A neurocognitive perspective on the relationship between exercise and reward: Implications for weight management and drug addiction

The impact of exercise on food reward is increasingly being discussed as an interplay between executive function (EF), homeostasis and mechanisms promoting or undermining intentional behaviour change. Integrating current knowledge of neurocognitive processes encompassing cognitive and affective networks within an energy balance framework will provide a more comprehensive account. Reward circuitry affected by recreational drugs and food overlap. Therefore the underlying processes explaining changes in drug-taking behaviour may offer new insights into how exercise affects the reward value of recreational drugs and food. EF is important for successful self-regulation, and training EF may boost inhibitory control in relation to food- and drug-related reward. Preclin- ical and clinical observations suggest that reward-seeking can transfer within and between categories of reward. This may have clinical implications beyond exercise improving metabolic health in people with obesity to un- derstanding therapeutic responses to exercise in people with neurocognitive deficits in non-food reward-based decision making such as drug dependence.


A R T I C L E I N F O
Handling Editor: Dr. S. Higgs

A B S T R A C T
The impact of exercise on food reward is increasingly being discussed as an interplay between executive function (EF), homeostasis and mechanisms promoting or undermining intentional behaviour change. Integrating current knowledge of neurocognitive processes encompassing cognitive and affective networks within an energy balance framework will provide a more comprehensive account. Reward circuitry affected by recreational drugs and food overlap. Therefore the underlying processes explaining changes in drug-taking behaviour may offer new insights into how exercise affects the reward value of recreational drugs and food. EF is important for successful selfregulation, and training EF may boost inhibitory control in relation to food-and drug-related reward. Preclinical and clinical observations suggest that reward-seeking can transfer within and between categories of reward. This may have clinical implications beyond exercise improving metabolic health in people with obesity to understanding therapeutic responses to exercise in people with neurocognitive deficits in non-food reward-based decision making such as drug dependence.
Foods and physical activities are natural sources of reward. Their pleasurable effects have evolved from their necessity for health, reproduction and survival. Recreational drugs have no such evolutionary advantages, yet exert their abuse potential via some of the same neural systems that mediate our enjoyment and pursuit of natural rewards. The distinction between natural and artificial rewards in modern life can be difficult to discern given the multitudinous formulations of foods and exercise that can be pursued. Nevertheless most foods and physical activities retain some benefit for biological functioning whereas recreational drugs are mostly used for their psychoactive effects. Food reward plays an important part in the control over food intake with a role in food seeking, consummatory behaviours and cessation of eating, in part through interaction with processes involved in hunger and satiety. This extension to the traditional homeostatic model has increasingly guided research on exercise and appetite towards the inclusion of food reward measures alongside more conventional appetite methodology to explain qualitative and quantitative changes in food intake in response to acute or chronic exercise. An increase in the reward value of food has been proposed as a mechanism driving compensatory eating after acute exercise, and excess energy intakes in response to highly palatable energy dense foods (Killgore et al., 2013). Elevated reward responses to food have also been proposed as a reason for poor weight loss and body composition outcomes during exercise training programs (Finlayson et al., 2011). Conversely, observational studies show that less active individuals show increased reward-responses to food, and may respond differently to food following an exercise bout (Beaulieu, Oustric, & Finlayson, 2020;Cornier, Melanson, Salzberg, Bechtell, & Tregellas, 2012). This implies that exercise modulates food reward via its impact on homeostatic regulation of nutrient intake and energy balance outcomes.
Exercise (defined here as structured, planned activity requiring physical effort and coordination) has ostensibly paradoxical effects in both elevating motivation to eat and enhancing satiety (King et al., 2009), but in the longer-term aligns energy intake with sustained increases in energy expenditure (Blundell, Gibbons, Caudwell, Finlayson, & Hopkins, 2015). Motivation to eat and food reward-driven behaviours also operate in the context of other non-food related motivations and rewards (e.g. health, wellbeing). Motivation to eat and food-based reward systems are largely subcortical and reactive, while decision-making can involve rationale conscious cortical processing, and complex motivations (i.e. planning to engage in exercise behaviours). Dual process theory suggests that decisions are motivated by separate cognitive (slow, reflective) and affective (fast, reactive) networks. This dichotomy tends to oversimplify the degree of interplay between cognitive, affective and metabolic systems and the neuroanatomical circuitry that links them (Keren & Schul, 2009).
