Funding agencies: This research was supported by the American Diabetes Association.
Relationships between direct and indirect measures of central and total adiposity in children: What are we measuring?
Article first published online: 11 JUN 2013
Copyright © 2013 The Obesity Society
Volume 21, Issue 10, pages 2055–2062, October 2013
How to Cite
Bigornia, S. J., LaValley, M. P., Benfield, L. L., Ness, A. R. and Newby, P.K. (2013), Relationships between direct and indirect measures of central and total adiposity in children: What are we measuring?. Obesity, 21: 2055–2062. doi: 10.1002/oby.20400
Disclosure: The authors declared no conflict of interest.
Relevant conflicts of interest/financial disclosures: Nothing to report. Full financial disclosures and author notes may be found in the online version of this article.
- Issue published online: 5 OCT 2013
- Article first published online: 11 JUN 2013
- Accepted manuscript online: 20 MAR 2013 02:14AM EST
- Manuscript Accepted: 17 JAN 2013
- Manuscript Received: 23 JUL 2012
- American Diabetes Association
Objective: The relationship between central and total fat measured by anthropometry, dual energy X-ray absorptiometry, and magnetic resonance imaging (MRI) with each other and systolic blood pressure (SBP) was examined.
Design and Methods: Participants of the Avon Longitudinal Study of Parents and Children were examined at ages 9, 11, 13, and 15 years (n = 3,796-6,567). MRI was available on a subset of children at 11 (n = 156) and 13 (n = 95).
Results: Body mass index (BMI) and waist circumference (WC) were highly correlated (r = 0.84-0.91, across ages), and total body fat mass (TBFM) and trunk fat mass (TFM) were very strongly correlated (r ≥ 0.98). Among boys, BMI vs. WC explained a similar degree of variation in TBFM and TFM (41-71% vs. 43-76%, across age and overweight groups); in girls, BMI accounted for 62-73% variance and WC 47-69%. Adiposity measures were generally similarly correlated with SBP within age groups. Further, the relationship between intra-abdominal adipose tissue (IAAT) volume and adiposity measures did not vary greatly at 11 (0.65-0.67) and 13 (0.64-0.67).
Conclusions: BMI and WC contain a large amount of overlapping information as evidenced by their high correlation and similarly sized associations with fat mass, SBP, and IAAT. This suggests that WC may be an inadequate marker of central adiposity during childhood.
In adults, central fat distribution is linked to obesity-related comorbidities [1, 2] and mortality . Over the past several decades, the prevalence of childhood obesity has increased  and evidence suggest that it tracks into adulthood . Atherosclerotic lesions have been identified in children and related to cardiovascular disease (CVD) risk factors . Further, risk factor levels in childhood are predictive of those in adulthood . Together, these findings suggest that obesity-related comorbidities later in life can be influenced by risk factors in childhood, stimulating research interest in the metabolic impact of fat mass and fat distribution in children.
Available evidence in children supporting a relationship between central fat and metabolic risk is complicated by the similar effect sizes of a variety of metabolic outcomes with central and total adiposity measures . These patterns are evident when examining associations between risk factors for CVD and/or insulin resistance with adiposity assessed by anthropometry [8, 10, 14, 15], dual energy X-ray absorptiometry (DXA) [9, 12, 15], magnetic resonance imaging (MRI), and/or computed tomography (CT) [9, 12, 13]; both MRI and CT can directly quantify intra-abdominal adipose tissue (IAAT). Further, observations in children and adolescents show strong correlations between body mass index (BMI) and waist circumference (WC) (r = 0.91)  and total body fat mass (TBFM) and BMI (r = 0.84-0.97) , and moderate to strong correlations between IAAT and BMI or WC or TBFM (r = 0.48-0.88) . The high inter-relatedness between total and central adiposity measures and the comparable effect sizes with various risk factors in children suggests that little additional information regarding metabolic risk may be gained from the use of central vs. total fat measures [26, 27].
Most previous studies that have reported associations between central and total fat measures did so in groups with large age ranges (e.g., ages 8-17 years  and 10-14 years ), which makes it difficult to determine if these correlations remain stable from childhood to adolescence. In light of the natural changes in fat distribution that occur during normal growth and development and differences in body composition between boys and girls, it is expected that these relationships may vary across subgroups and understanding when measures of central fat become meaningful is clinically important.
