Associations of QCT-derived body composition parameters with HOMA-IR-defined metabolic burden in patients with type 2 diabetes mellitus: a retrospective cross-sectional study

Abstract

Background

Insulin resistance is central to type 2 diabetes mellitus (T2DM) and is linked to abdominal adiposity, ectopic fat deposition, and skeletal muscle metabolism. However, HOMA-IR is a fasting surrogate measure that may be influenced by endogenous insulin secretion and treatment status in established T2DM. This study investigated associations between quantitative computed tomography (QCT)-derived body composition parameters and HOMA-IR-defined metabolic burden.

Methods

This retrospective cross-sectional study included 245 patients with T2DM who underwent abdominal QCT. Patients were categorized into low, middle, and high HOMA-IR groups by cohort-specific tertiles; high HOMA-IR was defined as the highest tertile (> 3.95). QCT parameters included visceral fat area (VFA), subcutaneous fat area (SFA), VFA/SFA ratio, intermuscular adipose tissue, skeletal muscle area and skeletal muscle index at L3 (SMI-L3), liver fat content, mean pancreatic fat content, and L1–L2 volumetric bone mineral density. Group comparisons, correlation analyses, and multivariable logistic regression were performed. Sensitivity analyses excluded basal insulin users and analyzed log-transformed HOMA-IR as a continuous outcome.

Results

Of the 245 patients, 82, 81, and 82 were assigned to the low, middle, and high HOMA-IR groups, respectively. VFA, SFA, VFA/SFA ratio, SMI-L3, and liver fat content differed significantly across tertiles. HOMA-IR was positively correlated with VFA, SFA, skeletal muscle area at L3, SMI-L3, and liver fat content. In the fully adjusted logistic regression model, VFA showed the largest point estimate for high HOMA-IR (odds ratio [OR] 2.450, 95% confidence interval [CI] 1.665–3.604, P < 0.001). Liver fat content (OR 1.618, 95% CI 1.180–2.217, P = 0.003), SMI-L3 (OR 1.659, 95% CI 1.054–2.611, P = 0.029), and VFA/SFA ratio (OR 1.501, 95% CI 1.065–2.115, P = 0.020) were also associated with high HOMA-IR. In sensitivity analyses, only VFA and liver fat content remained consistently significant.

Conclusions

VFA and liver fat content were the most consistent QCT-derived correlates of HOMA-IR-defined metabolic burden in patients with T2DM. Findings for VFA/SFA ratio and SMI-L3 were not reproduced in sensitivity analyses and should therefore be considered exploratory.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12902-026-02446-0.

Background

Type 2 diabetes mellitus (T2DM) is a major global metabolic disorder and continues to impose a substantial public health burden. According to the 11th edition of the International Diabetes Federation Diabetes Atlas, approximately 589 million adults aged 20–79 years were living with diabetes worldwide in 2024, and this number is projected to increase to 853 million by 2050. China has the largest number of adults with diabetes, with approximately 148.0 million affected adults aged 20–79 years in 2024 [1]. T2DM is a heterogeneous metabolic disorder characterized by impaired insulin secretion and insulin resistance [2]. Insulin resistance, a central metabolic abnormality in T2DM, reflects impaired insulin-mediated responses in key metabolic tissues, including adipose tissue, liver, and skeletal muscle, and contributes to hyperglycemia, dyslipidemia, ectopic lipid accumulation, and cardiometabolic complications [3].

Conventional anthropometric indices, such as body mass index (BMI) and waist circumference, are widely used in clinical practice but provide limited information on internal body composition. In particular, these measures cannot distinguish visceral from subcutaneous adiposity or adequately characterize skeletal muscle-related parameters and ectopic fat deposition [4]. This limitation is clinically relevant in patients with T2DM, because individuals with similar BMI may show markedly different patterns of abdominal fat distribution, skeletal muscle-related measures, and ectopic fat accumulation. Cross-sectional imaging, particularly computed tomography (CT)- or quantitative computed tomography (QCT)-based body composition analysis, provides an objective approach for quantifying abdominal adipose tissue distribution, skeletal muscle-related parameters, ectopic fat deposition, and bone-related measures within the same examination [5]. Previous CT- and QCT-based studies have evaluated body composition biomarkers, adipose tissue deposition, and pancreatic fat accumulation in diabetes, metabolic syndrome, and insulin resistance-related populations, supporting the potential value of imaging-derived body composition assessment beyond conventional anthropometric indices [6–9].

Among these compartments, abdominal fat distribution has received particular attention. Visceral adipose tissue is more closely associated with insulin resistance and cardiometabolic risk than subcutaneous adipose tissue [4, 10]. Visceral adiposity may contribute to insulin resistance through increased free fatty acid flux, adipose tissue dysfunction, chronic low-grade inflammation, and ectopic lipid deposition [4]. Longitudinal CT-based evidence further suggests that increases in intra-abdominal fat are associated with subsequent deterioration in insulin sensitivity [11]. Nevertheless, insulin resistance is not determined by abdominal adiposity alone. Because it involves multiple metabolic tissues, single-compartment assessment may be insufficient to describe the broader body composition phenotype associated with metabolic dysfunction in T2DM.

Ectopic fat deposition and skeletal muscle-related parameters may provide additional metabolic information. Liver fat accumulation has been relatively consistently linked to insulin resistance, whereas the metabolic significance of pancreatic fat remains less consistent across studies [7, 8, 12]. Skeletal muscle is a major insulin-responsive tissue responsible for a substantial proportion of insulin-mediated glucose disposal [13]. Accordingly, CT- or QCT-derived skeletal muscle area and height-normalized skeletal muscle index may provide opportunistic imaging markers of regional skeletal muscle quantity in metabolic studies. However, these area-based measures should be distinguished from muscle quality and function; intermuscular adipose tissue has increasingly been recognized as a metabolically relevant fat depot that may reflect muscle fat infiltration and impaired muscle quality [14–16]. Quantitative computed tomography (QCT) can also provide lumbar volumetric bone mineral density, although the relationship between insulin resistance and bone mineral density remains inconsistent, making volumetric bone mineral density an ancillary parameter in this setting [17]. Together, these considerations support a multi-compartment imaging approach for evaluating metabolic heterogeneity in patients with T2DM.

However, most previous imaging studies have examined individual compartments, such as visceral fat, liver fat, pancreatic fat, or skeletal muscle, rather than evaluating multiple QCT-derived body composition parameters within the same T2DM cohort. As a result, the relative associations of abdominal fat distribution, skeletal muscle-related characteristics, ectopic liver and pancreatic fat deposition, and lumbar volumetric bone mineral density with HOMA-IR-defined metabolic burden remain incompletely characterized. Accordingly, this study aimed to investigate the associations between QCT-derived body composition parameters and HOMA-IR-defined metabolic burden in patients with T2DM, and to identify which imaging-derived parameters showed more consistent associations after adjustment for clinical covariates and sensitivity analyses.

