The present study analyzes the prevalence of overweight/obesity in a sample of young adults from the University of the Basque Country (Spain), and tests the efficiency of the silhouettes to predict overweight/obesity. This cross-sectional study was conducted in a sample of volunteer university students from the University of the Basque Country (356 men and 745 women, age: 18–33 years), who came to the Physical Anthropology laboratory where a standardized questionnaire was administered and anthropometric measurements were taken by a well-trained anthropometrist. Height and weight were obtained. BMI was calculated as weight/height2 (kg/m2) and it was used as a reference method. Using a questionnaire, based on the standard figural stimuli, subjects were asked to choose the silhouette which was closest to his/her usual appearance (current body size). The accuracy of the Williamson et al.'s silhouettes as an overweight-obesity indicator was analyzed by gender-specific receiver operating curve (ROC). The cutoff figure to distinguish between nonoverweight and overweight-obese individuals corresponded to number 7 in men and 6 in women. These cutoff values matched optimal sensitivity and specificity, with few nonoverweight subjects selecting silhouettes bigger than 7 in the case of men or 6 for women. In conclusion, the figural stimuli allows the identification of populations at overweight/obesity risk with the simple use of silhouettes, at least in this rank of age, where the overweight and obesity are yet little frequent.
Obesity has increased during last decades in many countries. Although this nutritional problem affects the whole society, it seems that its incidence is rising significantly in children, adolescents, and young adults. These groups can be considered at risk to develop different diseases in the adulthood. Indeed, obesity has been related to many pathologies such as diabetes, hypertension, heart diseases, sleep apnea, and degenerative joint diseases. Besides, mortality rates rise with increasing BMI. Consequently, the development of more researches involving overweight and obesity particularly in children and young people is essential (1,2,3,4,5).
BMI is a practical and reproducible method to determine overweight in adults (1,6). In fact, the rapid estimation of overweight through BMI may contribute to an early diagnosis of a young “healthy” population in order to introduce both preventive measures and treatment of the overweight and/or obesity. However, sometimes the measure of large samples is not possible or it presents big complications. In those cases, the body image assessment could represent a useful option to predict overweight and obesity.
Many authors have studied body image during last decades. In fact, different collections of silhouettes have been developed (7,8,9,10,11). In 2000, Williamson et al. developed the “body image assessment for obesity (BIA-O)” as a figural stimulus method for assessing body image disturbances associated with eating disorders and obesity (11). The method consists of 18 silhouettes for each gender and it was enhanced from a previous body image assessment which only included 9 silhouettes (12). This kind of figures' collections has been widely used to study body image disturbances which are mainly related to abnormal eating behaviors and also to obesity (13,14). Most published studies have focused on the relationship between body image and BMI in subjects diagnosed as having eating disorders (15,16,17), mental disorders (18), or even those engaged in physical activities (19). Nevertheless, fewer studies have used this type of silhouettes' collections to assess overweight and obesity in general populations as an epidemiological study (20,21). Usually, the information on the relationship between figural stimuli and measured BMI is scarce and the available literature refers mostly to BMI data derived from self-reported weight and height.
The purpose of the present study was to analyze the prevalence of overweight/obesity in a sample of young adults from the University of the Basque Country (Spain), and to test the efficiency of the silhouettes to predict overweight/obesity. Whatever their actual body size, many young people in industrialized countries are concerned about weight and fatness. As Wardle et al. (22) have recently remarked “young adults are a particularly significant group in relation both to overweight and underweight,” and although in early adulthood obesity is still comparatively uncommon, adolescence and young adulthood are key risk times for body image problems and eating disorders (anorexia and bulimia), including the development of overweight and obesity in the future.
Methods and Procedures
Data were collected from 1,098 volunteer students of the University of the Basque Country (Spain). A cross-sectional sampling which included 353 men and 745 women aged 18–33 years (mean age 22.3 and 21.5 years, respectively) was performed. The aim of the present study and the kind of measurements were explained to the participants, who gave their informed written consent. The University Ethical Committee on Human Research approved the experimental protocols.
Anthropometry and figural stimuli
A well-trained anthropometrist performed all the measurements. Height was measured to the nearest 1 mm using a Harpenden stadiometer (Holtain, Crymych, Wales, UK). A standard beam balance was used for body weight measurements (Añó-Sayol; Atlántida, Añó Sayol, Barcelona, Spain). Readings were rounded to 0.1 kg. Both height (cm) and body weight (kg) were measured following the International Biological Program criteria (23). BMI was calculated as weight/height2 (kg/m2). Administration of the BIA-O (the standard figural stimuli developed by Williamson et al. (11)) involved shuffling the 18 cards (each with a figure of a different size) and presenting them in a different order. The participant was instructed to select the silhouette that most accurately depicted his/her body size as he/she perceived it to be. Each participant might choose only one silhouette and no rearrangements of the cards could be done to compare directly the silhouettes. The experimenter then recorded the card number (1 ± 18), which was written on the back of the card so that the participant could not use the number as an aid in selecting a body size estimate. This number was the score for current body size.
