The effects of the dopamine D₃ receptor antagonist GSK598809 on attentional bias to palatable food cues in overweight and obese subjects

Abstract

The mesolimbic dopamine system plays a critical role in the reinforcing effects of rewards. Evidence from pre-clinical studies suggests that D₃ receptor antagonists may attenuate the motivational impact of rewarding cues. In this study we examined the acute effects of the D₃ receptor antagonist GSK598809 on attentional bias to rewarding food cues in overweight to obese individuals (n=26, BMI mean=32.7±3.7, range 27–40 kg/m²) who reported binge and emotional eating. We also determined whether individual differences in restrained eating style modulated the effects of GSK598809 on attentional bias. The study utilized a randomized, double-blind, placebo-controlled cross-over design with each participant tested following acute administration of placebo and GSK598809 (175 mg). Attentional bias was assessed by the visual probe task and modified Stroop task using food-related words. Overall GSK598809 had no effects on attentional bias in either the visual probe or food Stroop tasks. However, the effect of GSK598809 on both visual probe and food Stroop attentional bias scores was inversely correlated with a measure of eating restraint allowing the identification of two subpopulations, low- and high-restrained eaters. Low-restrained eaters had a significant attentional bias towards food cues in both tasks under placebo, and this was attenuated by GSK598809. In contrast, high-restrained eaters showed no attentional bias to food cues following either placebo or GSK598809. These findings suggest that excessive attentional bias to food cues generated by individual differences in eating traits can be modulated by D₃ receptor antagonists, warranting further investigation with measures of eating behaviour and weight loss.

Introduction

Over-consumption of food is one of the leading factors contributing to the significant rise in the incidence of obesity (Hedley et al.2004). Obese individuals are more motivated to eat and find food (particularly palatable food) more reinforcing than do non-obese controls (Johnson, 1974; Saelens & Epstein, 1996). A number of cognitive theories may explain enhanced responsivity to food-related information in obese individuals, including maladaptive knowledge structures (i.e. schemas) (Williamson et al.2004), food preoccupation (Cox & Klinger, 2004), expectancy (Tiffany, 1990) and incentive salience (Berridge & Robinson, 1998; Berridge et al.2010). The latter theoretical view, which ascribes a key role to the mesolimbic dopaminergic system, has been particularly influential.

The mesolimbic dopaminergic system plays a critical role in the reinforcing effects of natural rewards like food (Bassareo and Di Chiara, 1997, 1999a, b; Wang et al.2004; Wise, 2006). Dopamine has also been shown to modulate hedonic and non-hedonic factors underlying motivation for food intake in animals (Swanson et al.1997; Szczypka et al.2001; Taber & Fibiger, 1997), while human imaging studies have linked changes in dopamine to motivational (Volkow et al.2002), restraint and emotionality processes regulating food intake (Volkow et al.2003). In particular, high-restrained eaters (i.e. individuals that intentionally restrict their intake of food and calories due to concern with weight and shape) show greater changes in dopamine in the dorsal striatum in response to food stimulation (Volkow et al.2003) and these authors suggested that higher restraint in this group could reflect a preventive adaptation strategy to minimize their exposure to salient food cues (Volkow et al.2003). This is consistent with theories that cognitive restraint may influence consumption of food (Herman & Mack, 1975) and with evidence that high-restrained eaters are more attuned to food cues in the environment, attempting to avoid these in order to control their body weight (Green & Rogers, 1993; Green et al.1997).

