Comparative analysis of dietary pattern indices and their associations with chronic kidney disease: a comprehensive analysis of NHANES data (2000–2020)

Abstract

Background

Although dietary patterns are recognized as modifiable risk factors for chronic kidney disease (CKD), comparative evidence on the differential impacts of commonly used dietary indices remains limited. This study aims to evaluate associations between four indices (Healthy Eating Index-2020, HEI-2020; alternative Mediterranean Diet Score, aMED; Dietary Approaches to Stop Hypertension, DASH; Dietary Inflammatory Index, DII) and CKD risk, and explore their population heterogeneity.

Methods

Utilizing cross-sectional data from NHANES (2000–2020), dietary scores were calculated for individuals with or without CKD. Logistic regressions estimated normalized odds ratio (ORs) per 25% scoring range increase. Predictive utility was assessed via marginal receiver operating characteristic (ROC) curves, and nonlinear associations were detected using restricted cubic splines (RCS). Subgroup analyses were conducted across different population characteristics. Component analyses were used to evaluate which components within each dietary index exert a significant effect on CKD risk.

Results

DASH (OR = 0.880, 95%CI: 0.812–0.954) and DII (OR = 1.099, 95%CI: 1.025–1.180) were significantly associated with CKD risk, only DII remained associated with CKD severity progression (OR = 1.264, 95%CI: 1.103–1.450). Dietary indices provided incremental utility second to comorbidities and age. Nonlinear analyses revealed that greater adherence to DASH/DII reduced CKD risk, with consistent results across subgroups of males, individuals over 65 years, Non-Hispanic Whites, both smokers and nonsmokers, family income-to-poverty ratio >3.5, and individuals with hypertension or without diabetes and cardiovascular diseases.

Conclusions

DASH and DII exhibited superior CKD risk discrimination versus other indices. Adopting dietary habits aligned with DASH/DII was most effective for reducing CKD risk in dietary interventions.

Introduction

Chronic kidney disease (CKD), characterized by progressive structural and functional renal impairment lasting for over three months, clinically manifests as proteinuria, edema, hypertension, and a reduced glomerular filtration rate (GFR) [1]. As a significant global public health issue, the prevalence of CKD has been steadily rising, with epidemiological studies estimating that approximately 10% of adults worldwide are affected by varying degrees of CKD [2,3]. CKD markedly increases the risks of renal failure, acute kidney injury, cardiovascular morbidity, and hospitalization, while also contributing to the growing burden of mortality related to non-communicable diseases [4,5].

The progression of CKD is influenced not only by non-modifiable factors such as age, gender, ethnicity, and genetic predisposition but also by modifiable metabolic risk factors such as hypertension, diabetes, and obesity. Among the modifiable variables, dietary pattern has emerged as a critical intervention target due to its easier adjustability and adaptability across populations [6]. Accumulating evidence underscores the dual role of dietary composition in both CKD pathogenesis and the management of its complications through mechanisms involving electrolyte homeostasis, inflammatory modulation, and oxidative stress [7]. Diets characterized by excessive animal fats, sodium, and phosphorus accelerate renal functional decline, whereas evidence-based nutritional interventions have been shown to effectively attenuate the deterioration of GFR [8]. Consequently, researchers have developed a variety of quantitative assessment systems to capture the complex associations between dietary patterns and kidney health.

In this study, we selected four dietary pattern scoring indices to compare their associations with CKD (including the Healthy Eating Index-2020 (HEI-2020), the alternative Mediterranean Diet Score (aMED), the Dietary Approaches to Stop Hypertension (DASH), the Dietary Inflammatory Index (DII)). The selection of these four dietary indices is grounded in their distinct conceptual frameworks and alignment with pathological mechanisms of CKD, including inflammation, metabolic dysregulation, and vascular dysfunction. For instance, the HEI-2020 quantifies adherence to the Dietary Guidelines for Americans (DGA) 2020–2025, emphasizing balanced intake of essential nutrient groups (e.g. vegetables, whole grains, dairy) that are critical for maintaining renal supportive nutrition [9]. The aMED highlights the consumption of olive oil, fish, and nuts, reflecting key principles of the Mediterranean diet; and these components may linked to anti-inflammatory and endothelial protective effects [10]. The DASH was originally developed for hypertension management but is particularly relevant here due to the high prevalence of hypertension in CKD populations; and DASH focuses on sodium restriction, potassium optimization, and alcohol moderation to improve metabolic balance and protect renal health [11]. The DII quantifies food components based on their pro-/anti-inflammatory potential [12], elucidating dietary modulation of chronic inflammation, which is a pivotal driver of renal fibrosis and disease progression. Together, these indices collectively provide diverse perspectives for assessing the associations between dietary patterns and CKD.

Although there is sufficient evidence linking dietary patterns to the occurrence and progression of CKD, comprehensive comparisons between commonly used dietary indices are still limited, which hindering the development of optimal dietary recommendations for reducing CKD risk. In this study, utilizing multi-cycle data from the National Health and Nutrition Examination Survey (NHANES, 2000–2020), we conducted a parallel assessment of four dietary indices (including HEI-2020, aMED, DASH, and DII) to evaluate their associations with CKD risk, and explore population heterogeneity across different subgroups, aiming to provide evidence-based precise dietary recommendations for clinical practice.

Methods

Study design and population

This study utilized publicly available data from NHANES between 2000 and 2020, focusing on individuals aged 18 years and older [13]. In the initial phase, structured household interviews were conducted to obtain detailed demographic information, including age, sex, ethnicity, educational level, smoking history, and the family income-to-poverty ratio (PIR). In the subsequent phase, participants who completed standardized biochemical and urine assessments provided valid measurements of urinary albumin-to-creatinine ratio (ACR) and serum creatinine levels. The classification of CKD was based on the criteria of estimated glomerular filtration rate (eGFR) and urinary ACR. Additionally, blood pressure, fasting plasma glucose, and body mass index (BMI) were measured to assess participants’ overall health status [14].

