Introduction

Evidence based neurocognitive models of addiction propose that addiction-related behaviors are the result of an imbalance of three neural systems: an impulsive neural system that promotes habitual and salient behaviors, interoceptive processes that are involved in uncertain risk and reward, and a reflective neural system for inhibitory control and decision-making [1][2]. Decision-making entails the cognitive process of making a choice after reflection on the consequences of that choice, and it is a key component of addiction development and maintenance in both substance use disorders (SUD) and behavioral addictions such as gambling disorder (GD). The assessment of decision-making is usually conducted through the Iowa Gambling Task (IGT), which simulates real life decision making strategies by factoring uncertainty, reward and punishment. The IGT is a relatively complex task as it is difficult to deconstruct into different cognitive constructs, however it measures decision-making with a high ecological validity[3] in a wide range of clinical and non-clinical groups[4]. Specifically, impairments in this task have been demonstrated in patients with ventromedial prefrontal cortex lesions (VMPC;[5]) and in different psychopathological conditions including addictions and eating disorders[6–8]. Distinctively, while clinical individuals with decision making impairment fail to learn the contingencies and prefer the choices that lead to higher long term losses, healthy individuals present a gradual learning across the IGT [4][5]. According to the Somatic Marker Hypothesis, individuals who perform poorly on the IGT have weaker physiological cues to guide risky choices and present what is referred to as “myopia for the future” [9].

Complementary, obesity (OB) is an increasing worldwide problem that shares similar patterns to addictions[6]. Individuals with obesity frequently decide to overeat despite being aware of its negative long-term health consequences and they usually put extra but unsuccessful efforts into controlling their eating behaviour. Neuropsychological studies support the hypothesis of an alteration on inhibitory control, emotion regulation and the executive function circuit in which one of the core cognitive traits appears to be decision making[10]. Accordingly, recent data shows that individuals with obesity are characterized by the tendency to engage in decisions that support a positive short-term reword even if it results in long term negative outcomes[7]. Furthermore when assessed by the IGT, individuals with obesity present significant decision making impairments in the overall task performance as well as in learning across the task [7,11,12]. Likewise GD and SUD individuals present a similar decision making pattern, being the overall task performance and learning across the task impaired[8]. For instance, a recent study conducted with GD participants describes a strong preference for choices featuring high rewards rather than higher losses during the IGT. The authors of this study suggest that this might reflect an hypersensitivity of their reward systems[13]. Additionally, GD and SUD are usually associated; being the decision-making patterns worse when both diagnoses are present[14]. Similarly, OB and GD display neurocognitive and clinical associations. For instance OB is associated with decision making and sustained attention impairments in gamblers, along with greater monetary losses due to gambling [15].

Direct comparisons of decision making profiles among SUD, GD and OB groups have yet to be conducted. However, direct comparisons can provide a valuable insight into the similarities and/or singularities of different addictive related behaviours.

Methods

Results

Table 1 includes the results of the χ² tests and the ANOVA to compare the sociodemographic variables between groups. Descriptive variables of our sample (Table 1) show the inherent differences in sex, age and education among clinical groups (GD, SUD and OB groups). Accordingly and for the statistical aims of this study we controlled for the age and education level variables and used them as covariates in the ANOVA when comparing IGT scores.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Sociodemographic descriptive.

https://doi.org/10.1371/journal.pone.0163901.t001

Table 2 includes the ANOVA model comparing the cognitive mean scores measured through the IGT, adjusted by the covariates age and years of education. The first step of the ANOVA procedure tested the interaction parameter ‘group-by-sex’. The absence of significant moderation effects indicates no sex significant differences in the diagnostic subtypes and it also indicates that differences between sexes were statistically equal among the diagnostic subtypes. Consequently, the interaction parameter term was excluded from the modeling and the main effects for the factors group and sex were obtained and interpreted. Regarding the group factor analysis, differences among groups emerged in all NET scores but in NET-1. Additionally, HC participants achieved higher mean scores in NET-total, NET-3 and NET-4 when compared to the other clinical conditions (GD, OB and SUD). The HC participants also achieved a higher mean score in NET-5 when compared to GD and SUD individuals and, GD participants obtained a lower mean score compared to OB and SUD individuals. Concerning NET-2, the OB group achieved the lowest mean score and it was statistically different when compared to the other clinical conditions. Finally, with reference to the sex factor analysis, differences between men and women emerged in all IGT scores except in NET-1 and NET-2 ones. Specifically, men achieved significantly higher means than women. Fig 1, displays the comparison of the mean scores in the cognitive learning across the task scales between groups.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Mean cognitive measures in learning across the task between groups.

https://doi.org/10.1371/journal.pone.0163901.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Comparison of the cognitive mean scores between diagnostic subtypes: ANOVA.

https://doi.org/10.1371/journal.pone.0163901.t002

Discussion

This study compares decision making patterns when assessed by the IGT in three different clinical samples (SUD, OB and GD) and a HC. Results display impairments in the overall performance and in the task learning process in all clinical samples when individually compared to the HC.