A key role of the dopamine system is to mediate the attribution of incentive salience to stimuli that are associated with reward (Berridge & Robinson, 1998; Berridge et al.2010). Attribution of high-incentive salience to a stimulus makes it attractive and ‘attention grabbing’ (Berridge et al.2010). Thus, attentional biases for food cues provide an objective cognitive index which has been linked to dopaminergic system functioning, and may underlie individual differences in the motivational salience of food cues and proneness to overeat and obesity. In support of this theoretical view linking attention biases and over-consumption of rewards, there is considerable evidence indicating that drug-dependent individuals show greater attentional bias to drug cues (Cox et al.2006; Field & Cox, 2008). Furthermore, consistent with dopaminergic models of overeating and obesity, a number of studies have shown enhanced attentional biases to food cues in obese and overweight individuals (Braet & Crombez, 2003; Castellanos et al.2009; Nijs et al.2010; Werthmann et al.in press). However, some have failed to find such biases (Phelan et al.2011; Pothos et al.2009) and variation in results across studies may arise as a function of methodological variables (Pothos et al.2009) and/or influences of eating style (Graham et al.2011). Although the relationship between body mass index (BMI) and attentional bias may not be a simple linear one (Pothos et al.2009), attentional biases for food cues have not only been associated with obesity (Braet & Crombez, 2003; Castellanos et al.2009; Nijs et al.2010), but have also been shown to predict change in BMI over a 1-yr period (Calitri et al.2010).

Dopamine receptors are potential therapeutic targets for modulating behavioural aspects of food consumption in conditions such as obesity. The dopamine D₃ receptor has recently been explored as potential drug target for the treatment of addiction (Heidbreder et al.2005). In animals, D₃ antagonists have been shown to reduce cue-controlled ‘drug seeking’, suggesting a selective role for D₃ receptors in the motivational impact of drug-related cues (Di Ciano et al.2003; Heidbreder et al.2005; Pilla et al.1999). Similarly, selective antagonism of D₃ receptors has been shown to reduce food intake and responses for food in obese Zucker rats (Thanos et al.2008). Clinical studies in human addict populations have also shown that mixed D₂/D₃ antagonists attenuate cue-induced craving (Gawin et al.1989) and reduce attentional biases on a modified Stroop task with drug-relevant stimuli (Ersche et al.2010; Franken et al.2004).

GSK598809 is a selective D₃ receptor antagonist in development for treatment of disorders of compulsive consumption including obesity (Searle et al.2010). In this study, we examined for the first time, the acute effects of GSK598809 on attentional bias to food-related cues in overweight to obese individuals who reported binge and emotional eating. This stratified sample was selected to ensure the obese population had characteristic overeating behaviours.

Attentional bias was tested with two commonly used tasks, the modified (food) Stroop task and the visual probe task. We also explored whether individual differences in eating style, namely restraint, modulated the effects of GSK598809 on attentional bias, given research noted earlier indicating that restraint influences dopamine response to food cues (Volkow et al.2003). The modified Stroop and visual probe (or dot probe) tasks have relative advantages and disadvantages as measures of attentional bias. The modified Stroop task provides a robust index of the personal motivational salience of aversive and appetitive stimuli (Cox et al.2006; Williams et al.1996) and has been shown to be predictive of weight gain in non-obese individuals (Calitri et al.2010). However, the modified Stroop index may reflect more than one underlying cognitive mechanism. For example, colour-naming interference effects may reflect enhanced attentional prioritizing of salient stimuli (attentional bias towards salient cues) and/or cognitive effort to suppress processing of them (‘avoidance’ of salient cues) (de Ruiter & Brosschot, 1994). In contrast, the visual probe task enables a more fine-grained analysis of the direction of attentional bias, because it allows for the differentiation between attention directed towards or away from a particular type of information (MacLeod et al.1986). The modified Stroop and the visual probe tasks also tap biases in different aspects of attentional processes, i.e. visual orienting vs. resolution of processing conflicts. Thus, it is informative to assess more than one type of attentional bias in order to evaluate its generality across different methodologies.

We hypothesized that, on both attentional tasks, GSK598809 would attenuate attentional bias to rewarding food-related cues. Specifically, we predicted that overweight/obese individuals would show an attentional bias towards food relative to control (non-food-related) cues in the placebo condition, but not in the GSK598809 condition. We also explored whether the effect of GSK598809 on attentional bias to food cues was modified by eating style, namely restrained eating.

