Association between diabetic kidney disease and visceral adiposity index: a meta-analysis

Abstract

Background

Visceral adiposity index (VAI) is a simple surrogate marker of visceral fat dysfunction and metabolic risk. Its association with diabetic kidney disease (DKD) remains unclear. This study aimed to evaluate the relationship between VAI and DKD risk through a meta-analysis of observational studies.

Methods

A systematic search of PubMed, Embase, Web of Science, and Cochrane Library was conducted up to March 14, 2025. Observational studies assessing the association between VAI and DKD were included. Two reviewers independently conducted study selection, data extraction, and quality assessment. Pooled estimates were calculated using STATA 15, and heterogeneity was assessed with the I² statistic. Sensitivity analyses were performed to test the robustness of the results. To identify potential sources of heterogeneity, we conducted subgroup analyses by study design, geographic region, index type (VAI vs. Chinese visceral adiposity index, CVAI), and DKD definition (composite vs. albuminuria-only), followed by meta-regression using these covariates.

Results

Nine studies involving 30,721 participants met the inclusion criteria. After adjusting for confounders, VAI was significantly higher in patients with DKD compared to those with diabetes alone (mean difference [MD] = 0.48; 95% confidence intervals (CI): 0.29–0.67; p < 0.001; I² = 96.2%). A dose-response analysis showed that each unit increase in VAI was associated with an 11% increase in DKD risk (odds ratio [OR] = 1.11; 95% CI: 1.03–1.19). Compared with the lowest VAI quartile (Q1), the odds of DKD were progressively higher in Q2 (OR = 1.17), Q3 (OR = 1.33), and Q4 (OR = 1.62). Meta-regression identified the DKD outcome definition (composite vs. albuminuria-only) as a potential source of heterogeneity (p = 0.047), while study design, region, and index type (VAI vs. CVAI) had no significant impact. Subgroup and sensitivity analyses produced consistent findings, indicating a robust positive association between VAI and DKD. However, given the inclusion of cross-sectional studies, causality cannot be established.

Conclusions

This meta-analysis demonstrates that higher VAI is associated with an increased risk of DKD in patients with diabetes. However, given the observational design of included studies, causality remains uncertain. Prospective cohort studies are warranted to confirm the predictive value of VAI for DKD.

Trial registration

The study received registration and approval from PROSPERO (registration number: CRD420251046770).

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12882-025-04741-9.

Background

The International Diabetes Federation (IDF) projects that global diabetes cases will exceed 783 million by 2045 [1], placing a heavy burden on healthcare systems. Diabetic kidney disease (DKD), a major complication of diabetes, is characterized by persistent proteinuria and a progressive decline in glomerular filtration function [2]. Beyond hyperglycemia, the primary driver of diabetic DKD, its development is closely associated with hypertension, dyslipidemia, obesity, chronic inflammation, and insulin resistance [3, 4]. Recently, visceral fat accumulation has drawn increasing attention for its role in DKD. The visceral adiposity index (VAI), a metabolic indicator that combines waist circumference, body mass index (BMI), triglycerides (TG), and high-density lipoprotein cholesterol (HDL-C) [5], indirectly reflects visceral fat distribution and function, and integrates lipid metabolism and insulin resistance to provide an overall assessment of metabolic risk [5, 6]. Several studies have shown a significant association between the VAI and the risk of diabetes and its complications, including DKD [7]. This association may be mediated by proinflammatory cytokines from visceral fat, such as IL-6 and TNF-α, which promote renal injury through oxidative stress and endothelial dysfunction [8–11]. However, the findings remain inconsistent. Some studies identified VAI as an independent predictor of DKD [12], while others have reported that the association weakens after adjusting for BMI, blood glucose, and other conventional factors, suggesting substantial population heterogeneity. Li et al. [13] reported a stronger VAI–DKD association among African American (hazard ratio [HR] = 1.25) and female patients with diabetes in the United States, though the effect diminished after adjustment. Similarly, Chen et al. [14] observed a stronger association in Chinese men and older adults, which lost statistical significance after adjustment. These inconsistencies may reflect differences in population characteristics, follow-up duration, and the confounding control across studies.

Given the conflicting evidence and growing interest in the VAI as a marker of cardiometabolic risk [15], this meta-analysis systematically evaluated its association with DKD.

Methods

We conducted this analysis following the PRISMA 2020 recommendations. Prior to implementation, the study received registration and approval from PROSPERO (registration number: CRD420251046770), thereby ensuring methodological clarity and reproducibility.

