Objective. The aim is to describe body composition in relation to body mass index (BMI; body weight/stature2) to provide health care professionals insight into the meaning, significance, and limitations of BMI as an index of adiposity during childhood.
Methods. Data from 387 healthy, white children 8 to 18 years of age from the Fels Longitudinal Study were analyzed. Measurements were scheduled annually and each child was examined 1 to 11 times, totaling 1748 observations. Total body fat (TBF) and fat-free mass (FFM) were determined from hydrodensitometry. Stature and weight were measured using standard methods and BMI and the components of BMI, TBF/stature2, and FFM/stature2 were calculated. Analyses included correlations between BMI and body composition variables; age-related patterns of BMI, TBF/stature2, and FFM/stature2; and annual changes in BMI, TBF/stature2, and FFM/stature2.
Results. Generally, correlations between BMI and body composition variables were strong and significantly different from zero. Means for BMI throughout childhood were similar for boys and girls, although significantly larger values were observed for girls at ages 12 to 13 years. Age-related patterns of TBF/stature2and FFM/stature2 differed between sexes. In each sex, annual increases in BMI were driven primarily by increases in FFM/stature2 until late adolescence, with increases in TBF/stature2 contributing to a larger proportion of the BMI increases in girls than in boys.
Conclusions. Unlike adults, annual increases in BMI during childhood are generally attributed to the lean rather than to the fat component of BMI. Because the properties of BMI vary during childhood, health care professionals must consider factors such as age and sex when interpreting BMI.
The statures and weights of children and adolescents are easily and reliably obtained in a wide variety of clinical settings. As such, the body mass index (BMI; weight [kg]/stature2 [m2]) is often used to determine overweight and obesity, usually by comparison of individuals to age- and sex-specific percentiles from a reference population. Using data from the first National Health and Nutrition Examination Survey, the Expert Committee on Clinical Guidelines for Overweight in Adolescent Preventive Services concluded that adolescents with BMI values >30 kg/m2 or ≥95th percentile for age and sex should be characterized as overweight, while those at or above the 85th percentile but <95th percentile should be considered at risk for overweight.1 A problem with this recommendation is that a specific prevalence of overweight is imposed without regard to health-related outcomes. More recently, participants at the International Obesity Task Force workshop suggested defining overweight and obesity at childhood percentiles that correspond to the adult values for overweight and obesity. The adult BMI values for overweight and obesity, 25 and 30 kg/m2, respectively, correspond to the 80th and 95th percentiles of the National Center for Health Statistics (NCHS) reference values for 18 years olds and are known to be related to morbidity and mortality.2 Although these recommendations were recently accepted, the validity of BMI as a measure of adiposity in children has not been established.
There are well-known limitations regarding the use of BMI. For example, BMI is generally defined in adults as an index of adiposity that is largely independent of stature; however, this property of BMI in adults does not necessarily hold true in children.3 Also, despite high correlations between BMI and total body fat (TBF) and percent body fat (%BF),4–7 BMI is also correlated with fat-free mass (FFM).3 In children, these relationships between BMI and the fat and fat-free components of the body are further complicated by varying growth rates and maturity levels.5,8 In addition, validity studies using BMI to identify children with excess adiposity have generally documented low to moderate sensitivities, which indicate only a poor to fair identification of those who are truly overweight, as determined from %BF.9–11 These findings imply that many children who are at risk for the health-related problems accompanying overweight and obesity may not be effectively targeted for intervention programs designed to treat or prevent the progression of obesity when selection for these programs is based solely on BMI. Collectively, the limitations of BMI foster ambiguous interpretations when using this index in children such that a more extensive examination of childhood body composition in relation to BMI is necessary to more fully understand the utility of BMI during childhood. This examination is especially important because the clinical use of BMI is likely to become more widespread with the recent release of BMI percentile charts by the NCHS.12
Because weight consists of TBF and FFM, BMI can be partitioned into the fat and lean components of TBF/stature2 and FFM/stature2, respectively. The examination of TBF/stature2 and FFM/stature2 across age in relation to BMI allows identification of the body composition changes that affect changes in BMI with age. Therefore, to provide health care professionals with information to assist the interpretation of BMI in children and adolescents in terms of body composition, we examined: 1) correlations between BMI and body composition components; 2) age- and sex-related patterns in BMI and the fat and lean components of BMI throughout childhood and adolescence; and 3) the annual changes in BMI and the BMI components, TBF/stature2 and FFM/stature2.