Materials and methods

Participants

Twenty-six otherwise healthy, overweight and obese participants (15 males, 11 females) aged between 18 and 45 yr (mean age=35.1±7.1 yr) and BMI ⩾27 kg/m² (mean=32.7±3.7, range 27–40 kg/m² inclusive) were recruited for this study. All participants had no history of psychiatric disorders, neurological disorders or eating disorders (including binge-eating disorder), substance abuse and significant weight loss (or gain) (defined as a change of ⩾5% of their body weight in the past 30 days) based on a physical examination and a clinical and psychiatric interview by a medical physician. Additionally, participants were only included if they reported current history (over the last 2 wk) of binge-eating behaviour [minimum one episode per week as assessed by the YBOCS-BE questionnaire (Q6) (Goodman et al.1989, McElroy et al.2007)] and emotional eating behaviour (by attaining a score of ⩾3 in at least one of the questions of the emotional eating scale (Q3, Q6, Q10) of the Three-Factor Eating Questionnaire (TFEQ-R18) (de Lauzon et al.2004). All participants gave written informed consent for participation in the study, which was approved by the Hounslow and Hillingdon Research Ethics Committee, UK.

Procedure

The study utilized a randomized, counterbalanced, double-blind, placebo-controlled, two-way cross-over design, where each participant was tested under two acute treatment conditions, separated by at least a 7-d washout period. The two treatment conditions were; GSK598809 (175 mg capsule) and placebo. The dose of GSK598809 was selected because it has been shown to be associated with >90% D_2/3 receptor occupancy (Searle et al.2010). Subjects attended the study on day 1. Dosing, and behavioural assessments and were conducted on day 1. Safety and tolerability were assessed on days 1, 3 and 4. Subjects were required to fast for approximately 15 h prior to testing. The attentional bias tasks were performed approximately 4–5 h post-dose [to coincide with the approximate time of peak plasma concentration (T_max) of GSK598809] and between 15 and 16 h following fasting.

Attentional bias tasks

Visual probe task

The main pictorial stimuli consisted of 20 colour photographs of food similar to those used by Brignell et al. (2009) and Hepworth et al. (2010) . Each food picture was paired with a photograph of another scene matched as closely as possible for content (e.g. number, colour and shape of items), but lacking any food cues (see Fig. 1 for examples). These stimuli were used as the main experimental and control stimuli for this task. An additional 20 picture pairs (unrelated to food) were used as fillers, and an additional 12 food-control picture pairs were used for practice and buffer trials. The computer tasks were presented on a computer using Presentation software and responses were recorded using a two-button response box.

The design was similar to that used by Bradley et al. (1998, 2003) and Hepworth et al. (2010) . Each trial commenced with a fixation cross displayed for 500 ms in the centre of the screen followed by a pair of pictures presented side by side for either 500 or 2000 ms (Fig. 1). A probe was then presented in the position of one of the preceding pictures until the participant gave a manual response. The probe was a single black dot (●). Participants pressed one of two buttons to indicate the location of the probe. They were instructed to look at the fixation cross at the start of each trial. The duration of the inter-trial interval (ITI) varied randomly between 500 and 1500 ms. There were 12 practice trials, and two blocks of 120 trials (160 critical trials and 80 filler trials in total), with a short break between the blocks. Each block was preceded by two buffer trials. The critical trials were made up from eight presentations of each of the 20 food-control picture pairs. There were an equal number of trials in each condition, as a function of picture duration, location of the food picture and probe location. The 20 filler pairs were presented four times each. All trials were presented in a new random order for each participant. When presented on the screen, each picture was 9 cm high by 12 cm wide, the distance between their inner edges was 6 cm and the distance between the two probe positions was 18 cm (visual angle of 11° as participants were seated 100 cm from the screen).

Modified (food) Stroop task

The food Stroop task was adapted from a modified version of the Stroop task used to probe addiction (Cox et al.2006) and further modified to include food-related words. Participants were asked to colour-name words, presented on a computer screen, as quickly and accurately as possible, while ignoring the word's meaning (see Fig. 2). Two different block-design paradigms were used: a food Stroop paradigm to measure attentional bias for food-related cues and a standard colour-word Stroop paradigm to measure cognitive (interference) control.

