Abstract
Background
The obesity epidemic brought a need for accessible methods to monitor body composition, as excess adiposity has been associated with cardiovascular disease, metabolic disorders, and some cancers. Recent 3-dimensional optical (3DO) imaging advancements have provided opportunities for assessing body composition. However, the accuracy and precision of an overall 3DO body composition model in specific subgroups are unknown.
Objectives
This study aimed to evaluate 3DO’s accuracy and precision by subgroups of age, body mass index, and ethnicity.
Methods
A cross-sectional analysis was performed using data from the Shape Up! Adults study. Each participant received duplicate 3DO and dual-energy X-ray absorptiometry (DXA) scans. 3DO meshes were digitally registered and reposed using Meshcapade. Principal component analysis was performed on 3DO meshes. The resulting principal components estimated DXA whole-body and regional body composition using stepwise forward linear regression with 5-fold cross-validation. Duplicate 3DO and DXA scans were used for test–retest precision. Student’s t tests were performed between 3DO and DXA by subgroup to determine significant differences.
Results
Six hundred thirty-four participants (females = 346) had completed the study at the time of the analysis. 3DO total fat mass in the entire sample achieved R2 of 0.94 with root mean squared error (RMSE) of 2.91 kg compared to DXA in females and similarly in males. 3DO total fat mass achieved a % coefficient of variation (RMSE) of 1.76% (0.44 kg), whereas DXA was 0.98% (0.24 kg) in females and similarly in males. There were no mean differences for total fat, fat-free, percent fat, or visceral adipose tissue by age group (P > 0.068). However, there were mean differences for underweight, Asian, and Black females as well as Native Hawaiian or other Pacific Islanders (P < 0.038).
Conclusions
A single 3DO body composition model produced accurate and precise body composition estimates that can be used on diverse populations. However, adjustments to specific subgroups may be warranted to improve the accuracy in those that had significant differences.
This trial was registered at clinicaltrials.gov as NCT03637855 (Shape Up! Adults).
Keywords: body composition, three-dimensional optical, diversity, DXA, accuracy
Introduction
Obesity has been a growing epidemic for the past few generations and has led to the need for more accessible methods to monitor adiposity [[1], [2], [3]]. Excess adiposity is linked to the development of cardiovascular disease, type 2 diabetes, and up to 20% of cancers [[4], [5], [6]]. BMI is generally used to classify obesity. However, BMI only uses height and weight and does not account for muscle or FM, and therefore, it is a poor method for nutritional assessment on an individual level. Instead, body composition methods have been developed to quantify FM and FFM (lean soft tissue + bone masses) to characterize health status [7,8]. Furthermore, regional body composition such as visceral adipose tissue (VAT) has shown to be a better predictor than BMI for adverse outcomes such as type 2 diabetes (OR: 2.17 vs. 1.66) and hypertension (OR: 2.08 vs. 1.77) [9]. Studies have also shown ethnic differences when it comes to fat distribution such as higher levels of VAT in Japanese-Americans compared with their ethnic counterparts [10] as well as age-related differences [11].
Imaging methods such as MRI and CT are among criterion methods for regional body composition [12,13]. However, these methods are expensive; require trained/certified technicians; use ionizing radiation (CT), limiting frequent use; and have limited access outside clinical settings. DXA, air displacement plethysmography (ADP), and BIA are among the more accessible modalities developed for body composition, but they also have limitations including requiring trained technicians (DXA and ADP), ionizing radiation (DXA), physiological calibration assumptions (BIA), and lack of regional/compartmental compositions (ADP) [[14], [15], [16]]. The ideal method would include total and regional composition to be accurate in subgroups by sex, BMI, age, and ethnicity; free of ionizing radiation; self-operating; and low cost to operate.
In the last 2 decades, body shape defined in detail using 3-dimensional optical (3DO) imaging has been intensely explored as a health descriptor [17]. 3DO scanners generally output a detailed 3D mesh with over 100,000 sample points that represents the person’s body shape and automated anthropometric estimates (ie, circumferences, lengths, surface areas, and volumes) [[18], [19], [20]]. From the early days until now, researchers validated the accuracy and precision of the automated anthropometry to criterion methods (eg, tape measurements to circumferences/lengths, underwater weighing or ADP to total volume) [[20], [21], [22]]. Researchers also showed that these automated anthropometric estimates, some of which would be difficult and tedious to obtain manually, were predictive of DXA FM and FFM [[23], [24], [25], [26]].
Since 3DO anthropometry was as good if not better than manual measurements, as health descriptors, researchers attempted to create more advanced shape descriptors using the 3DO mesh to utilize the entire body. Since then, the methodology to obtain these body shape descriptors have improved by incorporating automated processing methods, pose-independent body composition models, a 2D image to 3D mesh pipeline, and agnostic body composition models across multiple 3DO scanners so that the models were no longer device specific [[27], [28], [29], [30]].
Although much has been accomplished in this field in a short amount of time, body composition models created for 3DO scanners have not been interrogated for accuracy as a function of age, BMI, and ethnicity. Our hypothesis was that body shape is deterministically defined by the underlying distributions of fat and muscle, and when properly modeled, should accurately represent regional fat even when body shapes differ by sex, age, BMI, and ethnicity. The objective of this study was to evaluate the 3DO’s total and regional body composition accuracy and precision in a diverse population stratified by sex, age, BMI, and ethnicity as compared to DXA.
Methods
Study design
Shape Up! Adults was a cross-sectional study of healthy adults (NIH R01 DK109008, clinicaltrials.gov NCT03637855). This study was designed to investigate the associations between body shape and composition with various health markers. Participants underwent a series of measures that included whole-body 3DO scans, DXA scans, blood serum tests, and functional strength tests.
Participants
Participants were recruited at Pennington Biomedical Research Center (PBRC), University of Hawaii Cancer Center (UHCC), and University of California, San Francisco (UCSF). All participants provided informed consent. The study protocol was approved by the Institutional Review Boards at PBRC (IRB study #2016-053), UCSF (IRB #15-18066), and the University of Hawaii Office of Research Compliance (CHS #2017-01018). Volunteers were prescreened over the phone and were deemed ineligible if they were pregnant, breastfeeding, had missing limbs, nonremovable metal, previous body-altering surgery (eg, breast augmentation, liposuction), hair that could not be contained in a swim cap, were unable to stand still for 1 min, or unable to lay still for 3 min.
Pretesting preparations included an 8-h fast (water and prescribed medications were allowed) and no strenuous exercise 24 h prior to the study visit. Participants were stratified by age (18–39, 40–59, ≥60 y), ethnicity (non-Hispanic White [White], non-Hispanic Black [Black], Hispanic, Asian, and Native Hawaiian or Other Pacific Islander [NHOPI]), sex, and BMI (<18.5 [underweight], 18.5–24.9 [normal weight], 25–29.9 [overweight], >30 [obese] kg/m2) according to the World Health Organization [31]. Participants self-reported their ethnicity from the 5 ethnic subgroups. Height and weight were measured on a SECA 274 Stadiometer.
DXA
Participants received duplicate whole-body DXA scans with a Hologic Discovery/A or Horizon system according to International Society for Clinical Densitometry guidelines with repositioning [32]. Participants were scanned once, asked to get off the table, and laid back on the table for the second scan. The Hologic Whole-Body Phantom was scanned 10 times at each site with UCSF considered the gold standard site. Ratios between the gold standard site and the other sites were calculated. An ANOVA with a Dunnett test was applied to determine mean difference (MDs) between the sites and the gold standard site. There were no differences between UCSF and PBRC. There was a correction factor of 1.05036 between UCSF and UHCC for %fat. The Hologic Block Phantom was scanned daily for quality control to ensure the system did not fall out of calibration. No other corrections were needed. All scans were analyzed at UHCC by a single certified technologist using Hologic Apex version 5.6 with the NHANES Body Composition Analysis calibration option disabled [33]. Outputs included whole-body and regional FM and FFM measures.
3DO surface scans
Participants changed into form-fitting tights, a swim cap, and a sports bra if female. Duplicate 3DO surface scans were taken on the Fit3D ProScanner version 4.x within 10 min of each other and with repositioning. Participants grasped telescoping handles on the scanner platform and stood upright with shoulders relaxed and arms positioned straight and abducted from their torso. The platform rotates once around and takes approximately 45 s for the completion of the scan. Final point clouds were converted to a mesh connected by triangles with approximately 300,000 vertices and 600,000 faces to represent the body shape. Previous validation studies showed good agreement and precision for Fit3D’s automated anthropometry in comparison to manual circumference measurements [19,22,24].
3DO meshes were registered and digitally reposed by Meshcapade. Their algorithm registers each mesh to a 110,000-vertex template with full anatomical correspondence. This means each vertex corresponds to a specific anatomical location across all registered meshes. Without the registration, the number of vertices and their locations would be random. Thus, the variance would not be comparable for analysis. Meshcapade’s methods can be read in detail in Loper et al. [34]. All meshes were digitally reposed to a T-pose, where the person was standing straight, arms were brought horizontal and in plane with the body, and arms and legs were straightened. Previous work showed that reposing the mesh to a standardized position and pose reduced the precision error by approximately 50% [29].
Statistical shape modeling
Since vertices in the 3D meshes are highly correlated with neighboring vertices, principal component analysis was performed on the registered meshes to orthogonalize and reduce the dimensionality of the data so that fewer variables were needed to describe the data’s variance while creating sex-specific statistical shape models [35]. The resulting outputs were principal components (PCs) that could be used for analysis.
Manifold regression images stratified by ethnicity and sex (FIGURE 1, FIGURE 2) were created by adjusting the PCs with the mean FM, FFM, and age, shown in Table 1 [3]. The adjusted PCs were then averaged and inverted back into coordinate space (x, y, and z) to obtain the image. The manifold images were a demonstration on how average body shape may differ by ethnicity when FM, FFM, and age are the same. With this visual example, regional MDs were added to show how different ethnic groups may hold FM and FFM from other ethnic counterparts. P values for the regional differences were also displayed on the images.
FIGURE 1.
Female manifold regression images of different ethnic groups that were adjusted with the overall sample mean age and DXA total fat mass and fat-free mass (ie, 46.6 y, 25.4 kg, and 46.1 kg, respectively). This means each mesh, in theory, has the same age, fat, and fat-free masses. Regional differences (DXA – 3DO) are displayed to examine how different ethnic groups differ in body shape and composition even though they have the same total fat and fat-free masses. Bold print indicates a significant difference (P < 0.05) in total and regional fat masses between DXA and 3DO by Student’s t test. 3DO, 3-dimensional optical; DXA, dual-energy X-ray absorptiometry.
FIGURE 2.
Male manifold regression images of different ethnic groups that were adjusted with the overall sample mean age and DXA total fat mass and fat-free mass (ie, 43.5 y, 20.9 kg, and 66.6 kg, respectively). This means each mesh, in theory, has the same age, fat, and fat-free masses. Regional differences (DXA – 3DO) are displayed to examine how different ethnic groups differ in body shape and composition even though they have the same total fat and fat-free masses. Bold print indicates a significant difference (P < 0.05) in total and regional fat masses between DXA and 3DO by Student’s t test. 3DO, 3-dimensional optical; DXA, dual-energy X-ray absorptiometry.
