence the way people view food, and to ultimately decrease addiction in future generations. FA treatment does not always require specific "food abstinence" but it does warrant reduced exposure and harm reduction strategies. Given the strong evidence that neurobiological responses to food differ among people, personalized precision nutrition interventions are warranted. In order for these strategies to be successful, cultural shifts around food norms will be necessary. Furthermore, FA is both an individual and collective health problem, and should be addressed at the societal level with broad policy interventions. We propose that unregulated promotion of addictive foods by the food industry are major contributors of obesity, particularly in the face of disadvantage and distress. Government interdiction may be required to reduce the epidemic of obesity and the growing problem of food addiction. Multidisciplinary efforts using trauma-informed integrated biopsychosocial frameworks will be necessary to reverse obesity trends.

**Author Contributions:** Conceptual model, literature review, and first draft preparation (D.A.W.). Draft revisions (N.A. & M.G.). All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Food Addiction and Tobacco Use Disorder: Common**

**Liability and Shared Mechanisms**

**Laurie Zawertailo 1,2,\*,**†**, Sophia Attwells 2,**†**, Wayne K. deRuiter 2, Thao Lan Le 2, Danielle Dawson <sup>2</sup> and Peter Selby 2,3,4,5**


Received: 30 October 2020; Accepted: 10 December 2020; Published: 15 December 2020

**Abstract:** As food addiction is being more commonly recognized within the scientific community, parallels can be drawn between it and other addictive substance use disorders, including tobacco use disorder. Given that both unhealthy diets and smoking are leading risk factors for disability and death, a greater understanding of how food addiction and tobacco use disorder overlap with one another is necessary. This narrative review aimed to highlight literature that investigated prevalence, biology, psychology, and treatment options of food addiction and tobacco use disorder. Published studies up to August 2020 and written in English were included. Using a biopsychosocial lens, each disorder was assessed together and separately, as there is emerging evidence that the two disorders can develop concurrently or sequentially within individuals. Commonalities include but are not limited to the dopaminergic neurocircuitry, gut microbiota, childhood adversity, and attachment insecurity. In addition, the authors conducted a feasibility study with the purpose of examining the association between food addiction symptoms and tobacco use disorder among individuals seeking tobacco use disorder treatment. To inform future treatment approaches, more research is necessary to identify and understand the overlap between the two disorders.

**Keywords:** food addiction; nicotine; tobacco use disorder; comorbidity

#### **1. Introduction**

Smoking and obesity are the two most prevalent causes of preventable chronic disease morbidity and mortality worldwide [1,2]. Smoking and food addiction are both maladaptive behaviors in which an individual experiences loss of control and compulsive engagement in the behavior despite known harmful consequences. We have reached a pivotal time for understanding food addiction, similar to a time when tobacco use disorder was perceived as habit forming and not addictive. While food addiction does meet several of the American Psychiatric Association's diagnostic criteria outlining substance use disorder [3], more research is necessary to determine if certain foods are addictive and how to prevent and treat this condition.

The concept of food addiction represents a relatively new domain of research [4]. Food addiction refers to an "eating behavior involving the overconsumption of specific foods in an addiction-like manner". While the term "food addiction" was introduced to scientific literature in 1956 [5], research investigating the mechanisms, neurobiology, and genetics of food addiction was not pursued until the early 2000s. Food addiction in relation to certain foods has not been widely accepted or studied. As such, the diagnostic criteria of food addiction are not well established and are not formally recognized by the American Psychiatric Association as either a substance use disorder or a behavioral disorder in the Diagnosis and Statistical Manual of Mental Disorders 5 (DSM 5) [3]. Regardless, the criteria for food addiction have been modeled from substance use disorder criteria outlined in the DSM 5 [6].

In the DSM 5, substance use disorder is defined as a complex condition manifested by compulsive substance use despite harmful consequences. Furthermore, regular substance use may develop dependence and tolerance. The DSM 5 currently lists nine distinct disorders (e.g., tobacco use disorder and alcohol use disorder), but nearly all substances are diagnosed on the basis of the same overarching criteria [3]. Hallmark symptoms of food addiction include loss of control and frequent overconsumption, desire or repeated failed attempts to reduce or stop consumption, increased time spent in activities necessary to obtain and eat food, giving up on important activities such as physical exercise, continued consumption despite physical or psychological problems, and clinically significant impairment or distress [7].

Currently, limited research exists pertaining to food addiction and how it relates to other substance use disorders. Given that both unhealthy diets and smoking are among the leading risk factors for all-cause disability-adjusted life years, total deaths, and years lived with disability [8], a greater understanding of how food addiction and tobacco use disorder overlap with one another is necessary [9]. Furthermore, the role of food addiction in the common but understudied phenomenon of post-cessation weight gain among smokers trying to quit has not been researched. The purpose of this narrative review is to present and summarize the most up-to-date research findings on the prevalence, biology, psychology, and treatment of tobacco use disorder and food addiction as individual disorders, identifying overlap and commonalities that may help inform treatment approaches. It is important to understand each individual disorder in the context of the other since there is emerging evidence that the two disorders can develop concurrently or sequentially within individuals. To provide background on food addiction and tobacco use disorder, this review first discusses the prevalence of these disorders, measurements that assess severity, and theories of food addiction. This review then covers various subtopics such as biology, gut microbiome, psychosocial factors, and treatment options. The studies examined for this review mainly focus on food addiction rather than obesity and eating disorders, as these are two separate and complex conditions.

