Associations of Thigh and Abdominal Adipose Tissue Radiodensity with Glucose and Insulin in Non-Diabetic African-Ancestry Men

Abstract

Objective

Decreased radiodensity of adipose tissue (AT) located in the visceral (VAT), subcutaneous (SAT), and intermuscular (IMAT) abdominal depots is associated with hyperglycemia, hyperinsulinemia and insulin resistance independent of AT volumes. We sought to determine these associations in African Ancestry men, who have higher risk for type 2 diabetes and were underrepresented in previous studies.

Methods

This cross-sectional analysis included 505 non-diabetic men of African-Caribbean ancestry (median age: 61 years, median BMI: 26.8 kg/m²) from the Tobago Health Study. AT volumes and radiodensities were assessed using computed tomography, including abdominal (VAT and SAT) and thigh (IMAT) depots. We assessed associations between AT radiodensities with fasting serum glucose and insulin, and insulin resistance (HOMA2-IR).

Results

Higher radiodensity in any AT depot was associated with lower log-insulin and log-HOMA2-IR (β range: −0.16 to −0.19 for each, all p<0.0001). No AT radiodensity was associated with glucose. Thigh IMAT radiodensity associations were independent of, and similar in magnitude to, VAT radiodensities. Model fit statistics suggested AT radiodensity was a better predictor for insulin and insulin resistance compared to AT volumes in individuals with overweight and obesity.

Conclusions

AT radiodensities at multiple depots are significantly associated with insulin and insulin resistance in African Ancestry men.

Introduction

Despite the fact that obesity is a major driver of type 2 diabetes (T2D), T2D also depends on the distribution of adipose tissue (AT) throughout the body, especially the amount of fat around and within non-AT organs (known as ectopic AT) (1). The size of non-ectopic AT depots such as the subcutaneous AT (SAT) and visceral AT (VAT) depots are associated with worse glucose and insulin levels, with the VAT depot being more strongly associated than SAT (2–4). Additionally, ectopic AT depots such as abdominal and thigh intermuscular AT (IMAT) volume are also associated with impaired glucose and insulin levels (3, 5, 6).

In addition to AT distribution, novel surrogate markers of AT biology, such as computed tomography (CT)-derived average AT radiodensity, may indicate more pathogenic AT. Biopsy studies performed in rodents and small human trials suggest that AT radiodensity may capture other AT attributes such as cell size (7), lipid content (8), and vascularity (9). Indeed, AT radiodensity is emerging as a marker of increased risk for cardiometabolic disease independent of tissue volume (10–14). Several reports using the Framingham cohort observed lower abdominal VAT and SAT radiodensity associated with worse cardiometabolic profiles, including higher HOMA-IR (12), increased odds of impaired fasting glucose (11, 12), and increased glucose concentrations (10–12). Studies from the Multi Ethnic Study of Atherosclerosis (MESA) have also reported that individuals with lower vs. higher abdominal VAT, SAT, and IMAT radiodensity had greater levels of glucose and diabetes (13, 14).

These studies were performed in predominantly Caucasian cohorts or were not stratified by race; thus, there remains a paucity of data on associations of AT radiodensity with biomarkers of T2D risk in non-Caucasian racial/ethnic groups. This is an important area of inquiry, as African Ancestry individuals, who are at a higher risk of T2D independent of overall adiposity (15–17), also exhibit different ectopic AT distributions compared to their Caucasian counterparts, such as having lower VAT (18–21), higher abdominal SAT (18, 20, 21), and higher total IMAT (22, 23). Additionally, data on non-abdominal IMAT radiodensity, such as in the thigh, is very sparse. Skeletal muscle, which is insulin sensitive, is found in larger quantities in the thigh compared to the abdomen (24). The location of IMAT next skeletal muscle suggests a role for IMAT in insulin resistance; indeed, IMAT-secreted factors were shown to reduce insulin sensitivity in myotubes in vitro (25), and thigh IMAT volume is positively associated with both insulin resistance and risk of T2D independent of overall obesity (3, 26, 27). There are also racial/ethnic differences in regional IMAT distribution, with studies reporting greater thigh and calf IMAT in African Ancestry individuals compared to Caucasians (26, 28, 29) as opposed to similar (29, 30) or lower (31) levels of abdominal IMAT.