In a preliminary analysis, a Kolmogorov-Smirnov test was used to assess the normality of the data. Since data presented no normal distribution, the nonparametric Mann-Whitney's U-test, Kolmogorov-Smirnov's Z-test, and χ2-test were used to analyze sexual dimorphism. Receiver operating curves (ROCs) were generated using BMI as reference, in order to determine how the silhouettes performed in predicting overweight/obesity (24,25). The curves are a representation of sensitivity (true positive rate) against the corresponding values of [1 - specificity] (false positive rate). These curves can be used to obtain the optimal cutoff score for maximizing the true positive rate and minimizing the false positive rate. The area under the curve is used to test the validity of the indicator, then, the area represents the unit when all the subjects are correctly classified, whereas, an area under the curve equal to 0.5 means the results are due to chance (24,25). All statistical analyses were carried out using SPSS version 15.0 (SPSS, Chicago, IL).
Participants were divided into different categories based on their current BMI: underweight (BMI < 18.5), normal weight (BMI 18.5–24.9) and overweight/obesity (BMI ≥ 25) as the references suggested by the SEEDO (Spanish Association for the Study of Obesity, 26).
Table 1 shows the descriptive statistics (mean and s.d.) for anthropometric variables and decimal age in men and women. Men showed higher values for height, weight, and BMI. In fact, Mann-Whitney's U-test and Kolmogorov-Smirnov's Z-test found significant differences between sexes (P < 0.001), which means sexual dimorphism in body size.
Table 1. Descriptive statistics (mean and s.d.) for age and anthropometric variables in men and women and sexual dimorphism signification
The percentage of underweight individuals is low in relation to other categories (Table 2). Indeed, just 0.9% of men and 3.2% of women were underweight. Overweight and obesity were more common: 28.3% of men and 13.3% of women in the sample were overweight or obese. The prevalence of normal BMI was >70% in both sexes (70.8% in men and 83.5% in women). The χ2-test showed significant differences for overweight/obesity percentages between sexes.
Table 2. Prevalence of categories of nutritional status according to BMI of men and women and χ2-test
Table 3 shows the number of individuals who chose each of the 18 silhouettes as best representing their current appearance, and the mean and s.d. of each group's BMI. The more frequently chosen silhouette by men was number 5 (17.0%), the men's mean BMI was 22.4 (s.d. 1.9). This value was a little lower than the mean BMI of the male sample (23.8, s.d. 2.8). Otherwise, the silhouette number 5 was the most often selected by women (21.9%), these women's mean BMI was 22.0 (s.d. 2.0) which is almost the same as the mean BMI of female sample (22.1, s.d. 2.7).
Table 3. Number of individuals for each silhouette as best representing their CBS, the corresponding BMI, and percentage of overweight and obesity of each group by sex
In both sexes, mean BMI increased in value with silhouette number, except silhouettes 10 and 16 in men and silhouettes 12 and 14 in women, although the number of subjects who chose these silhouettes was very low. The percentage of overweight and obese individuals for each silhouette is shown in the last column of Table 3 for both sexes. Most individuals who chose silhouette 9 or bigger were overweight or obese.
The accuracy of the silhouettes as an overweight-obesity indicator was analyzed by gender-specific ROC. With regard to our silhouettes, the value of the area under the curve was high in both sexes (0.9 for men and 0.8 for women) (Figure 1a, b). The cutoff figure to distinguish between nonoverweight and overweight-obese individuals corresponded to number 7 in men and 6 in women. These cutoff values matched optimal sensitivity and specificity, with few nonoverweight subjects selecting silhouettes bigger than 7 in the case of men or 6 for women. The obtained sensibility for cutoff points was 0.6 in men and 0.7 in women.
In this study, BMI, used as a descriptive method of nutritional status, has shown a remarkable sexual dimorphism. The sexual differences have been due to the higher values for BMI found in men. Obtained mean values and variability were quite similar to those previously reported for university students from Spain and calculated through reported height and weight (27): 23.8 s.d. 2.8 (present study) vs. 23.1 s.d. 2.7 in men, and 22.1 s.d. 2.7 (present study) vs. 21.0 s.d. 2.6 in women.