TABLE 1.
Sample characteristics by age group and BMI category
| Female |
||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Ages 18–39 y |
Ages 40–59 y |
Ages ≥60 y |
Overall |
|||||||||||||
|
Underweight |
Normal |
Overweight |
Obese |
Underweight |
Normal |
Overweight |
Obese |
Underweight |
Normal |
Overweight |
Obese |
Underweight |
Normal |
Overweight |
Obese |
|
| (N = 9) | (N = 44) | (N = 38) | (N = 38) | (N = 10) | (N = 43) | (N = 33) | (N = 36) | (N = 6) | (N = 39) | (N = 26) | (N = 24) | (N = 25) | (N = 126) | (N = 97) | (N = 98) | |
| Age (y) | ||||||||||||||||
| Mean (SD) | 24.3 (5.98) | 28.0 (5.93) | 29.3 (5.75) | 31.1 (5.77) | 52.5 (4.55) | 49.6 (5.57) | 52.1 (5.81) | 48.3 (5.61) | 66.3 (2.73) | 66.4 (5.33) | 66.4 (3.79) | 64.6 (3.34) | 45.7 (17.8) | 47.2 (16.7) | 47.0 (16.2) | 45.6 (14.2) |
| [Min, Max] | [18.0, 35.0] | [18.0, 39.0] | [19.0, 39.0] | [18.0, 39.0] | [42.0, 56.0] | [41.0, 59.0] | [40.0, 59.0] | [40.0, 59.0] | [62.0, 70.0] | [60.0, 89.0] | [60.0, 75.0] | [60.0, 71.0] | [18.0, 70.0] | [18.0, 89.0] | [19.0, 75.0] | [18.0, 71.0] |
| Ethnicity | ||||||||||||||||
| Asian | 3 (33.3%) | 8 (18.2%) | 10 (26.3%) | 6 (15.8%) | 5 (50.0%) | 14 (32.6%) | 6 (18.2%) | 7 (19.4%) | 1 (16.7%) | 14 (35.9%) | 6 (23.1%) | 2 (8.3%) | 9 (36.0%) | 36 (28.6%) | 22 (22.7%) | 15 (15.3%) |
| Black | 0 (0%) | 9 (20.5%) | 6 (15.8%) | 11 (28.9%) | 1 (10.0%) | 7 (16.3%) | 9 (27.3%) | 6 (16.7%) | 1 (16.7%) | 7 (17.9%) | 6 (23.1%) | 9 (37.5%) | 2 (8.0%) | 23 (18.3%) | 21 (21.6%) | 26 (26.5%) |
| Hispanic | 1 (11.1%) | 12 (27.3%) | 4 (10.5%) | 6 (15.8%) | 0 (0%) | 7 (16.3%) | 6 (18.2%) | 5 (13.9%) | 0 (0%) | 3 (7.7%) | 2 (7.7%) | 2 (8.3%) | 1 (4.0%) | 22 (17.5%) | 12 (12.4%) | 13 (13.3%) |
| NHOPI | 0 (0%) | 3 (6.8%) | 6 (15.8%) | 7 (18.4%) | 0 (0%) | 4 (9.3%) | 3 (9.1%) | 4 (11.1%) | 0 (0%) | 1 (2.6%) | 0 (0%) | 1 (4.2%) | 0 (0%) | 8 (6.3%) | 9 (9.3%) | 12 (12.2%) |
| White | 5 (55.6%) | 12 (27.3%) | 12 (31.6%) | 8 (21.1%) | 4 (40.0%) | 11 (25.6%) | 9 (27.3%) | 14 (38.9%) | 4 (66.7%) | 14 (35.9%) | 12 (46.2%) | 10 (41.7%) | 13 (52.0%) | 37 (29.4%) | 33 (34.0%) | 32 (32.7%) |
| Height (cm) | ||||||||||||||||
| Mean (SD) | 163 (7.98) | 164 (6.29) | 163 (7.98) | 162 (6.76) | 163 (7.09) | 164 (6.86) | 163 (5.14) | 163 (6.21) | 164 (5.13) | 160 (7.45) | 160 (8.29) | 161 (5.57) | 163 (6.76) | 163 (7.05) | 162 (7.32) | 162 (6.27) |
| [Min, Max] | [151, 176] | [151, 180] | [149, 181] | [146, 174] | [155, 177] | [150, 181] | [154, 171] | [147, 176] | [157, 173] | [144, 178] | [149, 190] | [151, 171] | [151, 177] | [144, 181] | [149, 190] | [146, 176] |
| Weight (kg) | ||||||||||||||||
| Mean (SD) | 45.3 (5.94) | 58.3 (5.93) | 72.5 (7.31) | 98.4 (20.3) | 45.6 (4.95) | 59.2 (7.71) | 73.2 (5.91) | 98.1 (19.6) | 47.3 (4.20) | 54.8 (5.81) | 70.2 (9.02) | 87.2 (7.91) | 45.9 (5.03) | 57.5 (6.76) | 72.1 (7.41) | 95.5 (18.3) |
| [Min, Max] | [38.6, 55.4] | [44.5, 72.5] | [60.9, 88.8] | [70.5, 153] | [35.4, 53.0] | [44.2, 75.2] | [61.0, 85.0] | [66.3, 146] | [42.2, 52.3] | [41.7, 67.6] | [60.0, 103] | [73.6, 102] | [35.4, 55.4] | [41.7, 75.2] | [60.0, 103] | [66.3, 153] |
| BMI (kg/m2) | ||||||||||||||||
| Mean (SD) | 17.0 (1.11) | 21.7 (1.81) | 27.1 (1.18) | 37.5 (6.73) | 17.1 (1.26) | 21.9 (1.88) | 27.5 (1.22) | 36.8 (6.51) | 17.6 (0.905) | 21.4 (1.92) | 27.4 (1.33) | 33.6 (3.11) | 17.2 (1.11) | 21.7 (1.86) | 27.3 (1.24) | 36.3 (6.11) |
| [Min, Max] | [14.8, 18.4] | [18.6, 24.8] | [25.1, 29.6] | [30.0, 51.9] | [14.2, 18.4] | [18.7, 24.9] | [25.2, 29.9] | [30.1, 53.1] | [16.0, 18.4] | [18.6, 24.4] | [25.1, 29.6] | [30.0, 40.9] | [14.2, 18.4] | [18.6, 24.9] | [25.1, 29.9] | [30.0, 53.1] |
| Total FM (kg) | ||||||||||||||||
| Mean (SD) | 10.7 (2.81) | 16.4 (3.93) | 24.3 (4.12) | 39.7 (13.1) | 9.49 (2.08) | 17.5 (4.21) | 27.5 (3.64) | 40.3 (9.91) | 10.9 (1.74) | 17.0 (3.89) | 26.6 (4.16) | 36.7 (6.64) | 10.3 (2.31) | 17.0 (4.01) | 26.0 (4.18) | 39.2 (10.6) |
| [Min, Max] | [7.20, 14.6] | [8.31, 27.4] | [11.9, 32.7] | [22.7, 72.7] | [6.30, 12.2] | [9.31, 27.2] | [18.6, 35.0] | [27.9, 67.8] | [9.08, 13.1] | [10.4, 25.5] | [16.4, 34.7] | [22.4, 48.9] | [6.30, 14.6] | [8.31, 27.4] | [11.9, 35.0] | [22.4, 72.7] |
| Total FFM (kg) | ||||||||||||||||
| Mean (SD) | 34.4 (3.64) | 42.0 (4.55) | 47.9 (5.37) | 58.6 (9.14) | 36.1 (4.71) | 41.6 (5.45) | 45.7 (4.32) | 56.6 (9.88) | 36.1 (3.42) | 37.9 (4.51) | 43.2 (8.26) | 50.4 (4.00) | 35.5 (3.98) | 40.6 (5.17) | 45.9 (6.15) | 55.9 (9.02) |
| [Min, Max] | [30.0, 41.6] | [28.9, 52.5] | [39.8, 61.3] | [42.1, 80.4] | [28.6, 43.6] | [30.9, 53.8] | [37.1, 53.2] | [37.1, 77.5] | [32.2, 40.9] | [29.0, 47.5] | [33.7, 75.6] | [42.4, 58.0] | [28.6, 43.6] | [28.9, 53.8] | [33.7, 75.6] | [37.1, 80.4] |
| Percent body fat (%) | ||||||||||||||||
| Mean (SD) | 23.5 (3.91) | 27.9 (5.19) | 33.6 (4.43) | 39.6 (5.89) | 20.9 (4.23) | 29.5 (5.03) | 37.5 (3.66) | 41.4 (3.37) | 23.3 (2.93) | 30.9 (5.61) | 38.3 (5.48) | 41.9 (4.76) | 22.4 (3.90) | 29.4 (5.36) | 36.2 (4.90) | 40.8 (4.88) |
| [Min, Max] | [18.8, 29.6] | [15.2, 37.6] | [17.2, 39.2] | [28.7, 53.3] | [12.6, 27.1] | [18.7, 39.3] | [30.3, 45.2] | [34.9, 48.3] | [18.2, 26.6] | [20.4, 44.8] | [26.6, 45.6] | [30.5, 48.6] | [12.6, 29.6] | [15.2, 44.8] | [17.2, 45.6] | [28.7, 53.3] |
| VAT (kg) | ||||||||||||||||
| Mean (SD) | 0.11 (0.03) | 0.17 (0.07) | 0.30 (0.10) | 0.57 (0.31) | 0.11 (0.03) | 0.29 (0.14) | 0.56 (0.20) | 0.82 (0.28) | 0.13 (0.07) | 0.36 (0.18) | 0.65 (0.19) | 0.80 (0.23) | 0.11 (0.04) | 0.27 (0.16) | 0.48 (0.22) | 0.72 (0.30) |
| [Min, Max] |
[0.05, 0.15] |
[0.07, 0.34] |
[0.11, 0.51] |
[0.09, 1.37] |
[0.06, 0.17] |
[0.08, 0.61] |
[0.24, 1.14] |
[0.31, 1.32] |
[0.07, 0.23] |
[0.08, 0.90] |
[0.30, 1.14] |
[0.49, 1.22] |
[0.05, 0.23] |
[0.07, 0.90] |
[0.11, 1.14] |
[0.09, 1.37] |
| Males | ||||||||||||||||
|
(N = 2) |
(N = 37) |
(N = 59) |
(N = 34) |
(N = 1) |
(N = 82) |
(N = 32) |
(N = 38) |
(N = 1) |
(N = 22) |
(N = 22) |
(N = 17) |
(N = 4) |
(N = 82) |
(N = 113) |
(N = 89) |
|
| Age (y) | ||||||||||||||||
| Mean (SD) | 29.0 (5.66) | 26.6 (5.47) | 29.6 (6.11) | 30.4 (5.84) | 48.0 (NA) | 43.8 (18.1) | 48.9 (5.78) | 49.6 (6.20) | 67.0 (NA) | 67.6 (5.71) | 65.9 (4.78) | 63.6 (3.20) | 43.3 (18.5) | 43.8 (18.1) | 42.1 (15.5) | 45.0 (13.7) |
| [Min, Max] | 29.0 [25.0, 33.0] | 25.0 [18.0, 37.0] | 30.0 [19.0, 39.0] | 32.0 [18.0, 39.0] | 48.0 [48.0, 48.0] | 42.0 [18.0, 79.0] | 48.5 [40.0, 59.0] | 50.0 [40.0, 59.0] | 67.0 [67.0, 67.0] | 66.0 [60.0, 79.0] | 64.5 [60.0, 77.0] | 62.0 [60.0, 71.0] | 40.5 [25.0, 67.0] | 42.0 [18.0, 79.0] | 38.0 [19.0, 77.0] | 44.0 [18.0, 71.0] |
| Ethnicity | ||||||||||||||||