Thus, our primary objective was to determine if lower AT radiodensity, including thigh IMAT radiodensity, was associated with higher fasting serum levels of glucose and insulin, as well as insulin resistance, in middle-aged and older African Ancestry men without T2D. We hypothesized that less-dense AT would be associated with worse glucose and insulin levels.

Methods

Study Population

All men in this analysis were from the Tobago Health Study, which has been previously described (32). Briefly, the Tobago Health Study is a population-based, prospective cohort study of community-dwelling men aged 40 years and older, residing on the Caribbean island of Tobago, Trinidad and Tobago. Men from Tobago are of homogeneous African ancestry with low European admixture (<6%) (33). Participants in the Tobago Health Study were recruited without regard to health status and men were eligible if they were ambulatory, not terminally ill, and without a bilateral hip replacement. The baseline visit occurred from 2004–2007 and recruited 2,482 men; of these, a random subset (N=1,725) attended the first follow-up visit from 2010–2014. Men used in the current analysis attended an ancillary study visit from 2014–2018, when a convenience sub-sample of N=768 participants from the prior visit had computed tomography (CT) scans of the chest, abdomen, and mid-thigh for ectopic AT assessment. Exclusion from the current analysis included missing CT scans in the abdomen or in one or both thighs (N=33), having T2D (N=174), missing covariate data (N=12), non-African Caribbean ethnicity by self-report (N=43), and non-fasting serum samples (N=1). Individuals with T2D were excluded to better reflect potential associations of AT radiodensity on glucose and insulin levels without confounding effects of the later disease process. The final analytical sample included 505 individuals. Written informed consent was obtained from each participant using forms and procedures approved by the University of Pittsburgh Institutional Review Board, the U.S. Surgeon General’s Human Use Review Board, and the Tobago Division of Health and Social Services Institutional Review Board. This study was completed in accordance with the Declaration of Helsinki.

Computed Tomography Scans

Abdominal and thigh volumes and radiodensities were assessed on 3 mm thick slices and 500 mm display field of view from scans acquired using a GE dual slice, high-speed NX/I CT scanner (GE Medical Systems, Waukesha, WI) with 120 KVp, 250 mA, 0.7 second gantry speed, and pitch of 1.5:1. For participants with body weight greater than 200 lbs, the mA was increased to 300. CT contrast was not used. Only one CT scanner was used, and a single individual collected the scans for all participants. Scans were electronically transmitted to the central CT reading center at Vanderbilt University where image analysis and quality control were performed.

Image analysis was performed using methods as previously described (30, 34, 35). Briefly, images were analyzed using a dedicated imaging processing workstation with custom-programmed subroutines (OsiriX, Pixmeo, Geneva, Switzerland) and a dedicated pen computing display (Cintiq, Wacom Technology Corporation, Vancouver, WA, USA). A radiologist-trained analyst manually traced anatomical boundaries (skin, muscular fascia, muscle, and peritoneum) in CT scans. Tissue attenuation thresholds of −190 to −30 Hounsfield Units (HU) were used to distinguish AT voxels in these defined regions. For each tissue, the volume (mm³) and mean tissue attenuation (referred to here as AT radiodensity, in HU) were calculated.

Abdominal VAT and SAT were measured from CT scans of 3 contiguous slices of 3mm thickness centered at L4-L5. A lateral scout image was used to determine the z-axis location of the L4-L5 intervertebral space and that location and the slice immediately above and the slice immediately below were used to reconstruct a 9-mm thick single block of images. VAT was defined as AT located within the peritoneal cavity; SAT was defined as AT located beneath the skin and superficial to the abdominal muscular fascia.