Dietary intake information was gathered using a two-stage 24-h recall method. The initial recall took place in person during the examination, while the follow-up recall was conducted via telephone within a period of 3 to 10 days [15]. To address day-to-day dietary variability, individual-level intake values for each food component were calculated as the mean of the two 24-h recalls, providing a more stable measure of habitual intake. The study initially included 128,809 participants; however, after applying exclusion criteria, the final analytical sample was reduced to 46,742 individuals, and the study’s flow chart for the inclusion and exclusion of participants was shown in Figure 1. Exclusions were made for participants with incomplete demographic profiles, missing critical covariates data (such as smoking status, BMI, eGFR, urinary ACR), and comorbidities (such as hypertension, diabetes mellitus, cardiovascular disease (CVD) and kidney failure), as well as those lacking dietary records. All participants signed written informed consent, and as the NHANES data is de-identified and available through public repositories, no additional ethical approval was required.

Figure 1.

The study’s flow chart for the inclusion and exclusion of participants.

Dietary pattern assessments

Four commonly used dietary indices [16], including HEI-2020, aMED, DASH index, and DII, were used. Prior to index computation, participants with implausible energy intake values were excluded to ensure data quality (the average daily calorie intake is beyond three standard deviations of the average). This exclusion procedure aimed to minimize the influence of extreme or inconsistent dietary records on index calculations, ensuring that the final dataset reflected biologically plausible intake patterns for subsequent analyses.

Among these, HEI-2020 was developed based on the DGA, reflecting the most recent iteration of the Healthy Eating Index (2020–2025 version) [9]. HEI-2020 consists of 13 dietary components, which include various food groups such as vegetables, fruits, whole grains, dairy products, protein sources (e.g. meat and legumes), and fats (e.g. monounsaturated and saturated fats). Each component is assigned a score from 0 to 10 or 0 to 5, depending on the deviation of actual intake from the recommended levels. The total score of HEI-2020 ranges from 0 to 100, with higher scores indicating greater adherence to healthy dietary guidelines.

The aMED is a modified version of the Mediterranean Diet Score, designed to simplify and adapt the core principles of the traditional Mediterranean diet [10]. It evaluates key dietary components, including vegetables (excluding potatoes and fried potato products), fruits, nuts, whole grains, legumes, fish, the ratio of monounsaturated to saturated fats, red and processed meats, and alcohol consumption. Participants receive one point for each component if their intake surpasses the population median, while those below the median receive zero points. The scoring for red and processed meats, as well as alcohol, is reversed, with lower intake earning one point and higher intake receiving zero points. The total aMED score ranges from 0 to 9, with higher scores reflecting closer adherence to Mediterranean dietary patterns.

The DASH was specifically developed to assess dietary patterns aimed at preventing and managing hypertension [11]. It evaluates adherence to a diet emphasizing high consumption of fruits, vegetables, nuts, legumes, low-fat dairy products, and whole grains, while promoting limited intake of sodium, sugar-sweetened beverages, and red and processed meats. The DASH score is calculated based on quintile distributions within the studied population, with each dietary component receiving a score between 1 and 5, resulting in a total score ranging from 8 to 40. The index places particular emphasis on key minerals such as sodium, potassium, calcium, and magnesium, as higher scores are typically associated with lower sodium intake and higher levels of these essential nutrients.

The DII is a dietary assessment tool designed to quantify the inflammatory potential of an individual’s diet [12]. It evaluates the effects of various food components, including both anti-inflammatory nutrients (e.g. antioxidants, vitamins, and minerals) and pro-inflammatory factors (e.g. added sugars, saturated fats, and red meat). The calculation process of the DII is relatively complex, as it assigns scores to dietary components based on their established inflammatory impact. The index incorporates 45 food parameters, comparing each component’s intake with global reference values and assigning a score according to its overall effect on systemic inflammation. A positive DII score indicates a diet rich in pro-inflammatory foods (e.g. red meat, high-sugar, and high-fat foods), whereas a negative score suggests a diet abundant in anti-inflammatory foods (e.g. fruits, vegetables, and whole grains). The DII score typically ranges from −9 (highly anti-inflammatory) to +8 (highly pro-inflammatory).

Outcome definitions

The CKD status was diagnosed through standardized biochemical and urinary examinations [17]. eGFR was calculated according to the CKD Epidemiology Collaboration (CKD-EPI) equation, with CKD diagnosis defined as either (a) eGFR <60 mL/min/1.73 m² or (b) urinary ACR >30 mg/g. Disease staging followed international guidelines, categorizing CKD into five progressive stages based on eGFR thresholds: Stage I (eGFR ≥90), Stage II (60≤ eGFR <90), Stage III (30≤ eGFR <60), Stage IV (15≤ eGFR <30), and Stage V (eGFR <15) [18,19]. Kidney failure was obtained according to the reported or self-admitted physician diagnoses, which was assessed by the following question from the questionnaire of Kidney Conditions: “ever been told by a doctor or other health professional that (you/s/he) had weak or failing kidneys? Do not include kidney stones, bladder infections, or incontinence.” Any individuals confirmed as CKD at stage III or above, or CKD combined with kidney failure would be assigned to the moderate-severe CKD group, otherwise considered as mild CKD group.

Covariates assessment

A comprehensive set of covariates was collected as potential confounding factors, including demographic characteristics, lifestyle factors (smoking history), and health conditions (hypertension and diabetes). Demographic variables, obtained from household questionnaires, included age, sex, ethnicity, educational level, and the family PIR. BMI was derived from measured height and weight during physical examinations and categorized according to international standards: underweight (<18.5 kg/m²), normal weight (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²), and obesity (≥30 kg/m²) [20]. Smoking history was classified into ever-smokers (current or former smokers) and never-smokers.

