Abstract
Background
Body composition can be accurately quantified from abdominal computed tomography (CT) exams and is a predictor for the development of aging-related conditions and for mortality. However, reference ranges for CT-derived body composition measures of obesity, sarcopenia, and bone loss have yet to be defined in the general population.
Methods
We identified a population-representative sample of 4 900 persons aged 20 to 89 years who underwent an abdominal CT exam from 2010 to 2020. The sample was constructed using propensity score matching an age and sex stratified sample of persons residing in the 27-county region of Southern Minnesota and Western Wisconsin. The matching included race, ethnicity, education level, region of residence, and the presence of 20 chronic conditions. We used a validated deep learning based algorithm to calculate subcutaneous adipose tissue area, visceral adipose tissue area, skeletal muscle area, skeletal muscle density, vertebral bone area, and vertebral bone density from a CT abdominal section.
Results
We report CT-based body composition reference ranges on 4 649 persons representative of our geographic region. Older age was associated with a decrease in skeletal muscle area and density, and an increase in visceral adiposity. All chronic conditions were associated with a statistically significant difference in at least one body composition biomarker. The presence of a chronic condition was generally associated with greater subcutaneous and visceral adiposity, and lower muscle density and vertebrae bone density.
Conclusions
We report reference ranges for CT-based body composition biomarkers in a population-representative cohort of 4 649 persons by age, sex, body mass index, and chronic conditions.
Keywords: Age differences, Body composition, Computed tomography, Reference ranges, Sex differences
Body composition, specifically the quantity and distribution of adipose, muscle, and bone tissue within the body, is a fundamental and dynamic aspect of human health (1,2). Body composition plays a critical role as a risk factor for heart disease (3–5), certain cancers (6,7), metabolic conditions including diabetes and pancreatic dysfunction (8), and for outcomes following major surgery (9).
In clinical practice, body composition is typically measured only through calculation of the body mass index (BMI). However, BMI cannot discriminate between fat mass and lean body mass, and only provides indirect measures of other important components of body composition (eg, bone and muscle density) (10). For example, frailty in the form of decreased skeletal muscle mass may be difficult to perceive in older adults when masked by obesity (11,12).
Imaging-based measures of body composition, such as those from abdominal computed tomography (CT), provide submillimeter resolution of muscle and bone compartments and provide a more comprehensive assessment of body composition (13,14). Abdominal CT is also widely used in general medical practice. One study reported that nearly 17% of Medicare beneficiaries received a CT exam during a single year (15). Furthermore, many abdominal CTs are interpreted by radiologists as normal (one study reported that 34% of CTs obtained during emergency department visits were “normal”) (16).
Recent breakthroughs in automated CT image segmentation using deep learning allow for rapid calculation of body composition measures in large populations (17–21). This “opportunistic screening” offers the ability to gain further information about patient health beyond the indications for which the CT imaging was clinically ordered (22). Using automated techniques, many groups have reported reference ranges for body composition biomarkers drawn from various patient populations, particularly patients who have received CT exams to diagnose cancer (6,20,23–32).
Reference range values for CT-derived measures of body composition have yet to be conclusively established in the literature. On the one hand, most previous studies have been restricted to persons who had undergone CT imaging for a specific medical indication, and measurements of body composition may be systematically affected (biased) by such a selection. On the other hand, studies such as that by van Vugt et al. have reported reference ranges in a population of persons who were specifically chosen to be “healthy” (ie, renal donors). However, persons eligible for renal donation are not representative of the general population, particularly at older ages. Such a “healthy” cohort excludes persons diagnosed with conditions like hypertension, which are endemic in the general adult population (up to 30% prevalence) (28). Body composition is particularly difficult to define in older adults, where it is a dynamic process and where many “successful agers” have common chronic diseases (over 80% of persons older than age 65 have one or more chronic conditions) (33). For example, in the Baumgartner et al. landmark study, a cutoff for sarcopenia was defined in a referent group of men and women with an average age of 29 years. However, when this cutoff is applied to persons older than age 80 years, more than 50% of persons meet the definition for sarcopenia (34). Similarly, cutoffs for sarcopenia proposed by Prado et al. and Martin et al. (both now widely accepted) were originally defined in a highly selective cohort of people with obesity undergoing treatment for gastrointestinal or respiratory cancer (35,36). Such reference ranges have limited utility in accurately capturing the complex processes of aging.
To supplement the existing body of literature, we report reference range values in a cohort of persons specifically selected to be representative of a well-defined population residing in the Upper Midwest. We applied a previously validated deep-learning approach to comprehensively assess adiposity (27), sarcopenia (37), and osteopenia (38) from abdominal CT imaging.
Method
Identification of a Cohort Representing the REP Population
We used the Rochester Epidemiology Project (REP) medical records-linkage system to identify a representative population of persons on which to calculate body composition (39). The REP is a medical records-linkage system established in 1966 that captures health information for a large proportion of the population residing in a 27-county region of Southern Minnesota and Western Wisconsin (39,40). All data were acquired historically as part of routine clinical practice. For this study, informed consent was waived, and the study was approved by both the Mayo Clinic and Olmsted Medical Center Institutional Review Boards.
The REP was used to identify all persons residing in the 27-county region on January 1, 2015. We randomly sampled 35 men and 35 women from each single year of age from 20 to 89 years (n = 4 900 persons). We constructed this population sampling frame stratified by age and sex to optimize our ability to calculate reference ranges across the age-span from young to old (a simple random sample would have resulted in few persons in the oldest age categories). Race, ethnicity, education level, and region of residence were extracted using the REP electronic indexes. Additionally, presence of 20 chronic conditions was identified using International Classification of Diseases (ICD-9 and ICD-10) codes defined by the U.S. Department of Health and Human Services for studies of multimorbidity (Supplementary Table 1) (41). Diagnostic codes were extracted for the 5 years prior to January 1, 2015, and chronic conditions were defined based on having 2 or more codes separated by >30 days (42). Characteristics of this “Representative REP sample” are shown in Supplementary Table 2.
Creating a Matched CT Cohort
We used the REP to identify all persons who had a current procedural terminology (CPT) code that indicated an abdominal CT exam and who were residents of the 27-county region between January 1, 2010 and December 31, 2020 (11-year period) (40). Supplementary Figure 1 shows the process for selecting the final cohort included in our analyses. We excluded CT exams in persons younger than 20 years and older than 89 years. We also excluded exams that occurred within 30 days before a person’s death because the CT may have been performed for reasons directly related to the cause of death (final inclusion of n = 423 081 CT exams; Supplementary Figure 1, Phase 1).
To create a CT cohort that represented the 27-county region, we selected a corresponding person from the CT cohort for each of the 4 900 persons from the representative REP sample using propensity score matching. We fit a logistic regression propensity score model including race (White, Black, Asian, Other/mixed, or unknown), ethnicity (Hispanic or non-Hispanic), education (less than or equal to high school, more than high school, or unknown), the county of residency, and each of the 20 chronic conditions. The 20 chronic conditions in the CT cohort were defined based on the diagnosis codes received in the 5-year period before each anchoring CT date. For example, if the CT was performed on September 1, 2011, the diagnoses used to define the 20 chronic conditions were those from September 1, 2006, through August 31, 2011. Propensity matching was completed separately within each single year of age and sex stratum, so age and sex were matched by design. When persons from the CT group were successfully matched a first time, they were no longer eligible for inclusion if they received another CT exam at a later time (423 081 CT exams belonged to 181 187 unique persons).
