Obesity, characterised by excessive adiposity, is a considerable public health problem whose effects on morbidity and mortality are well known(Reference Boutari and Mantzoros1). According to the current obesity report of the WHO, 60·0 % of individuals in the European region are overweight or obese(2). According to Eurostat data, the prevalence of obesity is 16·5 % in Europe and 22·3 % in Turkey(3). Both genetic and environmental factors play a considerable role in body weight control(Reference Llewellyn and Fildes4). In particular, the developments in the food sector have caused delicious and energy-dense foods to become more accessible and cheaper. These changes in the food environment and decreased physical activity in modern life lead to positive energy imbalance and cause body weight gain(Reference Basolo, Bechi Genzano and Piaggi5).
The rapid increase in the prevalence of obesity worldwide is generally due to the obesogenic environment(Reference Folkvord and Hermans6). Individuals with various genes more responsive to extrinsic food cues or lower sensitivity to satiety have a higher risk of over-eating and obesity in response to the ‘obesogenic’ food environment(Reference Llewellyn and Fildes4). Changes in eating behaviour as a result of differences in appetite are responsible for the impact of environmental and genetic risks on body weight(Reference Sheikh, Nasrullah and Haq7). Regulation of eating behaviour is considered the primary goal for the development of interventions to prevent and treat obesity. Therefore, a better understanding of their aetiology and evolution throughout the life cycle is necessary to establish consistent measures of eating behaviours from childhood to adulthood(Reference Jacob, Tremblay and Fildes8).
In studies examining the relationship between appetite, eating behaviour and body weight in general, valid and reliable questionnaires such as Three-Factor Eating Questionnaire, Dutch Eating Behaviour Questionnaire and Adult Eating Behaviour (AEBQ) were used(Reference Buckland, Swinnerton and Ng9–Reference Braden, Emley and Watford11). The Three-Factor Eating Questionnaire consists of three sub-dimensions: Cognitive Restraint, Disinhibition and Hunger; Dutch Eating Behaviour Questionnaire consists of Emotional Eating, Restraint and External Eating sub-dimensions(Reference Stunkard and Messick12,Reference Van Strien, Frijtiers and Bergers13) . Unlike these scales, AEBQ examines appetitive features in more detail. The AEBQ, which was adapted from the Child Eating Behavior Questionnaire, includes four ‘food approaches’ (Hunger (H), Food Responsiveness (FR), Emotional Over-eating (EOE), and Enjoyment of Food (EF)) and four ‘food avoidance’ (Satiety Responsiveness (SR), Emotionally Under-eating (EUE), Food Fussiness (FF), and Slowness in Eating (SE)). Unlike AEBQ, the Three-Factor Eating Questionnaire measures emotional eating in the disinhibition sub-dimension. Although Dutch Eating Behaviour Questionnaire measures emotional eating, unlike AEBQ, it examines the effect of emotional state on eating(Reference Hunot, Fildes and Croker14).
Currently, the AEBQ is valid in adults and adolescents in the UK, Australia, China, and Mexico in adults, obese adolescents and bariatric surgery patients in the USA and adolescents in Poland(Reference Hunot, Fildes and Croker14–Reference Guzek, Skolmowska and Głąbska21). However, there is no validity and reliability in the Turkish version of this questionnaire in a Turkish-speaking population. Therefore, it is a clear requirement the AEBQ tool is adapted to Turkish society, taking into account cultural and linguistic differences. Therefore, the purposes of this study are (1) to confirm the factor structure of AEBQ-TR, the Turkish version of AEBQ, (2) to determine AEBQ-TR’s both internal and test–retest reliability and (3) to determine whether there is a relationship between AEBQ-TR and appetitive characteristics measured using BMI in the Turkish adult population.
