Association between red cell distribution width-to-albumin ratio and risk of diabetic kidney disease: a cross-sectional NHANES study

Abstract

Background

The red blood cell distribution width-to-albumin ratio (RAR) is a novel hematological biomarker that integrates information on inflammation and nutritional status. While RAR has been applied in assessing risks for various chronic diseases, its association with diabetic kidney disease (DKD) remains unclear. This study aims to evaluate the relationship between RAR and the risk of DKD in diabetic patients, exploring its potential value in early risk identification.

Methods

Utilizing data from the National Health and Nutrition Examination Survey (NHANES) spanning 2005 to 2020, we included 7,191 eligible adult participants diagnosed with diabetes. RAR was categorized into quartiles, and weighted multivariable logistic regression along with restricted cubic spline (RCS) models were employed to assess the association between RAR levels and the prevalence of DKD. Subgroup and sensitivity analyses were conducted to validate the robustness of the findings.

Results

The RAR level was significantly higher in the DKD group compared to the non-DKD group (p < 0.001), and the prevalence of DKD increased progressively across RAR quartiles (25.20%, 30.17%, 39.34%, and 43.33%, p < 0.001). In the unadjusted model, each one-unit increase in RAR was associated with a 76.4% higher risk of DKD (OR = 1.764, 95% CI: 1.541–2.017, p < 0.001). Participants in the highest RAR quartile (Q4) had a 2.269-fold increased risk of DKD compared to those in the lowest quartile (Q1) (95% CI: 1.809–2.846, p < 0.001). RCS analysis suggested a potentially linear association between RAR and DKD risk, with no significant non-linear trend observed after adjusting for covariates (p > 0.05). Subgroup analysis revealed that the association remained consistent across most strata, although significant interactions were found for sex and BMI (P for interaction < 0.05), with stronger associations observed in males and in participants with higher BMI. Sensitivity analyses confirmed the robustness of these findings.

Conclusion

Elevated RAR levels are significantly associated with increased risk of diabetic kidney disease among individuals with diabetes, suggesting its strong predictive potential. As a simple and cost-effective biomarker, RAR may serve as a useful tool for early DKD risk screening and stratification. However, prospective studies are warranted to further validate its clinical utility.

Introduction

Diabetes mellitus has emerged as a global epidemic, with a steadily increasing prevalence, and is now recognized as a major contributor to chronic complications such as cardiovascular disease, blindness, and renal failure [1]. Diabetic kidney disease (DKD), the most common microvascular complication, affects approximately 30% to 40% of patients with diabetes and has become the leading cause of end-stage renal disease (ESRD) [2, 3]. DKD not only results in irreversible decline in renal function but also substantially elevates the risk of cardiovascular events and all-cause mortality [4]. Therefore, identifying simple and reliable biomarkers for early risk assessment of DKD is crucial for delaying disease progression and improving clinical outcomes.

An increasing body of evidence suggests that systemic inflammation and nutritional imbalance play pivotal roles in the pathogenesis and progression of diabetic kidney disease (DKD) [5, 6]. Inflammatory markers such as C-reactive protein (CRP), interleukin-6 (IL-6), and tumor necrosis factor-alpha (TNF-α) may promote renal injury by activating immune responses and fibrotic pathways [7, 8]. Meanwhile, malnutrition—often characterized by hypoalbuminemia—is closely associated with proteinuria, inflammation, and poor prognosis in DKD patients [9, 10]. Red cell distribution width (RDW), a hematological parameter reflecting inflammation and oxidative stress, has been linked to increased risk of diabetes and its complications [3–12]. Although these individual markers offer some predictive value, they are influenced by multiple factors, limiting their stability and interpretability. As a result, composite indicators have gained growing attention. The red cell distribution width-to-albumin ratio (RAR) is a novel inflammation-nutrition composite biomarker that integrates RDW and serum albumin to more comprehensively reflect systemic inflammatory and nutritional status [13]. Recent studies have shown that elevated RAR is significantly associated with adverse outcomes in heart failure [14], stroke [15], metabolic syndrome [16], cancer [17, 18], and depression [19]. In diabetic populations, RAR has also been linked to increased risk of mortality, atherosclerosis, and diabetic retinopathy [20, 21]. However, no study has systematically explored the relationship between RAR and DKD risk, nor assessed its epidemiological characteristics and stratified predictive value in diabetic kidney disease.

Based on this background, the present study aimed to systematically evaluate the association between RAR and DKD among individuals with diabetes using large-scale data from the NHANES 2005–2020 cycles. We hypothesized a positive linear association between RAR and DKD risk, such that higher RAR levels would be associated with a progressively increased likelihood of developing DKD. Through this analysis, we seek to validate the potential of RAR as a predictive biomarker for DKD risk in diabetic populations and to provide novel insights into its role as a simple and cost-effective tool for risk stratification in diabetic kidney disease.

Materials and methods

Data and sample sources

This study was based on data from the National Health and Nutrition Examination Survey (NHANES) conducted from 2005 to 2020. NHANES is administered by the National Center for Health Statistics (NCHS) and employs a stratified, multistage probability sampling design to collect nationally representative health and nutrition data, including demographic characteristics, socioeconomic status, dietary habits, and various health-related indicators. The survey protocol was approved by the Ethics Review Board of the U.S. Centers for Disease Control and Prevention (CDC) NCHS, and all participants provided written informed consent. The data are publicly available at https://www.cdc.gov/nchs/nhanes/.

