A combined model integrating adiponectin and anthropometric indices for the identification of insulin resistance in patients with type 2 diabetes mellitus Authors: Xiuping Yin, Yunna Zhang, Cuiliu Li, Xinxin Yang, Jinxiu Xu, Fang Gao, Jie Ma, Renyong Huang, Guangya Wang Published: 2026 ABSTRACT BackgroundTo investigate the associations of adiponectin (APN), body mass index (BMI), and waist-to-hip ratio (WHR) with insulin resistance (IR) in patients with type 2 diabetes mellitus (T2DM), and to evaluate a combined model for discriminatory performance and potential clinical value for the identification of IR.MethodsA total of 394 hospitalized patients with T2DM were retrospectively analyzed. Participants were divided into low-IR and high-IR groups based on the median HOMA-IR. Clinical data were then analyzed by Spearman’s correlation, trend analysis, multiple linear regression, logistic regression, and receiver operating characteristic (ROC) curve analysis.ResultsCompared with the low-IR group, patients in the high-IR group exhibited a higher BMI and WHR but lower APN levels (P < 0.05 for all). Furthermore, APN was negatively correlated with HOMA-IR, whereas BMI and WHR were positively correlated (all P < 0.01). Multivariable analysis further revealed that APN (β = −0.335), BMI (β = 0.243), and WHR (β = 0.138) were independently associated with IR (P < 0.01 for all). The combined model showed better discrimination (AUC = 0.82) than APN (0.78), WHR (0.72), and BMI (0.69) alone.ConclusionsAPN, BMI, and WHR were independently associated with IR in patients with T2DM. Our combined model demonstrates superior discriminative performance to individual indicators and may be useful for identifying individuals at high risk of IR.Clinical trial numberNot applicable. Introduction Over recent years, population aging and lifestyle changes have contributed to the continued increase in the prevalence of type 2 diabetes mellitus (T2DM), a disease that now represents a major concern for global public health. According to the International Diabetes Federation (IDF) Diabetes Atlas (2025), approximately 589 million adults worldwide are now living with diabetes, and this number is projected to increase to 853 million by 2050 [1]. Notably, the proportion of overweight and obese individuals among patients with T2DM is increasing, with central obesity characterized by the accumulation of visceral fat playing a key role in the development and progression of disease [2]. Visceral adiposity is closely associated with metabolic abnormalities and significantly increases the risk of cardiovascular and cerebrovascular diseases [3]. Insulin resistance (IR), a core pathophysiological mechanism underlying T2DM, is strongly associated with both the amount and distribution of adipose tissue [4, 5]. In clinical and epidemiological studies, IR is commonly estimated using the homeostasis model assessment of insulin resistance (HOMA-IR), a practical surrogate index derived from fasting glucose and fasting insulin concentrations [6, 7]. In the present retrospective study, HOMA-IR was used as the operational measure for assessing IR. Excess visceral fat is known to promote metabolic dysfunction via multiple mechanisms, including the altered secretion of adipokines, chronic low-grade inflammation, and the ectopic deposition of lipids, thereby exacerbating IR and impairing glycemic control [8]. Adiponectin (APN), an adipokine secreted by adipocytes, plays a critical role in the metabolism of both glucose and lipid, and insulin sensitivity [9, 10]. In contrast, body mass index (BMI) reflects overall adiposity, whereas waist-to-hip ratio (WHR) serves as an indicator for the distribution of fat, particularly abdominal obesity [11, 12]. These indices capture distinct yet complementary aspects of metabolic status, including adipokine activity, total fat mass, and the distribution of fat. However, most previous studies have focused on individual indicators, and integrated assessment models combining adipokines with anthropometric measures for evaluating IR remain insufficiently explored. In clinical practice, exercise intervention, together with dietary and pharmacological approaches, represents a key strategy for metabolic management [13–15]. However, conventional interventions are often applied uniformly and may not adequately account for inter-individual differences in metabolic status and the distribution of fat [16]. A more refined stratification based on metabolic characteristics may help to optimize intervention strategies and improve metabolic outcomes. Accordingly, in this study, we aimed to investigate the associations between serum APN, BMI, and WHR with IR in patients with T2DM, and to develop and evaluate the discriminatory performance of a combined model integrating these indicators for identifying individuals with high HOMA-IR-defined IR, thereby providing a potential basis for metabolic risk stratification. Materials and methods Study population We enrolled patients with T2DM who had been hospitalized in the Department of Endocrinology, Cangzhou Central Hospital, Hebei Province, between December 2021 and December 2023. All