A prediction nomogram for mild cognitive impairment in type 2 diabetes mellitus based on the Chinese visceral adiposity index

Abstract

Visceral adiposity has been proposed to be closely linked to cognitive impairment. This cross-sectional study aimed to evaluate the predictive value of Chinese Visceral Adiposity Index (CVAI) for mild cognitive impairment (MCI) in patients with type 2 diabetes mellitus (T2DM) and to develop a quantitative risk assessment model. A total of 337 hospitalized patients with T2DM were included and randomly assigned to a training cohort (70%, n = 236) and a validation cohort (30%, n = 101). Demographic, clinical, and neuropsychological data were collected. CVAI levels were compared between patients with MCI and those with normal cognition. Associations between CVAI and cognitive performance were assessed using Spearman correlation and multivariable linear regression. Predictors of MCI were identified through Lasso regression followed by univariate and multivariate logistic regression analyses. A nomogram incorporating age, gender, education level, and CVAI was constructed and validated using calibration plots, ROC curve analysis, and decision curve analysis (DCA). Patients with MCI exhibited significantly higher CVAI values and lower MoCA and MMSE scores compared to those with normal cognition (all P < 0.001). CVAI was independently and negatively associated with MoCA and MMSE scores (β = -0.22, P < 0.001 for both) after adjustment. Multivariate logistic regression confirmed CVAI as an independent risk factor for MCI (P = 0.002). The nomogram demonstrated good discrimination, with an AUC of 0.765 in the training cohort and 0.690 in the validation cohort, and exhibited favorable clinical utility based on DCA. These findings suggest that CVAI is a valuable biomarker for the early identification and risk stratification of MCI in T2DM, and that the CVAI-based nomogram provides a practical tool for individualized clinical decision-making.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-17241-5.

Introduction

China has the largest number of people with diabetes in the world, with the figure projected to reach 164 million by 2030 and 175 million by 2045¹, and type 2 diabetes mellitus (T2DM) accounts for more than 90% of the cases^2,3. Cognitive impairment, as a complication of T2DM, has garnered increasing attention due to its profound impact on diabetes management and overall quality of life. The available clinical and epidemiological evidence suggests that individuals with T2DM face an elevated risk of developing mild cognitive impairment(MCI) and Alzheimer’s disease(AD)⁴. Studies have documented that the risk of cognitive decline in diabetic patients was 2–3 times greater than that in individuals without diabetes^5,6. The dysfunction of visceral adipose tissue contributes to the pathogenesis of insulin resistance and T2DM⁷. An epidemiological and Mendelian Randomization study has suggested that visceral adiposity may be causally linked to cognition⁸.

The Visceral Adiposity Index (VAI) is primarily applicable to European and American populations due to variations in body fat distribution among different ethnicities. In order to overcome this limitation, the Chinese Visceral Adiposity Index (CVAI) was developed and validated as a convenient and reliable tool for assessing visceral fat dysfunction specifically in Chinese individuals^9,10. The nomogram, an intuitive statistical tool, is widely used in clinical prediction models for accurate risk assessment and clinician-friendly decision support. While prior studies have examined the link between visceral obesity and cognitive decline, variations in sample composition, follow-up duration, and analytical methods limit comparability. Moreover, most research has focused on elderly or non-diabetic populations, limiting its relevance to individuals with diabetes.

This study aims to apply CVAI to T2DM patients and assess its correlation with diabetes-related cognitive impairment. How to quantitatively evaluate the risk of MCI in T2DM using widely applicable, non-invasive, and cost-effective assessments has been rarely addressed in previous studies. How to quantitatively evaluate the risk of MCI in T2DM is rarely involved in previous studies. Therefore, a nomogram model based on CVAI was constructed and validated to quantitatively evaluate the risk of MCI in T2DM for providing more intuitive and faster evaluation methods for patients and clinicians.

