Effectiveness of Physical Activity Related to Food Addiction in Obese and Overweight Healthy Adults: A Systematic Review

Obesity is one of the main causes of morbidity and mortality, and food addiction has been considered one of its phenotypes. The study aims to determine the effectiveness of physical activity (PA), exercise, and sedentary behaviour management in attenuating food addiction in obese and overweight healthy adults. A search strategy was undertaken using MEDLINE, CENTRAL, PsycINFO, and CINAHL until April 2019. A total of 295 studies were found from which seven studies, after duplicates were removed, were included. The acute effect of PA of moderate intensity had an effect on attentional bias, craving, and sweet taste perception, although no difference was observed between genders and body mass index groups. No effect was observed on the hedonic and reward value. The acute effects of PA of moderate intensity may be a useful tool to control the levels of food addiction. More research is needed to clarify if this effect also occurs during chronic application, at different doses and types of PA, and for all populations.


INTRODUCTION
In the literature, other paradigms are used to study the brain's reward system in the context of an eating-related addictive behaviour, which is correlated with reward elements. The attentional bias (AB) is a selective focus on personal relevant information over neutral information, and has been associated with levels of craving for various substances (Field, Mogg, & Bradley, 2005;Field, Mogg, Zetteler, & Bradley, 2004;Field, Munafo, & Franken, 2009), thus values of AB have been demonstrated as a useful instrument for predicting "wanting" (Finlayson & Dalton, 2012). The positive pleasure against sweet (i.e., sugar) or creamy (high-fat) has become a guarantee to energy intake (H. R. Berthoud, 2006) and that is the base for how preferences over sweet and creamy are related to obesity, even more, if both are taken together (Donaldson, Bennett, Baic, & Melichar, 2009). Finally, evidence from neural responses to food cues support the main idea of hyperactivation in specific brain areas to food cues in obese people (Val-laillet et al., 2015;Volkow, Wang, & Baler, 2012).
Physical activity (PA) and exercise performed over long periods reduce body weight (Donnelly et al., 2003;Jakicic, Marcus, Lang, Janney, & Forest, 2008;Ka, Hc, Rourke, & C, 2009). The benefits are not limited to weight loss; moderate PA also reduces the risk of developing some chronic diseases (González-Muniesa et al., 2017). However, the most important aims of this review are the effect of PA on the energy intake, by modulating the homeostatic and hedonic mechanisms (Beaulieu, Hopkins, Blundell, & Finlayson, 2016;Shook et al., 2015). According to the literature, sedentary behaviour promotes body weight gain by appetite deregulation and low energy expenditure (J E Blundell, 2011).
Indeed, FI through exercise and PA continues to be a controversial topic (Bilski, 2009).
To date, no systematic review study has been found that focuses its objectives on the effects of PA, exercise, and SB management of eating-related addictive behaviour outcomes, linking FA to HyPF in the overweight or obese population. This review aims to disclose relevant information on the conservative treatment of food addiction.
The purpose of the present study is to determine the effectiveness of the physical activity, exercise, and sedentary behaviour management in attenuating the food addiction of HyPF in obese and overweight healthy adults.
Secondarily, the aim is to determine the effectiveness influenced by the intensity or type of PA, and whether the difference in the body composition or gender could alter the power of the intervention.

MATERIAL AND METHODS
This systematic review was guided according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) criteria (Moher, Liberati, Tetzlaff, Altman, & Group, 2009). This review has not focused on lean people, so the outcomes of this group have been ignored.
However, since different BMI groups did not show statistically significant differences, all the data were combined into one main group.

Data sources
The electronic databases that were used for the main purpose of this research included: MEDLINE, The Cochrane Central Register of Controlled Trials (CENTRAL), CINAHL, and PsycINFO. In addition, snowballing was used by searching reference lists of retrieved articles and published reviews on the topic.

Literature search strategy
The search strategy was designed to capture articles with research relevant to the following topics in the field: "Obesity/overweight", "Physical Activity", "Addiction", and "Hyperpalatable Food". The results were filtered to include papers that have studied humans, including adult (18-65 years) participant. Searches were limited to English and Spanish. No publication date restrictions were imposed. (see detailed research strategy of MEDLINE search strategy, the search was adapted for the other electronic databases; Table 1).

Study selection
Once the criteria were established, the search was conducted by two reviewers (DPP and CBN) to identify titles and abstracts of the references obtained as a result of the search strategies previously exposed to exclude irrelevant reports. Then, after the removal of duplicates, the full-text articles of the remaining references were screened by DPP and CBN to identify studies for inclusion and classify and record reasons for the exclusion of ineligible studies. Both reviewers resolved different opinions for eligibility by discussion, and when the review authors did not agree, a third reviewer (FRC) made an arbitrary decision. The selection process has been reported in a PRISMA flow diagram (Figure 1).