Our objectives were to 1) examine the relationships between measures of central and total adiposity obtained by anthropometry, DXA, and MRI (11 and 13 years only) in children at 9, 11, 13, and 15 years of age and 2) compare how these were associated with a clinically meaningful risk factor, systolic blood pressure (SBP). Blood pressure is adversely associated with increased  adiposity, relates to subclinical CVD in children , and predicts future cardiovascular events . We hypothesized that measures of central and total adiposity would be highly correlated at each time-point and would differ by sex and weight groups. We also hypothesized that these adiposity measures would be similarly associated with SBP and IAAT by MRI.
The Avon Longitudinal Study of Parents and Children (ALSPAC) is a population-based prospective cohort study examining environmental effects on the health and development of children . In 1990, pregnant women (n = 14,541) residing in the former County of Avon located in South-West England with expected delivery dates between April 1991 and December 1992 were recruited. At age 7, in-house clinics were conducted on ∼8,000 children to collect more detailed information. For this study, we examined children with complete anthropometric and DXA adiposity measures. Participants were excluded from specific analyses conducted at ages 9, 11, 13, or 15 years if their age at a given clinic was >|0.5| years of the group mean to ensure that ages across adjacent clinics did not overlap. Sample sizes were 6,495 (91.8% of total sample), 6,567 (96.7%), 5,627 (96.1%), and 3,796 (93.7%) at ages 9, 11, 13, and 15, respectively. Tanner stage for pubertal growth was available on 5,323 (82.0%), 5,046 (76.8%), 4,296 (76.3%), and 2,555 (67.3%) children at 9, 11, 13, and 15, respectively. SBP was measured on 6,405 (98.6%), 6,468 (98.5%), 4,920 (87.4%), and 3,689 (97.2%) participants at the 4 ages, respectively.
A subset of children participated in MRI visits at 11 (n = 161) and 13 (n = 99). These individuals were a random sex (50% each) and BMI (above and below the mean) stratified sample of participants at the 11 years clinic and expressed interest in further sub-studies . At 11 years, there were no detectable differences in anthropometric and DXA adiposity measures between MRI and non-MRI participants. At 13 years, the MRI participants (vs. non-MRI) were younger (165.2 vs. 165.8 months, P = 0.001), had lower percent body fat (%BF) (22.3 ± 9.8% vs. 24.5 ± 10.3%, P = 0.03), and had a higher proportion of males (63% vs. 49%, P = 0.006). Ethical approval was obtained from the ALSPAC Law and Ethics Committee, the Local Research and Ethics Committees, and the Boston University Medical Center institutional review board.
Anthropometry and imaging
WC was measured at the midpoint between the lowest rib and the top of the iliac crest. Weight was obtained using a Tanita Body Fat Analyser (Model TBF 305) and height by a Harpenden Stadiometer. Using International Obesity Taskforce weight categories , participants were categorized as either overweight (OWT) or normal weight (NWT) if their BMI corresponded to a predicted 18 years BMI ≥ 25 kg/m2 or BMI < 25 kg/m2, respectively. Body composition was obtained by DXA by a Lunar Prodigy narrow fan beam densitometer. Images were reanalyzed as necessary to ensure the adequate positioning of borders defining the pre-established regions (i.e., head, arms, legs, and trunk). Trunk fat mass (TFM) included the chest, abdomen, and pelvic regions. MRI data were collected using a Philips Intera Pulsar 1.5 T system . Adipose tissue volume was obtained from 5 to 20 axial images of 10-mm-slice thickness within the abdominal and pelvic region (10 cm deep cylinder). Adipose tissue exhibits a high intensity signal allowing for the manual outlining of subcutaneous abdominal and visceral areas on a slice by slice basis from which IAAT volume (cm3) was calculated. IAAT included only that located within the abdominal cavity and excluded subcutaneous areas. At the MRI visit, children were 0.47 ± 0.17 years at 11 and 0.27 ± 0.17 years at 13 older than at the corresponding study visits.