The impact of exercise on food reward is increasingly being discussed as an interplay between reflective mechanisms promoting executive function and self-regulation, and reactive mechanisms undermining intentional behaviour change (Joseph, Alonso-Alonso, Bond, Pascual-Leone, & Blackburn, 2011). Integrating current knowledge of neurocognitive processes signalling food seeking (motivation to eat), consummation (appetitive behaviours) and cessation of eating (satiation and satiety) within the existing energy balance framework is likely to provide a more comprehensive account of the role of exercise on appetite control. This may have clinical implications beyond improving metabolic health in people with obesity to understanding therapeutic responses to exercise in people with neurocognitive deficits in non-food (e.g. drug) reward-based decision making. Neuroendocrine reward systems affected by drugs with abuse potential and food overlap. Therefore, the underlying processes explaining changes in drug-taking behaviour may offer new insights into how exercise affects the reward value of food.

A neurocognitive model of exercise and food reward
The impact of exercise on food reward has been of interest to researchers seeking to understand the interplay between physical activity, appetite control and weight management. Therefore it has typically been studied within an energy balance framework wherein appetite is modulated by interactions between endocrine, metabolic, and peripheral nervous system signals influenced, inter alia, by the composition of ingested foods, energy expenditure and feedback from the body composition of energy stores (Watts, Kanoski, Sanchez-Watts, & Langhans, 2022). These peripheral signals are integrated in a set of neural pathways and receptors extending from the brainstem to the arcuate nucleus in the hypothalamus and other hypothalamic structures responsible for the homeostatic regulation of appetite (in this context matching energy intake to changes in expenditure). These structures also interact with the reward system via direct and indirect projections from the hypothalamus to mesocorticolimbic areas responsible for motivational "wanting" and experiential "liking" for food rewards (Plassmann, Schelski, Simon, & Koban, 2022). Wanting is the process that ascribes motivational value to seek and ingest a food mediated by the release of dopamine (DA) from the ventral tegmental area (VTA) to the nucleus accumbens (Nacc) and amygdala. Liking is the sensory pleasure exerted while eating a food and is generated by endogenous opioid-binding in specialized clusters of neurons distributed along the reward pathway and particularly present in the Nacc shell (Smith, Mahler, Peciña, & Berridge, 2010). Both liking and wanting cue learning about future bodily states.
The numerous possible decisions around daily eating and physical activity vary greatly in terms of the cognitive or physical effort required to enact them. For most people, the ease and automaticity of rewarddriven decision-making leads to habits that favour consumption of readily-available energy-dense foods and the tendency to avoid discomfort or fatigue from exercise (Iso-Ahola, 2017). In contrast, effortful conscious control is often required for the decisions that change and sustain behaviours towards healthy eating, physical activity, and weight loss. Importantly, this type of effortful cognitive processingtermed Executive Function (EF) -is a limited-capacity resource (McGuire & Botvinick, 2010). Therefore, depending on the situation and the cognitive resources available, cognitive or affective neural systems are drawn upon differently to influence decisions and motivate their corresponding behaviour. The mechanisms behind affective and cognitive decision-making share many of the same brain regions, yet each domain has dissociable pathways. Significantly, the planning and execution of exercise and structured physical activity can impact on these pathways in ways that can affect decisions based on the reward value of food. The next section provides a short overview of the interplay between cognitive and affective networks involved in EF and their implication for understanding the impact of exercise on food reward.

Neurobiology of executive function and reward-driven exercise and eating behaviours
Seeking reward, or avoiding aversive experiences, is a key motivator for food and physical activity goal-based decisions. When reward is anticipated, DA neurons in the VTA project to the Nacc and further motivate the reward-seeking behaviour (Sugam, Day, Wightman, & Carelli, 2012). In familiar exercise or eating situations, the reward value will be representative of past experiences coded in the ventromedial prefrontal cortex (vmPFC) which is drawn on to inform decisions. If reward occurs as expected, the DA release encourages strengthening of synapses in the hippocampus, where these memories are processed and increases the likelihood of making the same decision in future (Stopper, Tse, Montes, Wiedman, & Floresco, 2014). In this way familiar patterns of exercise and eating behaviour become engrained habits requiring minimal cognitive resources to maintain. In novel or uncertain scenarios, such as attempts to break out of existing unhealthy habits towards healthier behaviours, the dorsolateral prefrontal cortex (dlPFC) is recruited to make predictions of the likelihood a new behaviour will yield reward (Schumacher, van Holstein, Bagrodia, Le Bouder, & Floresco, 2021). The dlPFC forms a neural network incorporating the hippocampus, amygdala and Nacc. During exercise, the dlPFC can inhibit negative emotions and memories and support new goal-directed behaviour (Bigliassi & Filho, 2022). Crucially, this anticipatory signalling occurs independently from the actual hedonic value of the behaviour, if a mismatch occurs between what is predicted and experienced, the hedonic value can shift from negative to positive or vice versa through repetition and cognitive reappraisal (Salamone et al., 2015).
