Low- and middle-income countries around the world are experiencing a rapid increase in the prevalence of overweight and obesity(Reference Popkin1, Reference Popkin2). This trend, also evident in Latin American countries(Reference Kain, Vio and Albala3, Reference Martorell, Khan and Hughes4), is associated with a nutrition transition, which entails marked changes in dietary habits and physical activity as a country moves from a traditional way of living to a more Westernised lifestyle. Accelerated increase in obesity rates among children is of special concern because childhood obesity not only has many short-term adverse effects(Reference Must and Strauss5), but is also likely to track into adulthood(Reference Singh, Mulder and Twisk6) and predict chronic diseases later in life(Reference Dietz7).
In addition to dietary intake and physical activity(Reference Gupta, Goel and Shah8, Reference Uauy and Diaz9), psychological factors including body image perception and weight-related concerns are associated with childhood obesity. For example, children may perceive their current body image as either heavier or thinner than it really is and may desire a thinner body image compared with the current one(Reference Pallan, Hiam and Duda10, Reference Al Sabbah, Vereecken and Abdeen11). These associations have been documented at ages as early as pre-school years(Reference Tremblay, Lovsin and Zecevic12).
Parental perception of the child's body image may play an important role in shaping the child's body perception and behaviours related to body image, through parents’ weight-related attitudes, diet-talk and feeding practices(Reference Helfert and Warschburger13–Reference Thompson and Sargent15). In addition to its potential influence on the child's own body image, parents’ perception of the child's body image may pose a serious obstacle for clinicians or public health professionals who are attempting to address childhood obesity(Reference Warschburger and Kroller16), particularly when interacting with parents who misperceive their child's actual weight(Reference Towns and D'Auria17–Reference Rietmeijer-Mentink, Paulis and van Middelkoop20).
Perceptions of body image by both the child and the parents depend on additional factors such as socio-economic status (SES) and exposure to mass media and acculturation towards a Westernised lifestyle that promotes stereotypical ideas of body shape(Reference Towns and D'Auria17, Reference Ward19, Reference Chen, Fox and Haase21–Reference Xu, Mellor and Kiehne24). Although associations between the child's weight status and body image perceptions in children have been previously investigated in developed countries, little is known in populations undergoing the nutrition transition. In these populations, both obesity rates and exposure to Westernised ideals of thinness are expected to continuously increase among adults and children.
The aim of the present study was to identify sociodemographic and anthropometric correlates of body image perception and attitudes among school-aged children from Colombia, a country undergoing the nutrition transition. In addition, we examined various psychosocial constructs of body image perception and examined their potential contribution to body image dissatisfaction in the child.
Methods
Study population and field methods
We conducted the present cross-sectional study in the context of the Bogotá School Children Cohort. This ongoing nutrition and health investigation of primary-school children in Bogotá, Colombia is described in detail elsewhere(Reference Arsenault, Mora-Plazas and Forero25). Briefly, in February 2006 we recruited 3202 children aged 5–12 years to represent all primary public schools in Bogotá by the end of 2005. The children came from 2981 households, when accounting for siblings. The majority of children attending public schools are from low- and middle-income families in the city(26); thus the study population is representative of these strata.
Parents or main caregivers filled out self-administered questionnaires (82 % response) during the first week of classes. These questionnaires were used to obtain information on several characteristics of the child, including sex, age, birthplace and number of hours spent weekly in front of a screen or engaging in outdoor physical activity. The questionnaires also elicited information on mother's age and birthplace, marital status, maternal parity and years of education. Finally, caregivers were asked to provide information on various indicators of socio-economic level, such as the number of home assets; household SES based on the government's classification for tax and planning purposes; and home ownership status.