Materials and methods

Ethical approval

This cross-sectional study was conducted in accordance with the guidelines outlined in the Declaration of Helsinki and obtained approval from the Medical Ethics Committee at Affiliated Zhongda Hospital of Southeast University (approval No. 2023ZDSYLL435-P01). All individuals involved in the experiment were fully informed of the study procedures and provided written informed consent.

Study population

Patients with T2DM who were admitted to the Department of Endocrinology at the Affiliated Zhongda Hospital of Southeast University between November 2021 and April 2023 were consecutively enrolled in this study. The inclusion criteria were as follows: (1) age ranging from 40 to 75 years; (2) T2DM diagnosis according to the World Health Organization’s standards established in 1999, with a duration of at least three years (to avoid reverse causality bias, i.e., cognitive impairment causing diabetes); (3) all patients having received education for a minimum of six years. The exclusion criteria were as follows: (1)any neurological disease that causes dementia, including AD, Parkinson’s disease, Huntington’s disease, normal pressure hydrocephalus, and brain tumors; (2) previous or current significant psychiatric disorder as well as any psychotropic medications; (3) acute and severe chronic diabetes complications, such as severe episodes of hypoglycemia, diabetic ketoacidosis, hyperosmolar nonketotic diabetic coma; (4) severe liver and kidney dysfunction; (5)other organic and mental diseases that may lead to brain dysfunction, such as acute cardiovascular or cerebrovascular events, cancer, thyroid dysfunction, or serious infections; (6) severe visual or hearing impairment. According to these criteria, a total of 337 eligible patients were included in the study. Based on their Montreal Cognitive Assessment (MoCA) scores, participants were categorized into two groups: the normal cognition (NC) group (n = 179) and the mild cognitive impairment (MCI) group (n = 158).

Data collection

Demographic data, including age, gender, height, weight, waist circumference (WC), duration of diabetes mellitus (DM), educational level, and the presence of hypertension, were collected upon admission. Fasting venous blood samples were collected to assess the levels of fasting plasma glucose (FPG), hemoglobin A1c (HbA1c), fasting C-peptide (FCP), total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), alanine aminotransferase (ALT), aspartate aminotransferase (AST), blood urea nitrogen (BUN), creatinine (Cr), and other biochemical parameters.

Variable calculation formula

The calculation formula is as follows:

BMI = weight (kg)/[height (m)]²¹¹,

CVAI(males) = − 267.93 + 0.68 × age + 0.03 × BMI + 4.00 × WC + 22.00 × lgTG − 16.32 × HDL and CVAI(females) = − 187.32 + 1.71 × age + 4.23 × BMI + 1.12 × WC + 39.76 ×lgTG − 11.66 × HDL⁹.

Neuropsychological tests

The Montreal Cognitive Assessment (MoCA) was administered to evaluate cognitive impairment in study participants, encompassing various domains including attention, concentration, executive functions, memory, language abilities, visuoconstructional skills, conceptual thinking, calculations, and orientation, while the assessment scores range from 0 to 30¹². MoCA score ≥ 26 is considered normal (NC group) and MoCA score <26 is considered MCI(MCI group)¹³. If the informed years of education were less than 12 years, an adjustment score was added to mitigate potential bias associated with educational level¹³. The cognitive function was also evaluated by the Mini-Mental State Examination (MMSE)^14,15. The CDT was conducted to analyze visual spatial function in accordance with the protocol of a previous study¹⁶. DST, VFT, TMTA, and TMTB were performed based on established methodologies to assess executive functions and information processing speed^17–19. AVLT-IR and AVLT-DR were administered to examine immediate and delayed memory function²⁰while LMT was utilized for measuring scene memory function²¹.

Statistical analysis

Statistical analyses were conducted by SPSS 24.0(IBM Inc., Chicago, IL, USA) and R(4.2.1). Sample size calculations were performed using PASS 15 (NCSS, LLC, Kaysville, UT). Random stratified sampling was employed to divide the data into training and validation sets, minimizing sampling bias.

