It has been described that eating behavior, measured through CEBQ in children, is a relatively stable trait over time, showing a good reproducibility and high internal consistency . A clear and graded association between CEBQ scores and BMI has been reported previously [10, 11, 27]. The present study confirms the existence of such association in Chilean children, especially in relation to the positive associations between obesity and "food-approach" ("pro-intake") subscales such as EF, FR and EOE. These results are similar to previous studies showing that children with increased BMI are highly responsive to environmental food cues. The inverse associations between body weight and scores of "food-avoidant" ("anti-intake") subscales such as SE and SR are similar to other studies. The CEBQ subscales DD, EUE and FF showed no association with childhood obesity [11, 27].
The original 8-factor structure was not perfectly replicated in our study since an important degree of overlapping has been found between SE and SR . However, we have analyzed the eight original factors for case-control associations to allow for comparisons with other studies [10, 11, 27]. The importance of focusing on behavioral traits is that eating behavior is susceptible to modification through adequate interventions to prevent and/or treat childhood obesity. In this context, changes in CEBQ scores can be also used to assess the effectiveness of such preventive/therapeutic actions . The results of our study indicate that the population of children under the age of 12 could be the target of intervention programs (education provided to families, promoting healthy diets and lifestyles) to improve the nutritional condition and the control of non-communicable diseases. On the other hand, the genetic influence on human feeding behavior has also been evaluated through studies of twins. TEDS (Twin Early Development Study) evaluated eating behavior at the age of 11 in monozygotic (MZ) and dizygotic (DZ) twins born in the United Kingdom through the questionnaire CEBQ . In addition of finding a high heritability for the BMI in childhood, a significantly higher correlation in MZ twins in relation to DZ twins in the subscales of enjoyment of food and satiety response was found in TEDS .
We have been evaluating the best way to represent graphically the multidimensionality of childhood eating behavior. After evaluating different options of multivariate charts, we have arrived to the so-called multivariate radar charts. We have used this type of chart to report child eating behavior to Chilean families, who understand it very well and consider useful. This type of chart defines two polygons: the upper polygon representing positive attitudes to food intake, and the lower polygon for negative attitudes to food intake. Although the ratio of areas of the upper and lower polygons might be a considered as a gross index for general children eating behavior, this index is in fact not adequate because the shapes and areas of the polygons are modified by only changing the order of the subscales. On the contrary, the sum of either "food-approach" or "food-avoidant" scores are not affected by this distorting effect and their ratio is used in Figure 1 as a summary index. In spite of the limitations commented above, we believe that radar charts are useful for health professionals in visualizing and interpreting children eating behavior.
There are few publications that have evaluated eating behavior in children through the CEBQ [10, 11, 27]. To our knowledge, this study is the first to assess the factor structure of the CEBQ and association with BMI in a Latin America population. Some advantages of our study were that children's weight and height were measured directly and not parentally reported and all interviews were conducted face-to-face by trained personnel. There are a number of limitations in our research derived from its limited sample size and the cross-sectional nature of this study. Significant associations detected in this study may in fact be due to reverse causation in which the disease (childhood obesity) caused mothers to change their perception of child's eating behavior. On the other hand, there is an inherent uncertainty related to the measurement of subjective eating behavior in humans through questionnaires, in contrast to direct measurements of eating behavior .