Material and methods
Data were collected through face-to-face interviews with 311 participants living in Ankara between March 2021 and July 2022. Participants between the ages of 20–65, who can communicate in Turkish, who are not diagnosed with eating disorders, who are not pregnant/lactating, and who do not have chronic diseases, were included in the study. The Ankara University Research Ethics Committee approved the study protocol (protocol number: 56786525–050·04·04/47090), and the Helsinki Declaration principles were applied in the research. Prior to the survey, each participant was informed about the study’s contents and signed an informed consent form, indicating their voluntary participation in the research.
Turkish adaptation protocol
Andrea Smith’s e-mail communication provided permission to translate the AEBQ. The English version of the AEBQ has been translated into Turkish. The back-and-forth translation method was used to complete the translation. Two translators fluent in English who were unaware of each other translated the questionnaire into Turkish using advanced translation. The two versions were checked, and any inconsistencies were resolved collaboratively by the research team. Another bilingual speaker who did not know the English version translated it back into English. AEBQ content validity was tested by a group of seven nutritionists, two psychologists and a nurse. Experts were asked to rate each question on its simplicity, clarity, relevance and necessity. The research team evaluated the scale with ten people with no major changes, and the questionnaire was finalised after necessary corrections were made.
Assessment of construct validity
Factor analysis was used to assess the scale’s construct validity. The sample size was determined by multiplying the number of scale items by at least 5–10 times(Reference Yurdugül22). Therefore, at least 175 participants were required to obtain an adequate sample size. Collected data included age, gender, educational status, marital status and BMI. Participants were asked to complete the translated version of the AEBQ. The original questionnaire includes thirty-five questions with a five-point Likert scale (1 = strongly disagree, 5 = strongly agree).
Items in the AEBQ are classified on eight scales, four of which are food approach and four that are food avoidance scales. Hunger, Food Responsiveness, Emotional Over-eating, Enjoyment of the Food scales food approach; Satiety Responsiveness, Emotional Under-eating, Food Fussiness, Slowness in Eating are food avoidance scales. Each scale consists of 3–5 items/questions. In previous validation studies, a better model fit was achieved by removing the eight-factor structure of AEBQ with the Hunger scale or combining it with the Food Responsiveness scale(Reference Hunot, Fildes and Croker14,Reference Mallan, Fildes and de la Piedad Garcia15,Reference Guzek, Skolmowska and Głąbska21) . Subscale scores were calculated using the averages of the items for each scale. Explanatory factor analysis followed the methods used in the original validation study, and varimax rotation principal component analysis was used to test the factor structures of thirty-five questions. The Kaiser–Meyer–Olkin and Bartlett’s sphericity tests were used to determine sample adequacy. Items with factor loadings less than 0·30 or that overlapped were removed from the scale. Internal consistency was assessed using item analysis and the reliability coefficient (Cronbach’s).
Data analysis
AMOS version 21 was used for confirmatory factor analysis. All statistical analyses were performed using SPSS software version 25.0. Chi-square, root mean error of approximation, comparative fit index and normed fit index were used to assess model fit. Chi-square P values greater than 0·05, root mean error of approximation less than 0·08, normed fit index and comparative fit index greater than 0·9 are all acceptable. Explanatory factor analysis investigated the AEBQ’s factorial structure. For the scale’s internal consistency, the standardised parameter Cronbach’s was used. The scale’s test–retest reliability was re-evaluated two weeks later. Besides, simple associations between appetitive trait means and BMI were examined using Pearson’s correlation coefficients. Multivariable linear regression analyses were used to test for associations between BMI and each appetitive trait, adjusting for sex, age, sex, education and marital status as statistical confounders. The statistical significance level was set at P < 0·05.
Results
The sociodemographic features of the individuals participating in the study are shown in Table 1. The mean age of the individuals was 29·3 ± 11·3 years. 69·5 % of individuals are between the ages of 18–29, 52·4 % are women, and 57·9 % have normal BMI. 62·7 % of the participants are high school graduates, and 70·1 % are single. AEBQ-TR was completed a second time by a total of seventy-two participants (36 males; 36 females) aged 31·5 ± 10·7 years.