participants met the diagnostic criteria for T2DM according to the relevant standards of the American Diabetes Association (ADA) [17]. The inclusion criteria were as follows: (1) fulfillment of the diagnostic criteria for T2DM; (2) aged ≥ 18 years; and (3) a complete set of clinical data, including height, weight, waist circumference, hip circumference, blood glucose, lipid parameters, and serum adiponectin. The exclusion criteria were as follows: (1) type 1 diabetes, gestational diabetes, or other specific types of diabetes; (2) severe hepatic or renal dysfunction, malignant tumors, or acute stress conditions; (3) severe infection or acute cardio-cerebrovascular events; (4) diseases that exert significant influence on adiponectin or lipid metabolism, such as severe inflammatory diseases or autoimmune disorders; and (5) the recent use of medications that may exert significant influence on metabolic status, such as glucocorticoids. Collection of clinical data and measurement of indicators For all subjects, we collected general clinical information, including age and sex. Clinical measurements, including height, weight, systolic blood pressure (SBP), diastolic blood pressure (DBP), waist circumference, and hip circumference, were obtained in the early morning after an overnight fast. BMI and WHR were then calculated. Medication information was reviewed from the electronic medical records when available. However, detailed data on statin type, dose, and treatment duration were not consistently recorded for all patients because of the retrospective nature of the study; therefore, statin therapy could not be fully evaluated in the present analysis. Blood samples were collected between 7:00 and 9:00 a.m. after an overnight fast of at least 12 h. Fasting blood glucose (FBG), triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C) were measured using standard enzymatic methods with an automatic biochemical analyzer (Hitachi 7600-120ISE, Hitachi, Japan). Fasting insulin (FINS) was measured using an electrochemiluminescence immunoassay. Serum adiponectin (APN) was measured using an enzyme-linked immunosorbent assay kit (Human ADPN ELISA Kit, Shanghai UNIQUE Biotechnology Co., Ltd., Shanghai, China). All measurements were performed in accordance with the manufacturers’ instructions and standard operating procedures. Grouping method The homeostasis model assessment of insulin resistance (HOMA-IR) was used to evaluate insulin resistance and was calculated as HOMA-IR = FBG × FINS / 22.5, where FBG was expressed in mmol/L and FINS in µIU/mL. Participants were divided into a low-IR group and a high-IR group according to the median HOMA-IR for between-group comparisons. In addition, according to HOMA-IR tertiles, participants were categorized into Q1 (low level), Q2 (moderate level), and Q3 (high level) groups for trend analysis. Statistical analysis Statistical analyses were performed using SPSS 26.0 software, and MedCalc software was used for ROC curve comparison. The normality of continuous variables was assessed using the Kolmogorov–Smirnov test and visual inspection of histograms and Q-Q plots. Normally distributed variables were expressed as mean ± standard deviation (x̄ ± s), whereas non-normally distributed variables were expressed as median (interquartile range). Categorical variables were expressed as number (percentage). For comparisons between two groups, normally distributed continuous variables with homogeneity of variance were analyzed using the independent-samples t test, and homogeneity of variance was assessed using Levene’s test. Non-normally distributed variables were compared using the Mann–Whitney U test, and categorical variables were compared using the χ² test. Trend analysis was performed using the Jonckheere–Terpstra test. Correlation analysis was conducted using Spearman’s rank correlation. After Ln transformation of the skewed HOMA-IR distribution, Ln(HOMA-IR) was included as the dependent variable in a multiple linear regression model to evaluate the independent associations between indicators and insulin resistance. Logistic regression analysis was further performed to identify independent factors associated with insulin resistance. ROC curves were then used to evaluate the discriminative ability of each indicator and the combined model for insulin resistance, and the area under the curve (AUC) was calculated. DeLong tests were used to compare differences in AUC between models. All tests were two-sided, and P < 0.05 was considered statistically significant. Results Clinical characteristics The baseline characteristics of the study population are shown in Table 1. Participants were divided by IR status into a low-IR group (n = 199) and a high-IR group (n = 195). Compared with the low-IR group, the high-IR group exhibited significantly higher BMI, WHR, DBP, FBG, FINS, TG, TC, LDL-C, and HOMA-IR (P < 0.05 for all), whereas HDL-C, HOMA-β, and APN were all significantly lower (P < 0.05 for all). In addition, there were significant differences between the two groups in terms of sex distribution, age, waist circumference, and hip circumference (P < 0.05 for all); there was no significant