Besides, hypertension was defined based on either self-reported physician diagnosis or clinical measurements, with systolic blood pressure ≥130 mmHg or diastolic blood pressure ≥80 mmHg meeting diagnostic criteria [21]. Similarly, diabetes mellitus was identified through self-reported diagnosis or laboratory-confirmed thresholds, specifically a fasting plasma glucose level ≥7.0 mmol/L or a glycated hemoglobin (HbA1c) level ≥6.5% [22]. CVD was defined according to the reported or self-admitted physician diagnoses. Self-admitted CVD was assessed by asking the following question: “Has a doctor or other health professional ever told you that you have a heart attack/coronary heart disease/angina/congestive heart failure/stroke?”. The answer could be “Yes/No/didn’t know”, and individuals who answered “Yes” were classified as having CVD; the “didn’t know” participants would be excluded [23].

Statistical analysis

Participants were categorized into CKD and non-CKD groups based on clinical diagnostic criteria. Descriptive analyses were conducted to summarize the basic characteristics of the study population, including age, sex, ethnicity, educational level, family PIR, BMI level, smoking history, hypertension, diabetes, CVD, and dietary indices. Continuous variables were reported as means ± standard deviations (SD), while categorical variables were expressed as frequencies (percentages). Comparisons were performed using independent t-tests or Mann-Whitney U tests for continuous variables and chi-square tests for categorical variables. We calculated a correlation coefficient matrix for various dietary indices to assess the magnitude of correlations between them.

Three types of multiple logistic regression models were employed to evaluate the associations between four dietary pattern scoring indices and CKD risk in sequence. Specifically: (1) four single-diet models were constructed, each including only one dietary index to independently assess its association with CKD risk; (2) for each dietary index, three double-diet models were constructed, adjusting for one additional dietary index at a time; (3) a full model incorporating all dietary indices together was established to evaluate their combined effects; meanwhile, the full model was also applied into patients with CKD to evaluate the combined effect of all dietary indices on the risk of CKD progression. To enhance comparability across indices, association estimates were standardized based on one-fourth of each index’s scoring range. Additionally, in the full model, we used variance inflation factors (VIF) to evaluate multicollinearity among all dietary indices, with VIF values less than 5 considered indicative of good independence. Model coefficients and odds ratios (ORs) with 95% confidence intervals (CIs) were visualized using forest plots, along with an overall composite model representation.

To quantify the predictive performance of dietary indices for CKD, receiver operating characteristic (ROC) curve analysis was conducted. Marginal differences in the area under the curve (AUC) were calculated between the full model and reduced models that systematically excluded individual covariates or dietary indices, allowing for a comparative assessment of their relative contributions [24]. Furthermore, we performed net reclassification improvement (NRI) and integrated discrimination improvement (IDI) analyses between each individual variable-excluded model and the full model. These analyses aimed to quantify the degree of improvement in the model’s classification performance when incorporating the excluded variable into the full model, thereby further evaluating the importance and statistical significance of the excluded variable.

Additionally, restricted cubic splines (RCS), with the Akaike information criterion (AIC) optimization program used to select the most suitable knots, were applied to explore potential nonlinear associations between different dietary indices and CKD risk in the total population and various sub-populations [25]; this approach allowed us to evaluate whether observed relationships remained consistent across diverse populations with inherently different dietary patterns and health profiles, serving as an alternative framework to assess result stability. In the RCS models, when focusing on the nonlinear association of one dietary pattern index, the other three indices were included in a linear form. Finally, a Supplementary analysis was conducted, specifically, based on the significance of the associations between each dietary index and CKD risk, we further explored the relationships between specific components of the corresponding indices and CKD risk. This analysis aimed to evaluate which components within each dietary index exert a significant effect on CKD risk, thereby providing targeted dietary recommendations in combination with their recommended intake levels. All results were expressed as ORs with 95% CIs to indicate the strength and direction of associations. Statistical analyses were performed using R version 4.4.1, with a two-tailed P value ≤0.05 considered statistically significant.

Results

This study analyzed a total of 46,742 individuals, with a mean age of 49.945 ± 17.814 years. Among them, 8,340 participants (17.84%) were diagnosed with CKD, while the remaining 38,402 (82.16%) did not have CKD. Table 1 presents the baseline characteristics of the study population stratified by CKD status. Notably, individuals who were older, identified as Non-Hispanic Whites, had lower educational levels, lower family PIR, and obesity, were smokers, and had hypertension, diabetes, or CVD were significantly more likely to have CKD (all p < 0.001). Regarding dietary patterns, participants with CKD exhibited significantly higher scores in HEI-2020 and DII compared to those without CKD. However, no statistically significant difference was observed for DASH (p = 0.290). Furthermore, Sup. Tables 1 and 2 in the Supplementary Materials present the cross-tabulated frequency distributions of CKD severity and kidney failure, as well as the correlation coefficient matrix among various dietary indices, respectively. Specifically, the results indicated that among CKD patients, particularly those with CKD severity at stage 3 or higher, the proportion of patients with comorbid kidney failure exceeded 10%, reaching a maximum of 86.275%. In terms of the correlation coefficient matrix of dietary indices, HEI-2020, aMED, and DASH exhibited high correlations with each other, while DII showed a moderate correlation with the other three dietary indices.

Table 1.

Description of demographic characteristics and associated factors for all participants divided by whether confirmed as CKD.