Table 1 Sociodemographic features of participants
Exploratory factor analysis
Table 2 shows the factor analysis results. Principal component extraction with varimax rotation was used for factor analysis. The established indicators of a high degree of interrelationship between the variables confirmed the relevance of the analysis: Bartlett’s test of sphericity 2 = 5601; P = 0·00 and KMO index was 0·83. In this study, thirty-five items in the AEBQ-TR version produced seven factors similar to the original questionnaire. The factor load showing the relationship of each item with the total score was over 0·30, and seven factors explained 62·2 % of the variance. The excluded factors were retained as described in the original article: Three ‘food approach’ and four ‘food avoidance’ scales were among the seven components. Hunger and Food Responsiveness (loaded on a single component), Emotional Over-eating, and Food Enjoyment were the ‘food approach’ scales. Satiety Responsiveness, Emotional Under-eating, Food Fussiness and Slowness in Eating were the four ‘food avoidance’ scales.
Table 2 Results of explanatory factor analysis of the AEBQ-TR
H: Hunger; FR: Food responsiveness; SR: Satiety responsiveness; EUE: Emotional under-eating; FF: Food fussiness; EOE: Emotional over-eating; EF: Enjoyment of food; SE: Slowness in eating.
* Factor loadings above 0·30 are presented. Extraction method: principal component analysis. Rotation method: Varimax with Kaiser Normalisation.
Confirmatory factor analysis
Table 3 shows the results of confirmatory factor analysis and goodness-of-fit indicators. According to these findings, the seven-factor model provided a suitable model fit in the Turkish data. (chi-square/degrees of freedom = 2·137, root mean error of approximation: 0·061, comparative fit index: 0·884, and normed fit index: 0·850).
Table 3 Results of confirmatory factor analysis of the AEBQ-TR
DF: Degrees of freedom, RMSEA: root mean square error of approximation, CFI: comparative fit index, NFI: normed fit index.
Table 4 indicates descriptive statistics (mean s d), internal validity (Cronbach’s alpha) and test–retest reliability for the eight-factor AEBQ-TR and the original eight-factor AEBQ validation study. The internal reliability of the AEBQ-TR shows adequate internal consistency of the questionnaire, with all Cronbach’s alphas greater than 0·70 except satiety responsiveness. Test–retest reliability was higher than 0·70 (0·95–0·98) for all subscales of the AEBQ-TR.
Table 4 Descriptive statistics and test–retest reliabilities of the AEBQ-TR and the original AEBQ validation
Associations between appetitive traits and BMI
The correlations between subscales are shown in Table 5. The ‘food approach’ subscales were, as expected, positively correlated with one another and generally negatively correlated with the ‘food avoidance’ subscales (Table 5), except for Hunger. Hunger was found to be related to Food Sensitivity, Emotional Over-eating and Food Enjoyment. Food Fussiness was not significantly associated with Slowness in Eating, but it was positively correlated with the ‘avoidance of food’ subscales.
Table 5 Pearson’s correlations and multivariable regression analyses between the AEBQ-TR scales and unadjusted and adjusted correlations with BMI in a Turkish sample
H: Hunger; FR: Food Responsiveness; SR: Satiety Responsiveness; EUE: Emotional Under-eating; FF: Food Fussiness; EOE: Emotional Over-eating; EF: Enjoyment of Food; SE: Slowness in Eating.
Adjusted for age, sex, education, and marital status.
* Correlation is significant at the 0·05 level (2-tailed).
** Correlation is significant at the 0·01 level (2-tailed).
† β (beta) values are unstandardised.
