Eating disorder symptoms and control-seeking behavior

Abstract

Objective: Eating disorders (EDs) are a heterogenous group of disorders characterized by disturbed eating patterns. Links have been made between ED symptoms and control-seeking behaviors, which may cause relief from distress. However, whether direct behavioral measures of control-seeking behavior correlate with ED symptoms has not been directly tested. Additionally, existing paradigms may conflate control-seeking behavior with uncertainty-reducing behavior.

Method: A general population sample of 183 participants completed part in an online behavioral task, in which participants rolled a die in order to obtain/avoid a set of numbers. Prior to each roll, participants could choose to change arbitrary features of the task (such as the color of their die) or view additional information (such as the current trial number). Selecting these Control Options could cost participants points or not (Cost/No-Cost conditions). Each participant completed all four conditions, each with 15 trials, followed by a series of questionnaires, including the Eating Attitudes Test-26 (EAT-26), the Intolerance of Uncertainty Scale, and the Obsessive-Compulsive Inventory-Revised (OCI-R).

Results: A Spearman's rank test indicated no significant correlation between total EAT-26 score and total number of Control Options selected, with only elevated scores on a measure of obsessions and compulsivity (OCI-R) correlating with the total number of Control Options selected (r_s = .155, p = .036).

Discussion: In our novel paradigm, we find no relationship between EAT-26 score and control-seeking. However, we do find some evidence that this behavior may be present in other disorders that often coincide with ED diagnosis, which may indicate that transdiagnostic factors such as compulsivity are important to control-seeking.

Keywords: behavioral task; control-seeking; eating disorders; intolerance of uncertainty; online.

FIGURE 1

Task trials. Participants start on the main page (top left), with a random die and three random target numbers. They are asked a series of five questions, which appear in a random order for each trial, regarding changes they can make (the Control Options): “Yes” or “No” must be answered to progress. The top right panel shows the page for changing the target numbers (the target number options appear in a random order), and the bottom right shows changing the die color (all color options appear in a random order). The middle page on the left shows the main page after the selection of additional information (showing current points and trial number). See Supporting Information for more information regarding the Control Option categories. The bottom left is the feedback screen—which would show a red cross or green tick, depending on if the task conditions have been met. The die was not shown on the “Rolling” or “Feedback” screen to minimize any association between colors and outcomes. After the feedback screen, participants return to the top left to begin a new trial with a new randomly assigned die and numbers, repeating this cycle until all four blocks had been completed: Avoid/No‐Cost; Avoid/Cost; Obtain/No‐Cost; and Obtain/Cost. Importantly, participants were told that all dice were fair, that they did not need to select any Control Options to continue, and the monetary reward would be at a fixed rate irrespective of any points won or spent. These instructions highlighted the arbitrary nature of all Control Options presented.

FIGURE 2

Effects of Condition and Position. Participants completed the die‐rolling task. On each trial, prior to rolling the dice, they could select various options (“Control Options”) that might allow them to gain information about their points total, change the visual appearance of the dice, or change the target numbers. The task had four different conditions: two separate framings (either participants were trying to avoid or obtain particular target numbers on dice rolls) and two different cost conditions (selecting control options either cost points or did not). In a two‐way repeated‐measures ANOVA, there was (a) no main effect of framing on the number of control options selected (F _(1, 182) = 0.00, p = .99), but there was (b) a main effect of Cost (F _(1, 182) = 25.53, p < .01). There was no interaction between Framing and Cost (F _(1, 182) = 3.55, p = .061). Left axis always shows total number of Control Options selected, dot plots show data points for all participants, and bar chart is drawn at the median. *p < .05.

FIGURE 3

H1–H3: (a) There was no correlation between EAT‐26 (r _s = .12, p = .87) and the total number of Control Options selected and (b) no correlation between Obtain–Avoid Control Option selection and EAT‐26 score (r _s = .061, p = .41). (c) There was no relationship between Cost minus No‐Cost Control Option selection and CPQ score (r _s = −.14, p = .054), though note that this was significant after the exclusion of those who responded that a confounding factor could have influenced their responses to the EAT‐26 (r _s = −.16, p = .034). (d) There was a significant positive relationship between OCI‐R scores and total Control Option selection, though this did not survive correction for multiple comparisons (r _s = .155, p = .036).

FIGURE 4

A path diagram showing a mediation analysis. We were interested in whether a potential relationship between EAT‐26 scores and control option selection was mediated by responses to the Intolerance of Uncertainty questionnaire. The overall effect was not significant, limiting any possible interpretation, although the relationship between the EAT‐26 and IUS‐12 scores was significant, as was the average causal mediation effect. Green boxes represent measured variables, and arrows show the direction of regressions, with annotations indicating the estimated size of effects and significance.