difference between the groups with respect to SBP. Table 1Baseline characteristics of participants according to insulin resistance statusVariableIR low (n = 199)IR high (n = 195)P valueGeneral characteristicsSex, n (%)Female: 117 (58.79%)Male: 82 (41.21%)Female: 92 (47.18%)Male: 103 (52.82%)0.021ᵇ*Age, years63.44 ± 8.8960.28 ± 8.760.001ᵇ**Body composition–related indicatorsBMI, kg/m²24.57 (22.67, 26.58)26.36 (25.19, 27.59)< 0.001ᵃ***Waist circumference, cm88.34 ± 9.3594.48 ± 10.810.012ᵇ*Hip circumference, cm98.08 (93.91, 105.00)94.92 (89.16, 102.40)0.018ᵃ*WHR0.89 (0.85, 0.93)0.97 (0.89, 1.09)< 0.001ᵃ***DBP, mmHg77.43 ± 7.5380.49 ± 7.720.002ᵇ**SBP, mmHg128.00 (121.00, 134.69)127.03 (119.74, 136.00)0.742ᵃGlucose, lipid, and adipokine metabolism indicatorsFBG, mmol/L6.05 (5.20, 7.80)10.08 (8.57, 11.75)< 0.001ᵃ***FINS, µIU/mL10.54 (9.54, 12.26)15.60 (13.51, 17.97)< 0.001ᵃ***TG, mmol/L1.47 (1.14, 2.05)2.48 (1.65, 3.05)< 0.001ᵃ***TC, mmol/L3.28 (1.83, 4.63)5.48 (3.91, 6.93)< 0.001ᵃ***HDL-C, mmol/L1.25 (0.97, 1.64)1.07 (0.86, 1.43)< 0.001ᵃ***LDL-C, mmol/L2.92 (2.27, 3.53)3.78 (2.88, 4.46)< 0.001ᵃ***HOMA-IR3.22 (2.29, 4.16)6.56 (5.64, 8.03)< 0.001ᵃ***HOMA-β87.47 (50.81, 130.25)47.58 (33.79, 61.94)< 0.001ᵃ***APN13.26 (11.45, 15.98)9.53 (7.39, 12.74)< 0.001ᵃ***Notes: Data are presented as mean ± standard deviation for normally distributed variables and median (interquartile range) for non-normally distributed variables. Categorical variables are expressed as n (%). Independent-samples t test was used for age, waist circumference, and DBP; Mann–Whitney U test was used for other continuous variables; χ² test was used for sex. *P < 0.05, **P < 0.01, ***P < 0.001. ᵃ Mann–Whitney U test. ᵇ Independent-samples t testBMI, body mass index; WHR, waist-to-hip ratio; SBP, systolic blood pressure; DBP, diastolic blood pressure; FBG, fasting blood glucose; FINS, fasting insulin; TG, triglycerides; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; APN, adiponectin Table 1Baseline characteristics of participants according to insulin resistance statusVariableIR low (n = 199)IR high (n = 195)P valueGeneral characteristicsSex, n (%)Female: 117 (58.79%)Male: 82 (41.21%)Female: 92 (47.18%)Male: 103 (52.82%)0.021ᵇ*Age, years63.44 ± 8.8960.28 ± 8.760.001ᵇ**Body composition–related indicatorsBMI, kg/m²24.57 (22.67, 26.58)26.36 (25.19, 27.59)< 0.001ᵃ***Waist circumference, cm88.34 ± 9.3594.48 ± 10.810.012ᵇ*Hip circumference, cm98.08 (93.91, 105.00)94.92 (89.16, 102.40)0.018ᵃ*WHR0.89 (0.85, 0.93)0.97 (0.89, 1.09)< 0.001ᵃ***DBP, mmHg77.43 ± 7.5380.49 ± 7.720.002ᵇ**SBP, mmHg128.00 (121.00, 134.69)127.03 (119.74, 136.00)0.742ᵃGlucose, lipid, and adipokine metabolism indicatorsFBG, mmol/L6.05 (5.20, 7.80)10.08 (8.57, 11.75)< 0.001ᵃ***FINS, µIU/mL10.54 (9.54, 12.26)15.60 (13.51, 17.97)< 0.001ᵃ***TG, mmol/L1.47 (1.14, 2.05)2.48 (1.65, 3.05)< 0.001ᵃ***TC, mmol/L3.28 (1.83, 4.63)5.48 (3.91, 6.93)< 0.001ᵃ***HDL-C, mmol/L1.25 (0.97, 1.64)1.07 (0.86, 1.43)< 0.001ᵃ***LDL-C, mmol/L2.92 (2.27, 3.53)3.78 (2.88, 4.46)< 0.001ᵃ***HOMA-IR3.22 (2.29, 4.16)6.56 (5.64, 8.03)< 0.001ᵃ***HOMA-β87.47 (50.81, 130.25)47.58 (33.79, 61.94)< 0.001ᵃ***APN13.26 (11.45, 15.98)9.53 (7.39, 12.74)< 0.001ᵃ***Notes: Data are presented as mean ± standard deviation for normally distributed variables and median (interquartile range) for non-normally distributed variables. Categorical variables are expressed as n (%). Independent-samples t test was used for age, waist circumference, and DBP; Mann–Whitney U test was used for other continuous variables; χ² test was used for sex. *P < 0.05, **P < 0.01, ***P < 0.001. ᵃ Mann–Whitney U test. ᵇ Independent-samples t testBMI, body mass index; WHR, waist-to-hip ratio; SBP, systolic blood pressure; DBP, diastolic blood pressure; FBG, fasting blood glucose; FINS, fasting insulin; TG, triglycerides; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; APN, adiponectin Baseline characteristics of participants according to insulin resistance status Female: 117 (58.79%) Male: 82 (41.21%) Female: 92 (47.18%) Male: 103 (52.82%) Notes: Data are presented as mean ± standard deviation for normally distributed variables and median (interquartile range) for non-normally distributed variables. Categorical variables are expressed as n (%). Independent-samples t test was used for age, waist circumference, and DBP; Mann–Whitney U test was used for other continuous variables; χ² test was used for sex. *P < 0.05, **P < 0.01, ***P < 0.001. ᵃ Mann–Whitney U test. ᵇ Independent-samples t test BMI, body mass index; WHR, waist-to-hip ratio; SBP, systolic blood pressure; DBP, diastolic blood pressure; FBG, fasting blood glucose; FINS, fasting insulin; TG, triglycerides; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; APN, adiponectin Spearman correlation analysis between HOMA-IR and indicators related to obesity and glucolipid metabolism Spearman correlation analysis showed that HOMA-IR was significantly positively correlated with BMI, WHR, DBP, FBG, and TG, with the strongest correlation observed for FBG (r = 0.84, P < 0.001). In contrast, HOMA-IR was significantly negatively correlated with HDL-C, HOMA-β, and APN (all P < 0.001), with