| Asian | 2 (100%) | 9 (24.3%) | 18 (30.5%) | 7 (20.6%) | 0 (0%) | 22 (26.8%) | 9 (28.1%) | 8 (21.1%) | 0 (0%) | 7 (31.8%) | 5 (22.7%) | 2 (11.8%) | 2 (50.0%) | 22 (26.8%) | 32 (28.3%) | 17 (19.1%) |
| Black | 0 (0%) | 8 (21.6%) | 8 (13.6%) | 9 (26.5%) | 1 (100%) | 16 (19.5%) | 7 (21.9%) | 8 (21.1%) | 1 (100%) | 3 (13.6%) | 8 (36.4%) | 6 (35.3%) | 2 (50.0%) | 16 (19.5%) | 23 (20.4%) | 23 (25.8%) |
| Hispanic | 0 (0%) | 8 (21.6%) | 9 (15.3%) | 5 (14.7%) | 0 (0%) | 12 (14.6%) | 2 (6.3%) | 3 (7.9%) | 0 (0%) | 1 (4.5%) | 1 (4.5%) | 1 (5.9%) | 0 (0%) | 12 (14.6%) | 12 (10.6%) | 9 (10.1%) |
| NHOPI | 0 (0%) | 2 (5.4%) | 6 (10.2%) | 4 (11.8%) | 0 (0%) | 2 (2.4%) | 1 (3.1%) | 4 (10.5%) | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 0 (0%) | 2 (2.4%) | 7 (6.2%) | 8 (9.0%) |
| White | 0 (0%) | 10 (27.0%) | 18 (30.5%) | 9 (26.5%) | 0 (0%) | 30 (36.6%) | 13 (40.6%) | 15 (39.5%) | 0 (0%) | 11 (50.0%) | 8 (36.4%) | 8 (47.1%) | 0 (0%) | 30 (36.6%) | 39 (34.5%) | 32 (36.0%) |
| Height (cm) | ||||||||||||||||
| Mean (SD) | 164 (12.7) | 176 (7.61) | 176 (7.02) | 176 (9.37) | 175 (NA) | 176 (7.45) | 174 (6.07) | 177 (7.18) | 151 (NA) | 175 (7.21) | 175 (7.35) | 176 (7.66) | 163 (12.1) | 176 (7.45) | 175 (6.79) | 177 (8.10) |
| [Min, Max] | [155, 173] | [162, 190] | [159, 192] | [147, 202] | [175, 175] | [159, 192] | [163, 188] | [161, 190] | [151, 151] | [159, 188] | [155, 188] | [158, 187] | [151, 175] | [159, 192] | [155, 192] | [147, 202] |
| Weight (kg) | ||||||||||||||||
| Mean (SD) | 47.1 (9.23) | 67.9 (8.28) | 85.1 (8.95) | 110 (17.2) | 56.1 (NA) | 69.2 (8.22) | 82.5 (7.21) | 110 (17.4) | 41.5 (NA) | 68.0 (7.57) | 83.2 (8.46) | 113 (15.4) | 48.0 (8.04) | 69.2 (8.22) | 84.0 (8.41) | 110 (16.8) |
| [Min, Max] | [40.6, 53.7] | [51.4, 81.2] | [63.9, 104] | [70.6, 149] | [56.1, 56.1] | [51.4, 90.2] | [69.1, 97.6] | [88.4, 174] | [41.5, 41.5] | [54.1, 82.4] | [62.1, 100] | [79.4, 149] | [40.6, 56.1] | [51.4, 90.2] | [62.1, 104] | [70.6, 174] |
| Weight (kg/m2) | ||||||||||||||||
| Mean (SD) | 17.5 (0.735) | 21.9 (1.79) | 27.5 (1.50) | 35.4 (4.76) | 18.4 (NA) | 22.3 (1.78) | 27.1 (1.43) | 34.8 (4.38) | 18.2 (NA) | 22.2 (1.82) | 27.2 (1.30) | 36.3 (4.37) | 17.9 (0.627) | 22.3 (1.78) | 27.3 (1.44) | 35.3 (4.51) |
| [Min, Max] | [17.0, 18.0] | [18.7, 25.0] | [25.0, 30.0] | [30.1, 49.2] | [18.4, 18.4] | [18.6, 25.0] | [25.0, 29.5] | [30.1, 52.6] | [18.2, 18.2] | [18.6, 24.6] | [25.0, 29.9] | [30.3, 47.3] | [17.0, 18.4] | [18.6, 25.0] | [25.0, 30.0] | [30.1, 52.6] |
| Total FM (kg) | ||||||||||||||||
| Mean (SD) | 8.93 (2.98) | 11.0 (3.39) | 18.7 (5.50) | 29.9 (10.6) | 8.30 (NA) | 12.6 (4.21) | 19.1 (4.04) | 31.7 (9.03) | 5.01 (NA) | 14.4 (4.16) | 20.1 (3.56) | 34.3 (8.70) | 7.79 (2.55) | 12.6 (4.21) | 19.1 (4.78) | 31.6 (9.58) |
| [Min, Max] | [6.82, 11.0] | [5.13, 20.0] | [9.05, 29.4] | [11.9, 45.9] | [8.30, 8.30] | [5.13, 26.5] | [7.44, 27.3] | [20.5, 66.5] | [5.01, 5.01] | [7.33, 21.5] | [14.3, 26.7] | [20.5, 48.9] | [5.01, 11.0] | [5.13, 26.5] | [7.44, 29.4] | [11.9, 66.5] |
| Total FFM (kg) | ||||||||||||||||
| Mean (SD) | 38.4 (6.28) | 57.2 (7.35) | 66.7 (7.76) | 81.0 (9.72) | 48.0 (NA) | 56.9 (7.17) | 63.5 (6.59) | 77.9 (10.9) | 35.7 (NA) | 54.3 (6.11) | 62.9 (7.15) | 78.4 (9.41) | 40.1 (6.51) | 56.9 (7.17) | 65.1 (7.46) | 79.1 (10.2) |
| [Min, Max] | [33.9, 42.8] | [43.7, 73.0] | [47.2, 86.0] | [65.1, 108] | [48.0, 48.0] | [40.6, 73.0] | [50.4, 78.2] | [62.6, 106] | [35.7, 35.7] | [40.6, 67.8] | [46.4, 74.2] | [58.6, 99.4] | [33.9, 48.0] | [40.6, 73.0] | [46.4, 86.0] | [58.6, 108] |
| Percent body fat (%) | ||||||||||||||||
| Mean (SD) | 18.6 (2.65) | 16.0 (4.26) | 21.8 (5.57) | 26.4 (7.16) | 14.7 (NA) | 18.0 (5.08) | 23.1 (4.53) | 28.7 (4.63) | 12.3 (NA) | 20.7 (4.58) | 24.3 (3.57) | 30.1 (4.97) | 16.1 (3.47) | 18.0 (5.08) | 22.6 (5.02) | 28.2 (5.84) |
| [Min, Max] | [16.7, 20.5] | [9.91, 25.1] | [10.6, 32.8] | [13.6, 38.6] | [14.7, 14.7] | [9.03, 32.6] | [10.2, 31.5] | [20.3, 38.6] | [12.3, 12.3] | [11.5, 31.6] | [16.6, 31.6] | [20.7, 38.4] | [12.3, 20.5] | [9.03, 32.6] | [10.2, 32.8] | [13.6, 38.6] |
| VAT (kg) | ||||||||||||||||
| Mean (SD) | 0.20 (0.01) | 0.23 (0.047) | 0.33 (0.09) | 0.52 (0.22) | 0.22 (NA) | 0.32 (0.16) | 0.49 (0.16) | 0.79 (0.32) | 0.16 (NA) | 0.46 (0.18) | 0.69 (0.24) | 0.91 (0.29) | 0.20 (0.03) | 0.32 (0.16) | 0.45 (0.20) | 0.72 (0.32) |
| [Min, Max] | [0.20, 0.21] | [0.16, 0.33] | [0.18, 0.63] | [0.22, 1.04] | [0.22, 0.22] | [0.16, 0.91] | [0.17, 0.73] | [0.29, 1.64] | [0.16, 0.16] | [0.20, 0.91] | [0.29, 1.22] | [0.35, 1.31] | [0.16, 0.22] | [0.16, 0.91] | [0.17, 1.22] | [0.22, 1.64] |
Abbreviations: NA, not applicable; NHOPI, Native Hawaiian and other Pacific Islanders; SD, standard deviation; VAT, visceral adipose tissue.
Body composition values measured from dual-energy X-ray absorptiometry.
Mean images (FIGURE 3, FIGURE 4) stratified by sex, BMI, and ethnicity were created by averaging the PCs of the specific strata and inverting back into coordinate space to show a generalized average shape of that group. Groups with <1 participant did not have a mean image.
FIGURE 3.
Female mean body shape images stratified by BMI and ethnicity. NA indicates that there was ≤ 1 participant, so a mean image could not be derived. BMI, body mass index.
FIGURE 4.
Male mean body shape images stratified by BMI and ethnicity. NA indicates that there was ≤ 1 participant, so a mean image could not be derived. BMI, body mass index.
Statistical analysis
Stepwise forward linear regression with 5-fold cross-validation was used to predict DXA body composition with the PCs. The reported results were the average result of the 5 folds. The independent variables were the PCs and demographics (ie, age and BMI) (PC+DEMO model), whereas the dependent variables were DXA whole-body and regional body composition measures. PC-only models were created first and then adjusted with age and BMI as potential covariates for the PC + DEMO models. Only variables with a significance of P < 0.5 remained in each model. Results were reported with R2 and RMSE. This model used the entire Shape Up! Adults cohort and was similar to previous models presented with an incomplete sample [28,29,35].
Test–retest precision, also known as short-term precision, was performed on the duplicate DXA and 3DO meshes. The %CV and RMSE were used to quantify the test–retest precision defined by Glüer et al. [36]. Subgroup precision was calculated using standard deviation between test–retest scans and averaged by group because RMSE between 2 scans could not be solved. Differences between subgroups were compared within the strata of age, BMI, and ethnicity with a 1-factor ANOVA. A P value < 0.05 was considered statistically significant.