Data from 387 healthy white children 8 to 18 years of age, a subsample of 201 boys and 186 girls selected from the ongoing Fels Longitudinal Study,13 were used for these analyses. Participants in the Fels Study reside in 43 of the 50 United States, though approximately two thirds of them live in Midwestern states. The NCHS US growth charts in use from 1977 to 2000 for children from birth to 36 months of age are based exclusively on data from participants in the Fels Longitudinal Study.14
Each child in the present subsample had concomitant measurements of stature, body weight, and body density. Stature and body weight were measured using the recommendations of the Anthropometric Standardization Conference15 and BMI was calculated as weight (kg)/stature2 (m2). Body density was determined from hydrodensitometry, and TBF (kg), FFM (kg), and %BF were calculated using Lohman's multicomponent model.16 This model allows for age and sex variation in the composition of FFM from 7 to 25 years of age. TBF and FFM were divided by stature2 to obtain the fat and lean BMI component variables. These are TBF/stature2(kg/m2) and FFM/stature2(kg/m2); that is BMI = weight/stature2 = (TBF+FFM)/stature2 = TBF/stature2 + FFM/stature2.
In the Fels Longitudinal Study, body composition measurements of children are scheduled annually until 18 years of age. Participants in this subsample had the required data from 1 to 11 examinations. Thus, these mixed longitudinal data, containing both cross-sectional and serial observations resulted in a total of 1748 observations available for analyses. The number of examinations for each participant depended on factors such as the age of the child when hydrodensitometry was implemented in the Fels Longitudinal Study (1976), the age at which the child first agreed to perform the hydrodensitometric procedure, and whether scheduled examinations were missed. Age 8 years is typically the youngest age at which hydrodensitometry data are collected in the Fels Longitudinal Study. Informed consent was obtained from each participant and from a parent or guardian for those younger than 18 years of age. All procedures used in this investigation were approved by the Institutional Review Board of Wright State University.
Age- and sex-specific means and standard deviations for stature, body weight, BMI, %BF, TBF, and FFM were calculated. When data were distributed normally, a Student's t test was used to determine significant sex differences within each age group; a Wilcoxon test was performed when data were not normally distributed. Normality of the distributions within age groups was determined using the Shapiro-Wilk test.17 Pearson product moment correlations were calculated to determine the linear relationship between BMI and the anthropometric and body composition variables of interest. Each age and sex group included only one observation per child; therefore, multiple samples of independent observations were used for statistical comparisons between sexes and for significance testing on correlations at each annual age. Statistical comparisons were not performed across age. Significance for the t test, Wilcoxon analyses, and Pearson correlations was adjusted for the multiple comparisons across age using Bonferroni's adjustment procedure. To obtain an overall α level of .05, the significance level for each comparison was set atP ≤ .005.
Data from children with consecutive examinations occurring within .75 and 1.25 years of each other were analyzed to determine annual changes in BMI, TBF/stature2, FFM/stature2, weight, TBF, and FFM. Data from 291 children met these criteria; these data totaled 1150 observations. All calculations of annual changes were adjusted for length of time between examinations. The statistical procedures performed in these analyses were conducted using SAS version 6.12 (SAS Institute, Inc, Cary, NC).
Means for weight and stature were compared with NCHS reference data14 for general descriptive purposes. Age-specific means for stature and weight for both boys and girls in the Fels sample were between the 50th and 75th percentiles of US national data.14
Correlations of BMI with weight, stature, %BF, TBF, and FFM are shown in Table 1. BMI was positively and significantly correlated with weight at all ages in each sex and was positively and significantly correlated with stature at ages 10 to 14 years in boys and at age 9 years in girls. Correlations were moderate to high between BMI and %BF and between BMI and TBF for boys at each age, ranging from .64 to .85 for %BF and .83 to .94 for TBF; these correlations indicate that BMI explains 41% to 88% of the variance in %BF or TBF. The corresponding correlations for girls were generally lower than those for boys, ranging from .37 to .78 for %BF and .67 to .90 for TBF, with BMI explaining 14% to 81% of the variance in %BF or TBF. Correlations between BMI and FFM ranged from .25 to .78 in boys and .39 to .72 in girls; BMI explained >25% of the variation in FFM at most ages. Between ages 9 and 18 years, all correlations of BMI with %BF, TBF, and FFM were significantly different from zero for each sex.
Correlations between stature with TBF, FFM, TBF/stature2, and FFM/stature2 are shown in Table 2. Stature was significantly correlated with TBF in boys from 10 to 14 years of age, and in girls from 9 to 13 years. As expected, FFM was positively and significantly correlated with stature at every age in both boys and girls. In boys, stature was also positively and significantly correlated with TBF/stature2 at ages 10 to 12 years and with FFM/stature2 at ages 12 to 15 years. Corresponding correlations in girls were not significant for TBF/stature2 at any age and only at age 12 years for FFM/stature2.