Thigh IMAT was measured from CT scans of 10 contiguous slices of 3mm thickness at the mid-thigh level in both legs. An anterior-posterior scout scan of the entire femur was used to localize the mid-thigh position, and that location and the four slices immediately above and five slices immediately below were used to reconstruct a 30-mm thick single block of images. IMAT was defined as AT located within thigh muscle groups. IMAT volume was defined as the total IMAT summed across both thighs, and IMAT radiodensity was the average AT attenuation across both thighs.

Intrareader technical error (TE) in re-analysis of a 5% oversampling of blinded scans was 0.6% for total abdominal volume, 1.4% for SAT volume, and 4.8% for VAT volume whereas TE for abdominal radiodensities were 0.7% for SAT and 0.5% for VAT. For thigh measures, TE was 1.5% for total thigh volume, 2.7% for IMAT volume, and 1.0% for IMAT radiodensity.

Anthropometric Measurements

Standing height was measured to the nearest 0.1 cm using a wall-mounted stadiometer. Body weight was recorded to the nearest 0.1 kg without shoes on a balance beam scale. BMI was calculated from body weight and standing height (kg/m²). BMI categories were defined as normal weight (18.5 ≤ BMI < 25), overweight (25 ≤ BMI < 30), and obese (BMI ≥ 30). Waist circumference was measured at the narrowest point of the waist using an inelastic tape. If there was no narrowest point, waist circumference was measured at the umbilicus.

Glucose and Insulin Measures

Fasting serum glucose and insulin measures were measured at the Advanced Research and Diagnostics Laboratory (ARDL), University of Minnesota. Fasting serum glucose was measured using an enzymatic procedure (interassay CV: 1.3–1.8%) and fasting serum insulin was measured using a sandwich immunoassay procedure (interassay CV: 3.1%) (assays manufacturer: Roche Diagnostics, Indianapolis, IN). Glucose was assessed in mg/dL units that were converted to SI units (mmol/L) by dividing by 18. Insulin was assessed in pmol/L units. The degree of insulin resistance was estimated by homeostasis model assessment (HOMA2-IR) and calculated using fasting glucose and insulin values and the HOMA2 Calculator v2.2.3 available from http://www.dtu.ox.ac.uk/homacalculator/ (36).

Other Measures

Information on lifestyle habits (current smoking [yes/no], number of hours walked per week, watching 14 or more hours of television per week [yes/no], current intake of alcohol of more than 4 drinks per week [yes/no]), ethnic self-identification (dichotomized as African-Caribbean Ethnicity [yes/no]), and medication use were assessed using standardized interviewer-administered questionnaires. Self-reported information on walking was recorded as walking is the predominant form of physical activity on the island of Tobago. Men were asked to bring all prescription medications taken in the past 30 days to their clinic visit. T2D was defined as currently taking an antidiabetic medication, regardless of fasting serum glucose level, or having a fasting serum glucose level of ≥7 mmol/l; all men with T2D were excluded from the current analyses.

Statistical Analyses

Men were stratified by median AT radiodensity, and comparisons of their baseline characteristics were made using two-sample t-test or Wilcoxon rank-sum test for continuous variables and Chi-square tests for categorical variables. Age-adjusted partial Spearman correlations were performed between AT radiodensities and outcomes/covariates.

Multiple linear regression models were performed separately for each outcome (glucose, insulin, HOMA2-IR), and separately for each predictor of interest (VAT radiodensity, SAT radiodensity, IMAT radiodensity). Insulin and HOMA2-IR were log-transformed to make residuals normal. Model covariates included age, weight, height, hours walked per week for exercise, watching 14+ hours of television per week, current smoking, and drinking 4+ alcoholic drinks per week. Identification of significant curvilinear relationships between AT volume and radiodensity (Figure 1) suggested that mutual adjustment for these variables would be inappropriate; thus, models including both AT volume and radiodensity were not included in our main analyses. A series of three models were run sequentially:

• Model 1: model covariates + a specific AT radiodensity

• Model 2: model covariates + VAT and SAT radiodensities

• Model 3: model covariates + VAT and IMAT radiodensity

Figure 1: Scatterplot of AT radiodensity by volume, per tissue depot.