Characteristics of the propensity-matched CT cohort are shown in Supplementary Table 2. To compare the balance of covariates in the CT cohort before and after (ie, representativeness), we used absolute standardized mean differences (ASMD) (43). For example, in the original CT cohort, 47% of persons had hypertension compared to only 30% of the representative REP sample (ASMD = 0.35; poorly balanced). After propensity matching, 30% of the matched CT cohort had hypertension (ASMD = 0.01, highly balanced). Improvements in representativeness across most other covariates were seen in our propensity-matched abdominal CT cohort. The “matched abdominal CT cohort” of 4 900 persons was used for all further analyses.
CT Quality Control Steps
In the matched CT cohort, 161 persons (3.2%) were excluded because they had a CPT code but did not have a CT exam available for analysis (Supplementary Figure 1; Phase 2). We also excluded persons with CT exams that lacked a 20-cm region of abdomen containing the L3 vertebra or that that lacked an axial series (n = 90; 1.8%). When a CT abdomen exam contained multiple eligible CT series, we selected the optimal series for body composition analysis based on slice thickness, reconstruction kernel, series description, and other criteria recorded in the DICOM image header. Additional information is included in Supplementary Methods, and Supplementary Table 3 shows the technical characteristics of included CT exams.
CT Body Composition Biomarkers
We used a validated deep-learning-based algorithm with a 3-dimensional U-Net Convolutional Neural Network (CNN) model, which has previously been described (18). The model segments subcutaneous adipose tissue, visceral adipose tissue, skeletal muscle, and bone from a 20-cm section of abdomen centered at the midpoint of the L3 vertebra; the volumetric region is averaged to derive a 2D area measurement (18). The algorithm uses a second CNN model to select the midpoint of the L3 vertebra (44).
We applied this deep-learning algorithm to CT exams from the matched CT cohort to compute subcutaneous adipose tissue area, visceral adipose tissue area, skeletal muscle area, and L3 vertebral bone area (additional details provided in Supplementary Methods) (25,45). All tissue area measurements in this article are reported as values indexed to height in meters squared (m2). Indexing skeletal muscle area and vertebral bone area to height may “equalize” measurements across persons with differences in stature. For example, this standardization may avoid misclassifying persons with a low body composition biomarker value because they are short. However, the impact of indexing adipose tissue area to height has yet to be completely understood. Therefore, we reported nonindexed values in Supplementary Table 4.
Additionally, we defined skeletal muscle density and vertebral bone density as the mean Hounsfield unit intensity of all voxels segmented into the respective compartments. Hounsfield units are a measure of radiologic x-ray attenuation, which is proportional to the density of an object, that is, higher attenuation equates to higher density. We made a semantic choice to refer to these constructs as “density” instead of “attenuation” because it is a more intuitive concept for clinical readers.
Deep-learning model inference was performed using the Tensorflow package version 2.6.0 (46) in Python 3.7.12. GPU acceleration was used for model inference with 4 NVIDIA T4 GPUs (CUDA version 11.2; NVIDIA Corp., Santa Clara, CA, USA). Portions of the CT quality control workflow and analysis of body composition biomarkers were performed using RStudio, R package version 4.0.3 (47). Model inference was performed using the VertexAI feature of Google Cloud Platform (Google Cloud Platform, Mountain View, CA, USA).
Statistical Analysis
We used quantile regression to model the 5th, 25th, 50th (median), 75th, and 95th percentiles of the 6 body composition biomarkers across age and separately for men and women (48). We plotted the body composition biomarkers and the quantile regression fits across age for men and women separately. In addition, we plotted each body composition biomarker separately by BMI category (underweight or normalweight: <25 kg/m2; overweight: 25 to <30 kg/m2; and obese: ≥30 kg/m2) and separately by the presence of 20 chronic conditions.
To measure the influence of preexisting chronic conditions on body composition biomarkers, we summarized differences in body composition biomarkers among persons with and without each of the 20 chronic conditions by fitting a regression model with both age (linear) and age-squared (quadratic) terms and comparing least squared means at fixed age points of 25, 55, and 85 years. Conditions affecting fewer than 100 persons were excluded from these comparisons. Statistical analyses were performed using SAS version 9.4 (Cary, NC, USA) and R version 4.1.2 at the standard 2-tailed alpha level of .05.
Results
Table 1 shows the demographic, clinical, and biomarker characteristics of the final cohort of 4 649 persons at the time of abdominal CT. Overall, 50% of the population were women, and 36% were 65 years of age or older. Most persons were White (94%) and non-Hispanic (97%), and the median BMI was 28.7 kg/m2. Hyperlipidemia (31%), hypertension (30%), and diabetes (16%) were the most common chronic conditions in the study population.
Table 1.
Characteristics of Persons Included in Analyses
| Summary, N (%) or median (Q1, Q3) | |||
|---|---|---|---|
| Characteristic | Men | Women | Men and Women |
| Total persons | 2 319 | 2 330 | 4 649 |
| Age at CT, y | |||
| 20 to 44 | 825 (35.6) | 832 (35.7) | 1 657 (35.6) |
| 45 to 64 | 663 (28.6) | 663 (28.5) | 1 326 (28.5) |
| 65 to 89 | 831 (35.8) | 835 (35.8) | 1 666 (35.8) |
| Race | |||
| White | 2 167 (93.4) | 2 195 (94.2) | 4 362 (93.8) |
| Black | 44 (1.9) | 32 (1.4) | 76 (1.6) |
| Asian | 22 (0.9) | 30 (1.3) | 52 (1.1) |