Continuous variables were presented as mean ± standard deviation (SD) or median with interquartile range (IQR), as appropriate. Group comparisons were conducted using the independent sample t-test (for normally distributed variables), the Mann–Whitney U test (for non-normally distributed variables), and the chi-squared test (for categorical variables). Spearman’s rank correlation analysis was used to examine the correlations between variables. Multivariable linear regression analysis was performed to investigate the relationship between CVAI and cognitive function. LASSO regression, along with univariate and multivariate logistic regression analyses, was performed to identify risk and protective factors associated with MCI. A nomogram was constructed based on LASSO and logistic regression analyses to provide a visual representation of the predictive model. Model calibration was evaluated using 500 bootstrap resampling, with calibration curves compared to the ideal 45° reference line²². The discriminative accuracy was assessed using a receiver operating characteristic curve (ROC), while the classification performance was evaluated by calculating the area under the curve (AUC)²³. The clinical utility of the nomogram was evaluated through decision curve analysis (DCA)²⁴. All statistical tests were two-tailed, and a p-value < 0.05 was considered statistically significant.

Results

Baseline characteristics of study population

A total of 337 participants were included in this study for model development. Based on a 7:3 ratio, 236 participants were allocated to the training set and 101 participants to the validation set. The incidence of MCI in the training and validation cohorts was 46.2% (109/236) and 48.5% (49/101), respectively, with no statistically significant difference. There were no significant statistical differences between the training and validation sets in terms of demographic and clinical characteristics, including age, gender, years of education, hypertension, biochemical indicators, glycemic indices, and cognitive function tests (P > 0.05, see Supplementary Table 1 for details).

To explore the risk of cognitive decline in T2DM patients, we compared demographic characteristics, clinical features, and neuropsychometric test outcomes between the MCI and NC groups in both the training and validation cohorts (Table 1). Compared to the NC group, patients in the MCI group were more likely to be female and older, with a significantly longer diabetes duration, higher CVAI, lower educational levels, and a higher prevalence of hypertension (P < 0.05). However, no significant differences were observed between the two groups in BMI, HbA1c, FCP, TG, TC, HDL, LDL, Cr, BUN, ALT, or AST (P > 0.05). Since this study included patients with recent poor glycemic control requiring hospitalization, differences in HbA1c and FCP did not reach statistical significance. Additionally, the results demonstrated a significant decline in MoCA, MMSE, DST, VFT, CDT, AVLT-IR, AVLT-DR, and LMT scores (P < 0.05) in the MCI group, accompanied by prolonged completion times for TMTA and TMTB.

Table 1.

Demographic, clinical and cognitive characteristics of T2DM patients in the training and validation cohorts.