Abbreviations: AT = Adipose Tissue, VAT = Visceral Adipose Tissue, SAT = Subcutaneous Adipose Tissue, IMAT = Intermuscular Adipose Tissue

Multicollinearity was assessed using condition indices and was not found to be an issue in any of these models; however, multicollinearity was an issue for models including both SAT and IMAT radiodensities, so models which included simultaneous addition of both of these depots were not reported, but can be found in Supplementary Material tables. Sensitivity analyses for regression models included further adjustment for respective total thigh and abdominal scan volumes (obtained from CT images), adjustment for respective AT depot volume, and adjustment for respective AT depot volume after stratification by BMI category. In the BMI-stratified analyses, four individuals were excluded for having a BMI below 18.5 kg/m².

Changes in model fit statistics (Akaike Information Criteria [AIC] and Bayesian Information Criteria [BIC]) were assessed when all three AT radiodensity or all three AT volumes were added to covariate-adjusted models; though collinearity can affect the estimates of individual predictors, overall model fit is not affected, allowing simultaneous inclusion of multiple AT depots. A change in fit ≤ −2 for either AIC or BIC was regarded as representing greater model fit. Interactions between radiodensity and age were assessed.

Statistical significance was based on an α = 0.05, and analyses were performed using SAS 9.4 software (SAS Institute, Inc., Cary, NC). Graphical 3-Dimensional Scatterplot was generated using the package scatterplot3d (37) in R version 3.5.2 (38).

Results

General Baseline Characteristics

Men had a median age of 61 years and a median BMI of 26.8 kg/m² (Table 1). There were low rates of reported physical activity and high reported sedentary behavior, but other lifestyle factors of current smoking and alcohol intake were relatively low. We next stratified our sample characteristics by median AT radiodensities (Supplementary Table S1). Overall, individuals with a higher AT radiodensity in either the VAT, SAT, or IMAT depots had significantly lower anthropometric measures (weight, BMI, and waist circumference), lower AT depot volumes, and higher AT depot radiodensities in any depot (all p <0.0001).

Table 1:

Characteristics of African-Caribbean men (N=505)

Variable	Median (IQR), Mean (SD), or N (%)
Age (years)	61.0 (56.0, 68.0)
Weight (kg)	82.5 (73.4, 93.1)
Height (cm)	175.6 (6.7)
BMI (kg/m2)	26.8 (24.2, 29.9)
Underweight [BMI < 18.5] (%)	4 (0.8%)
Normal Weight [18.5 ≤ BMI < 25] (%)	159 (31.5%)
Overweight [25 ≤ BMI < 30] (%)	219 (43.4%)
Obese [BMI ≥ 30] (%)	123 (24.4%)
Waist Circumference (cm)	95.9 (89.0, 104.0)
Lifestyle and Comorbidities
Current Smoker (%)	36 (7.1%)
Drinks Alcohol 4+/week (%)	67 (13.3%)
Hours Walked per Week	2.0 (0.0, 5.0)
Watches television ≥ 14 hours/week (%)	242 (47.9%)
CT-Derived Measures
VAT Volume (cm3)	78.3 (49.0, 118.0)
SAT Volume (cm3)	174.4 (118.6, 234.1)
IMAT Volume (cm3)	108.0 (84.4, 140.5)
VAT Radiodensity (HU)	−89.6 (−94.7, −82.1)
SAT Radiodensity (HU)	−99.7 (−103.3, −94.0)
IMAT Radiodensity (HU)	−70.9 (−73.3, −68.0)
Total Abdominal Volume (cm3)	563.3 (465.1, 668.0)
Total Thigh Volume (cm3)	1625.3 (1396.4, 1849.4)
Glucose and Insulin Metabolism
Glucose (mmol/L)	4.8 (4.4, 5.2)
Insulin (pmol/L)	52.0 (34.0, 81.0)
HOMA2-IR	1.0 (0.6, 1.5)

Results reported as Median (IQR) or Mean (SD) for continuous variables and N (%) for categorical.

