Association between the novel dietary index for gut microbiota and diabetic kidney disease among patients with diabetes: evidence from the NHANES

Abstract

Background

Emerging evidence suggests that gut microbiota-modulating dietary patterns may influence the development of diabetic kidney disease (DKD), but population-based evidence remains limited. This study aimed to investigate the associations between the Dietary Index for Gut Microbiota (DI-GM) and DKD in a nationally representative sample.

Methods

We analyzed data from the National Health and Nutrition Examination Survey (NHANES), which included 5,560 adults with diabetes. The DI-GM was calculated on the basis of dietary components known to modulate the gut microbiota. DKD was defined by the urinary albumin-to-creatinine ratio or estimated glomerular filtration rate. Multivariate logistic regression, restricted cubic spline (RCS), and subgroup analyses were performed to assess associations while adjusting for sociodemographic, lifestyle, and clinical factors.

Results

Logistic regression analyses revealed that higher DI-GM scores were significantly associated with reduced DKD risk (adjusted OR = 0.96, 95% CI: 0.92–0.99). The RCS results demonstrated a nonlinear relationship between DI-GM and DKD, with a threshold effect at DI-GM = 5. When the DI-GM was < 5, no significant association was observed (OR = 1.04, 95% CI: 0.97–1.11; P = 0.31). However, when DI-GM was ≥ 5, each unit increase in DI-GM was associated with a 12% reduction in DKD risk (OR = 0.88, 95% CI: 0.81–0.95; P < 0.01). The association remained consistent across subgroups (all P-interaction > 0.05) and proved robust in sensitivity analyses accounting for missing data and additional clinical confounders.

Conclusion

Our findings indicate a nonlinear association between the DI-GM and DKD risk. Specifically, a protective effect was observed only when the DI-GM reached a threshold of 5. These results suggest that maintaining a higher DI-GM score, particularly above 5, may be an effective dietary strategy to lower the risk of DKD in patients with diabetes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-26371-y.

Introduction

Diabetic kidney disease (DKD) is a common microvascular complication of diabetes and poses a significant threat to global public health [1, 2]. Approximately 30–50% of diabetic patients are estimated to develop DKD during their lifetime [3–5]. DKD not only leads to a progressive decline in renal function but also increases the risk of cardiovascular disease and all-cause mortality. The pathophysiology of DKD is complex and multifactorial and involves metabolic abnormalities, inflammatory responses, oxidative stress, and genetic predispositions induced by chronic hyperglycemia [5, 6]. Traditional risk factors, such as poor glycemic control, hypertension, and dyslipidemia, have been extensively studied and are well recognized in the development and progression of DKD [7]. However, DKD still occurs in individuals with well-controlled traditional risk factors, highlighting the importance of identifying additional risk factors for the prevention and management of DKD [8].

The gut microbiota, a vast and diverse microbial community residing in the gastrointestinal tract, has gained increasing attention for its role in regulating host metabolism and immune function [9]. Gut dysbiosis, defined as alterations in the composition and function of the gut microbiota, has been associated with various diseases, including diabetes [9], obesity [10], and cardiovascular disease [11]. In recent years, a growing body of research has emphasized the potential link between the gut microbiota and DKD, suggesting its critical role in the pathophysiology of DKD [12–14]. For example, patients with DKD often exhibit distinct gut microbial characteristics, such as a reduction in beneficial bacteria and an increase in pathogenic bacteria. These changes may lead to increased intestinal permeability, allowing bacterial products to enter the systemic circulation and subsequently activate inflammatory and immune responses, thereby exacerbating Kidney injury [12, 15, 16]. Diet is widely recognized as a key modulator of the gut microbiota. Different dietary patterns and nutrients can significantly shape the composition and metabolic activities of the gut microbiota [17]. Diets rich in fiber, fruits, vegetables, and whole grains have been shown to promote the growth of beneficial gut bacteria and enhance microbial diversity, whereas diets high in saturated fats, processed foods, and added sugars may contribute to gut dysbiosis [18, 19].

Given the close relationship between diet and the gut microbiota, the dietary index for gut microbiota (DI-GM) has emerged as a promising tool for assessing the overall quality of dietary patterns in relation to gut microbiota health [20]. The DI-GM integrates multiple dietary components that are either beneficial or detrimental to the gut microbiota. This 14-component index evaluates the intake of beneficial (e.g., fermented dairy products, dietary fiber) and harmful (e.g., processed meat, sugars) foods, providing a comprehensive score that reflects an individual’s dietary habits and their potential impact on the gut microbiota [20]. Previous studies have shown that higher DI-GM scores are associated with a lower risk of kidney diseases, including chronic kidney disease (CKD) [21–23], hypoalbuminemia in CKD patients [24], and Kidney stones [25, 26]. Furthermore, higher DI-GM scores are linked to a reduced risk of metabolic disorders, such as metabolic syndrome [27–30], diabetes [31–33], and dyslipidemia [34]. However, the relationship between DI-GM and the risk of DKD remains unclear. Despite these insights, the specific relationship between DI-GM and the risk of DKD remains unexplored. In particular, it is unclear whether dietary patterns specifically designed to modulate the gut microbiota can counteract the complex pathophysiology of renal injury within the high-risk metabolic milieu of diabetes. Moreover, population-level evidence regarding potential non-linear associations or optimal dietary thresholds for renal protection in diabetic patients is currently lacking.

In this study, we aimed to investigate the association between DI-GM and the risk of DKD using data from the National Health and Nutrition Examination Survey (NHANES). These findings may contribute to a deeper understanding of the potential role of dietary interventions targeting the gut microbiota in the prevention and management of DKD, thereby offering new insights for the development of preventive and therapeutic strategies for DKD.