Puberty and blood pressure
Tanner stage (5 point scale) for pubic hair growth was self-reported by postal questionnaire sent to the parent of the child . Questionnaires contained schematic drawings and descriptions of secondary sexual characteristics. Self-reported Tanner staging correlates with physical examination [34, 35] and relates to hormonal maturity . Tanner staging was collapsed to pubertal stage (pre = 1, early = 2-3, and late = 4-5). Pubertal assessment and attendance at the corresponding clinic visit were not necessarily concurrent so analyses including puberty were restricted to subjects who completed questionnaires <|12| months of the clinic to obtain an accurate measure of maturity. SBP was measured using a Dinamap 9301 Vital Signs Monitor (Morton Medical London). Before blood pressure attainment, children were given a simple explanation of what would happen during the blood pressure measurement session. Two readings of blood pressure were obtained and the mean value used in all analyses.
To prepare the data for analysis, we conducted several tests to observe the distributions of all of our variables and check for statistical and biological outliers. Scatter-plots and residuals were examined for linearity between measures. Observations with studentized residuals >|3| that were distinctly separate from other values were considered outliers and excluded. From this, five and four observations at 11 and 13 years, respectively, were excluded from analyses of children with MRI data.
Descriptive analyses examined differences in demographic characteristics between boys and girls using the independent t, Wilcoxon-Mann-Whitney, or the chi-square tests as appropriate. BMI, WC, TBFM, TFM, and IAAT were skewed and the log-transformation of these measures used in correlation and regression analyses. We tested the strength of associations between adiposity measures in the full cohort using Pearson's partial correlation coefficients adjusted for age, sex, and height; %BF exhibited curvilinear relationships; thus, Spearman's correlation coefficients were used for those analyses.
We determined the contribution of sex to the strength of association between adiposity measures using multivariable linear regression models at each age. Specifically, BMI, WC, TBFM, TFM, and %BF were used as dependent variables in sets of height-, age- and sex-adjusted regression models with one of the other fat measures set as the independent variable. For example, we fit the model TBFM = β0 + β1[BMI] + β2[age] + β3[height] + β4[sex]. Previous reports [17, 36] suggest that infer-fat associations strengthen with greater BMI; therefore, analyses were stratified by OWT status. There was evidence of effect modification by sex (P < 0.05) for the majority of associations and subsequent analyses were stratified.
The amount of variation in an adiposity measure explained by another was determined in sex- and weight-stratified and age-, height-, and puberty-adjusted linear regression models using the modeling strategy described above. Partial variances were reported, which represent the unique contribution of an adiposity measure to the variability in another fat measure in models containing terms for age, height, and pubertal stage. The partial variances of pubertal stage in each model were explored to examine the influence of maturity level on these inter-fat associations. We used Pearson's correlation coefficients adjusted for age and height to examine how central and total fat measures related to SBP.
Finally, we examined the association of IAAT with anthropometric and DXA fat measures and SBP in sex-stratified, age- and height-adjusted analyses using partial Pearson's correlation coefficients in the MRI subset. Due to small sample sizes (n = 156 at 11 years and n = 95 at 13 years), these correlations were not weight-stratified. For analyses including pubertal stage or SBP or IAAT, samples were restricted to participants with complete data for these variables. Confidence limits for correlations were derived as follows: 1) Fisher's z-transformation of the correlation, r, 2) calculate the two-sided CI limits using the z-transformed value, and 3) back transform values derived in step 2. The CI can be calculated in SAS using the Fisher option in the Proc Corr procedure. Data were analyzed using SAS software package (version 9.2; SAS Institute, Cary, NC). Alpha was set at 0.05 for all analyses.
WC, BMI, TBFM, and TFM increased across ages (Table 1). Boys had higher WC compared with girls from 9 to 13 years (P < 0.05) (data not shown). Conversely, girls had larger %BF and BMI at all ages (P < 0.0001). The majority of participants were prepubescent at 9 and late-pubescent at 15 years. Inter-fat partial correlations adjusted for age at clinic, sex, and height were moderate to very strong at all ages (P < 0.0001, Figure 1). Correlations between BMI and WC remained stable from 9 to 11 years (r = 0.91-92) with attenuation at 13 (r = 0.88) and 15 years (r = 0.84). A similar pattern was observed for TBFM and TFM correlations with BMI and WC. Conversely, TBFM, TFM, and %BF correlations were consistent across ages. WC and BMI were similarly associated with TBFM and %BF within age groups.
