A systematic review and meta-analysis of cognitive bias to food stimuli in people with disordered eating behaviour

Abstract

Aim

Maladaptive cognitions about food, weight and shape bias attention, memory and judgment and may be linked to disordered eating behaviour. This paper reviews information processing of food stimuli (words, pictures) in people with eating disorders (ED).

Method

PubMed, Ovid, ScienceDirect, PsychInfo, Web of Science, Cochrane Library and Google Scholar were searched to December 2009. 63 studies measured attention, memory and judgment bias towards food stimuli in women with ED.

Results

Stroop tasks had sufficient sample size for a meta-analyses and effects ranged from small to medium. Other studies of attention bias had variable effects (e.g. the Dot-Probe task, distracter tasks and Startle Eyeblink Modulation). A meta-analysis of memory bias studies in ED and RE yielded insignificant effect. Effect sizes for judgment bias ranged from negligible to large.

Conclusions

People with ED have greater attentional bias to food stimuli than healthy controls (HC). Evidence for a memory and judgment bias in ED is limited.

Introduction

Dysfunctional attitudes regarding weight shape and eating, for example concerns about food, weight gain, misperceptions regarding the body and obsessions about thinness are diagnostic criteria for eating disorders (ED) (American Psychiatric Association, 1994). Robust prospective evidence exists for the contribution of weight concern to the development and maintenance of ED (Goldfein et al., 2000, Gowers & Shore, 2001, Jacobi et al., 2004).

Cognitive theories related to ED (Fairburn et al., 2003, Vitousek & Orimoto, 1993) propose that elaborate, inaccurate cognitive structures or maladaptive schemata underpin common concerns about weight, shape and eating. Overuse of maladaptive schemata may cause these concerns to become habitual and automatic (Williamson, Muller, Reas, & Thaw, 1999), biasing attention, memory and judgment regarding ED-related stimuli such as food. Processing biases related to food stimuli may contribute to the development and maintenance of ED (Vitousek & Orimoto, 1993), and susceptibility to cognitive biases may act as a biomarker for EDs.

There is some evidence that bulimia nervosa (BN) is a culturally-bound Western phenomenon, whereas anorexia nervosa (AN) is likely to be more strongly linked to heritable factors (Keel & Klump, 2003). Some disordered eating behaviour may therefore be linked to an abundance of food-related stimuli (e.g. advertisements and the availability of highly palatable food) in Western societies. Cognitive processing of food-related stimuli might increase dopamine release in the brain associated with incentive saliency – the ability of a stimulus to elicit a reward-driven response (Berridge, 2009, Wise, 2006).

Biased information processing of food-related stimuli (e.g. related to attention, memory and judgment) may be concomitant with differential psychological function (e.g. excessive cognitive restriction and control of eating versus impulsivity) in those who are at risk of developing AN or BN. For example, for those who are at risk of developing AN, a cognitive bias to food stimuli might reflect rigid cognitive strategies linked to cognitive restriction of appetitive processes. Conversely, those who have impulsive appetitive tendencies and are at risk of developing BN might experience disruption to cognitive processes in response to food stimuli, as they may be less able to maintain attention, memory and judgment.

Various experimental paradigms examine information processing to food stimuli both in people with an ED and in people at high risk but without physiological complications that may influence cognitions. For example, restrained eaters (RE) whose weight is in the normal range but who restrict food intake. Attention bias refers to how salient stimuli are preferentially processed so that the focus of attention influences responses (Macleod, Mathews, & Tata, 1986). Other measures of attention bias use distracter tasks to determine whether disease-related stimuli can draw attention away from a target stimulus (e.g. Smeets, Roefs, Van, & Jansen, 2008).

Memory and judgment biases are the two other cognitive domains extensively studied in ED (Williamson et al., 1999), but somewhat neglected in previous reviews. Information related to an individual's concerns may be encoded more readily in memory and easier to recall. Judgment bias refers to the perception of body- and food-related stimuli and how perception is altered in relation to an individual's current concerns or maladaptive schemata.

Earlier reviews have reported attentional bias in ED populations and at risk non-clinical groups using Stroop tasks. One qualitative review found that in people with AN interference is most robust for food-related stimuli, whereas in people with BN it is strongest for body-related words and in people with non-clinical RE interference for food words is equivalent to women with AN but for body-related words is virtually non-existent (Faunce, 2002).

A meta-analysis of 28 Stroop studies with positive and negative body- and food-related stimuli in AN and BN and non-clinical RE (Dobson & Dozois, 2004) found that both body- and food-related stimuli cause equivalent attentional bias in people with BN, but greater effect for body/weight than food stimuli in people with AN; attentional bias in RE to any stimuli was virtually non-existent.

A later meta-analysis examined Stroop studies using food and body-related stimuli with negative overtones (e.g. related to weight concerns) in 27 studies: ED, non-eating disordered women with concerns about their bodies and eating, and HC were examined. (Johansson, Ghaderi, & Andersson, 2005). The mean interference score across body- and food-related stimuli in women with ED was significantly larger than in non-ED women with body and shape concerns and normal controls.

Another meta-analysis extended these findings with attention and memory biases in ED and non-clinical groups using Stroop (n = 27), Dot-Probe (n = 4) and memory tasks (n = 6) (Lee & Shafran, 2004). ED populations and non-clinical samples with ED pathology (e.g. restrained eating, high drive for thinness and chronic dieters) take longer to name eating-, weight- and shape-related words. Dot-Probe studies are comparatively sparse, but more robust than the Stroop at measuring disorder severity. There was some indication of memory biases for eating-related stimuli in ED and non-clinical samples.

This review extends previous reviews by considering paradigms examining the salience of disease-related stimuli in people with disordered eating behaviour, that is, the degree to which the relevance of stimuli influences information processing (attention, memory and perceptual judgment). Other paradigms examine the valence of disease-related stimuli in people with disordered eating behaviour, that is, whether their attitude or mood towards the stimulus is positive or negative. For example, the Implicit Association Test (IAT: Greenwald, 1998) and the Affective Simon Task (De Houwer, 2003) are two popular methods. Valency encompasses an evaluation of the merit of a stimulus, whereas saliency involves the ‘strength of relevance’ (e.g. whether the stimulus disrupts cognitive function or not). Food- and not body-related stimuli were included in this review, in order to summarise the effects of appetitive processes on cognitive functions in people with ED and those at risk (e.g. RE), as opposed to processes related to concerns about the body.

This review differs from previous reviews (Dobson & Dozois, 2004, Johansson et al., 2005, Lee & Shafran, 2004) on five counts; a) it considers cognitive bias to food stimuli only, as there is some evidence that bias towards food stimuli is more robust than body image stimuli at detecting a difference between ED groups (Faunce, 2002); b) it examines a broader sample of EDs (e.g. AN, BN and those at risk but without physical complications: RE); c) it summarises the most widely studied cognitive biases of salience in ED (e.g. attention, memory and judgment bias); d) it includes two meta-analyses (for attention and memory bias); e) there is extensive quantitative information to aid interpretation of the differences between patient and control groups (e.g. Cohen's d, Confidence Intervals, mean scores and probability values for significance).

The aim of this review is to synthesize and analyze information processing (attention, memory and judgment) towards food stimuli in people with an ED and people at high risk of developing an ED, e.g. restrained eaters (RE).