| Other/mixed | 49 (2.1) | 51 (2.2) | 100 (2.2) |
| Unknown | 37 (1.6) | 22 (0.9) | 59 (1.3) |
| Hispanic ethnicity | 79 (3.4) | 79 (3.4) | 158 (3.4) |
| Geographic location* | |||
| Olmsted county | 433 (18.7) | 404 (17.3) | 837 (18.0) |
| 8 Near counties | 791 (34.1) | 784 (33.6) | 1 575 (33.9) |
| 18 Far counties | 1 095 (47.2) | 1 142 (49.0) | 2 237 (48.1) |
| Height (cm) | 177.8 (172.7, 181.9) | 163.0 (158.0, 167.4) | 170.0 (162.5, 178.0) |
| Weight (kg) | 90.5 (79.4, 104.0) | 75.2 (63.8, 89.8) | 83.5 (70.3, 97.7) |
| BMI (kg/m2) | 28.8 (25.6, 32.9) | 28.5 (24.2, 33.7) | 28.7 (24.9, 33.2) |
| Normal or underweight (<25 kg/m2) | 485 (20.9) | 700 (30.0) | 1 185 (25.5) |
| Overweight (25 to <30 kg/m2) | 892 (38.5) | 683 (29.3) | 1 575 (33.9) |
| Obese (≥30 kg/m2) | 942 (40.6) | 947 (40.6) | 1 889 (40.6) |
| DHHS chronic conditions† | |||
| Hypertension | 709 (30.6) | 687 (29.5) | 1 396 (30.0) |
| Congestive heart failure | 96 (4.1) | 71 (3.0) | 167 (3.6) |
| Coronary artery disease | 287 (12.4) | 123 (5.3) | 410 (8.8) |
| Cardiac arrhythmias | 323 (13.9) | 239 (10.3) | 562 (12.1) |
| Hyperlipidemia | 755 (32.6) | 686 (29.4) | 1 441 (31.0) |
| Stroke | 85 (3.7) | 58 (2.5) | 143 (3.1) |
| Arthritis | 298 (12.9) | 347 (14.9) | 645 (13.9) |
| Asthma | 67 (2.9) | 124 (5.3) | 191 (4.1) |
| Cancer | 160 (6.9) | 130 (5.6) | 290 (6.2) |
| Chronic kidney disease | 146 (6.3) | 101 (4.3) | 247 (5.3) |
| COPD | 124 (5.3) | 115 (4.9) | 239 (5.1) |
| Dementia or delirium | 63 (2.7) | 58 (2.5) | 121 (2.6) |
| Depressive disorders | 217 (9.4) | 382 (16.4) | 599 (12.9) |
| Diabetes | 412 (17.8) | 339 (14.5) | 751 (16.2) |
| Hepatitis | 13 (0.6) | 10 (0.4) | 23 (0.5) |
| Osteoporosis | 18 (0.8) | 126 (5.4) | 144 (3.1) |
| Schizophrenia and psychosis | 27 (1.2) | 17 (0.7) | 44 (0.9) |
| Substance abuse disorders | 93 (4.0) | 48 (2.1) | 141 (3.0) |
| Number of 20 DHHS chronic conditions | 1 (0, 3) | 1 (0, 3) | 1 (0, 3) |
| Abdominal CT biomarkers | |||
| Subcutaneous adipose tissue area (cm2) | 208.0 (147.1, 294.9) | 273.4 (178.4, 386.9) | 235.3 (158.9, 348.6) |
| Subcutaneous adipose tissue area indexed‡ (cm2/m2) | 66.3 (46.9, 93.4) | 104.0 (67.8, 145.3) | 81.4 (54.4, 124.1) |
| Visceral adipose tissue area (cm2) | 212.0 (108.1, 311.3) | 102.3 (42.6, 176.1) | 148.6 (64.3, 247.7) |
| Visceral adipose tissue area indexed‡ (cm2/m2) | 67.6 (34.4, 99.2) | 39.1 (16.0, 66.4) | 52.7 (23.0, 83.4) |
| Skeletal muscle area (cm2) | 188.9 (165.6, 212.5) | 131.7 (116.5, 147.4) | 156.5 (129.5, 191.3) |
| Skeletal muscle area indexed‡ (cm2/m2) | 60.0 (53.4, 67.4) | 49.8 (44.2, 55.6) | 54.3 (47.7, 62.4) |
| Skeletal muscle density (HU) | 30.0 (19.0, 40.8) | 24.3 (14.0, 34.6) | 27.0 (16.4, 37.7) |
| Vertebral bone area (cm2) | 29.0 (25.6, 32.5) | 23.5 (20.7, 26.3) | 26.0 (22.3, 29.9) |
| Vertebral bone area indexed‡ (cm2/m2) | 9.2 (8.2, 10.3) | 8.9 (7.8, 10.0) | 9.0 (8.0, 10.1) |
| Vertebral bone density (HU) | 298.1 (260.8, 338.9) | 308.4 (260.7, 354.5) | 302.4 (260.8, 346.3) |
Notes: BMI = body mass index; COPD = chronic obstructive pulmonary disease; CT = computed tomography; DHHS = Department of Health and Human Services; HU = Hounsfield units; Q1 = first quartile (25th percentile); Q3 = 3rd quartile (75th percentile).
*The 8 near counties include Dodge, Fillmore, Freeborn, Goodhue, Mower, Steele, Wabasha, and Waseca counties in Minnesota. The 18 far counties include Blue Earth, Brown, Faribault, Houston, Le Sueur, Martin, Nicollet, Rice, Watonwan, and Winona Counties in Minnesota and Barron, Buffalo, Chippewa, Dunn, Eau Claire, La Crosse, Pepin, and Trempealeau Counties in Wisconsin.
†Autism and HIV are not shown (no persons had autism diagnostic codes and only 1 person had HIV diagnostic codes).
‡Indexed by dividing by height in meters squared.
Body Composition Biomarkers by Age and Sex
Table 2 provides reference values for each body composition biomarker by age and sex (nonindexed values are provided in Supplementary Table 4). Figure 1 shows sex and age trends for each CT biomarker in the full cohort. All biomarkers differed between men and women and by age (p < .001 for both). Women had higher subcutaneous adipose tissue area and higher vertebral bone area measures than men. Men had higher vertebral bone density, skeletal muscle area, and skeletal muscle density than women. Vertebral bone density changed with age in both men and women. Women had higher vertebral bone density measures than men until age 60 years.
Table 2.
The Reference Percentiles for the Six Tissue Biomarkers Separately by Sex and 5-Year Age Groups
| Sex/Age Group | Tissue Biomarker Reference Percentiles (5th, 25th, 50th, 75th, 95th)* | |||||
|---|---|---|---|---|---|---|
| Subcutaneous Adipose Tissue Area Index (cm2/m2) | Visceral Adipose Tissue Area Index (cm2/m2) | Skeletal Muscle Area Index (cm2/m2) | Skeletal Muscle Density (HU) | Vertebral Bone Area Index (cm2/m2) | Vertebral Bone Density (HU) | |
| Women | ||||||
| 20–24 y | (24.1, 55.4, 94.2, 138.7, 240.8) | (0.1, 3.9, 12.5, 27.9, 57.9) | (40.0, 45.5, 50.3, 55.6, 64.9) | (21.6, 32.4, 39.6, 45.7, 53.2) | (3.6, 5.6, 8.2, 10.8, 12.9) | (300.7, 334.0, 361.4, 389.9, 448.6) |
| 25–29 y | (26.6, 56.9, 103.6, 154.6, 240.9) | (1.1, 7.1, 19.3, 38.7, 73.1) | (39.2, 46.1, 51.0, 56.8, 66.1) | (19.0, 29.2, 37.0, 43.2, 51.9) | (3.5, 5.6, 8.2, 11.0, 13.2) | (291.4, 330.7, 362.2, 392.6, 454.0) |
| 30–34 y | (30.1, 60.9, 110.6, 163.0, 232.4) | (2.3, 10.2, 25.9, 48.6, 85.3) | (39.3, 46.7, 52.0, 57.6, 66.9) | (17.0, 27.2, 34.6, 40.7, 50.1) | (3.6, 5.7, 8.4, 11.1, 13.3) | (280.0, 324.9, 358.6, 389.3, 451.5) |
| 35–39 y | (32.9, 65.6, 114.8, 164.3, 227.1) | (3.2, 13.0, 31.6, 56.8, 94.8) | (39.6, 46.9, 52.5, 57.9, 68.0) | (14.9, 25.6, 32.5, 38.8, 48.6) | (3.7, 5.8, 8.5, 11.3, 13.5) | (266.5, 315.9, 351.3, 382.0, 444.9) |
| 40–44 y | (34.8, 69.9, 116.4, 160.3, 224.7) | (3.9, 15.6, 37.0, 63.5, 103.2) | (39.6, 46.8, 52.3, 57.8, 69.1) | (12.6, 23.7, 30.4, 37.2, 47.1) | (3.9, 6.0, 8.7, 11.4, 13.7) | (251.7, 302.6, 339.9, 370.7, 434.0) |