Trained research assistants visited the schools and obtained the children's height and weight measurements, using standard protocols(Reference Lohman, Roche and Martorell27). Height was measured without shoes to the nearest 1 mm with a wall-mounted portable Seca 202 stadiometer, and weight was measured in light clothing to the nearest 0·1 kg on a Tanita HS301 solar-powered electronic scale. In a random sub-sample of 629 children, information was obtained on their body image perceptions and attitudes with the use of child-adapted Stunkard figure rating scales. These sex-specific scales, formally validated in children(Reference Mciza, Goedecke and Steyn28), portray eight figurines in increasing order of body girth from 1 (thinnest) to 8 (heaviest). Using these scales, children were asked to indicate specific silhouette numbers in response to questions regarding their perceptual body image, i.e. to identify the silhouette that resembled their current body image, and their attitudinal body image, i.e. the silhouette that represented the body image they desired to have. In addition, children were asked to select silhouettes that represented psychosocial constructs related to body image, including the silhouette that ‘looked normal’, ‘looked best’, ‘had more friends’ and ‘looked strongest’. Mothers of 367 children in this sub-sample were asked to use the same child-adapted figurine scales to answer questions about the way they perceived their child's body image, including the silhouette that resembled the child's current body image and the one the mother desired for her child.
The study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving the participants were approved by the Ethics Committee of the National University of Colombia Medical School. The Health Sciences and Behavioral Sciences Institutional Review Board at the University of Michigan approved the use of data from the study. Prior to enrolment, informed consent was obtained in writing from parents or primary caregivers of all children.
Data analysis
The 629 children in the body image sub-sample did not differ from the rest of the cohort in terms of sex, age, SES or anthropometric characteristics. We examined three main outcomes of interest: (i) the silhouette representing the child's current body (body image perception); (ii) the silhouette representing the body desired by the child; and (iii) the difference between the two, which we identified as a body image dissatisfaction (BID) score and represents attitudinal body image(Reference Mciza, Goedecke and Steyn28). A negative BID score was obtained when the silhouette selected to represent the desired body image was greater than that chosen to represent current body image, i.e. a desire for a larger body size. A zero BID score was obtained when the silhouettes representing desired and current body images were equal, indicating satisfaction with body image. Finally, a positive BID score represented a desired body image that was leaner than the perceived current body image, a reflection of the desire for a thinner body.
We first examined the distribution of child and maternal characteristics. Height- and BMI-for-age Z-scores (HAZ and BAZ, respectively) were estimated based on the 2007 WHO growth references(Reference de Onis, Onyango and Borghi29). Maternal BMI was calculated from measured height and weight in 40 % of the mothers and from self-reported data in all other mothers. On the basis of a sum from among six items (refrigerator, bicycle, blender, television, stereo and washing machine), we created for analysis an index of home asset ownership. Household socio-economic stratum was determined according to the local government's classification of public service fees, which are higher in wealthier neighbourhoods than in poorer ones (1–4 in the sample, with 1 being lowest).
Next, we compared the distribution of each of the three body image perception outcomes (silhouettes indicated by the child as current and desired body image, and BID score) by categories of sociodemographic and anthropometric predictors by estimating means (and standard deviations). Differences in the outcomes’ distribution by dichotomous predictors were tested using Wilcoxon's rank-sum test. For ordinal predictors, a test for trend was obtained by introducing a continuous variable representing the ordinal categories of the predictor into a linear regression model. We used multivariable linear regression to examine the associations of child's BAZ with each outcome. Explanatory variables included BAZ and potential confounders such as child's age, HAZ and home ownership, an indicator of SES. Because the association of BAZ with these outcomes was non-linear, we used restricted cubic splines(Reference Durrleman and Simon30) to accurately represent its curvilinear shape. One linear term and three spline terms for BAZ were used in the models. These analyses were conducted separately for boys and girls.
Finally, we analysed the associations of the two components of child's BID, i.e. silhouette numbers indicating perceived (‘current’) and desired body image, with the child's responses to additional body image perception questions that represented psychosocial constructs of body image. In the subset of 367 children with data on maternal perception of the child's body, we studied the relationships of maternal and child body image perceptions and attitudes. Because these ordinal variables were not normally distributed, we calculated Spearman correlation coefficients between each pair of variables. Partial correlations were obtained by adjustment for child's age. These analyses were stratified by child's sex and weight status (thin if BAZ<−1; normal if BAZ ≥ −1 and <1; heavy if BAZ ≥ 1), on the assumption that body image perception depends on those characteristics.
All analyses were conducted with the use of the statistical software package SAS version 9·2.