Results are in agreement with previous studies reporting impaired decision making in SUD and GD individuals [13,14] as well as in individuals with OB[10,12] when independently assessed and compared to a HC group. Additionally, each of the three clinical groups of our sample present more impaired learning across the task than the HC group. Data shows how the HC group presents a preference for the short term wins but after their first set of selections they progressively learn to choose the decks that will lead to bigger long term wins. This is an expected pattern in healthy individuals[4]. In line with previous studies, the decision-making pattern of our HC sample cannot be extended to the clinical groups where learning starts later and/or with a more erratic progress. For instance, OB individuals maintain negative scores until the third set (i.e.: NET_3) where learning starts but to a lesser extent than in the HC group. Previous studies have also reported similar impairments in OB[7,18]. The SUD participants show an earlier learning across the task but improvement progression is slower when compared to the HC ones. The GD individuals display more erratic or random choices. The learning effect is quite small and ends with a score similar to the initial one. Accordingly, previous studies demonstrate how SUD individuals tend to present an adaptive shift in decision-making performance towards the end of the task[19,20] whereas GD individuals do not[21,22]. Therefore, the decision-making impairment of individuals with SUD is probably more associated to a learning delay strategy rather than an inability to learn from the task and this pattern seems not to be extended to GD participants. Finally, no significant differences are found in the overall performance among the clinical groups. All groups achieve low but still positive scores. Results can reflect that to some extent the three clinical samples respond to the punishment cues (although probably to a lesser extent than the HC group) and also that their deficit could be more due to difficulties in reward processing.

Results also display no sex influence among groups. However, some differences are observed when comparing the IGT performance between men and women in the whole sample. Specifically, men present a better performance on the IGT. Previous studies report similar results and suggest that men tend to have a better performance than women in general population[23] and in SUD[24]. Although sex factor was controlled in our study, it would be interesting to explore further in future studies to specifically explore sex differences in the studied samples.

Our data adds to the current literature more evidence of decision making patterns in SUD, OB and GD by comparing the three clinical samples. Importantly, in the analyses conducted none of the differences among the groups could be due to sex or other sociodemographic variables, which adds extra value to the findings reported here. Our study supports the hypothesis that OB shares specific cognitive and neurobiological patterns with SUD and GD which are known to suffer from impairments in dopaminergic pathways that regulate neural systems associated with reward sensitivity and incentive motivation[25]. Accordingly, food addiction has been associated with obesity in previous behavioral and neuroimaging studies[26] and the food intake management difficulties observed in some of the individuals with obesity could be associated with decision making impairments. The clinical implications of these results lie on the identified cognitive patterns that suggest the value of parallel therapeutic approaches among OB, SUD and GD individuals[27,28]. Importantly, the SUD and OB groups display some improvement in decision making; however, they are far from the HC group results (see Fig 1). Hence, cognitive stimulation protocols (through executive function working tools) could potentially benefit these individuals by enhancing their adaptive decision making strategies when treating their disorder. Additionally, some specific differences among the clinical groups could be considered for improving current treatments. For instance, OB individuals take longer to learn the relationship between decks but once they do, their behavior conforms to that of HC. Still, their learning pattern remains less adaptive than the HC one and it should be considered for treatment approaches. Specifically, results may suggest that OB individuals could probably benefit from an early treatment extra emphasis on gaining more adaptive decision making strategies. On the other hand, SUD tend to learn faster than OB individuals and more similar to HC but they display very little improvement when compared to HC. This could be due to more risky behaviors and thus more emphasis should be given to these difficulties. Finally, GD individuals display even more risky behaviors and move faster towards bigger rewards. These difficulties should be targeted in treatment. Future studies should test this hypothesis and further explore these domains in order to (1) help disentangle to what extend these differences are generalizable or constrained to our sample, (2) test the predictive or mediating role of decision making impairments in a study which compares the here studied clinical samples. Furthermore, future studies could further explore sex differences within the studied groups.

The present study has some limitations. Firstly, there are sociodemographic differences concerning education, sex and age across the groups which are representative of the studied disorders. This is an expected result when working with a consecutive clinical sample referred to GD, SUD and OB treatments. However, we have controlled these variables in all the statistical analysis. Although in this study we have paid specific attention to one of the most relevant neuropsychological factors (namely decision making), other cognitive functioning variables or intelligence measures not assessed here may better explain specific differences (more than commonalities) among the clinical groups. Finally, although participants did not present withdrawal symptoms or presented any life time mood-anxiety disorder or mental disorder that could hinder the assessment, the analyses conducted do not explicitly control for the participants’ hunger, anxiety or sadness feelings and these could also play a role in the participants’ performance on the task.

To our knowledge, we present the first study that compares decision making in substance and behavioral related addictions, obese individuals and healthy controls. Results show similar impairments in decision making in the three clinical groups. These impairments are statistically significant when compared to the healthy control group but not different among the three clinical groups. Finally, the clinical groups present significant difficulties in learning across the task when compared to the healthy controls and also some specific differences when comparing clinical groups.