Abbreviations: VAT = Visceral Adipose Tissue, SAT = Subcutaneous Adipose Tissue, IMAT = Intermuscular Adipose Tissue

Association of AT Depot Attenuation with Anthropometric Measures and Levels of Glucose and Insulin

Table 2 shows interrelationships of AT radiodensities with anthropometric, CT-based, and glucose/insulin variables. In general, AT radiodensities were strongly positively correlated with each other and strongly negatively correlated with anthropometry and AT volume measures, with stronger associations seen between VAT and SAT radiodensity (all p < 0.0001). AT radiodensity in any depot was strongly negatively correlated with insulin levels and HOMA2-IR (p < 0.0001), while only VAT and SAT radiodensity were significantly (though weakly) associated with glucose.

Table 2:

Interrelationship of all adiposity and metabolic measures used in our analyses (age-adjusted Spearman correlations)

Variable	VAT Radiodensity	SAT Radiodensity	IMAT Radiodensity
Weight (kg)	−0.57‡	−0.60‡	−0.39‡
BMI (kg/m2)	−0.62‡	−0.68‡	−0.44‡
Waist Circumference (cm)	−0.64‡	−0.72‡	−0.47‡
VAT Volume (cm3)	−0.82‡	−0.70‡	−0.47‡
SAT Volume (cm3)	−0.64‡	−0.84‡	−0.57‡
IMAT Volume (cm3)	−0.52‡	−0.61‡	−0.57‡
Total Abdominal Scan Volume (cm3)	−0.67‡	−0.74‡	−0.48‡
Total Thigh Scan Volume (cm3)	−0.55‡	−0.67‡	−0.49‡
VAT Radiodensity (HU)	--	0.74‡	0.48‡
SAT Radiodensity (HU)	0.74‡	--	0.66‡
IMAT Radiodensity (HU)	0.48‡	0.66‡	--
Glucose (mmol/L)	−0.14†	−0.13†	−0.08
Insulin (pmol/L)	−0.58‡	−0.63‡	−0.47‡
HOMA2-IR	−0.58‡	−0.63‡	−0.47‡

^†= <0.05

^‡= <0.0001

Abbreviations: VAT = Visceral Adipose Tissue, SAT = Subcutaneous Adipose Tissue, IMAT = Intermuscular Adipose Tissue

The relationship between each AT depot’s radiodensity and corresponding volume was explored visually using a scatterplot (Figure 1). Relationships followed an inverse association, where tissues at higher radiodensities tended to be of lower volume while tissues at lower radiodensities tended to have much higher volumes. Though the IMAT depot followed a similar curvilinear relationship as the VAT and SAT depots, it increased in volume at a much higher radiodensity as compared to the other depots, and did not reach the levels of low radiodensity that the other depots did.

The results from multiple linear regression analyses are shown in Table 3. Multiple linear regression model building indicated high collinearity between SAT and IMAT radiodensities; when regressing on one another, the model R² for SAT and IMAT radiodensities was ~0.70. Given these strong interrelationships, reported regression models did not include additional adjustment for AT volume, nor did they include mutual adjustment for SAT and IMAT radiodensities (though sensitivity analyses including these adjustments can be found in the Supplementary Tables S2–S4).

Table 3:

Difference in fasting glucose and insulin levels and HOMA2-IR per 1 SD (95% CI) increase in AT radiodensity