#### **2. Methods**

To locate relevant publications on the relationship between food addiction and tobacco use disorder, the following databases were searched: Ovid Medline, PsycInfo, and Embase. The Medline search strategy included both relevant medical subject headings (MESH) and keywords for the concepts of food addiction and tobacco use disorder/cessation. Food addiction-related terms included "food adj3 addict", "eat adj3 addict", "food adj3 dependence", "food use disorder", "compulsive eating", "Yale Food Addiction Scale", "obes", "diabetes", and "binge eating disorder". Smoking terms included "tobacco use disorder", "tobacco depend", "nicotine", "((nicotine or tobacco or smoking) adj3 (cessation or quit or quitting or quits or give up or giving up or stop or stopping or stopped or stops))", "((nicotine or tobacco or smoking) adj3 withdraw)", and "smoking cessation/or tobacco use cessation". The Medline search strategy was adapted for the controlled vocabulary of the other databases searched. All of the search results were limited to English language publications, but no date or study type limitations were applied. Furthermore, the database search was complemented by a manual review of reference lists from the retrieved articles. Articles related to the purpose of this narrative were retrieved from the database search. Studies examining eating disorders and obesity in the absence of food addiction were not reviewed as this was outside of the scope of this review.

#### **3. Discussion**

#### *3.1. Prevalence*

#### 3.1.1. Prevalence and Severity Measurement of Tobacco Use Disorder

In 2017, 15% of the Canadian population (about 4.6 million people) were current smokers. Within that population, 11% and 4% reported being daily smokers and occasional smokers, respectively [10]. Daily smokers averaged approximately 14 cigarettes per day, with a higher percentage of smokers being male (17%) than female (13%) [10]. Furthermore, in 2018, nearly 34% of Canadian youth reported trying an electronic cigarette (e-cigarette) at least once in their lifetime [11]. Evidence suggests that e-cigarette use in youth increases the risk of developing tobacco use disorder later in life [12]. Similar statistics are reported in the United States, where approximately 12% of the population (or 40 million people) are daily smokers [13].

The DSM 5 replaced the DSM-IV's categories of nicotine abuse and dependence with tobacco use disorder. Within this context, problematic patterns of tobacco use must cause significant impairment and distress for at least two symptoms within a 1 year period to be considered a disorder. Other common measures of nicotine dependence severity include the Fagerstrom Test for Nicotine Dependence (FTND), Heaviness of Smoking Index (HSI), and time to first cigarette (TTFC). The FTND contains six questions scored 0–3 for a total possible score of 0–10. Higher scores indicate greater the physical dependence on nicotine. Similarly, HSI contains two measures: TTFC of the day and daily consumption of cigarettes. These metrics have been used to predict behavioral and biochemical indices of smoking, including ability to quit and cancer incidence and mortality [14]. Both measures are reliable over time and are important predictors of quitting [15,16].

#### 3.1.2. Prevalence and Severity Measurement of Food Addiction

Food addiction is commonly measured using the Yale Food Addiction Scale (YFAS) [17], a 25-item tool developed in accordance with the substance dependence criteria of the DSM-IV [7,18]. The YFAS applies these criteria to the concept of food addiction for the purpose of identifying individuals who possess a predisposition toward the overconsumption of highly palatable foods within the previous 12 months [7,19]. The YFAS is capable of providing two scoring measures: (i) a cutoff score (yes/no) is achieved when an individual fulfills a minimum of three criteria and satisfies a clinical significant impairment or distress criterion, and (ii) a symptom count (0–7) in which the total number of endorsed criteria, with the exception of clinically significant impairment or distress criteria, is added together [7,18]. The YFAS represents a valid and reliable tool in identifying individuals who meet criteria for food addiction [7,20]; however, a major limitation of the YFAS is the reliance on self-reporting of symptoms. To date, there are no biological measurements to confirm the presence or absence of food addiction. Therefore, in conjunction with the lack of formal recognition by the DSM 5, a clinical diagnosis for food addiction does not exist at this time.

With the release of the DSM 5, criteria for diagnosing substance-related and addiction disorders have changed. Consequently, the 35-item YFAS 2.0 was developed to accurately align the concept of food addiction with the new DSM 5 substance use disorder criteria [21]. Another adaptation, the modified version of the YFAS (mYFAS), is an abbreviated nine-item version of the YFAS. Within the mYFAS, each of the seven DSM-4 substance dependence criteria is represented by one question [18]. The remaining two items of the mYFAS pertain to whether food or eating causes an individual clinically significant impairment or distress [18]. Both YFAS 2.0 and mYFAS have demonstrated marginal to good psychometric properties [20,22].

The observed prevalence rate of food addiction in a meta-analysis of 20 studies from North American and European countries was 19.9% (range: 16.3% to 24.0%) [23]. In a more recent meta-analysis of 36 articles, Burrows et al. (2018) concluded that the prevalence of individuals with mental health symptoms who met the cutoff for food addiction was 16.2% (range: 13.6% to 19.3%) [22]. A sex difference was also observed, where higher prevalence of food addiction appeared among females (12.2%) compared to males (6.4%) [23].

Although many individuals may not meet the threshold for food addiction, it is not uncommon for individuals to report specific symptoms. On average, individuals with mental health symptoms endorse approximately three symptoms for food addiction [22,23]. Of the seven YFAS symptoms for food addiction, "persistent desire or unsuccessful attempts to cut down or control eating" is one of the most frequently endorsed symptoms of food addiction [19,23,24].

#### 3.1.3. Theories of Food Addiction

Several theories link specific high-caloric and palatable foods to food addiction. Similar to addictive substances, these foods exist on a spectrum of addiction [7]. For example, individuals meeting the cutoff for food addiction convey significant problems with highly palatable foods such as chocolate, doughnuts, cookies, cake, candy, white bread, pasta, rice, crackers, French fries, and hamburgers compared to their counterparts without food addiction [25,26]. In addition, more frequent consumption of hamburgers, candy bar, milk chocolate, butter, pizza, and low-calorie beverages was associated with meeting the cutoff for food addiction [26]. Conversely, more frequent consumption of dark chocolate, homemade cookies, white rice, and sugar-sweetened beverages was negatively associated with food addiction [26]. Other theories suggest that food itself is not addictive, but the manner in which the food is consumed is an addictive behavior. Specifically, the repetitive behavior of food restriction and dieting leads to periods of overeating and binging. While controversy surrounds the addictive properties of foods, it is important to consider that tobacco took several decades before it was declared an addictive substance [27]. Addictive substances including alcohol, tobacco, and cannabis have gained acceptance at some point and it was not until these substances were recognized as being addictive that society implemented changes that would provide opportunities for individuals to receive treatment [27].