Method

Research data and research population

This study employed a cross-sectional design utilizing publicly available data from the nationally representative National Health and Nutrition Examination Survey (NHANES) (https://www.cdc.gov/nchs/nhanes/index). Briefly, the NHANES is a nationwide cross-sectional survey program conducted by the National Center for Health Statistics (NCHS) under the Centers for Disease Control and Prevention (CDC), designed to assess the health and nutritional status of the U.S. population. The survey adopts a multistage, stratified probability sampling design, enrolling approximately 5,000 noninstitutionalized civilian residents annually from 15 representative locations across the United States, with oversampling of ethnic minorities (e.g., African Americans, Mexican Americans) and older adults. Multidimensional data, including demographic characteristics, physical examinations, laboratory tests, and dietary interviews, were collected. All participants provided written informed consent upon enrollment. The study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement, which conforms to reporting standards for observational epidemiological research.

To maximize the sample size, data from ten survey cycles spanning 1999–2018 were synthesized, yielding an initial sample of 101,316 individuals. Participants were excluded stepwise on the basis of the following criteria: (1) did not have a diagnosis of diabetes (n = 91,748); (2) had missing data on the urinary albumin‒creatinine ratio (UACR) or estimated glomerular filtration rate (eGFR) (n = 1,182); and (3) had missing data on the DI‒GM index (n = 2,826). Ultimately, 5,560 adult patients with diabetes were included in the analysis. A detailed flowchart of the participant selection process is presented in Fig. 1.

Fig. 1.

Study population screening flowchart

Definitions of diabetes mellitus and DKD

The diagnosis of diabetes mellitus was based on the criteria established by the American Diabetes Association (ADA) [35] and referenced questionnaire data from the National Health and Nutrition Examination Survey (NHANES). An individual was defined as having diabetes mellitus if they met any of the following criteria: (1) fasting blood glucose (FBG) ≥ 126 mg/dL; (2) glycated hemoglobin (HbA1c) ≥ 6.5%; (3) 2-hour blood glucose concentration ≥ 200 mg/dL during a 75 g oral glucose tolerance test (OGTT); (4) self-reported diagnosis of diabetes mellitus; and (5) self-reported use of insulin or other antidiabetic medications.

The diagnosis of DKD was based on markers of renal injury and functional indicators, strictly adhering to the guidelines of the American Diabetes Association and the National Kidney Foundation and conforming to internationally recognized standards: urinary albumin-to-creatinine ratio (UACR) ≥ 30 mg/g or estimated glomerular filtration rate (eGFR) < 60 mL·min⁻¹·1.73 m⁻² [36, 37]. Since the NHANES does not directly measure the glomerular filtration rate, the eGFR was calculated via the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation, with serum creatinine values standardized and calibrated via isotope dilution mass spectrometry (IDMS), and was modeled by incorporating variables such as age, sex, and ethnicity [38]. This definition is consistent with that used in multiple previously published NHANES studies, encompassing the microvascular injury characteristics of diabetic glomerulosclerosis, and is widely applied for the identification of DKD in cross-sectional studies [39, 40].

Definition of the DI - GM

The dietary index for the gut microbiota (DI-GM) is an a priori scoring system developed by Kase et al. (2024) and was constructed through a systematic review of gut microbiota intervention studies [20]. This index includes 14 food components categorized into beneficial components (fermented dairy products, chickpeas, soybeans, whole grains, dietary fiber, cranberries, avocados, broccoli, coffee, and green tea) and unfavorable components (red meat, processed meat, refined grains, and a high-fat diet). On the basis of 24-hour dietary recall data, intake levels were stratified by sex-specific medians: 1 point was assigned for each beneficial component with intake ≥ the median, and 1 point was assigned for each unfavorable component with intake ≤ the median. In this study, the calculation of the DI-GM strictly followed the component definitions and scoring rules outlined in the original literature. The scores of individual components were summed to form a composite index ranging from 0 to 14, where higher scores indicate a diet more conducive to gut microbiota diversity.

Covariates

In accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines [41] and based on biological plausibility, potential confounding factors were identified and adjusted in this study. These covariates were categorized into three domains: demographic and socioeconomic characteristics, lifestyle factors, and clinical and laboratory measurements.

Demographic and socioeconomic characteristics

Data regarding age (continuous), sex (male or female), and ethnicity were obtained through standardized household interviews. Ethnicity was categorized as non-Hispanic White, non-Hispanic Black, Mexican American, and other ethnicities. Socioeconomic indicators included educational attainment (below high school, high school or equivalent, and above high school), marital status (married/cohabiting, widowed/divorced/separated, and never married), and the family poverty-income ratio (PIR). Following established protocols, PIR was stratified into three levels: low (< 1.3), middle (1.3–3.5), and high (≥ 3.5) [42].

Lifestyle factors

Lifestyle variables included dietary quality, physical activity, smoking status, and alcohol consumption. Dietary intake was assessed using the average of two 24-hour dietary recall interviews; the first was conducted in person at the Mobile Examination Center (MEC), and the second was completed via telephone 3 to 10 days later. Overall dietary quality was quantified via the Healthy Eating Index-2015 (HEI-2015), a 100-point scale assessing adherence to the Dietary Guidelines for Americans [43]. Physical activity was evaluated using the Global Physical Activity Questionnaire (GPAQ). Total physical activity was calculated as Metabolic Equivalent of Task (MET) minutes per week [44]. Smoking status was classified into three categories: never smokers (< 100 cigarettes in lifetime), former smokers (≥ 100 cigarettes but quit), and current smokers (≥ 100 cigarettes and still smoking). Alcohol consumption was classified into five distinct categories: (1) Never drinkers, defined as individuals who had fewer than 12 drinks in their lifetime; (2) Former drinkers, defined as those who had consumed 12 or more drinks in any single year or in their lifetime but had not consumed alcohol in the past year; (3) Mild drinkers, defined as current drinkers consuming up to 1 drink per day for females or up to 2 drinks per day for males; (4) Moderate drinkers, defined as current drinkers consuming 2 drinks per day for females or 3 drinks per day for males, or those reporting binge drinking on 2 to 4 days per month; and (5) Heavy drinkers, defined as current drinkers consuming 3 or more drinks per day for females or 4 or more drinks per day for males, or those reporting binge drinking (defined as 5 or more drinks within a two-hour period) on 5 or more days per month.