The two trial types were administered in a block design consisting of two blocks: food Stroop and colour-word Stroop, which differed from each other in the type of words presented. Each block consisted of 64 trials. The food Stroop block presented a list of 16 palatable food-related target words, a list of 16 non-palatable food-related target words and two lists of 16 neutral words (matched for length and frequency). The colour-word Stroop block consisted of 16 words including eight colour-incongruent colour words (e.g. the word blue written in green ink) and eight congruent colour words (e.g. the word red written in red ink), each matched with 16 colour-unrelated words. Each experimental trial lasted 2.2 s and included the presentation of a word for 1.9 s followed by a presentation of a fixation for 0.3 s, which was then immediately followed by the next trial.

Eating style

TFEQ-R18

The TFEQ-R18 measures three aspects of eating behaviour: cognitive restraint, uncontrolled eating, and emotional eating (Karlsson et al.2000; Stunkard & Messick, 1988). Restrained eating or cognitive restraint measures dietary restraint, i.e. the conscious restriction of food intake in order to control body weight or to promote weight loss. The primary reason for including the TFEQ-R18 was to explore the relationship between restrained eating and drug effects on attentional bias and the results reported here focus on this measure, given previous literature, discussed earlier, linking restraint with dopamine response to food cues (Volkow et al.2003). However, additional measures of eating style were included for further exploratory analyses, which are described in Supplementary Tables S1 and S2 (available online)†.

Safety and tolerability assessments

Safety and tolerability were assessed and included spontaneously reported adverse effects (AEs), cardiovascular variables, movement disorder, temperature and respiratory rate.

Data preparation and statistical analysis

For the visual probe task, attentional bias scores were calculated for each participant, session and stimulus exposure duration. These were calculated by subtracting the mean reaction time (mRT) when the probe replaced the food cue from the mRT when the probe replaced the non-food cue. Positive values indicate an attentional bias towards food. Before the calculation of bias scores, reaction times were excluded if they occurred on error trials or if they were outliers (i.e. ⩽200 ms, ⩾2000 ms, and/or >3 standard deviations (s.d.) above each participant's mRT in each session; Hepworth et al.2010).

For the food Stroop and colour-word Stroop task, attentional bias scores for food-related cues was measured by comparing the time taken to name the colour of food-related words and the time taken for matched neutral words. An interference score was calculated by subtracting each participant's mRT for correct responses to neutral words from their mRT for correct responses to target (i.e. food words). Greater colour naming interference for food-related words is interpreted as greater attentional bias to food-related words. For the colour-word Stroop, the interference score was similarly derived for the colour words (i.e. mRT of correct responses to neutral (colour-unrelated) words was subtracted from the mRT of correct responses to incongruent colour words). Greater interference in the colour-word Stroop indicates less cognitive control or conflict monitoring.

A 2×2 analysis of variance (ANOVA) model was fitted for attentional bias scores with drug (GSK598809, placebo) and stimulus duration (500 ms, 2000 ms) as within-subjects independent variables. An ANOVA was also performed for the Stroop interference scores and error data with drug (GSK598809, placebo) as a within-subjects independent variable.

For the correlational analyses, the effect of GSK598809 on attentional bias for the visual probe and modified Stroop tasks was summarized by a contrast term formed by subtracting the bias score in the GSK598809 condition from the bias score in the placebo condition. Positive values indicate larger attentional bias to food in the placebo than drug condition; i.e. reflecting the predicted effect of drug reducing attentional bias to food. Correlational analyses were conducted between the questionnaire measure of restraint at screening (time 1) and the drug-effect contrast term and overall attentional bias (averaged across all conditions) for the visual probe and modified Stroop tasks.

Results

Safety and tolerability assessments

The safety and tolerability data are reported in the Supplementary material (available online).

Visual probe task

Rates of errors and outliers were low (mean percentages of trials with errors and outliers were 0.6 and 1.6%, respectively) and mRT was 436 ms, with no significant differences in error or outlier rate, or overall mRT between GSK598809 and placebo conditions (F_1,25<1). One-sample Kolmogorov–Smirnov tests indicated that the distributions of the bias scores did not differ significantly from normality. However, attentional bias scores from the 2000-ms placebo condition appeared skewed due to a large positive score for one participant; so analyses were repeated without this individual (similar results were obtained irrespective of whether or not this participant was included).