#### *3.2. Biology*

#### 3.2.1. Neurobiological Parallels between Food Addiction and Tobacco Use Disorder

Within the past several decades, neuroimaging has allowed the quantification of specific proteins and neurotransmitter receptors, the investigation of food and tobacco cues and their effects on neural activation, and the investigation of the integrity of gray and white matter, using positron emission tomography (PET), functional magnetic resonance imaging (fMRI), and structural MRI, respectively. Although no specific neuroimaging study has investigated the neural correlates of tobacco use disorder with comorbid food addiction, when investigated separately, the effects of food addiction on the brain often resemble those of tobacco use disorder. Specifically, highly palatable foods, such as those with high fat and sugar content, can activate the dopaminergic (DAergic) reward pathways [7,28–30], and specific conditioned food cues such as its sight, smell, and taste may trigger the desire or craving of eating [31]. This suggests that common neural substrates exist for both food and tobacco use disorder, both of which depend on DAergic pathways.

To date, no PET studies have investigated dopamine (DA) receptor availability in individuals with food addiction. Although obesity and food addiction are distinct disorders, most neurobiological literature surrounding food addiction is derived from obesity studies since the two conditions often co-occur. The first human neuroimaging study to examine striatal DA D2 receptor availability in relation to obesity was Wang et al. 2001 [32]. Measured by [11C]raclopride PET, striatal DA D2 receptor availability was significantly lower in obese individuals compared to healthy controls, and body mass index (BMI) was negatively correlated with D2 receptor availability. Similarly, low levels of DA are often reported in individuals addicted to drugs including cocaine [33], alcohol [34], opiates [35], and nicotine [36], and low DA receptor levels are associated with addictive behaviors irrespective of food or addictive drugs [32]. DA deficiency may perpetuate pathological eating to replenish the mesolimbic DAergic pathway, and feeding has been shown to increase extracellular DA levels in the

nucleus accumbens (NAcc) [37], a region thought to contribute to the reinforcing effects of euphoria [38]. It is possible that chronic overconsumption of food leads to increases in DA, resulting in DA D2 receptor downregulation. This produces a feed-forward cyclical pattern where overconsumption of food must then be sustained to replenish DA levels to avoid food cravings and withdrawal symptoms [39].

DA receptor availability in individuals with tobacco use disorder has been extensively investigated with [11C]raclopride PET. For example, a 26% to 37% reduction in binding potential, indicative of greater DA release, was observed in the left ventral caudate, NAcc, and left ventral putamen in cigarette smokers compared to nonsmokers [40]. In contrast, several other PET studies using tobacco cigarettes and alternative methods of nicotine administration, such as nicotine nasal sprays and nicotine gum, found no significant changes in binding potential within smokers [41–44]. More recent nonhuman primate PET studies found [11C]PHNO ([11C]-(+)-propyl-hexahydro-naphtho-oxazin) to be more sensitive to nicotine-induced DA release compared to [11C]raclopride [45]. To date, three studies have utilized [11C]PHNO PET in relation to nicotine administration and smoking-associated cues in humans. Specifically, following cigarette smoking, a 12% to 15% reduction in D2 and D3 receptor binding potential was observed compared to control conditions [46]. These findings are likely influenced by genetics, where, during abstinence, slow metabolizers of nicotine had lower [11C]PHNO-binding potential compared to fast metabolizers within the D2 regions of the striatum [47]. Interestingly, there was no change in [11C]PHNO-binding potential in the striatum of nicotine-dependent individuals following the presentation of tobacco-associated cues [48].

The effects of food cues and craving on neural activity and DA receptor binding have been widely investigated. fMRI studies demonstrated that food cues activate the amygdala, insula, orbitofrontal cortex, and striatum brain regions compared to neutral cues [49,50]. These cues of highly palatable foods activate similar reward neurocircuitry to tobacco use disorder [51]. Furthermore, food cravings are associated with increased bold-oxygen-level-dependent (BOLD) signals in the hippocampus, insula, and caudate [50], regions involved in craving, motivation, and memory [52]. PET studies demonstrated a positive association of food cravings and increased dorsal caudate and putamen regional cerebral blood flow [53], as well as an association of DA ligand binding within the dorsal striatum and feeding [54]. In summary, molecular imaging studies food cues provide supportive evidence of DAergic pathway activation.

Similar to food addiction, fMRI studies have examined the neuronal activation patterns produced by nicotine. For example, dose- and time-dependent BOLD signal increases were observed within the anterior cingulate cortex, dorsolateral prefrontal cortex, and medial prefrontal cortex brain regions in cigarette smokers [55]. This pattern of brain activation is consistent with DAergic pathways innervating the frontal cortex, as well as evidence supporting acute nicotine's role in positively enhancing reaction time, short-term memory, working memory, and attention [56]. Furthermore, smoking-cue fMRI studies demonstrated that nicotine-dependent smokers exhibited more BOLD signal activation than nonsmokers in the prefrontal cortex, ventral striatum, and NAcc brain regions [57,58]. In addition, contextual factors such as cigarette availability can affect neural activity, and variation in tobacco use disorder severity and genotype can modulate cue-induced activity [59].