Clinical and laboratory measurements

Anthropometric measurements, including height and weight, were collected by trained health technicians in the MEC to calculate Body Mass Index (BMI). Glycated hemoglobin (HbA1c) was measured using high-performance liquid chromatography (HPLC) and certified by the National Glycohemoglobin Standardization Program (NGSP). Hypertension was defined as meeting any of the following: a self-reported physician diagnosis, current use of antihypertensive medication, systolic blood pressure ≥ 130 mmHg, or diastolic blood pressure ≥ 80 mmHg [45]. All variable definitions and laboratory procedures were consistent with NHANES official documentation and previously validated studies.

Statistical analysis

All analyses in this study accounted for the complex multistage sampling design of the National Health and Nutrition Examination Survey (NHANES) and were weighted in accordance with the methods recommended on the official NHANES website, ensuring that the study results accurately reflected the characteristics of the U.S. adult population.

First, differences in baseline characteristics were compared between groups stratified by the presence or absence of DKD. For continuous variables, survey-weighted linear regression was used to calculate P values for comparing mean differences between the two groups; for categorical variables, survey-weighted chi-square tests were employed to compute P values for evaluating distributional differences across groups. To explore the association between DI-GM and DKD risk in depth, three multivariable weighted logistic regression models were constructed: the crude model was used to analyze the unadjusted raw association between DI-GM and DKD without any covariate adjustment; Model I was further adjusted for sex and age; Model II was comprehensively adjusted for a set of potential confounding factors, including age, sex, height, weight, glycated hemoglobin (HbA1c), ethnicity, educational attainment, marital status, PIR, alcohol consumption status, smoking status, and a history of hypertension. Additionally, on the basis of previously established grouping criteria [46], DI-GM was categorized into three groups (low: 0–4; moderate: 5–6; high: ≥7), and a trend test was conducted to assess the trend of the association between DI-GM and DKD risk with increasing DI-GM levels. To ensure the stability and reliability of the multivariable logistic regression models, multicollinearity screening was performed for all covariates included in the fully adjusted model (Model II). Multicollinearity was assessed using the Variance Inflation Factor (VIF). In this study, a VIF value greater than 5 was pre-specified as the threshold for indicating significant multicollinearity among the independent variables. Our results demonstrated that all adjusted covariates, including demographic, socioeconomic, and clinical factors, yielded VIF values well below the threshold of 5, suggesting that the regression estimates were not substantially influenced by multicollinearity. The detailed VIF values for each variable are provided in Supplementary Table 1.

Given that the trend test suggested a potential nonlinear association between DI-GM and DKD risk, restricted cubic spline (RCS) analysis was further performed to fit the dose‒response curve of DI-GM versus DKD risk, with adjustment for all covariates in Model II. The significance of nonlinearity was determined via the likelihood ratio test (P-nonlinear < 0.05). In addition, in response to the nonlinear association indicated by the RCS, a two-stage linear regression model was used to identify the inflection point: the optimal inflection point position was determined via iterative likelihood statistics, and a piecewise logistic regression model was constructed to verify the heterogeneity of associations on both sides of the inflection point.

To evaluate potential effect modification, subgroup analyses were conducted by stratifying variables, including age (dichotomized at 60 years: ≥60 years vs. <60 years), sex, body mass index (BMI), ethnicity (reclassified as non-Hispanic White vs. other ethnicities), educational attainment (below high school, high school or equivalent, above high school), marital status (married/cohabiting, widowed/divorced/separated, never married), PIR (low, moderate, high), alcohol consumption status (current drinkers vs. never/former drinkers), and smoking status (current smokers vs. never/former smokers). Specifically, BMI was dichotomized at 30 kg/m² (≥ 30 vs. <30) to stratify participants by obesity status according to the standard definitions for the U.S. population [47]. Covariates in Model II were adjusted for in these analyses (except for the stratification variable itself when it served as the effect modifier). The statistical significance of the interaction terms (DI-GM × subgroup variable) was assessed via weighted Wald tests.

To verify the robustness of the study results, multiple imputation was used to handle missing data, generating 4 imputed datasets, and the analyses presented in Table 2 were repeated. To further validate the robustness of the threshold effect observed in our primary analysis, we conducted a series of rigorous sensitivity analyses among participants with DI-GM ≥ 5 (Supplementary Table 3). Four additional adjustment models were constructed: Sensitivity-1 incorporated physical activity scores into Model II (with missing values treated as an independent category to preserve sample size); Sensitivity-2 further adjusted for hyperlipidemia status; Sensitivity-3 accounted for the use of medications, including antihypertensive, glucose-lowering, and lipid-lowering agents; and Sensitivity-4 simultaneously adjusted for all variables included in Sensitivity-1 to Sensitivity-3.

Table 2.

