A cross-sectional subject sample was evaluated over a ten year period as part of research programs investigating body composition and related metabolic disorders at the New York Obesity Research Center. Whole-body estimates of VAT, SAT, and TAT were used to generate sex-specific age plots that were then evaluated with or without adjustment for potential influencing factors.

Each subject was screened with a medical history, physical examination, and blood studies. Subjects were retained in the current database who were healthy, free of acute and serious chronic diseases, and who were not engaged in high level physical activity training programs. Subjects weighing more than 300 pounds (136 kg) were excluded from the study due to the MRI system weight limit. There were no race restrictions and race was established by self-report. A lower age limit of 5 years was set as children below this threshold are usually unable to comfortably tolerate the imaging procedure. Once medically cleared, each subject completed the whole-body MRI scan for measurement of adipose tissue volumes.

Body weight was measured to the nearest 0.1 kg and height to the nearest 0.1 cm using appropriately calibrated scales and stadiometers.

Whole-body MRI scans were acquired, as previously reported by our group 1617, using a 1.5 T General Electric system (6X Horizon, Milwaukee, WI). Seventy-one children (ages 5 to 17 years) and all 352 adults (age ≥ 18 years) were scanned by protocol 1, which acquires cross-sectional images of 10 mm thickness at 40 mm intervals from fingers to toes with the subject in either a prone or supine position, resulting ~40 images in each subject. The L₄-L₅ inter-vertebral disc was set as the point of origin. A subset of 76 children was scanned by protocol 2, which acquires cross-sectional images of 10 mm thickness without any gaps, resulting ~200 images in each subject. In each subject scanned with protocol 2, a subset of images was selected by including every slice and by skipping every four slices. The selection of this subset of images from protocol 2 includes ~40 images, with an image slice interval and thickness that exactly matches that of protocol 1. As this subset of images from protocol 2 was used for data generation, all subjects can be considered as using the same scanning protocol (i.e., protocol 1). Following acquisition, VAT and SAT were segmented at the New York Image Reading Center by trained and quality-controlled technicians using image analysis software (SliceOmatic, Tomovision Inc., Montreal, Canada). Visceral adipose tissue volumes were calculated using all slices between the dome of the liver and the bottom of the pelvis (abdominopelvic region). Adipose tissue compartment volume (V) was calculated as:

where A_i is each scan's cross-sectional area, h is the between-slice interval, t is the thickness of each slice, and N is the number of total slices. The intra-class correlation coefficient based on 2-way mixed effects model for volume rendering of SAT and VAT for the same scan by different analysts are 0.99, and 0.95, respectively 18.

Group data are presented as the mean ± SD for children (5–17 years) and adults (age ≥ 18 years). Visceral adipose tissue and SAT volumes were plotted against age for males and females separately.

Regression models were used to investigate the relationships between age and adipose tissue compartments. Nonlinear effects of age on VAT and SAT were tested by including quadratic and cubic terms and their interactions with sex. Covariates making significant contributions were retained in the model. Regression analyses were also conducted for VAT and SAT volumes after adjustment for TAT, ethnicity, and menopause status for females.

Cubic or quadratic lines, depending on the final regression model for VAT and SAT with and without adjustment, were then fitted separately for males and females for the age range of 2.5 to 97.5 percentile (i.e., age 7 to 78 years).

Regression analyses were also conducted separately for pre- and post- menopausal women, and for prepubertal and pubertal children. The intra-class correlation coefficient was calculated with subjects as a random effect and MRI analyst as a fixed effect based on a 2-way mixed effects model (i.e., model 3, 1) 18.

All statistical analyses were carried out using SPSS (SPSS for Windows, 16.0, SPSS Inc., Chicago, USA) and we used two-tailed (α = 0.05) tests of significance.

The subject demographic and body composition data are presented in Table 1. The multiethnic group included 147 children (ages 5 – 17 years; n = 88 males and 59 females) and 352 adults (age ≥ 18 years; n = 164 males and 188 females). The adult and child groups had mean BMIs or BMI percentiles approximating those observed in the general US population 1920. Among the adults, males were younger than females (P < 0.001) and there were no sex differences in BMI. Age and BMI percentiles in the child group did not differ significantly between males and females.

Total adipose tissue was significantly smaller in the adult males than in the females (18.8 ± 8.1 L vs. 25.2 ± 11.9 L; p < 0.001). TAT was also significantly smaller in the boys than in the girls (13.4 ± 9.5 L vs. 17.7 ± 12.5 L; p = 0.020).

VAT was significantly greater in the adult males than in the females (2.1 ± 1.8 L vs. 1.4 ± 1.2 L; p < 0.001). There were no sex differences in VAT in the children (0.6 ± 0.5 vs. 0.6 ± 0.5 L; p = 0.762).

SAT volume was significantly smaller in the adult males than in the females (16.5 ± 6.9 L vs. 23.7 ± 11.1 L; p < 0.001) and significantly smaller in the boys than in the girls (12.2 ± 8.5 vs. 16.4 ± 11.6 L; p = 0.013).