Methods

Study participants

This retrospective observational study used existing clinical, biochemical, and QCT data from consecutive patients with T2DM who visited the Affiliated Hospital of Gansu University of Chinese Medicine between November 17, 2025 and May 7, 2026 and underwent abdominal quantitative computed tomography (QCT) examination. The study was reviewed and approved by the Medical Ethics Committee of the Affiliated Hospital of Gansu University of Chinese Medicine (approval number: 2026–087). The requirement for written informed consent was waived by the ethics committee because of the retrospective nature of the study and the use of anonymized clinical and imaging data.

The inclusion criteria were as follows: (1) age between 40 and 80 years; (2) diagnosis of T2DM according to the Chinese Guideline for the Prevention and Treatment of Type 2 Diabetes Mellitus, 2020 edition [18], with reference to the American Diabetes Association Standards of Care in Diabetes—2026 [19], defined as fasting plasma glucose (FPG) ≥ 7.0 mmol/L, 2-hour plasma glucose during a 75-g oral glucose tolerance test ≥ 11.1 mmol/L, random plasma glucose ≥ 11.1 mmol/L in the presence of classic hyperglycemic symptoms, or glycated hemoglobin (HbA1c) ≥ 6.5%; confirmatory testing on a separate day was required for patients without classic symptoms; (3) continuous follow-up or treatment for at least 1 year after the diagnosis of T2DM; and (4) ability to cooperate with breath-holding instructions during abdominal QCT scanning. Antidiabetic medication use at the time of QCT examination was reviewed from the electronic medical records. Permitted antidiabetic regimens included metformin, α-glucosidase inhibitors, dipeptidyl peptidase-4 inhibitors, glinides, sulfonylureas, and basal insulin therapy, either alone or in combination according to clinical records. Sodium-glucose cotransporter 2 inhibitors and thiazolidinediones were not permitted because these agents may affect body weight, adipose tissue distribution, or ectopic fat deposition [20]. Patients receiving glucagon-like peptide-1 receptor agonists were also not included because this drug class may influence visceral adiposity and liver fat content [21]. Medication use in the final analyzed cohort was summarized across HOMA-IR tertiles.

Exclusion criteria were: (1) diabetes other than T2DM, or acute metabolic complications of diabetes, including diabetic ketosis, diabetic ketoacidosis, or hyperosmolar hyperglycemic state; (2) diseases that could substantially affect body composition, bone metabolism, muscle metabolism, or nutritional status, such as thyroid dysfunction, parathyroid disease, rheumatoid arthritis, Cushing syndrome, bone metastases, severe hepatic or renal dysfunction, or malabsorption; (3) a history of pathological fracture, or long-term use within the preceding 3 months of medications that could affect bone metabolism, muscle metabolism, or body composition; and (4) severe systemic illness, psychiatric disorders, cognitive impairment, inability to cooperate with QCT examination, or refusal to participate.

A total of 434 patients were initially screened. After excluding patients without eligible QCT examinations, those with incomplete scan coverage or failed QCT measurements, incomplete clinical or key biochemical data, clinical ineligibility, or non-permitted antidiabetic medication use, 245 patients with T2DM were included in the final analysis. The patient selection process is shown in Fig. 1.

Fig. 1.

Study flowchart of patient selection and grouping. Abbreviations: QCT, quantitative computed tomography; CT, computed tomography; T2DM, type 2 diabetes mellitus; HOMA-IR, homeostasis model assessment of insulin resistance

Demographic and clinical biochemical parameters

Demographic and clinical data were collected using the electronic medical record system, including age, sex, body mass index (BMI), diabetes duration, and blood pressure. Systolic blood pressure (SBP, mmHg) and diastolic blood pressure (DBP, mmHg) were measured after rest, and the average of at least two measurements was recorded. Biochemical parameters were obtained from the first fasting venous blood samples collected after admission and were measured using routine laboratory assays by the Department of Laboratory Medicine at the Affiliated Hospital of Gansu University of Chinese Medicine. The collected variables included fasting glucose (mmol/L), fasting insulin (µU/mL), fasting C-peptide (ng/mL), triglyceride (TG, mmol/L), total cholesterol (TC, mmol/L), high-density lipoprotein cholesterol (HDL-C, mmol/L), low-density lipoprotein cholesterol (LDL-C, mmol/L), alanine aminotransferase (ALT, U/L), aspartate aminotransferase (AST, U/L), serum creatinine (SCr, µmol/L), and uric acid (UA, µmol/L).

Assessment of HOMA-IR

The Homeostasis Model Assessment of Insulin Resistance (HOMA-IR) was used as a fasting surrogate index to characterize HOMA-IR-defined metabolic burden. HOMA-IR was calculated as fasting insulin (µU/mL) × fasting plasma glucose (mmol/L) / 22.5, according to the original homeostasis model assessment method [22]. Because this was a retrospective study based on routine clinical laboratory records, assay-specific information regarding cross-reactivity of the fasting insulin assay with exogenous insulin preparations or insulin analogues could not be reliably retrieved. Therefore, basal insulin therapy at the time of QCT examination was recorded from the electronic medical records, medication distribution was summarized across HOMA-IR tertiles, and a sensitivity analysis was performed after excluding patients receiving basal insulin therapy. Participants were categorized into three groups according to tertiles of HOMA-IR in the study cohort: low HOMA-IR (≤ 1.89), middle HOMA-IR (1.90–3.95), and high HOMA-IR (> 3.95). For logistic regression analysis, participants in the highest tertile were defined as the high HOMA-IR group, whereas those in the lower and middle tertiles were combined as the non-high HOMA-IR group. This threshold was cohort-derived and was used for exploratory risk-stratified analysis rather than as a general clinical cutoff for insulin resistance.