Logistic regression analyses for the associations between DI-GM and DKD in different models

	OR (95%CI)
Exposure	Crude Model	Adjust I	Adjust II
DI-GM	0.96 (0.93, 0.99)	0.94 (0.91, 0.97)	0.96 (0.92, 0.99)
DIGM (subgroup)
Low	1.0(Ref)	1.0(Ref)	1.0(Ref)
Moderate	1.04 (0.93, 1.17)	1.00 (0.88, 1.13)	1.08 (0.94, 1.25)
High	0.85 (0.72, 1.00)	0.75 (0.64, 0.89)	0.82 (0.67, 0.99)
p-trend	0.1706	0.0064	0.1896

Abbreviations OR odds ratio

The crude model was adjusted for none

Model I adjusts for age and sex

Model II adjusts for age, sex, height, weight, HbA1c, ethnicity, education, marital status, PIR, drinking status, smoking status and hypertension history

All analyses were performed via R software (version 4.2.0), and key steps were reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. A two-tailed P value < 0.05 was considered statistically significant.

Results

Baseline characteristics

The baseline characteristics of the 5,560 study participants, stratified by the presence of DKD, are summarized in Table 1. A total of 3,445 participants were classified as non-DKD, while 2,115 were identified as having DKD. Compared to the non-DKD group, participants with DKD were significantly older (65.37 ± 13.05 vs. 57.14 ± 14.34 years; P < 0.001) and exhibited poorer glycemic control, as indicated by higher HbA1c levels (7.49 ± 1.90% vs. 7.05 ± 1.59%; p < 0.001). While body weight did not differ significantly between the two groups (P = 0.304), participants in the DKD group were slightly shorter (P = 0.010). Notably, lifestyle and dietary indicators showed marked disparities: the DKD group had significantly lower physical activity scores (median: 1,440 vs. 1,560 MET-min/week; P = 0.038) and lower HEI-2015 scores (53.88 ± 12.74 vs. 55.29 ± 13.40; P < 0.001). Furthermore, the DI-GM was significantly lower in the DKD group (4.70 ± 1.65 vs. 4.81 ± 1.72; P = 0.019), suggesting an association between poorer gut-microbiota-related dietary quality and DKD.

Table 1.

Baseline characteristics of participants stratified by the presence of DKD

	Non-DKD	DKD	P value
Participants	3445	2115
Age, years	57.14 (14.34)	65.37 (13.05)	< 0.001
Weight, kg	89.78 (23.50)	89.11 (23.49)	0.304
Height, cm	166.25 (10.40)	165.52 (10.01)	0.010
HbA1c, %	7.05 (1.59)	7.49 (1.90)	< 0.001
Physical activity score	1560.00 (600.00-4290.00)	1440.00 (560.00-3795.00)	0.038
HEI-2015 score	55.29 (13.40)	53.88 (12.74)	< 0.001
DI-GM	4.81 (1.72)	4.70 (1.65)	0.019
Sex			0.070
Female	1689 (49.03%)	984 (46.52%)
Male	1756 (50.97%)	1131 (53.48%)
Ethnicity			< 0.001
Non-Hispanic White	1153 (33.47%)	845 (39.95%)
Non-Hispanic Black	835 (24.24%)	507 (23.97%)
Mexican American	652 (18.93%)	358 (16.93%)
Other races	805 (23.37%)	405 (19.15%)
Education			0.008
Below high school	1137 (33.03%)	774 (36.65%)
High school or equivalent	780 (22.66%)	485 (22.96%)
Above high school	1525 (44.31%)	853 (40.39%)
Marital status			< 0.001
Married/cohabiting	2138 (63.42%)	1169 (55.59%)
Widowed/divorced/separated	887 (26.31%)	774 (36.80%)
Never married	346 (10.26%)	160 (7.61%)
PIR			< 0.001
Low	1111 (35.55%)	748 (39.02%)
Medium	980 (31.36%)	671 (35.00%)
High	1034 (33.09%)	498 (25.98%)
Smoking status			< 0.001
Never	1776 (52.42%)	1002 (47.65%)
Former	1045 (30.84%)	786 (37.38%)
Now	567 (16.74%)	315 (14.98%)
Drinking status			< 0.001
Never	526 (17.06%)	372 (19.58%)
Former	685 (22.22%)	564 (29.68%)
Mild	1058 (34.32%)	580 (30.53%)
Moderate	366 (11.87%)	164 (8.63%)
Heavy	448 (14.53%)	220 (11.58%)
Hypertension history			< 0.001
No	1262 (36.69%)	418 (19.80%)
Yes	2178 (63.31%)	1693 (80.20%)

Abbreviations DKD Diabetic kidney disease, HbA1c Glycated hemoglobin, HEI-2015 Healthy Eating Index-2015, DI-GM Dietary index for the gut microbiota, PIR Poverty income ratio

Significant differences were also observed in demographic and socioeconomic factors. The DKD group comprised a higher proportion of non-Hispanic White individuals (39.95% vs. 33.47%) and a lower proportion of those with education above high school (40.39% vs. 44.31%; P = 0.008). Socioeconomic vulnerability was more pronounced in the DKD group, which had a higher prevalence of individuals in the low PIR category (39.02% vs. 35.55%; P < 0.001) and a higher percentage of widowed, divorced, or separated individuals (36.80% vs. 26.31%; P < 0.001). Regarding lifestyle behaviors, smoking and drinking statuses varied significantly (P < 0.001), with a higher prevalence of former smokers (37.38%) and former drinkers (29.68%) in the DKD group. Lastly, hypertension was markedly more prevalent among participants with DKD (80.20% vs. 63.31%; P < 0.001), highlighting its critical role as a comorbidity.