Inclusion and exclusion criteria

The inclusion criteria were: (1) observational study design (cohort, cross-sectional, or case-control studies); (2) Eligible participants were adults (≥ 18 years) with type 1 or type 2 diabetes who met the diagnostic criteria for DKD, defined as history of diabetes with persistent albuminuria (urinary albumin-creatinine ratio [UACR] ≥ 30 mg/g or urinary albumin excretion ≥ 30 mg/24 h) and/or estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m². (3) studies providing original or grouped data for VAI (reported as quartiles, tertiles, or continuous variables); (4) studies reporting effect sizes (OR, HR, or Risk Ratio [RR]) with their 95% CI for the relationship between VAI and DKD.

Exclusion criteria were: (1) studies not related to DKD (e.g.,hypertensive nephropathy) or involving special populations (e.g.,gestational diabetes, pediatric or adolescent diabetes); (2) duplicate publications or studies without accessible full-text articles; (3) publications lacking extractable original data; (4) animal studies, conference abstracts, case reports, reviews, letters, commentaries, and expert opinions.

In the PECOS framework of this meta-analysis, the study population (P) consisted of patients with diabetes from China, the United States, and Iran. The exposure (E) is VAI levels, categorized as unclassified, Q1, Q2, Q3, and Q4. The comparison (C) is the difference in VAI levels between DKD patients and diabetes patients without DKD. The main outcome (O) examined was the association between VAI and DKD. The study design (S) includes cohort and cross-sectional studies assessing the relationship between VAI and DKD.

Literature search method

This study systematically searched PubMed, Embase, Web of Science, and the Cochrane Library to collect research articles on the association between the VAI and DKD. The search covered the period from the earliest records in each database to March 13, 2025. To improve both sensitivity and specificity, a search strategy combining subject headings (MeSH/Emtree) and free-text terms was used. To ensure comprehensive inclusion, no limitations were set on the study location or publication date. Example search keywords included: Diabetic Nephropathies, Nephropathies Diabetic, and others (For detailed search strategy, see Table 1).

Table 1.

The search strategy

Datebases	Searching strategy	Literature number
Pubmed	#1 “diabetic nephropathies“[MeSH Terms] #2 ((((((((((((((((Nephropathies, Diabetic[Title/Abstract]) OR (Nephropathy, Diabetic[Title/Abstract])) OR (Diabetic Kidney Disease[Title/Abstract])) OR (Diabetic Kidney Diseases[Title/Abstract])) OR (Kidney Disease, Diabetic[Title/Abstract])) OR (Kidney Diseases, Diabetic[Title/Abstract])) OR (Diabetic Nephropathy[Title/Abstract])) OR (Diabetic Glomerulosclerosis[Title/Abstract])) OR (Glomerulosclerosis, Diabetic[Title/Abstract])) OR (Intracapillary Glomerulosclerosis[Title/Abstract])) OR (Kimmelstiel-Wilson Disease[Title/Abstract])) OR (Kimmelstiel Wilson Disease[Title/Abstract])) OR (Nodular Glomerulosclerosis[Title/Abstract])) OR (Glomerulosclerosis, Nodular[Title/Abstract])) OR (Kimmelstiel-Wilson Syndrome[Title/Abstract])) OR (Kimmelstiel Wilson Syndrome[Title/Abstract])) OR (Syndrome, Kimmelstiel-Wilson[Title/Abstract]) #3 #1 OR #2 #4 ((((visceral adiposity index) OR (visceral adipose index)) OR (visceral fat indexes)) OR (VAI)) OR (VFI) #5 #3 AND #4	38
Web of science	#1 TS=(Diabetic Nephropathies) OR TS=(Nephropathies, Diabetic) OR TS=(Nephropathy, Diabetic) OR TS=(Diabetic Kidney Disease) OR TS=(Diabetic Kidney Diseases) OR TS=(Kidney Disease, Diabetic) OR TS=(Kidney Diseases, Diabetic) OR TS=(Diabetic Nephropathy) OR TS=(Diabetic Glomerulosclerosis) OR TS=(Glomerulosclerosis, Diabetic) OR TS=(Intracapillary Glomerulosclerosis) OR TS=(Kimmelstiel-Wilson Disease) OR TS=(Kimmelstiel Wilson Disease) OR TS=(Nodular Glomerulosclerosis) OR TS=(Glomerulosclerosis, Nodular) OR TS=(Kimmelstiel-Wilson Syndrome) OR TS=(Kimmelstiel Wilson Syndrome) OR TS=(Syndrome, Kimmelstiel-Wilson) #2 TS=(visceral adiposity index) OR TS=(visceral adipose index) OR TS=(visceral fat indexes) OR TS=(VAI) OR TS=(VFI) #3 #2 AND #1	66
Embase	#1 ‘diabetic nephropathy’/exp #2 ‘diabetic nephropathies’:ab, ti #3 ‘nephropathies, diabetic’:ab, ti #4 ‘nephropathy, diabetic’:ab, ti #5 ‘diabetic kidney disease’:ab, ti #6 ‘diabetic kidney diseases’:ab, ti #7 ‘kidney disease, diabetic’:ab, ti #8 ‘kidney diseases, diabetic’:ab, ti #9 ‘diabetic nephropathy’:ab, ti #10 ‘diabetic glomerulosclerosis’:ab, ti #11 ‘glomerulosclerosis, diabetic’:ab, ti #12 ‘intracapillary glomerulosclerosis’:ab, ti #13 ‘intracapillary glomerulosclerosis’:ab, ti #14 ‘kimmelstiel-wilson disease’:ab, ti #15 ‘kimmelstiel wilson disease’:ab, ti #16 ‘nodular glomerulosclerosis’:ab, ti #17 ‘glomerulosclerosis, nodular’:ab, ti #18 ‘kimmelstiel-wilson syndrome’:ab, ti #19 ‘kimmelstiel wilson syndrome’:ab, ti #20 ‘syndrome, kimmelstiel-wilson’:ab, ti #21 #1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12 OR #13 OR #14 OR #15 OR #16 OR #17 OR #18 OR #19 OR #20 #23 ‘visceral adiposity index’ #24 ‘visceral adipose index’ #25 ‘visceral fat indexes’ #26 ‘vai’ #27 ‘vfi’ #28 #23 OR #24 OR #25 OR #26 OR #27 #29 #21 AND #28	21
Cochrane	#1MeSH descriptor: [Diabetic Nephropathies] explode all trees #2(Diabetic Nephropathies): ti, ab, kw OR (Nephropathies, Diabetic): ti, ab, kw OR (Nephropathy, Diabetic): ti, ab, kw OR (Diabetic Kidney Disease): ti, ab, kw OR (Diabetic Kidney Diseases): ti, ab, kw #3(Kidney Disease, Diabetic): ti, ab, kw OR (Kidney Diseases, Diabetic): ti, ab, kw OR (Diabetic Nephropathy): ti, ab, kw OR (Diabetic Glomerulosclerosis): ti, ab, kw OR (Glomerulosclerosis, Diabetic): ti, ab, kw #4(Intracapillary Glomerulosclerosis): ti, ab, kw OR (Kimmelstiel-Wilson Disease): ti, ab, kw OR (Kimmelstiel Wilson Disease): ti, ab, kw OR (Nodular Glomerulosclerosis): ti, ab, kw OR (Glomerulosclerosis, Nodular): ti, ab, kw #5(Kimmelstiel-Wilson Syndrome): ti, ab, kw OR (Kimmelstiel Wilson Syndrome): ti, ab, kw OR (Syndrome, Kimmelstiel-Wilson): ti, ab, kw #6#1 or #2 or #3 or #4 or #5 #7(visceral adiposity index) OR (visceral adipose index) OR (visceral fat indexes) OR (VAI) OR (VFI) #8#6 and #7	7