The relationships between BMI and the sub-dimensions of the scale are presented as three different models: (i) unadjusted relationships (Pearson’s correlations); (ii) unadjusted multivariate regressions; and (iii) multivariate regressions adjusted for gender, age, education level and marital status. Both higher Emotional Over-eating (β = 0·149, CI 0·074, 0·223, P < 0·001) and Enjoyment of the Food (β = 0·228, CI 0·069, 0·387, P < 0·001) were associated with higher BMI. Higher BMI was associated with lower Satiety Responsiveness (β = –0·162, Cl –0·295, –0·028, P < 0·001) and Emotional Under-eating (β = –0·084, Cl –0·160, –0·09, P < 0·001).
Discussion
The AEBQ is a valid and reliable scale developed to assess adult appetitive traits. However, this comprehensive, valid and reliable measurement tool was not studied in Turkish. As a result, we performed the first Turkish validity and reliability analysis of the AEBQ, which is widely used in research and clinical practice.
The KMO value and the Bartlett sphericity test were used to determine the sample’s suitability for factor analysis. The value of the Bartlett sphericity test was determined to be statistically significant (P < 0·05) in this study, and the KMO value was greater than 0·60 (KMO = 0·83), both of which meet the criteria for performing factor analysis(Reference Kaiser23).
According to factor analysis, the AEBQ-TR is a seven-dimensional scale with thirty-five items that explain 62·2 per cent of the variance. The factor loading of all scales was greater than 0·40 in this study. The factor load of all items in the original scale was higher than 0·30, and there were seven sub-dimensions(Reference Hunot, Fildes and Croker14). The first scale, Hunger and ‘Food Responsiveness’, has nine items. The second scale is ‘Satiety Responsiveness’, which includes four items; the third is ‘Emotional Under-eating’, which includes five items; the fourth is ‘Food Fussiness’, which includes five items; the fifth is ‘Emotional Over-eating’, which includes five items; the sixth is ‘Enjoyment of Food’, which includes three items; and the final scale is ‘Slowness in Eating’, which includes four items. The original questionnaire’s construct validity is provided by the Turkish questionnaire. According to the literature, the structure determined by explanatory factor analysis should also be examined with confirmatory factor analysis for the scale’s validity and reliability(Reference Burke Johnson24). In this study, the eight-factor structure of the AEBQ, including the Hunger scale, and the seven-factor model, which eliminates the Hunger scale, was tested with confirmatory factor analysis. It was determined that the confirmatory factor analysis fit indices of the seven-factor model had better model fit than the fit indices of the eight-factor model.
Moreover, the Australian and the Spanish studies stated that separating these scales provided the best model fit when there was a strong relationship between Hunger and Food responsiveness, but no relationship between hunger and weight(Reference Mallan, Fildes and de la Piedad Garcia15,Reference Hunot-Alexander, Arellano-Gómez and Smith17) . Similarly, Hunot et al. (2016) found that the seven-factor structure improved model fit. Although Hunger and Food Responsiveness seem to be overlapping constructs, a validation study conducted in China determine that the eight-factor model including Hunger had a better model fit than a seven-factor model including ‘Food Responsiveness’ and ‘Hunger’ scales loaded on a subscale(Reference He, Sun and Zickgraf18). However, confirmatory factor analysis has not been performed to evaluate a model without the ‘Hunger’ scale suggested in the Chinese study. The findings of this study recommended that the ‘Hunger’ scale included in the original AEBQ be removed from future studies using AEBQ-TR. ‘Hunger’ items may be associated with internal states instead of a trait. Therefore, temporal factors such as the time of the last meal may have a greater influence on these items.
A measurement tool must be trustworthy in order to be valid(Reference Taber25). Cronbach’s reliability coefficient, which measures the consistency of the scale’s sub-dimensions with the overall scale, was defined as reliable when it was greater than 0·60 in all sub-dimensions. In the initial study, Cronbach’s alpha values were reported in the range of 0·75–0·90 for all scales of the questionnaire(Reference Hunot, Fildes and Croker14). The AEBQ-TR scale showed internal consistency characteristics that were similar to those of the original scale. The AEBQ-TR is quite reliable, as indicated by the reliability coefficient.