Associations between the DI-GM and DKD

Logistic regression analyses (Table 2) revealed a significant inverse association between DI-GM and DKD. In the crude model, each unit increase in DI-GM was associated with a 4% reduction in the odds of DKD (OR = 0.96, 95% CI: 0.93–0.99). This association remained significant after adjusting for age and sex (Model I: OR = 0.94, 95% CI: 0.91–0.97) and further adjustment for additional covariates, including anthropometric, metabolic, sociodemographic, and lifestyle factors (Model II: OR = 0.96, 95% CI: 0.92–0.99). When DI-GM was categorized into subgroups (low, moderate, and high), a significant protective effect was observed in the high DI-GM group in Model I (OR = 0.75, 95% CI: 0.64–0.89; p-trend = 0.0064). However, in the fully adjusted model (Model II), the trend test lost statistical significance (p-trend = 0.1896), although the high DI-GM group still exhibited a significant protective effect (OR = 0.82, 95% CI: 0.67–0.99). The lack of a significant linear trend (p-trend = 0.1896) suggests that the relationship between DI-GM and DKD may not follow a simple linear pattern but rather a nonlinear association. These findings suggest that higher DI-GM scores may be associated with a reduced risk of DKD, particularly after basic demographic factors are accounted for. However, the attenuation of significance in the fully adjusted model indicates that the relationship may be partially confounded by clinical and lifestyle variables.

Dose‒response relationship between DI-GM and DKD

The nonlinear relationship between the DI-GM and DKD was further characterized via restricted cubic splines and a two-piecewise linear regression model (Fig. 2). The spline curve demonstrated a significant overall association (P for overall < 0.001) with notable nonlinearity (P for nonlinearity = 0.018), revealing a threshold effect at DI-GM = 5. Further two-stage linear regression analysis (Table 3) revealed that, below the inflection point, there was no significant association between DI-GM and DKD (OR = 1.04, 95% CI: 0.97–1.11; P = 0.31). However, above this threshold, each unit increase in DI-GM was associated with a 12% reduction in DKD risk (OR = 0.88, 95% CI: 0.81–0.95; P < 0.01). The log-likelihood ratio test confirmed the superiority of the two-piecewise model over linear regression (P < 0.01), supporting the existence of a nonlinear dose‒response relationship. These findings suggest that the protective effect of DI-GM on DKD becomes clinically meaningful only when dietary quality exceeds a certain threshold, providing important insights for targeted dietary interventions.

Fig. 2.

Dose‒response relationships between the DI-GM and DKD

Table 3.

Results of the two-piecewise linear regression model

Outcome: diabetic kidney disease	OR (95%CI)	P value
Fitting model by standard linear regression	0.96 (0.92, 1.00)		0.04
Inflection points of DI-GM	5
DI-GM < 5	1.04 (0.97, 1.11)		0.31
DI-GM ≥ 5	0.88 (0.81, 0.95)		< 0.01
P for log-likelihood ratio test	< 0.01

The model was adjusted as Model II

Subgroup analysis

Subgroup analyses were conducted to evaluate potential effects on the association between DI-GM and DKD (Table 4). The inverse association between DI-GM and DKD remained consistent across all the examined subgroups, with no statistically significant interactions observed (all P-interaction > 0.05). Notably, the protective effect appeared slightly stronger in non-Hispanic ethnic minorities (OR = 0.93, 95% CI: 0.89–0.98) than in non-Hispanic Whites (OR = 1.00, 95% CI: 0.94–1.06), although the interaction test did not reach statistical significance (P-interaction = 0.1069). Similarly, the association tended to be more pronounced in individuals with a BMI < 30 kg/m² (OR = 0.94, 95% CI: 0.89–1.00) than in those with a BMI ≥ 30 kg/m² (OR = 0.97, 95% CI: 0.92–1.02) and in high-income individuals (OR = 0.94, 95% CI: 0.87–1.00) than in lower income individuals. However, none of these subgroup differences were statistically significant. The consistency of the associations across demographic and clinical strata suggests that the protective effect of DI-GM on DKD is robust and not substantially modified by common population characteristics.

Table 4.

Association between DI-GM and DKD stratified by age, sex, BMI, ethnicity, education, marital status, PIR, drinking status, smoking status and hypertension history

Subgroup	No. of participants	Adjusted OR (95%CI)	P-interaction
Age			0.7609
< 60 years	2451	0.97 (0.91, 1.04)
≥ 60 years	3109	0.96 (0.92, 1.01)
Sex			0.4934
Female	2673	0.95 (0.89, 1.00)
Male	2887	0.97 (0.92, 1.03)
BMI			0.4960
< 30 kg/m2	2314	0.94 (0.89, 1.00)
≥ 30 kg/m2	3169	0.97 (0.92, 1.02)
Ethnicity			0.1069
Non-Hispanic White	1998	1.00 (0.94, 1.06)
Other races	3562	0.93 (0.89, 0.98)
Education			0.3410
Below high school	1911	1.00 (0.93, 1.07)
High school or equivalent	1265	0.93 (0.86, 1.00)
Above high school	2378	0.95 (0.90, 1.01)
Marital status			0.9208
Married/cohabiting	3307	0.96 (0.91, 1.01)
Widowed/divorced/separated	1661	0.97 (0.90, 1.04)
Never married	506	0.94 (0.83, 1.08)
PRI			0.6820
Low	1859	0.97 (0.91, 1.04)
Medium	1651	0.97 (0.91, 1.04)
High	1532	0.94 (0.87, 1.00)
Drinking status			0.9494
Current	2147	0.96 (0.90, 1.02)
Never/past	2836	0.96 (0.91, 1.01)
Smoking status			0.4643
Current	4609	0.95 (0.91, 1.00)
Never/past	882	0.99 (0.90, 1.10)
Hypertension history			0.7834
No	1680	0.97 (0.89, 1.05)
Yes	3871	0.96 (0.92, 1.00)