Results
Mean age of the children was 8·6 (sd 1·7) years; 51 % were boys (Table 1). Boys had a higher mean BAZ than girls (0·23 (sd 1·17) v. 0·04 (sd 0·99), P = 0·10). Nevertheless, compared with boys, girls chose a slimmer silhouette as their desired body image (3·5 (sd 1·3) v. 3·8 (sd 1·4), P = 0·02) and had a slightly higher BID score (0·2 (sd 1·7) v. 0·0 (sd 1·8), P = 0·17), representing a greater desire to be thinner (Table 2). In the subset with information on maternal body image perception, mothers also chose heavier silhouettes to reflect the desired body image for their sons (4·3 (sd 0·7)) compared with their daughters (3·9 (sd 0·8), P = 0·002). Maternal dissatisfaction with the child's body image score reflected a desire to have heavier children, especially for boys.
Table 1 Characteristics of school-aged children and their mothers, Bogotá, Colombia, 2006Footnote *
* Values are presented as means and standard deviations unless noted otherwise.
† According to the WHO 2007 growth references(Reference de Onis, Onyango and Borghi29).
‡ From a list that included bicycle, refrigerator, blender, television, stereo and washing machine.
§ Stratum 1 or 2 of a maximum 4 in the sample and a maximum of 6 in total, according to the government's classification for tax and planning purposes.
Table 2 Body image perception and attitudes of school-aged children and their mothers in Bogotá, Colombia, 2006Footnote *
* Silhouette number from child-adapted Stunkard scales consisting of eight silhouettes in increasing order of girth from 1 (thinnest) to 8 (heaviest).
† Body image dissatisfaction was calculated as the silhouette number representing the desired body image subtracted from the silhouette representing the current body image.
‡ In a sub-sample of 367 children (186 girls and 181 boys).
Silhouettes chosen by the children to indicate their current and desired body images were each positively associated with age (P for trend = 0·01 and 0·002, respectively; Table 3). Compared with shorter children (HAZ<−2), taller children (HAZ ≥ 1) reported a current silhouette that was heavier by an average of 0·6 units and had a mean BID score that was higher by 1·2 units (P for trend = 0·006 and 0·002, respectively). The silhouette indicated by the children to represent their current body and the BID score were both positively associated with the child's BAZ, with the mother's BMI and with her dissatisfaction with the child's body image. By contrast, the silhouette selected to indicate the desired body was strongly associated with these three characteristics in an inverse manner. In the subgroup of 367 children with information on maternal body image perception, the mother–child BID correlation (Spearman) was 0·46 (P < 0·0001). Compared with children whose families did not own their house, children from homeowners indicated a desired silhouette that was 0·2 units slimmer and had a 0·2 units significantly higher BID score (P = 0·03 and P = 0·04, respectively). The household's SES classification was not associated with body image perception of the child.
Table 3 Body image perception in school-aged children (n 629) according to sociodemographic and anthropometric characteristics, Bogotá, Colombia, 2006
*Totals may be <629 due to missing values.
†Silhouette number from child-adapted Stunkard scales consisting of eight silhouettes in increasing order of girth from 1 (thinnest) to 8 (heaviest).
‡From Wilcoxon's rank-sum test for dichotomous variables. For all other variables (ordinal), P was for a test of linear trend when a variable that represented the ordinal categories of the predictor was introduced into a linear regression model as continuous (Wald test).
§According to the WHO 2007 growth references(Reference de Onis, Onyango and Borghi29).
||In a sub-sample of 367 children. Body image dissatisfaction (BID) was calculated as the silhouette indicated by the mother as the desired body image for her child subtracted from the silhouette representing her perception of her child's current body image. A negative BID score represents a desire for a larger body. A BID score of zero represents body image satisfaction. A positive BID score represents a desire for a thinner body.
¶According to the government's classification for tax and planning purposes.
