Toward a Romanian version of the Three-Factor Eating Questionnaire-R21 for children and adolescents (CTFEQ-R21): Preliminary psychometric analysis and relation with body composition

The continuous worldwide progression of overweight and obesity in pediatric populations is becoming one of the most critical global public health concerns [1,2]. While most westernized countries have been working on public health strategies aimed at promoting a healthy lifestyle, developing countries are currently facing the alarming progression of physical inactivity, sedentary behaviors and unhealthy eating behaviors. In 2013 the World Health Organization (WHO) and the European Childhood Obesity Surveillance Initiative (ECOSI) reported an overweight/obesity prevalence of 26.8% (11.6% for obesity) in 8-year-old children from Romania (based on the WHO reference curves) [3], which placed Romania among the worst countries in the world in term of excess weight [4]. More recently, Emandi and collaborators reported that almost one in four Romanian children aged 6-19 years, were overweight or obese between 2006 and 2015 [5].

Although valid universal tools are available to objectively and properly measure physical activity (e.g., accelerometers), valid methods to assess the cognitive and behavioral dimensions of eating behaviors, as well as attitudes towards food, remain to be developed in Romania. The Three Factor Eating Questionnaire (TFEQ) [6] has been specifically developed to assess individuals’ eating behaviors, and its accuracy has been shown in different populations [7, 8, 9, 10]. Recently, a short form of the TFEQ composed of 21 items (CTFEQ-R21) has been developed to measure the cognitive, behavioral and emotional dimensions of eating behaviors [10, 11, 12]. These three factors refer to “Cognitive Restraint” (CR), i.e. conscious effort to control what an individual ingests in order to maintain or decrease body weight, “Uncontrolled Eating” (UE), i.e. a tendency to over-consume food in response to the loss of control over the food itself; and “Emotional Eating” (EE), i.e. the need to overeat when individuals do not manage to cope with emotionally negative events and mood. This shorter TFEQ form has been widely used among adults [11, 13, 14, 15], with a growing interest in children and adolescents [9, 10].

Some studies have been interested in the relationship between weight status or body composition with eating behaviors in children and adolescents [6, 9, 10, 11, 12, 16, 17]. Using the original 51 -item version of the TFEQ, it has been shown that cognitive or restrained eating was positivity associated with children and adolescents’ body mass index (BMI) and body mass [6, 18]; however, the relationships between adiposity and emotional or external/ uncontrolled eating are unclear [19]. Recently, Bryant et al. developed and validated a pediatric version of the TFEQ (CTFEQ-R17), showing that CR was associated with a higher body weight, BMI and BMI percentile and that high UE and EE scores were related to a preference for high fat savoury and high fat sweet foods in 12-year-old children and adolescents [9]. Martin-Garcia and collaborators also found that CR was strongly and positively related to body composition in Spanish children and adolescents using an adapted Spanish version of the TFEQ-R based on 21 times (TFEQ-R21) [10].

While there is a growing interest in the prevention and treatment of pediatric overweight and obesity in Romania, there is also a need for adapted and validated tools to properly assess children and adolescents’ eating behaviors. The purpose of the present study was to validate a Romanian version of the TFEQ for children and adolescents and to analyze its psychometric properties and factor structure. A second aim was to explore the relationships between Romanian children’s and adolescents’ eating behaviors and their body composition.

This study aimed to test the psychometric properties of a Romanian version of the CTFEQ-R21 to assess eating behaviors in children and adolescents (CTFEQ-R21). This Romanian version of the CTFEQ-21 provides a satisfactory assessment of childrens and adolescents’ eating behaviors, with an internal-consistency Cronbach’s alpha coefficient of 0.78. When analyzed separately, CR, EE and UE showed coefficients of 0.55, 0.76 and 0.75, respectively.