Particularly when food and physical activity decisions have more complex long-term implications (such as personal health and wellbeing), this slower reflective route of decision making must be utilised. Activation of the vmPFC in response to palatable food cues is modulated by the dlPFC in situations where healthiness cues are also present and self-control is engaged (van Meer, Charbonnier, & Smeets, 2016). This implies that health-promoting cues from planning and engagement with physical activity and exercise can support decisions that overcome immediate sensory gratification from consuming palatable foods to align with long-term health goals. Whilst reflective accounts of decision making aptly depict conscious choices, they overlook the role that affective signalling plays on the process most of the time. The amygdala and vmPFC are both involved in producing emotional states, with the amygdala responding to immediate events in the environment, whereas the vmPFC draws on stored memories and knowledge (Bechara, 2005). The affective neural network can also bypass cognitive appraisal entirely to mediate rapid impulsive behaviours. This network therefore has the capacity to 'short-circuit' conscious health-related choices. This fast subcortical route, channels external sensory information from the sensory cortex directly to the amygdala where it generates a response in the brainstem. The extent to which emotional input influences choice is determined by the functional connectivity between the amygdala and PFC. Interestingly, acute and chronic aerobic exercise has been shown to increase the functional connectivity of these brain regions and may partly explain the association between exercise and performance on EF tasks (Moore, Jung, Hillman, Kang, & Loprinzi, 2022). It remains to be tested whether the affective value of exercise or exercise-induced improvement in EF affects post exercise energy intake or food reward.
The implication here for exercise and food reward is that the reward value and emotional valence an individual ascribes to a food after acute or chronic exercise is to some extent routed through EF and associated circuits in the PFC to account for a broader range of psychological motives than simply its immediate sensory-hedonic impact. There is considerable scope for motivational and behavioural tensions to arise between multiple competing aspects of reward related to habitual unhealthy versus new healthy behaviours an individual is intending to implement (Greaves, Poltawski, Garside, & Briscoe, 2017). The outcome of these tensions may influence whether the new healthy behaviour is maintained. The impact of exercise on food reward cannot be understood as a simple function of the intensity and duration of exercise as a physiological stimulus. For example, the affective value ascribed to a 30-min bout of exercise might be positive, while the sensory experience itself is physically unpleasant. Likewise, the affective value of a piece of cake can be diminished if accompanied by feelings of regret, even when its sensory properties are highly rewarding. A further possibility is that exposure to the setting and execution of planned exercise could inhibit affective-reward based activations in response to palatable food.
It is likely the aggregate evaluative outcome of these different reward-aversion associations that determines the ability of an individual to adopt and maintain a new healthy behaviour or to relapse into prior learned habits. With reflective processing, these broader motives can be activated in the moments before a decision is made and they can be entrained (through repetition and practice), eventually becoming nonconscious and automatic. Interestingly, the importance of practice and training in exercise and eating behaviours, as separate from 'will-power' and motivation, is not often explicitly recognised in weight management interventions. The effect of exercise on food reward may then be influenced by a number of inputs including the actual experience of the intervention, prior learned behaviour and the goals of the individual. Aligning positive aspects in all of these domains to minimise 'aggregate discomfort' and maximise 'aggregate pleasure' may provide a strong basis for learned cues that reinforce healthy intended behaviours and offset the motivation to seek unhealthy rewards like energy-dense food or recreational drugs. Recognition of these phenomena potentially opens up pathways for increasing the effectiveness of behaviour change interventions by making them specific to the motivations, values and rewards of individual participants. This does not mean that metabolic signals concerned with energy balance regulation are unimportant in understanding exercise and food reward. Evidence is accumulating to show how diet, physical activity and appetite-related gastrointestinal hormones can affect cognitive processes involved in executive function that in turn modulate reward-driven behaviours (Berthoud, 2004;Higgs et al., 2017;Shin, Zheng, & Berthoud, 2009).