Repeated-measures ANOVA of attentional bias scores showed no overall main effects of drug or stimulus duration, or drug×stimulus duration interaction (F_1,25<1). (There were no significant main or interactive effects of session order on attentional bias.) A one-sample t test showed that, across all participants and conditions, the mean attentional bias score did not differ significantly from zero (mean bias=2.6 ms, s.d.=7.9, t₂₅=1.7, p=0.10); i.e. no significant attentional bias for food cues relative to non-food cues.

The effect of GSK598809 on attentional bias correlated significantly and negatively with TFEQ restraint score (r=−0.45, p=0.02, two-tailed; Fig. 3a), suggesting that the effect of GSK598809 on attentional bias was greater in individuals with lower restraint. The overall attentional bias score was not significantly correlated with TFEQ restraint (r=−0.09, p=0.67). The correlation between restraint and the drug effect on attentional bias remained significant after Bonferroni correction of the threshold for significance for the two bias measures from this task at p<0.05/2=0.025.

The correlational finding indicated that the effect of GSK598809 on attentional bias for food cues was significantly influenced by individual differences in restraint. However, this analysis does not indicate the precise nature of this interaction effect of GSK598809 and restraint on attentional bias. Thus, to clarify the correlational results, the sample was divided into low- and high-restraint groups (TFEQ restraint scores: mean=12.5, median=11.5, s.d.=3.7, range 7–20, n=26; low-restraint group: TFEQ restraint scores ⩽12, n=15; high-restraint group: TFEQ restraint scores >12, n=11). An independent-sample t test showed no significant difference in mean BMI between the low- and high-restraint groups (low restraint=32.3, high restraint=33.1; t₂₄=−0.52, p=0.6). The categorization of low- vs. high-restrained eating using a cut-off of 12 is consistent with previous studies (Yeomans et al.2004, 2008). A 2×2×2 ANOVA model was fitted to attentional bias scores, with drug (GSK598809, placebo) and stimulus duration (500 ms, 2000 ms) as within-subjects independent variables, and restraint group (high, low) as a between-subjects independent variable. There was a significant drug×restraint interaction (F_1,24=4.45, p<0.05) (see Fig. 4a), which corresponds to findings from the correlational analysis.

One-sample t tests showed that only the results of the low-restraint group were consistent with the primary hypothesis (corrected p values=uncorrected p multiplied by 2, as hypothesis-driven tests were conducted for each group separately): the low-restraint group had a marginally significant attentional bias towards food in the placebo condition (t₁₄=2.46; uncorrected p=0.027, corrected p=0.054), but not in the GSK598809 condition (t₁₄=0.45; uncorrected p=0.66, corrected p=1). The high-restraint group showed no bias in either placebo (t₁₀=1.23; uncorrected p=0.25, corrected p=0.49) or GSK598809 conditions (t₁₀=1.57; uncorrected p=0.15, corrected p=0.29) (Fig. 4a).

Modified Stroop task

Mean percent of correct trials in the food Stroop task was 87% (s.d.=17). Mean percent correct in the standard Stroop task was 84% (s.d.=19). Two subjects performed with an accuracy level greater than 2 s.d. below the mean and their data was excluded from further analyses. Mean overall RT for the standard Stroop task was 875 ms (s.d.=169). Mean overall RT for the food Stroop task was 874 ms (s.d.=169). Two variables were calculated from RT scores; food Stroop interference effect (calculated by subtracting mRT to neutral food words from mRT to palatable food words) and the standard Stroop incongruency effect (i.e. interference) (calculated by subtracting mRT to colour-unrelated neutral words from mRT to incongruent colour words). These variables were calculated for the GSK598809 and placebo conditions separately. One-sample Kolmogorov–Smirnov tests indicated that the distributions of these variables did not differ significantly from normality.