#### 3.2.2. Neurobiology Unique to Tobacco Use Disorder

Nicotine is the main psychoactive component of tobacco, and it specifically acts as an agonist of nicotinic acetylcholine receptors (nAChRs) in the brain. nAChRs containing the α4 and β2 subunits are critical for mediating nicotine reinforcement, nicotine sensitivity, reward motivation, and DA release [60–65]. nAChRs are located throughout the brain, with highest density within the thalamus, basal ganglia, frontal cortex, cingulate cortex, occipital cortex, and insula [66,67]. Most importantly, nicotine stimulates the release of DA in the mesolimbic area, the corpus striatum, and the frontal cortex [68,69]. These DAergic pathways are critical in nicotine-induced rewarding behaviors [70], as well as in regulating reward, motivation, decision-making, learning, and memory [71].

Several preclinical and clinical studies of tobacco use disorder examined the effects of cigarette smoking and smoking-related behaviors on brain function, specifically β2-nAChR desensitization and subsequent upregulation [72]. Preclinical studies assessing nicotine administration in animals and postmortem human studies of smokers demonstrated β2-nAChR upregulation throughout the striatum, frontal cortex, anterior cingulate cortex, temporal cortex, occipital cortex, and cerebellum [73], suggesting greater levels of β2-nAChR desensitization and inactivation produced by long-term smoking or nicotine administration [74]. Brain imaging studies examining β2-nAChR availability in human smokers mimicked these preclinical finding [74–77]. Dysregulation of these brain regions following drug use is commonly associated with processing of drug cues and loss of inhibitory control, the primary contributing factor to relapse [78–80]. Furthermore, human postmortem studies of smokers with variable lifelong smoking histories and former smokers demonstrated that nAChR upregulation was reversible following abstinence [81]. Taken together, many preclinical and clinical brain imaging studies support the theory that long-term nicotine administration or chronic smoking can lead to nAChR desensitization and upregulation in smokers [82], but this upregulation is reversible following extended periods of smoking abstinence [81,83].

#### 3.2.3. Neurobiology Unique to Food Addiction

With the exception of the DAergic system, the literature on other neurocircuits implicated in food addiction is limited. To date, only a few fMRI studies have been conducted in individuals who met the YFAS cutoff threshold for food addiction. The first study by Gearhardt et al. (2011) found a positive correlation between food addiction scores and neural activation in the anterior cingulate cortex, medial orbitofrontal cortex, and amygdala when participants anticipated highly palatable foods [84]. Furthermore, upon tasteless food cue presentation, a negative correlation was observed between food addiction scores and activation in the caudate, a region implicated in reward motivation [84]. In a more recent fMRI study, Schulte et al. (2019) investigated food-cue effects on neural activity in obese women who either met the YFAS 2.0 threshold cutoff or did not [85]. When presented with highly palatable foods, participants with food addiction exhibited moderate, elevated activation in the superior frontal gyrus. Decreased activations were observed when minimally processed food cues were presented. Interestingly, participants in the control group had opposite responses in this region [85]. Most of the literature on the neurobiology of food addiction is derived from studies examining obesity; however, the findings from Schulte et al. (2019) presented food addiction as a unique phenotype within obesity.

Provided the limited neuroimaging research, it is theorized that dysregulation in the hypothalamus may also contribute to food addiction given its role as the main homeostatic regulation center for feeding behaviors. The hypothalamus integrates different hormonal and neuronal signals to control appetite and energy. This regulation system monitors body adiposity by using hormones such as leptin, insulin, and ghrelin [86]. Ghrelin, the "hunger peptide", stimulates DAergic reward pathways, whereas leptin and insulin inhibit these circuits [49]. Several brain regions, such as the amygdala, hippocampus, insula, orbitofrontal cortex, and striatum, are also involved with the regulation of feeding and appetite [49]. These brain structures are involved in learning about food, allocating attention and effort towards food, conditioning reward with specific food cues in the environment, and integrating homeostatic information such as hunger with availability of food in the environment [49,71]. For a recent review of potential mechanisms for food addiction (in the presence of obesity) using a systems approach, see [87].

#### *3.3. Role of the Gut Microbiome*

3.3.1. Parallels in the Role of Gut Microbiome in Both Tobacco Use Disorder and Food Addiction

As discussed in the section above, both food addiction and tobacco use disorder reflect an imbalance in the extended reward system in response to environmental stimuli. Peptides that regulate appetite such as glucagon-like peptide 1 (GLP-1), ghrelin, leptin, peptide YY, and neuromedin U are expressed throughout the brain reward circuitry, providing strong evidence that food addiction and tobacco use disorder share overlapping gut–brain axis mechanisms [88]. Endocrine signals play a significant role in reward regulation and dysregulation, which is a hallmark feature of all addictive disorders. The neuropeptides that have been studied most extensively are ghrelin and GLP-1.

There were only a few studies exploring the mechanism via which the gut microbiome affects the behavioral response to drugs of abuse [89–91]. Nonetheless, there is preliminary clinical and preclinical evidence of bacterial dysbiosis in response to drugs of abuse, which requires further investigation [92–96].

#### 3.3.2. Tobacco Use Disorder and the Gut Microbiome

The effect of smoking on the brain–gut axis and its behavioral implications has been largely unexplored. However, it has been shown that smoking induces specific changes in the microbiome. Furthermore, evidence suggests that smoking cessation induces an increase in microbial diversity [97,98], thereby reversing the negative effects of smoking and tobacco dependence on gut microbiota. A study by Biedermann et al. (2013) examined the association between smoking and gut microbiota in smokers without specific diseases [98] and showed that smoking cessation induced an increase in Firmicutes and Actinobacteria and a decrease in Bacteroidetes and Proteobacteria [98]. However, this study was conducted in only 10 subjects, and most of the participants developed an increased BMI following smoking cessation [98]. Previous studies showed that higher BMI is associated with increased Firmicutes and decreased Bacteroidetes in the gut compared to normal BMI [99,100]. Therefore, changes in gut microbiota following smoking cessation [98] might be associated with cessation-induced weight gain, as well as with smoking itself.