Data extraction
Data was extracted from the admitted articles, and relevant information according to the established inclusion criteria was recorded on a data extraction form that has been developed for this review (see in appendices section; Table 2). Data extracted included the following items: authors, participants (including sample size, sex, age, and BMI), baseline status (including PA status, abstain status, and FA status), settings (including study design and details of intervention and control group), relevant outcomes, measurement tools, and adverse effects.

Risk of bias assessment
The risk of bias was assessed using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins, 2011). The risk of bias was graded for each domain as high, low, or unclear. Unclear means there is a risk of bias, but not enough information is available to determine whether an important risk of bias is present, or there is a lack of clarity whether an identified problem will introduce bias. The high risk was determined when there was at least one important risk of bias for a domain.

RESULTS
Two-hundred and ninety-five potentially relevant articles were identified from the database search and other sources. After the removal of duplicates, a total of 258 (87.5%) were screened on the basis of their titles and abstracts. The full text was retrieved for 25 (8.5%, from the total) articles, 18 were excluded for the following reasons: results not posted (n = 3), not the population of interest (n = 3), not the intervention of physical activity (n = 6), not the study design of interest (n = 6). Finally, seven articles were included in the qualitative synthesis. Figure 1 illustrates the PRISMA flow diagram of the data collection process.

Risk of bias
The risk of bias from the included studies was presented descriptively (see figure 3). All the following items showed an unclear risk of bias: 1) Random sequence generation; 2) Allocation concealment; 3) Blinding of participants and personnel; 4) Blinding of outcome assessment; 5) Selective outcome reporting. Finally, the item "Selective outcome reporting" had a low risk of bias, and no other bias were detected. According the nature of studies, the main bias was the information of allocation concealment and random sequence which was missed in the main of studies. Also, the blinding of participants and personnel were not possible too due to knowledge of which intervention a participant received. However, the selective outcome reporting had a low risk of bias in all studies except Alkahtani (Alkahtani, Byrne, Hills, & King, 2014b).

Participants
The sample size for the included studies ranged from 12 to 58 (n = 137). To assess obesity and overweight, all the included studies used BMI. No participants were underweight. Three articles conducted the study with both genders (Ledochowski, Ruedl, Taylor, & Kopp, 2015;Martins et al., 2015), the other two articles included just females (Oh & Taylor, 2013)  Three articles (Martins et al., 2015;Oh & Taylor, 2013;Westerterp-Plantenga et al., 1997) classified their participants as inactive, untrained or sedentary. Only one paper did not report any information on their PA (Ledochowski et al., 2015). Finally, two studies (Oh & Taylor, 2013;Westerterp-Plantenga et al., 1997) included different groups according to their BMI, and both reported no significant differences between groups; therefore, all the data were combined.

Authors Participants Randomization
Intervention group versus control

Effect adverse
Oh, H. and Taylor

Interventions
All of studies founded focused their interventions on the acute effects of a single bout of different types and doses of exercise. There wasn't any article that analysed the effect of exercise or physical activity management follow up on time. All studies (Alkahtani et al., 2014b(Alkahtani et al., , 2014aLedochowski et al., 2015;Martins et al., 2017Martins et al., , 2015Oh & Taylor, 2013;Westerterp-Plantenga et al., 1997) conducted their intervention at moderate intensity, and four studies (Alkahtani et al., 2014b, 2014a, Martins et al., 2017 included two conditions based on high intensity (HIIT). The modality of physical activity were brisk walk in two studies (Ledochowski et al., 2015;Oh & Taylor, 2013) and cycling in the rest. According of variable used in the studies, the intensity of HIIT were between 85-90% of HRmax (until 250kcal) or 85-90% VO2max. On the other hand, MIIT were between 70% HRmax or 40% VO2max, also 20% below and 20% above max fat oxidation were measured to determinate the intensity. The duration and frequency varied between studies.

Outcomes
The included articles showed different eating-related addictive behaviour outcomes linked to  Table 2 summarises the information related to all the outcomes included.