relatively strong negative correlations for APN (r = − 0.55, P < 0.001) and HOMA-β (r = − 0.54, P < 0.001). These correlations were integrated into the Spearman correlation heatmap and are presented in Fig. 1. Fig. 1Spearman correlation heatmap of clinical variables, including HOMA-IR and selected metabolic indicators. Only the lower triangle of the correlation matrix is displayed. Notes: Values in the cells represent Spearman correlation coefficients (r), and asterisks indicate statistical significance (*P < 0.05, **P < 0.01, ***P < 0.001). Darker shading indicates stronger correlations based on the absolute value of r Spearman correlation heatmap of clinical variables, including HOMA-IR and selected metabolic indicators. Only the lower triangle of the correlation matrix is displayed. Notes: Values in the cells represent Spearman correlation coefficients (r), and asterisks indicate statistical significance (*P < 0.05, **P < 0.01, ***P < 0.001). Darker shading indicates stronger correlations based on the absolute value of r Trends of APN and obesity indicators across HOMA-IR groups After grouping participants according to HOMA-IR tertiles, APN levels progressively decreased, whereas BMI and WHR progressively increased with increasing insulin resistance, and all differences were statistically significant. Specifically, APN decreased from 14.15 (11.97, 16.52) in the Q1 group to 8.79 (7.15, 10.96) in the Q3 group, showing a marked downward trend (P < 0.001). BMI increased from 24.35 (22.23, 26.37) in the Q1 group to 26.38 (25.44, 27.78) in the Q3 group (P = 0.001), and WHR also showed a gradual upward trend (P = 0.023). These results are presented in Table 2, and the between-group distribution is shown in Fig. 2. Table 2Trend analysis of APN, BMI, and WHR across tertiles of HOMA-IRVariableQ1Q2Q3P for trend(J–T)APN14.15 (11.97,16.52)11.81 (9.49,14.29)8.79 (7.15,10.96)< 0.001BMI24.35 (22.23,26.37)25.76 (24.05,27.07)26.38 (25.44,27.78)0.001WHR0.88 (0.85,0.91)0.91 (0.85,0.99)1.01 (0.92,1.11)0.023Notes: Data are presented as median (interquartile range). Q1–Q3 represent tertiles of HOMA-IR, where Q1 indicates the lowest insulin resistance level and Q3 indicates the highest. P for trend was calculated using the Jonckheere–Terpstra testAPN, adiponectin; BMI, body mass index; WHR, waist-to-hip ratio; HOMA-IR, homeostasis model assessment of insulin resistance Table 2Trend analysis of APN, BMI, and WHR across tertiles of HOMA-IRVariableQ1Q2Q3P for trend(J–T)APN14.15 (11.97,16.52)11.81 (9.49,14.29)8.79 (7.15,10.96)< 0.001BMI24.35 (22.23,26.37)25.76 (24.05,27.07)26.38 (25.44,27.78)0.001WHR0.88 (0.85,0.91)0.91 (0.85,0.99)1.01 (0.92,1.11)0.023Notes: Data are presented as median (interquartile range). Q1–Q3 represent tertiles of HOMA-IR, where Q1 indicates the lowest insulin resistance level and Q3 indicates the highest. P for trend was calculated using the Jonckheere–Terpstra testAPN, adiponectin; BMI, body mass index; WHR, waist-to-hip ratio; HOMA-IR, homeostasis model assessment of insulin resistance Trend analysis of APN, BMI, and WHR across tertiles of HOMA-IR Notes: Data are presented as median (interquartile range). Q1–Q3 represent tertiles of HOMA-IR, where Q1 indicates the lowest insulin resistance level and Q3 indicates the highest. P for trend was calculated using the Jonckheere–Terpstra test APN, adiponectin; BMI, body mass index; WHR, waist-to-hip ratio; HOMA-IR, homeostasis model assessment of insulin resistance Fig. 2Distribution of APN, BMI, and WHR across tertiles of HOMA-IR. Notes: Different letters (a, b, c) indicate statistically significant differences between groups (P < 0.05). The box represents the interquartile range (IQR), the horizontal line inside the box indicates the median, and the whiskers represent the range excluding outliers. Q1–Q3 represent tertiles of HOMA-IR Distribution of APN, BMI, and WHR across tertiles of HOMA-IR. Notes: Different letters (a, b, c) indicate statistically significant differences between groups (P < 0.05). The box represents the interquartile range (IQR), the horizontal line inside the box indicates the median, and the whiskers represent the range excluding outliers. Q1–Q3 represent tertiles of HOMA-IR Analysis of independent influencing factors for HOMA-IR After adjustment for age, sex, and other confounding factors, multiple linear regression analysis revealed that serum APN was significantly and negatively associated with Ln(HOMA-IR) (β = −0.335, P < 0.001), whereas BMI (β = 0.243, P < 0.001) and WHR (β = 0.138, P = 0.003) were significantly and positively associated with Ln(HOMA-IR). In addition, age was negatively associated with Ln(HOMA-IR) (β = −0.135, P = 0.001), and sex was also negatively associated with Ln(HOMA-IR) (β = −0.105, P = 0.010). APN had the largest absolute standardized regression coefficient, indicating the strongest association with Ln(HOMA-IR) (Table 3). Table 3Multivariate linear regression analysis of factors associated with Ln(HOMA-IR)VariableBSEβ95% CIP valueAPN-0.0420.006-0.335(-0.054, -0.030)< 0.001BMI0.0470.0080.243(0.030, 0.063)< 0.001WHR0.5210.1750.138(0.178, 0.865)0.003Age-0.0080.002-0.135(-0.012, -0.003)0.001Sex-0.1060.041-0.105(-0.186, -0.026)0.010Notes: