La restriction cognitive (RC) désigne l’attitude des personnes qui restreignent délibérément leur consommation alimentaire dans le but de perdre du poids ou pour éviter d’en prendre. Il existe plusieurs outils de mesure de la RC, mais qui limitent leur évaluation à sa dimension comportementale. Depuis 1998, le Groupe de Réflexion sur l’Obésité et le Surpoids (G.R.O.S.) analyse les déterminants de ces comportements à partir des modélisations récentes en thérapies cognitivo-comportementales et émotionnelles. Le G.R.O.S. a ainsi développé un outil de mesure global, le Cognitive Restraint Assessment Questionnaire (CRAQ), en 2009, dont la validation pose cependant question et qui doit être reprise.

Nous proposons ici une validation de construit des dix dimensions du CRAQ et la comparaison statistique de trois regroupements factoriels retrouvés dans la littérature ou la clinique.

Cette validation s’appuie sur des analyses factorielles confirmatoires (AFC) et de comparaisons de performance de modèles non imbriqués à partir des données d’un échantillon français de 452 personnes.

Les AFC confirment les dix dimensions. De plus, deux organisations factorielles sont confirmées, une première organisation unifactorielle confirme l’intérêt du score total, une seconde organisation fondée sur la clinique définit cinq facteurs apportant une finesse dans l’orientation des prises en charge.

Cette première étape dans la validation du CRAQ demande à être poursuivie et confirmée dans d’autres populations.

Cognitive restraint (CR) is a behavioural pattern characterised by individuals deliberately restricting their food intake to achieve weight loss or prevent weight gain. CR is currently the most comprehensive theoretical framework for elucidating the ineffectiveness of conventional weight loss strategies and the associated cognitive, emotional, and behavioural challenges, including the dysregulation of eating patterns that can contribute to the development of eating disorders. While various tools exist to assess CR, these typically focus solely on its behavioural aspects. Since its inception in 1998, the Groupe de Réflexion sur l’Obésité et le Surpoids (G.R.O.S.) has been investigating the underlying factors influencing these behaviours, drawing upon contemporary cognitive-behavioural and emotional therapy models. In 2009, the G.R.O.S. developed the Cognitive Restraint Assessment Questionnaire (CRAQ), a comprehensive measurement instrument whose validity has been subject to scrutiny and requires further examination.

This study aims to validate the ten dimensions of the CRAQ and conduct a statistical comparison of three distinct factorial groupings identified in existing literature or clinical practice: a unifactorial grouping (CRAQ total score), a three-factor statistical grouping retrieved by Rabemampianina et El-Hage in 2009, and a five-factor clinical grouping.

To validate the proposed structure of the 10-dimensional measure, a three-step process was employed. Firstly, principal component analyses (PCA) were conducted to explore the underlying factorial structure of the 10 dimensions independently. Secondly, a series of confirmatory factor analyses (CFA) were performed to test the hypothesised factor structures. The first series of CFA examined the 10 dimensions, and their underlying structure identified in the PCA at step 1. Subsequently, a second series of CFA was conducted to assess the three-factorial structure, based on the 10 factors retrieved during the first series of CFA. Finally, a comparison of the performance for the three non-nested models (five-factor, three-factor, and a single-factor model) was undertaken to determine the optimal model. Data were collected from a French sample of 452 individuals. Participants were recruited from two sources: (1) psychological files of patients routinely followed at the Clinique Saint-Louis de Poissy, and (2) individuals who voluntarily agreed to participate in the study through social media platforms (LinkedIn, Facebook, Twitter/X).

Step 1: a principal component analysis (PCA) was conducted to examine the underlying factor structure of the 10-dimensional measure. Prior to the analysis, assumptions were tested to ensure the suitability of the 10 dimensions for factor analysis. The PCA revealed a complex factor structure. Six dimensions demonstrated a clear one-factor structure. Three dimensions (Fear, Cognitive Restraint, and Strategies of Control) exhibited a two-factor structure. The Eating Sensation dimension displayed a four-factor structure. Step 2: a series of CFAs were conducted to test the hypothesised factor structures derived from the PCA (first-ordered factorial structure). All dimensions and their underlying factorial structure were confirmed, except for the problematic four-factor structure of the Eating Sensation dimension. To address this issue, a post-hoc analysis was conducted to determine the most parsimonious factor structure for this dimension (exclusion of the dimension, one-factor or four-factor structure). The results indicated that a one-factor structure for Eating Sensation contributed significantly to the prediction of the total CRAQ score. Moreover, with the second series of CFA, all second-ordered factorial structures were confirmed (five-factor, three-factor, and a single-factor model). Step 3: performance comparison of the three non-nested models highlighted the importance of clinically based organisation that delineates five factors, offering a more nuanced approach to intervention.

This study aimed to validate the ten-dimensional structure of the CRAQ and to assess the psychometric properties of three alternative factor structures. Our findings provide strong empirical support for the construct validity of the CRAQ. The initial PCA revealed an underlying complex factor structure of the ten dimensions, with some exhibiting multi-factorial patterns. Additionally, the second-ordered five-factor clinical organisation demonstrated superior fit compared to a three-factor or single-factor model. The five-factor structure offers a particularly valuable tool for the nuanced assessment of CR determinants, facilitating more tailored patient management. This approach has the potential to mitigate patient attrition during therapy by focusing on the dysfunctional dimension, and to reduce relapse rates due to a more effective intervention. Given the pivotal role of CR, encompassing food beliefs and control strategies, in the maintenance of both syndromic and subsyndromic eating disorders, a thorough assessment of this dimension is particularly essential to highlight potential vulnerability. This initial step in validating the CRAQ necessitates further investigation and confirmation in diverse populations.