Acute effects of Physical Activity on Food Addiction
Six included studies were focused on the effect of a single bout of exercise and one of them studied four week training intervention (Alkahtani et al., 2014b). None of the included studies reported or measured sedentary behaviour management. No significant differences between groups of intervention were reported; therefore, all the data were taken together, as described above.
The Leeds Food Preference Questionnaire were used to measure the hedonic value of food using a 100-mm visual analogue scale (VAS) (Martins et al., 2017(Martins et al., , 2015 and the reward value. Two studies with 32 participants, 25 males and 7 females, found that there were no significant differences in the hedonic value per conditions, groups, and time to different types of drinks and concentrations of sweet taste (Westerterp-Plantenga et al., 1997, Martins et al. 2017), buffet styles, and the same for the reward value of high-fat foods relative to low-fat for all intensities (Martins et al., 2015).
A modified visual dot-probe task with chocolate-related pictures were used in one study (Oh & Taylor, 2013) of 58 female participants in order to measure AB (Oh & Taylor, 2013). Each trial began with a central black fixation cross on a white background for 1000 ms, followed by a pair of images, which were presented for either 200 ms or 1000 ms, to capture initial The variable "linking" and "wanting" were assessed by 4 studies (Alkahtani et al., 2014a, 2014b, Martins et al., 2017. Only in the Alkahtani study there was a significant effect on explicit liking for all food categories except high fats nonsweet (Alkahtani et al., 2014a).

DISCUSSION
This systematic review tried to determine the influence of exercise doses in relation to eatingrelated addictive behaviour, craving, or neural activation to HyPF linked to FA in overweight or obese people.
Obese are more prone to show preferences for sweet and fatty flavours (Donaldson et al., 2009) and they also show greater neuronal activation to food cues (Val-laillet et al., 2015;Volkow et al., 2012). Two studies (Oh & Taylor, 2013;Westerterp-Plantenga et al., 1997) that included different groups according to their BMI did not show differences in the results to the STP (Westerterp-Plantenga et al., 1997), the hedonic value (Westerterp-Plantenga et al., 1997), the AB (Oh & Taylor, 2013), or neither of the craving values (Oh & Taylor, 2013).