The accuracy of 3DO total FM, FFM, and %fat was evaluated at the subgroup level with the stratifications mentioned previously. MDs were calculated between 3DO estimates from the PC + DEMO model and DXA (3DO – DXA). Paired Student’s t test determined if the differences were statistically significant (P < 0.05). If significant, the %MD would be calculated [(DXA – 3DO) / DXA × 100]. Percent MDs under 2% were considered small, 2% to 5% moderate, and >5% large. All analyses were performed in R version 4.2.1 (R Core Teams).
Results
At the time of this analysis, 883 participants were available. Two hundred forty-nine participants were excluded for different study protocol (n = 181), dropouts (n = 7), missing or unusable 3DO scan (n = 48), or invalid DXA scan (n = 13) (Supplemental Figure 1). After exclusions, 634 participants remained in the analysis (Table 1). In total, 225, 302, and 108 participants were recruited from UCSF, PBRC, and UHCC, respectively. The female and male ethnic distributions were similar to NHANES for White and Black participants. However, the Shape Up! Adults recruitment had about 10% more Asian and 10% less Hispanic participants compared with NHANES [37].
The first 3 PC modes captured 95% of the shape variance in each of the female and male shape models. 3DO PC-only equations were highly correlated to DXA body composition values (Table 2). 3DO total FM and FFM in females achieved R2s of 0.94 and 0.92 with RMSEs of 2.91 kg and 2.76 kg, whereas males achieved R2s of 0.94 and 0.94 with RMSEs of 3.04 kg and 2.97 kg, respectively. Female and male 3DO %fat had moderate correlations with DXA (R2: 0.75 and 0.73; RMSE: 3.82% and 3.31%, respectively). Regional 3DO FM and FFM estimates (ie, arms, legs, and trunk) had moderate to strong correlations with DXA (R2 range: 0.79–0.95, RMSE range: 0.27–1.86 kg) for females and males. After possible adjustments for age and BMI as covariates (PC + DEMO), there were marginal improvements to the R2s and RMSEs.
TABLE 2.
3D optical body composition models from stepwise forward linear regression to estimate to DXA body composition (mean results from 5-fold cross-validation)
| Sex | Outcome | PC-only |
PC + Demographics |
||||||
|---|---|---|---|---|---|---|---|---|---|
| Variable | Coefficient | R2 | RMSE | Variable | Coefficient | R2 | RMSE | ||
| Females | Total FM | Intercept | 26.5689 | 0.94 | 2.91 | Intercept | 16.4494 | 0.94 | 2.88 |
| PC1 | 0.1282 | PC1 | 0.1108 | ||||||
| PC2 | 2.0930 | PC2 | 1.5903 | ||||||
| PC3 | -1.0575 | PC3 | -0.7970 | ||||||
| PC4 | 0.7487 | PC4 | 0.5483 | ||||||
| PC5 | 0.5329 | PC5 | 0.4035 | ||||||
| PC6 | -0.4188 | PC6 | -0.2978 | ||||||
| PC7 | -1.3200 | PC7 | -0.9192 | ||||||
| PC9 | -0.4824 | PC9 | -0.7256 | ||||||
| PC10 | 1.6580 | PC10 | 1.1085 | ||||||
| PC11 | 1.8068 | PC11 | 1.4661 | ||||||
| PC13 | -2.7090 | PC13 | -2.3351 | ||||||
| BMI | 0.3597 | ||||||||
| Total FFM | Total Mass - Total Fat Mass | 0.92 | 2.76 | Total Mass - Total Fat Mass | 0.92 | 2.63 | |||
| Percent Fat | Total Fat Mass / Total Mass | 0.75 | 3.82 | Total Fat Mass / Total Mass | 0.75 | 3.76 | |||
| VAT | Intercept | 0.4516 | 0.72 | 0.15 | Intercept | 0.3334 | 0.78 | 0.14 | |
| PC2 | 0.0408 | PC2 | 0.0431 | ||||||
| PC3 | -0.0226 | PC3 | -0.0254 | ||||||
| PC4 | 0.0151 | PC4 | 0.0142 | ||||||
| PC5 | -0.0332 | PC5 | -0.0252 | ||||||
| PC6 | 0.0554 | PC6 | 0.0417 | ||||||
| PC7 | 0.0349 | PC7 | 0.0382 | ||||||
| PC8 | -0.0680 | PC8 | -0.0931 | ||||||
| PC12 | -0.0809 | PC12 | -0.0301 | ||||||
| PC13 | -0.0348 | PC13 | -0.0569 | ||||||
| PC14 | -0.0489 | PC14 | -0.0555 | ||||||
| PC15 | -0.0570 | PC15 | -0.0490 | ||||||
| Age | 0.0028 | ||||||||
| Arm FM | Intercept | 1.6889 | 0.86 | 0.34 | Intercept | 0.6964 | 0.87 | 0.33 | |
| PC1 | 0.0073 | PC1 | 0.0058 | ||||||
| PC2 | 0.1508 | PC2 | 0.1023 | ||||||
| PC3 | -0.0723 | PC3 | -0.0508 | ||||||
| PC4 | 0.0354 | PC4 | 0.0242 | ||||||
| PC7 | -0.0391 | PC7 | -0.0176 | ||||||
| PC8 | -0.0928 | PC8 | -0.0755 | ||||||
| PC10 | 0.0857 | PC10 | 0.0553 | ||||||
| PC11 | 0.1248 | PC11 | 0.0901 | ||||||
| PC13 | -0.0635 | PC13 | -0.0259 | ||||||
| PC14 | 0.1282 | PC14 | 0.0937 | ||||||
| BMI | 0.0356 | ||||||||
| Arm FFM | Intercept | 2.4555 | 0.81 | 0.27 | Intercept | 1.7392 | 0.81 | 0.27 | |
| PC1 | 0.0159 | PC1 | 0.0141 | ||||||
| PC2 | 0.0791 | PC2 | 0.0423 | ||||||
| PC4 | -0.0283 | PC4 | -0.0308 | ||||||
| PC5 | 0.0653 | PC5 | 0.0427 | ||||||
| PC6 | -0.0587 | PC6 | -0.0469 | ||||||
| PC7 | 0.0945 | PC7 | 0.0748 | ||||||
| PC8 | -0.0871 | PC8 | -0.0601 | ||||||
| PC9 | 0.0919 | PC9 | 0.0999 | ||||||
| PC11 | -0.0678 | PC11 | -0.0745 | ||||||
| PC12 | 0.1752 | PC12 | 0.1223 | ||||||
| PC13 | 0.1557 | PC13 | 0.1900 | ||||||
| PC14 | 0.1555 | PC14 | 0.1657 | ||||||
| BMI | 0.0253 | ||||||||
| Leg FM | Intercept | 4.9806 | 0.89 | 0.77 | Intercept | -0.7249 | 0.9 | 0.7 | |
| PC1 | 0.0301 | PC1 | 0.0225 | ||||||
| PC2 | 0.3257 | PC2 | 0.0640 | ||||||
| PC3 | -0.1496 | PC3 | -0.0395 | ||||||
| PC4 | 0.1742 | PC4 | 0.0844 | ||||||
| PC5 | 0.2336 | PC5 | 0.1576 | ||||||
| PC6 | -0.3811 | PC6 | -0.3458 | ||||||
| PC7 | -0.6359 | PC7 | -0.5000 | ||||||
| PC8 | 0.2286 | PC8 | 0.3740 | ||||||
| PC9 | -0.1582 | PC9 | -0.1925 | ||||||
| PC10 | 0.6106 | PC10 | 0.3756 | ||||||
| PC11 | 0.4500 | PC11 | 0.2694 | ||||||
| PC13 | -0.6444 | PC13 | -0.4967 | ||||||
| Age | 0.0090 | ||||||||
| BMI | 0.1886 | ||||||||
| Leg FFM | Intercept | 7.6220 | 0.86 | 0.69 | Intercept | 2.1961 | 0.88 | 0.61 | |
| PC1 | 0.0535 | PC1 | 0.0432 | ||||||
| PC2 | 0.2321 | PC2 | -0.0418 | ||||||
| PC3 | -0.0540 | PC3 | 0.0506 | ||||||
| PC4 | -0.0477 | PC4 | -0.0836 | ||||||
| PC5 | 0.2668 | PC5 | 0.1021 | ||||||
| PC6 | -0.2926 | PC6 | -0.1679 | ||||||
| PC12 | 0.5220 | PC12 | 0.2473 | ||||||
| PC13 | 0.2352 | PC13 | 0.3576 | ||||||
| PC14 | 0.4598 | PC14 | 0.3194 | ||||||
| Age | -0.0064 | ||||||||
| BMI | 0.2051 | ||||||||
| Trunk FM | Intercept | 12.2398 | 0.95 | 1.45 | Intercept | 12.2398 | 0.95 | 1.45 | |
| PC1 | 0.0523 | PC1 | 0.0523 | ||||||
| PC2 | 1.1211 | PC2 | 1.1211 | ||||||
| PC3 | -0.6125 | PC3 | -0.6125 | ||||||
| PC4 | 0.3434 | PC4 | 0.3434 | ||||||
| PC6 | 0.3615 | PC6 | 0.3615 | ||||||
| PC8 | -0.7211 | PC8 | -0.7211 | ||||||
| PC10 | 0.3308 | PC10 | 0.3308 | ||||||
| PC11 | 0.6288 | PC11 | 0.6288 | ||||||
| PC13 | -1.3325 | PC13 | -1.3325 | ||||||
| PC15 | -0.5152 | PC15 | -0.5152 | ||||||
| Trunk FFM | Intercept | 23.3186 | 0.9 | 1.55 | Intercept | 15.7269 | 0.9 | 1.56 | |
| PC1 | 0.1303 | PC1 | 0.1205 | ||||||
| PC2 | 0.7044 | PC2 | 0.3419 | ||||||
| PC3 | -0.3264 | PC3 | -0.1800 | ||||||
| PC4 | -0.2970 | PC4 | -0.3386 | ||||||
| PC5 | 0.3798 | PC5 | 0.2091 | ||||||
| PC6 | 0.1300 | PC6 | 0.2489 | ||||||
| PC7 | 0.6023 | PC7 | 0.6429 | ||||||
| PC8 | -0.5750 | PC8 | -0.3958 | ||||||
| PC11 | 0.4349 | PC11 | 0.1498 | ||||||
| PC12 | 0.7638 | PC12 | 0.5748 | ||||||
| PC13 | 0.2979 | PC13 | 0.4630 | ||||||
| PC14 | 0.7890 | PC14 | 0.6475 | ||||||
| PC15 | 0.5148 | PC15 | 0.4326 | ||||||
| BMI | 0.2745 | ||||||||
