Objective. To evaluate whether body mass index (BMI) is a representative equivalent measure of body fatness independent of age, race, gender, sexual maturation, and distribution of fat in children and adolescents.
Study Design. Cross-sectional study of 192 healthy subjects (100 boys and 92 girls, 103 white and 89 black) age 7 to 17 years.
Methods. Height and weight were measured in the standard fashion, and BMI (kg/m2) was calculated from these values. Fat mass and percent body fat were determined using dual-energy x-ray absorptiometry. Sexual maturation was evaluated by physical assessment. Distribution of fat was determined by the waist:hip ratio.
Results. There were no significant differences by gender and ethnic group for any of the demographic or anthropometric variables, except waist:hip ratio, which was higher in white compared with black boys. BMI was significantly and positively correlated with age, stage of maturation, and all of the anthropometric variables in each race-sex group. The correlation of maturation stage with BMI was stronger than the correlation between age and BMI. A multiple regression analysis showed that BMI, gender, race, sexual maturation, and distribution of fat were all significant independent correlates of the percent body fat (multiple R2 = .77). The percent body fat-BMI relationship was dependent on the stage of sexual maturation, gender (for an equivalent BMI, girls have greater amounts of body fat than boys), race (for equivalent BMI, whites have higher body fat than blacks), and waist:hip ratio (for equivalent BMI, those with central obesity have greater body fatness than those with peripheral obesity).
Conclusion. BMI is not an equivalent measure of the percent body fat for each race-sex group. When BMI is used as a measure of body fatness in a research or clinical setting, particularly when comparisons are made across race and gender, it may be important to consider the maturation stage, race, gender, and distribution of body fat in the interpretation of the results.
Obesity in children and adolescents has become an increasing clinical and public health concern. There appears to be a secular trend toward increasing obesity in children in the United States.1 This trend appears to be resulting in increased morbidity in childhood2 and may result in premature mortality in adulthood.3 Obesity is the result of excess adipose tissue. Unfortunately, the accurate measurement of total fat mass requires sophisticated and often expensive methods that have limited applicability in the clinical setting. Many clinicians and investigators have used measures of body weight standardized for height as an approximation to measurement of adipose tissue mass. The most commonly used measure of body composition is the body mass index (BMI) which is body weight (kg) divided by height (m) squared. This measure was first described by the mathematician Lambert Adolphe Jacques Quetelet and has sometimes been referred to as the Quetelet index.4 A variety of studies have suggested that BMI is a useful measurement of obesity, and it has been recommended for the clinical evaluation of obesity in adolescents.5 Gallagher et al6 found that BMI is age- and sex-dependent when used as an indicator of body fatness in adults. However, it is not known whether this measure of adiposity is independent of the effects of age, race, gender, and sexual maturation in children and adolescents. The use of dual-energy x-ray absorptiometry (DEXA) has been shown to be a reliable and accurate measure of fat mass.7-9 It has close correlation with hydrodensitometry, which has been the gold standard for evaluation of percent body fat.10 Therefore, DEXA provides a useful comparison with which to evaluate the utility of BMI in the evaluation of fatness.
The purpose of this investigation was to evaluate whether BMI is representative of body fatness independent of age, race, gender, and sexual maturation in children and adolescents. This was accomplished by comparing BMI to the measurement of body fat using DEXA.
The study included 201 subjects with an age range of 7 to 17 years, who were recruited from the local school system. Nine subjects with BMI greater than 30 were excluded from the study because of potential methodologic concerns with the reliability of determination of fat mass by DEXA in very obese subjects. Thus, the study group included 192 subjects, 92 girls (40 black, 52 white) and 100 boys (49 black, 51 white). The ethnic group was determined by self report, with concordance between the child and parents required. Subjects were healthy at the time of investigation and had no chronic or acute illnesses. The study was approved by the Institutional Review Board of the Children's Hospital Medical Center, Cincinnati, OH. Subjects were studied after obtaining informed consent from the parents or guardian.
Height was measured with subjects in socks and their head in the Frankfurt plane by means of custom-made portable stadiometers. Weight was measured with a Health-O-Meter electronic scale (Model 482 Continental Scale, Bridgeview, IL). All measurements were taken twice and repeated a third time if the first two measurements differed by more than .5 cm (height) or .3 kg (weight).
