The present study aimed to investigate the validity and reliability of the Persian version of the AEBQ in Kermanshah. The Item Content Validity Index (I-CVI) and Scale Content Validity Index (S-CVI) for relevancy and clarity were above 80%, except for the relevancy of question 6, indicating adequate content validity of this questionnaire. However, content experts reported a relevance index of approximately 60% for question 6, suggesting the need for further research on this question before using the questionnaire in future populations(18–21).
The results of this study in adults over 18 years of age in Kermanshah revealed a 7-factor construct that could explain 97% of the variance. The findings suggest that this questionnaire is a valid and reliable tool for measuring appetitive traits in the population older than 18 years in Kermanshah. Compared to the factor construct of the original version, the "Food responsiveness" and "Hunger" scales in the present study were combined into one factor due to their strong correlation, which increased the internal correlation. However, in the validity assessment of the questionnaire by Kimberley M. Mallan in Australia and the main study in the UK, combining hunger and food responsiveness as one scale decreased the goodness of fit indices, and the 7-factor model (with hunger and food responsiveness combined) was not deemed a good model (6, 22).
The internal consistency (Cronbach’s alpha) of the whole questionnaire was calculated to be 85% based on information obtained from the available sample of 30 individuals. For the subscales, internal consistency ranged from 45–97%. According to Helmstadter, a scale with an internal reliability coefficient greater than 50% is considered to have acceptable reliability and can be used in studies that aim to compare groups (23). Based on the internal consistency coefficient of 85%, it can be concluded that the Adult Eating Behavior questionnaire has acceptable internal reliability, as this value is higher than the suggested threshold of 50%.
Bernstein and Nunnally have set the minimum internal reliability coefficient at 70%(24). Considering this benchmark, the Adult Eating Behavior Questionnaire instrument exhibits acceptable internal reliability. However, the results of the analysis of test-retest factors at two-week intervals showed that the coefficients calculated for some subscales, such as hunger (69%), were slightly lower than the test-retest coefficients reported in the original research (6). Landis and Koch reported that intra-cluster test-retest reliability coefficients greater than or equal to zero indicate non-correlation, while values ranging from 0-0.20, 0.21–0.40, 0.41–0.60, 0.61–0.80, and 0.81-1.0 indicate partial, fair, moderate, significant, and almost complete correlation, respectively(11). Based on this criteria, the satiety responsiveness, food responsiveness, and food fussiness scales exhibited moderate intra-cluster coefficients, while the hunger, slowness in eating, emotional under-eating, emotional overeating, and enjoyment of food scales showed significant intra-cluster correlation coefficients. These findings demonstrate the favorable validity of the Iranian version of the Adult Eating Behavior Questionnaire at a 2-week interval.
The factor analysis of the original version of the AEBQ produced two different constructs: an 8-factor construct and a 7-factor construct. In the 7-factor construct, the subscales for hunger and food responsiveness were combined into one construct (as was done in our present study). However, in the 8-factor construct, these factors were loaded as separate subscales and introduced as a superior model. Rafael Jacob et al.'s validation studies of the Adult Eating Behavior Questionnaire in the French population aged 18 and above in 2021 also support this conclusion (4, 25).
In contrast to the main study, which used confirmatory factor analysis, the present study proposes a 7-factor construct as a better model through exploratory factor analysis. These factors include enjoyment of food, emotional overeating, emotional under-eating, food fussiness, food responsiveness combined with hunger as one factor, satiety responsiveness, and slowness in eating. Together, these factors explain 97% of the variance (6). In other words, while maintaining the individual subscales, the present study combines the food responsiveness and hunger factors. Maintaining the 8-factor model allows for flexibility in using the entire questionnaire or eliminating the hunger scale and presenting a 7-factor model with 30 items, similar to validation studies conducted on Mexican adolescents and adults by Kimberley M. Malan in Australia in 2017, as well as Claudia Honut Alexander's study on the Adult Eating Behavior Questionnaire in a Spanish sample in 2021(4, 22).
The correlation analysis of the questionnaire subscales revealed a significant positive correlation between all subscales of the food approach, with the strongest correlation observed between food responsiveness and enjoyment of food. Additionally, a positive internal correlation was observed between the other subscales of the food avoidance approach, except for slowness in eating and food fussiness, which showed a negative correlation. With the exception of the correlation between slowness in eating and emotional under-eating, as well as satiety responsiveness and food fussiness, all other correlations were statistically significant at the 5% level. These findings suggest that, in most cases, the AEBQ subscales are consistent with previous studies as expected (4–6, 9, 22, 25, 26).
Strengths and Limitations
The strengths of this study included a large sample size and the fact that it was population-based. However, one limitation was that a Health Integrated System (HIS) was used to recruit participants, and incomplete data may have resulted in a non-representative sample. Additionally, questionnaires were completed via telephone interviews, potentially limiting participant confidence and leading to inaccurate responses. Self-reported measurements of weight and height may have also contributed to overestimation of height and underestimation of weight. The cross-sectional nature of the study precluded any inference of a causal relationship. Furthermore, the study was conducted during the Covid-19 epidemic, which may have influenced the results. Finally, the study did not account for potentially confounding factors such as demographic characteristics, lifestyle habits, home diet, or health conditions such as diet-related diseases, which could affect the results.