| 45–49 y | (36.0, 73.6, 115.4, 153.9, 225.2) | (4.4, 18.0, 41.8, 68.6, 110.1) | (39.2, 46.5, 51.5, 57.3, 69.7) | (10.0, 21.4, 28.4, 35.8, 45.7) | (3.9, 6.0, 8.6, 11.4, 13.6) | (237.9, 286.0, 324.3, 355.6, 419.0) |
| 50–54 y | (36.3, 76.3, 113.6, 149.1, 228.0) | (4.8, 20.0, 45.6, 72.1, 115.5) | (38.5, 45.9, 50.6, 56.6, 69.7) | (7.4, 18.9, 26.3, 34.2, 44.1) | (3.9, 5.9, 8.6, 11.4, 13.6) | (226.4, 269.6, 305.8, 337.8, 400.9) |
| 55–59 y | (36.1, 78.3, 112.5, 147.7, 230.8) | (5.1, 21.8, 48.6, 74.8, 119.7) | (37.5, 45.1, 50.2, 56.0, 69.2) | (4.7, 16.1, 24.0, 32.0, 42.1) | (4.0, 6.1, 8.8, 11.5, 13.8) | (216.6, 255.6, 287.3, 320.0, 382.8) |
| 60–64 y | (35.7, 79.2, 111.8, 149.5, 230.0) | (5.4, 23.3, 50.8, 76.5, 122.6) | (36.2, 44.3, 49.5, 55.4, 68.2) | (2.0, 13.3, 21.4, 29.1, 39.6) | (4.2, 6.3, 9.0, 11.7, 14.0) | (208.3, 244.8, 274.6, 307.9, 371.2) |
| 65–69 y | (35.4, 78.5, 110.1, 150.0, 224.2) | (5.7, 24.4, 52.0, 77.4, 124.2) | (34.9, 43.4, 48.5, 54.6, 66.8) | (−0.5, 10.6, 18.7, 25.9, 36.7) | (4.4, 6.5, 9.2, 12.0, 14.2) | (201.5, 237.5, 269.8, 303.6, 368.4) |
| 70–74 y | (34.8, 75.2, 105.4, 144.2, 213.2) | (5.9, 25.2, 52.3, 77.3, 124.4) | (34.1, 42.4, 47.2, 53.5, 64.9) | (−3.0, 7.8, 15.9, 22.8, 33.3) | (4.6, 6.7, 9.4, 12.2, 14.4) | (196.0, 233.2, 270.6, 304.7, 372.1) |
| 75–79 y | (33.3, 69.5, 97.8, 132.3, 197.8) | (6.0, 25.6, 51.6, 76.3, 123.3) | (33.9, 41.8, 46.4, 52.1, 62.7) | (−5.3, 5.1, 13.1, 19.9, 29.9) | (4.8, 7.0, 9.7, 12.4, 14.7) | (191.7, 231.7, 270.3, 304.8, 375.1) |
| 80–84 y | (30.6, 62.5, 89.7, 119.8, 178.8) | (6.1, 25.7, 50.0, 74.4, 120.8) | (34.1, 41.8, 46.3, 51.3, 61.8) | (−7.5, 2.4, 10.2, 17.2, 26.9) | (5.1, 7.2, 9.9, 12.7, 15.0) | (187.1, 230.0, 265.1, 300.3, 373.3) |
| 85–89 y | (26.8, 55.5, 84.0, 113.7, 158.6) | (6.1, 25.4, 47.4, 71.6, 117.0) | (34.7, 42.4, 47.1, 52.7, 64.3) | (−9.6, −0.3, 7.3, 14.5, 24.3) | (5.1, 7.2, 9.9, 12.7, 15.0) | (176.9, 223.4, 254.9, 290.8, 366.2) |
| Men | ||||||
| 20–24 y | (9.0, 22.6, 41.8, 71.6, 160.6) | (0.0, 5.0, 16.4, 33.7, 72.9) | (43.1, 52.1, 58.2, 64.6, 75.8) | (31.3, 40.3, 46.3, 50.6, 58.5) | (3.7, 5.7, 8.3, 10.9, 13.0) | (287.3, 321.3, 347.3, 385.8, 453.5) |
| 25–29 y | (14.5, 34.4, 57.8, 92.2, 179.2) | (2.9, 14.2, 30.7, 53.3, 96.1) | (43.9, 54.0, 60.6, 67.5, 78.8) | (27.1, 36.8, 43.7, 48.9, 56.7) | (3.8, 5.9, 8.6, 11.3, 13.6) | (270.4, 308.9, 340.9, 375.7, 438.8) |
| 30–34 y | (17.7, 42.9, 66.0, 99.8, 183.5) | (6.0, 22.8, 43.8, 69.7, 109.9) | (44.9, 55.7, 62.4, 69.6, 80.5) | (23.7, 34.1, 41.2, 47.0, 54.6) | (3.9, 6.0, 8.6, 11.4, 13.6) | (256.6, 297.4, 333.0, 364.0, 421.7) |
| 35–39 y | (19.9, 48.2, 69.6, 101.2, 184.0) | (7.7, 30.5, 54.8, 81.7, 119.0) | (45.8, 56.8, 63.6, 70.9, 81.7) | (20.5, 31.7, 38.9, 45.1, 52.6) | (3.8, 5.9, 8.6, 11.4, 13.6) | (245.4, 287.1, 324.1, 353.7, 408.8) |
| 40–44 y | (22.0, 51.0, 71.5, 101.9, 181.1) | (8.4, 37.5, 63.8, 90.5, 127.3) | (46.4, 57.4, 64.3, 71.5, 82.4) | (17.4, 29.2, 36.6, 43.0, 50.7) | (3.9, 6.0, 8.7, 11.4, 13.7) | (235.9, 277.4, 313.7, 344.5, 399.2) |
| 45–49 y | (24.1, 51.8, 72.0, 102.1, 175.4) | (9.0, 43.6, 71.0, 97.2, 134.9) | (46.8, 57.5, 64.3, 71.5, 82.5) | (14.3, 26.8, 34.3, 40.8, 48.9) | (4.1, 6.2, 8.9, 11.6, 13.9) | (228.4, 268.7, 302.7, 336.5, 393.2) |
| 50–54 y | (26.1, 51.8, 71.8, 102.0, 168.7) | (9.6, 48.6, 76.7, 102.8, 141.8) | (46.9, 57.0, 63.7, 70.8, 82.2) | (11.4, 24.4, 31.8, 38.5, 47.1) | (4.3, 6.4, 9.1, 11.8, 14.1) | (222.5, 260.8, 292.8, 329.7, 390.5) |
| 55–59 y | (28.1, 51.9, 71.6, 101.3, 162.5) | (10.1, 52.5, 81.3, 107.6, 147.8) | (46.8, 56.1, 62.6, 69.6, 81.4) | (8.4, 21.9, 29.2, 36.1, 45.3) | (4.5, 6.6, 9.3, 12.1, 14.3) | (217.8, 253.9, 285.4, 324.0, 389.5) |
| 60–64 y | (30.0, 52.4, 71.3, 99.8, 156.6) | (10.8, 55.4, 84.7, 111.7, 152.9) | (46.2, 55.0, 61.2, 68.2, 80.1) | (5.6, 19.3, 26.6, 33.5, 43.1) | (4.7, 6.8, 9.6, 12.3, 14.6) | (213.1, 248.4, 280.8, 319.3, 387.3) |
| 65–69 y | (31.3, 52.6, 71.1, 97.3, 150.2) | (11.4, 57.2, 87.0, 115.0, 157.0) | (45.1, 53.7, 59.7, 66.5, 78.2) | (2.8, 16.6, 23.8, 30.8, 40.6) | (5.0, 7.1, 9.8, 12.6, 14.8) | (207.6, 244.6, 278.9, 315.5, 382.3) |
| 70–74 y | (31.7, 52.0, 70.6, 93.8, 142.3) | (12.0, 57.9, 87.8, 117.3, 160.1) | (43.4, 52.2, 58.1, 64.7, 76.0) | (0.1, 13.8, 21.0, 28.0, 37.8) | (5.2, 7.3, 10.0, 12.8, 15.1) | (201.3, 241.5, 278.4, 311.9, 374.7) |
| 75–79 y | (31.0, 50.5, 68.7, 88.9, 132.8) | (12.7, 57.6, 87.2, 118.7, 162.3) | (41.3, 50.5, 56.4, 62.7, 73.3) | (−2.6, 10.6, 18.1, 25.0, 34.6) | (5.4, 7.5, 10.2, 13.0, 15.3) | (192.8, 236.5, 275.0, 306.8, 364.9) |
| 80–84 y | (29.2, 48.0, 65.0, 82.6, 122.0) | (13.4, 56.2, 84.9, 119.3, 163.7) | (38.8, 48.6, 54.7, 60.7, 70.5) | (−5.2, 7.2, 15.1, 22.0, 31.2) | (5.5, 7.7, 10.4, 13.2, 15.5) | (182.0, 228.3, 266.5, 299.6, 357.1) |
| 85–89 y | (26.3, 44.6, 59.3, 74.8, 109.7) | (14.1, 53.7, 80.9, 118.9, 164.3) | (35.6, 46.5, 52.9, 58.5, 67.8) | (−7.8, 3.3, 12.0, 18.8, 27.4) | (5.6, 7.8, 10.5, 13.3, 15.6) | (168.3, 216.5, 252.5, 289.7, 356.9) |
Notes: Bolded values indicate 50th percentiles. HU = Hounsfield units.
*Tissue biomarkers representing areas are indexed to height in meters squared.
Figure 1.
Scatter plot for 6 body composition biomarkers, plotted across age and separately for women and men. The lines are the quantile regression spline fits for the 5th, 25th, 50th (median), 75th, and 95th percentiles.
Older age was associated with a decrease in both skeletal muscle area and skeletal muscle density. By contrast, older age was associated with a marked increase in visceral adipose tissue area. Subcutaneous adipose tissue area increased through age 45 years in both men and women, and then decreased thereafter. Finally, vertebral bone area increased, but vertebral bone density decreased with older age.
Body Composition by BMI Strata