Characteristic	Training cohort			Validation cohort
	NC, N = 127	MCI, N = 109	P value	NC, N = 52	MCI, N = 49	P value
Age(year)	57 (51, 64)	64 (59, 69)	< 0.001	60 (52, 66)	63 (56, 69)	0.03
Gender, n (%)
Female	33 (26%)	49 (45%)	0.002	15 (29%)	17 (35%)	0.528
Male	94 (74%)	60 (55%)		37 (71%)	32 (65%)
Education(year)	12 (12, 15)	11 (9, 12)	< 0.001	12 (9, 15)	12 (9, 12)	0.012
DM Duration (year)	10 (5, 15)	12 (8, 20)	0.006	10 (5, 17)	10 (7, 17)	0.56
HTN, n (%)
No	69 (54%)	40 (37%)	0.007	24 (46%)	19 (39%)	0.454
Yes	58 (46%)	69 (63%)		28 (54%)	30 (61%)
HTN Duration (year)	0 (0, 10)	5 (0, 10)	0.05	5 (0, 10)	5 (0, 20)	0.809
Smoking history, n (%)
No	80 (63%)	75 (69%)	0.348	34 (65%)	40 (82%)	0.065
Yes	47 (37%)	34 (31%)		18 (35%)	9 (18%)
Alcohol use, n (%)
No	103 (81%)	86 (79%)	0.673	41 (79%)	42 (86%)	0.367
Yes	24 (19%)	23 (21%)		11 (21%)	7 (14%)
BMI	24.9 (23.0, 27.2)	24.6 (22.6, 26.7)	0.673	24.35 (23.08, 26.54)	24.10 (22.80, 26.40)	0.912
CVAI	126 (98, 155)	143 (114, 168)	< 0.001	120 (95, 140)	138 (109, 167)	0.005
HbA1c(%)	8.12 (7.17, 9.50)	8.60 (7.60, 9.68)	0.097	8.50 (7.26, 9.83)	8.44 (7.56, 9.89)	0.82
FBG(mmol/L)	6.91 (5.46, 8.73)	7.19 (6.13, 8.58)	0.713	7.73 (6.54, 9.40)	6.33 (5.61, 8.11)	0.007
FCP(nmol/L)	0.54 (0.39, 0.71)	0.50 (0.32, 0.74)	0.369	0.49 (0.37, 0.77)	0.54 (0.39, 0.62)	0.465
TG(mmol/L)	1.44 (1.00, 1.98)	1.33 (0.99, 1.98)	0.355	1.44 (1.04, 2.16)	1.20 (0.98, 1.54)	0.969
TC(mmol/L)	4.18 ± 1.08	4.33 ± 1.21	0.307	4.40 ± 0.91	4.46 ± 1.07	0.765
HDL(mmol/L)	1.00 (0.86, 1.20)	1.01 (0.89, 1.20)	0.787	1.00 (0.88, 1.19)	0.98 (0.89, 1.21)	0.507
LDL(mmol/L)	2.34 (1.76, 2.81)	2.30 (1.78, 3.09)	0.587	2.60 (2.15, 3.02)	2.39 (1.93, 3.30)	0.32
Cr(µmol/L)	70 (58, 84)	68 (56, 87)	0.254	68 (59, 79)	68 (56, 80)	0.302
BUN(mmol/L)	6.30 (5.60, 7.40)	6.60 (5.70, 7.80)	0.918	5.95 (5.00, 7.24)	6.40 (5.40, 7.30)	0.308
UA(µmol/L)	303 (252, 371)	300 (245, 369)	0.607	318 (265, 373)	272 (230, 356)	0.337
ALT(U/L)	17 (12, 22)	16 (12, 21)	0.246	18 (13, 26)	16 (12, 23)	0.931
AST(U/L)	17 (14, 20)	16 (13, 21)	0.139	17.0 (14.0, 21.0)	15.0 (14.0, 20.0)	0.461
ALP(U/L)	70 (59, 84)	72 (59, 87)	0.713	75 (66, 88)	75 (63, 84)	0.691
GGT(U/L)	20 (15, 32)	19 (16, 34)	0.669	23 (16, 34)	18 (14, 29)	0.489
MoCA	28.00 (27.00, 29.00)	24.00 (23.00, 25.00)	< 0.001	28.00 (27.00, 29.00)	24.00 (22.00, 25.00)	< 0.001
MMSE	29.00 (28.00, 30.00)	27.00 (26.00, 28.00)	< 0.001	29.00 (27.00, 30.00)	26.00 (25.00, 28.00)	< 0.001
DST	12.3 (11.0, 14.0)	11.0 (9.0, 13.0)	< 0.001	12.00 (11.00, 14.00)	11.00 (8.00, 12.00)	< 0.001
VFT	18.0 (16.0, 21.0)	16.0 (14.0, 19.0)	< 0.001	18.0 (15.0, 20.0)	15.0 (14.0, 18.0)	0.002
CDT	4.00 (3.00, 4.00)	3.00 (2.00, 4.00)	< 0.001	3.00 (3.00, 4.00)	3.00 (2.00, 4.00)	0.027
TMTA	59 (45, 75)	69 (54, 88)	< 0.001	64 (51, 81)	66 (57, 80)	0.465
TMTB	136 (110, 166)	168 (123, 204)	< 0.001	145 (113, 175)	167 (126, 202)	0.01
AVLT-IR	17.0 (14.0, 21.0)	14.0 (11.0, 18.0)	< 0.001	17.4 (13.8, 20.3)	14.0 (11.0, 18.0)	0.035
AVLT-DR	6.0 (3.0, 8.0)	5.0 (3.0, 7.0)	0.03	6.5 (4.0, 8.0)	3.0 (2.0, 6.0)	< 0.001
LMT	6.0 (4.0, 9.0)	4.0 (3.0, 8.0)	0.047	5.0 (3.0, 9.0)	4.5 (2.0, 6.0)	0.223