We further studied the associations of child body image dissatisfaction components (silhouettes chosen to indicate current or desired body, and BID score) with actual current weight status (BAZ), using multivariable linear regression adjusted for child's age, HAZ and home ownership, an indicator of SES. The silhouette chosen to indicate current body weight was positively associated with BAZ in boys and girls (Fig. 1). A steep linear relationship was observed above BAZ > −1. Compared with boys, girls chose a heavier current silhouette at BAZ values of about <−0·5 or >0·5; however, this difference was statistically significant only at BAZ values >1·5. The mean estimated silhouette value for BAZ = 0 was about the same for boys and girls (3·9 (se 0·1) v. 3·8 (se 0·1), respectively; mean difference = 0·1, 95 % CI −0·2, 0·4). There was a weak, positive association of desired body image score with BAZ at BAZ values <0 and a weak, inverse relationship at BAZ > 0. This association did not vary significantly by sex.
Fig. 1 Silhouettes representing school-aged children's current (
, boys; 
, girls) and desired (
, boys; 
, girls) body image according to BMI-for-age Z-score (BAZ), Bogotá, Colombia, 2006. BAZ was calculated according to the WHO 2007 growth references(Reference de Onis, Onyango and Borghi29). Children (n 629) selected one of eight silhouettes to indicate their current (perceived) or desired body image, on child-adapted Stunkard scales. Curves were constructed using linear regression models. In these models, the outcome was the silhouette number selected to represent either current or desired body image, whereas predictors included linear and spline terms for BAZ, child's age and height-for-age Z-score (HAZ), and home ownership status. Values are plotted at the population mode for age (11–12 years), mean HAZ and no home ownership. Robust estimates of variance were used in all models
The association of BID score with BAZ followed a ‘J shape’ in boys and girls (Fig 2). BID was larger in girls than boys at any level of BAZ, although this difference was statistically significant only at BAZ values between 1·2 and 1·9. Estimated mean BID score at BAZ = 0 was −0·1 (se 0·2) in boys and 0·0 (se 0·2) in girls (mean difference = −0·1, 95 % CI −0·6, 0·3).
Fig. 2 Body image dissatisfaction (BID) score of school-aged children (
, boys; 
, girls) according to BMI-for-age Z-score (BAZ), Bogotá, Colombia, 2006. BAZ was calculated according to the WHO 2007 growth references(Reference de Onis, Onyango and Borghi29). BID was calculated as the silhouette number selected by the children (n 629) as their desired body image subtracted from the number representing their perceived (‘current’) body image. Curves were constructed using linear regression models. In these models, the outcome was BID, whereas predictors included linear and spline terms for BAZ, child's age and height-for-age Z-score (HAZ), and home ownership status. Values are plotted at the population mode for age (11–12 years), mean HAZ and no home ownership. Robust estimates of variance were used in all models
We next examined the correlations of body image dissatisfaction perception components (‘current’ and ‘desired’ body image) with other psychosocial constructs of body image, after adjusting for child's age. The silhouette chosen by the boys to reflect their perceived body image (‘current’) did not correlate with any of the other psychosocial body image constructs (Table 4). Nevertheless, it was positively correlated with the silhouette indicated as ‘normal’ in thin girls and with the one indicated as ‘best-looking’ among normal-weight girls.
Table 4 Correlations of body image perception components in school-aged children, Bogotá, Colombia, 2006Footnote *, Footnote †
* Spearman correlations are adjusted for child's age. Statistically significant correlation coefficients (P ≤ 0·05) are noted by §.
† Based on the interviewee's rating of child-adapted Stunkard scales consisting of eight silhouettes in increasing order of girth from 1 (thinnest) to 8 (heaviest). Child's body image perception information: n 629. Maternal body image perception: n 367.
‡ According to the WHO 2007 growth reference(Reference de Onis, Onyango and Borghi29).
The silhouette selected to indicate the desired body image was positively correlated with the silhouette chosen as the ‘best-looking’ in all boys and among normal-weight girls. It was also positively correlated with the silhouette chosen to indicate a ‘normal’ body image in normal-weight boys and girls; and strongly among thin girls (r = 0·7). The silhouette selected to represent the desired body image was also positively associated with that chosen as ‘having more friends’ in normal-weight boys. Among thin boys, the silhouette representing the desired body image was positively correlated with that chosen to indicate the ‘strongest’ body (r = 0·4).