Literature screening and data extraction

Literature screening and data extraction were carried out independently by two researchers. Titles and abstracts were first reviewed to filter out non-relevant studies, and then the full texts of potentially eligible articles were analyzed to verify whether they met the predefined inclusion criteria. The following information was collected: (1) Basic study characteristics: the first author, publication year, study location, design, total sample size, mean age, and gender distribution of participants; (2) Study-specific data: including the sample size for each group, and the mean and standard deviation of VAI. If additional data were provided, the mean and standard deviation were calculated using relevant formulas; (3) Analytical outcomes: focusing on VAI classification data adjusted for confounders, and the OR with 95% CI for the relationship between VAI and DKD. Throughout the process, researchers worked independently and cross-checked their results. Any disagreements were settled by group discussions until consensus was achieved.

Quality assessment method

Two researchers independently evaluated the potential sources of bias across the selected studies and conducted cross-validation. Given that the included studies comprised both cross-sectional and cohort designs, appropriate quality assessment tools were applied to each study type. For cross-sectional studies, the 11-item quality assessment scale developed by the AHRQ (Agency for Healthcare Research and Quality) was used. Studies were classified into three quality levels based on the proportion of criteria met: low quality (compliance rate < 30%), moderate quality (compliance rate 30%-60%), and high quality (compliance rate > 60%) [16]. For cohort studies, to evaluate cohort studies, the Newcastle-Ottawa Scale (NOS) was applied, focusing on three principal aspects of study quality: (1) selection of study population (up to 4 points), (2) comparability between groups (up to 2 points), and (3) outcome measurement (up to 3 points). A total score of 9 points or more was considered high-quality research [17, 18].