Abbreviations: BMI Body mass index, HTN Hypertension, DM Diabetes mellitus, CVAI Chinese visceral adiposity index, HbA1c, hemoglobin A1c, FPG Fasting plasma glucose, FCP Fasting C-Peptide, TG Triglycerides, TC Total cholesterol, LDL Low density lipoprotein, HDL High density lipoprotein, Cr creatinine, BUN blood urea nitrogen, UA Uric Acid, ALT Alaninetransaminase, AST Aspartate aminotransferase, ALP Alkaline Phosphatase, GGT Gamma-Glutamyl Transferase, MoCA Montreal Cognitive Assessment, MMSE Mini-mental State Examination, DST Digit Span Test, VFT Verbal Fluency Test, CDT Clock Drawing Test, TMTA Trail Making Test-A, TMTB Trail Making Test-B, AVLT-IR, Auditory Verbal Learning test- immediate recall; AVLT-DR, Auditory Verbal Learning test-delayed recall; LMT, logical memory test.

Association between CVAI and cognitive function

The results of the study showed that CVAI was enriched considerably in MCI group of T2DM patients, so the relationship between CVAI and neuropsychometric test results was conducted.

CVAI was negatively associated with MoCA, MMSE and AVLT-DR (P < 0.001), whereas positively associated with TMTA in the MCI group (P < 0.05) by spearman correlation analyses(Fig. 1).

Fig. 1.

Association between CVAI and cognitive function.

Furthermore, stepwise multivariable linear regression was performed with MoCA and MMSE as dependent variables to identify factors predicting cognitive function (Table 2). Standardized CVAI analysis indicated that each standard deviation increase in CVAI was associated with a 0.22–0.25 point decrease in MoCA scores (Model 3: β = − 0.22, 95% CI: − 0.35 to − 0.10, P < 0.001). When categorized, compared to the low-level CVAI group, the middle-level group exhibited a significant decline in MoCA scores in Model 1 (β = − 1.00, 95% CI: − 1.84 to − 0.16, P = 0.021) and Model 3 (β = − 0.90, 95% CI: − 1.72 to − 0.08, P = 0.033). The high-level group showed a more pronounced decline across all models (Model 3: β = − 1.08, 95% CI: − 1.91 to − 0.26, P = 0.011). Trend analysis further supported a dose-response relationship between CVAI and MoCA decline (P < 0.05). Similar trends were observed for MMSE. After adjusting for potential confounders in three sequential models with increasing levels of covariate adjustment, CVAI remained an independent predictor of cognitive decline, as assessed by MoCA and MMSE.

Table 2.

Multivariable linear regression analysis of MoCA/MMSE across CVAI tertiles in T2DM.