| Males | Total FM | Intercept | 23.4226 | 0.94 | 3.04 | Intercept | 24.2945 | 0.94 | 2.85 |
| PC1 | -0.1924 | PC1 | -0.1883 | ||||||
| PC2 | 2.0747 | PC2 | 1.9407 | ||||||
| PC3 | -1.4689 | PC3 | -1.3345 | ||||||
| PC4 | -1.2207 | PC4 | -1.2314 | ||||||
| PC5 | 0.9020 | PC5 | 0.9058 | ||||||
| PC6 | 0.6394 | PC6 | 0.5996 | ||||||
| PC7 | -2.3710 | PC7 | -1.8262 | ||||||
| PC8 | 0.9488 | PC8 | 0.8188 | ||||||
| PC9 | -1.5091 | PC9 | -1.6708 | ||||||
| PC10 | -1.4176 | PC10 | -0.6422 | ||||||
| PC11 | 2.8085 | PC11 | 2.6083 | ||||||
| PC12 | 2.0674 | PC12 | 1.5622 | ||||||
| PC14 | 1.1338 | PC14 | 1.1794 | ||||||
| PC15 | -1.2783 | PC15 | -1.1086 | ||||||
| Age | -0.0311 | ||||||||
| Total FFM | Total Mass - Total Fat Mass | 0.94 | 2.97 | Total Mass - Total Fat Mass | 0.95 | 2.63 | |||
| Percent Fat | Total Fat Mass / Total Mass | 0.73 | 3.31 | Total Fat Mass / Total Mass | 0.77 | 3.17 | |||
| VAT | Intercept | 0.5050 | 0.73 | 0.14 | Intercept | 0.3881 | 0.75 | 0.14 | |
| PC1 | -0.0025 | PC1 | -0.0024 | ||||||
| PC2 | 0.0360 | PC2 | 0.0382 | ||||||
| PC3 | -0.0256 | PC3 | -0.0280 | ||||||
| PC4 | -0.0489 | PC4 | -0.0390 | ||||||
| PC5 | 0.0712 | PC5 | 0.0616 | ||||||
| PC6 | 0.0489 | PC6 | 0.0401 | ||||||
| PC7 | 0.0919 | PC7 | 0.0633 | ||||||
| PC9 | -0.0505 | PC9 | -0.0340 | ||||||
| PC10 | 0.0453 | PC10 | 0.0206 | ||||||
| PC11 | 0.0859 | PC11 | 0.0641 | ||||||
| PC13 | -0.0786 | PC13 | -0.0813 | ||||||
| PC14 | 0.0598 | PC14 | 0.0437 | ||||||
| PC15 | 0.0984 | PC15 | 0.0930 | ||||||
| Age | 0.0030 | ||||||||
| Arm FM | Intercept | 1.4169 | 0.86 | 0.3 | Intercept | 1.4169 | 0.86 | 0.3 | |
| PC1 | -0.0109 | PC1 | -0.0109 | ||||||
| PC2 | 0.1403 | PC2 | 0.1403 | ||||||
| PC3 | -0.0888 | PC3 | -0.0888 | ||||||
| PC4 | -0.0633 | PC4 | -0.0633 | ||||||
| PC5 | 0.0538 | PC5 | 0.0538 | ||||||
| PC7 | -0.1132 | PC7 | -0.1132 | ||||||
| PC8 | 0.0640 | PC8 | 0.0640 | ||||||
| PC11 | 0.1479 | PC11 | 0.1478 | ||||||
| PC12 | 0.1454 | PC12 | 0.1454 | ||||||
| PC15 | -0.0818 | PC15 | -0.0818 | ||||||
| Arm FFM | Intercept | 4.3909 | 0.79 | 0.43 | Intercept | -2.2843 | 0.88 | 0.36 | |
| PC1 | -0.0246 | PC1 | -0.0089 | ||||||
| PC2 | 0.1095 | PC2 | -0.1596 | ||||||
| PC3 | 0.0468 | PC3 | 0.1405 | ||||||
| PC4 | 0.0861 | PC4 | 0.2015 | ||||||
| PC5 | -0.2594 | PC5 | -0.1733 | ||||||
| PC6 | -0.1831 | PC6 | -0.1072 | ||||||
| PC7 | 0.2161 | PC7 | 0.1650 | ||||||
| PC8 | -0.1722 | PC8 | -0.2115 | ||||||
| PC9 | 0.2050 | PC9 | 0.3574 | ||||||
| PC11 | -0.3739 | PC11 | -0.4795 | ||||||
| PC12 | 0.1182 | PC12 | -0.0916 | ||||||
| PC14 | -0.2356 | PC14 | -0.1463 | ||||||
| PC15 | 0.1522 | PC15 | 0.1853 | ||||||
| Age | 0.0070 | ||||||||
| BMI | 0.2198 | ||||||||
| Leg FM | Intercept | 3.8794 | 0.88 | 0.7 | Intercept | 3.8794 | 0.88 | 0.7 | |
| PC1 | -0.0329 | PC1 | -0.0329 | ||||||
| PC2 | 0.3032 | PC2 | 0.3032 | ||||||
| PC3 | -0.2220 | PC3 | -0.2221 | ||||||
| PC4 | -0.1629 | PC4 | -0.1629 | ||||||
| PC7 | -0.8775 | PC7 | -0.8775 | ||||||
| PC8 | 0.1725 | PC8 | 0.1725 | ||||||
| PC9 | -0.3007 | PC9 | -0.3007 | ||||||
| PC10 | -0.4316 | PC10 | -0.4316 | ||||||
| PC11 | 0.4213 | PC11 | 0.4213 | ||||||
| PC12 | 0.4360 | PC12 | 0.4360 | ||||||
| PC15 | -0.6294 | PC15 | -0.6294 | ||||||
| Leg FFM | Intercept | 11.0904 | 0.84 | 0.84 | Intercept | 6.5031 | 0.85 | 0.81 | |
| PC1 | -0.0678 | PC1 | -0.0537 | ||||||
| PC2 | 0.2644 | PC2 | 0.0743 | ||||||
| PC4 | 0.0760 | PC4 | 0.1364 | ||||||
| PC5 | -0.4727 | PC5 | -0.4295 | ||||||
| PC6 | -0.4029 | PC6 | -0.3450 | ||||||
| PC11 | -0.2748 | PC11 | -0.2650 | ||||||
| PC12 | 0.4749 | PC12 | 0.3277 | ||||||
| PC14 | -0.3186 | PC14 | -0.2001 | ||||||
| BMI | 0.1572 | ||||||||
| Trunk FM | Intercept | 11.6548 | 0.94 | 1.66 | Intercept | 12.2359 | 0.94 | 1.54 | |
| PC1 | -0.1024 | PC1 | -0.1009 | ||||||
| PC2 | 1.1661 | PC2 | 1.1352 | ||||||
| PC3 | -0.8335 | PC3 | -0.8090 | ||||||
| PC4 | -0.7817 | PC4 | -0.7977 | ||||||
| PC5 | 0.8185 | PC5 | 0.8424 | ||||||
| PC6 | 0.5521 | PC6 | 0.5651 | ||||||
| PC7 | -0.3729 | PC7 | -0.2483 | ||||||
| PC8 | 0.4690 | PC8 | 0.3791 | ||||||
| PC9 | -0.8747 | PC9 | -0.9111 | ||||||
| PC10 | -0.5862 | PC10 | -0.4334 | ||||||
| PC11 | 1.6527 | PC11 | 1.5991 | ||||||
| PC12 | 0.9004 | PC12 | 0.7377 | ||||||
| PC14 | 0.8207 | PC14 | 0.7938 | ||||||
| Age | -0.0152 | ||||||||
| Trunk FFM | Intercept | 32.6704 | 0.91 | 1.86 | Intercept | 18.9138 | 0.91 | 1.83 | |
| PC1 | -0.1891 | PC1 | -0.1519 | ||||||
| PC2 | 0.9060 | PC2 | 0.3550 | ||||||
| PC3 | -0.2653 | PC3 | -0.0705 | ||||||
| PC4 | 0.3630 | PC4 | 0.5238 | ||||||
| PC5 | -0.5713 | PC5 | -0.3227 | ||||||
| PC6 | -0.7049 | PC6 | -0.4851 | ||||||
| PC7 | 1.2259 | PC7 | 1.1442 | ||||||
| PC11 | -0.6362 | PC11 | -0.6084 | ||||||
| PC12 | 0.8325 | PC12 | 0.1553 | ||||||
| PC14 | -0.6104 | PC14 | -0.4821 | ||||||
| BMI | 0.4737 | ||||||||
Abbreviations: DXA, dual-energy X-ray absorptiometry; PC, principal component; RMSE, root mean squared error; VAT, visceral adipose tissue.
PC-only: stepwise forward linear regression with only PCs as possible predictors of DXA body composition outputs.
PC+Demographics: stepwise forward linear regression with PCs, age, and BMI as possible predictors of DXA body composition outputs.
3DO test–retest precision was comparable to DXA (Table 3). 3DO total FM and FFM achieved a %CV (RMSE) of 1.76% (0.44 kg) and 0.96% (0.44 kg) in females, whereas DXA had a %CV (RMSE) of 0.98% (0.24 kg) and 0.59% (0.27 kg), respectively. In females, 3DO VAT, arm FM, and trunk FM (%CV: 4.4%, 2.72%, and 1.55%; RMSE: 0.02 kg, 0.04 kg, and 0.18 kg, respectively) achieved better test–retest precision than DXA (%CV: 8.08%, 3.16%, and 2.04%; RMSE: 0.03 kg, 0.05 kg, and 0.23 kg, respectively). Test–retest precision in males showed similar results. However, only 3DO VAT had better test–retest precision than DXA in males. When comparing the precision across subgroups of age, BMI, and ethnicity with 1-factor ANOVA, there were no significant differences for all 3DO whole-body and regional body composition outputs (data not shown, P > 0.06).
TABLE 3.
Test–retest precision of 3D optical and DXA in %CV (RMSE) (n = 639)
| Variable | DXA Female | 3DO Female | DXA Male | 3DO Male |
|---|---|---|---|---|
| Total FM, kg | 0.98 (0.24) | 1.76 (0.44) | 1.37 (0.28) | 2.93 (0.60) |
| Total FFM, kg | 0.59 (0.27) | 0.96 (0.44) | 0.55 (0.36) | 0.91 (0.60) |
| Percent Fat, % | NA (0.33) | NA (0.66) | NA (0.31) | NA (0.74) |
| VAT, kg | 8.08 (0.03) | 4.40 (0.02) | 6.88 (0.03) | 4.78 (0.02) |
| Arm FM, kg | 3.16 (0.05) | 2.72 (0.04) | 3.43 (0.04) | 3.99 (0.05) |
| Arm FFM, kg | 2.14 (0.05) | 2.57 (0.06) | 1.89 (0.08) | 1.94 (0.08) |
| Leg FM, kg | 1.47 (0.07) | 2.90 (0.14) | 2.34 (0.08) | 4.29 (0.14) |
| Leg FFM, kg | 1.13 (0.08) | 1.95 (0.14) | 1.18 (0.13) | 1.22 (0.13) |
| Trunk FM, kg | 2.04 (0.23) | 1.55 (0.18) | 2.42 (0.25) | 3.03 (0.31) |
| Trunk FFM, kg | 1.16 (0.26) | 1.52 (0.34) | 1.07 (0.34) | 1.20 (0.38) |
Abbreviations: 3DO, 3-dimensional optical; CV, coefficient of variation; DXA, dual-energy X-ray absorptiometry; RMSE, root mean squared error; VAT, visceral adipose tissue.