One-sample t tests on the standard and food Stroop interference scores in the placebo condition showed that there was the standard interference effect in the standard Stroop (mean interference=172 ms, s.d.=86, t₂₃=9.7, p<0.001), but no significant interference effect of food cues in the modified Stroop (mean interference=18, s.d.=63, t₂₃=1.4, p=0.17. Repeated-measures ANOVA showed no overall effect of GSK598809 on either the food interference effect or standard Stroop incongruency effect (F_1,23<1.2) (there were no significant main or interactive effects of session order on the Stroop effects). The effect of GSK598809 on food interference effect correlated significantly and negatively with TFEQ restraint (r=−0.47, p=0.02; Fig. 3b), suggesting that the effect of GSK598809 on the food interference effect was greater in individuals with lower restraint scores. The overall food interference effect also correlated significantly and negatively with TFEQ restraint score (r=−0.56, p=0.004). These correlations between restraint and the two attentional bias measures remained significant after Bonferroni correction of the threshold for significance for the two bias measures from this task at p<0.05/2=0.025.

To clarify the correlational results concerning the GSK598809 effect on the food interference effect, the sample was divided into low- and high-restraint groups (as described above for analysis of the visual probe task). A 2×2 ANOVA was fitted to the food interference scores, with drug (GSK598809, placebo) as a within-subjects independent variable and restraint group (high, low) as a between-subjects independent variable. There was a significant drug×restraint interaction (F_1,22=6.7, p<0.05), which corresponds to the findings from the correlational analysis. One-sample t tests (corrected p values are unadjusted p values multiplied by 2, as hypothesis-driven tests were conducted for each group) showed that the low-restraint group had a significant food interference effect in the placebo condition (t₁₂=4.56; uncorrected p=0.001, corrected p=0.002), but not in the GSK598809 condition (t₁₂=−0.85; uncorrected p=0.41, corrected p=0.82). The high-restraint group showed no food interference effect in either placebo (t₁₀=−0.77; uncorrected p=0.46, corrected p=0.92), or GSK598809 (t₁₀=0.78; uncorrected p=0.46, corrected p=0.92), conditions (Fig. 4b).

Correlations between the effects of GSK598809 on visual probe and modified Stroop tasks

Correlational analysis revealed that there was a positive correlation between GSK598809 effects on the food Stroop and attentional bias (r=0.43, p<0.05), but not between GSK598809 effects on the standard Stroop and attentional bias (r=0.33, p=0.12).

Discussion

The dopamine D₃ receptor has recently been explored as potential drug target for the treatment of disorders of compulsive consumption including obesity. Until now, there has not been any study in humans investigating the efficacy of D₃ receptor antagonists in experimental models of food reinforcement. We report for the first time, the acute effects of the D₃ receptor antagonist, GSK598809, on attentional bias to food-related cues in a cohort of overweight/obese binge and emotional eaters, using two commonly used tasks of attentional bias, the food Stroop task and the visual probe task. Overall, GSK598809 had no effect on attentional bias to food cues. However, individual differences in eating styles impacted on the effect of GSK598809 on attentional bias, as evident from a significant negative correlation between restraint and the effect of GSK598809 on attentional bias for each task. More specifically, low-restrained eaters showed an attentional bias to food cues in the placebo condition (this bias was significant in the food Stroop task and marginally significant in the visual probe task), but not in the GSK598809 condition. The high-restraint group showed no significant attentional bias in either condition.

Dopamine and attentional bias

Dopamine has been shown modulate attentional bias through activation of D₂/D₃ receptors. Studies conducted in human addict populations have also shown that D₂/D₃ antagonists can reduce attentional biases on a Stroop task with drug-relevant stimuli (Ersche et al.2010; Franken et al.2004). In the current study, low-restrained eaters showed an attentional bias towards food cues in the placebo, but not the GSK598809 treatment condition. This pattern of results was significant on the modified Stroop task and marginally significant on the visual probe task. Together, these findings provide additional support that antagonism of D₃ receptors may attenuate attentional processing of salient or rewarding cues. Given that obese individuals have been shown to have greater attentional bias to food cues (Braet & Crombez, 2003; Castellanos et al.2009; Nijs et al.2010) and food-related cognitive biases have been shown to predict changes in BMI (Calitri et al.2010), it is possible that D₃ receptor antagonists may have efficacy in reducing behavioural aspects of food intake, such as bingeing or overeating in a subgroup of obese individuals by modulating cognitive processing (i.e. attentional bias) and food cue-induced craving, possibly leading to weight loss. These hypotheses require testing in future studies.