Mean values for BMI generally increased with age in both sexes (Fig 1A). Girls tended to have larger means for BMI than boys between ages 11 and 16 years; however, significant sex differences between means for BMI occurred only at ages 12 and 13 years after accounting for multiple comparisons. BMI means for boys and girls were between the 50th and 85th percentiles for age and sex from the first National Health and Nutrition Survey18,19 and were generally nearer the 50th percentile.
With few exceptions, the means for TBF/stature2for girls increased steadily with age (Fig 1B). In boys, means for TBF/stature2 increased slightly from 8 to 14 years of age, after which they decreased from 14 to 16 years of age and then increased slightly but did not reach the mean level observed at age 14 years by age 18 years. Means for TBF/stature2 were larger for girls than for boys at all ages, with significant differences between means at ages 10 to 18 years.
The pattern of means for FFM/stature2 with age differed considerably from that observed for TBF/stature2 (Fig 1C). Boys and girls showed similar small increases in FFM/stature2 until age 14 years, after which there were only small increases in the means for girls, while those for boys increased markedly until 17 years of age. Significant sex differences between means for FFM/stature2 were observed at ages 15 to 18 years, with boys exhibiting larger values than girls.
Annual changes in BMI, TBF/stature2, and FFM/stature2 for boys are shown in Fig 2A. Between the ages of 9 and 12 years, increases in the BMI of boys were accompanied by increases in both FFM/stature2 and TBF/stature2. Between 8 and 9 years and between 12 and 17 years of age, however, mean annual increases in BMI resulted exclusively from increases in FFM/stature2, with TBF/stature2 decreasing at these ages. Between 17 and 18 years of age, increases in both FFM/stature2 and TBF/stature2 contributed to the increase in BMI. The annual changes in weight, TBF, and FFM are shown in Fig 2B. Annual increases in weight for boys were largely explained by increases in FFM. Mean changes in TBF were slightly negative between 14 and 15 and between 16 and 17 years of age.
The corresponding data for girls (Fig 3A) show that annual changes in both FFM/stature2 and TBF/stature2 reflected the net change in BMI from 8 to 9 years and 10 to 16 years of age; FFM/stature2 accounted for most of the mean increase in BMI from ages 10 to 14 years. As girls approached adulthood, ie, between ages 16 and 18 years, BMI changes were almost exclusively accounted for by increases in TBF/stature2, although slight positive changes in FFM/stature2 were also observed. From 9 to 10 years of age, a decrease in FFM/stature2 was exhibited. Consequently, despite a large increase in TBF/stature2, the increase in BMI was smaller than that in the annual intervals immediately before or after 9 to 10 years. These data correspond closely to each of the annual changes observed for weight, TBF, and FFM throughout childhood in girls (Fig 3B). Generally, increases in FFM accounted for most of the mean increase in weight until age 16 years, after which weight increases were largely explained by increased TBF.
The age- and sex-specific correlations of BMI with TBF and %BF were moderate to high in the present study and are similar to the correlations of childhood BMI with fat indices reported by others using anthropometry,20 dual energy x-ray absorptiometry, and hydrodensitometry.6 The strength of the associations between BMI with TBF and %BF indicates that BMI is a suitable measure of adiposity. In each sex, age- and sex-specific correlations between BMI and FFM were generally not as high as the correlations for the fat variables, but with one exception, all were high and significantly different from zero (P ≤ .005). Interestingly, the correlations between BMI and FFM were stronger than the correlations between BMI and %BF in girls between 8 and 13 years of age. Thus, these correlations show that both fat and lean components of body mass must be considered when interpreting BMI.
Moderate but statistically significant correlations were observed between BMI and stature in boys at ages 10 to 14 years, and in girls at age 9 years. Although correlations between BMI and stature were not significant in girls at ages 10 to 13 years using the Bonferroni-adjusted P value of .005, the correlations indicate a trend toward a weak relationship between BMI and stature in girls at these ages. Thus, unlike in adults where BMI is generally uncorrelated with stature, the present findings show that BMI and stature are related in children, particularly during early adolescence in boys; in girls, the relationship between BMI and stature is less clear. Health care professionals should, therefore, note that children and younger adolescents, particularly boys, who are tall for their ages may have large BMI values as a consequence of stature rather than excess adiposity. It should also be kept in mind, however, that children with increased adiposity are generally taller than their lean peers at the same age and sex.21 In girls 9 to 14 years of age, Himes and Roche22 showed that statures were increasingly larger with increasingly higher levels of adipose tissue thickness until a threshold was reached beyond which further differences between statures were no longer observed.