In a more recent large-scale cross-sectional study that included current, former, and never male smokers, smoking status influenced gut microbiota composition. Specifically, current smokers had a higher proportion of Bacteroidetes compared to never and former smokers, as well as lower proportions of Firmicutes and Proteobacteria compared with never smokers [89]. There were no observed differences in the composition of gut microbiota between never and former smokers, suggesting that smoking cessation allows gut microbiota composition to recover to pre-smoking status. The three groups did not differ significantly in terms of BMI or nutrient intake, thereby providing stronger evidence for the reversal of gut microbiota changes to normal upon smoking cessation.

Furthermore, a recent study compared the oral and gut microbiota in current smokers, current e-cigarette users, and healthy controls [101]. Tobacco smoking was associated with significant differences in the bacterial profiles in fecal, buccal, and saliva samples, while the e-cigarette users were no different to healthy controls. In keeping with previous studies, tobacco smokers had higher relative abundance of *Prevotella* and lower relative abundance of *Bacteroides* in their gut microbiota. This is in accordance with existing data demonstrating gut microbiotal changes following smoking cessation [98,102]

#### 3.3.3. Food Addiction and the Gut Microbiome

Eating behavior is regulated by both homeostatic and hedonic mechanisms in the central nervous system (CNS). These mechanisms involve orchestrated signaling from several sources including gut peptides, endocrine signals, and neuronal impulses, as well as signals from the gut microbiota. For example, ghrelin signals hunger and craving, putatively via amplification of DA signaling [103]. On the other hand, satiety is signaled by other intestinal hormones such as glucagon-like peptide 1 (GLP-1) and peptide YY [104]. Insulin also triggers hunger and increases the palatability of sugar. There is evidence that the gut microbiota can regulate insulin sensitivity through various mechanisms [105]. As discussed in the previous section, normal eating behavior is under the control of the extended reward network, which is involved in the processing of all rewarding stimuli including but not limited to food-related behaviors. These processes become maladaptive when the salience of a specific type of reward such as highly palatable food is greater than that of other stimuli and becomes

preferred at the expense of other rewards, thereby leading to addiction-type behavior. At this point, the hedonic system becomes more prominent than the homeostatic system in the regulation of food intake. Therefore, eating behavior becomes driven predominantly by activation of the salience network of the brain, whereby food cues activate this network leading to increased attentional bias to the food cues at the expense of other cues. This in turn results in the uncontrolled overconsumption of highly palatable food.

While there is currently no evidence in humans that food addiction is caused by an altered gut microbiome or that it is driven by particular gut microbes or microbial metabolites, there is substantial evidence from rodent models that point to a role of the gut microbiome in food addiction-like behaviors. However, there is overwhelming evidence that a high-sugar, high-fat diet results in changes to the gut microbiome, which further "supports" addictive-like eating behaviors [87].

The few studies that examined the relationship between the gut microbiome and its metabolites with addictive-like eating behaviors have shown that tryptophan metabolites are implicated in modulating brain–gut–microbiome interactions [106]. In a recent study [107], the association between microbial profiles and tryptophan metabolites with food addiction was examined in a sample of human females with high BMI. The study found that there was a difference in the gut microbiome of females with food addiction versus those without, whereby levels of *Bacteroides* and *Akkermansia* were negatively associated with food addiction.

#### *3.4. Psychological*

#### 3.4.1. Childhood Adversity

Adverse childhood experiences (ACEs) have been shown to have deleterious effects on adult health [108–110]. To standardize the operationalization of childhood adversity within studies, Felitti et al. (1998) developed the ACE Survey, which quantifies an individual's reports of exposure to abuse, neglect, and household dysfunction before the age of 18 [111]. These questions surveyed ACEs by asking about behaviors rather than subjective experiences of trauma. This survey originally encompassed seven categories including three of abuse and four of household dysfunction. More recent versions of the ACE study capture two additional categories of neglect and parental separation or divorce [112]. Cronholm et al. proposed the expansion of the concept of ACEs to include experiences such as witnessing violence, feeling discrimination, living in an unsafe neighborhood, experiencing bullying, and living in foster care to fully understand the influence of childhood adversity on adult substance use [113].

#### 3.4.2. Overlap in Childhood Adversity of Tobacco Use Disorder and Food Addiction

Childhood adversity influences various physiological and behavioral mechanisms that contribute to adult addictive behaviors. Studies demonstrate that chronic stress in childhood may cause changes in the nervous, endocrine, and immune systems. Alterations in these systems may lead to impairments in cognitive, social, and emotional development that predispose individuals with ACEs to adopt addictive behaviors [114–116]. The original ACE study reported that individuals who experienced four or more categories of adversity were 2.2 times more likely to be a current smoker and 1.6 times more likely to have a BMI ≥ 35 (severe obesity) [111]. This study did not evaluate food addiction. While the focus here is on tobacco use disorder and food addiction, individuals who report ACEs also engage in other addictive behaviors including alcohol abuse and drug use [111]. This suggests that childhood adversity may have significant and varying downstream effects on numerous adult health behaviors. While there is research on the relationships between childhood adversity and food addiction and between childhood adversity and tobacco use disorder, research is needed to better understand the overlap of food addiction and tobacco use disorder.

#### 3.4.3. Childhood Adversity and Tobacco Use Disorder

Using the ACE Survey, Felitti et al. (1998) produced a landmark paper describing the gradient relationships between the number of categories of childhood adversity and the prevalence of current smoking. Specifically, 6.8% of participants who reported no adversity (zero categories) and 16.5% of participants who reported four or more categories of adversity were current smokers. Participants who reported four or more categories of adversity were 2.2 times more likely to smoke than those who reported no adversity (odds ratio (OR) adjusted for age, gender, race, and educational attainment) [111]. Since then, a systematic review and a meta-analysis of 37 studies found that individuals who reported at least four categories of ACEs were more than twice as likely to be current smokers. Furthermore, there was a moderate association between childhood adversity and smoking [113].