|Age at clinic, y|
|Sample characteristics||n = 6,495||n = 6,567||n = 5,627||n = 3,796|
|Age, y||9.8 ± 0.2||11.7 ± 0.2||13.8 ± 0.2||15.4 ± 0.2|
|Height, cm||139.6 ± 6.2||150.7 ± 7.1||163.3 ± 7.7||169.2 ± 8.4|
|Weight, kg||33.0 (29.4-38.2)||41.6 (36.4-49.0)||52.8 (47.0-60.4)||59.2 (53.4-66.3)|
|BMI, kg/m2||17.0 (15.7-19.0)||18.3 (16.6-20.9)||19.6 (17.9-22.0)||20.6 (19.0-22.7)|
|Waist, cm||61.0 (57.4-66.3)||66.0 (61.7-73.0)||70.0 (65.7-76.1)||75.0 (70.7-81.0)|
|Body fat, %||22.4 (16.2-29.9)||24.6 (18.3-32.6)||24.0 (16.0-32.0)||24.5 (14.5-32.2)|
|Total body fat, kg||7.2 (4.8-11.0)||9.9 (6.7-15.3)||12.0 (7.6-18.2)||13.5 (8.3-19.4)|
|Trunk fat, kg||2.7 (1.7-4.4)||4.0 (2.5-6.6)||5.1 (3.0-8.2)||6.0 (3.5-9.1)|
|Pubertal stage, % (pre/early/late)c|
|Systolic blood pressure, mm Hgd||102.6 ± 9.3||105.5 ± 9.9||106.4 ± 9.4||123.0 ± 10.8|
The contribution of sex to these inter-fat associations was explored using multivariable linear regression (data not shown). Among NWT participants in BMI models predicting TBFM, sex explained 10.6%, 7.4%, 19.4%, and 34.2% variation in fat mass at 9, 11, 13, and 15 years, respectively; it accounted for less TBFM variation in the OWT group at 2.6%, 1.4%, 3.8%, and 7.5%. In similar models with WC as the independent variable, among NWT individuals sex explained 21.5%, 14.5%, 35.2%, and 44.5% TBFM variance at 9, 11, 13, and 15 years, respectively; in the OWT group, it accounted for 11.9%, 10.8%, 19.6%, and 21.8%, respectively. Results were very similar in models with TFM as the dependent variable.
In most circumstances, the variance in BMI, WC, TBFM, and TFM explained by select adiposity measures decreased from 9 to 15 years within sex and weight groups (Table 2). In boys, WC and BMI explained a comparable amount of TBFM variance across ages. In girls, BMI accounted for more TBFM variation compared with WC: 9.6-22.5% more for NWT and 9.3-21.0% for OWT across ages. Sex- and weight-specific trends in TBFM variance explained by select adiposity measures were similar to those in %BF analyses. However, the variation in %BF accounted for by the other adiposity measures was lower in comparison. BMI and WC explained a greater degree of variation in TBFM and TFM among OWT vs. NWT boys (3.0-24.5% more), whereas among girls this gap was smaller. Between OWT and NWT girls, differences in TBFM or TFM variances accounted by BMI or WC ranged from −8.0% to 8.0%. Variance in fat measures explained by pubertal stage across ages ranged from 0.0% to 11.6% for NWT boys, 0.0% to 6.6% for OWT boys, 0.0% to 5.9% for NWT girls, and 0.0% to 3.3% for OWT girls (data not shown).
|Partial variance (%) in adiposity explained by select adiposity measures|
|Normal weight||Overweight||Normal weight||Overweight|
|Age at clinic, y||9||11||13||15||9||11||13||15||9||11||13||15||9||11||13||15|
SBP correlations with fat measures at ages 9 and 11 years were similar to one another within sex and weight groups (Figures 2 and 3). Among boys, SBP correlations between the adiposity measures became more heterogeneous (Figure 2) after age 11, whereas among girls associations remained similar (Figure 3). For example, correlations among NWT boys ranged from 0.14 to 0.20 at 11 years and 0.06 to 0.17 at 13 years, whereas among NWT girls, correlations ranged from 0.16 to 0.21 and 0.18 to 0.23 at 11 and 13 years, respectively.