Variables	Overall (N = 46,742, 100%)	Without CKD (N = 38,402, 82.157%)	With CKD (N = 8,340, 17.843%)	P value
Sex				0.372
Male	22241 (47.582)	18310 (47.680)	3931 (47,134)
Female	24501 (52.418)	20092 (52.320)	4409 (52.866)
Age	49.945 ± 17.814	46.984 ± 16.660	63.579 ± 16.561	<.001
Ethnicity				<.001
Mexican American	7910 (16.923)	6726 (17.515)	1184 (14.197)
Other hispanic	3870 (8.279)	3323 (8.653)	547 (6.559)
Non-hispanic whites	21089 (45.118)	16870 (43.930)	4219 (50.588)
Non-hispanic blacks	9716 (20.786)	7901 (20.574)	1815 (21.763)
Other race	4157 (8.894)	3582 (9.328)	575 (6.894)
Educational level				<.001
Less than 9th grade	4843 (10.361)	3606 (9.390)	1237 (14.832)
9-11th grade	6507 (13.921)	5177 (13.481)	1330 (15.947)
High school graduate	10834 (23.178)	8761 (22.814)	2073 (24.856)
Some college or AA	13804 (29.532)	11532 (30.030)	2272 (27.242)
College graduate	10754 (23.007)	9326 (24.285)	1428 (17.122)
BMI Level				<.001
Underweight or normal	13176 (28.189)	11186 (29.129)	1990 (23.861)
Overweight	15731 (33.655)	13032 (33.936)	2699 (32.362)
Obesity	17835 (38.156)	14184 (36.936)	3651 (43.777)
Smoking History	21033 (44.998)	16900 (44.008)	4133 (49.556)	<.001
Family PIR	2.591 ± 1.624	2.639 ± 1.640	2.370 ± 1.526	<.001
Hypertension	23068 (49.352)	16779 (43.693)	6289 (75.408)	<.001
Diabetes Mellitus	7673 (16.416)	4594 (11.963)	3079 (36.918)	<.001
Cardiovascular disease	5068 (10.842)	2821 (7.346)	2247 (26.942)	<.001
Kidney failure	1397 (2.989)	497 (1.294)	900 (10.791)	<.001
Dietary pattern indices
HEI-2020 index	51.112 ± 12.143	50.972 ± 12.179	51.756 ± 11.954	<.001
aMED index	3.527 ± 1.416	3.533 ± 1.427	3.498 ± 1.367	0.042
DASH index	22.468 ± 5.027	22.456 ± 5.075	22.521 ± 4.804	0.290
DII index	2.472 ± 2.036	2.404 ± 2.048	2.784 ± 1.949	<.001

Note: Independent Student-t test was used to test the difference between participants with CKD and without CKD groups for continuous variables. Chi-square test was used to test the difference for categorical variables. Abbreviations: Some college or AA, some college or associates (AA) degree; BMI, Body Mass Index; Family PIR, the ratio of family income to poverty. The ranges of BMI for underweight, normal, overweight and obesity were <18.5, 18.5–24.9, 25.0–29.9, and ≥30 kg/m², respectively.

As presented in Figure 2, multiple logistic regression analyses incorporating a single dietary pattern index showed that all four dietary indices (HEI-2020, aMED, DASH, and DII) were significantly associated with CKD risk. Specifically, for a one-quarter increase in each index’s maximum score range, the ORs were 0.801 (95% CI: 0.756–0.849) for HEI-2020, 0.828 (95% CI: 0.791–0.867) for aMED, 0.805 (95% CI: 0.768–0.844) for DASH, and 1.242 (95% CI: 1.171–1.317) for DII, respectively. The double-dietary pattern model analyses demonstrated that aMED, DASH, and DII remained significantly negatively associated with CKD risk even after accounting for the influence of other dietary patterns. However, the association between HEI-2020 and CKD became unsignificant when DASH was included in the model. Findings from the full logistic regression model (as details also shown in Table 2 and Sup. Figure 1 of the Supplementary), which incorporated all four dietary indices, further supported these results, with DASH and DII showing significant associations, while HEI-2020 and aMED exhibited no statistically significant associations with CKD risk. Additionally, when the full model applied to individuals with CKD to explore the association between all dietary indices and the severity of CKD, only DII still remained statistically significant, with an OR of 1.264 (95%CI: 1.103–1.450), as shown in Sup. Table 3 and Sup. Figure 2 of the Supplementary materials.

Figure 2.

Forest Plot of the associations between the four dietary indices and CKD risk in single-diet, double-diet, and overall-diet models.

Table 2.

Multiple logistic regression analysis of the association between all dietary indices and CKD risk adjusted for all covariates.

Variables	Coefficients	Std.error	Z.value	OR	95% CI of OR		P value
					Lower	Upper
Sex (Ref = male)	0.133	0.029	4.566	1.143	1.079	1.210	<.001*
Age	0.049	0.001	47.030	1.050	1.048	1.052	<.001*
Ethnicity (Ref = Mexican American)
Other hispanic	−0.178	0.062	−2.870	0.837	0.740	0.945	0.004*
Non-hispanic whites	0.148	0.045	3.262	1.159	1.061	1.267	0.001*
Non-hispanic blacks	0.066	0.049	1.353	1.069	0.971	1.177	0.176
Other race	0.148	0.064	2.313	1.160	1.022	1.314	0.021*
Educational level (Ref = less than 9th grade)
9–11th grade	0.027	0.053	0.506	1.027	0.926	1.140	0.613
High school graduate	−0.02	0.050	−0.395	0.980	0.888	1.082	0.693
Some college or AA	0.007	0.051	0.146	1.007	0.912	1.113	0.884
College graduate	−0.071	0.058	−1.226	0.932	0.832	1.043	0.220
BMI level (Ref = underweight or normal)
Overweight	−0.137	0.036	−3.754	0.872	0.812	0.937	<.001*
Obesity	−0.018	0.036	−0.500	0.982	0.915	1.054	0.617
Smoking history	−0.071	0.028	−2.515	0.931	0.881	0.984	0.012*
Family PIR	−0.086	0.010	−8.653	0.918	0.900	0.936	<.001*
Hypertension	0.491	0.032	15.373	1.634	1.535	1.739	<.001*
Diabetes mellitus	0.844	0.032	26.738	2.325	2.186	2.473	<.001*
Cardiovascular disease	0.569	0.036	15.995	1.767	1.648	1.895	<.001*
Dietary pattern indices
HEI-2020 index	−0.019	0.046	−0.426	0.981	0.897	1.073	0.670
aMED index	−0.052	0.036	−1.430	0.949	0.884	1.019	0.153
DASH index	−0.127	0.041	−3.091	0.880	0.812	0.954	0.002*
DII index	0.095	0.036	2.635	1.099	1.025	1.180	0.008*

Notes: * presented as statistical significant; OR was the odds ratios in full model; for each dietary pattern scoring index, the coefficients, std.error, OR and 95%CI of OR were all estimated with the standardized per one-fourth of each index’s scoring range. Abbreviations: Some college or AA, some college or associates (AA) degree; BMI, Body Mass Index; Family PIR, the ratio of family income to poverty.