B represents the unstandardized regression coefficient; SE represents the standard error; β represents the standardized regression coefficient; 95% CI indicates the 95% confidence interval; P value indicates statistical significance. The model was adjusted for age, sex, BMI, WHR, and APN Table 3Multivariate linear regression analysis of factors associated with Ln(HOMA-IR)VariableBSEβ95% CIP valueAPN-0.0420.006-0.335(-0.054, -0.030)< 0.001BMI0.0470.0080.243(0.030, 0.063)< 0.001WHR0.5210.1750.138(0.178, 0.865)0.003Age-0.0080.002-0.135(-0.012, -0.003)0.001Sex-0.1060.041-0.105(-0.186, -0.026)0.010Notes: B represents the unstandardized regression coefficient; SE represents the standard error; β represents the standardized regression coefficient; 95% CI indicates the 95% confidence interval; P value indicates statistical significance. The model was adjusted for age, sex, BMI, WHR, and APN Multivariate linear regression analysis of factors associated with Ln(HOMA-IR) Notes: B represents the unstandardized regression coefficient; SE represents the standard error; β represents the standardized regression coefficient; 95% CI indicates the 95% confidence interval; P value indicates statistical significance. The model was adjusted for age, sex, BMI, WHR, and APN Multivariable logistic regression analysis of insulin resistance status A multivariable logistic regression analysis was conducted, with the insulin resistance status (IR group) designated as the dependent variable. Following adjustment for age, sex, and other potential confounding factors, APN, BMI, and WHR were determined to be independent predictors of insulin resistance status. Specifically, elevated APN levels were associated with a reduced risk of high insulin resistance status (odds ratio [OR] = 0.79, P < 0.001), while higher BMI (OR = 1.18, P = 0.001) and WHR (OR = 13.21, P = 0.015) were associated with an increased risk of high insulin resistance status. Additionally, sex was also identified as an independent predictor (OR = 1.75, P = 0.022), whereas age was not significantly associated with insulin resistance status (P = 0.14). The detailed results are presented in Table 4. Table 4Logistic regression analysis for insulin resistance (IR group)VariableBOR (Exp(B))95% CIP valueAge-0.020.98(0.95, 1.01)0.14BMI0.171.18(1.07, 1.31)0.001**WHR2.5813.21(1.64, 106.41)0.015*APN-0.240.79(0.73, 0.85)< 0.001***Sex0.561.75(1.08, 2.82)0.022*Notes: B, regression coefficient; OR, odds ratio; 95% CI, 95% confidence interval; P value, statistical significance. *P < 0.05, **P < 0.01, ***P < 0.001 Table 4Logistic regression analysis for insulin resistance (IR group)VariableBOR (Exp(B))95% CIP valueAge-0.020.98(0.95, 1.01)0.14BMI0.171.18(1.07, 1.31)0.001**WHR2.5813.21(1.64, 106.41)0.015*APN-0.240.79(0.73, 0.85)< 0.001***Sex0.561.75(1.08, 2.82)0.022*Notes: B, regression coefficient; OR, odds ratio; 95% CI, 95% confidence interval; P value, statistical significance. *P < 0.05, **P < 0.01, ***P < 0.001 Logistic regression analysis for insulin resistance (IR group) Notes: B, regression coefficient; OR, odds ratio; 95% CI, 95% confidence interval; P value, statistical significance. *P < 0.05, **P < 0.01, ***P < 0.001 Discriminative performance for IR status (ROC analysis) To further investigate the discriminatory capacity of each indicator for IR status, we next conducted receiver operating characteristic (ROC) curve analysis. WHR, BMI, and APN all exhibited a certain level of discriminatory ability. Of these, APN exhibited the best performance (area under the curve [AUC] = 0.78), followed by WHR (AUC = 0.72) and BMI (AUC = 0.69). Of the single indicators, BMI exhibited relatively high sensitivity (82.56%), while APN exhibited relatively high specificity (80.40%). The construction of a combined model incorporating age, sex, BMI, WHR, and APN led to enhanced discriminatory performance (AUC = 0.82), with relatively high specificity (86.43%) (Table 5; Fig. 3). DeLong tests further indicated that the AUC of the combined model was significantly higher than that of WHR (ΔAUC = 0.10, P < 0.001), BMI (ΔAUC = 0.12, P < 0.001), and APN (ΔAUC = 0.03, P = 0.01) (Table 6). Table 5Diagnostic performance of WHR, BMI, APN, and the combined model for identifying insulin resistanceVariableAUC95% CICut-offSensitivity (%)Specificity (%)WHR0.72(0.67, 0.77)0.9268.7269.35BMI0.69(0.64, 0.74)24.9282.5655.78APN0.78(0.74, 0.83)10.9867.6980.40Combined model0.82(0.77, 0.86)0.6066.1586.43Notes: AUC, area under the receiver operating characteristic curve; 95% CI, 95% confidence interval. The optimal cut-off value was determined using the Youden index. For APN, the cut-off is presented on the original scale, while discrimination was calculated using the inverse direction because lower APN indicates higher insulin resistance risk Table 5Diagnostic performance of WHR, BMI, APN, and the combined model for identifying insulin resistanceVariableAUC95% CICut-offSensitivity (%)Specificity (%)WHR0.72(0.67, 0.77)0.9268.7269.35BMI0.69(0.64, 0.74)24.9282.5655.78APN0.78(0.74, 0.83)10.9867.6980.40Combined model0.82(0.77, 0.86)0.6066.1586.43Notes: AUC, area under the