#### 3.4.4. Childhood Adversity and Food Addiction

There is limited research into food addiction. One study identified childhood abuse as a risk factor for food addiction. This study of 57,321 adult women examined the association between child abuse (specifically, physical and sexual child abuse) and food addiction (defined as three or more clinically significant symptoms on the mYFAS). In this sample, over 8% of participants reported severe physical abuse in childhood, 5.3% reported severe sexual abuse, and 8% met the criteria for food addiction. Findings indicated that women with food addiction had a higher BMI than women without food addiction. Furthermore, severe physical and severe sexual abuse was associated with about 90% increased risk for food addiction (physical abuse: relative risk (RR) 1.92, 95% confidence interval (CI) 1.76 to 2.09; sexual abuse: RR 1.87, 95% CI 1.69 to 2.05). The RR for combined severe physical abuse and sexual abuse was 2.40 (95% CI 2.16 to 2.67) demonstrating the additive effects of adversity [117]. Childhood adversity has also been linked to disordered eating, including food addiction, obesity, and binge eating, which has overlapping characteristics with food addiction [111,118–122].

#### 3.4.5. Attachment Insecurity

Attachment theory describes how individuals internalize experiences with their caregivers to form mental representations of themselves and others. Individuals with high attachment anxiety tend to have negative self-views, are concerned about rejection, magnify their expressions of distress, and prefer close proximity to and support from a partner [123,124]. Individuals with high attachment avoidance tend to report positive self-views [123], suppress expressions of distress, and prefer emotional distance in relationships [124]. Individuals can be characterized by varying levels of attachment anxiety and avoidance, and both types of insecurity can co-occur [124]. The experience of childhood trauma is associated with both attachment anxiety and attachment avoidance in adulthood [125]. Attachment theory provides another framework to understand tobacco use disorder.

#### 3.4.6. Overlap in Attachment Insecurity of Tobacco Use Disorder and Food Addiction

Attachment insecurity may influence several processes that may in turn contribute to tobacco use disorder and food addiction. While there is research on the relationships between attachment insecurity and food addiction and between attachment insecurity and tobacco use disorder, research is needed to better understand the overlap of food addiction and tobacco use disorder. Attachment insecurity is related to affect regulation [126], i.e., "the process by which individuals influence which emotions they have, when they have them, and how they experience and express these emotions" [127]. Emotional regulation involves efforts to up- and downregulate positive and negative emotions [127]. High levels of attachment insecurity are associated with a deficit in affect regulation [128]. Individuals with high attachment insecurity may feel less capable of disengaging from negative feelings and in turn attempt to calm themselves through food or other substances. Food or other substances (e.g., tobacco, alcohol, or drugs), when consumed in order to reduce feelings of insecurity, have been called "external regulators of affect" [129]. Individuals with high attachment

insecurity are more likely to use these "external regulators of affect" instead of utilizing more adaptive emotional regulation strategies. Attachment insecurity has been studied in the context of disordered eating, eating disorders, and obesity, but there is still limited research on food addiction [130–132].

#### 3.4.7. Attachment Insecurity and Tobacco Use Disorder

Insecure attachment patterns, specifically attachment anxiety, are associated with the use of tobacco and other drugs [133–136]. Attachment anxiety was associated with increased use of tobacco to reduce stress in college students [135]. In undergraduate and graduate students, significant differences in attachment patterns in tobacco users and nonusers were observed [122]. In a study of adults, findings suggested that attachment anxiety, but not attachment avoidance, was associated with current smoking [137]. In contrast, a study in adult women found that attachment avoidance was associated with being a current smoker [133]. It is currently unclear whether a particular dimension of attachment insecurity, such as attachment avoidance or attachment anxiety, is associated with tobacco use disorder.

#### 3.4.8. Attachment Insecurity and Food Addiction

Few studies examined the association of attachment insecurity and food addiction. One study of a national nonclinical sample of 1841 respondents in the Czech Republic found that attachment insecurity was associated with increased scores of mYFAS 2.0 [138]. In a study of 195 adult women from an eating disorder treatment center, the prevalence of food addiction was 83.6%, and the most frequently reported food addiction criteria were "clinically significant impairment or distress in relation to food", "craving", and "persistent desire or repeated unsuccessful attempts to cut down". Within this sample, no differences in attachment insecurity were found between those meeting the criteria for food addiction and those who did not fulfill the criteria for food addiction [132]. As such, there may be a similarity between food addiction and eating disorders in terms of attachment patterns. The literature suggests that attachment insecurity is associated with unhealthy eating, including anorexia nervosa, bulimia nervosa, nonclinical levels of disordered eating, and obesity [130,131,139–148].

#### *3.5. Treatment*

#### 3.5.1. Treating Tobacco Use Disorder

There is a wide array of smoking cessation interventions that are well-established by evidence from several systematic reviews. Individualized treatments such as nicotine replacement therapy (NRT), varenicline, bupropion, behavioral supports, e-cigarettes, and combination therapies have positive outcomes on cessation rates and sustained abstinence. In addition, community- or government-level efforts such as prohibiting smoking in public spaces, advertising restrictions, and health warning labels contribute to reductions in smoking.

NRT is a common first-line, over-the-counter cessation aid that is available in forms such as the patch, lozenge, inhaler, and gum. NRT has a >6.5% sustained abstinence rate after 6 months, more than double that of placebo [149,150]. Compared to placebo, NRT shows no statistically significant differences in adverse events, except nausea, which has been listed as a common side effect [150]. Using the NRT patch together with another type of NRT (e.g., gum, lozenge, mist, or inhaler), can increase that rate by an additional 15–36% [151]. NRT dose and duration also affect quit rate. Quit success is positively correlated with higher-dose NRT patches (25 mg worn over 16 h or 21 mg worn over 24 h) compared to smaller-dose patches (15 mg over 16 h or 14 mg over 24 h) [151].