IAAT volume among boys at 11 and 13 were median 73 (inter-quartile range 50-124) cm3 and 135 (93-207) cm3; among girls, values were 105 (71-199) cm3 and 152 (132-199) cm3. Correlations between IAAT and BMI, WC, TBFM, and TFM were moderate to strong at 11 and 13 years (Table 3). Correlations with anthropometric and DXA measures were similar, though in girls at 13 years, WC was more correlated with IAAT (r = 0.60) compared with the other adiposity measures (r = 0.42 for BMI, TBFM, and TFM). TBFM and TFM correlations with IAAT were nearly identical in boys and girls at each age.
|Age at clinic, y|
|r (CI)||r (CI)|
|Total Cohortb||n = 156||n = 95|
|BMI||0.66 (0.56, 0.74)||0.64 (0.50, 0.75)|
|WC||0.67 (0.57, 0.75)||0.60 (0.45, 0.72)|
|TBFM||0.65 (0.55, 0.73)||0.67 (0.54, 0.77)|
|TFM||0.65 (0.55, 0.73)||0.67 (0.54, 0.77)|
|SBPc||0.17 (0.01, 0.32)||0.29 (0.09, 0.48)|
|Boys||n = 88||n = 60|
|BMI||0.61 (0.46, 0.73)||0.72 (0.56, 0.82)|
|WC||0.62 (0.46, 0.73)||0.67 (0.49, 0.79)|
|TBFM||0.59 (0.43, 0.71)||0.71 (0.55, 0.82)|
|TFM||0.57 (0.41, 0.7)||0.70 (0.54, 0.81)|
|SBP||0.12 (−0.1, 0.33)||0.32 (0.06, 0.54)|
|Girls||n = 68||n = 35|
|BMI||0.68 (0.52, 0.79)||0.42 (0.10, 0.67)|
|WC||0.72 (0.58, 0.82)||0.60 (0.32, 0.78)|
|TBFM||0.71 (0.57, 0.81)||0.42 (0.09, 0.67)|
|TFM||0.73 (0.59, 0.83)||0.42 (0.09, 0.67)|
|SBP||0.19 (−0.06, 0.41)||0.15 (−0.23, 0.48)|
In support of our hypothesis, indirect and direct measures of total and central adiposity were strongly related to each other in this study of children during adolescence. Specifically, WC, a widely used indicator of central adiposity, was highly correlated with BMI, suggesting that these measures contain a large amount of overlapping information. In support of this, both explained comparable amounts of variance in directly assessed total fat as well as being positively correlated with SBP. In addition, BMI and WC shared strikingly similar correlations with directly measured IAAT in a smaller sample of children. The magnitudes of these associations were found to vary to some degree by age, sex, and overweight status.
We found that correlations between central and total adiposity measures obtained by anthropometry and DXA were moderate to strong across ages, in agreement with previous studies [17, 19, 20]. Very strong correlations (r > 0.90) between BMI or WC and TBFM or TFM have been reported among 8-11 year olds , overweight 8 year olds , and 8-17 year olds . As well, in our study, BMI and WC explained a similar amount of variation in body fat, more so for boys than girls, suggesting that WC can be interpreted as a proxy for total adiposity. BMI may account for a somewhat higher variation in body fat in girls than boys due to the preferential deposition of adipose tissue in the lower extremities and the overall higher fat mass associated with normal human growth in females.
The strength of association between WC and BMI and DXA measures generally attenuated after age 11. The weaker relationships at ages 13 and 15 as compared with 9 and 11 imply that less variation in WC is explained by total adiposity as children age. Thus, other factors may account for this unexplained variance, such as IAAT. Our results are consistent with findings from Huang et al. , who demonstrated in a CT study of 8.1 ± 1.6 year olds followed for up to 5 years that IAAT increased with age even after adjustment for subcutaneous abdominal fat in mixed linear regression models. It is possible that the weaker correlations with WC are because it is measured with greater error as children progress through adolescence. However, no work to our knowledge, including our study, supports this conclusion and research is needed to determine the effects of age and growth on WC measurement error during adolescence.