In the subgroup of 367 children with information on maternal body image perception and attitudes, the silhouettes chosen by the mother and the child to indicate their current perception of the child's body were positively correlated among normal-weight children only. The silhouettes chosen by the mother and the child to represent the desired body for the child were positively associated in normal-weight boys and, even more strongly, in thin and heavy girls.
Discussion
We examined sociodemographic and anthropometric correlates of body image perception and attitudes in Colombian schoolchildren. Child's body image perception and dissatisfaction were associated with child's weight and height, as well as with the mother's BMI and her dissatisfaction with the child's body. In addition, the constructs associated with the components of body image dissatisfaction varied according to the child's sex and weight status.
The strongest predictor of body image dissatisfaction was the children's actual weight status, consistent with previous reports in other populations(Reference Pallan, Hiam and Duda10, Reference Stice31–Reference Dalton, Johnston and Foreyt35). Both thin and heavy children expressed a desire to be of average size; thus, the association mostly reflects the children's perception of their truly thinner or heavier bodies. We also found strong positive relationships between maternal BMI and the child's body image perception. These associations may result from a high correlation between maternal and child BMI due to common causes. However, it is also possible that the child's perception of his or her mother's actual weight has an independent effect on the child's own body image perception(Reference Keery, Eisenberg and Boutelle36, Reference Cooley, Toray and Wang37).
Child's height was also a significant correlate of body image perception in the present study. This finding is especially noteworthy, because the child-adapted scales depict silhouettes of identical height. It is possible that taller children perceive themselves as being heavier, independent of their actual weight status. Both men and women tend to overestimate their height(Reference Connor Gorber, Tremblay and Moher38, Reference Engstrom, Paterson and Doherty39) and so do short children and their parents(Reference Chaplin, Kristrom and Jonsson40). Yet very few studies have addressed the role of height on body image concerns. Among patients with eating disorders(Reference Gupta and Johnson41, Reference Kollei, Schieber and de Zwaan42) body image dissatisfaction depends also on constructs not related to weight, including height. In a study of adolescents and young adults in China(Reference Chen and Jackson43), physical stature was an important correlate of body image dissatisfaction because tallness conforms to local norms of desired physical appearance, independent of weight. Our results extend the findings from these studies by suggesting that height might play an important role in shaping weight perceptions among children.
Consistent with previous studies(Reference Xanthopoulos, Borradaile and Hayes33, Reference Bully and Elosua44–Reference Wilkosz, Chen and Kenndey47), we found that an SES indicator, house ownership, was associated with body image dissatisfaction. SES is also directly related to obesity in populations at the early stages of the nutrition transition(Reference Monteiro, Moura and Conde48), including ours(Reference McDonald, Baylin and Arsenault49). Hence, the association between SES and body weight could be mediated in part by body image perception. The association of house ownership with BID seemed driven by a desire for a slimmer body, possibly reflecting exposure to Westernised ideals of thinness among wealthier children(Reference Chen, Fox and Haase21, Reference Xanthopoulos, Borradaile and Hayes33, Reference Chaplin, Kristrom and Jonsson40–Reference Chen and Jackson43).
Girls in our study perceived themselves as heavier compared with boys, regardless of their actual weight. In consequence, mean BID score was higher in girls than boys, in line with findings from developed countries(Reference Xanthopoulos, Borradaile and Hayes33, Reference Bully and Elosua44). Our findings also suggest that, compared with boys, psychosocial constructs may play a larger role in girls’ body image perception and attitudes. For example, while perception of current image was unrelated to looking ‘normal’, ‘best’ or ‘strongest’ in boys, it was positively related to looking ‘normal’ in girls. Similarly, the desired body image was highly correlated with looking ‘normal’ in girls only. Previous studies of Taiwanese(Reference Chen, Fox and Haase21) and Australian(Reference Knauss, Paxton and Alsaker50) children found that girls expressed more awareness and internalisation of socio-cultural or media body ideals than boys. Similarly, a study of Chinese adolescents(Reference Xie, Chou and Spruijt-Metz51) found that among girls, but not boys, BID was associated with greater exposure to foreign media and with concerns regarding physical appearance. A longitudinal study among US adolescents(Reference Carlson Jones52) found that social interactions and social norms contributed more to BID in girls, whereas BID stemmed mainly from ideals of muscularity in boys. In our study, a correlation of desired body image with the silhouette looking ‘strongest’ was observed among thin boys only. It is possible that normal-weight or overweight boys are more concerned about adiposity than muscularity. Inconsistencies on the relationships of body image with muscularity in previous studies of males(Reference McCabe and Ricciardelli53) may be the result of failing to consider their actual weight status.