Varenicline is well established as achieving the highest quit and sustained abstinence rate of all cessation aids, with minimal risk for adverse effects (see, for example, [149–151]). Varenicline more than doubles the chances of quit compared to placebo [152,153], helping approximately 50% more people to quit and have sustained abstinence than the NRT patch, tablets, spray, lozenge, and inhaler, and 70% more than NRT gum [152]. The antidepressant bupropion has similarly high quit success, making it 52–71% more likely that a person will quit [154]. There have been concerns surrounding

varenicline and bupropion's linkage to psychiatric adverse events. A systematic review did not support this for varenicline, where the most frequently reported adverse event was nausea [153]. However, there is high-certainty evidence that unwanted mental health side-effects and adverse events linked to taking bupropion lead to lower medication adherence [154]. Furthermore, highest quit and sustained abstinence rates occur when pharmacotherapies are used in combination with behavioral supports. Behavioral therapy has been shown to increase effectiveness of pharmacotherapies by 83% to 97% across different care settings [155]. One systematic review compared the effects of brief physician advice to quit with offering assistance in the form of behavioral support or medication [156]. Physicians who offered assistance generated more quit attempts than those who gave advice to quit on medical grounds. Furthermore, when assistance was delivered in the form of motivational interviewing, a goal-oriented and patient-centered counseling approach that elicits motivation for change, abstinence rates were statistically significant and demonstrated up to 45% greater odds of smoking abstinence than control groups [157].

In recent years, e-cigarettes have become commonly used smoking cessation aids. While evidence is currently limited, the literature suggests the promise of e-cigarettes being an effective smoking cessation aid [158,159]. There is moderate-certainty evidence [using Grading of Recommendations Assessment, Development and Evaluation (GRADE)] from Cochrane systematic review that e-cigarettes with nicotine are more effective at helping people stop smoking for at least 6 months than NRT (three studies; 1498 participants), nicotine-free e-cigarettes (three studies; 802 participants), and no support or behavioral support alone (four studies; 2312 participants) [158]. Of the adverse events reported for e-cigarettes, throat and/or mouth irritation was the most commonly reported. Moderate-certainty evidence indicates the potential for results to change when more evidence becomes available. More research is needed to determine if e-cigarettes with nicotine are the preferred option for smoking cessation.

In summary, combination therapies that include either pharmacotherapy and behavioral support or a lower-risk nicotine product such as NRT or e-cigarettes with behavioral support increased quit attempts and boosted the level of sustained abstinence after one year. Organizational interventions also played a role in the sale and ease of promoting healthier smoking behaviors.

#### 3.5.2. Treatment Options for Food Addiction

There is currently no well-established treatment model for intervening on food addiction. This is unsurprising given that there is no formal recognition of food addiction as a neurological or behavioral disease within the DSM 5 due to the debate of perceiving food addiction as a (non)substance use disorder or as a behavioral addiction. Furthermore, the heterogeneity of addiction makes it difficult to translate addiction into a working treatment model.

In its contextualization along the spectrum of eating disorders within the scientific literature, there are notable behavioral treatment options. Psychosocial interventions for food addiction include reducing access to processed foods, reducing habit-based eating, removing restrictions on eating healthy foods, and behavioral therapies to improve emotional regulation and to help combat submission to cravings and emotional eating [160]. Furthermore, participation in an integrative and psychological weight management group demonstrated promise in treatment efficacy [161]. Learning about mindful eating, keeping a food diary and keeping track of body weight, creating and maintaining an exercise plan, and planning for social eating are tactics that are taught to enable the maintenance of healthy body weight [161]. A systematic review revealed that additional research is needed to develop and test the efficacy of these types of interventions within the context of food addiction [162].

Noninvasive brain stimulation has been used most frequently for the treatment of addictions such as tobacco use disorder [163]. Transcranial direct current stimulation (tDCS) in particular, is a safe, economical, and accessible means of modifying neural activity [163]. A systematic review highlighted that tDCS significantly improved the symptoms of food addiction by reducing food cravings brought on by visual stimuli [164]. Craving was measured before and after stimulation using visual analogue scales, eye tracking, or the Food Craving Questionnaire—State, and tDCS was found to significantly repress the desire to eat, leading to less food consumption [164]. However, due to underpowered studies and the complexity of addiction, more research is necessary to make any definitive comments about the success of tDCS on food addiction. Tailoring neuromodulating interventions to individuals or subgroups, on the basis of cognitive and neural profiling, might prove to be useful [163]. Since food addiction often presents with comorbidities, current research suggests using evidence-based interventions to address other conditions first [162].

#### **4. Current and Future Research Directions**

Evidence demonstrates that individuals can expect to gain an average of 4 to 5 kg of weight after successfully achieving smoking cessation [165,166]. For many tobacco users, this potential increase in weight can be a substantial obstacle when attempting smoking cessation [167,168], thereby leading to continued tobacco use. One possible explanation for post-smoking cessation weight gain is that quitting smoking increases the desire to consume highly palatable, high-calorie foods. Consequently, the authors of this review developed a feasibility study with the purpose of examining the association between food addiction symptoms and smoking behavior among individuals seeking treatment for tobacco use disorder. This feasibility study recruited individuals from the Nicotine Dependence Clinic at the Centre for Addiction and Mental Health (CAMH) who had yet to begin their smoking cessation treatment. Those individuals who provided consent to participate in the study were asked to complete a survey which included the FTND and mYFAS questionnaires.

The sample of this feasibility study included 51 participants seeking treatment for tobacco use disorder. The majority of participants in this convenience sample were male (58.8%). Two-thirds of the sample consumed <20 cigarettes per day (CPD). The prevalence of individuals meeting the cutoff for food addiction was 11.8%. The average symptom count for food addiction was 1.5 ± 1.8 (standard deviation (SD)) symptoms.