Statistical analysis

STATA 15.0 software was utilized to perform all statistical analyses in this study. The relationship between VAI and the risk of DKD was assessed through meta-analysis. For effect size estimation, OR and 95% CI were used for dichotomous variables, while standardized mean differences (SMD) or weighted mean differences (WMD) were used for continuous variables. The effect model was determined based on the results of the heterogeneity test: a fixed-effect model was applied when no significant statistical or clinical heterogeneity was found (p > 0.10 and I² < 50%), while a random-effects model was used if substantial heterogeneity was detected (p < 0.10 or I² ≥ 50%). To identify potential sources of heterogeneity, we conducted subgroup analyses and meta-regression to assess the effects of covariates. Sensitivity analyses were performed by sequentially excluding individual studies to test the robustness of the results; the findings remained stable when heterogeneity was unchanged. Funnel plots and Egger’s test were used to evaluate publication bias, with P > 0.05 indicating no bias and P < 0.05 suggesting its presence.

Results

Study selection

Systematic searches of PubMed, Web of Science, Embase, and the Cochrane Library retrieved a total of 132 articles. Reference management was performed using EndNote X9. The screening process included: (1) removing 38 duplicates and excluding 10 studies based on type (animal studies = 3, conference abstracts = 4, meta-analyses = 2, reviews = 1); (2) excluding 70 irrelevant articles after title and abstract screening based on the PICOS framework; (3) assessing the full text of 14 remaining articles, excluding 4 due to missing key data and 1 due to unclear DKD diagnosis; and (4) evaluating the final 9 studies using the AHRQ and NOS tools, all of which met moderate or high quality criteria (score ≥ 6). Nine studies [7, 13, 19–25]met the inclusion criteria and were included in the meta-analysis. The study selection process is depicted in Fig. 1.

Fig. 1.

Study selection flowchart

Study characteristics

A total of nine observational studies [7, 13, 19–25]were included, comprising four cross-sectional and five cohort studies, published between 2020 and 2025. The pooled sample included 30,721 patients with diabetes, of whom 6,753 had DKD and 23,968 had diabetes without DKD. Among the four cross-sectional studies, two were conducted in China [7, 19], one in Iran [22], and one in the United States (U.S. NHANES) [13]. All five cohort studies were conducted in China [20, 21, 23–25].

Data sources included the NHANES survey published by the U.S. Centers for Disease Control and Prevention (CDC) [13], diabetes-related data from hospital outpatient clinics [21–25] and inpatient departments [7], as well as community-based [19]and urban population-based studies [20]. Among the analyzed patients with diabetes, the majority were male, with a mean age over 55 years. Diagnoses of diabetes and DKD were based on clinical, laboratory, or combined diagnostic criteria.

In these studies, the VAI was calculated using a standardized formula:

VAI classification varied across studies: two used quartiles [19, 23], three used tertiles [13, 20, 21], and four did not specify a classification method [7, 22, 24, 25]. The primary clinical finding was a positive association between elevated VAI levels and an increased risk of DKD in patients with diabetes. Detailed baseline characteristics are presented in Table 2. The specific definitions of DKD used across the included studies are summarized in Table S1.

Table 2.

Baseline characteristics of included studies

Study	Year	Study design	Country	Sample size	Number of DKD	Gender(M/F)	Mean age	Regression modle
CY Li [13]	2024	cross-sectional	USA	2508	945	1336/1172	58.76	Logistic regression
H Wang [19]	2020	cross-sectional	China	3824	893	1949/1875	67.15	Logistic regression
M Sun [20]	2025	cohort study	China	4172	748	3462/710	53.8	Cox regression
ZY WU [21]	2022	cohort study	China	8948	467	6154/2794	53.35	Cox regression
PP Zhao [7]	2024	cross-sectional	China	1176	548	746/430	59.48	Linear regression
S Ali [22]	2023	cross-sectional	Iran	2934	526	1380/1554	56.46	Logistic regression
ZZ Sun [23]	2023	cohort study	China	2659	1080	1307/1352	57.74	Cox regression
YL Ou [24]	2021	cohort study	China	1872	706	809/1063	64	Logistic regression
ZZ Sun [25]	2022	cohort study	China	2628	840	1356/1272	55.9	Cox regression

Quality assessment of included studies

The quality of the included studies was thoroughly evaluated using standardized tools. The four cross-sectional studies were assessed using the AHRQ scale (www.ncbi.nlm.nih.gov), with three classified as high quality and one as moderate quality. The five cohort studies were evaluated using the NOS, and all were deemed high quality. Both tools assess studies across multiple domains, including study design, sample selection, data collection, and statistical analysis. All included studies demonstrated sound methodological rigor and reliable results, with no major methodological flaws or high risk of bias identified. These findings indicate an overall high quality of the included literature, providing a solid foundation for the subsequent meta-analysis. The detailed quality assessment is presented in Table 3.

Table 3.