We examined the correlations of child body image perception and attitudes between mothers and children in a subset of 367 pairs. Perceptions of current body image were positively correlated only among normal-weight children. This may be because body image perceptions reflect less accurately actual status among overweight or thin children than they do among children of normal weight(Reference Towns and D'Auria17–Reference Rietmeijer-Mentink, Paulis and van Middelkoop20). Therefore, discordant misperceptions of child's actual weight between mothers and their thin or heavy children may result in attenuated mother–child correlations. Maternal–child correlations of attitudes toward the child's body image were stronger for girls than boys, consistent with previous studies(Reference Helfert and Warschburger13, Reference Mchiza, Goedecke and Lambert14, Reference Gualdi-Russo, Manzon and Masotti54–Reference Cutting, Fisher and Grimm-Thomas57), and were apparent only among thin or heavy girls. This suggests that the mother–child agreement on body shape ideals is larger than the agreement on the perception of current body image.
Of note, whereas child's mean BID score reflected a desire to be thinner, the average maternal score indicated a desire for heavier children, even though children were already heavier than the international reference. This discrepancy may represent intergenerational differences in the appraisal of weight as a health attribute, which are characteristic of societies undergoing the nutrition transition. The maternal desire for a heavier child may reflect traditional concepts associating improved health and physical appearance with heaviness as a result of better economic conditions or health literacy(Reference Xie, Chou and Spruijt-Metz51). By contrast, children may be more likely to associate these qualities with a leaner body due to internalisation of Westernised ideals of thinness and therefore tend to desire a slimmer body for themselves.
Our study has several strengths. Participants represented low- and middle-income children from a population experiencing the nutrition transition. We were able to investigate various aspects of body image perception and attitudes according to the child's sex and weight, as well as mother–child correlations. There are also some limitations. First, the study's cross-sectional design limits the possibility of inferring the temporal sequence of the associations observed. Second, because study participants were of low- and middle-SES levels, generalisation of the results to higher SES groups is limited. Third, the selection of a given silhouette to represent ‘current’ body image is subject to variability. Because there are no normative BMI values assigned to the silhouette drawings in the Stunkard scale, it is not possible to account for this variability on the children's perception of their current body. Finally, although the child-adapted figurine rating scale has been formally validated in comparable settings(Reference Mciza, Goedecke and Steyn28, Reference Stevens, Story and Becenti58), it has not been validated in our population.
Conclusion
We found that body image perception in children is associated with weight and height, as well as with maternal BMI and dissatisfaction with the child's body. The psychosocial foundations of body image perception and attitudes vary by sex and actual weight status of the child. Additional research is needed to further clarify the influences of parental perceptions and attitudes on child's body image and health. Longitudinal studies are warranted to enhance our understanding of the impact of body image perception on weight trajectories in the child and whether interventions on these perceptions may affect growth patterns during childhood and into adolescence.
Acknowledgements
Sources of funding: The Bogotá School Children Cohort Study is currently sponsored by the ASISA Research Fund at the University of Michigan. The funder did not contribute to the study design, conduct of the study, the analysis of samples or data, the interpretation of findings, or the preparation of the manuscript. Conflicts of interest: None of the authors have any conflicts of interest or financial disclosures. Authors’ contributions: O.D. contributed to data analyses, interpretation of the results and writing of the first draft of the manuscript; M.M-P. and C.M. contributed to the study design and data collection; A.B. contributed to the study design, data collection and data interpretation; J.M.L. and C.M.d.L. contributed to data interpretation; E.V. contributed to the study design, data collection and analyses, interpretation of the results and preparation of the manuscript. All authors critically reviewed the manuscript.