receiver operating characteristic curve; 95% CI, 95% confidence interval. The optimal cut-off value was determined using the Youden index. For APN, the cut-off is presented on the original scale, while discrimination was calculated using the inverse direction because lower APN indicates higher insulin resistance risk Diagnostic performance of WHR, BMI, APN, and the combined model for identifying insulin resistance Notes: AUC, area under the receiver operating characteristic curve; 95% CI, 95% confidence interval. The optimal cut-off value was determined using the Youden index. For APN, the cut-off is presented on the original scale, while discrimination was calculated using the inverse direction because lower APN indicates higher insulin resistance risk Fig. 3Receiver operating characteristic (ROC) curves of WHR, BMI, APN, and the combined model for identifying insulin resistance. Notes: The diagonal dotted line represents the line of no discrimination (AUC = 0.5). The y-axis indicates sensitivity, and the x-axis represents 1 − specificity Receiver operating characteristic (ROC) curves of WHR, BMI, APN, and the combined model for identifying insulin resistance. Notes: The diagonal dotted line represents the line of no discrimination (AUC = 0.5). The y-axis indicates sensitivity, and the x-axis represents 1 − specificity Table 6Comparison of AUCs using DeLong TestComparisonΔAUC95% CIP valueCombined vs. WHR0.10(0.05–0.15)< 0.001Combined vs. BMI0.12(0.07–0.17)< 0.001Combined vs. APN0.03(0.01–0.06)0.01Notes: ΔAUC represents the difference in area under the curve between models. 95% CI indicates the 95% confidence interval. Differences in AUCs were compared using the DeLong testNotes: The diagonal dotted line represents the line of no discrimination (AUC = 0.5). The y-axis indicates sensitivity, and the x-axis represents 1 − specificity Table 6Comparison of AUCs using DeLong TestComparisonΔAUC95% CIP valueCombined vs. WHR0.10(0.05–0.15)< 0.001Combined vs. BMI0.12(0.07–0.17)< 0.001Combined vs. APN0.03(0.01–0.06)0.01Notes: ΔAUC represents the difference in area under the curve between models. 95% CI indicates the 95% confidence interval. Differences in AUCs were compared using the DeLong testNotes: The diagonal dotted line represents the line of no discrimination (AUC = 0.5). The y-axis indicates sensitivity, and the x-axis represents 1 − specificity Comparison of AUCs using DeLong Test Notes: ΔAUC represents the difference in area under the curve between models. 95% CI indicates the 95% confidence interval. Differences in AUCs were compared using the DeLong test Notes: The diagonal dotted line represents the line of no discrimination (AUC = 0.5). The y-axis indicates sensitivity, and the x-axis represents 1 − specificity Discussion Over recent years, the global burden of T2DM has continued to increase, accompanied by a rising prevalence in excessive body weight, obesity, and central obesity among affected individuals [18, 19]. Even in patients with normal body weight, abnormal fat distribution and metabolic disturbances are not uncommon [20]. Given that IR is a central pathophysiological mechanism linking obesity to multiple metabolic abnormalities [21], we stratified our patients according to the degree of IR and investigated the patterns of change in relevant indicators across different metabolic states. Our analysis showed that the high-IR group had significantly higher BMI, WHR, DBP, FBG, and TG levels than the low-IR group, thus suggesting that worsening IR is closely accompanied by increased adiposity and more pronounced metabolic disturbance [22]. These findings are consistent with previous studies suggesting that reduced insulin sensitivity is closely related to obesity and glucolipid metabolic abnormalities [23, 24]. Visceral adiposity may impair insulin signaling through abnormal lipid metabolism, lipotoxicity, and low-grade inflammation [24, 25]. Enhanced triglyceride synthesis and the accumulation of lipid-derived metabolites, including free fatty acids (FFAs) and diacylglycerol (DAG), can disrupt insulin signaling, activate NADPH oxidase and NF-κB pathways, and subsequently induce oxidative stress and inflammation [26–28]. In addition, under physiological conditions, insulin may exert vasodilatory and anti-inflammatory effects through the activation of endothelial vasodilation pathways; therefore, IR may blunt these protective vascular effects and contribute to elevated blood pressure and low-grade inflammation [29]. IR may also increase circulating TG levels by decreasing lipoprotein lipase (LPL) activity, thereby impairing the clearance of triglyceride-rich lipoproteins [30]. These mechanisms may partly explain the observed positive associations of HOMA-IR with BMI, WHR, DBP, and TG in the present study. To further clarify these associations, we next performed correlation and trend analyses. We found that APN was negatively correlated with HOMA-IR, whereas BMI, WHR, FBG, and TG were positively correlated with HOMA-IR. These data indicated that adiponectin level, overall adiposity, central fat distribution, and glucolipid metabolic abnormalities vary in parallel with IR. These findings