QCT examinations and image analysis

All participants underwent abdominal quantitative computed tomography (QCT) scanning in the supine position with both arms raised above the head. Scanning was performed using a dual-source CT scanner (SOMATOM Force; Siemens Healthineers, Germany), with a Mindways QCT calibration phantom (Mindways Software, Inc., Austin, TX, USA) placed beneath the participant during image acquisition. The scanning range covered the abdomen from the xiphoid process to the pubic symphysis. Acquisition parameters were as follows: tube voltage, 120 kV; tube current-time product, 200 mAs; gantry rotation time, 0.5 s; pitch, 1.0875; matrix, 512 × 512; and field of view, 500 mm. CT images were reconstructed using a standard algorithm with a slice thickness of 1.5 mm and a reconstruction interval of 1.0 mm. Images were transferred to a QCT workstation and analyzed using QCT Pro software, version 6.10 (Mindways Software, Inc., Austin, TX, USA). All QCT measurements were performed according to a standardized protocol by a trained radiologist and were reviewed by another radiologist with experience in abdominal imaging. Both radiologists were blinded to the HOMA-IR grouping and clinical laboratory results during image analysis. Disagreements, if any, were resolved by consensus. Trabecular volumetric bone mineral density (vBMD) of the lumbar spine was measured at L1–L2 using QCT. The mean L1–L2 vBMD was calculated as the average value of the two vertebral bodies. For each vertebra, an elliptical region of interest (ROI) was placed within the anterior two-thirds of the trabecular compartment, with the ROI margin kept at least 3 mm away from the cortical bone to minimize cortical partial-volume effects and avoid the basivertebral venous plexus. The ROI was adjusted to include as much trabecular bone as possible, with an area of approximately 250 mm² and a thickness of 9 mm (Fig. 2A).

Fig. 2.

Diagram illustrating QCT parameter measurements. (A) Trabecular volumetric bone mineral density was measured at L1–L2 using elliptical regions of interest placed within the vertebral trabecular bone. (B) Liver fat content was measured using three regions of interest placed on the same representative axial slice of the liver parenchyma. (C1–C3) Pancreatic fat content was measured using regions of interest placed in the pancreatic head, body, and tail, respectively. Abbreviations: QCT, quantitative computed tomography; ROI, region of interest; vBMD, volumetric bone mineral density

Liver fat content was measured using the QCT tissue measurement module. Measurements were performed on the largest cross-sectional image of the liver near the entry of the right portal vein branch. Three ROIs were placed in the liver parenchyma on the same representative axial slice, with one ROI placed in each of the left lobe, right anterior lobe, and right posterior lobe. Each ROI had an area of approximately 300 mm² and a thickness of 9 mm. Large intrahepatic vessels, bile ducts, hepatic margins, and obvious artifacts were carefully avoided. The mean value of the three ROIs was used as the liver fat content for subsequent analysis (Fig. 2B).

Pancreatic fat content was measured using the same QCT tissue measurement module. Three round ROIs were placed in the pancreatic head, body, and tail on representative axial images. Each ROI was positioned within the pancreatic parenchyma and avoided the main pancreatic duct, intrapancreatic vessels, peripancreatic fat, calcifications, cystic lesions, and obvious artifacts. Each ROI had an area of approximately 100–150 mm². The fat content values obtained from the pancreatic head, body, and tail were averaged and defined as the mean pancreatic fat content for subsequent analysis (Fig. 2C).

Body composition was assessed at the mid-L3 vertebral level using threshold-based segmentation. Fat tissue was defined using a threshold range of − 190 to − 30 Hounsfield units (HU), and skeletal muscle was defined using a threshold range of − 29 to 150 HU. Regions of interest were manually delineated at the L3 level to include the abdominal wall muscles, rectus abdominis, paraspinal muscles, psoas muscles, subcutaneous adipose compartment, and visceral adipose compartment.

Skeletal muscle area (SMA-L3), intermuscular adipose tissue (IMAT), subcutaneous fat area (SFA), and visceral fat area (VFA) were calculated after segmentation. SMA-L3 was used as an opportunistic QCT-derived measure of regional skeletal muscle quantity at the L3 level. The skeletal muscle index at L3 (SMI-L3) was calculated as SMA-L3 (cm²) divided by height squared (m²) to account for differences in body height. The V/S ratio was calculated as VFA divided by SFA. Representative L3-level body composition segmentation is shown in Fig. 3.

Fig. 3.

Representative L3-level abdominal adipose tissue and muscle segmentation. (A) Segmentation of adipose tissue at the mid-L3 vertebral level. The blue area inside the green contour line indicates visceral fat area, and the blue area outside the green contour line but within the outer abdominal boundary indicates subcutaneous fat area. (B) Delineation of visceral compartment at the same level. The abdominal wall compartment was derived by subtracting the visceral compartment ROI from the total abdominal ROI. Within this abdominal wall compartment, skeletal muscle area was calculated from pixels within the skeletal muscle attenuation range, and intermuscular adipose tissue was calculated from pixels within the adipose-tissue attenuation range. Abbreviations: L3, third lumbar vertebral level; ROI, region of interest; VFA, visceral fat area; SFA, subcutaneous fat area; SMA, skeletal muscle area; IMAT, intermuscular adipose tissue

Statistical analysis

All statistical analyses were performed using R software version 4.6.0 (R Foundation for Statistical Computing, Vienna, Austria) and IBM SPSS Statistics version 26.0 (IBM Corp., Armonk, NY, USA). Participants were categorized into low, middle, and high HOMA-IR groups according to cohort-specific HOMA-IR tertiles. The Shapiro–Wilk test was used to assess the normality of continuous variables. Continuous variables were described as the mean ± standard deviation (SD) for normally distributed data and the median (Q1, Q3) for non-normally distributed data. Categorical variables were expressed as n (%). For comparisons among the three HOMA-IR groups, one-way analysis of variance (ANOVA) was used for normally distributed continuous variables with homogeneity of variance, Welch’s ANOVA was used for normally distributed variables with unequal variances, and the Kruskal–Wallis test was used for non-normally distributed continuous variables. Categorical variables were compared using the chi-square test or Fisher’s exact test, as appropriate. When the overall comparison among the three HOMA-IR groups was statistically significant, post hoc pairwise comparisons were performed. Pairwise comparisons after ANOVA or Welch’s ANOVA were conducted using appropriate post hoc tests, and pairwise comparisons after the Kruskal–Wallis test were performed with Bonferroni adjustment.

Spearman correlation analysis was used to evaluate the associations between HOMA-IR and QCT-derived quantitative parameters. Binary logistic regression analyses were performed to evaluate the associations between QCT-derived quantitative parameters and high HOMA-IR status. The dependent variable was high HOMA-IR status, defined as the highest cohort-specific tertile of HOMA-IR (> 3.95), whereas non-high HOMA-IR was defined as HOMA-IR ≤ 3.95. Each QCT-derived quantitative parameter was entered separately as the independent variable in the logistic regression models. Three models were constructed: Model 1 was unadjusted; Model 2 was adjusted for age and sex; and Model 3 was adjusted for age, sex, BMI, and diabetes duration. QCT-derived quantitative parameters were standardized as z scores before logistic regression; therefore, odds ratios represent the change in odds of high HOMA-IR status per 1-standard-deviation increase in each parameter.