The regression equations for each adipose tissue component are presented in Table 2. The smoother line for VAT and SAT volumes versus age are shown in Figure 1 and Figure 2, respectively. The figure shows that VAT and SAT are both larger with greater age across the entire age range except for a reverse trend before age 17 years in females (i.e., VAT at age 7, 0.70 L, VAT age 17, 0.48 L). Males had a larger amount of VAT than females after age 12 years and this difference remained throughout the remaining age range (Figure 1). In contrast, females had a larger SAT volume than males across the whole age range (Figure 2). In both males and females, the slope of the regression line between SAT and age was smaller for subjects over the age of 50 years older than for their younger counterparts (slopes in males, 0.003 vs 0.204; slopes in females, 0.021 vs 0.322).

After adjustment for TAT volume and ethnicity, in males VAT was smaller with greater age before the age of 12 years and VAT was larger after the age of 12 years throughout the remaining age range (Figure 3). After adjustment for TAT, ethnicity, and menopause status in females, VAT was smaller with greater age before the age of 26 years and larger after the age of 26 years throughout the remaining age range (Figure 3). After adjustment for TAT volume and ethnicity, SAT was larger for older age before the age of 17 years in males and smaller after the age of 17 years throughout the remaining age range (Figure 4). In females SAT was larger with increasing age before the age of 35 years and SAT was smaller with increasing age after the age of 35 years (Figure 4).

When regression analyses for both VAT and SAT were completed in pre- and post- menopausal women separately, age² (P = 0.003, P < 0.001 for VAT and SAT respectively) and age³ (P = 0.001, P < 0.001 for VAT and SAT respectively) significantly entered the model in pre-menopausal women. However, there were non-significant relationships between age and SAT (P = 0.130) and age and VAT (P = 0.872) in postmenopausal women. When pre- and post- menopausal women were pooled, menopausal status was no longer significant for both VAT and SAT, either as an individual term (VAT, P = 0.279; SAT, P = 0.272) or as interactions with age (VAT, P = 0.284; SAT, P = 0.440).

The increase of VAT was calculated for every age decade. In women, the VAT increases from ages 40 to 50 years (i.e., mean volume, 0.55 L) and from age 50 to 60 years (i.e., mean volume, 0.45 L) were the largest observed.

There were a total of 115 children (i.e., ~78%) with pubertal status available. One 16 year old child with Tanner Stage 1 was excluded from the data analysis as an outlier. When regression analyses for both VAT and SAT were completed in pre-pubertal and pubertal children separately, age did not significantly (P = 0.495, P = 0.337 for VAT and SAT, respectively) enter the model in pre-pubertal children when controlled for sex (Table 3). On the other hand, age borderline significantly or significantly (P = 0.068, P = 0.004 for VAT and SAT, respectively) entered the model in pubertal children when controlled for sex (Table 3). Sex did not significantly enter the models in prepubertal children for VAT (P = 0.235), and SAT (P = 0.259). In pubertal children, sex was significant in the SAT model (P = 0.004) but not in the VAT model (P = 0.722). When pre-pubertal and pubertal children were pooled, age significantly entered the model for VAT (P = 0.021) and SAT (P = 0.001). In the pooled data, pubertal status was not significant for both VAT and SAT, either as an individual term (VAT, P = 0.907; SAT, P = 0.718) or as interactions with age (VAT, P = 0.802; SAT, P = 0.415).

There are several limitations of this study. First, we cannot exclude the possibility of the existence of a subject selection bias. Although ideally our relatively large and ethnically diverse sample along with controlling for ethnicity minimize selection bias, individuals who continued to increase abdominal adiposity throughout their life span may have developed diseases and would have been excluded from participation in this study because we only included healthy subjects. As the prevalence of obesity has increased over the past two decades the possibility also exists that younger subjects are more obese than older subjects and therefore the cross-sectional results from the study may be different from what would have been observed had we conducted a longitudinal study. Similarly, the increase in the prevalence of obesity may also induce changes in adipose tissue distribution if adipose tissue distribution is a function of total adiposity. This possibility implies that younger subjects' adipose tissue distribution may be different from older subjects' adipose tissue distribution secondary to greater adiposity in the younger subjects. However, a recent study showed that the relationship between BMI and waist circumference as characterized by the slope of the linear regression of waist circumference on BMI does not appear to be changing appreciably over time 43. This implies that fat distribution may not change with total adiposity at a population level.

Second, we did not control for cardiorespiratory fitness, which is also related to visceral adiposity 3544. Our study was confined to non-athletes and measures of cardiorespiratory fitness were unavailable to us.

Third, we did not control for hormone replacement therapy that may affect abdominal adiposity.

Fourth, compared to most epidemiological studies, the scale of our study is relatively small (i.e., N = 499). Our results should therefore not be used as a population reference data.

Finally, there are race differences in adipose tissue distribution 45. Although we controlled for race differences in this study, we did not have enough subjects in each ethnic group to examine differences in adipose tissue partitioning differences across the whole life span.

Despite the above mentioned limitations, our study is unique in depicting the preliminary trends in the relationships between adipose tissue partitioning and age based on a diverse cross-sectional sample. The results generated from the present study can be used to guide future longitudinal studies adipose tissue growth and its change with aging.