As expected, both subcutaneous and visceral adipose tissue areas were higher in persons with higher BMI. A higher BMI was associated with higher skeletal muscle area but lower skeletal muscle density. BMI category was not associated with vertebral bone area, but higher BMI was associated with lower vertebral bone density. Supplementary Figure 2 shows sex and age trends for each CT biomarker within BMI categories (normal weight, overweight, and obese).
Body Composition by Chronic Condition Strata
Supplementary Figure 3 shows plots for sex and age trends for each of 6 CT biomarkers (6 blocks on each page) within each of 18 chronic condition strata (a separate page for each of the 18 chronic conditions), and Figure 2 summarizes how CT biomarkers were significantly affected by the presence of chronic conditions across age (only conditions that affected 100 persons or more are shown). All chronic conditions were associated with a statistically significant difference in at least one body composition biomarker. On average, the presence of a chronic condition was associated with an increase in subcutaneous adipose tissue area, visceral adipose tissue area, and skeletal muscle area, but a decrease in skeletal muscle density. Associations differed in men and women and by age. For example, hypertension, coronary artery disease, cardiac arrythmias, hyperlipidemia, arthritis, asthma, and osteoporosis were associated with a decrease in vertebral bone density in women aged 55 years but not in men aged 55 years.
Figure 2.
The presence of chronic conditions was significantly associated with differences in CT-based body composition biomarkers. The effect of chronic conditions was explored separately for women (top) and men (bottom) and at 3 ages representing young age (25), middle age (55), and old age (85). Biomarkers that are statistically significantly increased are indicated by orange shading whereas biomarkers that are significantly decreased are indicated by purple shading. Biomarkers that did not differ in persons with and without a chronic condition have no color shading. The percent displayed in each cell represents the change in the body composition biomarker in persons with the condition as compared to those without the condition. Only conditions affecting 100 persons or more are shown.
Discussion
We have summarized CT-based body composition reference ranges for a population of men and women from ages 20 to 89 years who were specifically selected to be representative of a well-defined population living in the Upper Midwest region of the United States. We found that body composition biomarkers differed between men and women, across age, by BMI strata, and by the presence of 20 chronic conditions. This comprehensive evaluation of abdominal body composition in a population-based cohort deepens our understanding of how body composition differs by sex, changes across age, and how body composition measures are associated with common chronic conditions.
Other investigators have previously reported CT-based body composition measures across subgroups of the population. Our results are similar to several previous studies (Table 3). However, it should be noted that cutoffs for sarcopenia originally proposed by Prado et al. and Martin et al. (which are now widely accepted) were originally defined in a highly selective cohort of people with obesity undergoing treatment for gastrointestinal or respiratory cancer (35,36). More recently, van der Werf et al. and Magudia et al. have defined reference ranges for skeletal muscle area, visceral adipose tissue area, and subcutaneous adipose tissue area in healthy living kidney donors and adult outpatients (20,45). We have expanded on previous studies by creating a cohort of persons with CT scans that specifically represented a well-defined underlying population. Therefore, our results represented CT measures for the “average” person living in an Upper Midwest region. In addition, we have described population-based measures of vertebral bone area and vertebral bone density from the same CT cohort images.
Table 3.
Comparison of Reference Range Values to Previous Studies
| Characteristic | Our Study | Prado (2008) | Martin (2013) | Magudia (2021) | Navin (2021) | Pickhardt (2022) | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Sample size, N | 4 649 | 250 | 1 473 | 12 128 | 692 | 9 223 | ||||||
| Study population | CT cohort matched to be representative of the general population | Patients with obesity, incident respiratory tract, colorectal, or gastrointestinal cancers | Patients with obesity, with incident respiratory tract, colorectal, or gastrointestinal cancers | Adult outpatients at Brigham and Women’s Hospital and Dana Farber | Healthy individuals recruited for bone density study | Patients undergoing colorectal cancer screening at the University of Wisconsin | ||||||
| Women, % | 50% | 46% | 44% | 57% | 54% | 56% | ||||||
| Age in years, mean | 55 | 64 | 65 | 52 | 55 | 57 | ||||||
| Non-White race, % | 7% | Not reported | Not reported | 17% | 2% | Not reported | ||||||
| Obese, % | Median BMI = 28.7 Obese = 40.6% |
100% | 17% | Mean BMI = 29 | 30% | Mean BMI = 28.9 | ||||||
| Single CT protocol? | No | No | No | No | Yes | Yes | ||||||
| Men | Women | Men | Women | Men | Women | Men | Women | Men | Women | Men | Women | |
| Subcutaneous adipose tissue area index (cm2/m2) | 66.3 | 104.0 | — | — | — | — | 66.5 | 90.7 | 71.0 | — | — | |
| Visceral adipose tissue area index (cm2/m2) | 67.6 | 39.1 | — | — | — | — | 66.5 | 28.5 | 58.7 | 38.8 | — | — |
| Skeletal muscle area index (cm2/m2) | 60.0 | 49.8 | 59.1 | 48.8 | 51.5 | 41.3 | 58.7 | 44.9 | 51.7 | 39.7 | — | — |
| Skeletal muscle density (HU) | 30.0 | 24.3 | 30.6 | 35.5 | 34.5 | — | — | — | — | 28.9 | ||
| Vertebral bone area index (cm2/m2) | 9.2 | 8.9 | — | — | — | — | — | — | — | — | — | — |
| Vertebral bone density (HU) | 298.1 | 308.4 | — | — | — | — | — | — | — | — | 171.2 | |
Note: BMI = body mass index; CT = computed tomography; HU = Hounsfield units.