Variable	Characteristic	Model 1				Model 2				Model 3
		N	Beta	95% CI1	P value	N	Beta	95% CI1	P value	N	Beta	95% CI1	P value
MoCA	CVAI (per-SD)	236	− 0.25	− 0.37, − 0.12	< 0.001	236	− 0.23	− 0.35, − 0.11	< 0.001	236	− 0.22	− 0.35, − 0.10	< 0.001
	CVAI
	Low (< 114)	79	–	–		79	–	–		79	–	–
	Medium (114 ~ 150)	78	− 1	− 1.84, − 0.16	0.021	78	− 0.5	− 1.29, 0.29	0.219	78	− 0.9	− 1.72, − 0.08	0.033
	High(> 150)	79	− 1.27	− 2.10, − 0.43	0.003	79	− 0.92	− 1.72, − 0.11	0.027	79	− 1.08	− 1.91, − 0.26	0.011
	P for trend				0.003				0.027				0.011
MMSE	CVAI (per-SD)	236	− 0.26	− 0.38, − 0.13	< 0.001	236	− 0.2	− 0.32, − 0.08	< 0.001	236	− 0.22	− 0.34, − 0.10	< 0.001
	CVAI
	Low (< 114)	79	–	–		79	–	–		79	–	–
	Medium (114 ~ 150)	78	− 0.79	− 1.36, − 0.22	0.007	78	− 0.4	− 0.94, 0.13	0.142	78	− 0.7	− 1.25, − 0.14	0.014
	High(> 150)	79	− 0.94	− 1.50, − 0.37	0.001	79	− 0.58	− 1.13, − 0.04	0.037	79	− 0.76	− 1.32, − 0.20	0.008
	P for trend				0.001				0.037				0.008

¹CI = Confidence Interval, SD = 45.

Model 1 : No covariates were adjusted.

Model 2 : Adjusted for Age.year., Gender, and Education.year.

Model 3 :Additionally adjusted for DM.Duration.year. and HTN.

LASSO and logistic regression analyses of risk factors influencing MCI

Baseline characteristics encompassing 24 demographic and metabolic parameters (Age, Gender, Education, DM duration, HTN, HTN duration, Smoking history, Alcohol use, BMI, CVAI, HbA1c, FBG, FCP, TG, TC, HDL, LDL, Cr, BUN, UA, ALT, AST, ALP) were analyzed. LASSO regression with 10-fold cross-validation (λ selected at 1 standard error from minimum, Fig. 2) identified four robust predictors in the training cohort: age, gender, education, and CVAI. This parsimonious model achieved optimal regularization while preserving predictive accuracy. Additionally, univariable and multivariable logistic regression analyses were conducted to identify factors associated with mild cognitive impairment (MCI), with results presented in Table 3. The dependent variable was MCI status, while independent variables were selected based on prior group analysis. Both univariable and multivariable analyses identified age, gender, CVAI, and education level as significant factors associated with MCI. Notably, stepwise multivariable regression confirmed that CVAI independently contributed to MCI risk in T2DM patients, underscoring its potential role as a key predictor.

Fig. 2.

(A) Lasso Regression Cross-Validation Plot (λ = 0.0694302737352021). (B) Lasso regression coefficient path plot (λ = 0.0694302737352021).

Table 3.

Univariate and multivariate logistic regression analyses of MCI risk in training cohort.

Characteristic	Univariable					Multivariable
	N	Event N	OR1	95% CI1	P value	N	Event N	OR1	95% CI1	P value
Age(year)	236	109	1.09	1.06, 1.13	< 0.001	236	109	1.06	1.02, 1.10	0.004
Gender
Male	154	60	–	–		154	60	–	–
Female	82	49	2.33	1.35, 4.02	0.003	82	49	2.05	1.08, 3.89	0.028
Education (year)	236	109	0.82	0.75, 0.91	< 0.001	236	109	0.88	0.79, 0.98	0.017
DM duration (year)	236	109	1.05	1.01, 1.09	0.007	236	109	1.02	0.97, 1.06	0.444
HTN
No	109	40	–	–		109	40	–	–
Yes	127	69	2.05	1.22, 3.46	0.007	127	69	1.30	0.72, 2.37	0.384
CVAI	236	109	1.01	1.01, 1.02	< 0.001	236	109	1.01	1.00, 1.02	0.002

OR odds ratio, CI confidence interval.

Construction and evaluation of nomogram scoring system

As shown in Fig. 3A, we constructed a nomogram and novel scoring system which included Age, Gender, CVAI, and educational level to predict MCI of patients with T2DM based on these regression analyses. All risk factors that were considered have been assigned numeric scores for quantification and the final risk score was calculated by summing up the score of each item using the nomogram depicted. Then we could predict MCI in T2DM patients according to total points of all risk factors.

Fig. 3.