Quality assessment of included studies

Cross-sectional
Study	Is the source of the data clearly specified?	Are the inclusion and exclusion criteria for the exposed group and the unexposed group listed, or are previous publications referred to?	Has the time period for identifying patients been given?	If the study subjects are not sourced from the general population, are the study subjects continuous?	Does the subjective factor of the evaluator obscure other aspects of the study subjects' situations?	Describe any evaluations carried out to ensure quality.	Explained the reasons for excluding any patients from the analysis.	Describe the measures on how to evaluate and/or control confounding factors.	If possible, explain how the missing data were handled in the analysis.	Summarized the response rate of the patients and the completeness of data collection.	If there is a follow-up, ascertain the percentage of expected incomplete patient data or the follow-up results.
1. Quality assessment of individual studies using Agency for Healthcare Research and Quality
CY Li	yes	yes	yes	yes	unclear	no	yes	yes	unclear	yes	unclear
H Wang	yes	yes	yes	yes	unclear	no	yes	yes	unclear	yes	unclear
PP Zhao	yes	yes	yes	yes	unclear	no	yes	yes	unclear	yes	unclear
S Ali	yes	yes	yes	yes	unclear	no	no	yes	unclear	yes	unclear

Cohort study
Study	Representativeness of the exposed group	Selection of non-exposed groups	Determination of exposure factors	Identification of outcome indicators not yet to be observed at study entry	Comparability of exposed and unexposed groups considered in design and statistical analysis	design and statistical analysis	Adequacy of the study's evaluation of the outcome	Adequacy of follow-up in exposed and unexposed groups	Total scores
2. Quality assessment of individual studies using Newcastle-Ottawa Scale
M Sun	1	1	1	0	2	1	1	1	8
ZY WU	1	1	1	0	2	1	1	1	8
ZZ Sun	1	1	1	0	2	1	1	1	8
YL Ou	1	1	1	0	2	1	1	1	8
ZZ Sun	1	1	1	0	2	1	1	1	8

Meta-analysis results

VAI levels in the diabetes-only group vs. The DKD group

In this meta-analysis, seven studies [13, 19, 21–25] reported VAI levels in both diabetes (n = 19,916) and DKD groups (n = 5,457). Due to high heterogeneity (I² = 96.2%, P < 0.001), a random-effects model was used. Pooled results showed significantly higher VAI levels in DKD patients than in those with diabetes alone (MD = 0.48, 95% CI: 0.29–0.67; P < 0.001). See Fig. 2(a).

Fig. 2.

VAI Levels: Diabetes-Only Group vs. DKD Group. (a). Forest Plot; (b). Sensitivity Analysis

To explore sources of heterogeneity, subgroup analyses were conducted by study design, country, index type, and outcome definition (Table 4a). A consistent positive association between VAI and DKD was observed across study designs (cross-sectional and cohort), countries (United States, China, Iran), index types (VAI vs. CVAI), and outcome definitions (composite vs. albuminuria-only). Meta-regression (Table 5a) indicated that outcome definition was a potential contributor to heterogeneity (P = 0.047), while study design, country, and index type were not significant (all P > 0.05). These results suggest that, beyond outcome definition, heterogeneity likely arises from unmeasured factors.

Table 4a.

Summary of subgroup analyses. subgroup analysis of VAI levels in the diabetes-only group vs. the DKD group

Analysis	WMD (95% CI)	I-squared(%)	P(Q-test)	Model
Overall	0.48 (0.29, 0.67)	96.2	0.000	Random
Study design
Cross-sectional	0.57 (0.32, 0.82)	90.1	0.000	Random
Cohort	0.38 (0.17, 0.60)	91.7	0.000	Random
Country
USA	0.60 (0.59, 0.61)	-	-	-
China	0.39 (0.22, 0.56)	89.1	0.000	Random
Iran	1.37 (0.70, 2.04)	-	-	-
Indicator type
VAI	0.51 (0.30, 0.72)	96.4	0.000	Random
CVAI	0.31 (0.19, 0.43)	-	-	-
Outcome definition
Composite	0.43 (0.24, 0.62)	96.7	0.000	Random
Albuminuria-only	1.37 (0.70, 2.04)	-	-	-

Table 5a.

Summary of meta-regression analyses. Meta-regression of VAI levels in the diabetes-only group vs. the DKD group

Variable	Coefficient	95% CI	SE	P
Study design
Cross-sectional	1.211	（0.679 2.160）	0.286	0.448
Cohort	0.825	（0.463 1.472）	0.195	0.448
Country
USA	1.138	（0.481 2.693）	0.401	0.726
China	0.673	（0.361 1.256）	0.171	0.171
Iran	2.565	（0.912 7.214）	1.084	0.067
Indicator type
VAI	1.231	（0.520 2.919）	0.434	0.576
CVAI	0.812	（0.343 1.925）	0.286	0.576
Outcome definition
Composite	0.390	（0.139 1.096）	0.165	0.047
Albuminuria-only	2.565	（0.912 7.214）	1.084	0.047

Result robustness was assessed using a leave-one-out sensitivity analysis (Fig. 2b). The pooled WMD was 0.48 (95% CI: 0.29–0.67). Sequential exclusion of individual studies yielded stable estimates (0.29–0.67) with overlapping CI (0.24–0.72), confirming result stability.