Metabolic signals of energy balance and executive function
As previously mentioned, EF is a critical domain of cognitive processing that encompasses the mental ability to perform actions involving working memory, inhibitory control, set-shifting (cognitive flexibility), attention and planning. EF is also important for successful selfregulation and training EF may boost self-control in relation to exercise-food-and drug-related reward (Hofmann, Schmeichel, & Baddeley, 2012). Leptin, an adipocyte-secreted hormone, has a primary role in energy balance and is known to regulate energy expenditure and appetite-signalling in the hypothalamus, but leptin receptors are also expressed in the PFC and thought to have an impact on a diversity of functions including EF (Miller, Lee, & Lumeng, 2015). Elevated concentrations of leptin are associated with poorer EF (Smith et al., 2019) even after accounting for other relevant metabolic and inflammatory markers (Van Dyken & Lacoste, 2018). Leptin has been proposed as a potential mediator between elevated Body Mass Index (BMI) and EF (Smith et al., 2019). Furthermore, exercise training reduces leptin concentrations (Tehfe, Elkhansa, Fu, & Tehfe, 2021) and individuals with obesity who regularly exercise showed improved performance on a visuospatial attention task, which was inversely correlated with leptin concentrations (Tsai et al., 2019), suggesting that the potential mechanism of neurocognitive facilitation by regular exercise could be a reduction in serum leptin concentrations. Ghrelin is a circulating peptide hormone mainly secreted from the stomach. Acute bouts of high intensity exercise transiently supress serum acylated ghrelin concentrations (King, Wasse, Stensel, & Nimmo, 2013), and exogenous ghrelin administration enhances the hedonic response to palatable foods (Goldstone et al., 2014). Serum ghrelin concentrations may be a biomarker of EF and have been shown to be a significant predictor of EF impairment in type 2 diabetes mellitus patients (Chen et al., 2017) and patients with bipolar disorders (Chen et al., 2021). In a study in bariatric surgery patients, lower leptin and higher ghrelin concentrations were linked to improved attention and EF at 12-month follow-up after surgery (Alosco et al., 2015). Insulin, a major metabolic hormone, is released by beta cells in the pancreatic islets of Langerhans and involved in glucose homeostasis and appetite. Exercise training improves insulin sensitivity by upregulation of insulin transporters in insulin-dependent cells and improvement of insulin signal transduction (Yaribeygi, Atkin, Simental-Mendia, & Sahebkar, 2019). In older adults, insulin resistance is linked to poorer EF (Abbatecola et al., 2004). In patients with alcohol dependence, lower blood insulin concentrations were linked to worse EF (Han, Bae, Won, Lim, & Kim, 2015). Compared to placebo groups, intranasal administration of insulin in individuals with bipolar disorder showed a significant improvement in EF (McIntyre et al., 2012). GLP-1 receptor agonists (GLP-1 RAs) appear to influence cellular pathways involved in neuroinflammation, neuroplasticity. According to mounting data, GLP-1 RAs, such as Liraglutide, have been shown in preclinical re Meanwhile, Liraglutide has also been demonstrated to improve EF and memory in individuals with major depressive and bipolar disorder (Mansur et al., 2017). These lines of evidence suggest that there are interactions between exercise, peripheral endocrine signals (many of which are associated with energy balance status) and executive function. search to improve various cognitive domains, including EF, spatial learning and memory, and recognition memory (Flintoff, Kesby, Siskind, & Burne, 2021). Liraglutide was shown to be effective in reversing deficits in amphetamine-induced EF, hyperlocomotion, and spatial learning and memory deficits (Chaves et al., 2020).