Sensitivity analyses were performed to evaluate the robustness of the main logistic regression findings. First, to assess the potential influence of basal insulin therapy on HOMA-IR-based classification, the fully adjusted multivariable logistic regression analyses were repeated after excluding patients receiving basal insulin therapy. The results of this analysis were reported as odds ratios (ORs) with 95% confidence intervals (CIs) and are presented in Supplementary Table S3. Second, an outcome-specification sensitivity analysis was performed to reduce reliance on the highest cohort-specific HOMA-IR tertile used to define high HOMA-IR. In this analysis, HOMA-IR was log-transformed and analyzed as a continuous outcome using multivariable linear regression models adjusted for age, sex, BMI, and diabetes duration. The results were reported as regression coefficients (B) with 95% CIs and are presented in Supplementary Table S4. In both sensitivity analyses, each standardized QCT-derived quantitative parameter was entered separately into the model.

Multicollinearity was assessed using variance inflation factors (VIFs). Each QCT-derived quantitative parameter was entered separately into a fully adjusted logistic regression model together with the prespecified covariates, and model-specific VIF values were calculated. A VIF value < 5 was considered to indicate no substantial collinearity between the tested QCT-derived parameter and the adjustment covariates. A two-sided P value < 0.05 was considered statistically significant.

Results

Demographic and clinical biochemical parameters of participants

A total of 434 participants were initially screened. After exclusion of 189 participants according to the predefined eligibility, imaging, biochemical, and medication criteria, 245 participants were ultimately enrolled in the study, including 154 men (62.9%) and 91 women (37.1%). According to tertiles of the HOMA-IR index, the enrolled participants were divided into the low HOMA-IR group (≤ 1.89, n = 82), middle HOMA-IR group (1.90–3.95, n = 81), and high HOMA-IR group (> 3.95, n = 82) (Fig. 1). Differences among the three HOMA-IR groups were observed in age, BMI, diabetes duration, fasting glucose, fasting insulin, fasting C-peptide, triglyceride, total cholesterol, HDL-C, uric acid, ALT, and AST (all P < 0.05). Compared with the low HOMA-IR group, patients in the high HOMA-IR group had higher BMI, fasting glucose, fasting insulin, triglyceride, ALT, and AST levels, and lower HDL-C levels (P < 0.05). Patients in the high HOMA-IR group had the shortest diabetes duration but the highest fasting C-peptide level among the three groups, whereas patients in the low HOMA-IR group had a longer diabetes duration and the lowest fasting C-peptide level. No statistically significant differences were observed in sex, LDL-C, serum creatinine, SBP, and DBP among the three groups (all P > 0.05) (Table 1). Antidiabetic medication use across HOMA-IR tertiles is summarized in Supplementary Table S2. Metformin use was common and similarly distributed across the three tertiles. No patient in the final analyzed cohort had documented GLP-1 receptor agonist, sodium-glucose cotransporter 2 inhibitor, or thiazolidinedione use. Basal insulin therapy was more frequent in the low HOMA-IR group than in the middle and high HOMA-IR groups.

Table 1.

Baseline characteristics according to HOMA-IR tertiles

Variable	Total N = 245	HOMA-IR			P value
		Low (≤ 1.89)	Middle (1.90–3.95)	High (> 3.95)
		n = 82	n = 81	n = 82
Age, years	63.00 (55.00, 70.00)	66.50 (61.00, 71.25)	61.00 (54.50, 68.00)a	62.00 (53.50, 70.00)a	0.003
Sex, n (%)					0.178
Female	91 (37.1)	24 (29.3)	32 (39.5)	35 (42.7)
Male	154 (62.9)	58 (70.7)	49 (60.5)	47 (57.3)
BMI, kg/m²	24.20 (22.33, 26.38)	23.22 (21.17, 24.66)	24.38 (22.85, 25.95)a	25.53 (23.59, 27.76)a	< 0.001
Diabetes duration, years	8.00 (3.25, 15.25)	12.38 (6.21, 20.02)	7.58 (2.34, 12.63)a	5.58 (1.56, 10.50)a	< 0.001
Fasting glucose, mmol/L	8.80 (7.50, 10.79)	8.00 (7.00, 10.23)	8.20 (7.10, 10.35)	9.95 (8.00, 12.00)ab	< 0.001
Fasting insulin, µU/mL	7.43 (3.96, 11.80)	3.17 (1.82, 4.22)	7.64 (6.57, 9.13)a	14.85 (11.05, 18.63)ab	< 0.001
Fasting C-peptide, ng/mL	2.26 (1.54, 3.37)	1.34 (0.85, 1.70)	2.36 (2.06, 2.92)a	3.53 (2.86, 4.31)ab	< 0.001
Triglyceride, mmol/L	1.43 (1.04, 2.11)	1.10 (0.82, 1.50)	1.61 (1.07, 2.1)a	1.98 (1.32, 2.94)ab	< 0.001
Total cholesterol, mmol/L	4.41 ± 1.11	4.16 ± 1.01	4.48 ± 1.08	4.59 ± 1.19a	0.035
HDL-C, mmol/L	1.08 (0.94, 1.27)	1.14 (0.95, 1.28)	1.13 (0.95, 1.28)	1.01 (0.90, 1.15)ab	0.005
LDL-C, mmol/L	2.65 ± 0.95	2.49 ± 0.91	2.73 ± 0.96	2.74 ± 0.97	0.167
SCr, µmol/L	69.00 (61.00, 78.50)	68.50 (59.75, 77.00)	67.00 (61.00, 78.00)	71.50 (60.50, 80.00)	0.937
UA, µmol/L	299.14 ± 78.99	282.71 ± 68.82	301.88 ± 80.76	312.87 ± 84.53a	0.046
ALT, U/L	21.40 (16.40, 32.00)	20.20 (15.35, 24.68)	21.10 (15.50, 27.85)	26.25 (18.05, 46.00)ab	< 0.001
AST, U/L	21.50 (18.35, 27.05)	20.25 (17.93, 24.50)	20.80 (18.45, 28.10)	23.65 (19.58, 33.70)a	0.012
SBP, mmHg	126.00 (122.00, 132.00)	126.00 (123.75, 132.00)	126.00 (120.00, 130.00)	126.00 (120.75, 134.00)	0.416
DBP, mmHg	78.00 (72.00, 84.00)	78.00 (70.00, 82.50)	78.00 (73.50, 85.00)	78.00 (71.75, 82.50)	0.777

Note: Data are expressed as the mean ± SD for continuous variables with a normal distribution, as the median (Q1, Q3) for continuous variables that significantly deviated from a normal distribution, and as n (%) for categorical variables. Pvalues were derived from one-way ANOVA, Welch’s ANOVA, or the Kruskal–Wallis test for continuous variables, and the chi-square test for categorical variables, as appropriate. Post hoc pairwise comparisons were performed when the overall comparison was statistically significant