Dual-energy X-ray Absorptiometry
DEXA is a relatively new method of assessing body composition. This technology can be used to measure the mass of body fat as well as the bone mineral content. It uses two radiograph beams that traverse the body. The energy that is collected by the detector is attenuated by the bone and soft tissue through which it has passed. The number of photons per unit area is corrected for soft tissue by a linear two-dimensional interpolation, and the corrected values are summed to estimate the total body mineral content. Soft tissue is resolved using mass attenuation coefficients from tissue equivalent standards for fat and fat-free tissue. DEXA has been shown to provide accurate and precise measurements of total body bone mineral content and total body fat mass.9 DEXA has been validated in adults and children against the hydrodensitometry method that has previously been established as the most valid and reliable method for measuring fat mass.10
Subjects in the study were scanned using a Hologic (QDR-1000/W) whole body scanner. For data interpretation, body composition is divided into bone mass and soft tissue mass. Soft tissue mass can be divided into fat mass and lean body mass. The percent body fat is calculated by dividing the fat mass by the body mass.
Pubertal staging was determined by physical assessment. Examiners of boys were male and examiners of girls were female. For boys, the maturation assessment included pubic hair stage as defined by Marshall and Tanner11 and determination of testicular volume by means of the Prader orchidometer. Pubertal staging was defined by the criteria of Biro et al,12 combining those two elements to produce a scale of 1 to 4. For girls, maturation staging was performed with staging criteria developed by Garn and Falkner that incorporate pubic hair and areolar development based on Tanner staging principles,13 which were modified to make them suitable for girls of all ethnic groups and body habitus. These methods have been described by Morrison et al14 and result in a scale of 1 to 3, where 1 is prepubertal, 2 is pubertal but premenarcheal, and 3 is postmenarcheal.
Data were analyzed using SAS statistical programs (SAS Institute, Cary, NC). Data are presented as mean and standard deviation for continuous variables and as proportions for categorical variables. Differences between ethnic groups and gender were tested using Student's t test. Pearson correlation coefficients were used to evaluate the relationships between BMI and age, sexual maturation, and body composition variables. Multiple linear regression analysis was used to investigate the influence of age, sexual maturation, gender, and ethnicity on the relationship between the percent body fat and BMI. In these analyses, the percent body fat was used as the dependent variable and BMI, age, sexual maturation stage, gender, ethnic group, and their interaction terms were used as independent variables. These analyses were repeated using fat mass as the dependent variable. A P value less than .05 was considered to indicate statistical significance. For the purpose of regression analysis, females were coded as 0, males as 1; blacks were coded as 0′, whites as 1.
The characteristics of the study population are presented in Table1. There were no significant differences by gender and ethnic group for any of the variables.
The correlations of BMI with age, maturation stage, and body composition variables by gender and race are presented in Table2. BMI was significantly and positively correlated with all of the variables in each race-sex group. For the group as a whole, the correlation of maturation stage with BMI was stronger than the correlation between age and BMI.
To evaluate whether the relationship of body fatness with BMI was independent of age or maturation stage, multiple regression analysis was performed (Table 3). In this analysis, the percent body fat was the dependent variable and BMI and either age or stage of maturation were the independent variables. The analyses were performed separately for each race-sex group. When both age and maturation stage are considered in the regression model, the coefficient for maturation stage is significant, although the coefficient for age is not. This suggests that the stage of sexual maturation is a more important determinant of percent body fat than age. The negative regression coefficient for maturation stage indicates that there is a relatively lower body fat percentage in more sexually mature adolescents compared with less sexually mature children of similar BMI.
Next, sex was added to the regression model including BMI and maturation stage as independent variables. This was done to determine whether the relationship between percent of body fat and BMI is significantly different in boys and girls (Table 4). A significant negative regression coefficient was found for both black and white subjects. This indicates that for a similar BMI and maturation stage, boys had a lower percent body fat than girls.
The next step was to add race to the multiple regression model. The regression coefficient for race was significant and indicates that white subjects have a higher percent body fat than blacks for a given BMI after controlling for gender and maturation stage.
The distribution of fat may be an important determinant of the ability of BMI to estimate adiposity. The waist:hip ratio provides a measure of central adiposity. The addition of waist:hip ratio to the regression model resulted in a significant positive regression coefficient for waist:hip ratio and increased the multiple R2from .74 to .77. The final regression model is presented in Table5. These results indicate that gender, race, maturation stage, and waist:hip ratio all have a significant independent effect on the relationship between BMI and body fatness. Further evaluation determined that there was no significant interaction between the class variables in the final regression models. When fat mass (kg) was used as the dependent variable in the multiple regression analysis, the relationships with BMI and the other independent variables were similar to those using percent body fat as the dependent variable.