BMI indicated that over 70% of men and over 83% of women were normal weighted. These results showed a considerable frequency of individuals with overweight or obesity in both sexes (28.3% of men and 13.3% of women). Among men, the data were similar to those from the Pan-European Survey on Body Weight and Physical Activity (28) which included Spanish subjects aged 15–34 years (29% of the Spanish men were overweight or obese), whereas, women from that research presented a higher prevalence of overweight/obesity than women of the present study (18.7% vs. 13.3%). However, mean BMIs as well as percentages of overweight and obesity were lower than those obtained recently in a sample of the Spanish general population with a similar age range (29). In Aranceta's study (29), men's mean BMI was 24.9 s.d. 3.4 and women's mean BMI was 23.2 s.d. 3.6, and prevalence of overweight and obesity was 43.1% in men and 25.2% in women. These values are far from those obtained in the present study. Furthermore, it is necessary to take into account that the studied sample has been obtained in a university sample and it is not representative of the general population of the same cohort. Not only the obesity is less frequent in young adulthood, but also in this case, educational level and social class were higher than in other groups from the same age, the formers are factors which have decisive influence on the maintenance of an appropriate body weight (27).
The need to develop epidemiological studies in large samples has led investigators to look for simple methods which optimize economic and human resources, without losing information. This research has tried to relate measured BMI to body image represented by the collection of silhouettes developed by Williamson et al. (11), by getting a cutoff score to describe those individuals who presented overweight/obesity through silhouettes. The accuracy of silhouettes for the identification of overweight/obesity has been high in both men and women, as the areas under the ROCs have shown (0.9 for men and 0.8 for women). Consequently, we are confident that Williamson's silhouettes can be a useful tool in nutritional assessment in a similar population, being the silhouette number 7 in men and the number 6 in women the cutoff points to describe overweight/obesity. Men showed higher prevalence of overweight/obesity than women did. This fact can be due to the inability of the BMI to differentiate between lean and fat mass. Besides, men will identify bigger silhouettes with muscularity, whereas, women will associate those with overweight/obesity (fatness). Due to the former differences of interpretation, the cutoff point to describe overweight/obesity with silhouettes obtained by ROCs was higher in men than in women.
Most of men and women classified as normal weight by BMI are well classified by silhouettes. Whereas, some cases of overweight individuals were included in the normal weight group by BIA-O method, although the sensibility of the test was high. This matter, particularly in men, could be due to the low sensibility of BMI to distinguish between body fat and fat-free mass classifying individuals with high fat-free mass percentage as overweight.
As an exception, even though nonsignificant, two men and two women chose figures really bigger than those which corresponded to their current BMI. This fact indicated a “misperception” of their body size. Men were clearly overweight, but in view of the fact that BMI is not able to distinguish between muscle mass and fat mass (30), they probably perceived a more muscular body instead of an excess of fat. In the case of men, the development of a “muscular” body is very important, especially in youth, which induces those normal weighted individuals to choose silhouettes associated with higher BMIs. This probably represented a desire for an increase in muscle mass, not an increase in body weight per se (31). Women's election of such different silhouettes from their current BMIs seems surprising, taking into account the fact that although young normal weighted women are likely to overestimate themselves, overweight women are more prone to underestimate their weight status (32,33,34). The overestimation of the body size could be considered as a sign of an eating disorder in these two women, even though it is necessary to take into account the multidimensional nature of body image assessment which often results in difficulties for the interpretation.
Although some authors have pointed out that the specific factors related to a more favorable perception of body silhouette are: overweight, small height, young age, and a low level of education (35), the present research showed that (leaving apart the discussed exceptions) each subject was able to select the silhouette which corresponded to his/her current BMI accurately, as it has been shown in similar studies with other silhouettes collections and populations (20,36). These results showed up that this approach allows the identification of populations at risk with the simple use of silhouettes, at least in this rank of age, where the overweight and obesity are yet little frequent.
The present study does not intend to propose the replacement of anthropometry by the body image technique, but it suggests that this method can be useful under particular circumstances, i.e., in fieldworks and researches which include large samples of population like the study done by Bellisle et al. (27), where scales or stadiometers might not be readily available, or when the assessment occurs in disadvantaged populations (ageing people, some ethnic groups, developing countries) where self-reported height and weight are not as accurate as in the high socioeconomic level populations (20,21). Although the method of silhouettes has been assessed as suitable, in disadvantaged populations a validation should be done to adapt it to particular circumstances.
This investigation has been carried out with two biannual project support, 1/UPV 00154.310-E-13972/2001 and 1/UPV 00101.125-15283/2003, besides a predoctoral scholarship from the University of the Basque Country associated with these projects.