When comparing 3DO total FM, FFM, %fat, and VAT to DXA by age subgroups (Table 4), there were no MDs in either females or males (P > 0.068). For the BMI subgroups (Table 5), females with underweight had significant differences for total FM, FFM, and %fat (MDs: 1.23 kg, −1.23 kg, and 3.12%, respectively, P < 0.014) as well as %fat for the females with obesity (MD = −0.79%, P = 0.023). There were no MDs across the male BMI subgroups (P > 0.314). In the ethnicity subgroups (Table 6), female Asian, Black, and NHOPI subgroups had MDs in total FM (MDs: 0.64 kg, −0.81 kg, −1.55 kg%, respectively), FFM (MDs: −1.6 kg, 1.7 kg, and 3 kg%, respectively), and %fat (MDs: 1.38%, −1.11%, and −2.24%, respectively) (P < 0.017). Additionally, the male NHOPI subgroup had MDs for total FM, FFM, and %fat (MDs: −1.45 kg, 1.45 kg, and −1.88%, respectively; P < 0.038). If ethnicity was added to the model (data not shown), differences in ethnicity subgroups were no longer significant (P > 0.99). However, underweight females still had a MD (P = 0.029).
TABLE 4.
3D optical vs. DXA body composition by age group
| Sex | Groups | n | Variable | Mean Difference | % Mean Difference | P |
|---|---|---|---|---|---|---|
| Female | Young | 126 | Total FM | -0.28 | N.S. | 0.303 |
| Middle Aged | 119 | Total FM | 0.05 | N.S. | 0.828 | |
| Senior | 89 | Total FM | -0.03 | N.S. | 0.901 | |
| Young | 126 | Total FFM | 0.28 | N.S. | 0.303 | |
| Middle Aged | 119 | Total FFM | -0.05 | N.S. | 0.828 | |
| Senior | 89 | Total FFM | 0.03 | N.S. | 0.901 | |
| Young | 126 | %Fat | -0.23 | N.S. | 0.559 | |
| Middle Aged | 119 | %Fat | 0.19 | N.S. | 0.566 | |
| Senior | 89 | %Fat | 0.04 | N.S. | 0.932 | |
| Young | 126 | VAT | -0.01 | N.S. | 0.204 | |
| Middle Aged | 119 | VAT | 0.02 | N.S. | 0.068 | |
| Senior | 89 | VAT | -0.01 | N.S. | 0.438 | |
| Male | Young | 127 | Total FM | -0.01 | N.S. | 0.960 |
| Middle Aged | 93 | Total FM | 0.25 | N.S. | 0.373 | |
| Senior | 61 | Total FM | -0.36 | N.S. | 0.288 | |
| Young | 127 | Total FFM | 0.01 | N.S. | 0.960 | |
| Middle Aged | 93 | Total FFM | -0.25 | N.S. | 0.373 | |
| Senior | 61 | Total FFM | 0.36 | N.S. | 0.288 | |
| Young | 127 | %Fat | 0.03 | N.S. | 0.923 | |
| Middle Aged | 93 | %Fat | 0.35 | N.S. | 0.261 | |
| Senior | 61 | %Fat | -0.37 | N.S. | 0.384 | |
| Young | 127 | VAT | -0.00 | N.S. | 0.893 | |
| Middle Aged | 93 | VAT | 0.01 | N.S. | 0.508 | |
| Senior | 61 | VAT | -0.01 | N.S. | 0.589 |
Abbreviations: 3DO, 3-dimensional optical; DXA, dual-energy X-ray absorptiometry; %Fat, percent fat; N.S., not significant; VAT, visceral adipose tissue.
Mean differences = 3DO – DXA; P value: Student’s t test between 3DO and DXA
PC+Demo model was used in this comparison. FM, FFM, and VAT are measured in kg.
Young (18–39 y old), middle aged (40–59 y old), senior (≥60 y old).
TABLE 5.
3D optical vs. DXA body composition by BMI
| Sex | Groups | n | Variable | Mean Difference | %Mean Difference | P |
|---|---|---|---|---|---|---|
| Female | Underweight | 24 | Total FM | 1.23 | 12 | 0.014∗ |
| Normal | 119 | Total FM | -0.01 | N.S. | 0.958 | |
| Overweight | 94 | Total FM | -0.13 | N.S. | 0.610 | |
| Obese | 96 | Total FM | -0.54 | N.S. | 0.099 | |
| Underweight | 24 | Total FFM | -1.23 | -3.4 | 0.014∗ | |
| Normal | 89 | Total FFM | 0.01 | N.S. | 0.958 | |
| Overweight | 94 | Total FFM | 0.13 | N.S. | 0.610 | |
| Obese | 96 | Total FFM | 0.54 | N.S. | 0.099 | |
| Underweight | 24 | %Fat | 3.12 | — | 0.010∗ | |
| Normal | 119 | %Fat | -0.10 | N.S. | 0.785 | |
| Overweight | 94 | %Fat | -0.18 | N.S. | 0.616 | |
| Obese | 96 | %Fat | -0.79 | — | 0.023∗ | |
| Underweight | 24 | VAT | -0.01 | N.S. | 0.772 | |
| Normal | 119 | VAT | -0.01 | N.S. | 0.238 | |
| Overweight | 94 | VAT | 0.01 | N.S. | 0.286 | |
| Obese | 96 | VAT | 0.00 | N.S. | 0.988 | |
| Male | Underweight | 3 | Total FM | 2.97 | N.S. | 0.155 |
| Normal | 80 | Total FM | 0.11 | N.S. | 0.701 | |
| Overweight | 112 | Total FM | -0.06 | N.S. | 0.779 | |
| Obese | 85 | Total FM | -0.16 | N.S. | 0.633 | |
| Underweight | 3 | Total FFM | -2.97 | N.S. | 0.155 | |
| Normal | 80 | Total FFM | -0.11 | N.S. | 0.701 | |
| Overweight | 112 | Total FFM | 0.06 | N.S. | 0.779 | |
| Obese | 85 | Total FFM | 0.16 | N.S. | 0.633 | |
| Underweight | 3 | %Fat | 7.12 | N.S. | 0.168 | |
| Normal | 80 | %Fat | 0.21 | N.S. | 0.601 | |
| Overweight | 112 | %Fat | -0.06 | N.S. | 0.804 | |
| Obese | 85 | %Fat | -0.29 | N.S. | 0.356 | |
| Underweight | 3 | VAT | -0.04 | N.S. | 0.407 | |
| Normal | 80 | VAT | -0.01 | N.S. | 0.347 | |
| Overweight | 112 | VAT | 0.01 | N.S. | 0.380 | |
| Obese | 85 | VAT | 0.00 | N.S. | 0.936 |
Abbreviations: 3DO, 3-dimensional optical; BMI, body mass index; DXA, dual-energy X-ray absorptiometry; %Fat, percent fat; N.S., not significant; VAT, visceral adipose tissue.
Mean differences = 3DO - DXA; P value: Student’s t test between 3DO and DXA
PC+Demo model was used in this comparison. FM, FFM, and VAT are measured in kg.
BMI classifications: Underweight (BMI <18.5), Normal (18.5–24.9), Overweight (25–29.9), Obese (BMI ≥ 30)
∗ p-values less than 0.05.
TABLE 6.
3D optical vs. DXA body composition by ethnic group
| Sex | Groups | n | Variable | Mean Difference | %Mean Difference | P-Value |
|---|---|---|---|---|---|---|
| Female | Asian | 80 | Total FM | 0.64 | 3.5 | 0.017∗ |
| Black | 70 | Total FM | -0.81 | -2.6 | 0.015∗ | |
| Hispanic | 47 | Total FM | 0.36 | N.S. | 0.386 | |
| NHOPI | 28 | Total FM | -1.55 | -5.4 | 0.003∗ | |
| White | 107 | Total FM | 0.01 | N.S. | 0.980 | |
| Asian | 80 | Total FFM | -0.64 | -1.6 | 0.017∗ | |
| Black | 70 | Total FFM | 0.81 | 1.7 | 0.015∗ | |
| Hispanic | 47 | Total FFM | -0.36 | N.S. | 0.386 | |
| NHOPI | 28 | Total FFM | 1.55 | 3 | 0.003∗ | |
| White | 107 | Total FFM | -0.01 | N.S. | 0.980 | |
| Asian | 80 | %Fat | 1.38 | — | 0.006∗ | |
| Black | 70 | %Fat | -1.11 | — | 0.010∗ | |
| Hispanic | 47 | %Fat | 0.63 | N.S. | 0.305 | |
| NHOPI | 28 | %Fat | -2.24 | — | 0.002∗ | |
| White | 107 | %Fat | -0.01 | N.S. | 0.979 | |
| Asian | 80 | VAT | -0.00 | N.S. | 0.908 | |
| Black | 70 | VAT | -0.02 | N.S. | 0.189 | |
| Hispanic | 47 | VAT | 0.01 | N.S. | 0.627 | |
| NHOPI | 28 | VAT | -0.00 | N.S. | 0.942 | |
| White | 107 | VAT | 0.01 | N.S. | 0.362 | |
| Male | Asian | 71 | Total FM | 0.29 | N.S. | 0.360 |
| Black | 62 | Total FM | 0.42 | N.S. | 0.164 | |
| Hispanic | 32 | Total FM | -0.65 | N.S. | 0.226 | |
| NHOPI | 16 | Total FM | -1.45 | -6.5 | 0.038∗ | |
| White | 98 | Total FM | -0.01 | N.S. | 0.958 | |
| Asian | 71 | Total FFM | -0.29 | N.S. | 0.360 | |
| Black | 62 | Total FFM | -0.42 | N.S. | 0.164 | |
| Hispanic | 32 | Total FFM | 0.65 | N.S. | 0.226 | |
| NHOPI | 16 | Total FFM | 1.45 | 1.9 | 0.038∗ | |
| White | 98 | Total FFM | 0.01 | N.S. | 0.958 | |
| Asian | 71 | %Fat | 0.50 | N.S. | 0.258 | |
| Black | 62 | %Fat | 0.57 | N.S. | 0.093 | |
| Hispanic | 32 | %Fat | -0.57 | N.S. | 0.374 | |
| NHOPI | 16 | %Fat | -1.88 | — | 0.013∗ | |
| White | 98 | %Fat | -0.08 | N.S. | 0.789 | |
| Asian | 71 | VAT | -0.01 | N.S. | 0.421 | |
| Black | 62 | VAT | -0.03 | N.S. | 0.105 | |
| Hispanic | 32 | VAT | 0.02 | N.S. | 0.159 | |
| NHOPI | 16 | VAT | 0.02 | N.S. | 0.669 | |
| White | 98 | VAT | 0.02 | N.S. | 0.159 |
Abbreviations: 3DO, 3-dimensional optical; DXA, dual-energy X-ray absorptiometry; %Fat, percent fat; NHOPI, Native Hawaiian or other Pacific Islander; N.S., not significant; VAT, visceral adipose tissue.