BMI has been recommended as an appropriate measure of adiposity for use in the clinical setting. It is an attractive measure because it is relatively easy to obtain in a variety of settings and is reliable. In addition, sex-specific national reference data exist for determination of percentiles of BMI.5,15,16 These percentiles can be used for the evaluation of individual children and adolescents to determine their place in the distribution. The use of BMI to evaluate the fatness of children assumes that BMI represents adiposity independent of age, race, and sex, meaning that all subjects of similar BMI have the same degree of fatness, regardless of their age, race, or sex. Gallagher et al and others have demonstrated that this assumption may be erroneous for adults.6 These questions have not been systematically addressed in children and adolescents.
The present study found that the relationship between BMI and body fatness is dependent on maturation stage, race, gender, and waist:hip ratio in children and adolescents. This suggests that caution must be used in studies in which BMI is used for race or gender comparisons of adiposity. These results also indicate that BMI may have limitations for determining adiposity in the clinical setting. The findings in the present study indicate that the stage of sexual maturation is a more important correlate of percent body fat than is age. This means that the percent body fat at a given BMI will differ, depending on the level of sexual maturation. Among subjects who have similar BMI, the child who is more advanced in sexual maturation will have a lower percent body fat. This is different than the findings in adults, in whom there is an increase in relative fatness with aging. For example, Gallagher et al6 found that when comparing older adults with younger adults with similar BMI, the older person will have a greater percent of body weight as fat. The age range for subjects in their study was 20 to 94 years. The timing of the onset of an increase in relative fatness with increasing age is not known precisely but may occur in the late teenage or early young adult years.
We also found that BMI cannot be used as an equivalent measure of fatness in girls and boys or in black and white subjects. Boys and girls differ substantially in body fatness given a similar BMI, with girls having greater total body fat than boys at the same level of BMI. These results are similar to those of several studies of adults in which BMI was shown to underestimate fatness in adult men compared with women.6,17-19 These findings taken together indicate that this gender difference in the relationship of BMI to body fatness may be present throughout life.
We also found an ethnic difference, with whites having higher body fatness for a given BMI than black subjects. The results of studies of ethnic differences in adults have not yielded consistent results. For example, Gallagher et al found no differences by race in the BMI/body fatness relationship in adults.6 Our findings are similar to those of Kleerekoper et al who did find racial differences in the relationship of BMI to percent body fat in older women.20
Finally, the waist:hip ratio was also a significant independent predictor of percent body fat after including the other variables in the model. The waist:hip ratio is an indicator of central obesity, with a higher ratio associated with greater central adiposity and possibly greater visceral adiposity. Our findings suggest that in subjects with similar BMI, those with higher waist:hip ratios have greater percent body fat. Thus, BMI may underestimate the level of fatness in children and adolescents with a central fat distribution. It has been shown that central obesity may be a more important risk factor for cardiovascular outcomes than peripheral obesity. Central obesity is associated with increased insulin resistance, higher circulating insulin levels, elevation of blood pressure, and decreased HDL-cholesterol.21 The results of the present study suggest that the use of BMI to evaluate obesity in the clinical setting may provide a false reassurance, because in individuals with equivalent percent body fat, those with more central obesity will have lower BMI compared with those with more peripheral obesity.
The results of this investigation indicate that there are important factors to consider when using body index as a measure of fatness in children and adolescents, particularly when the objective is to make comparisons across race and gender. When BMI is used in a research or clinical setting to evaluate body fatness, the maturation stage, race, gender, and distribution of fat should be considered in the interpretation of the results.
- Received August 1, 1996.
- Accepted November 18, 1996.
Reprint requests to (S.R.D.) Division of Cardiology, Children's Hospital Medical Center, 3333 Burnet Avenue, Cincinnati, OH 45229.
- BMI =
- body mass index •
- DEXA =
- dual-energy x-ray absorptiometry
- ↵Quetelet LAJ. A treatise on man and the development of his faculties. In: Comparative Statistics in the 19th Century. Edinburgh, Scotland: William and Robert Chambers; 1842
- Himes JH,
- Dietz WH
- Gallagher D,
- Visser M,
- Sepulveda D,
- Pierson RN,
- Harris T,
- and Heymsfield SB
- ↵Lohman TG. Advances in Body Composition Assessment. Champaign, IL: Human Kinetics Publishers; 1992
- Marshall WA,
- Tanner JM
- ↵Tanner JM. Growth at Adolescence. 2nd ed. Oxford, UK: Blackwell Scientific Publications; 1962
- Must A,
- Dallal GE,
- Dietz WH
- ↵Must A, Dallal GE, Dietz WH. Reference data for obesity: 85th and 95th percentiles of body mass index (wt/ht2)—a correction. Am J Clin Nutr. 1991;54:773 (rapid communication)
- Freedman D,
- Srinivasan S,
- Burke G,
- et al.
- Copyright © 1997 American Academy of Pediatrics