Spearman rank correlation coefficients were conducted to examine the association between tobacco use disorder, specifically CPD, and food addiction symptom count, as measured by the mYFAS. No significant association was observed between CPD and food addiction symptom count for the overall sample (*r*<sup>s</sup> = 0.21, *p* = 0.14) or by sex (male: *r*<sup>s</sup> = 0.20, *p* = 0.28; female: *r*<sup>s</sup> = 0.24, *p* = 0.31).

The results of this feasibility study, even though underpowered, represent an important initial step in developing future research. Using a cross-sectional study design, we were unable to observe a significant association between food addiction symptom count and CPD. The findings from another cross-sectional study examining the association between smoking and food addiction found that male smokers report twice as many YFAS symptoms compared to male nonsmokers [169]. However, among females, no relationship was observed between food addiction symptoms and smoking [169]. While this study provides some evidence of an association between smoking and food addiction, more research is needed. Applying a longitudinal study design and examining how changes in smoking behavior relate to changes food addiction symptoms could provide greater insight into the relationship between smoking cessation and weight gain. Furthermore, the feasibility study also revealed that the prevalence of food addiction may be lower among individuals seeking smoking cessation treatment compared to the general population, suggesting, perhaps, that food addiction may have a minor role in post-cessation weight gain. Therefore, examining other concepts such as the role that the gut microbiome may have on smoking cessation weight gain could be promising. However, it is also plausible that achieving smoking cessation could result in the adoption of an addictive behavior involving the overconsumption of highly palatable foods, a process known as addiction transfer [170].

E-cigarettes have been demonstrated to be effective smoking cessation aids [158]. As outlined above, e-cigarette users have gut and oral microbiomes that are more similar to healthy controls than to tobacco smokers [103]. Given that *Bacteriodes*, the bacterium that is most commonly significantly decreased in smokers compared to healthy controls and e-cigarette users, is also implicated in

obesity [171], there is the potential that switching smokers to e-cigarettes may decrease their risk of cessation-related weight gain. This hypothesis has not yet been tested, and there is no clear evidence that smokers who switch to e-cigarettes avoid weight gain. As such, these are important research gaps to address.

#### **5. Limitations**

This was a narrative review; thus, limitations with respect to the scope of literature covered are present. The authors made every attempt to be systematic in their initial literature search, but they may have missed some important publications. Narrative reviews are also subject to bias, but the authors did their best to mitigate this by using systematic approaches to their literature search.

#### **6. Conclusions**

Food addiction and tobacco use disorder share similar but not identical neurological, physiological, and behavioral abnormalities. We attempted to summarize these similarities between the two disorders where there is evidence of their existence. Differences between the two disorders should not lead one to conclude that food addiction is not a "real" disorder. As argued in a recent perspective [172], core features of addiction can differ dramatically depending on which substance is being used, reflecting different underlying neurobiological processes at play. For example, the pattern of consumption and symptoms of withdrawal exhibited in cocaine use disorder completely different to what is seen in tobacco use disorder.

Addiction is a complex disorder that we are only beginning to understand at a system level. The role of the gut–brain axis in brain reward mechanisms related to addiction needs to be further researched. In addition, more attention needs to be paid to the role that early life events have on the risk of developing an addictive disorder. Taking learnings from different types of addictive behaviors such as food addiction, as well as newer "behavioral addictions" such as internet gaming, will help to further our understanding of the underlying mechanisms common to all addictive disorders. This will in turn pave the road toward effective treatment options. Exploring the commonalities between tobacco use disorder and food addiction is a step in this direction.

**Author Contributions:** L.Z. conceptualized the basis and structure of the review. L.Z., S.A., T.L.L., W.K.d., and D.D. wrote the initial drafts and subsequent drafts of the paper. L.Z., S.A., T.L.L., W.K.d., D.D., and P.S. approved the final version. All authors have read and agreed to the published version of the manuscript.

**Funding:** L.Z.: T.L.L., W.K.d., D.D., and P.S. are paid employees of the Center for Addiction and Mental Health. S.A. is a postdoctoral fellow. Funding for infrastructure was from the Center for Addiction and Mental Health.

**Conflicts of Interest:** S.A., T.L.L., and D.D. report no conflicts of interest. L.Z. has received peer-reviewed funding from Canadian Institutes of Health Research, Canadian Cancer Society, Ontario Ministry of Health of Long-term Care, including salary support from Pfizer Inc. (GRAND Award) and the Health Services Research Fund from the Ontario Ministry of Health. W.K.d. reports receiving grants from the Public Health Agency of Canada and Pfizer Inc. W.K.d. is also a shareholder of Abbott Laboratories. P.S. reports receiving grants and/or salary and/or research support from the Center for Addiction and Mental Health, Health Canada, Ontario Ministry of Health and Long-term care, Canadian Institutes of Health Research, Canadian Center on Substance Use and Addiction, Public Health Agency of Canada, Ontario Lung Association, Medical Psychiatry Alliance, Extensions for Community Healthcare Outcomes, Canadian Cancer Society Research Institute, Cancer Care Ontario, Ontario Institute for Cancer Research, Ontario Brain Institute, McLaughlin Center, Academic Health Sciences Center, Workplace Safety and Insurance Board, National Institutes of Health, and the Association of Faculties of Medicine of Canada. P.S. also reports receiving funding and/or honoraria from the following commercial organizations: Pfizer Inc./Canada, Shoppers Drug Mart, Bhasin Consulting Fund Inc., Patient-Centered Outcomes Research Institute, ABBVie, and Bristol-Myers Squibb. Furthermore, P.S. reports receiving consulting fees from Pfizer Inc./Canada, Evidera Inc., Johnson & Johnson Group of Companies, Medcan Clinic, Inflexxion Inc., V-CC Systems Inc., MedPlan Communications, Kataka Medical Communications, Miller Medical Communications, Nvision Insight Group, and Sun Life Financial.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