Mean differences = 3DO - DXA; P value: Student’s t test between 3DO and DXA
PC+Demo model was used in this comparison. FM, FFM, and VAT are measured in kg.
∗ p-values less than 0.05.
Average shape images were created for both males and females by ethnic group (FIGURE 1, FIGURE 2) to show MDs by region. Asians had significant differences in the trunk and arm measures, Blacks had differences in the head measures, and NHOPIs had differences in the head and leg measures. Although difficult to see in the figures, Black females on average had longer legs and shorter torsos, whereas NHOPI females held more mass in the torso, compared with their Asian counterparts. White males were taller on average, whereas NHOPIs held more mass in their abdomen on average, compared with other ethnicities.
Discussion
In this study, 3DO body composition accuracy and precision were evaluated at the subgroup level of age, BMI, and ethnicity. Accuracy of subgroups was compared to DXA, whereas precision was an intragroup comparison. The analysis showed MDs in accuracy among females with underweight, NHOPI females and males, and Asian and Black females, whereas all other groups showed no differences. We detected no significant differences in precision among the subgroups. Differences between 3DO and DXA body composition measures were not observed in 19 out of the 24 subgroups that were evaluated. However, it is worth noting that 3 of the 5 subgroups where differences were observed were among female ethnicity subgroups. However, it is important to keep in mind the differences with small to moderate %MD. Specific equations may be needed for those groups with differences.
Overall, the accuracy and precision were similar to previous work done by our group on interim analysis of the cohort that used the Fit3D [29]. The previous publication reported total FM R2s of 0.90 and 0.95 for males (n = 159) and females (n = 202), respectively, which was similar to the current analysis. The follow-up analysis used 3 different 3DO scanners to build an agnostic model and reported total FM R2s of 0.93 and 0.94 for males and females, respectively [28]. However, we had not previously presented subgroup analysis due to the limited statistical power.
In the female underweight group, there was a large difference between 3DO and DXA for total FM and a moderate difference for FFM (%MD = 12% and −3.4%, respectively). The %MD was proportionately higher despite a small difference due to the lower FM in the underweight group (approximately 10 kg). Another factor could be ethnicity because models were not initially adjusted for this. Female Asians, Blacks, and NHOPIs had MDs, and 44% of the underweight group were either Asian or Black. However, after adjustment for ethnicity, 3DO still overestimated total FM in underweight females. Thus, our models may not have seen enough shape variance in the underweight sample (underrepresentation) or other unique shape features from this group. The female subgroup with obesity also had a MD in %fat (−0.79%). However, this MD was very small, especially in a group with the highest mean %fat (41%). Since the difference was small and the total FM and FFM were not significant, the %fat difference may have been due to chance. It is worth noting that females with obesity may exhibit greater shape variability due to differences in how fat is distributed throughout their bodies. Specifically, the location of fat deposition may vary among the abdomen, hips, or appendages [38]. The largest differences were seen in the NHOPI group for both males and females. The differences may be driven by the low representation in comparison to other ethnic groups. In addition, differences in shape can also be driving these differences. Certain shape characteristics in Asian, Black, and NHOPI participants could be different from the majority of the group, which led to under- or overestimations as shown in Figure 1.
Compared with other studies that reported the accuracy by subgroups, Graybeal et al. [39] examined the accuracy of 3DO body composition estimates derived from smartphone-based applications (apps) with respect to the 4C model. For the HALO app, the authors reported a MD for total FM in males (MD: 2.1 kg; P < 0.001) but not in females (MD: 0.2 kg). In addition, they reported MDs for total FM in Whites and Blacks (MD: 0.9 kg and 1.0 kg, respectively; P < 0.05). In comparison to the current study, we did not observe MDs by sex or in the White ethnic group. However, we also reported MDs in the Black, female ethnic group. It is worth pointing out that our criterion was DXA, whereas Graybeal et al. used the 4C model.
Although not related to the Fit3D, other investigators have explored novel methods of using 2D imaging for body composition from accessible smartphone technology. The 2D methods generally capture an image(s) of a person and utilize machine learning methods to estimate body composition values to criterion methods such as DXA or the 4C model. Our previous study used 2D images taken from consumer-grade cameras and used the silhouettes to create a 3D mesh, which was used to estimate the body composition. We reported total FM R2s of 0.96 and 0.94 in male and female test sets, respectively [30]. Nana et al. [40] reported 3DO body composition accuracy by sex from another smartphone app, Body Composition Technologies, with respects to DXA. The authors reported a %fat MD of 0.1% and −0.1% for males and females, respectively, which was similar to our findings. Farina et al. [41] developed a method to estimate DXA body composition using demographics and lateral surface image occupancy. The authors achieved R2s of 0.97 and 0.95 for females and males, respectively, which was marginally better than ours. However, our sample size was about 5.4 times larger. Aside from traditional, commercial 3DO scanners, 2D methods showed promising results. Further investigation is warranted to evaluate the 2D methods’ ability to monitor changes and accuracy across different ethnic, age, and BMI subgroups.
Other groups have also examined the precision of 3DO devices. Tinsley et al. [42] evaluated the precision the Fit3D ProScanner, Styku S100, Naked, and Size Stream SS20. Although the precision reported was not broken into subgroups, the overall precision of the devices achieved a %CV that ranged from 2.5% to 4.3% for total FM and 0.7% to 1.4% for total FFM with the Fit3D ProScanner at the high end for FM and FFM. Compared to the results in Table 3, our modeling method achieved a lower %CV with a larger sample size. Additionally, we did not observe differences in precision by subgroups using our models. The precision of the device/model will determine the ability to monitor smaller changes [43]. Additional evaluation of subgroup precision is necessary to validate the monitoring of FM and FFM changes in specific subgroups using commercial body composition systems.
The strength of the study was the scrutiny at the subgroup level. Although other studies have looked at accuracy by sex, BMI, and ethnicity, the current analysis examined more strata. In addition, only a single equation was needed for the majority of the subgroups that were explored. A subgroup-specific equation may be needed for those with MDs. However, there were also limitations. Some subgroups were underrepresented (ie, NHOPI and underweight). As such, the outcomes and interpretations of these groups should be cautioned as they were not as powered. The current analysis was completed with a healthy sample, so the results may not be applicable to patients with body composition-altering diseases. Future work can focus on populations with diseases that need constant body composition monitoring (younger populations [birth to 17 y]), and more accessible 3DO methods such as apps from smartphones [44,45].
In conclusion, 3DO body composition has advanced greatly in accuracy and precision in recent years. This work provides further insight of 3DO’s ability to estimate body composition. Although 19 of the 24 subgroups had MDs from DXA, the majority of the subgroups did not display MDs using a single 3DO body composition model. Specific equations for the 5 subgroups that had MDs may improve the agreement. Beyond clinical settings, the accessibility and safety of these devices are appealing for those that want to monitor body composition frequently.
Acknowledgments
We thank all the participants for graciously giving their time to be part of the studies, our collaborators at each site for their part in the study, Tyler Carter and Greg Moore at Fit3D for providing us the 3DO data, and Naureen Mahmood and Talha Zaman for providing us the application program interface to register and repose our 3DO data.
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.ajcnut.2023.07.010.
Author contributions
The authors’ responsibilities were as follows – MCW, JAS: designed and conducted the research; MCW, BQ, JAS: were part of the data analysis; MCW, YEL, NNK, CM, AKG, GM, SBH, JAS: were in charge of their respective study recruitment and protocols; MCW, JAS: drafted the manuscript and were responsible over the final content; and all authors: read and approved the final manuscript.
Conflict of interest
SBH reports his role on the Medical Advisory Boards of Tanita Corporation, Amgen, and Medifast; he is also an Amazon Scholar. JAS reports his role as a scientific advisor for Styku and Hologic. All other authors report no conflicts of interest.
Funding
Phases of this study were funded by the National Institute of Diabetes and Digestive and Kidney Diseases (NIH R01DK109008).
Data availability
Data described in the manuscript, code book, and analytic code will be made available upon request pending researchers who meet the criteria for access to confidential data. To request an application or to use our statistical shape model to transform your Meshcapade processed mesh into usable principal components, please contact John Shepherd (johnshep@hawaii.edu) or visit www.shapeup.shepherdresearchlab.org.