Many human and animal studies have found that hormones involved in energy homeostasis may be effective in the treatment of drug addiction. Exogenous leptin injection in rats reduces dopaminergic activity in the VTA, reducing cocaine's rewarding effects through manipulating leptin signalling (Shen, Jiang, Liu, Wang, & Ma, 2016), which may contribute to the development of addiction treatment strategies. The majority of preclinical studies suggest that ghrelin receptor antagonists reduce alcohol consumption (Lee et al., 2020), whereas ghrelin intravenous infusion increases alcohol craving in humans when compared to a placebo administration (Leggio et al., 2014). The use of intranasal insulin administration to noninvasively deliver and target insulin to the brain could lead to improvements in the treatment of addiction (Kashyap, Hanson, & Frey, 2020). Similarly, GLP-1 reduces the reinforcing effects of various drugs (Reddy, Stanwood, & Galli, 2014), which have been suggested as an addiction treatment (Sorensen et al., 2015). Taken together the evidence suggests that there is overlap in motivational and consummatory reward pathways and because of this, energy homeostasis hormones or their modulation by exercise training, could be used to produce beneficial effects on other non-nutritional consummatory behaviours. Metabolic hormones involved in energy homeostasis, and can be modulated by physical activity, improve EF which could be useful for treatment of drug addiction.

Executive function, food choice and exercisevicious circles and virtuous loops
Diet is a powerful environmental risk factor for obesity and metabolic disorders but similarly implicated in cognitive decline and neurodevelopmental diseases (Flores-Dorantes, Diaz-Lopez, & Gutierrez-Aguilar, 2020). In children and adolescents, a systematic review found there is a positive relationship between consumption of healthier foods (e.g. whole grains, fish, fruits and/or vegetables) and EF, whereas less-healthy snack foods, sugar-sweetened beverages, and red/processed meats were inversely related to executive function (Cohen, Gorski, Gruber, Kurdziel, & Rimm, 2016). The authors concluded that changes in brain structure and function, particularly in frontal cortex regions involved in synapsis, neurogenesis, myelination, and glucose management, may result from unhealthy dietary choices (Cohen et al., 2016). Davidson and colleagues (Davidson & Stevenson, 2022) propose a 'vicious circle' model to describe the relationship between poor diet and impaired EF. They show that a high-fat, high-sugar diet causes damage to the hippocampus and decreases hippocampal-dependent learning and memory in rats. They hypothesise that in humans 'Western-style' diets may interfere with hippocampal-dependent processes of appetite such as interoception of hunger and satiety cues and a reduced ability to inhibit memories of food reward outcomes, leading to overconsumption, obesity and further cognitive decline (Stevenson et al., 2020). This implies that food reward could be involved as both a cause and consequence of impaired EF. As reflective reward-based decision-making in appetite behaviours can be effortful (as described above), reduced EF would leave individuals, particularly those with obesity, more susceptible to reactive process influencing food choices.
A similar bi-directional relationship has been proposed between exercise and EF (Audiffren & Andre, 2019). They argue that a person who performs exercise on a regular basis creates a virtuous loop that improves components of EF (working memory, inhibitory control, self-regulation) leading to a reciprocal facilitation of exercise adherence, maintenance and habit formation. In a meta-analysis, Ludyga, Gerber, Puhse, Looser, and Kamijo (2020) found a slight increase in cognitive performance across age groups following long-term exercise, and advises practitioners on how to optimise this advantage. All forms of exercise appear to be beneficial, but when compared to endurance, resistance, and mixed varieties, coordinative exercise offers more noticeable cognitive benefits. The beneficial effects of exercise cannot be boosted simply by increasing the overall dose (Ludyga et al., 2020). Scientific data in recent years has accumulated to suggest that exercise can produce adjustments in the programming of genes (epigenomic manifestations). These upregulate synaptic and cognitive plasticity, which is critical for the prevention and treatment of neurological and psychiatric disorders (Fernandes, Arida, & Gomez-Pinilla, 2017). Hypothetical mechanisms by which exercise could therefore help to reduce reward-driven food intake is via improvement in EF and self-regulation (Andrade et al., 2010) or alternatively by shifting the balance of the aggregate reward outcome away from food towards other goals i.e. fulfilment of psychological needs through non-food-based behaviours. The extent to which exercise-EF relationships affect other reward-driven behaviours such as drug taking are also being investigated.

Can exercise facilitate the transfer of reward between drugs and food?