Table 3.

The comparison results of ROC, AUC, and model improvement indicators between the full model and the specific-factor-excluded models.

Models	AUC	95% CI	MD	NRI index		IDI index
				Value	P value	Value	P value
Exclude: Sex	0.796	0.791–0.802	0.000	0.0003	0.836	0.0002	0.138
Exclude: Ethnicity	0.796	0.790–0.801	0.000	−0.0005	0.725	0.0013	<.001
Exclude: BMI level	0.796	0.791–0.802	0.000	0.0022	0.091	0.0003	0.016
Exclude: Smoking history	0.796	0.791–0.802	0.000	−0.0007	0.524	0.0002	0.012
Exclude: Family PIR	0.795	0.790–0.801	0.001	0.1181	<.001	0.0019	<.001
Exclude: Dietary indices	0.795	0.790–0.801	0.001	0.0947	<.001	0.0026	<.001
Exclude: Cardiovascular disease	0.794	0.788–0.799	0.002	0.0024	0.240	0.0078	<.001
Exclude: Hypertension	0.792	0.786–0.797	0.004	0.0051	0.058	0.0044	<.001
Exclude: Diabetes mellitus	0.786	0.780–0.791	0.010	0.0276	<.001	0.0176	<.001
Exclude: Age	0.758	0.753–0.764	0.038	0.5725	<.001	0.0576	<.001
Overall model	0.796	0.791–0.802	0.000	–	–	–	–

Notes: AUC is the area under the curve for each model; 95%CI is the confidence interval of AUC; NRI is the net reclassification improvement for the overall model compared with each specific-factor-excluded model; IDI is the integrated discrimination improvement for the overall model compared with each specific-factor-excluded model. Abbreviations: MD, Marginal differences of AUC between specific-factor-excluded model and overall model; BMI, Body Mass Index; Family PIR, the ratio of family income to poverty.

As presented in Table 2 and Sup. Figure 1 of the Supplementary, after mutual adjustment for covariates, the full model identified several significant risk factors for the CKD. Specifically, sex (female: OR = 1.143), ethnicity (Non-Hispanic Whites: OR = 1.159; Other Hispanic: OR = 0.837; Other Race: OR = 1.160), BMI level (overweight: OR = 0.872), smoking history (OR = 0.931), higher family PIR (OR = 0.918), hypertension (OR = 1.634), diabetes mellitus (OR = 2.325), advancing age (OR = 1.050 per year), and CVD (OR = 1.767) remained statistically significant (p < 0.05). However, the statistical significance of educational level was not maintained. Assessment of multicollinearity among the four dietary pattern indices yielded VIF of 2.728, 2.829, 3.569, and 1.582, respectively, suggesting a lower degree of the collinearity concerns.

Analyses of ROC curves demonstrated stable predictive performance across models that excluded individual covariates, with the exception of age, whose removal led to the most pronounced decline in AUC, as presented in Figure 3. Specifically, Figure 3A displays the distributions of ROC curves for the full model versus individual variable-excluded models, while Figure 3B presents bar charts illustrating the marginal differences in AUC; these results are also summarized in Table 3. Furthermore, Table 3 also presents the analyses of NRI and IDI comparing the full model with each individual variable-excluded model. Collectively, these findings revealed that among the risk factors associated with CKD risk, only family PIR, dietary indices, CVD, hypertension, diabetes, and age exhibited non-zero marginal AUC differences, with statistically significant NRI and IDI metrics for each. Notably, dietary indices ranked next to age, hypertension, diabetes, and CVD in terms of their contributions, which highlighting that dietary indices constitute a notable and intervention-worthy category of risk factors.

Figure 3.

The ROC and AUC comparisons in the full model and the specific-factor-excluded models. Figure 3A and B presented the ROC and AUC comparisons respectively.

In the section of nonlinear analyses, we evaluated the nonlinear dose-response associations between dietary indices and the risk of CKD using RCS with knots optimized via AIC value, shown in Sup. Figure 3 of the Supplementary Materials. For all indices, the values of AIC were minimized in the model with 3 knots, and these configurations were retained for subsequent analyses. As presented in Figure 4, the results from the non-linear models were generally consistent with those from the linear full model. Specifically, after adjusting for the other three dietary indices and setting the reference point as the median value of corresponding index, the HEI-2020 and aMED were not statistically associated with the risk of CKD. The DASH and DII exhibited statistically significant associations with the risk of CKD, with P values of 0.002 and <0.001, respectively. Among them, the DASH exhibited a linear association (P value of nonlinear test was 0.102) and the DII showed a nonlinear association (P value of nonlinear test was 0.011). From the nonlinear dose-response associations for DASH and DII on the risk of CKD, using the median value (22.000 and 2.611, respectively) as the reference point, both indices presented significance in the segment of better dietary pattern adherence (22–40 for DASH, and −5 − 2.61 for DII), but lost significance in the poorer adherence segment (8–22 for DASH and 2.61 − 8 for DII).

Figure 4.

Nonlinear associations between the four dietary indices and CKD. The multiple logistic regressions with RCS were used to estimate the non-linear associations by adjusting for confounders and other dietary indices in a linear form. A lower score for the HEI-2020, aMED, and DASH indicated a poorer dietary habit, whereas the opposite was true for the DII.