Outcome	Model	VAT Radiodensity (SD = 8.44 HU)	SAT Radiodensity (SD = 10.71 HU)	IMAT Radiodensity (SD = 5.70 HU)
Glucose (mmol/L)	M + Depot	−0.04 (−0.11, 0.03)	−0.05 (−0.12, 0.02)	−0.01 (−0.07, 0.05)
	M + VAT + SAT	−0.03 (−0.10, 0.05)	−0.03 (−0.11, 0.04)	--
	M + VAT + IMAT	−0.04 (−0.11, 0.03)	--	−0.00 (−0.07, 0.07)
Log Insulin (pmol/L)	M + Depot	−0.16 (−0.21, −0.11) ‡	−0.18 (−0.24, −0.13) ‡	−0.16 (−0.21, −0.11) ‡
	M + VAT + SAT	−0.10 (−0.16, −0.05) †	−0.14 (−0.20, −0.08) ‡	--
	M + VAT + IMAT	−0.12 (−0.17, −0.07) ‡	--	−0.13 (−0.18, −0.08) ‡
Log HOMA2-IR	M + Depot	−0.16 (−0.21, −0.11) ‡	−0.18 (−0.24, −0.13) ‡	−0.16 (−0.20, −0.11) ‡
	M + VAT + SAT	−0.10 (−0.16, −0.05) †	−0.14 (−0.20, −0.08) ‡	--
	M + VAT + IMAT	−0.12 (−0.17, −0.07) ‡	--	−0.13 (−0.18, −0.08) ‡

M = age, weight, height, alcohol intake, smoking, hours walked/week, and television watching ≥ 14 hours/week

Significant P-values

^†= <0.05

^‡= <0.0001

Abbreviations: AT = Adipose Tissue, VAT = Visceral Adipose Tissue, SAT = Subcutaneous Adipose Tissue, IMAT = Intermuscular Adipose Tissue

After adjustment for age, weight, height, alcohol intake, smoking, hours walked per week, and television watching, higher radiodensity in any tissue was associated with significantly lower serum insulin and lower HOMA2-IR (all p <0.0001). SAT and IMAT radiodensity associations persisted even after VAT radiodensity adjustment. To provide context using HOMA2-IR results, the smallest reported effect size of β=−0.11 (for VAT radiodensity, after adjustment for SAT radiodensity) indicates a 10.6% lower HOMA2-IR for every 1 SD increase in VAT radiodensity; in contrast, the largest effect size of β=−0.19 (for SAT adjustment alone) indicates a 17.3% lower HOMA2-IR for every 1 SD increase in SAT radiodensity. Notably, no AT radiodensity was statistically significantly associated with glucose, though there was a tendency towards lower glucose at higher AT radiodensities. Results remained similar in a number of sensitivity analyses including adjustment for all 3 AT radiodensities, adjustment for respective AT volumes, and adjustment for total abdominal and thigh CT scan volume (Supplementary Tables S2–S4).

We additionally explored the potential interactions of AT radiodensity-by-age in the main analysis models. No significant AT radiodensity-by-age interactions were identified for any outcome.

Model Fit Statistics

Changes in model fit statistics (AIC and BIC) were evaluated for the linear regression models (Table 4), where covariate-adjusted models were further adjusted for either all three AT radiodensities or all three AT volumes. For linear regression models, the simultaneous addition of VAT, SAT, and IMAT radiodensities to log-insulin and log-HOMA2-IR models greatly improved model fit by both AIC and BIC criteria. Compared to volume-alone or volume- and radiodensity-adjusted models, the radiodensity-alone models accounted for a greater increase in model fit. Glucose models had worsening model fit by inclusion of AT radiodensities and/or volumes. Changes in model AIC and BIC for individual AT radiodensities and volumes remained similar to simultaneous AT inclusion (Supplementary Table S5).

Table 4:

Change in multiple linear regression model fit statistics after inclusion of all AT depot radiodensities

Outcome	Δ AIC			Δ BIC
	All Radiodensities	All Volumes	Radiodensities & Volumes	All Radiodensities	All Volumes	Radiodensities & Volumes
Glucose (mmol/L)	2.77	3.07	3.89	3.00	3.30	4.43
Log-Insulin (pmol/L)	−57.45 †	−29.73 †	−58.95 †	−57.21 †	−29.50 †	−58.41 †
Log-HOMA2-IR	−58.01 †	−29.47 †	−59.48 †	−57.78 †	−29.24 †	−58.94 †

^†= improvement of fit (Δ ≤ −2) compared to covariate-alone model

Compares inclusion/exclusion of (1) all 3 AT radiodensities simultaneously, (2) all three AT volumes simultaneously, or (3) all 3 AT radiodensities AND volumes, to the base covariate model [age, weight, height, alcohol intake, smoking, hours walked/week, and television watching ≥ 14 hours/week]