Understanding of the processes involved in exercise and food reward has progressed thanks to the concepts and methodologies derived from neurobiological theories of drug addiction (Finlayson, King, & Blundell, 2007;Stice, Figlewicz, Gosnell, Levine, & Pratt, 2013). Of particular relevance are the constructs of liking and wanting thought to serve as a basis for learning behaviours that lead to the acquisition of energy and essential nutrients (Berridge & Kringelbach, 2008;Berridge & Robinson, 2003). The incentive sensitisation theory (Robinson & Berridge, 1993) describes how intense stimulation from drugs (to an intensity that far exceeds any food) can cause dysfunction of the reward system, including the sensitisation of mesolimbic dopamine neurons involved in wanting. Therefore a core manifestation of drug addiction is the disproportionate focus on drug seeking behaviours at the expense of those unrelated to drugs (Hyman, Malenka, & Nestler, 2006). That is, the aggregate evaluative outcome of reward-aversion associations (drawing on input from affective and cognitive networks) favours drug-related rewards compared to non-drug rewards including food (Diekhof, Falkai, & Gruber, 2008;Feltenstein & See, 2010). Based on this concept, two key substance-use disorder treatment goals are to 1) reduce behaviours that maintain drug-use and 2) increase adaptive non-drug related rewarding behaviours such as eating food (Banks, 2017). Therefore, interventions are needed that not only address incentive sensitisation and increase executive function and inhibitory control in response to drug cues, but also reinstate reward-seeking and consummation of food and non-drug stimuli during withdrawal.
Alcohol, nicotine and some drugs of abuse can affect appetite (Abel, 1975;Grunberg, 1982;Hetherington, Cameron, Wallis, & Pirie, 2001). One study found that higher BMI was associated with lower probability of illegal drug use, supporting the hypothesis that food and drugs may compete for the same brain reward circuit (Blueml et al., 2012). At the same time, palatable food activates the brain reward circuit through rapid sensory input and absorption after ingestion (such as increasing the concentration of glucose in the blood and brain), while drugs activate these same pathways mainly through direct pharmacological effects on the reward circuit (Volkow & Wise, 2005). This leads to the hypothesis that when people with drug-dependence cannot obtain their drug of choice and alternative drugs are unavailable, they may transfer their reward-seeking towards other rewarding stimuli including palatable foods leading to passive overeating and increased body weight over time. This reduced sensitivity to food rewards in drug dependence is not for all categories of food. Animal experiments found that the preference of rats for sweet water exceeded that of cocaine, even those with a history of drug abuse (Cantin et al., 2010), indicating that sweets have a strong reward effect, and the strong desire and preference for sugary food and the anhedonia in eating other kinds of food are separated (Dileone, Taylor, & Picciotto, 2012). The same phenomenon has been found in human studies. Among people with opioid-dependence, previous studies have reported a strong desire, preference and intake of sweet foods during withdrawal (Emilia, Bartosz, Agnieszka, Dorota, & Magdalena, 2005;Morabia et al., 2010;Nolan & Scagnelli, 2009). Substantial changes are also observed in the consumption habits of patients with alcohol-dependence in the early stage of abstinence, especially in mental stimulants such as coffee, cigarettes, chocolate and other sweets. Higher alcohol cravings were related to the desire for these substances, which supports the idea that in the early stages of abstinence, reward-seeking can transfer from alcohol to other substances, especially sweets (Junghanns, Veltrup, & Wetterling, 2000).