Appendix A. Supplementary data
The following is the Supplementary data to this article:
References
- 1.Bennett J., Wong M.C., McCarthy C., Fearnbach N., Queen K., Shepherd J., et al. Emergence of the adolescent obesity epidemic in the United States: five-decade visualization with humanoid avatars. Int. J. Obes. (Lond) 2022;46(9):1587–1590. doi: 10.1038/s41366-022-01153-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Hales C.M., Carroll M.D., Fryar C.D., Ogden C.L. 2020. Prevalence of obesity and severe obesity among adults: United States, 2017-2018.https://www.cdc.gov/nchs/products/databriefs/db360.htm [Internet] Available from: [PubMed] [Google Scholar]
- 3.Wong M.C., McCarthy C., Fearnbach N., Yang S., Shepherd J., Heymsfield S.B. Emergence of the obesity epidemic: 6-decade visualization with humanoid avatars. Am. J. Clin. Nutr. 2022;115(4):1189–1193. doi: 10.1093/ajcn/nqac005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Calle E.E., Rodriguez C., Walker-Thurmond K., Thun M.J. Overweight, obesity, and mortality from cancer in a prospectively studied cohort of U.S. adults. N. Engl. J. Med. 2003;348(17):1625–1638. doi: 10.1056/NEJMoa021423. [DOI] [PubMed] [Google Scholar]
- 5.Després J.P., Lemieux I. Abdominal obesity and metabolic syndrome. Nature. 2006;444(7121):881–887. doi: 10.1038/nature05488. [DOI] [PubMed] [Google Scholar]
- 6.Grundy S.M. Obesity, metabolic syndrome, and cardiovascular disease. J. Clin. Endocrinol. Metab. 2004;89(6):2595–2600. doi: 10.1210/jc.2004-0372. [DOI] [PubMed] [Google Scholar]
- 7.Müller M.J., Lagerpusch M., Enderle J., Schautz B., Heller M., Bosy-Westphal A. Beyond the body mass index: tracking body composition in the pathogenesis of obesity and the metabolic syndrome. Obes. Rev. 2012;13(13):6–13. doi: 10.1111/j.1467-789X.2012.01033.x. suppl 2. [DOI] [PubMed] [Google Scholar]
- 8.Ward L.C. Human body composition: yesterday, today, and tomorrow. Eur. J. Clin. Nutr. 2018;72(9):1201–1207. doi: 10.1038/s41430-018-0210-2. [DOI] [PubMed] [Google Scholar]
- 9.Direk K., Cecelja M., Astle W., Chowienczyk P., Spector T.D., Falchi M., et al. The relationship between DXA-based and anthropometric measures of visceral fat and morbidity in women. BMC Cardiovasc. Disord. 2013;13(1):25. doi: 10.1186/1471-2261-13-25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lim U., Monroe K.R., Buchthal S., Fan B., Cheng I., Kristal B.S., et al. Propensity for intra-abdominal and hepatic adiposity varies among ethnic groups. Gastroenterology. 2019;156(4):966–975.e10. doi: 10.1053/j.gastro.2018.11.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Maskarinec G., Shvetsov Y.B., Wong M.C., Garber A., Monroe K., Ernst T.M., et al. Subcutaneous and visceral fat assessment by DXA and MRI in older adults and children. Obesity (Silver Spring) 2022;30(4):920–930. doi: 10.1002/oby.23381. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Heymsfield S.B., Wang Z., Baumgartner R.N., Ross R. Human body composition: advances in models and methods. Annu. Rev. Nutr. 1997;17(1):527–558. doi: 10.1146/annurev.nutr.17.1.527. [DOI] [PubMed] [Google Scholar]
- 13.Lukaski H.C. Methods for the assessment of human body composition: traditional and new. Am. J. Clin. Nutr. 1987;46(4):537–556. doi: 10.1093/ajcn/46.4.537. [DOI] [PubMed] [Google Scholar]
- 14.Shepherd J.A., Ng B.K., Sommer M.J., Heymsfield S.B. Body composition by DXA. Bone. 2017;104:101–105. doi: 10.1016/j.bone.2017.06.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.McCrory M.A., Gomez T.D., Bernauer E.M., Molé P.A. Evaluation of a new air displacement plethysmograph for measuring human body composition. Med. Sci. Sports Exerc. 1995;27(12):1686–1691. doi: 10.1249/00005768-199512000-00016. [DOI] [PubMed] [Google Scholar]
- 16.Abu Khaled M., McCutcheon M.J., Reddy S., Pearman P.L., Hunter G.R., Weinsier R.L. Electrical impedance in assessing human body composition: the BIA method. Am. J. Clin. Nutr. 1988;47(5):789–792. doi: 10.1093/ajcn/47.5.789. [DOI] [PubMed] [Google Scholar]
- 17.Treleaven P., Wells J. 3D body scanning and healthcare applications. Computer. 2007;40(7):28–34. doi: 10.1109/MC.2007.225. [DOI] [Google Scholar]
- 18.Kennedy S., Hwaung P., Kelly N., Liu Y.E., Sobhiyeh S., Heo M., et al. Optical imaging technology for body size and shape analysis: evaluation of a system designed for personal use. Eur. J. Clin. Nutr. 2020;74(6):920–929. doi: 10.1038/s41430-019-0501-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Kennedy S., Smith B., Sobhiyeh S., Dechenaud M.E., Wong M., Kelly N., et al. Digital anthropometric evaluation of young children: comparison to results acquired with conventional anthropometry. Eur. J. Clin. Nutr. 2022;76(2):251–260. doi: 10.1038/s41430-021-00938-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Heymsfield S.B., Bourgeois B., Ng B.K., Sommer M.J., Li X., Shepherd J.A. Digital anthropometry: a critical review. Eur. J. Clin. Nutr. 2018;72(5):680–687. doi: 10.1038/s41430-018-0145-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Rumbo-Rodríguez L., Sánchez-SanSegundo M., Ferrer-Cascales R., García-D’Urso N., Hurtado-Sánchez J.A., Zaragoza-Martí A. Comparison of body scanner and manual anthropometric measurements of body shape: a systematic review. Int. J. Environ. Res. Public Health. 2021;18(12):6213. doi: 10.3390/ijerph18126213. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Tinsley G.M., Moore M.L., Dellinger J.R., Adamson B.T., Benavides M.L. Digital anthropometry via three-dimensional optical scanning: evaluation of four commercially available systems. Eur. J. Clin. Nutr. 2020;74(7):1054–1064. doi: 10.1038/s41430-019-0526-6. [DOI] [PubMed] [Google Scholar]
- 23.Bennett J.P., Liu Y.E., Quon B.K., Kelly N.N., Wong M.C., Kennedy S.F., et al. Assessment of clinical measures of total and regional body composition from a commercial 3-dimensional optical body scanner. Clin. Nutr. 2022;41(1):211–218. doi: 10.1016/j.clnu.2021.11.031. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Ng B.K., Hinton B.J., Fan B., Kanaya A.M., Shepherd J.A. Clinical anthropometrics and body composition from 3D whole-body surface scans. Eur. J. Clin. Nutr. 2016;70(11):1265–1270. doi: 10.1038/ejcn.2016.109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Wong M.C., Ng B.K., Kennedy S.F., Hwaung P., Liu E.Y., Kelly N.N., et al. Children and adolescents’ anthropometrics body composition from 3-D optical surface scans. Obesity (Silver Spring) 2019;27(11):1738–1749. doi: 10.1002/oby.22637. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Harty P.S., Sieglinger B., Heymsfield S.B., Shepherd J.A., Bruner D., Stratton M.T., et al. Novel body fat estimation using machine learning and 3-dimensional optical imaging. Eur. J. Clin. Nutr. 2020;74(5):842–845. doi: 10.1038/s41430-020-0603-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Sobhiyeh S., Borel N., Dechenaud M., Graham C.A., Wong M., Wolenski P., et al. Fully automated pipeline for body composition estimation from 3D optical scans using principal component analysis: a Shape UP study. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. 2020;2020:1853–1858. doi: 10.1109/EMBC44109.2020.9175211. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Tian I.Y., Wong M.C., Kennedy S., Kelly N.N., Liu Y.E., Garber A.K., et al. A device-agnostic shape model for automated body composition estimates from 3D optical scans. Med. Phys. 2022;49(10):6395–6409. doi: 10.1002/mp.15843. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Wong M.C., Ng B.K., Tian I., Sobhiyeh S., Pagano I., Dechenaud M., et al. A pose-independent method for accurate and precise body composition from 3D optical scans. Obesity (Silver Spring) 2021;29(11):1835–1847. doi: 10.1002/oby.23256. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Tian I.Y., Ng B.K., Wong M.C., Kennedy S., Hwaung P., Kelly N., et al. Predicting 3D body shape and body composition from conventional 2D photography. Med. Phys. 2020;47(12):6232–6245. doi: 10.1002/mp.14492. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Consultation, W. H. O. Vol. 894. 2000. pp. 1–253. (“Obesity: preventing and managing the global epidemic.” World Health Organization technical report series). [PubMed] [Google Scholar]
- 32.Hangartner T.N., Warner S., Braillon P., Jankowski L., Shepherd J. The Official Positions of the International Society for Clinical Densitometry: acquisition of dual-energy X-ray absorptiometry body composition and considerations regarding analysis and repeatability of measures. J. Clin. Densitom. 2013;16(4):520–536. doi: 10.1016/j.jocd.2013.08.007. [DOI] [PubMed] [Google Scholar]
- 33.Ng B.K., Liu Y.E., Wang W., Kelly T.L., Wilson K.E., Schoeller D.A., et al. Validation of rapid 4-component body composition assessment with the use of dual-energy X-ray absorptiometry and bioelectrical impedance analysis. Am. J. Clin. Nutr. 2018;108(4):708–715. doi: 10.1093/ajcn/nqy158. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Loper M., Mahmood N., Romero J., Pons-Moll G., Black M.J., SMPL a skinned multi-person linear model. ACM Trans. Graph. 2015;34(6):1–16. doi: 10.1145/2816795.2818013. [DOI] [Google Scholar]
- 35.Ng B.K., Sommer M.J., Wong M.C., Pagano I., Nie Y., Fan B., et al. Detailed 3-dimensional body shape features predict body composition, blood metabolites, and functional strength: the Shape UP! studies. Am. J. Clin. Nutr. 2019;110(6):1316–1326. doi: 10.1093/ajcn/nqz218. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Glüer C.C., Blake G., Lu Y., Blunt B.A., Jergas M., Genant H.K. Accurate assessment of precision errors: how to measure the reproducibility of bone densitometry techniques. Osteoporos. Int. 1995;5(4):262–270. doi: 10.1007/bf01774016. [DOI] [PubMed] [Google Scholar]
- 37.Lee H., Kim D., Jung A., Chae W. Ethnicity, social, and clinical risk factors to tooth loss among older adults in the U.S., NHANES 2011–2018. Int. J. Environ. Res. Public Health. 2022;19(4):2382. doi: 10.3390/ijerph19042382. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Duhon B.H., Phan T.T., Taylor S.L., Crescenzi R.L., Rutkowski J.M. Current mechanistic understandings of lymphedema and lipedema: tales of fluid, fat, and fibrosis. Int. J. Mol. Sci. 2022;23(12):6621. doi: 10.3390/ijms23126621. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Graybeal A.J., Brandner C.F., Tinsley G.M. Visual body composition assessment methods: a 4-compartment model comparison of smartphone-based artificial intelligence for body composition estimation in healthy adults. Clin. Nutr. 2022;41(11):2464–2472. doi: 10.1016/j.clnu.2022.09.014. [DOI] [PubMed] [Google Scholar]
- 40.Nana A., Staynor J.M.D., Arlai S., El-Sallam A., Dhungel N., Smith M.K. Agreement of anthropometric and body composition measures predicted from 2D smartphone images and body impedance scales with criterion methods. Obes. Res. Clin. Pract. 2022;16(1):37–43. doi: 10.1016/j.orcp.2021.12.006. [DOI] [PubMed] [Google Scholar]
- 41.Farina G.L., Spataro F., De Lorenzo A., Lukaski H. A smartphone application for personal assessments of body composition and phenotyping. Sensors (Basel) 2016;16(12):2163. doi: 10.3390/s16122163. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Tinsley G.M., Moore M.L., Benavides M.L., Dellinger J.R., Adamson B.T. 3-Dimensional optical scanning for body composition assessment: a 4-component model comparison of four commercially available scanners. Clin. Nutr. 2020;39(10):3160–3167. doi: 10.1016/j.clnu.2020.02.008. [DOI] [PubMed] [Google Scholar]
- 43.Shepherd J.A., Lu Y. A generalized least significant change for individuals measured on different DXA systems. J. Clin. Densitom. 2007;10(3):249–258. doi: 10.1016/j.jocd.2007.05.002. [DOI] [PubMed] [Google Scholar]
- 44.Stark E., Haffner O., Kučera E. Low-cost method for 3D body measurement based on photogrammetry using smartphone. Electronics. 2022;11(7):1048. doi: 10.3390/electronics11071048. [DOI] [Google Scholar]
- 45.Majmudar M.D., Chandra S., Yakkala K., Kennedy S., Agrawal A., Sippel M., et al. Smartphone camera based assessment of adiposity: a validation study. NPJ Digit. Med. 2022;5(1):79. doi: 10.1038/s41746-022-00628-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data described in the manuscript, code book, and analytic code will be made available upon request pending researchers who meet the criteria for access to confidential data. To request an application or to use our statistical shape model to transform your Meshcapade processed mesh into usable principal components, please contact John Shepherd (johnshep@hawaii.edu) or visit www.shapeup.shepherdresearchlab.org.