Regardless of differences by age, BMI, WC, and DXA measures remained highly correlated with one another from 9 to 15 years. Moreover, this makes it difficult to disentangle the individual contributions of WC and BMI to metabolic risk as they contain a large degree of shared information . Indeed, in our study, SBP correlations with WC, BMI, and TBFM were comparable in size, consistent with previous reports in children and adolescents [14, 15]. This corroborates findings from a recent prospective study from ALSPAC, which showed that change in BMI, WC, and TBFM from 9-12 years to 15-16 years were consistent at predicting 15-16 year risk for a number of CVD outcomes including SBP . In our study, we additionally found that IAAT was uniformly correlated with WC, BMI, TBFM, and TFM. This is in line with a recent study of 6-7 year olds, where IAAT was similarly correlated with abdominal and suprailiac skinfold thicknesses, WC, hip circumference, and BMI (r = 0.64-0.72) . Further, among older children and adolescents (8-17 years), IAAT correlations with BMI and WC were 0.84 and 0.88, respectively .
Together with our results, these studies support conclusions from a recent systematic review, in which the investigators determined that neither BMI nor WC was superior to the other adiposity measure in identifying children and adolescents with adverse metabolic risk profiles . In other words, BMI and WC effectively measure the same thing in children: thus, WC does not provide a measure of “central adiposity” per se, as in adults. This is an important consideration given that increased WC in adults captures meaningful biological differences that adversely affects metabolic risk beyond general adiposity alone [39, 40].
In addition, we found that the association between TBFM and TFM was strong and stable across ages and each of these adiposity measures exhibited comparable correlations with SBP and IAAT with little variation across sex and weight strata. Based on the current available evidence including our results shown here, TFM may not be an appropriate indicator of central adiposity in children. Additionally, BMI and WC associations with TBFM and TFM were stronger in OWT boys as compared with NWT boys, consistent with results shown elsewhere [17, 36].
Our study has several strengths. Successive cross-sectional analyses conducted during the transition from late childhood to early adolescence allowed for the elucidation of age-related effects on the association between central and total fat measures, which has not been done previously using a large sample size and various direct and indirect measures of adiposity. Our sample size provided adequate statistical power to conduct prespecified subgroup analyses between sex and weight strata. Directly assessed adiposity using gold standard assessments of body composition (i.e., DXA and MRI) yielded valid measures of fat by which to compare anthropometric techniques.
Alternatively, this study was conducted in a relatively lean and predominantly White population and, due to biological differences in body composition, these findings may not be generalizable to other groups including children with extreme obesity . SBP was the only CVD risk factor examined in the current study. Therefore, our findings are not applicable to other CVD risk factors. However, investigators elsewhere have demonstrated that measures of central and total adiposity have similarly sized associations with multiple other risk factors . MRI was conducted on a small subset of children and only at 11 and 13 years. Thus, we could not adequately capture age-related changes in the association between IAAT and DXA and anthropometric measures. Further, children were 6 and 3 months older at the MRI visits compared with the study visits at 11 and 13 years, respectively. Associations between IAAT and anthropometry or DXA may weaken as the gap between the MRI and clinic visits widens limiting our ability to compare IAAT correlations between 11 and 13 years. However, this does not affect the validity of comparing IAAT correlations within age groups.
In conclusion, we demonstrated that WC and BMI remained strongly correlated from 9 to 15 years of age, with some attenuation occurring after age 11. This likely explains the similarly sized associations that these anthropometric measures had with directly assessed total fat and IAAT and SBP. These findings should be considered in future studies of central obesity in children, as they suggest that WC as well as TFM, can be interpreted as measures of total adiposity, and alternatively are poor indicators of central adiposity. This highlights the need to consider the metabolic effects of both central and total adiposity in children as they may be very similar depending upon age, sex, and level of adiposity.
We are appreciative of the families who participated in ALSPAC, the midwives for their help in recruiting participants, and the ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses. Support for ALSPAC comes from the UK Medical Research Council (Grant ref: 74882), the Welcome Trust (Grant ref: 076467), and the University of Bristol. This study was supported by a grant from the American Diabetes Association.