Association between ungrouped VAI and DKD risk

Four studies [7, 22, 24, 25] compared unclassified VAI levels between DKD and diabetes-only groups. Meta-analysis revealed that higher VAI significantly increased DKD risk in patients with diabetes (OR = 1.11, 95% CI: 1.03–1.19). Given substantial heterogeneity (I² = 88.7%, P < 0.001), a random-effects model was applied (Fig. 3a).

Fig. 3.

Ungrouped VAI and DKD Risk. (a). Forest Plot; (b). Sensitivity Analysis

To identify sources of heterogeneity, subgroup analyses were performed by study design, country, and outcome definition (Table 4b). All subgroups showed a significant positive association. Meta-regression (Table 5b) revealed that none of these covariates significantly explained the heterogeneity (all P > 0.05).

Table 4b.

Summary of subgroup analyses. Subgroup analysis of the association between ungrouped VAI and DKD risk

Analysis	OR(95% CI)	I-squared(%)	P(Q-test)	Model
Overall	1.11 (1.03, 1.19)	88.7	0.000	Random
Study design
Cross-sectional	1.26 (0.82, 1.93)	91.7	0.001	Random
Cohort	1.10 (1.05, 1.16)	55.9	0.132	Random
Country
China	1.15 (1.05, 1.27)	81.0	0.005	Random
Iran	1.03 (1.01, 1.05)	-	-	-
Outcome definition
Composite	1.15 (1.05, 1.27)	81.0	0.005	Random
Albuminuria-only	1.03 (1.01, 1.05)	-	-	-

Table 5b.

Meta-regression of the association between ungrouped VAI and DKD risk

Variable	Coefficient	95% CI	SE	P
Study design
Cross-sectional	1.092	(0.389 3.063)	0.262	0.748
Cohort	0.916	(0.326 2.568)	0.219	0.748
Country
China	1.089	(0.767 1.548)	0.089	0.405
Iran	0.918	(0.646 1.305)	0.075	0.405
Outcome definition
Composite	1.089	(0.767 1.548)	0.089	0.405
Albuminuria-only	0.918	(0.646 1.305)	0.075	0.405

Result robustness was evaluated using a leave-one-out sensitivity analysis (Fig. 3b). The pooled OR was 1.11 (95% CI: 1.03–1.19). Sequential exclusion of individual studies produced stable estimates (1.02–1.17) with overlapping CI(0.99–1.24), confirming result stability.

VAI index (Q2 vs. Q1)

In the initial meta-analysis, the study by Li Chunyao et al. [13] contributed 99% of the total weight, potentially exerting undue influence on the pooled estimate. To ensure robustness, it was excluded, and the effect size and forest plot were recalculated. After exclusion, four studies [19, 20, 21, 23] compared DKD risk between VAI Q2 and Q1 groups, showing a 17% higher risk in Q2 patients (OR = 1.17, 95% CI: 1.05–1.30) (Fig. 4a).

Fig. 4.

VAI Quartile Comparison: Q2 vs. Q1. (a). Forest Plot; (b). Sensitivity Analysis

Subgroup analysis (Table 4c) indicated a consistent association across study designs and index types. Meta-regression (Table 5c) found no significant impact of either factor on the pooled estimates (all P > 0.05).

Table 4c.

Summary of subgroup analyses. Subgroup analysis of VAI index (Q2 vs. Q1)

Analysis	OR (95% CI)	I-squared(%)	P(Q-test)	Model
Overall	1.17 (1.05, 1.30)	18.5	0.297	Fixed-effect
Study design
Cross-sectional	1.39 (1.06, 1.80)	48.6	0.163	Fixed-effect
Cohort	1.13 (1.00, 1.27)	0.0	0.610	Fixed-effect
Indicator type
VAI	1.17 (1.00, 1.35)	54.2	0.113	Fixed-effect
CVAI	1.17 (0.99, 1.37)	0.0	0.461	Fixed-effect

Table 5c.

Meta-regression of VAI index (Q2 vs. Q1)

Variable	Coefficient	95% CI	SE	P
Study design
Cross-sectional	1.223	(0.585 2.559)	0.284	0.449
Cohort	0.817	(0.391 1.710)	1.182	0.449
Indicator type
VAI	0.963	(0.606 1.532)	0.140	0.815
CVAI	1.038	(0.653 1.650)	0.151	0.815

Result robustness was evaluated using a leave-one-out sensitivity analysis (Fig. 4b). With low heterogeneity (I² = 18.5%, P = 0.297), a fixed-effects model produced a pooled OR of 1.17 (95% CI: 1.05–1.30). Sequential exclusion of individual studies yielded stable estimates (1.02–1.27) with overlapping CI(0.97–1.36) and no directional shift, confirming result stability.