Abbreviations: HOMA-IR, homeostasis model assessment of insulin resistance; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; DBP, diastolic blood pressure; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; SBP, systolic blood pressure; SCr, serum creatinine; T2DM, type 2 diabetes mellitus; UA, uric acid

aP< 0.05 vs the Low HOMA-IR group (≤1.89)

bP< 0.05 vs the Middle HOMA-IR group (1.90–3.95)

QCT quantitative parameters according to HOMA-IR tertiles

QCT quantitative parameters according to HOMA-IR tertiles are summarized in Table 2. VFA, SFA, V/S ratio, SMI-L3, and liver fat content differed significantly among the three groups. The high HOMA-IR group had significantly higher VFA than both the low and middle HOMA-IR groups (P < 0.05). SFA was significantly higher in the middle and high HOMA-IR groups than in the low HOMA-IR group (P < 0.05). SMI-L3 and liver fat content were higher in the high HOMA-IR group than in the low HOMA-IR group (P < 0.05). Although SMA-L3 tended to be higher in the high HOMA-IR group, the overall difference among the three groups did not reach statistical significance. No significant differences were observed in L1–L2 vBMD, IMAT, or mean pancreatic fat content among the three groups (all P > 0.05).

Table 2.

QCT-derived parameters according to HOMA-IR tertiles

Variable	Total N = 245	HOMA-IR			P value
		Low (≤ 1.89)	Middle (1.90–3.95)	High (> 3.95)
		n = 82	n = 81	n = 82
L1–L2 vBMD, mg/cm³	110.46 ± 33.97	107.09 ± 32.03	113.94 ± 35.62	110.38 ± 34.25	0.437
VFA, cm²	197.26 ± 77.16	169.15 ± 69.17	192.03 ± 72.42	230.55 ± 77.59ab	< 0.001
SFA, cm²	114.60 (86.25, 155.20)	90.80 (64.60, 120.05)	126.80 (99.35, 171.50)a	127.85 (104.03, 180.95)a	< 0.001
V/S ratio	1.61 (1.24, 2.14)	1.78 (1.30, 2.34)	1.47 (1.04, 1.96)a	1.66 (1.24, 2.16)	0.029
IMAT, cm²	15.80 (12.80, 19.00)	16.00 (12.35, 18.70)	15.90 (13.30, 19.05)	15.50 (12.78, 19.05)	0.915
SMA-L3, cm²	123.40 ± 28.18	117.95 ± 22.28	124.10 ± 27.60	128.15 ± 33.01	0.052
SMI-L3, cm²/m²	43.49 ± 8.02	41.33 ± 6.56	43.31 ± 7.42	45.84 ± 9.27a	0.002
Liver fat content, %	12.70 (9.90, 15.80)	11.30 (8.90, 13.25)	13.00 (10.60, 15.55)	14.80 (10.75, 18.33)a	< 0.001
Mean pancreatic fat content, %	16.73 (13.42, 19.57)	16.67 (12.35, 19.78)	16.73 (13.47, 18.78)	16.58 (13.55, 20.09)	0.738

Note: Data are expressed as the mean ± SD for continuous variables with a normal distribution and as the median (Q1, Q3) for continuous variables that significantly deviated from a normal distribution. P values were derived from one-way ANOVA, Welch’s ANOVA, or the Kruskal–Wallis test, as appropriate

Abbreviations: QCT, quantitative computed tomography; T2DM, type 2 diabetes mellitus; HOMA-IR, homeostasis model assessment of insulin resistance; BMD, bone mineral density; VFA, visceral fat area; SFA, subcutaneous fat area; V/S, VFA/SFA ratio; IMAT, intermuscular adipose tissue; SMA-L3, skeletal muscle area at the third lumbar vertebra; SMI-L3, skeletal muscle index at the third lumbar vertebra; vBMD, volumetric bone mineral density

aP< 0.05 vs the Low HOMA-IR group (≤1.89)

bP< 0.05 vs the Middle HOMA-IR group (1.90–3.95)

Correlations between HOMA-IR and QCT quantitative parameters

Spearman correlation analysis showed that HOMA-IR was positively correlated with VFA (r = 0.358, P < 0.001), SFA (r = 0.346, P < 0.001), SMA-L3 (r = 0.132, P = 0.039), SMI-L3 (r = 0.230, P < 0.001), and liver fat content (r = 0.303, P < 0.001). No significant correlations were observed between HOMA-IR and L1–L2 vBMD, V/S ratio, IMAT, or mean pancreatic fat content (all P > 0.05) (Table 3).

Table 3.

Correlations between HOMA-IR and QCT-derived parameters

Variables	Correlation	P value
L1–L2 vBMD, mg/cm³	−0.012	0.847
VFA, cm²	0.358	< 0.001
SFA, cm²	0.346	< 0.001
V/S ratio	−0.043	0.505
IMAT, cm²	−0.003	0.965
SMA-L3, cm²	0.132	0.039
SMI-L3, cm²/m²	0.230	< 0.001
Liver fat content, %	0.303	< 0.001
Mean pancreatic fat content, %	0.046	0.471

Note: Spearman correlation analysis was used to assess univariate associations between HOMA-IR and QCT-derived quantitative parameters. P values are two-tailed. Values of P < 0.05 were considered statistically significant

Abbreviations: HOMA-IR, homeostasis model assessment of insulin resistance; QCT, quantitative computed tomography; T2DM, type 2 diabetes mellitus; vBMD, volumetric bone mineral density; VFA, visceral fat area; SFA, subcutaneous fat area; V/S, VFA/SFA ratio; IMAT, intermuscular adipose tissue; SMA-L3, skeletal muscle area at the third lumbar vertebra; SMI-L3, skeletal muscle index at the third lumbar vertebra

Logistic regression analyses for high HOMA-IR

Binary logistic regression analyses were performed to evaluate the associations between QCT-derived quantitative parameters and high HOMA-IR. High HOMA-IR status was used as the dependent variable, and each standardized QCT-derived quantitative parameter was entered separately as the independent variable. The results are presented in Table 4, and the fully adjusted associations are shown in Fig. 4. In Model 1, the unadjusted model, VFA, SFA, SMI-L3, and liver fat content were significantly associated with high HOMA-IR. After adjustment for age and sex in Model 2, VFA, SFA, SMA-L3, SMI-L3, and liver fat content remained significantly associated with high HOMA-IR. After further adjustment for age, sex, BMI, and diabetes duration in Model 3, VFA, V/S ratio, SMI-L3, and liver fat content remained associated with high HOMA-IR.

Table 4.