Furthermore, RCS analyses were also used in subgroups to explore the population heterogeneity of the associations between different dietary indices and CKD risk. The summarized results are shown in Table 4 and Sup. Figures 4 to 26 of the Supplementary Materials. The results of subgroup analyses showed that in the subgroups of males, aged over 65 years, non-Hispanic Whites, both smokers and no smokers, the family PIR over 3.5, and individuals with hypertension or without diabetes and CVD, the associations for those dietary indices were more consistent with the associations found in the full model (Subgroup analyses of categorical covariates were performed according to their categories. For continuous variables (age and family PIR), age was analyzed in subgroups according to whether the participants were elderly (age >65 years old) [26], the family PIR was divided into ≤1.3, 1.3 to 3.5, and >3.5, according to the guidelines of whether participants are eligible for the Supplemental Nutrition Assistance Program (SNAP) [27,28]).

Table 4.

Summary of the significance results of all non-linear analyses of dietary indices and CKD in the total and subgroup populations.

Variables	HEI-2020 index		aMED index		DASH index		DII index
	P for index	P for non-linear	P for index	P for non-linear	P for index	P for non-linear	P for index	P for non-linear
Total population	0.189	0.076	0.070	0.070	0.002	0.102	<0.001	0.011
Sub groups
Sex
Male	0.097	0.046	0.046	0.015	0.043	0.023	0.002	0.181
Female	0.821	0.547	0.276	0.979	0.003	0.889	0.185	0.066
Age
Age ≤65	0.184	0.185	0.551	0.302	0.014	0.085	0.281	0.238
Age >65	0.318	0.197	0.184	0.174	0.043	0.147	<0.001	0.017
Ethnicity
Mexican American	0.712	0.678	0.248	0.580	0.756	0.459	0.445	0.562
Other hispanic	0.572	0.595	0.434	0.252	0.764	0.713	0.069	0.142
Non-hispanic whites	0.886	0.625	0.949	0.811	0.017	0.909	0.029	0.571
Non-hispanic blacks	0.005	0.004	0.415	0.197	<0.001	<0.001	0.003	<0.001
Other race	0.725	0.606	0.002	0.025	0.659	0.594	0.747	0.863
Without smoking history	0.270	0.112	0.171	0.477	0.049	0.272	0.085	0.150
With smoking history	0.853	0.620	0.436	0.208	0.040	0.287	0.022	0.049
Family PIR
PIR ≤1.3	0.346	0.181	<0.001	0.013	0.088	0.040	0.455	0.278
PIR between 1.3 and 3.5	0.725	0.453	0.972	0.817	0.031	0.668	0.195	0.837
PIR >3.5	0.343	0.154	0.315	0.128	0.120	0.217	<0.001	0.008
BMI level
Underweight or normal	0.860	0.719	0.287	0.366	0.188	0.307	0.711	0.917
Overweight	0.987	0.908	0.235	0.624	0.645	0.753	<0.001	0.060
Obesity	0.102	0.033	0.488	0.231	0.005	0.178	0.170	0.066
Without hypertension	0.331	0.374	0.106	0.355	0.165	0.060	0.068	0.218
With hypertension	0.161	0.148	0.287	0.132	<0.001	0.513	0.024	0.030
Without diabetes mellitus	0.407	0.346	0.151	0.165	0.013	0.380	0.002	0.048
With diabetes mellitus	0.021	0.041	0.353	0.193	0.036	0.031	0.145	0.069
Without cardiovascular disease	0.739	0.438	0.304	0.272	0.008	0.325	0.016	0.065
With cardiovascular disease	0.068	0.031	0.125	0.080	0.131	0.111	0.061	0.049

Notes: BMI, Body Mass Index; Family PIR, the ratio of family income to poverty.

Finally, we conducted further component analyses on the two dietary indices (DASH and DII) that showed significant associations with CKD risk in the study, with details provided in Sup. Tables 4 to 6 of the Supplementary materials. Specifically, Sup. Table 4 explains the meaning of each component within the DASH and DII indices, while Sup. Table 5 presents the differences in the distribution of component scores of DASH and DII between the CKD and non-CKD groups. For DASH, all components exhibited significant distributional differences between the two groups, whereas for DII, all components showed significant intergroup differences except for β-Carotene, Flavan-3-ol, Flavonones, Anthocyanidins, and Isoflavones. Further, Sup. Table 6 displays the results of logistic regression analyses, specifically, in the DASH index, all components were statistically significant except for the intake of red and processed meats; while in the DII index, components such as Alcohol, Carbohydrate, Cholesterol, Fe, Mg, Protein, Niacin, Vitamin A, and Thiamin primarily exerted significant effects on CKD. By further integrating the recommended intake levels of these specific components that are significantly associated with CKD risk, more precise prevention and control strategies can be formulated for individuals.

Discussions

This study aimed to compare the associations of four dietary indices with CKD risk, assess the importance of dietary patterns among CKD-related factors, and explore population heterogeneity. In single-diet models, all four indices were significantly associated with CKD risk. However, in the full model incorporating all indices (which including both linear and nonlinear analyses), only DASH and DII showed robust associations, with aMED’s role remaining unclear and HEI-2020 showing no significant association. Notably, only DII was significantly associated with CKD severity. Nonlinear analyses further revealed that DASH exhibited a linear association with CKD risk, while DII showed a nonlinear relationship, with both indices demonstrating significance in segments reflecting better dietary pattern adherence. The incremental utility of dietary indices in CKD risk prediction ranked second only to age and comorbidities (including hypertension, diabetes, and CVD). Subgroup analyses indicated consistent associations with the full model in specific populations, including males, individuals over 65 years, Non-Hispanic Whites, both smokers and nonsmokers, those with family PIR >3.5, and individuals with hypertension or without diabetes and CVD. Component analyses of DASH and DII further supported their links to CKD by identifying specific constituent factors associated with risk. Collectively, these findings suggest that incorporating DASH and DII into CKD risk assessment may help optimize prevention strategies, underscoring their potential as actionable targets for intervention.