These results are in line with those of Bryant et al. who proposed the adapted version of the TFEQ for children and adolescents showing an internal consistency coefficient of 0.81 [9]. In their study, Martin-Garcia and collaborators proposed a Spanish version of this CTFEQ-R21 among 8- to 17-year- old youth and also found a Cronbach’s alpha coefficient of 0.73 [10]. Similarly to the present work, both the English [9] and Spanish versions [10] of this CTFEQ-R21 found lower internal consistency for the CR dimension compared with UE and EE. As illustrated by the heatmap model, the results showed moderate inter-item correlations, which also supports the available literature [9, 10]. While both the English and Spanish versions of the questionnaire found significant correlations between UE and both EE and CR [9, 10], the results observed in the present study showed a significant correlation only between UE and EE, with CR showing low correlations with UE and EE. In the present study, the three factors explained 43% of the total variance, although this remains moderate compared to the 51.6% variance observed in the English version of the CTFEQ-R21 [9]. Nevertheless, this is better than the 34.4% variance obtained in the study from Martin-Garcia et al. [10].

This Romanian version of the CTFEQ-R21 showed moderate correlations between each dimension of the questionnaire (CR, UE, EE) and body weight, BMI and FFM. However, fat mass was not correlated with CR, UE, or EE. Although Bryant and colleagues also observed significant positive correlations between CR, body weight and BMI (particularly in girls) in a similar population, UE and EE were not related to anthropometric measures in their work [9]. Previously published studies also found significant correlations between CR and anthropometric values [7, 8, 10]. In addition, the present results are in line with other evidence showing significant correlations between body weight/BMI and EE [22], UE [10, 23] or both [24, 25].

In the present study, we also divided data according to tertiles of CTFEQ-R21 eating behavior traits that discriminate against children and adolescents showing the lowest, intermediate and highest values for each of the three dimensions. Accordingly, higher body weight, BMI, fat-free mass and fat mass (kg only) were observed with higher tertiles of eating behaviors. In their work, Martin-Garcia et al. also analyzed their results according to tertiles and found similar results for CR only [10]. According to the later study, children and adolescents in the lower UE fertile showed a significantly higher body weight with no difference for BMI and total fat mass. However, they did not find any difference between anthropometric and body composition measures between the tertiles of EE scores [10]. Such divergent results could be explained by the higher heterogeneity of their sample in terms of weight status with a large majority of overweight/obese participants (51 overweight, 83 obese and 54 lean) [10].

Although this work is the first to propose a version of the CTFEQ-R21 for Romanian children, its results must be interpreted in the light of some limitations. First, the composition of our sample must be considered. The present sample is mainly composed of normal-weight children and adolescents, so further investigations should be conducted among overweight and obese ones. Moreover, the sample presents quite a broad age range, which has to be considered when interpreting our results. The unequal proportion of boys and girls is certainly one of the main limitations of our sample. Indeed, we did not observe gender differences during our analysis, which is certainly due to this disproportionate number of boys and girls. This is of particular importance, since the literature clearly points out differences between boys and girls when it comes to eating behaviors, with girls usually showing significantly higher average results in each of such scale dimensions. Nevertheless, even if our analyses had been controlled for age and sex, it would have been more precise to also evaluate the participants’ biological maturation using Tanner stages. The use of an impedance analyzer to assess body composition might also be considered as a limitation, since more accurate methods, such as Dual-Energy X-ray absorptiometry (DXA) could be used, as in the study of Martin-Garcia et al. [10]. However, the Tanita MC780 BIA analyzer used in the present study has been shown to provide accurate measurements of body composition in children and adolescents [21].

The proposed Romanian version of the CTFEQ-R21 for children and adolescents shows satisfactory psychometric properties and internal consistency, with anthropometric values, as well as FM and FFM, being correlated with the three eating behavior traits that compose the whole questionnaire. According to the present results, this first version of the Romanian CTFEQ-R21 proposes an interesting tool that needs to be further developed and improved. Further studies enrolling a larger sample should be conducted to specifically question the validity of this questionnaire among overweight and obese Romanian children and adolescents. Other international tools used to assess eating behaviors should be validated in developing countries that are today facing the progression of pediatric obesity which require effective public health strategies.