Modulation of attention bias by GSK598809 depends on restrained eating

The attenuating effect of GSK598809 on attentional bias was greater in overweight/obese individuals who had lower levels of restrained eating. These findings suggest that, in low-restrained eaters, D₃ receptor antagonism with GSK598809 reduces selective attention allocated to rewarding food cues, which is manifest in both visual orienting and resolution of processing conflict (on visual probe and food Stroop tasks, respectively). Interestingly, no significant correlation between restraint eating and interference scores in the standard Stroop was found (data not reported), suggesting that the effects of the drug on attentional bias are not due to a general attenuation of cognitive interference, but rather, to interference (or conflict) caused by the salient food-related cues. Previously it has been argued that the interference effect in the Stroop task may reflect either a bias towards or away from food words (de Ruiter & Brosschot, 1994). In contrast, the visual probe task allows for the differentiation between attention directed towards or away from a particular type of information (MacLeod et al.1986). In the current study, reaction times in the visual probe task were faster when the probe replaced the food cues (in the placebo condition) relative to non-food cues, indicating that attentional bias was directed towards food cues in the low-restraint group, whereas this food-directed attentional bias was no longer evident after administration of GSK598809.

The lack of overall effect of GSK598809 on attentional bias seems likely to be explained by the fact that the high-restraint group did not show an attentional bias to food cues during placebo treatment. It has been suggested that restrained eaters are more attuned to food cues in the environment and attempt to avoid these in order to control their body weight (Green & Rogers, 1993; Green et al.1997). The lack of attentional bias to food cues in the high-restrained eaters and the inability of GSK598809 to modulate attentional bias in this group may be related to high-restrained eaters adopting cognitive control strategies that suppress processing of food cues in an attempt to counteract the high saliency that food cues may have for them. This is consistent with the evidence that high-restrained eaters show greater changes in dopamine neurotransmission in the dorsal striatum to food stimulation (Volkow et al.2003), possibly reflecting a preventive adaptation strategy to minimize their exposure to salient food cues (Volkow et al.2003). Conversely, low-restrained eaters could be characterized by low dopamine release in dorsal striatum. Interestingly, preclinical studies indicate that D₃ antagonists can act as pro-dopaminergic agents by blocking D₃ autoreceptors expressed in mesencephalic dopaminergic neurons that project to the nucleus accumbens and striatum, increasing the release of dopamine (Collo et al.2008; Schwarz et al.2007). According to this interpretation, in low-restrained eaters GSK586809 would produce a generalized activation of the dopamine release non-contingent to the food cue, resulting in the attenuation of the salience of the cue and of the cue-induced attentional bias. Consistent with hypothesis, the D₃ selective antagonist SB-277011A has been shown to reduce food intake and responses for food in an operant task (i.e. fixed ratio schedule of reinforcement) (Thanos et al.2008) and D₃ receptor-deficient mice have been shown to become obese when fed a high fat diet (McQuade et al.2004), providing a link between D₃ receptors and obesity.