The superior performance of DASH and DII in full model was inseparable from their targeted dietary assessment perspective and the physiological mechanism behind them. For the DASH index, several studies, including NHAENS [29,30], the Korean National Health and Nutrition Examination Survey (KNHANES) [31], and the observational multicenter German CKD (GCKD), have found that high adherence to the DASH diet is significantly associated with a lower risk of CKD or end-stage renal disease (ESRD) [32], which are all consistent with the findings of our study. The renal benefits of the DASH dietary pattern stem from its integrated approach to nutrient modulation. By limiting sodium intake, saturated fats, and alcohol, it reduces intravascular volume and renin-angiotensin system activation, lowering glomerular hydrostatic pressure and mitigating endothelial damage, which were important drivers of glomerular hypertension and tubular injury [33–35]. Concurrently, potassium intake stimulates sodium excretion, which improves vascular compliance and preserves glomerular filtration barrier integrity [36], while DASH’s emphasis on magnesium-rich foods (e.g. nuts, legumes) suppresses vascular smooth muscle cell proliferation and reactive oxygen species production, further attenuating tubulointerstitial fibrosis [37,38]. Calcium contributes to phosphate homeostasis, preventing mineral metabolism disturbances that accelerate renal fibrosis [39,40]. Clinically, this dual strategy (restriction of harmful components and supplementation of protective nutrients) synergistically preserves kidney function through multiple pathways, such as lowering the urinary albumin to creatinine ratio, maintaining eGFR, and reducing oxidative stress [41,42]. Observational studies consistently link DASH adherence to slower CKD progression, likely mediated by its combined effects on blood pressure regulation, acid-base balance, and tubular epithelial protection [43,44].

On the other hand, the DII quantifies systemic inflammation modulated by pro-inflammatory nutrients (e.g. trans-fats, refined carbohydrates), which exacerbates endothelial dysfunction and renal fibrosis via NF-kB activation and cytokine overproduction, and promotes transcription of pro-fibrotic cytokines (TGF-β, IL-1β) while enhancing macrophage infiltration into the interstitium [37,45,46]. Chronic low-grade inflammation, quantified by elevated C-reactive protein (CRP), tumor necrosis factor (TNF) or interleukin 6 (IL-6) levels, has been causally linked to accelerated eGFR decline in cohort studies [47–49]. Additionally, anti-inflammatory nutrients like polyphenols (olive oil, green tea) and omega-3 fatty acids inhibit NLRP3 inflammasome activation and reduce monocyte chemoattractant protein-1 (MCP-1), which blunt renal inflammatory cascades and fibrosis [50,51]. In contrast, HEI-2020 and aMED prioritize broader dietary quality metrics (e.g. whole grain diversity in HEI-2020, olive oil consumption in aMED) that may lack specificity for renal pathomechanisms, potentially diluting their predictive power in adjusted models.

Moreover, while the previous investigation [52] comprehensively evaluated four dietary indices (HEI-2020, aMED, DASH, DII) in association with CKD risk using both observational and Mendelian approaches, our study advances these findings by critically adjusting for multiple dietary patterns. While their excellent work established individual associations between each index and the risk of CKD, the lack of combined exposure modeling left unresolved questions about collinearity and clinical prioritization. Our study advances Huang et al.’s work in several key aspects. In terms of scope, while their study mainly evaluated individual associations between each dietary index and the risk of CKD, our research incorporates all four dietary indices into multiple logistic regression models. This comprehensive approach allows us to explore the collinearity among these indices and prioritize them clinically. In terms of model adjustment, our analysis critically adjusts for multiple dietary patterns, which was not fully addressed in their study.

Through this adjustment, we found that the significance of HEI-2020 and aMED diminished in the fully adjusted models, in contrast to their reported individual significance. This difference may be due to the overlapping protective components in these indices, such as the shared emphasis on whole grains and vegetables, which could lead to residual confounding in single-index effect estimates. Additionally, cultural and demographic factors in our study population (such as predominance of Non-Hispanic Whites) might limit the relevance of aMED (rooted in Mediterranean dietary patterns) and HEI-2020 (designed for U.S. dietary guidelines) to specific subgroups, as dietary adherence and relevance can vary by ethnicity or regional food environments [53]. Importantly, this does not negate their potential utility in other studies, for instance, aMED may retain predictive value in populations with stronger Mediterranean dietary adherence, while HEI-2020 could perform better in studies focusing on broader metabolic health outcomes beyond CKD. Future research in diverse cohorts is warranted to clarify their contextual specificity and expand the generalizability of dietary index applications in renal risk assessment.

Notably, our finding that DII, but not DASH, was significantly associated with CKD severity may reflect distinct mechanistic pathways through which these dietary indices influence CKD progression. As a dietary index explicitly designed to quantify the inflammatory potential of diet, DII captures components known to modulate systemic inflammation, such as pro-inflammatory nutrients (such as saturated fats, refined carbohydrates) and anti-inflammatory compounds (such as certain vitamins, fiber), which could directly impact renal pathophysiology [45]. Chronic low-grade inflammation is a well-established driver of CKD progression, contributing to glomerular damage, tubulointerstitial fibrosis, and oxidative stress in renal tissues [46]. The specific association of DII with CKD severity may thus stem from its ability to reflect dietary patterns that either exacerbate or mitigate this inflammatory cascade, with higher DII scores potentially accelerating structural and functional decline in already impaired kidneys [49].

In contrast, DASH, while associated with CKD occurrence, focuses primarily on dietary factors linked to blood pressure regulation and cardiovascular health (such as sodium reduction, increased potassium and fiber intake). Its lack of association with CKD severity might suggest that while DASH-related dietary patterns are protective against incident CKD [34,35], likely through hemodynamic effects such as reducing hypertension-induced renal strain, they may have a more limited role in modulating the inflammatory or fibrotic processes that drive progression in established disease. This divergence highlights the possibility that distinct dietary attributes influence different stages of CKD, that is DASH may act as a preventive factor by addressing early hemodynamic risks, whereas DII captures dietary pro-inflammatory potential that directly impacts the progression of existing renal damage.