Although mean BMI values with age were similar in boys and girls, sex differences were clearly evident in the age-related patterns of the BMI components. In girls, the patterns of means for FFM/stature2 showed little increase after age 14 years, whereas those for TBF/stature2 generally showed a linear increase with age until 17 years. In boys, means for FFM/stature2 showed a rapid increase during adolescence with a concomitant decrease in means for TBF/stature2. Thus, as shown by data presented here, the contribution of each BMI component to the overall age-related trend in BMI differs between sexes. One should keep in mind, however, that the relationships observed for group data may not reflect changes within individuals. Because mixed longitudinal data were used for these analyses, these data more closely approximate patterns of individual growth than do group data alone.
The decrease observed in TBF/stature2 of boys corresponds to longitudinal findings for fat patterns during childhood.8,23,24 Gasser et al23,24 reported pubertal decreases in the skinfold thickness of boys that coincide with the peak of the pubertal growth spurt. In the Fels sample, the pubertal growth spurt in boys, represented by the age at maximal velocity of BMI, occurs at age 14.3 years25 and corresponds to the age of 14 years at which decreases in mean TBF/stature2 values were first observed. Gasser et al23,24 also reported decreases in specific skinfold thickness in girls during the pubertal growth spurt, but corresponding decreases in mean TBF/stature2 were not observed for girls in the present investigation. In addition, sex differences in TBF/stature2 were observed at most ages in the current study. Although sex differences in TBF during puberty are well known, our data are generally consistent to those of Taylor et al26 who, using dual energy x-ray absorptiometry, reported adiposity differences between sexes before puberty.
The annual changes in TBF/stature2 and FFM/stature2 for individuals show that mean age-related increases in the BMI of boys are primarily driven by increases in the lean component of BMI. In fact, FFM/stature2 exclusively drives the BMI increases in boys from ages 12 to 17 years and corresponds to a time when changes in TBF/stature2 are small. Annual changes in the BMI components of girls indicate that, generally, both TBF/stature2 and FFM/stature2 drive the increases observed in BMI until age 16 years, after which these increases are almost exclusively driven by increases in the fat component of BMI.
One interesting finding was the change in BMI observed for girls from 9 to 10 years of age (Fig 3A). This change was the result of a dramatic increase in TBF/stature2 and corresponding decrease in FFM/stature2. This apparently surprising result is explained by a disproportionate increase in stature relative to FFM, resulting in a negative change in FFM/stature2. For further clarification, the annual change of BMI between ages 9 and 10 years in relation to the components of BMI was calculated as follows:Thus, a large increase in stature relative to a small increase in FFM can yield a negative value for FFM/stature2because a large denominator at age 10 years relative to age 9 years can make the ratio at this age smaller than that at age 9 years. This is true despite an absolute increase in FFM between ages 9 and 10 years (Fig 3B). Thus, the disparity between increases in stature and FFM accounts for the mean decrease in FFM/stature2observed in girls between ages 9 and 10 years. This finding may be explained by a lag in muscle growth behind that of bone elongation and linear growth.
The analyses for annual changes suggest that BMI is an appropriate measure of adiposity in late adolescence, particularly in girls. As can be seen in Fig 3A, changes in FFM/stature2contributed only minimally to the changes observed in BMI in girls between 16 and 18 years of age, whereas the fat component of BMI exerted the predominate effect on the changes. A similar finding was present in boys between 17 and 18 years of age, although the contribution of the lean component to the change in BMI was larger than that observed for girls.
The annual changes of the BMI components and concomitant changes in BMI have important clinical implications. Namely, when interpreting changes in BMI in children and adolescents from one year to the next, increases in BMI should not be simply interpreted as increases in adiposity, but also as changes in body composition that are associated with growth. As seen from the data presented here, age-related increases in BMI are driven by increases in the lean component of BMI (FFM/stature2), the fat component of BMI (TBF/stature2), or both BMI components. The extent to which each component contributes to the change in BMI depends on the sex and age of the individual. Because maturity also affects body composition,5,8 maturational stage should also be considered in clinical assessment. At similar BMI values, %BF will be lower in the more mature children.5 In addition, Schaefer et al27 showed that in boys, obesity was predicted reasonably well by BMI in the upper third of %BF percentiles, although not in the lower two thirds of the sample. Because children in the present study had mean statures and weights within the third quartile of national data, it is reasonable to speculate that BMI may be an even more effective indicator of adiposity at the upper ranges of the distribution.