Trends in Body Composition in the General Population
Computed tomography body composition reference values matched well-established patterns of body habitus with increased aging (1,12,31): women had higher subcutaneous adipose tissue area, whereas men had higher visceral adipose tissue area, higher skeletal muscle area, and higher skeletal muscle density. Our data show that vertebral bone area increased with older age, but that vertebral bone density decreased with older age. We also found that women had a slightly higher vertebral bone density (indicating greater bone mass) than men until age 50–60 years after which vertebral bone density decreased. In older ages, degenerative changes in the spine including compression fractures increase bone area.
As expected, persons with higher BMI had higher subcutaneous and visceral adipose tissue area. However, regardless of BMI category, subcutaneous adipose tissue area remained relatively constant throughout adult life, whereas visceral adipose tissue area increased consistently with older age. The trend in visceral adiposity was observed in all BMI groups and may reflect a shift toward central obesity in middle age even in normal-weight persons. Persons with higher BMI also had higher skeletal muscle area, but lower skeletal muscle density; this trend suggests that increased skeletal muscle area is driven by a greater degree of fatty infiltrate in muscle tissue, rather than true muscle mass. The relationship between BMI and skeletal muscle area is also well accepted. Martin et al. reported a higher cutoff value to define sarcopenia for skeletal muscle area in men with a BMI above 25 kg/m2 and Derstine et al. recommend adjusting skeletal muscle area for BMI, as well as height (49). We also found a strong association of decreased skeletal muscle density with older age in both sexes.
Trends in Body Composition in Persons With Chronic Conditions
All chronic conditions were associated with a change in at least one body composition biomarker, and most conditions were associated with a difference in 4 or more biomarkers. In general, the presence of a chronic condition was associated with increased adiposity (higher subcutaneous and visceral adipose tissue area) and lower skeletal muscle density and vertebral bone density. Persons aged 25 and 55 with a chronic condition had higher subcutaneous and visceral adipose tissue area, higher skeletal muscle area, but lower skeletal muscle density compared to persons of the same age without a chronic condition. However, the trend was often reversed in older adults aged 85 years. In this age group, a chronic condition was often associated with lower adiposity, and lower skeletal muscle area (at least in men). We hypothesize that persons with a chronic condition who survive to age 85 years tend to be those with a favorable body composition, and the reversed trends in cross-sectional body composition biomarkers at this age may be a reflection of survival bias. Finally, many chronic conditions were associated with lower vertebral bone density in women, but not in men. In aging adults (>85), degenerative changes in the spine and even compression fractures increase bone area and density and this appears to increase in patients with chronic conditions.
Strengths and Limitations
Strengths of this study include the construction of a CT cohort, which is representative of a well-defined general population, and the use of a previously validated CT body composition approach (18). In addition, we included comprehensive data for 6 body composition biomarkers and demonstrated that these biomarkers are influenced by sex, age, BMI, and the presence of common chronic conditions.
One limitation, shared by all CT body composition studies, is the impact of CT quality factors on body composition measurements. Persons with metal artifacts due to spine implants or other hardware were excluded due to issues with deep-learning model quality. In persons with extremely high BMI, portions of the abdomen may appear outside of the range-of-view of the CT scanner, which decreases subcutaneous adipose tissue area measurements. Skeletal muscle density and vertebral bone density measures may be affected by intravenous (IV) contrast. Unfortunately, data regarding the presence or absence of IV contrast was not readily available for the images used in this project. We expect the presence of IV contrast in some images to increase the variability of the density measurements, which would bias our comparisons of density measurements in persons with and without chronic conditions toward no association. However, we note that the percentiles for skeletal muscle and vertebral bone density biomarkers display a narrow and pronounced trend with age, suggesting that failing to correct for the presence of IV contrast had a minimal effect on study results (50). Finally, our study population was drawn from a geographically defined region of the Upper Midwest and reflects the characteristics of the population residing in this region. The proportion of persons from racial and ethnic minority populations in our study is lower than some other populations in the United States (51). Further studies focused specifically on persons from minority races and ethnicities are needed to quantify the effects of race and ethnicity on the body composition biomarkers.
Conclusion
We report reference ranges for CT-based body composition biomarkers in a population-representative cohort of 4 649 persons. Our results show the significant impact of sex, age, BMI, and a wide range of chronic conditions on these novel imaging biomarkers.
Supplementary Material
Acknowledgments
The authors would like to thank Barb Abbott and Susan Weston for help with data retrieval, analysis, and advice on study design and statistical support. The authors would like to acknowledge our sponsors, particularly NIH Institute on Aging, the Mayo Clinic Research Committee, and support from all users who access REP data.
Contributor Information
Alexander D Weston, Digital Innovation Lab, Department of Quantitative Health Sciences, Mayo Clinic, Jacksonville, Florida, USA.
Brandon R Grossardt, Division of Clinical Trials and Biostatistics, Department of Quantitative Health Sciences, Mayo Clinic, Rochester, Minnesota, USA.
Hillary W Garner, Division of Musculoskeletal Radiology, Department of Radiology, Mayo Clinic, Jacksonville, Florida, USA.
Timothy L Kline, Department of Radiology, Mayo Clinic, Rochester, Minnesota, USA.
Alanna M Chamberlain, Division of Epidemiology, Department of Quantitative Health Sciences, Mayo Clinic, Rochester, Minnesota, USA; Department of Cardiovascular Medicine, Mayo Clinic, Rochester, Minnesota, USA.
Alina M Allen, Division of Gastroenterology and Hepatology, Mayo Clinic, Rochester, Minnesota, USA.
Bradley J Erickson, Mayo Artificial Intelligence Laboratory, Department of Radiology, Mayo Clinic, Rochester Minnesota, USA.
Walter A Rocca, Division of Epidemiology, Department of Quantitative Health Sciences, Mayo Clinic, Rochester, Minnesota, USA; Department of Neurology, Mayo Clinic, Rochester, Minnesota, USA.
Andrew D Rule, Division of Epidemiology, Department of Quantitative Health Sciences, Mayo Clinic, Rochester, Minnesota, USA; Department of Medicine, Mayo Clinic, Rochester, Minnesota, USA.
Jennifer L St. Sauver, Division of Epidemiology, Department of Quantitative Health Sciences, Mayo Clinic, Rochester, Minnesota, USA; The Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery, Mayo Clinic, Rochester, Minnesota, USA.