All abbreviations are listed in Table 1

Sensitivity analysis

To ensure the robustness of our findings, we conducted a two-fold sensitivity analysis. First, multiple imputation was employed to account for missing covariates across four independently imputed datasets; the results (Table 5) were highly consistent with the primary analysis, with a stable OR range of 0.95–0.96 (95% CI: 0.92–0.99) per unit increase in the DI-GM, suggesting that missing data did not substantially bias our estimates. Second, to further validate the observed threshold effect, additional sensitivity analyses were performed specifically among participants with DI-GM ≥ 5 (Supplementary Table 3). After sequentially adjusting for physical activity (with missing values treated as an independent category), hyperlipidemia status, and the use of medications (including antihypertensive, glucose-lowering, and lipid-lowering agents), the inverse association between the DI-GM and DKD remained remarkably stable. In these advanced models, a high DI-GM continued to be associated with a 14% reduction in DKD risk (continuous OR: 0.86; all P < 0.01), and the trend tests remained statistically significant across all adjustment levels (P -trend range: 0.0066–0.0082). Collectively, these analyses reinforce the conclusion that the DI-GM on DKD is robust and independent of potential biases from missing data or residual confounding from clinical and lifestyle factors.

Table 5.

Sensitivity analysis of the association between DI-GM and diabetic kidney disease

	OR (95% CI)
	Sensitivity-1	Sensitivity-2	Sensitivity-3	Sensitivity-4
DI-GM	0.95 (0.92, 0.99)	0.96 (0.92, 0.99)	0.95 (0.92, 0.99)	0.95 (0.92, 0.99)
DI-GM (subgroup)
Low	1.0(Ref)	1.0(Ref)	1.0(Ref)	1.0(Ref)
Moderate	1.04 (0.92, 1.18)	1.05 (0.92, 1.19)	1.04 (0.92, 1.18)	1.04 (0.92, 1.18)
High	0.81 (0.68, 0.97)	0.82 (0.68, 0.97)	0.81 (0.68, 0.96)	0.81 (0.68, 0.96)
p-trend	0.090	0.103	0.084	0.081

(1) Sensitivity-1 to sensitivity‐4 were repeated analyses using data after multiple imputations of missing covariates

The NHANES data were screened on the basis of the study design to select eligible participants

All abbreviations are listed in Table 1. Restricted cubic splines were adjusted for the Model II model adjusted for age, sex, height, weight, HbA1c, ethnicity, education, marital status, PIR, drinking status, smoking status and hypertension history

Discussion

On the basis of the nationally representative sample from the NHANES, this study is the first to systematically investigate the association between the DI-GM and the risk of DKD. The results indicated a significant negative correlation between DI-GM and DKD risk—higher DI-GM scores were associated with lower DKD risk, and this association remained significant after adjusting for multiple confounding factors. Further RCS analysis revealed a nonlinear threshold effect: when DI-GM ≥ 5, each 1-unit increase in the score was associated with a 12% reduction in DKD risk, whereas no significant association was observed when DI-GM < 5. This finding not only provides a new quantifiable target for the nutritional prevention of DKD but also further expands the epidemiological evidence base for the “diet–gut microbiota–renal health” axis.

As a novel dietary assessment tool focused on gut microbiota modulation, DI-GM’s core design logic is derived from the systematic integration of 14 dietary components known to influence the composition and function of the gut microbiota. The selection of these components is based on evidence of “diet‒gut microbiota” associations confirmed in previous longitudinal intervention studies [20]. For example, beneficial components incorporated in DI-GM—such as fermented dairy products, dietary fiber, and whole grains—have been validated in multiple studies to significantly increase gut microbiota α diversity and enrich short-chain fatty acid (SCFA)-producing bacteria (e.g., Bifidobacterium and Lactobacillus) [48–52]. In contrast, the restriction of unfavorable components (e.g., processed meat and high-sugar foods) can reduce the abundance of proinflammatory gut microbiota (e.g., certain Firmicutes species) and the production of harmful metabolites such as trimethylamine-N-oxide (TMAO) [20, 50]. Theoretically, by quantifying “microbiota-friendly” dietary patterns, DI-GM serves as a standardized tool for evaluating the potential impact of diet on gut microecology, laying a plausible mechanistic foundation for investigating its association with DKD risk.

In recent years, the role of the gut microbiota in the development and progression of diseases has been extensively explored. Emerging studies have demonstrated that gut microbiota dysbiosis is closely linked to the onset and progression of DKD [12, 15, 16], while improving dietary structure can effectively reduce DKD risk [53–56]. Our results align with this emerging evidence, suggesting that dietary patterns modulating the composition of the gut microbiota may play a role in metabolic and renal health. For example, previous studies have reported that diets rich in dietary fiber, fermented foods, and polyphenols—key components of DI-GM—are associated with improved glycemic control and reduced systemic inflammation, both of which are involved in the pathogenesis of DKD [57–60]. Additionally, multiple prior studies have shown significant negative correlations between DI-GM and conditions such as chronic Kidney disease and Kidney stones. However, to our knowledge, this is the first study to quantify the relationship between a gut microbiota-specific dietary index and DKD risk using a large population-based dataset. The threshold effect we identified (DI-GM = 5) provides new insights into the potential dose‒response relationship between diet and DKD risk, which has not been previously reported. This finding corroborates the core conclusions of prior studies on diet‒chronic disease associations while further expanding the evidence for the “diet‒gut microbiota‒renal health” axis. For example, a NHANES-based study by Liu et al. [61] showed that dietary flavonoids (especially flavan-3-ols and flavones) can reduce DKD risk; these components (e.g., catechins in green tea and anthocyanins in berries) are important “beneficial dietary factors” in the DI-GM, and their regulatory effects on the gut microbiota (e.g., promoting Bifidobacterium growth and inhibiting pathogen colonization) may serve as key mediating pathways for their renoprotective effects [20]. Notably, evidence from studies on DI-GM and metabolic syndrome or diabetes also supports such negative associations, with RCS analyses indicating mostly linear correlations. However, conflicting evidence exists: Niu et al. reported that DI-GM ≥ 3 was required to significantly reduce metabolic syndrome risk in a cross-sectional study [30], and similarly, Song et al. [62] reported a threshold effect of “DI-GM ≥ 4 for significant reduction in constipation risk” in their study on constipation. These results are highly consistent with the nonlinear association observed in our study, suggesting that the protective effect of DI-GM in certain diseases may depend on an “effective modulation threshold” of diet on the gut microbiota—only when the cumulative intake of beneficial dietary components reaches a critical level can gut microbiota dysbiosis in pathological states be reversed, thereby exerting health-protective effects.