VAI index (Q3 vs. Q1)

Five studies [13, 19–21, 23] compared DKD risk between VAI Q3 and Q1 groups. Meta-analysis showed a 33% higher risk in Q3 patients (OR = 1.33, 95% CI: 1.20–1.47). Given moderate heterogeneity (I² = 53.8%), a random-effects model was applied (Fig. 5a).

Fig. 5.

VAI Quartile Comparison: Q3 vs. Q1. (a). Forest Plot; (b). Sensitivity Analysis

To identify sources of heterogeneity, subgroup analyses were performed by study design, country, and index type (Table 4d). The increased risk persisted across all subgroups. Meta-regression (Table 5d) showed no significant effect of these factors on heterogeneity (all P > 0.05).

Table 4d.

Summary of subgroup analyses. Subgroup analysis of VAI index (Q3 vs. Q1)

Analysis	OR (95% CI)	I-squared(%)	P(Q-test)	Model
Overall	1.33 (1.20, 1.47)	53.8	0.055	Random
Study design
Cross-sectional	1.36 (1.36, 1.36)	0.0	0.954	Random
Cohort	1.32 (1.05, 1.67)	74.6	0.019	Random
Country
USA	1.36 (1.36, 1.36)	-	-	-
China	1.33 (1.13, 1.57)	53.3	0.073	Random
Indicator type
VAI	1.28 (1.12, 1.47)	64.9	0.036	Random
CVAI	1.49 (1.23, 1.79)	14.8	0.279	Random

Table 5d.

Meta-regression of VAI index (Q3 vs. Q1)

Variable	Coefficient	95% CI	SE	P
Study design
Cross-sectional	1.079	(0.702 1.660)	0.167	0.649
Cohort	0.927	(0.603 1.425)	0.144	0.649
Country
USA	1.040	(0.630 1.718)	0.188	0.893
China	0.962	(0.582 1.588)	0.174	0.893
Indicator type
VAI	0.839	(0.503 1.398)	0.154	0.393
CVAI	1.192	(0.715 1.987)	0.219	0.393

Result robustness was assessed using a leave-one-out sensitivity analysis (Fig. 5b). With moderate heterogeneity (I² = 53.8%, P = 0.055), the pooled OR was 1.33 (95% CI: 1.20–1.47). Sequential exclusion of individual studies produced stable estimates (1.18–1.46) with overlapping CI (1.11–1.53) and no directional shift, confirming result robustness.

VAI index (Q4 vs. Q1)

Two studies [19, 23] compared DKD risk between VAI Q4 and Q1 groups. Meta-analysis showed a 62% higher risk in Q4 patients (OR = 1.62, 95% CI: 1.15–2.29). Given substantial heterogeneity (I² = 76.8%, P = 0.013), a random-effects model was applied (Fig. 6a).

Fig. 6.

VAI Quartile Comparison: Q4 vs. Q1. (a). Forest Plot; (b). Sensitivity Analysis

Due to the limited number of studies, subgroup and meta-regression analyses could not be performed. Sensitivity analysis (Fig. 6b), using a leave-one-out approach, showed stable estimates (1.07–2.15) with overlapping CI (0.82–2.46) and consistent direction, indicating moderate stability. However, because of the small sample size and high heterogeneity, these findings should be interpreted with caution and validated in future research.

In summary, this meta-analysis of nine studies [7, 13, 19–25] evaluated the association between the VAI and DKD risk. Higher VAI levels were positively associated with an increased DKD risk in observational data (OR = 1.11, 95% CI: 1.03–1.19). Quartile analyses revealed a stepwise increase in risk compared with Q1: 17% for Q2 (OR = 1.17, 95% CI: 1.05–1.30), 33% for Q3 (OR = 1.33, 95% CI: 1.20–1.47), and 62% for Q4 (OR = 1.62, 95% CI: 1.15–2.29). These findings suggest a potential dose–response relationship between VAI and DKD risk. However, as some included studies were cross-sectional, causality cannot be inferred, and prospective studies are warranted for confirmation.

Publication bias

Publication bias was evaluated using funnel plots (Fig. 7) and Egger’s tests: VAI levels between diabetes and DKD groups (Fig. 7a, P = 0.210), unclassified VAI (Fig. 7b, P = 0.065), Q2 vs. Q1 (Fig. 7c, P = 0.222), Q3 vs. Q1 (Fig. 7d, P = 0.753), and Q4 vs. Q1 (Fig. 7e, P = 0.061). All P-values exceeded 0.05, indicating a low risk of publication bias. However, the limited number of studies may reduce statistical power and undetected bias cannot be entirely excluded.

Fig. 7.