Lewis A Lipsitz, (Medical Sciences Section).
Funding
This study used the resources of the Rochester Epidemiology Project (REP) medical records-linkage system, which is supported by the National Institute on Aging (NIA; AG 058738), by the Mayo Clinic Research Committee, and by fees paid annually by REP users. Additional support was provided by National Institute on Aging (NIA; AG 081223). The content of this article is solely the responsibility of the authors and does not represent the official views of the National Institutes of Health (NIH) or the Mayo Clinic.
Conflict of Interest
None.
Author Contributions
A.D.W. and B.R.G. are responsible for analysis and data curation. A.D.W., B.R.G., H.W.G., and J.S.S. are responsible for the initial draft of this manuscript. T.L.K., A.M.C., A.M.A., and B.J.E. are responsible for conceptualization and clinical input. W.A.R., A.R., and J.S.S. are responsible for project supervision. All authors contributed to and approved the final version of this manuscript.
References
- 1. Morley JE, Anker SD, von Haehling S.. Prevalence, incidence, and clinical impact of sarcopenia: facts, numbers, and epidemiology—update 2014. Springer; 2014. 253–259. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Abelson P, Kennedy D.. The obesity epidemic. American Association for the Advancement of Science; 2004. 1413–1413. [DOI] [PubMed] [Google Scholar]
- 3. Fukuda T, Bouchi R, Takeuchi T, et al. Ratio of visceral‐to‐subcutaneous fat area predicts cardiovascular events in patients with type 2 diabetes. J Diabetes Investig. 2018;9(2):396–402. 10.1111/jdi.12713 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Ryckman EM, Summers RM, Liu J, Del Rio AM, Pickhardt PJ.. Visceral fat quantification in asymptomatic adults using abdominal CT: is it predictive of future cardiac events? Abdom Imaging. 2015;40(1):222–226. 10.1007/s00261-014-0192-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Kaess B, Pedley A, Massaro J, Murabito J, Hoffmann U, Fox C.. The ratio of visceral to subcutaneous fat, a metric of body fat distribution, is a unique correlate of cardiometabolic risk. Diabetologia. 2012;55(10):2622–2630. 10.1007/s00125-012-2639-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. de Paula NS, Chaves GV.. Percentiles for body composition parameters based on computed tomography in patients with endometrial cancer. Nutrition. 2020;79-80:110873. 10.1016/j.nut.2020.110873 [DOI] [PubMed] [Google Scholar]
- 7. Welch C, Greig CA, Masud T, Pinkney T, Jackson TA.. Protocol for understanding acute sarcopenia: a cohort study to characterise changes in muscle quantity and physical function in older adults following hospitalisation. BMC Geriatr. 2020;20(1):1–11. 10.1186/s12877-020-01626-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Pickhardt PJ, Jee Y, O’Connor SD, del Rio AM.. Visceral adiposity and hepatic steatosis at abdominal CT: association with the metabolic syndrome. Am J Roentgenol. 2012;198(5):1100–1107. 10.2214/AJR.11.7361 [DOI] [PubMed] [Google Scholar]
- 9. Loosen SH, Schulze-Hagen M, Püngel T, et al. Skeletal muscle composition predicts outcome in critically ill patients. Crit Care Explor. 2020;2(8):e0171. 10.1097/CCE.0000000000000171 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Lee DH, Giovannucci EL.. Body composition and mortality in the general population: a review of epidemiologic studies. Exp Biol Med. 2018;243(17-18):1275–1285. 10.1177/1535370218818161 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Prado CM, Gonzalez MC, Heymsfield SB.. Body composition phenotypes and obesity paradox. Curr Opin Clin Nutr Metab Care. 2015;18(6):535–551. 10.1097/MCO.0000000000000216 [DOI] [PubMed] [Google Scholar]
- 12. Newman AB, Haggerty CL, Goodpaster B, et al. ; Health Aging And Body Composition Research Group. Strength and muscle quality in a well‐functioning cohort of older adults: the Health, Aging and Body Composition Study. J Am Geriatr Soc. 2003;51(3):323–330. 10.1046/j.1532-5415.2003.51105.x [DOI] [PubMed] [Google Scholar]
- 13. Pickhardt PJ, Summers RM, Garrett JW.. Automated CT-based body composition analysis: a golden opportunity. Korean J Radiol. 2021;22(12):1934–1937. 10.3348/kjr.2021.0775 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Bachrach L. Dual energy X-ray absorptiometry (DEXA) measurements of bone density and body composition: promise and pitfalls. J Pediatr Endocrinol Metab. 2000;13:983–988. [PubMed] [Google Scholar]
- 15. Moreno CC, Hemingway J, Johnson AC, et al. Changing abdominal imaging utilization patterns: perspectives from Medicare beneficiaries over two decades. J Am Coll Radiol. 2016;13(8):894–903. 10.1016/j.jacr.2016.02.031 [DOI] [PubMed] [Google Scholar]
- 16. Mwinyogle AA, Bhatt A, Ogbuagu OU, Dhillon N, Sill A, Kowdley GC.. Use of CT scans for abdominal pain in the ED: factors in choice. Am Surg. 2020;86(4):324–333. [PubMed] [Google Scholar]
- 17. Takahashi N, Sugimoto M, Psutka SP, Chen B, Moynagh MR, Carter RE.. Validation study of a new semi-automated software program for CT body composition analysis. Abdom Radiol (NY). 2017;42(9):2369–2375. 10.1007/s00261-017-1123-6 [DOI] [PubMed] [Google Scholar]
- 18. Weston AD, Korfiatis P, Kline TL, et al. Automated abdominal segmentation of CT scans for body composition analysis using deep learning. Radiology. 2019;290(3):669–679. 10.1148/radiol.2018181432 [DOI] [PubMed] [Google Scholar]
- 19. Pickhardt PJ, Graffy PM, Zea R, et al. Utilizing fully automated abdominal CT-based biomarkers for opportunistic screening for metabolic syndrome in adults without symptoms. Am J Roentgenol. 2021;216(1):85–92. 10.2214/ajr.20.23049 [DOI] [PubMed] [Google Scholar]
- 20. Magudia K, Bridge CP, Bay CP, et al. Population-scale CT-based body composition analysis of a large outpatient population using deep learning to derive age-, sex-, and race-specific reference curves. Radiology. 2021;298(2):319–329. 10.1148/radiol.2020201640 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Naser MA, Wahid KA, Grossberg AA, et al. Deep learning auto-segmentation of cervical skeletal muscle for sarcopenia analysis in patients with head and neck cancer. Front Oncol. 2022;12:930432. 10.3389/fonc.2022.930432 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Pickhardt PJ. Value-added opportunistic CT screening: state of the art. Radiology. 2022;303(2):241–254. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Navin P, Moynagh M, Atkinson E, et al. Establishment of normative biometric data for body composition based on computed tomography in a North American cohort. Clin Nutr. 2021;40(4):2435–2442. 10.1016/j.clnu.2020.10.046 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Somasundaram E, Castiglione JA, Brady SL, Trout AT.. Defining normal ranges of skeletal muscle area and skeletal muscle index in children on CT using an automated deep learning pipeline: implications for sarcopenia diagnosis. Am J Roentgenol. 2022;219(2):326–336. 10.2214/AJR.21.27239 [DOI] [PubMed] [Google Scholar]