Compared with traditional dietary assessment tools (e.g., the Mediterranean diet score (DASH) diet score), the DI-GM is unique in its core design logic centered on “gut microbiota modulation,” which makes it more useful for elucidating the mechanism underlying the association between diet and DKD [20, 63]. For example, while the Mediterranean diet has been proven to reduce the risk of diabetes mellitus (DM) [64–66] and chronic Kidney disease [67, 68], its scoring criteria primarily rely on macronutrients (e.g., olive oil, fish intake) and food group classifications, lacking quantification of specific effects on the gut microbiota [20]. In contrast, by explicitly incorporating components that directly act on the gut microbiota (e.g., fermented dairy products, dietary fiber, and coffee), DI-GM more accurately reflects the “diet‒microbiota‒host Kidney” interaction [20]. Mechanistically, the protective effect of DI-GM on DKD may be mediated through multiple gut microbiota-dependent pathways, which is highly consistent with conclusions from previous basic research. First, “microbiota-friendly” diets corresponding to high DI-GM scores can promote the production of SCFAs in the gut [69]. As key metabolites of the gut microbiota, SCFAs can regulate renal hemodynamics via G protein-coupled receptors (GPR41/43) to reduce glomerular filtration pressure [70, 71] and inhibit histone deacetylase (HDAC) activity to alleviate oxidative stress and apoptosis in intrinsic renal cells (e.g., mesangial cells and podocytes) [72, 73]. Second, the restriction of unfavorable dietary components in DI-GM can reduce the production of nephrotoxic metabolites such as TMAO and indoxyl sulfate [74]. These metabolites have been confirmed to activate the renin‒angiotensin‒aldosterone system (RAAS) and induce endoplasmic reticulum stress in renal tubular epithelial cells, exacerbating renal fibrosis [75–78]; thus, increased DI-GM scores may indirectly delay DKD progression by reducing the sources of such harmful metabolites.

Notably, the threshold effect of DI-GM (≥ 5 points) observed in this study holds important clinical translational value, which may be underpinned by biological bases such as a “gut microbiota diversity threshold” or a “metabolite concentration threshold.” When DI-GM < 5, insufficient intake of beneficial dietary components may fail to reverse preexisting gut microbiota dysbiosis in diabetes (e.g., reduced abundance of beneficial bacteria, decreased microbiota diversity); at this stage, the gut microbiota remains in a “proinflammatory phenotype,” and its metabolites and immunomodulatory effects are insufficient to counteract renal injury. In contrast, when DI-GM ≥ 5, the cumulative intake of beneficial dietary components reaches a critical level, sufficient to reshape the gut microbiota structure (e.g., increasing the proportion of SCFA-producing bacteria and restoring microbiota diversity); the protective effect then exceeds the injury threshold, thereby exerting a significant protective effect against DKD. These findings suggest that the DI-GM threshold can serve as a quantifiable target for nutritional interventions in DKD patients, providing a reference for clinicians to develop individualized dietary plans.