Diagnostic nomogram and calibration for MCI in T2DM. (A) Diagnostic nomogram for identifying MCI from T2DM. (B) Calibration curves for internal validation of the nomogram. Ideal line representing the ideal prediction, apparent line representing the predictiveness curves, bias-corrected line representing the calibration curve, the bias-corrected line was close to the ideal line, which indicated that the nomogram was well calibrated.

The calibration plots demonstrated excellent concordance between the predictions generated by the nomogram and the observed outcomes (Fig. 3B). The optimism-corrected C-index for assessing the discriminative performance of the nomogram model was 0.747 (95% CI: 0.695–0.800), indicating a robust and accurate estimate of model performance. Additionally, the Hosmer–Lemeshow goodness-of-fit test showed good agreement between the predicted and observed outcomes (χ² = 3.3909, P = 0.9075), further confirming the model’s calibration.

Decision curve analysis (DCA) was employed to evaluate the clinical applicability of the diagnostic nomogram. The net benefit was calculated by quantifying the trade-off between true-positive and false-positive rates. The decision curve demonstrated that using the nomogram to guide MCI screening in T2DM patients provided greater net benefit compared to “treat-all” or “treat-none” strategies across a wide threshold probability range (Fig. 4A, B). The predictive performance was further validated using ROC curves. The nomogram achieved superior discriminative ability compared to CVAI alone, with AUC values of 0.765 (95% CI: 0.705–0.826) in the training cohort and 0.690 (95% CI 0.588–0.792) in the validation cohort (Fig. 4C, D). Clinical utility analysis demonstrated consistent screening benefit optimization when applying the nomogram at probability thresholds > 0.1 in both the training and validation cohorts, particularly enhancing the identification of T2DM patients with MCI (Fig. 4).

Fig. 4.

Decision curve analysis and ROC Curves for the nomogram model. (A) DCA curve for the training cohort. (B) DCA curve for the validation cohort. The DCA curves evaluate the clinical benefit and application range of the nomogram. The x-axis representsthe risk threshold probability, ranging from 0 to 1. The y-axis shows the calculated net benefit corresponding to a given threshold probability. “All” represented net benefit of intervening all patients, and “None” meant net benefit of no patient with intervention, whose net gain was zero. (C) ROC curves for the training cohort. (D) ROC curves for the validation cohort .

Discussion

Type 2 diabetes mellitus has been recognized as an independent risk factor for the development of MCI and dementia, as well as the progression from MCI to dementia^25–28. Visceral adiposity plays an important role in the progression of insulin resistance and type 2 diabetes^7,29due to increased inflammation in various metabolically active tissues both locally and systemically^30–32. CVAI is a visceral fat assessment index based on Asian body fat distribution, developed using a binary linear regression model⁹. Several studies have suggested that the CVAI is a more reliable indicator of visceral fat dysfunction than traditional adiposity indices, such as BMI, WC, WHR, VAI, and LAP^33–36. Moreover, CVAI is noninvasive, convenient, and easily applicable in clinical settings.

This study sought to investigate the association between CVAI and cognitive function in patients with T2DM. Our findings revealed that elevated CVAI was significantly associated with worse cognitive performance, particularly in memory and executive functions. Both univariate and multivariate logistic regression analyses identified CVAI as an independent risk factor for MCI in T2DM patients. Notably, the negative correlation between CVAI and cognitive scores (MoCA and MMSE) remained robust even after adjusting for potential confounders, including age, gender, education level, duration of diabetes, and hypertension.

These findings are consistent with previous reports that visceral adiposity is linked to cognitive decline among patients with diabetes³⁷. The specific mechanisms by which obesity affects the nervous system are yet to be elucidated. In fact, insulin resistance plays a pivotal role in the detrimental effects on hippocampal cognitive function³⁸. Theoretically, insulin resistance is likely associated with inflammatory responses, and consequently, both insulin resistance and inflammation may contribute to cognitive decline^39,40.