The test–retest methodology is yet another approach for assessing internal consistency(Reference Rattray and Jones26). In this study, the variation of the reliability coefficients of the scale sub-dimensions between 0·95 and 0·98 supports the test–retest reliability of the scale. Similarly, both the original validation and the Spanish study had test–retest reliability > 0·70(Reference Hunot, Fildes and Croker14,Reference Hunot-Alexander, Arellano-Gómez and Smith17) .
In general, food approach (H, FR, EOE and EF) and food avoidance (SR, EUE, FF and se) scales showed a positive correlation within themselves. In addition, mostly negative correlations were found between the food approach and avoidance scales. Unlike the original study, the statistically insignificant weak negative correlation between ‘Emotional Under-eating’ and ‘Hunger’in this study. In addition, contrary to the original study, weak positive relationships were between the ‘Food Fussiness’ and ‘Hunger’ scales, as well as the Emotional Over-eating and Slowness in Eating scales(Reference Hunot, Fildes and Croker14). Similarly, the negative correlation between ‘Emotional Under-eating’ and ‘Hunger’ was also shown in the study in Chinese adults, but the other four studies found positive correlations(Reference Mallan, Fildes and de la Piedad Garcia15,Reference Hunot-Alexander, Arellano-Gómez and Smith17–Reference Zickgraf and Rigby19) . Especially in the original study, the positive correlation between Emotional Under-eating and Hunger made it difficult to explain the Hunger scale(Reference Hunot, Fildes and Croker14). In this study, the weak but negative correlation between ‘Emotional Under-eating’ and ‘Hunger’ scales is not surprising because the ‘Hunger’ scale aims to measure physiological hunger(Reference Karlsson, Persson and Sjöström27). Therefore, these individuals, aware of the hunger-satiety signals, can manage their food intake in the face of their emotions(Reference Fuente González, Chávez-Servín and de la Torre-Carbot28). Although the ‘Hunger’ scale is an important aspect of appetite that is not associated with emotional and restrictive situations, the inability to distinguish this expressed physiological hunger due to eating regulation or various eating attitudes (disinhibition, etc.) may explain the positive correlation between ‘Emotional Over-eating’ and ‘Hunger’(Reference Hunot, Fildes and Croker14). Results related to the Hunger scale raise concerns about the validity of this scale and its use in Turkish society. Therefore, the ‘Hunger’ scale was combined with the moderately correlated ‘Food Responsiveness’ scale. In addition, the positive correlation between ‘Hunger’ and ‘Food Fussiness’ in this study may be because ‘Food Fussiness’ reflects selectivity in food selection, unlike other food avoidance scales that reflect increased sensitivity to satiety cues(Reference Mallan, Fildes and de la Piedad Garcia15). Furthermore, the unexpected positive correlation between ‘Emotional Over-eating’ and ‘Slowness in Eating’ may be due to the high eating speed of the majority of individuals in Turkish society.
Several studies showed a positive correlation between emotional eating and body weight(Reference Van Strien, Herman and Verheijden29,Reference Konttinen, Haukkala and Sarlio-Lä Hteenkorva30) . In this study, there was only a relationship between higher Emotional Over-eating and higher BMI among the food approach features, while a correlation was found between higher Emotional Over-eating and Enjoyment of Food and higher BMI after adjusting age, gender, education and marital status. Various studies found similar results. Hunot-Alexander et al. (2022) found the relationship between higher ‘Emotional Over-eating’ and higher BMI after adjusting age, gender and data collection method. Mallan et al. (2017) determined a similar relation when adjusting age, gender, educational status, marital status and employment status 17. Unlike this study, Mallan et al. (2017) found a statistically significant higher BMI and lower ‘Hunger’, ‘Food Fussiness’, ‘Slowness in Eating’ relationship. On the other hand, Hunot et al. (2022) found a statistically significant relation between higher BMI and lower ‘Slowness in Eating’. Similar to this study, it is interesting that no relation was found between Food Responsiveness and BMI in both the Australian and Mexican studies. This may be due to the lack of awareness of the obesogenic environment.