Logistic regression analyses for high HOMA-IR

QCT parameter	Model 1		Model 2		Model 3
	OR (95% CI)	P value	OR (95% CI)	P value	OR (95% CI)	P value
Adipose distribution
VFA, cm²	2.011 (1.485–2.723)	< 0.001	2.506 (1.761–3.566)	< 0.001	2.450 (1.665–3.604)	< 0.001
SFA, cm²	1.688 (1.274–2.235)	< 0.001	1.648 (1.219–2.227)	0.001	1.404 (0.999–1.972)	0.051
V/S ratio	1.109 (0.854–1.441)	0.438	1.351 (0.985–1.853)	0.062	1.501 (1.065–2.115)	0.020
Skeletal muscle-related parameters
IMAT, cm²	1.042 (0.801–1.355)	0.761	1.104 (0.847–1.440)	0.463	1.167 (0.887–1.537)	0.270
SMA-L3, cm²	1.289 (0.987–1.683)	0.063	1.667 (1.180–2.356)	0.004	1.262 (0.746–2.136)	0.386
SMI-L3, cm²/m²	1.563 (1.187–2.060)	0.001	1.796 (1.313–2.458)	< 0.001	1.659 (1.054–2.611)	0.029
Ectopic fat deposition
Liver fat content, %	1.802 (1.347–2.410)	< 0.001	1.777 (1.324–2.385)	< 0.001	1.618 (1.180–2.217)	0.003
Mean pancreatic fat content, %	1.078 (0.828–1.403)	0.576	1.077 (0.825–1.406)	0.586	1.020 (0.774–1.344)	0.886
Bone mineral density
L1–L2 vBMD, mg/cm³	0.997 (0.764–1.300)	0.980	0.854 (0.602–1.211)	0.375	0.972 (0.667–1.418)	0.884

Note: Model 1 was unadjusted. Model 2 was adjusted for age and sex. Model 3 was adjusted for age, sex, BMI, and diabetes duration. All QCT-derived parameters were standardized as z scores before logistic regression; therefore, ORs represent the change in odds of high HOMA-IR status per 1-standard-deviation increase in each parameter

Abbreviations: QCT, quantitative computed tomography; HOMA-IR, homeostasis model assessment of insulin resistance; OR, odds ratio; CI, confidence interval; BMI, body mass index; VFA, visceral fat area; SFA, subcutaneous fat area; IMAT, intermuscular adipose tissue; SMA-L3, skeletal muscle area at the third lumbar vertebra; SMI-L3, skeletal muscle index at the third lumbar vertebra; vBMD, volumetric bone mineral density

Fig. 4.

Fully adjusted associations between QCT parameters and high HOMA-IR. Forest plot showing the adjusted odds ratios (ORs) and 95% confidence intervals (CIs) of QCT-derived parameters for high HOMA-IR. High HOMA-IR was defined as the highest tertile of HOMA-IR (> 3.95), whereas non-high HOMA-IR was defined as HOMA-IR ≤ 3.95. The fully adjusted model was adjusted for age, sex, body mass index, and diabetes duration. All QCT quantitative parameters were standardized as z scores before logistic regression; therefore, ORs represent the change in the odds of high HOMA-IR per 1-standard-deviation increase in each parameter. Abbreviations: QCT, quantitative computed tomography; T2DM, type 2 diabetes mellitus; HOMA-IR, homeostasis model assessment of insulin resistance; OR, odds ratio; CI, confidence interval; VFA, visceral fat area; SFA, subcutaneous fat area; V/S ratio, visceral-to-subcutaneous fat area ratio; IMAT, intermuscular adipose tissue; SMA-L3, skeletal muscle area at the third lumbar vertebra; SMI-L3, skeletal muscle index at the third lumbar vertebra; vBMD, volumetric bone mineral density

Because QCT-derived quantitative parameters were standardized before logistic regression, the odds ratios (ORs) represent the change in odds of high HOMA-IR per 1-standard-deviation increase in each parameter. In Model 3, VFA showed the largest point estimate among the significant parameters (OR 2.450, 95% CI 1.665–3.604, P < 0.001). Liver fat content (OR 1.618, 95% CI 1.180–2.217, P = 0.003), SMI-L3 (OR 1.659, 95% CI 1.054–2.611, P = 0.029), and V/S ratio (OR 1.501, 95% CI 1.065–2.115, P = 0.020) were also significantly associated with high HOMA-IR in the fully adjusted model. SFA showed a borderline association in Model 3 (OR 1.404, 95% CI 0.999–1.972, P = 0.051). In contrast, L1–L2 vBMD, IMAT, SMA-L3, and mean pancreatic fat content were not significantly associated with high HOMA-IR in the fully adjusted model.

In the sensitivity analysis excluding patients receiving basal insulin therapy, VFA remained associated with high HOMA-IR after adjustment for age, sex, BMI, and diabetes duration (OR 2.039, 95% CI 1.148–3.619, P = 0.015), as did liver fat content (OR 1.891, 95% CI 1.155–3.096, P = 0.011). In contrast, the associations of V/S ratio and SMI-L3 were no longer statistically significant. The remaining QCT-derived parameters were not significantly associated with high HOMA-IR in this sensitivity analysis (Supplementary Table S3). When HOMA-IR was analyzed as a continuous log-transformed outcome, VFA remained significantly associated with log-HOMA-IR after adjustment for age, sex, BMI, and diabetes duration (B = 0.274, 95% CI 0.110–0.439, P = 0.001), as did liver fat content (B = 0.167, 95% CI 0.035–0.299, P = 0.014). V/S ratio and SMI-L3 were not significantly associated with log-HOMA-IR in this continuous-outcome analysis (Supplementary Table S4).

Multicollinearity assessment

Model-specific collinearity diagnostics were performed because each QCT-derived quantitative parameter was entered separately into a fully adjusted logistic regression model with age, sex, BMI, and diabetes duration. The VIF values for all tested QCT-derived parameters were below 5, suggesting no substantial collinearity between each individual QCT-derived parameter and the adjustment covariates (Supplementary Figure S1 and Supplementary Table S1).

Discussion

This study used QCT to quantify abdominal fat distribution, skeletal muscle-related parameters, ectopic fat deposition, and lumbar vBMD in patients with T2DM, and examined their associations with HOMA-IR-defined metabolic burden. The main finding was that VFA and liver fat content were the most consistent QCT-derived correlates of high HOMA-IR. In the full cohort, VFA, V/S ratio, SMI-L3, and liver fat content were associated with high HOMA-IR in the fully adjusted logistic regression model. However, only VFA and liver fat content remained significant after excluding patients receiving basal insulin therapy and when HOMA-IR was analyzed as a continuous log-transformed outcome. Thus, visceral adiposity and liver fat accumulation appear to be the most robust imaging correlates in this cohort, whereas the associations involving V/S ratio and SMI-L3 were less consistent and should be interpreted as exploratory.