Obesity can influence insulin signaling in the brain through mechanisms such as chronic low-grade inflammation, ectopic lipid deposition, and oxidative stress⁴¹. Each component of metabolic syndrome is independently linked to vascular endothelial damage, which impairs cerebral blood flow and increases the risk of cerebrovascular infarction⁴². Furthermore, obesity-induced metabolic abnormalities may directly or indirectly compromise gray matter thickness and white matter integrity through white matter damage, ultimately affecting cognitive function⁴³. Given the difficulty in quantifying cognitive risk in patients with T2DM using traditional clinical indicators alone, we developed a predictive nomogram incorporating CVAI alongside key demographic and clinical variables. The model exhibited satisfactory discrimination and calibration, as demonstrated by internal validation through calibration plots, DCA and ROC curves. This nomogram provides a practical tool for the early identification of T2DM patients at high risk of MCI, potentially enabling timely interventions to delay or prevent cognitive decline.

A key strength of this study is the identification of CVAI as a novel, noninvasive, and easily accessible predictor of cognitive impairment in T2DM. To our knowledge, this is among the first studies to quantitatively assess the risk of MCI using visceral adiposity-based indicators in a Chinese diabetic population, offering new insights into personalized risk stratification and clinical decision-making.

The nomogram was constructed to implement a scoring system, with higher total scores indicating an increased risk of MCI. Internal validation confirmed the model’s reliability and clinical utility, as evidenced by favorable ROC, calibration, and DCA curves. Furthermore, our findings suggest that CVAI, when combined with conventional clinical parameters, contributes meaningfully to individualized risk assessment in diabetic cognitive dysfunction. Despite these strengths, several limitations should be acknowledged. First, the cross-sectional design limits our ability to establish a causal relationship between CVAI and cognitive impairment. Second, although the model demonstrated good internal validity, external validation using independent and more diverse populations is necessary to assess its generalizability. While our sample was derived from a single tertiary care hospital, it is relatively representative of the Chinese T2DM population in terms of demographic and clinical characteristics. However, multicenter and prospective studies are needed to further validate our findings across different ethnic groups, healthcare settings, and glucose tolerance statuses. Additionally, the model did not incorporate neuroimaging data, traditional cognitive biomarkers⁴⁴lifestyle factors, or other metabolic indicators, which may also play critical roles in the development of cognitive decline. Future studies should integrate these variables to enhance predictive performance and deepen mechanistic understanding. Overall, our study highlights the potential of CVAI as an independent and clinically relevant predictor of MCI risk in T2DM patients. The proposed nomogram enables individualized risk assessment and supports the development of targeted preventive strategies, offering practical insights for the management of diabetes-associated cognitive decline.

Conclusion

CVAI was identified as an independent risk factor for MCI in T2DM patients and showed a significant negative association with cognitive function. The nomogram based on CVAI may provide valuable guidance for clinical decision-making regarding T2DM patients with MCI in the Chinese population. Large-scale multi-center studies and prospective analyses are necessary to validate these findings. Subsequently, we could optimize key factors that exhibit both statistical and clinical significance to develop a more precise and practical forecasting model applicable to diverse diabetic populations. However, obesity, as a modifiable and controllable factor influencing cognitive impairment, plays a pivotal role in facilitating effective interventions or delaying the progression of diabetic cognitive impairment. Our results are expected to inform future studies aimed at tailoring interventions for diabetic cognitive dysfunction.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

X.L. Z. and S.H.W. desinged the study; T.N. and D.D.Y. examined the blood samples and collected the clinical data; X.L.Z. analyzed the data, wrote the manuscript and prepared all the Figures and Tables; S.H.W. and Dandan Yu interpretated the data and revised the manuscript.

Funding

This study was supported by the National Natural Science Foundation of China (grant number: 81870568, Shaohua Wang).

Data availability

The data that support the findings of this study are available on request from the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

This research complies with the principle of the Helsinki Declaration. All procedures performed in studies involving human participants were authorized by the Research Ethics Committee, Affiliated ZhongDa Hospital of Southeast University.

Informed consent

Informed consent was obtained from all individual participants included in the study.