The RCS analyses offer critical clinical insights by delineating dose-response relationships between dietary indices and CKD risk. DASH demonstrated a linear association, with consistent risk reduction observed for scores ranging from 22 to 40, which reflects better adherence. This range provides a tangible benchmark for dietary counseling, as it indicates that incremental improvements in DASH adherence correspond to reduced CKD risk. Conversely, DII exhibited a nonlinear pattern, with significant associations confined to scores between −5 and 2.61, signifying less pro-inflammatory diets. Scores exceeding 2.61 suggest elevated risk, highlighting the importance of addressing pro-inflammatory dietary components such as refined sugars and excessive alcohol intake in these cases. Derived from our study data, these thresholds empower clinicians to translate abstract dietary indices into actionable targets, facilitating personalized dietary recommendations based on patients’ current dietary profiles.

The predictive value of dietary indices, which ranked second only to age and key comorbidities including hypertension, diabetes, and CVD, underscores their significance in clinical screening. Importantly, dietary modification offers distinct cost-effectiveness advantages and multiple organ benefits over pharmacological management of diabetes, hypertension or CVD [54,55]. On the other hand, our results also highlighted the significant heterogeneity of dietary indices across populations with different characteristics, and suggested that the associations for DASH and DII on CKD in the subgroups of males, individuals over 65 years, Non-Hispanic Whites, both smokers and nonsmokers, those with family PIR >3.5, and individuals with hypertension or without diabetes and CVD, are more consistent with the results of the overall model, which suggests that it is more beneficial to consider DASH and DII diet improvement strategies in these populations. For instance, consider a 65-year-old non-Hispanic White male with hypertension, a subgroup in which our analyses revealed consistent associations with the full model. Incorporating DASH and DII scores into his risk assessment could enhance clinical decision-making. A DASH score below 22 or a DII score above 2.61 would prompt intensified monitoring or targeted dietary counseling, complementing traditional risk factors such as blood pressure. Supported by our subgroup findings, this approach highlights how dietary indices can refine risk stratification, particularly among populations most likely to benefit, such as males and individuals with hypertension.

Component analyses further deepen these insights. For DASH, all components except red and processed meats were correlated with CKD risk, reinforcing the emphasis on fruits, vegetables, whole grains, and low-fat dairy in dietary counseling. Sodium reduction, a central tenet of the DASH diet, emerges as a key recommendation. For DII, significant contributors to CKD risk included alcohol, refined carbohydrates, cholesterol, and micronutrients such as iron, magnesium, and vitamins A and thiamin. These findings underscore the need to limit pro-inflammatory dietary elements while optimizing micronutrient intake. Collectively, these results enable clinicians to move beyond generic dietary advice, tailoring interventions to promote fiber-rich foods and reduce red meat consumption for DASH, and to moderate alcohol and refined carbohydrate intake while prioritizing micronutrient-dense options for DII. This specificity enhances the practicality of dietary modifications, bridging research findings with actionable clinical strategies.

While this investigation provides novel insights into dietary-CKD associations, several methodological limitations inherent in population-based surveys and observational designs are worth consideration. First, the cross-sectional nature of NHANES precludes delineation of temporal relationships between dietary exposures and CKD occurrence, leaving residual uncertainty regarding causal directionality, a limitation shared by prior nutritional epidemiology studies. While cross-sectional analyses cannot establish temporality, they provide critical initial evidence on the comparative discriminatory power of dietary indices for CKD risk and heterogeneity, informing hypotheses for future longitudinal studies. Second, despite employing telephone interviews to enhance recall accuracy, the reliance on 24-h dietary assessments introduces potential measurement error, as single-day recalls may inadequately represent habitual intake patterns compared to food frequency questionnaires or biomarker-validated methods. Therefore, our findings offer preliminary insights into the relative discriminative power of these indices under real-world assessment constraints, while underscoring the need for future studies with more comprehensive dietary assessments to corroborate these associations and strengthen evidence of habitual diet-CKD relationships. Third, although we adjusted for a number of demographic, clinical, and socioeconomic confounders, uncontrolled residual confounding from unassessed variables (e.g. genetic predisposition, environmental toxin exposure) remains plausible. Finally, while NHANES ensures national representativeness for U.S. adults, extrapolation to global populations requires caution given documented transcontinental variations in both dietary practices and CKD etiological profiles. These limitations underscore the necessity for multi-center prospective cohorts (e.g. integrating UK Biobank dietary data with serial eGFR measurements) and mechanistic studies employing mediating biomarkers (e.g. urinary F2-isoprostanes for oxidative stress quantification) to confirm observed associations.

Conclusions

In our study, comparative analyses of four dietary indices highlight DASH and DII as the most robustly associated with CKD risk, with DII uniquely linked to CKD severity. DASH exhibited a linear association with risk reduction in higher adherence ranges, while DII showed nonlinear relationships tied to pro-inflammatory dietary patterns. The incremental predictive utility of these indices ranked second only to age and key comorbidities, with consistent associations in specific subgroups (such as males, individuals over 65, those with hypertension). Component analyses further identified actionable dietary factors within DASH and DII linked to CKD risk. In contrast, HEI-2020 showed no significant association, and aMED’s role remained unclear. These findings support integrating DASH and DII into CKD risk assessment to optimize prevention strategies, underscoring their potential as targeted interventions for reducing CKD risk and progression.

Funding Statement

The authors declare that no financial support was received for the research, authorship, and publication of this article.

Informed consent statement

Informed consent was obtained from all subjects involved in the NHANES survey; and ethical review and approval were waived for this study due to fully anonymous patient data.

Disclosure statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. And the authors declare that the Deepseek-R1 in the writing process to assist our text modification (20%).

Data availability

Data are available at the NHANES website (https://wwwn.cdc.gov/nchs/nhanes/Defa- ult.aspx).