The interpretation of HOMA-IR in patients with established T2DM requires caution. HOMA-IR is derived from fasting glucose and fasting insulin and is a surrogate rather than a direct measure of tissue-specific insulin sensitivity [22, 23]. In treated patients with established T2DM, lower HOMA-IR values do not necessarily indicate normal insulin sensitivity, because they may also reflect lower endogenous insulin secretion or diminished β-cell reserve. This issue may explain why patients in the high HOMA-IR group had the shortest diabetes duration in the present cohort. These patients had the highest fasting C-peptide level, whereas patients in the low HOMA-IR group had the longest diabetes duration and the lowest fasting C-peptide level. This pattern suggests that patients with shorter diabetes duration may have retained greater endogenous insulin secretory capacity and therefore showed higher fasting insulin, C-peptide, and HOMA-IR values in the presence of insulin resistance. Conversely, lower HOMA-IR values in patients with longer diabetes duration may partly reflect reduced endogenous insulin secretion or β-cell reserve rather than better insulin sensitivity [24]. Therefore, HOMA-IR tertiles in this study should be interpreted as cohort-specific categories of HOMA-IR-defined metabolic burden rather than as sequential stages of diabetes severity or as normal-versus-abnormal insulin sensitivity categories.

The findings for abdominal fat distribution are consistent with previous imaging-based studies showing that visceral adiposity is more closely related to insulin resistance and adverse metabolic profiles than subcutaneous adiposity [10, 25]. In this study, VFA showed the most consistent association with HOMA-IR-defined metabolic burden, whereas the association of SFA was attenuated after adjustment for BMI and other clinical covariates. Previous findings regarding abdominal subcutaneous fat have been inconsistent. Niijima et al. reported an inverse association between subcutaneous abdominal fat area and HOMA-IR in a small cohort not specifically designed for patients with T2DM [26], whereas Hayashi et al. found that CT-measured intra-abdominal fat, but not abdominal subcutaneous fat, predicted future increases in insulin resistance indices among Japanese Americans [27]. Similarly, Liu et al. reported that longitudinal changes in visceral adiposity were more closely related to future insulin sensitivity than changes in abdominal subcutaneous fat [11]. Although previous studies have linked the visceral-to-subcutaneous fat ratio to cardiometabolic risk beyond BMI and visceral fat volume [28], the V/S ratio showed less consistent associations in the present study. Its association with high HOMA-IR was observed in the fully adjusted dichotomous model of the full cohort, but was not retained after excluding patients receiving basal insulin therapy and was not significant when HOMA-IR was analyzed as a continuous log-transformed outcome. Therefore, V/S ratio may provide exploratory distributional information on relative abdominal fat distribution, but it should not be interpreted as a robust standalone correlate in this cohort.

Liver fat content showed a more consistent association with HOMA-IR-defined metabolic burden than mean pancreatic fat content. Liver fat content differed significantly across HOMA-IR tertiles, was positively correlated with HOMA-IR, and remained associated with high HOMA-IR after adjustment for age, sex, BMI, and diabetes duration. This result is directionally consistent with a QCT-based study by Li et al. in patients with new-onset T2DM, in which estimated liver fat, but not pancreatic fat, remained associated with HOMA-IR after adjustment for age, sex, and BMI [7]. Mechanistically, this finding is consistent with the central role of the liver in glucose and lipid homeostasis and insulin action. Hepatic lipid accumulation may promote hepatic insulin resistance, lipotoxicity, and inflammatory responses, thereby contributing to the metabolic disturbances linking nonalcoholic fatty liver disease and T2DM [29]. In addition, Yang et al. reported that, among overweight or obese individuals, hepatic fat had greater value than pancreatic fat as a risk marker for impaired glucose regulation [30]. Although the present study focused on HOMA-IR in patients with T2DM rather than impaired glucose regulation itself, these findings support a relatively stable association between liver fat content and HOMA-IR-defined metabolic burden within the present QCT-based framework.

By contrast, mean pancreatic fat content was not significantly associated with either continuous HOMA-IR or high HOMA-IR. This negative finding should be interpreted in light of the heterogeneous evidence regarding pancreatic fat and metabolic dysfunction. Previous reviews have emphasized that associations between pancreatic fat, insulin resistance, β-cell function, and glucose dysregulation may vary according to obesity severity, liver fat accumulation, disease stage, and measurement methods [31]. Some QCT studies have reported positive correlations between pancreatic fat deposition and HOMA-IR in patients with T2DM, suggesting that the metabolic relevance of pancreatic fat may depend on population characteristics and analytical approach [8]. In the present study, mean pancreatic fat content was derived from ROI measurements in the pancreatic head, body, and tail. Although practical, this approach may not fully capture regional heterogeneity in pancreatic fat deposition. Pancreatic fat may also be more closely related to β-cell function, insulin secretory capacity, or diabetes stage than to HOMA-IR alone. The absence of a significant association in this study may therefore reflect both biological heterogeneity and limitations of ROI-based mean pancreatic fat assessment.

Skeletal muscle has a central role in insulin-mediated glucose uptake and whole-body glucose homeostasis [13], and previous cohort evidence has linked lower skeletal muscle mass and accelerated muscle loss to impaired glucose metabolism and incident diabetes [32]. This provides a rationale for including muscle-related QCT measures in a metabolic body-composition assessment, while also underscoring their limitations. Regional muscle area measured by QCT does not directly represent muscle quality, contractile function, or skeletal-muscle-specific insulin-mediated glucose disposal. In this study, SMI-L3 was associated with high HOMA-IR in the fully adjusted logistic regression model of the full cohort, whereas SMA-L3 was no longer significant after full adjustment. However, the association of SMI-L3 was not retained after excluding patients receiving basal insulin therapy and was not significant when HOMA-IR was analyzed as a continuous log-transformed outcome. Given the cross-sectional design, no causal relationship between SMI-L3 and HOMA-IR can be inferred. This finding should therefore not be interpreted as evidence that greater skeletal muscle mass promotes insulin resistance. Rather, SMI-L3 may reflect residual differences related to sex, body size, adiposity, weight-related mechanical loading, or disease-stage heterogeneity that were not fully captured by BMI and diabetes duration. In addition, HOMA-IR is a fasting surrogate index and does not directly measure skeletal-muscle-specific insulin sensitivity, which may partly explain the imperfect correspondence between L3-level muscle area and HOMA-IR-defined metabolic burden. Accordingly, SMI-L3 should be regarded as an exploratory regional body-composition marker rather than direct evidence of skeletal-muscle-specific insulin resistance.