Despite the shortcomings of BMI, health care professionals can effectively use BMI to monitor changes in body mass. Although specific BMI values may be associated with a number of varying body compositions depending on age, sex, and individual differences, comparisons with age- and sex-specific national percentiles can provide relative BMI values that can be easily tracked over time. Upward deviations across rather than along BMI percentiles may indicate disproportionate increases in weight relative to stature that are likely attributed to excess adiposity. A few individuals may have large BMI values because of large amounts of lean versus fat tissue, but the moderate to high correlations between BMI and the body fat variables cannot be overlooked. Therefore, the use of age- and sex-specific percentiles to compare the BMI values of pediatric patients is a reasonable approach for screening those at risk for excess adiposity. Such screening is especially useful for children at higher percentiles because these children are likely to become obese in adulthood.28However, using age- and sex-specific percentiles does not overcome all the limitations inherent in the BMI measurement.
In clinical assessments of children and adolescents, care should be taken to prevent misclassification of adiposity status to avoid undue psychological trauma. If indicated by the presence of other health risk factors, additional measurements of body composition should be taken in conjunction with BMI to assess adiposity. However, because high BMI values during childhood are associated with adverse health effects in adulthood as well as with clinical problems during childhood and adolescence,29–31 additional assessments of body composition may not be cost-effective for individuals at the uppermost percentiles. This is especially true in light of the simplicity and reliability with which BMI measurements are obtained.
The most important finding of this investigation was the determination of body composition changes during annual intervals. These data uniquely demonstrate that increases in BMI throughout childhood are driven primarily by increases in FFM/stature2, particularly in adolescent boys. Increases in TBF/stature2 were generally observed before and after adolescence in both boys and girls, with TBF/stature2 generally contributing to a greater proportion of the annual BMI increases in girls than in boys throughout childhood. Thus, the extent to which each component contributes to the change in BMI depends on the sex and age of the individual. These data should assist health care professionals in their interpretations of the BMI status of their pediatric patients.
This work was supported by the National Institute of Child and Human Development, National Institutes of Health Grant R01-12252.
- Received December 7, 1999.
- Accepted May 25, 2000.
Reprint requests to (L.M.M.) Division of Human Biology, Wright State University School of Medicine, 3171 Research Blvd, Kettering, OH 45420-4014. E-mail:
- BMI =
- body mass index •
- NCHS =
- National Center for Health Statistics •
- TBF =
- total body fat •
- %BF =
- percent body fat •
- FFM =
- fat-free mass
- Himes J,
- Dietz W
- Bellizzi MC,
- Dietz WH
- Garn S,
- Leonard W,
- Hawthorne V
- Goran MI,
- Driscoll P,
- Johnson R,
- Nagy TR,
- Hunter G
- Daniels SR,
- Khoury PR,
- Morrison JA
- Roche A,
- Siervogel R,
- Chumlea W,
- Webb P
- Lazarus R,
- Baur L,
- Webb K,
- Blyth F
- Kuczmarski RJ, Ogden CL, Grummer-Strawn LM, et al. CDC Growth Charts: United States. Advance data from vital and health statistics. No. 314. Hyattsville, MD: National Center for Health Statistics; 2000
- Roche AF. Growth, Maturation, and Body Composition: The Fels Longitudinal Study 1929–1991. Cambridge, England: Cambridge University Press; 1992
- Hamill P, Drizd T, Johnson C, Reed R, Roche A. NCHS growth curves for children, birth to 18 years, United States. Vital Health Stat 11. 1977;165
- Gordon CC, Chumlea WC, Roche AF. Stature, recumbent length, and weight. In: Lohman TG, Roche AF, Martorell R, eds. Anthropometric Standardization Reference Manual. Abridged ed. Champaign, IL: Human Kinetics; 1991:3–7
- Lohman T
- Shapiro SS,
- Wilk MB
- Must A,
- Dallal G,
- Dietz W
- Must A,
- Dallal G,
- Dietz W
- Guo SS,
- Chumlea WC
- Freedman DS,
- Dietz WH,
- Srinivasan SR,
- Berenson GS
- Must A, Jacques P, Dallal G, Bajema C, Dietz W. Long-term morbidity and mortality of overweight adolescents: a follow-up of the Harvard Growth Study of 1922 to 1935. N Engl J Med. 1992:327:1350–1355
- Copyright © 2001 American Academy of Pediatrics