Castellanos et al. (Castellanos et al., 2009) did not show a difference in AB between obese
and normal-weight subjects who were in a state of hunger. In contrast, when the participants had eaten, the obese showed a higher AB value. This could be the reason why the subjects did not show differences, since none of the two studies submitted the subjects to a period of abstinence, with the exception of one study where the subjects were restrained only to one type of food (chocolate snacks) (Oh & Taylor, 2013). The inclusion criteria for this review may have also influenced the lack of differences between the BMI groups. In addition, these results allow hypothesising whether the cause of FA is actually obesity per se, or the baseline level of the PA, where all the groups of both studies (Oh & Taylor, 2013;Westerterp-Plantenga et al., 1997) were considered inactive. The evidence is still ambiguous and cannot reach a firm conclusion. Moreover, two of the studies included subjects of both genders (Ledochowski et al., 2015;Martins et al., 2015), but none analysed the differences between them and rather included all the subjects of the same data which could give a wrong view of the results considering that men and women have different behaviours and responses related to craving (Hallam et al., 2016) and food addiction (Pursey, Stanwell, Gearhardt, Collins, & Burrows, 2014), with women being most affected. This might be an effect that could be correlated with higher levels of overweight in this gender (World Health Organization, 2018).
The effectiveness of different intensities of exercise related to different mechanisms in FI behaviour is not clear; studies have been contradictory with those who have reported significant differences in energy intake for obese adolescents (Thivel, Isacco, Montaurier, Boirie, & Duche, 2012) and satiety for overweight men (Rosenkilde et al., 2013), and those who have not been able to observe any difference in the effect on appetite, "liking" and "wanting" in overweight males (Alkahtani et al., 2014a) with different intensities. This lack of clarity is also represented in the articles selected, where the only study (Martins et al., 2015) that compared several intensities with a rest control group did not observe any effect on the reward value assessed with the Leeds Food Preference Questionnaire (LFPQ). Furthermore, this study also saw no effect of exercise on hunger and satiety through all the conditions, an the theory of self-determination (SDT) (Heather & Geoffrey C, 2012). This theory sustains that individuals who exercise for self-determined or autonomous reasons did not favour the consumption of HyPF than those who did it for more controlled reasons (Beer, Dimmock, Jackson, & Guelfi, 2017;West, Guelfi, Dimmock, & Jackson, 2017). Therefore, self-selection of the intensities, participation in activities that subjects enjoy, and also the assessment of the acute affective responses to interventions could be a good idea to keep in mind. This hypothesis could explain why subjects who underwent sessions of the same intensity (moderate intensity) and duration, but continuously or intermittently, showed a reduction in appetite and an increase in satiety for the second condition, compared to those who did it continuously (Holmstrup, Fairchild, Keslacy, Weinstock, & Kanaley, 2013), where it was less demanding for them, less stressful, and more enjoyable.
This study provides relevant information that seems to indicate that some eating-related addictive behavioural outcomes are influenced by several elements that take part in the term called food addiction, after a single bout of exercise, at moderate intensities in obese or overweight healthy subjects. One of the most studied cases has been the AB, where Oh and Taylor (Oh & Taylor, 2014) noted that after short sessions of moderate and high-intense exercise, there was a reduction in the initial AB in response to the images of cigarettes and HyPF, instead of only the high-intensity exercise having an effect on the maintained AB.
Then, any exercise would have an effect in reducing the attention to such images, but only the high-intensity exercises had an effect in maintaining the gaze and interest from drifting towards the salient images of cigarettes and snack food. In this review, we have included an article (Oh & Taylor, 2013) that corroborates part of the results, observing that moderateintensity exercises have a great effect in the IAB, and a lesser effect in MAB, although being significant with respect to the resting control group. Moreover, although not enough to reach a definitive conclusion, it seems to indicate that single bouts of exercise can have an effect in the first phases when those substances that cause craving are perceived. AB values have also been associated with craving in different studies in variate substances (Field et al., 2005(Field et al., , 2004(Field et al., , 2009). The studies included in this review provide relevant information that seem to indicate that the craving values for HyPF are reduced after a single bout of exercise at moderate intensity in obese or overweight subjects (Ledochowski et al., 2015;Oh & Taylor, 2013, Alkahtani et al., 2014a, results that would be in accordance with those presented on AB. This effect becomes more solid after observing how it can be seen against other substances, such as craving in cigarettes (Taylor, Ussher, & Faulkner, 2007;Thayer, Peters, & Takahashi, 1993) and alcohol (Ussher et al., 2004) with large investigations supporting them.
Several studies (Taylor & Oliver, 2009;Thayer et al., 1993) also observed how an exercise session reduced craving in non-obese subjects to certain types of food.
In a recent review (Coltell et al., 2019), the relationship between taste perception and body weight, BMI, and waist circumference in older patients was studied, indicating an inverse association. Only a single article (Westerterp-Plantenga et al., 1997) was found that analysed the perception of sweet taste after applying a moderate exercise session and observed an increase in the perception of sweet taste, compared to the control group at lower concentrations of sucrose in a simple solution, in a mixture, and in a sports drink after exercise compared to after rest, although this change did not occur at other concentrations.
These findings suggest that those subjects who do not feel the taste that much, need more doses to achieve the same pleasure (Donaldson et al., 2009).
The results found on the hedonic value (Westerterp-Plantenga et al., 1997) and the food reward (Martins et al., 2015), indicated no acute effect of a single bout of exercise on these outcomes. These findings are not in accordance with what has been described in the literature, where a reduction in the neural response was observed to visual stimuli in non-obese adults in principal cerebral areas, related to reward after a bout of exercise (Evero, Hackett, Clark, Phelan, & Hagobian, 2012). It was also observed after the chronic application of exercise in overweight/obese subjects (Luo, O'Connor, Belcher, & Page, 2018). In another study (Luo et al., 2018), a correlation between greater time in moderate to vigorous physical activity (MVPA) and lower in SB time and decreased brain responses to HyPF was observed in obese and normal-weight adult subjects. The lack of results may lie in the type of population used for the studies, the interventions, or more probably, the difference in the evaluation methods between the studies.
As it has been discussed, the mechanisms behind the effect of PA and exercise on FA are still not well understood; therefore, generic aspects, such as energy consumption or macronutrients, appetite, hunger, satiety and satiation, combined with more specific elements to homeostatic components (i.e., hormones), and those for food addiction (i.e., craving, homeostatic hunger, "liking", "wanting", taste perception, AB and food reward), even neuroimaging studies, could help to identify those aspects where it is most useful.

Limitations
The systematic review is under a set of limitations due to the structure of the methodology itself. All included studies were based on a cross-over design, this methodology is considered as a type of randomised controlled trial, with particular characteristics where the subjects receive several treatments consecutively, this limits long-term conditions analysis, and the effects could be transferred to the following interventions; nevertheless, it allows to obtain a larger sample (Sibbald & Roberts, 1998).

Implications for current practice
The PA and exercise have been described as essential components for weight loss and maintenance programs (Donnelly et al., 2003;Jakicic et al., 2008;Ka et al., 2009;Murray et al., 2014), especially in combination with healthy eating habits (González-Muniesa et al., 2017). The current knowledge also could associate a partial effect in the reduction of food intake (Beaulieu et al., 2016;J E Blundell, 2011;Shook et al., 2015), and as described before, could be by several elements of eating-related addictive behaviour. This knowledge gives those professionals, who deal with patients suffering from obesity/overweight, a useful conservative tool to help them.

Conclusion
The literature seems to indicate that the acute effects of PA at moderate intensity may be useful as a complement to other therapies to help control the levels of addiction to HyPF.
Further research is needed to determine the effect of chronic interventions, different doses and types of PA applications. It is also unknown if only obese/overweight subjects could benefit from this effect since there does not seem to be much difference between body composition, which means that it could serve as a preventive treatment. The unknown pathways and mechanisms on the topic open many lines to research.

Declaration of interest
The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.