Recent imaging studies have increasingly emphasized muscle quality, particularly myosteatosis, rather than muscle quantity alone [14, 33]. Myosteatosis generally refers to ectopic fat infiltration within skeletal muscle, but its imaging assessment is heterogeneous and may involve reduced muscle attenuation, low-attenuation muscle area, intermuscular adipose tissue, intramuscular fat, or composite CT- or MRI-derived muscle-quality indices [14, 33]. In a large abdominal CT-based study, higher HOMA-IR was associated with greater odds of CT-defined myosteatosis [34], supporting the concept that qualitative alterations in muscle composition may be more closely related to metabolic dysfunction than muscle area alone. In contrast, L3-level IMAT was not associated with high HOMA-IR after adjustment in the present study. This may be partly because IMAT area captures only one component of muscle fat infiltration and does not reflect muscle attenuation, intramuscular fat content, muscle strength, or physical performance [14, 15]. Thus, L3-level muscle area and IMAT should be interpreted as regional imaging markers rather than direct measures of skeletal muscle metabolic function in patients with T2DM.

Finally, lumbar vBMD showed no significant relationship with HOMA-IR-defined metabolic burden. L1–L2 vBMD did not differ across HOMA-IR tertiles, was not correlated with continuous HOMA-IR, and was not significantly associated with high HOMA-IR in either unadjusted or adjusted logistic regression models. This finding is consistent with evidence showing that the relationship between HOMA-based insulin resistance indices and BMD remains inconsistent, with no clear overall association across previous studies [17]. However, this negative result should not be interpreted as evidence against diabetes-related skeletal impairment, because patients with T2DM may have preserved or even higher BMD despite increased fracture risk [35]. In the present QCT-based framework, lumbar vBMD did not appear to be a major imaging correlate of HOMA-IR-defined metabolic burden. Instead, the most consistent imaging correlates were VFA and liver fat content, whereas V/S ratio and SMI-L3 should be interpreted as exploratory findings.

This study has several strengths. QCT allowed the assessment of abdominal fat distribution, skeletal muscle-related parameters, ectopic fat deposition, and lumbar vBMD within the same T2DM cohort, enabling a compartment-specific comparison beyond BMI alone. Image analysis followed a standardized QCT protocol, and radiologists were blinded to HOMA-IR grouping and clinical laboratory results during image review. In addition, the associations between QCT-derived parameters and HOMA-IR-defined metabolic burden were evaluated using complementary analyses, including HOMA-IR tertile-based comparisons, correlation analyses, multivariable logistic regression with standardized QCT-derived parameters, and sensitivity analyses excluding patients receiving basal insulin therapy and analyzing HOMA-IR as a continuous log-transformed outcome.

Nevertheless, several limitations should be considered. First, this was a single-center retrospective cross-sectional study, and causal relationships between QCT-derived body composition parameters and HOMA-IR-defined metabolic burden cannot be established. In particular, the association between SMI-L3 and high HOMA-IR in the primary logistic regression model should not be interpreted as evidence that greater skeletal muscle mass causes or promotes insulin resistance. Second, HOMA-IR is a fasting surrogate index and does not directly measure tissue-specific insulin sensitivity; moreover, high HOMA-IR was defined using the highest cohort-specific tertile rather than a universal clinical cutoff. Although medication distribution was reported and sensitivity analyses excluding patients receiving basal insulin therapy and using log-transformed HOMA-IR as a continuous outcome were performed, residual confounding related to antidiabetic medication use, endogenous insulin secretion, and β-cell reserve cannot be fully excluded. In addition, assay-specific cross-reactivity between the fasting insulin assay and exogenous insulin preparations or insulin analogues could not be confirmed from the retrospective laboratory records. Third, QCT measurements were based on selected anatomical levels and ROI-based assessments, and formal intraobserver and interobserver reproducibility analyses were not performed. Therefore, L3-level muscle and fat areas and ROI-based liver and pancreatic fat measurements may not fully represent whole-body composition, regional ectopic fat heterogeneity, or measurement variability. Finally, muscle attenuation, intramuscular fat content, muscle strength, physical performance, and lifestyle-related variables were not available, which limits the interpretation of skeletal muscle-related parameters. Future prospective multicenter studies incorporating functional muscle assessment, longitudinal metabolic outcomes, and more comprehensive medication and β-cell function data are needed to validate these findings.

Conclusions

In patients with T2DM, QCT-derived body composition parameters showed differential associations with HOMA-IR-defined metabolic burden. VFA and liver fat content were the most consistent imaging correlates of high HOMA-IR, including in sensitivity analyses excluding patients receiving basal insulin therapy and analyzing HOMA-IR as a continuous log-transformed outcome. Associations involving V/S ratio and SMI-L3 were less consistent and should be interpreted as exploratory. These findings support the potential value of QCT-based multi-compartment body composition assessment for characterizing metabolic heterogeneity in patients with T2DM, while emphasizing that cohort-derived HOMA-IR categories should not be interpreted as universal clinical cutoffs or as direct measures of tissue-specific insulin sensitivity.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ALT

Alanine aminotransferase

ANOVA

Analysis of variance

AST

Aspartate aminotransferase

BMI

Body mass index

CI

Confidence interval

CT

Computed tomography

DBP

Diastolic blood pressure

DPP-4

Dipeptidyl peptidase-4

FPG

Fasting plasma glucose

GLP-1

Glucagon-like peptide-1

HbA1c

Glycated hemoglobin

HDL-C

High-density lipoprotein cholesterol

HOMA-IR

Homeostasis model assessment of insulin resistance

HU

Hounsfield unit

IMAT

Intermuscular adipose tissue

LDL-C

Low-density lipoprotein cholesterol

OR

Odds ratio

QCT

Quantitative computed tomography

ROI

Region of interest

SBP

Systolic blood pressure

SCr

Serum creatinine

SD

Standard deviation

SFA

Subcutaneous fat area

SGLT2

Sodium-glucose cotransporter 2

SMA-L3

Skeletal muscle area at the third lumbar vertebra

SMI-L3

Skeletal muscle index at the third lumbar vertebra

T2DM

Type 2 diabetes mellitus

TC

Total cholesterol

TG

Triglyceride

UA

Uric acid

VFA

Visceral fat area

V/S ratio

Visceral-to-subcutaneous fat area ratio

vBMD

volumetric bone mineral density

VIF

Variance inflation factor

Author contributions

Authors’ contributions ML and LW conceived and designed the study; ML collected the clinical and biochemical data, performed the statistical analysis, and drafted the manuscript; ML and SD performed the QCT image measurements; JY and SD reviewed the imaging data and contributed to measurement quality control. LW supervised the study, provided methodological guidance, interpreted the results, and critically revised the manuscript for important intellectual content.

Funding

No funding was received for this study.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was reviewed and approved by the Medical Ethics Committee of the Affiliated Hospital of Gansu University of Chinese Medicine (approval number: 2026–087). The study was conducted in accordance with the Declaration of Helsinki. The requirement for written informed consent was waived by the ethics committee because of the retrospective nature of the study and the use of anonymized clinical and imaging data.

Consent for publication

Not applicable. No identifiable individual information or images are included in this article.

Competing interests

The authors declare no competing interests.