Strengths and limitations

Nevertheless, this study has several limitations that require caution when interpreting the results. First, the cross-sectional design prevents the establishment of a causal relationship between the DI-GM and DKD. Although we controlled for potential confounders (e.g., age, sex, BMI, hypertension) through multivariable adjustment, residual confounding factors (e.g., unmeasured genetic factors, baseline gut microbiota status) may still exist. Future studies should use prospective cohort designs or randomized controlled trials (e.g., DI-GM-based dietary intervention trials) to further validate causal inferences. Second, DI-GM was calculated on the basis of 24-hour dietary recall data from the NHANES, which may be subject to recall bias. Additionally, the NHANES database lacks specific intake data for original DI-GM components such as chickpeas and cranberries, potentially leading to slight underestimation of DI-GM scores. Although sensitivity analyses revealed that missing data did not significantly affect the results, future studies could improve the accuracy of DI-GM assessment via more detailed dietary records (e.g., 7-day food diaries) or food frequency questionnaires. Third, this study did not directly measure participants’ gut microbiota composition or metabolites (e.g., SCFAs and TMAO), precluding direct verification of the mediating role of the gut microbiota. Subsequent studies could integrate fecal microbiota sequencing and serum metabolomics data to further elucidate the “DI-GM-gut microbiota-DKD” mechanistic pathway. Fourth, although physical activity is a known determinant of metabolic health, it was not adjusted for in our primary multivariable models due to a substantial proportion of missing data (39.06%). Including this variable would have significantly reduced the statistical power and potentially introduced selection bias. While multiple imputation was considered, the high degree of missingness rendered it inappropriate for the primary analysis. Consequently, residual confounding by physical activity cannot be entirely excluded. Future studies with more complete lifestyle records are needed to further clarify these associations. Fifth, the duration of diabetes, a critical determinant of microvascular outcomes, was not included in our multivariable analysis. This decision was necessitated by the data structure of NHANES, which provided extremely limited information on the age at diabetes diagnosis, with a missingness rate of approximately 90% across the included study population. Adjusting for this variable would have led to an unacceptable loss of statistical power and potential selection bias. Although we adjusted for age and HbA1c to partially account for the effects of cumulative metabolic burden and glycemic status, the absence of precise disease duration remains a limitation. Future longitudinal studies with more complete clinical histories are needed to confirm the independent association between DI-GM and DKD risk while controlling for disease duration. Sixth, the use of specific renoprotective medications, such as renin-angiotensin system (RAS) inhibitors, sodium-glucose cotransporter-2 (SGLT2) inhibitors, and mineralocorticoid receptor (MR) antagonists, was not adjusted for in the primary models due to significant data missingness and inconsistent recording across NHANES cycles. However, we performed a sensitivity analysis adjusting for the use of antihypertensive, glucose-lowering, and lipid-lowering medications, and the core findings remained robust. Nonetheless, the potential for residual confounding from unmeasured pharmacological factors cannot be entirely ruled out. Future prospective studies with more granular and longitudinal medication data are warranted to confirm our observations. Finally, the diagnosis of DKD in this study was based on single-point measurements of UACR and eGFR, which may not strictly adhere to the clinical requirement of persistent abnormalities for at least three months. Due to the cross-sectional nature of the NHANES, we could not account for transient fluctuations in albuminuria or eGFR caused by acute illness, heavy exercise, or medication, potentially leading to an overestimation of DKD prevalence. Furthermore, since renal biopsy data were unavailable, we could not distinguish DKD from non-diabetic kidney diseases. However, the use of calibrated eGFR and UACR remains the standard approach in large-scale epidemiological research and provides meaningful insights into the association between dietary patterns and renal health in the diabetic population.

Despite these limitations, this study still provides new perspectives and evidence for the nutritional prevention of DKD. From a public health perspective, the “DI-GM ≥ 5” threshold proposed in this study can serve as a simple and feasible dietary guideline for the primary prevention of DKD—by promoting dietary patterns rich in dietary fiber, fermented foods, and whole grains, diabetic patients can help maintain DI-GM above the protective threshold, thereby reducing DKD risk. From a clinical practice perspective, the DI-GM can act as a convenient dietary assessment tool to assist clinicians in developing individualized nutritional plans for diabetic patients. Particularly for patients with well-controlled traditional risk factors (e.g., blood glucose and blood pressure) who still face DKD risk, DI-GM assessment may provide an additional basis for risk stratification.

Conclusion

In this cross-sectional study based on a nationally representative sample, we demonstrated for the first time a significant inverse association between the DI-GM and the risk of DKD. Importantly, a clear threshold effect of DI-GM on DKD protection was observed, suggesting that maintaining a DI-GM score above 5 may serve as a practical dietary goal to reduce the risk of DKD onset in adult patients with diabetes. These findings provide valuable insights for developing gut microbiota-targeted dietary interventions aimed at preventing DKD.

Abbreviations

ADA

American diabetes association

BMI

Body mass index

CKD

Chronic kidney disease

CKD-EPI

Chronic kidney disease epidemiology collaboration

CI

Confidence interval

DI-GM

Dietary index for gut microbiota

DKD

Diabetic kidney disease

eGFR

Estimated glomerular filtration rate

FBG

Fasting blood glucose

HbA1c

Glycated hemoglobin

HDAC

Histone deacetylase

HPLC

High-performance liquid chromatography

IDMS

Isotope dilution mass spectrometry

KDIGO

Kidney disease: improving global outcomes

NHANES

National health and nutrition examination survey

NGSP

National glycohemoglobin standardization program

OGTT

Oral glucose tolerance test

OR

Odds ratio

PIR

Poverty‒income ratio

RAAS

Renin–Angiotensin–Aldosterone system

VIF

Variance inflation factor

RCS

Restricted cubic spline

SCFAs

Short-chain fatty acids

STROBE

Strengthening the reporting of observational studies in epidemiology

TMAO

Trimethylamine-N-oxide

UACR

Urinary Albumin-to-Creatinine ratio

Authors’ contributions

JP-L was responsible for conceptualization, methodology, supervision, and project administration. JJ-Q, HT-Z, and JY-F contributed to writing the original draft. JE-L, S-X, HX Z, YY-Z, SQ-Y, LX-F, YJ-Z, and J-Z participated in data collection and analysis. All the authors reviewed and edited the manuscript and approved the final version. JJ-Q, HT-Z, and JY-F contributed equally to this work and share first authorship.

Funding

This work was supported by the National Natural Science Foundation of China (Grant Nos. 82160162 and 81760150), the Key Research, Development Program of Jiangxi Province (20243BBI91008), the Project of the Second Affiliated Hospital of Nanchang University (2022efyA04) and the Jiangxi Province Key Laboratory of Molecular Medicine (No. 2024SSY06231).

Data availability

The datasets analyzed during the current study are publicly available in the National Health and Nutrition Examination Survey (NHANES) repository. Data can be accessed through the official website of the Centers for Disease Control and Prevention (CDC) at: [https://www.cdc.gov/nchs/nhanes/index.htm]

Declarations

Ethics approval and consent to participate

The National Center for Health Statistics (NCHS) Research Ethics Review Board approved the research protocols, and all participants provided written informed consent. All the methods included in this study were in accordance with the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.