Abbreviations: AT=Adipose Tissue

Models Stratified by BMI-Category

To further disentangle the relationship between AT volume, radiodensity, and T2D risk factors, we investigated the consistency of these relationships within each by BMI category. Models adjusted for covariates age, hours walking for exercise, TV watching, smoking, drinking, and a specific AT depot’s volume and radiodensity together. The regression coefficients for each depot’s radiodensity and volume are plotted in Supplementary Figure S1–S3. Briefly, an overall trend was observed in which, after adjustment for AT volume, coefficients for AT radiodensity were significantly and inversely associated with log-insulin and log-HOMA2-IR in individuals with overweight and obesity, but not in individuals with normal weight status. Conversely, AT volumes were significantly and positively associated with log-insulin and logHOMA2-IR in individuals with normal weight status, but not in individuals with overweight or obesity, after adjustment for AT radiodensity. Only lower VAT radiodensity and higher SAT volume were significantly associated with small increases in glucose, and only in individuals with normal weight status. These results were supported by assessing changes in model fit statistics (Supplementary Table S6).

Discussion

In this population study of non-diabetic middle-aged and older African ancestry men, we identified relationships between higher AT radiodensity and better levels of insulin and insulin resistance. These relationships were overall similar but not entirely consistent with those seen in predominantly white or race -adjusted cohorts. Our study included the novel addition of thigh IMAT radiodensity, an AT depot in peripheral skeletal muscle, which showed associations that were independent of and as strong as VAT radiodensities. Importantly, insulin and HOMA2-IR model fits were largely improved with the addition of AT radiodensities, demonstrating that AT radiodensity may be most closely linked to insulin compared to glucose levels in these men.

AT radiodensity is a CT-derived measure indicating the tissue’s opacity to X-rays. As such, it is not a direct measure of AT biology, which can only be assessed through invasive techniques such as biopsies. Nonetheless, studies indicate that AT radiodensity can reflect structural aspects of AT (7–9) and is associated with metabolic health (10–14). Importantly, there is a lack of information on AT radiodensity in African Ancestry populations, who are at higher risk for development of T2D independent of overall obesity (15–17). This is an important knowledge gap, as there are known racial/ethnic differences in AT distribution which may impact T2D risk, and suggestions of racial/ethnic differences in AT radiodensity as well (13). While cohort studies of Framingham and MESA have laid the groundwork for the associations of AT radiodensity with cardiometabolic health, further studies of these associations in African Ancestry individuals, specifically, and focusing additionally on relevant non-abdominal AT depots are needed to demonstrate consistency of findings in this high-risk racial/ethnic group.

While explorations of AT radiodensity and diabetes-related biomarkers are relatively novel, we are the first study to report on thigh IMAT radiodensity. MESA found that levels of glucose and prevalence of diabetes increased across decreasing quartiles of abdominal IMAT radiodensity (independent of age, gender, and race/ethnicity) (14), indicating that IMAT radiodensity may play a role in diabetes risk. In contrast, our study found no associations between lower thigh IMAT radiodensity and fasting glucose, but did find significant associations with higher insulin and insulin resistance, even after adjustment for VAT radiodensity. It is important to note that regional IMAT accumulation may differentially impact the relationship between IMAT and metabolic health (39). Comparisons of regional distribution of IMAT are not well studied; however, results from Ruan et al (29) suggest that the thigh may have a greater amount of IMAT compared to the waist. Additionally, studies suggest that IMAT in African Ancestry individuals tends to be found in larger amounts in the thigh (28, 29) but not in the abdomen (29–31) when compared to Caucasian individuals. These differences in IMAT distribution and their metabolic consequences suggest a need for greater understanding of non-abdominal IMAT radiodensity and T2D risk factors.