are consistent with the study by Yang et al., who reported that serum adiponectin levels were negatively associated with insulin resistance in women with polycystic ovary syndrome [31]. Although the study population differed from ours, both studies support an inverse relationship between adiponectin and HOMA-IR-related metabolic dysfunction. In addition, Stojanovic et al. suggested that adiponectin-related parameters were associated with blood pressure status in patients with metabolic syndrome, further supporting the link between adipokine dysregulation and cardiometabolic abnormalities [32]. Similarly, an elevated BMI and an abnormal central fat distribution have been recognized as important correlates of IR [33]. Our trend analysis further revealed that APN decreased progressively across rising HOMA-IR tertiles, whereas BMI and WHR increased progressively. These data support the view that IR is not merely a binary state but rather a continuous metabolic abnormality with graded phenotypic expression [34]. Several inter-related pathophysiological mechanisms may underlie these observations. Adiponectin is an important endogenous insulin-sensitizing adipokine that enhances peripheral glucose uptake and lipid oxidation via several pathways, including the activation of AMP-activated protein kinase, thereby improving insulin sensitivity [35, 36]. When APN levels decline, the efficiency of insulin signaling is correspondingly impaired. Furthermore, overall obesity, particularly central fat accumulation, may aggravate IR through ectopic lipid deposition, enhanced triglyceride synthesis, and the accumulation of lipid-derived metabolites such as FFAs and DAG, together with chronic low-grade inflammation and adipokine imbalance [26–28, 37]. These processes interact with one another and may form a vicious cycle in which reduced adiponectin and adipose tissue dysfunction jointly promote the progression of IR. Previous studies have also provided support for these mechanisms from the perspectives of inflammatory activation and adipokine dysregulation [38]. Based on our correlation and trend analyses, we next utilized multiple linear regression to evaluate independent associations between each factor and the severity of IR. After adjustment for confounding factors, we found that APN was independently and negatively associated with Ln(HOMA-IR), whereas BMI and WHR were independently and positively associated. Collectively, these results indicate that both adipokine status and body fat distribution are independently related to IR. Notably, APN showed the largest absolute standardized regression coefficient, suggesting that APN may have had a more prominent contribution within the multivariable model. Unlike BMI and WHR, which predominantly reflect overall adiposity and the distribution of fat, APN more directly reflects adipose endocrine function and insulin sensitivity. APN may therefore be more closely linked to the core metabolic abnormalities underlying IR [39, 40]. Adiponectin is closely linked to insulin sensitivity through its effects on insulin signaling, carbohydrate metabolism, and inflammation [35, 36, 41]. It has been reported to activate AMP-activated protein kinase and related downstream pathways, thereby promoting glucose uptake, fatty acid oxidation, and improved metabolic homeostasis [35, 36]. In addition, its anti-inflammatory effects may attenuate obesity-related low-grade inflammation, which is closely associated with impaired insulin signaling [38, 42]. These mechanisms may partly explain why APN showed a stronger standardized effect than anthropometric indices alone and may also help explain the negative association between APN and HOMA-IR observed in the present study. To further assess the associations of these indicators with IR status, we next performed logistic regression analysis. After adjustment for confounding factors, we found that higher APN levels were associated with a lower risk of a high IR status, whereas a higher BMI and WHR were associated with an increased risk. Although the OR value for WHR was relatively high, WHR is a continuous variable with a narrow numerical range; its effect estimate may therefore appear magnified in regression models. Accordingly, the magnitude of WHR should be interpreted with caution and in the context of its scale characteristics. Nevertheless, our present findings still support a significant association between WHR and IR status. These results are supported by the review by Han et al., which summarized recent advances regarding the role of adiponectin in obesity and related metabolic syndromes [10]. In addition, Tchernof and Després described the pathophysiological contribution of visceral obesity to metabolic dysfunction, supporting the relevance of WHR as an indicator of central fat distribution in IR-related metabolic risk [43]. Collectively, these findings suggest that a combined assessment, incorporating adipokines and anthropometric indices, may improve the identification of individuals at high metabolic risk. Next, we performed ROC curve analysis to evaluate the discriminative performance of each indicator and the combined model for