- 25. Kong M, Geng N, Zhou Y, et al. Defining reference values for low skeletal muscle index at the L3 vertebra level based on computed tomography in healthy adults: a Multicentre Study. Clin Nutr. 2022;41(2):396–404. 10.1016/j.clnu.2021.12.003 [DOI] [PubMed] [Google Scholar]
- 26. Yoon JK, Lee S, Kim KW, et al. Reference values for skeletal muscle mass at the third lumbar vertebral level measured by computed tomography in a healthy Korean population. Endocrinol Metab. 2021;36(3):672–677. 10.3803/EnM.2021.1041 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. MacLean MT, Jehangir Q, Vujkovic M, et al. Quantification of abdominal fat from computed tomography using deep learning and its association with electronic health records in an academic biobank. J Am Med Inform Assoc. 2021;28(6):1178–1187. 10.1093/jamia/ocaa342 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. van Vugt JL, van Putten Y, van der Kall IM, et al. Estimated skeletal muscle mass and density values measured on computed tomography examinations in over 1000 living kidney donors. Eur J Clin Nutr. 2019;73(6):879–886. 10.1038/s41430-018-0287-7 [DOI] [PubMed] [Google Scholar]
- 29. Kim EH, Kim KW, Shin Y, et al. Reference data and T-scores of lumbar skeletal muscle area and its skeletal muscle indices measured by CT scan in a healthy Korean population. J Gerontol A Biol Sci Med Sci. 2021;76(2):265–271. 10.1093/gerona/glaa065 [DOI] [PubMed] [Google Scholar]
- 30. Bahat G, Turkmen BO, Aliyev S, Catikkas NM, Bakir B, Karan MA.. Cut-off values of skeletal muscle index and psoas muscle index at L3 vertebra level by computerized tomography to assess low muscle mass. Clin Nutr. 2021;40(6):4360–4365. 10.1016/j.clnu.2021.01.010 [DOI] [PubMed] [Google Scholar]
- 31. Gonzalez MC, Xiao J, Disi IR.. Sex-, age-, and ethnicity-dependent variation in body composition: can there be a single cutoff? In: Tandon P, Montano-Loza AJ, eds.. Frailty and sarcopenia in cirrhosis. Springer; 2020:119–126. [Google Scholar]
- 32. Derstine BA, Holcombe SA, Ross BE, Wang NC, Su GL, Wang SC.. Skeletal muscle cutoff values for sarcopenia diagnosis using T10 to L5 measurements in a healthy US population. Sci Rep. 2018;8(1):1–8. 10.1038/s41598-018-29825-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Rocca WA, Boyd CM, Grossardt BR, et al. Prevalence of multimorbidity in a geographically defined American population: patterns by age, sex, and race/ethnicity. 2014;89:1336–1349. 10.1016/j.mayocp.2014.07.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Baumgartner RN, Stauber PM, McHugh D, Koehler KM, Garry PJ.. Cross-sectional age differences in body composition in persons 60+ years of age. J Gerontol A Biol Sci Med Sci. 1995;50(6):M307–M316. 10.1093/gerona/50a.6.m307 [DOI] [PubMed] [Google Scholar]
- 35. Prado CM, Lieffers JR, McCargar LJ, et al. Prevalence and clinical implications of sarcopenic obesity in patients with solid tumours of the respiratory and gastrointestinal tracts: a population-based study. Lancet Oncol. 2008;9(7):629–635. 10.1016/S1470-2045(08)70153-0 [DOI] [PubMed] [Google Scholar]
- 36. Martin L, Birdsell L, MacDonald N, et al. Cancer cachexia in the age of obesity: skeletal muscle depletion is a powerful prognostic factor, independent of body mass index. J Clin Oncol. 2013;31(12):1539–1547. 10.1200/JCO.2012.45.2722 [DOI] [PubMed] [Google Scholar]
- 37. Pahor M, Manini T, Cesari M.. Sarcopenia: clinical evaluation, biological markers and other evaluation tools. J Nutr Health Aging. 2009;13(8):724–728. 10.1007/s12603-009-0204-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Pickhardt PJ, Pooler BD, Lauder T, del Rio AM, Bruce RJ, Binkley N.. Opportunistic screening for osteoporosis using abdominal computed tomography scans obtained for other indications. Ann Intern Med. 2013;158(8):588–595. 10.7326/0003-4819-158-8-201304160-00003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. St Sauver JL, Grossardt BR, Yawn BP, et al. Data resource profile: the Rochester Epidemiology Project (REP) medical records-linkage system. Int J Epidemiol. 2012;41(6):1614–1624. 10.1093/ije/dys195 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Rocca WA, Grossardt BR, Brue SM, et al. Data resource profile: expansion of the Rochester Epidemiology Project medical records-linkage system (E-REP). Int J Epidemiol. 2018;47(2):368–368j. 10.1093/ije/dyx268 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Goodman RA, Posner SF, Huang ES, Parekh AK, Koh HK.. Defining and measuring chronic conditions: imperatives for research, policy, program, and practice. Prev Chronic Dis. 2013;10:E66. 10.5888/pcd10.120239 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. St Sauver JL, Chamberlain AM, Bobo WV, et al. Implementing the US Department of Health and Human Services definition of multimorbidity: a comparison between billing codes and medical record review in a population-based sample of persons 40–84 years old. BMJ Open. 2021;11(4):e042870. 10.1136/bmjopen-2020-042870 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity‐score matched samples. Stat Med. 2009;28(25):3083–3107. 10.1002/sim.3697 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Weston AD. Automated segmentation of CT abdomen for quantifying body composition using deep learning. College of Medicine-Mayo Clinic; 2019. [Google Scholar]
- 45. Van der Werf A, Langius J, De Van Der Schueren M, et al. Percentiles for skeletal muscle index, area and radiation attenuation based on computed tomography imaging in a healthy Caucasian population. Eur J Clin Nutr. 2018;72(2):288–296. 10.1038/s41430-017-0034-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Abadi M, Barham P, Chen J, et al. Tensorflow: large-scale machine learning on heterogeneous distributed systems. arXiv preprint arXiv. 2016. 10.48550/arXiv.1603.04467 [DOI] [Google Scholar]
- 47. Team RC. R: A language and environment for statistical computing. 2013. [Google Scholar]
- 48. Muggeo VMR, Muggeo MVM.. Package ‘quantregGrowth’. Stat Med. 2023;25:1369–1382. [Google Scholar]
- 49. Derstine BA, Holcombe SA, Ross BE, Wang NC, Su GL, Wang SC.. Optimal body size adjustment of L3 CT skeletal muscle area for sarcopenia assessment. Sci Rep. 2021;11(1):1–10. 10.1038/s41598-020-79471-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Boutin RD, Kaptuch JM, Bateni CP, Chalfant JS, Yao L.. Influence of IV contrast administration on CT measures of muscle and bone attenuation: implications for sarcopenia and osteoporosis evaluation. Am J Roentgenol. 2016;207(5):1046–1054. 10.2214/AJR.16.16387 [DOI] [PubMed] [Google Scholar]
- 51. Sauver JLS, Grossardt BR, Leibson CL, Yawn BP, Melton LJ III, Rocca WA.. Generalizability of epidemiological findings and public health decisions: an illustration from the Rochester Epidemiology Project. Elsevier; 2012:151–160. [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.