Publication Bias Assessment: Funnel Plots. (a). VAI Levels (Diabetes vs. DKD); (b). Ungrouped VAI; (c). Q2 vs. Q1; (d). Q3 vs. Q1; (e). Q4 vs. Q1

Discussion

This meta-analysis, the first to examine the association between VAI and DKD risk, included nine clinical studies [7, 13, 19–25] with a total of 30,721 patients. Pooled results showed that each one-unit increase in VAI was associated with an 11% higher risk of DKD (OR = 1.11), and patients in the highest quartile (Q4) had a 62% higher risk than those in Q1 (OR = 1.62). The robustness of this gradient relationship was confirmed in sensitivity analyses. Furthermore, subgroup analyses revealed consistent positive associations across study designs, geographic regions, and DKD definitions, thereby strengthening the reliability of the main findings. Meta-regression indicated that study design, region, and index type (VAI vs. CVAI) were not significant sources of heterogeneity, while DKD outcome definition (composite vs. albuminuria-only) may contribute to the observed variability (P = 0.047).

This study has several strengths. It adhered to PRISMA 2020 guidelines and was prospectively registered on PROSPERO, ensuring methodological transparency. Comprehensive, unrestricted database searches minimized selection bias. Data from nine studies involving a total of 30,721 patients provided robust statistical power. The stepwise increase in DKD risk across VAI quartiles indicates a monotonic exposure–response relationship, offering valuable evidence for risk stratification.

From a pathophysiological standpoint, VAI may impair renal function through several mechanisms. Visceral fat dysfunction leads to systemic metabolic disturbances, including increased insulin resistance and chronic hyperglycemia, which directly or indirectly damage renal cells [5, 25, 26]. Visceral adipose tissue further contributes by releasing proinflammatory cytokines (IL-6, TNF-α) [8–11, 28], activating the renin–angiotensin–aldosterone system (RAAS) [29], and inducing oxidative stress [27] and lipotoxicity [30], collectively promoting renal inflammation, fibrosis, and dysfunction. Additionally, adipocyte-derived exosomal microRNAs may mediate renal cell injury by regulating apoptosis-related genes[31].

This study has several limitations. First, the number of studies included in certain comparisons (such as the highest quartile, Q4) is limited, which may affect the stability of the results. Second, the observed significant heterogeneity may stem from inconsistencies in DKD diagnostic criteria, lack of standardized VAI cutoffs, and differences in population characteristics. Third, the VAI cutoffs derived from the aggregated data (Q1: <0.91; Q2: 0.91–2.65; Q3: 2.66–4.30; Q4: >4.30) should not be directly applied as clinical diagnostic criteria due to the general absence of sex-stratified data and the predominance of Asian populations in the included studies. Fourth, the diagnosis of DKD was based solely on clinical indicators (albuminuria and eGFR) rather than renal biopsy, which may have led to the inclusion of non-diabetic kidney disease cases. Finally, residual confounding cannot be entirely ruled out due to the observational nature of the study design.

Future research should prioritize several areas. Large-scale, multiethnic prospective studies are needed to establish causality, validate standardized VAI cutoffs, and clarify underlying molecular mechanisms. Because most existing studies focus on overweight or obese individuals, future work should include BMI-stratified analyses—especially among underweight populations—to assess potential U-shaped associations and define optimal VAI ranges for DKD risk prediction. Further studies should also examine whether combining VAI with renal biomarkers such as UACR and eGFR can improve early detection and risk stratification for DKD.

Conclusion

This meta-analysis demonstrates a significant association between higher VAI and increased risk of DKD in patients with diabetes, with a clear dose–response pattern. However, the observational nature of the included studies precludes causal inference. As most evidence comes from Asian populations, further validation in multiethnic cohorts is needed. Future studies should evaluate VAI’s utility in early DKD detection and risk stratification across the full BMI spectrum, including underweight individuals, to better define its clinical relevance.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

VAI

Visceral Adiposity Index

DKD

Diabetic kidney disease

CVAI

Chinese visceral adiposity index

MD

Mean difference

CI

Confidence intervals

OR

Odds ratio

IDF

International Diabetes Federation

BMI

Body mass index

TG

Triglycerides

HDL-C

High-density lipoprotein cholesterol

HR

Hazard ratio

UACR

Urinary albumin-creatinine ratio

eGFR

Estimated glomerular filtration rate

RR

Risk Ratio

AHRQ

Agency for Healthcare Research and Quality

NOS

Newcastle-Ottawa Scale

SMD

Standardized mean differences

WMD

Weighted mean differences

CDC

Centers for Disease Control and Prevention

RAAS

Renin–angiotensin–aldosterone system

Author contributions

Conceptualization: Yichan Huang, Shuguang Sun; Methodology: Yichan Huang; Formal analysis and investigation: Yichan Huang; Writing - original draft preparation: Yichan Huang; Writing - review and editing: Shuguang Sun; Resources: Dali University; Supervision: Shuguang Sun.All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

All data generated or analysed during this study are included in this published article and its supplementary information files.

Declarations

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.