Among the food avoidance scales, ‘Satiety Responsiveness’, ‘Emotional Under-eating’ and ‘Slowness in Eating’ scales are associated with lower BMI, while ‘Food Fussiness’ scale is associated with higher BMI. Similarly, a relationship was found between lower Satiety Responsiveness and Slowness in Eating and higher BMI in the Australian population after adjusting age, gender and data collection method and in the Mexican study after adjusting age, gender, educational status, marital status and employment(Reference Mallan, Fildes and de la Piedad Garcia15,Reference Hunot-Alexander, Arellano-Gómez and Smith17) . These results support the hypotheses suggesting that obese individuals have higher ‘Food Fussiness’ features or that ‘Food Fussiness’ causes excessive energy intake due to the presence of palatable nutrients(Reference Guss and Kissileff31,Reference Schachter32) . In various AEBQ validity studies, unrelated or negative/positive relationships were found between FF and BMI(Reference Mallan, Fildes and de la Piedad Garcia15,Reference Hunot-Alexander, Arellano-Gómez and Smith17–Reference Zickgraf and Rigby19) . Results may differ depending on sampling (e.g. paediatric, bariatric surgery, etc.) or population habits. In addition, showing different results between ‘food approach’ and ‘food avoidance’ scales and BMI may be due to the fact that anthropometric measurements are based on self-reported in different studies(Reference Hunot, Fildes and Croker14,Reference He, Sun and Zickgraf18) .
Limitations
This study is a scale translated into Turkish to evaluate the AEBQ scale in adults. In addition, this scale may guide future research in Turkish adults. Although this article provides considerable data on AEBQ, it also has some limitations. Firstly, the results may not be applicable to the general population due to the cross-sectional design. Further studies with larger sample groups and other disease subgroups should be examined. Secondly, this study was carried out in a single city (adult Ankara residents), which may have resulted in bias. Future studies should assess its suitability for use in clinical and research settings (via clinical practice and/or other modes of administration such as face-to-face interviews) and with a wider range of populations, including clinical populations.
Conclusion
The results of our study evaluating the relationship between AEBQ and BMI in adults in Turkey showed that AEBQ-TR is a valid and reliable scale for this population. AEBQ-TR has determined that appetitive properties related to the aetiology of weight change and especially obesity risk may be useful in research on adult population as in adolescents and children. Researchers recommend that further studies be conducted to examine population differences between AEBQ-TR scales and BMI in early and late adulthood. Also, repeating this scale in a larger sample and using longitudinal designs is recommended in future research.
Acknowledgements
The authors are grateful to the individuals for participating in this study.
Financial support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Conflict of interest
There are no conflicts of interest.
Authorship
Idea/concept: constructing the hypothesis or idea of research and/or article: H.Y., N.E. and N.N.A.C. Design, planning methodology to reach the conclusions: N.E., N.N.A.C. Control/supervision, organising, supervising the course of progress and taking the responsibility of the research/study: H.Y., N.E., and N.N.A.C. Data collection and/or processing, collection of relevant materials, data management and reporting, execution of the experiments: N.E. and N.N.A.C. Analysis and/or interpretation, taking responsibility in logical interpretation and conclusion of the result: N.N.A.C. Literature review, taking responsibility in necessary literature review for the study: N.E. and H.Y. Writing the article, taking responsibility in the writing of the whole or important parts of the study: H.Y., N.E., N.N.A.C. and H.Y. Critical review, reviewing the article before submission scientifically besides spelling and grammar: H.Y., N.E. and N.N.A.C.
Ethics of human subject participation
This study was conducted according to the guidelines set forth in the Declaration of Helsinki, and all procedures involving research study participants were approved by Ankara University. Written informed consent was obtained from all subjects/patients. Where verbal consent was obtained, succinct consent was subsequently witnessed and formally recorded.
Open access