IR status. Of the single indicators, APN showed relatively favorable discriminative ability, whereas BMI and WHR also exhibited moderate discriminative value. After integration into a combined model, we found that the AUC increased significantly, thus indicating improved discriminative performance when compared with any individual indicator alone. The DeLong test further confirmed the superiority of the combined model over each single indicator. ROC and AUC analysis are established tools for evaluating model discrimination [44] and have been widely applied in studies of metabolic disease [45, 46]. Zahid et al. evaluated biochemical, anthropometric, and combined indices for metabolic syndrome prediction and found that integrated approaches may improve risk identification compared with single indicators [47]. Similarly, Park et al. reported that anthropometric indicators had predictive value for metabolic syndrome in Korean adults, further supporting the usefulness of body composition-related measures in metabolic risk assessment [48]. Our results extend this evidence by suggesting that the integration of adipokine and anthropometric information may capture metabolic heterogeneity more comprehensively and improve the identification of individuals with high IR. In current clinical practice, lifestyle and exercise interventions for patients with T2DM are often individualized according to broad clinical characteristics such as obesity, cardiovascular disease, and other comorbidities [49]. Although this approach is practical, it predominantly relies on external phenotypes and may not fully capture the underlying metabolic heterogeneity among patients. Given that APN, BMI, and WHR were all independently associated with IR and that the combined model exhibited better discriminative performance, our findings suggest that metabolic stratification beyond conventional clinical phenotypes may be helpful in identifying subgroups of patients with different degrees of adiposity-related metabolic impairment. Kanaley et al. emphasized the importance of exercise and physical activity in the management of patients with T2DM [49]. In addition, Mateo-Gallego et al. reported that high-intensity interval training improved glucose metabolism, cardiorespiratory fitness, and weight control in individuals with diabetes [50], whereas Wang et al. found that resistance training enhanced metabolic and muscular health and reduced systemic inflammation in middle-aged and older adults with T2DM [51]. Moreover, recent systematic reviews and meta-analyses have suggested that exercise may further reduce cardiometabolic risk by modulating inflammatory adipokines, attenuating chronic low-grade inflammation, and optimizing body composition and fat distribution [52, 53]. However, because the present study was retrospective and observational, we cannot directly determine whether this model improves exercise prescription or intervention outcomes. Rather, our findings suggest that the combined assessment of APN, BMI, and WHR may provide a potential basis for future studies to explore stratified lifestyle or exercise interventions in patients with T2DM, particularly those with overweight, obesity, or central adiposity. This study has several limitations that need to be considered. First, because of its retrospective observational design, causal relationships could not be established. In addition, detailed information on statin therapy, including statin type, dose, and treatment duration, was not consistently available in the medical records. Because statins are commonly prescribed to patients with T2DM and may influence lipid profiles and adiponectin levels, their potential confounding effects could not be fully evaluated in this retrospective study. Therefore, residual confounding related to medication use, particularly statin therapy, cannot be excluded. Second, our study cohort was derived from a single center; this may limit the generalizability of our findings. Third, direct measurements of visceral fat and skeletal muscle mass were not available. Fourth, the practical efficacy of stratified exercise interventions was not evaluated. Therefore, future multicenter prospective studies, ideally with interventional follow-up, are now needed to further validate the clinical utility of metabolic stratification strategies based on adipokine and anthropometric indicators. Conclusion Analyses revealed that APN, BMI, and WHR were closely associated with IR in patients with T2DM. APN was independently and negatively associated with IR, whereas BMI and WHR were independently and positively associated with IR. Our combined model demonstrated superior performance in identifying IR when compared with individual indicators, thus suggesting that the integration of adipokines with anthropometric indices may more comprehensively reflect metabolic abnormalities and improve risk identification. Collectively, these findings may help to identify metabolically distinct subgroups of patients with T2DM and provide a potential basis for risk stratification. However, their clinical applicability requires further validation in multicenter prospective studies.