Another interesting finding regarding IMAT radiodensity was its high collinearity with SAT, but not VAT, radiodensity. IMAT volume increases with increasing total AT, and racial/ethnic differences suggest that African American individuals have a greater IMAT increase at higher levels of adiposity compared to Asian and white individuals (22). We and others have demonstrated that while IMAT increases with aging, the rate of IMAT accumulation depends on weight gain status, with greater accumulation in weight-gainers and lesser accumulation in weight-losers (5, 40); this suggests that IMAT accumulation is influenced by overall adiposity, a large proportion of which is comprised of SAT. The origins of IMAT adipocytes is hypothesized to be derived from muscle-based mesenchymal progenitor cells (41); however, a recent study in mice demonstrated the ability of SAT-derived adipocyte progenitors to be released from SAT and take up residence in muscle as IMAT in response to nutrient overload (42). People with obesity have a five times higher level of circulating progenitor cells (43). We report that IMAT volume increases at relatively higher radiodensities compared to SAT, perhaps indicating an increase in cell number rather than cellular hypertrophy as the driving force behind increased IMAT volume. Further research into human IMAT cellularity and origins is warranted.

Our findings of associations between abdominal AT radiodensities and glucose and insulin were similar, but not entirely consistent, with cross-sectional reports in men from the Framingham study (12). Notably, we reported stronger associations between SAT radiodensity than Framingham after adjustment for weight and height; additionally, our associations with glucose were not statistically significant and were markedly smaller than those reported in Framingham. These differences may be due to systematic differences between measurement collections, as Framingham scanned a much larger abdominal area; or they may also be attributable to analyzed population differences, given the overall differences between the cohorts (Caucasian vs. African Ancestry, ~10 year average older age in our study) and our exclusion of individuals with T2D.

In addition to examining the relationship between AT radiodensities and diabetesrelated biomarkers, we were also able to determine the relative importance of AT radiodensity compared to AT volume, measures which are both derived from the same CT scan and reflect different aspects of tissue structure. We found that any specific AT depot’s volume and radiodensity was strongly and inversely associated and followed a curvilinear pattern, similar to previous reports (13, 44). Given the strength of the curvilinear relationships in our study sample, we felt that mutual adjustment for AT radiodensity and volume would not be appropriate. However, to get at the relative importance of AT radiodensity compared to volume, we compared changes in model fit statistics. We report consistent and large improvements of model fit with the addition of AT radiodensities to our models of insulin and HOMA2-IR. Importantly, these improvements in model fit were remarkably larger than those seen in models using AT volumes, and they weren’t substantially improved by simultaneous addition of AT radiodensities and volumes. However, stratification by BMI category revealed a pattern whereby AT volume was a more significant and informative marker in individuals with normal weight but not in individuals with overweight or obesity, and AT radiodensity exhibited the opposite pattern. That the associations with radiodensity were present only in individuals with overweight and obesity after volume adjustment may indicate dysfunctional AT growth in these individuals.

Our study has several potential limitations. Given our use of an all-male cohort, we were unable to examine associations in women. Another limitation is the use of self-reported physical activity, which may not be the most accurate assessment of physical activity (45). However, our study also has several strengths. The use of multi-slice CT at multiple anatomical locations allowed us to obtain and compare volumetric and radiodensity data across multiple AT depots. Additionally, the use of an African-ancestry cohort provides information on an understudied and high-risk ethnic group.

In conclusion, higher AT radiodensities are significantly associated with lower insulin and HOMA2-IR levels in a cohort of African-ancestry men. These associations are independent of adjustment of other depot radiodensities. Our novel thigh IMAT findings highlight the importance of this depot’s radiodensity in diabetes risk. Future studies investigating changes in IMAT radiodensity with insulin metabolism and mechanistic studies of AT radiodensity are warranted.

What is already known about this subject?

• Abdominal adipose tissue radiodensity (visceral, subcutaneous, and intermuscular) is inversely associated with glucose and insulin levels, with most studies conducted in predominantly Caucasian cohorts

What does your study add?

• Adipose tissue radiodensities, including thigh intermuscular adipose tissue, are inversely associated with fasting insulin and insulin resistance in African Ancestry individuals

• Adipose tissue radiodensities may be more informative predictors of insulin resistance than adipose tissue volumes