To our knowledge, this is the first time the CEBQ is tested in China. Also, this study is the first attempt to assess eating behaviour pattern among young Chinese children. We examined whether eating behaviours measured by the CEBQ are associated with children's weight in a sample of Chinese children at the age of 12-18 months.
Compared with the eight-factor structure in the original study , our study revealed seven CEBQ sub-scales (with two separate FR factors), and two factors could not be detected in our sample (EF and SR). Each sub-scale contained the items originally belonging to it, with an exception of the sub-scale factor SE, which in our study also had items originally belonging to another sub-scale (EF), such as item 5 (my child is interested in food).
Surprisingly, the original FR factor was split into two factors, each having two items loaded onto it, called FR 1 and FR 2 respectively. The items loaded onto FR 1 were 'Even if my child is full up, s/he finds room to eat his/her favourite food' and 'If given the chance, my child would always have food in his/her mouth'. The items loaded onto FR 2 were 'My child is always asking for food' and 'Given the choice, my child would eat most of the time'. However, one possible reason for two FRs could be the interpretation for 'always have food in the mouth'. Younger children are prone to keep the food in the mouth, and not swallowing immediately compared to older children.
We only show 19 items in table 2, although the original CEBQ scale has 35 items. The other missing16 items either did not load on any factors (loadings < 0.6) or were excluded even though loadings > 0.6 (item 5), or because they originally belonged to the two missing factors (EF and SR). The comparison of factor structure of this study with the original scale structure was indicated in additional file 1. When we lowered the loading from 0.6 to 0.4, it brought more items loaded on each factor. However, it didn't change the final factor results.
Regarding the two missing sub-scales, compared to the original factor structure of the CEBQ (EF and SR), they might not be applicable for the early age of the Chinese study group. The impact of the Chinese history and culture may also be taken into account of. The concept that 'a chubby baby is a healthy baby' is still presents today in spite of the growing awareness of childhood obesity in China. The one-child policy in China in the 1980s aggravated the situation when the only child became the cosset in the family. Moreover, in China, many grandparents take care of their only grandchild in an excessive way, traditionally reflecting in overfeeding practices. As a result, they provide their grandchildren with high energy-dense foods in far greater than needed amounts, increasing the potential risk for adiposity . Consequently, young children could be tired of the food from an early age since they are always overfed, which could explain the missing EF and SR sub-scales.
Cronbach's alpha coefficient normally ranges between 0 and 1 but actually no lower limit to the coefficient is existed and the closer it is to 1, the stronger consistency the items in the scale . In our study, low scores of the two FRs could be due to the inapplicability among particularly young children.
An interesting finding of this study is a significant gender difference in the FF and the two FR factors, suggesting that boys might be more interested in food than girls and girls might be more 'picky' than boys at an early age. Contradictory gender results in picky eating have been reported so far: either no gender differences [31, 32] or boys scoring higher of food fussiness than girls . However, another study has just recently reported the gender difference of 'food responsiveness' when using the Baby Eating Behaviour Questionnaire: male infants were slightly more responsive to food than female infants . The issue is complicated by many aspects, for example, age. Addessi and others showed that this behaviour peaks at the age of 2 - 6, decreasing later all the way through adolescence .
In our study sample we found an unusually high proportion of overweight and obesity being almost half of the sample size. Although this should be checked in a bigger population, we still would like to report this. As mentioned above, this could be due to a common feeding practice in China today together with the lack of physical activity since small children are held by parents most of the time. No significant associations between children's relative weight (BMI SDS) and eating behaviour were observed in the present study. We also considered if adjusting parental weight could affect the correlation between children's BMI SDS and CEBQ scales. Therefore we also checked the simple correlation without controlling anything. The result remains the same and we therefore suggest that this could probably be explained by the young age of children in this study, as some eating behaviours are harder to detect at early age.
Moreover, different eating behavior in Chinese and European children might also contribute to this negative correlation, because CEBQ was developed on the basis of European children's eating behavior. This was also one main intention to conduct our study.
This study has several limitations. First, the sample population might not be representative for the whole Chinese children at the age of 12-18 months as it was tested only in two cites. Secondly, the parental weight and height were self-reported. In addition, the original CEBQ used oblimin rotation in factor analysis, we used varimax rotation instead, though no difference presented.
Despite of the drawbacks, this investigation pioneered a new field of studying young children's eating behaviours by using CEBQ in a totally different setting (China) and in younger children compared to the original and other studies. Based on our results we suggest certain adjustments in the Chinese version of CEBQ, such as to exclude the questions belonging to EF and SR scales, as they might be irrelevant for such an early age in Chinese children. In addition, our important observations regarding children's gender difference in eating behaviour such as FF and two FRs could have a further study.