The deleterious impact of drug use on EF has been investigated in relation to physical activity. In a cross-sectional study (Weinstock et al., 2021), examined 35,548 non-institutionalised US adults who had provided measures of attention and EF using a 12-item self-report EF Index (Aharonovich, Shmulewitz, Wall, Grant, & Hasin, 2017) along with physical activity using the International Physical Activity Questionnaire (short form) (Lee, Macfarlane, Lam, & Stewart, 2011) and substance use in the past year by the Alcohol Use Disorder and Associated Disabilities Interview Schedule-5 (Grant et al., 2015). The findings showed that frequent drug users who engage in light-to-moderate physical activity may have poorer cognitive performance than those who engage in vigorous physical activity. Although the study was strengthened by a large sample size, the self-report nature of the variables in their analyses may have affected the outcomes. Furthermore, the correlational approach and large number of uncorrected comparisons conducted requires some caution in interpretation. Impairments to EF have been found in individuals with Methamphetamine (MA)-dependence compared with non-MA users (Panenka et al., 2013). Additionally, when assessed early in the abstinence period compared to that prior to withdrawal, people with MA-dependence performed better on EF (Volkow et al., 2001). Likewise, long-term MA abstinence is associated with improvement in EF (Salo et al., 2009). Based on the concept of reward transfer and evidence that exercise improves EF in populations with cognitive impairment, a program was developed to examine the impact of exercise on food and drug-related reward in men and women with MA-dependence. In the first examination of exercise on food reward in this population, aerobic exercise was performed using a bicycle ergometer while keeping participants' heart rate at one of two exercise intensities: moderate 65-75% or high ≥85% of their estimated maximum heart rate. A control group (no exercise) from the same population were required to read about drug abuse treatments, including exercise-and fitness-related information, in a quiet room for 35 min. Food reward was measured using the Leeds Food Preference Questionnaire, adapted for Chinese adults (Zhou et al., 2023). We found that both moderate and high intensity acute aerobic exercise increased food preference  and reward-related neuronal responses (Wang et al., 2019) for high-fat savoury foods, and stimulated appetite in men with MA-dependence. Furthermore, among women with MA-dependence, findings showed that acute moderate-intensity dance and aerobic exercise may equally decrease cue-induced MA craving and increase food reward and appetite responses, with an increase in wanting and decrease in inhibition-related neuronal responses for high-fat savoury food and decreased wanting for high-fat sweet food (Zhou et al., 2021). These findings support the use of aerobic exercise as a potential adjunct therapy in MA-dependence in that it may realign reward circuits towards non-drug based stimuli via improvements in EF during an integrated program of management for people with addiction.

Future directions
In view of the overlap of food, exercise and recreational drugs on reward and the recruitment of brain networks involved in energy homeostasis, emotion and executive function, a holistic approach with common methodologies is needed to investigate therapeutic strategies designed to support the maintenance of physical activity, healthy food choice and possible extension to abstinence from drugs of abuse. The aggregate evaluative outcome of these different reactive reward-aversion associations is heavily informed by reflective cognitive processes which are themselves impacted by metabolic signals of homeostasis. Assessment of these three italicized domains may predict the ability of an individual to adopt and maintain a new healthy behaviour or to relapse into prior learned habits. If consensus can be reached on such a unifying framework, this conceptual model and its application to eating behaviour should be tested in future research. We encourage the investigation of long-term exercise and pharmacological combination therapies that target metabolic hormones (e.g. (Jensen et al., 2019)) and may promote neuroplasticity. Clinical trials are needed to evaluate the safety and efficacy of such programs to enhance EF and inhibit emotional reward-based decisions leading to drug-or unhealthy food seeking behaviours.

Conclusions
The impact of exercise on food reward may be better understood using a neurocognitive-energy balance framework that accounts for affective and cognitive processes involved in decision-making. EF and exercise are linked in a virtuous loop; while EF, unhealthy food choices and overconsumption form a vicious circle. When EF is impaired by metabolic dysregulation (such as occurs in obesity), entrained habits and cues associated with reactive pleasure-driven food seeking dominate with minimal reflective processing. Exercise may improve components of EF, in turn enabling food reward decisions to incorporate a greater composite of cognitive motives beyond immediate sensory or affective impact. Neurocognitive models of addiction also point to the possibility that impaired EF is involved in maintenance of addictive behaviours via direct pharmacological effects of drugs on the brain. Incentive sensitisation to drug-related cues and loss of reward motivation for food and other non-drug stimuli might also be re-balanced after improvements in EF with certain forms of exercise. Furthermore, metabolic hormones involved in energy homeostasis give some mechanistic insight into the impact of exercise on EF, which could be a useful tool for multicomponent interventions treating obesity, drug addiction and possibly other disorders involving impaired decision-making. The overlap of food and drug reward circuitry along with preclinical and clinical observations suggests that reward-seeking can transfer within and between categories of reward. This has implications for the use of exercise as adjunct to cognitive-behavioural strategies during drug abstinence or dieting to control body weight. Understanding how tensions between multiple competing aspects of reward related to habitual versus new behaviours arise may be an important area for future research seeking to understand why people disengage from interventions to improve their health. Identifying and aligning the positive aspects to maximise aggregate pleasure and minimise aggregate discomfort could be a strategy to reinforce motivation to exercise and offset the motivation to seek unhealthy rewards like energy-dense food or recreational drugs.

Ethical statement
No ethical approval was required for this invited commentary.