Restraint and attentional bias

The overall attentional bias measured using the modified Stroop task (i.e. food interference effect averaged across treatment conditions) was negatively correlated with TFEQ-restrained eating suggesting that, at least in this population of overweight to obese participants, high-restraint eating was associated with lower attentional bias to food-related cues. As discussed earlier this may be related to high-restrained eaters adopting cognitive control strategies to minimize their processing of salient food cues. This finding also seems conceptually compatible with results of a recent eye-tracking study which indicated that higher levels of restrained eating were associated with reduced attentional bias in visual orienting to high-calorie sweet foods in overweight individuals, which the authors suggested may have been due to such foods having reduced reward value for overweight restrained eaters (Graham et al.2011). However, some other studies using the modified Stroop task have shown that restrained eaters may show greater attentional bias to food cues, although these studies were performed in healthy-weight subjects (Green & Rogers, 1993; Overduin et al.1995; Stewart & Samoluk, 1997) or patients with anorexia and bulimia with high drive for thinness (Perpiñá et al.1993) and hence the latter findings cannot be directly compared to our findings in a overweight and obese population. If restrained eating is associated with a controlled strategy aimed at minimizing processing of food cues, it could be argued that this should be more evident in the longer than shorter stimulus exposure condition of the visual probe task (which would be manifest in an interaction effect of restraint×stimulus exposure on attentional bias). The overall attentional bias score measured using the visual probe task was not associated with restrained eating, which is consistent with a previous study (Ahern et al.2010) and there was also no significant interaction effect of restraint×stimulus exposure, which suggests caution in the interpretation of these null results. Nevertheless, as discussed above, for both tasks there was a significant inverse relationship between TFEQ-restrained eating and drug-induced changes in attentional bias indicating that the effect of D₃ receptor antagonism on attentional bias may be dependent on individual trait difference in cognitive/motivational processes, with attenuation of bias observed only in the low-restrained eating group.

Methodological considerations

There are some methodological factors in the study that should be highlighted. (1) The effects of GSK598809 were observed under conditions of hunger (i.e. 15–16 h of fasting) when increases in food craving and appetitive behaviours are expected. Previous studies have found that both obese and normal-weight individuals show an attentional bias to food cues in a state of hunger compared to satiety (Castellanos et al.2009; Nijs et al.2010) consistent with our findings in low-restraint overweight/obese individuals (in the placebo condition). However, differential effects of hunger vs. satiety on attentional bias may depend on the specific bias measure (Nijs et al.2010) and, given that the effects of GSK598809 were only tested in the state of hunger, our findings cannot be extended to conditions of satiety. (2) We report the acute effects of GSK598809 on attentional bias to salient food cues. It is unknown if these effects translate to a reduction in reinforcing effects of foods and ultimately a decrease in food intake. The findings generated from this study provide encouraging evidence in support of a proof-of-concept study in patients with obesity exploring the chronic effects of GSK598809 on food-related attentional bias, food intake and weight loss. (3) This study only examined the effects of a single dose of GSK598809 on attentional bias. While one could argue that there may be dose-related effects, the dose used in this study (175 mg) has been shown to occupy more than 90% of D₃ receptors (Searle et al.2010), and hence it is unlikely that higher doses would impact further on attentional bias (i.e. D₃ receptor mediated). (4) This study recruited a highly stratified group of obese patients who reported binge and emotional eating behaviour and hence these findings may not be generalizible to the wider obese population. (5) It should be highlighted that while restrained eating may be associated with impaired dopamine neurotransmission (Volkow et al.2003), the exact mechanisms of action of D₃ antagonists in humans (including the role of D₃ receptors) are unknown.

In summary, the current study provides evidence that the D₃ receptor antagonist, GSK598809, has a moderating effect on attentional bias to food-related cues in overweight and obese individuals, which is dependent on individual differences in eating styles, with a stronger drug effect observed in those with lower restrained eating. These findings highlight that individual differences in eating style can modulate the effects of D₃ receptor antagonism on responses to food cues and warrant further attention in future experimental studies investigating the effects D₃ receptor antagonists on measures of eating behaviour and body weight.

Note

Supplementary material accompanies this paper on the Journal's website.

Acknowledgements

The authors acknowledge the contributions made by the GSK598809 project team.

Statement of Interest

This study was funded and conducted by GlaxoSmithKline (GSK) Pharmaceuticals (ClinicalTrials.gov identifier: NCT01039454). All authors except K.M., B.B. and P.C.F. are employees of GSK and hold shares in the company. B.B. and K.M.'s primary employer, the University of Southampton, received compensation from GSK for their work on this project. B.B. and K.M. have received consultancy fees from GSK for their contribution to other research.

References

Author notes