The initial sample included 76,469 participants. We excluded individuals under 20 years of age and those with missing data on RAR, serum creatinine, or diabetes diagnosis. After applying these criteria, a total of 7,191 adults with diabetes were included in the final analysis, among whom 2,689 were diagnosed with diabetic kidney disease (DKD). The detailed selection process is illustrated in Fig. 1.

Fig. 1.

Flow chart of the participants selection process

Exposure factors

The exposure variable in this study was the red blood cell distribution width-to-albumin ratio (RAR). In NHANES, serum albumin concentration was measured using the bromocresol purple method, while red cell distribution width (RDW, %) was determined using a Coulter blood cell analyzer on peripheral blood samples collected at mobile examination centers. RAR was calculated using the following formula: RAR = RDW (%)/Serum Albumin (g/dL).

Outcome variables

The primary outcome variable of this study was diabetic kidney disease (DKD). The diagnosis of diabetes was based on glycemic indicators, including glycated hemoglobin (HbA1c), fasting plasma glucose (FPG), and the 2-hour oral glucose tolerance test (OGTT), as well as self-reported questionnaire data. Diagnostic criteria were adopted from the 2023 Standards of Medical Care in Diabetes issued by the American Diabetes Association (ADA) and relevant studies [22, 23]. Participants were classified as having diabetes if they met any of the following conditions: (1) HbA1c ≥ 6.5%; (2) FPG ≥ 7.0 mmol/L; (3) OGTT ≥ 11.1 mmol/L; (4) answered “yes” to the question “Has a doctor or other health professional ever told you that you have diabetes?”; or (5) were currently taking insulin or oral hypoglycemic agents.

Based on this, the diagnosis of kidney disease was defined as meeting at least one of the following criteria: a reduced estimated glomerular filtration rate (eGFR < 60 mL/min/1.73 m²) or an elevated urine albumin-to-creatinine ratio (UACR ≥ 30 mg/g), consistent with previous literature [24]. eGFR was calculated using the standardized creatinine-based equation from the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) [25]. According to these criteria, participants were coded as 1 for DKD and 0 for non-DKD.

Covariates

Based on existing literature and clinical relevance [26, 27], we included a wide range of potential confounders in the analysis, including age, sex, race/ethnicity, education level, marital status, poverty-income ratio (PIR), body mass index (BMI), systolic and diastolic blood pressure (SBP and DBP), triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), white blood cell count (WBC), red blood cell count (RBC), smoking status, alcohol consumption, coronary heart disease (CHD), and hypertension. Demographic data were obtained from the NHANES demographic questionnaire module, while information on CHD and hypertension was collected through standardized interviews conducted by trained medical personnel. All covariates were extracted from the NHANES database and standardized prior to inclusion in the regression models to minimize potential confounding effects.

Additionally, the handling of missing data is as follows: Missing rates for SBP, DBP, BMI, PIR, TC, HDL-C, UA, and urine albumin-to-creatinine ratio (UACR) were all below 15%, and these missing values were imputed using the k-nearest neighbors (KNN) method. For TG, LDL-C, and INS, which had missing rates above 15%, the original missing values were retained without imputation.

Statistical methods

All statistical analyses in this study accounted for the complex sampling design of NHANES, and weighted methods were applied to ensure the representativeness and robustness of the results. Sampling weights (WTMECPRP), stratification (SDMVSTRA), and clustering (SDMVPSU) recommended by the National Center for Health Statistics (NCHS) were incorporated to reflect the survey’s design complexity.

Participants were grouped by DKD status and RAR quartiles. For continuous variables with normal distribution, weighted Student’s t-tests (for two-group comparisons) or weighted one-way ANOVA (for multi-group comparisons) were used; for non-normally distributed variables, weighted Mann–Whitney U tests or Kruskal–Wallis tests were applied. Categorical variables were compared using the weighted chi-square test. Continuous variables are presented as weighted means ± standard deviations, and categorical variables are presented as unweighted frequencies and weighted percentages.

To assess the potential effect modification of RAR across subpopulations, subgroup analyses were conducted using multivariable logistic regression models stratified by sex, race/ethnicity, education level, and other key demographic factors. To explore the association between RAR and DKD, three logistic regression models were constructed. Prior to modeling, variance inflation factor (VIF) analysis was performed to assess multicollinearity among covariates. Model 1 was unadjusted; Model 2 was adjusted for age, sex, and race/ethnicity; and Model 3 was further adjusted for education, marital status, PIR, BMI, SBP, DBP, TG, TC, HDL-C, LDL-C, WBC, RBC, smoking status, alcohol use, CHD, and hypertension. RAR was analyzed both as a continuous variable and in quartiles. Odds ratios (ORs) and 95% confidence intervals (CIs) were reported for each model.

To examine the potential nonlinear association between RAR and DKD risk, restricted cubic spline (RCS) analysis was performed. Sensitivity analyses were conducted by adjusting for different combinations of covariates, removing extreme values, and applying stepwise regression to verify the robustness of the models. All statistical tests were two-sided with a significance threshold of p < 0.05. Analyses were conducted using Python 3.9 and DecisionLinnc 1.0 software [28], with multiple verification procedures performed to ensure result accuracy.

Results

Baseline characteristics comparison between DKD and Non-DKD groups

A total of 7,191 participants with diabetes were included in this study, with a mean age of 59.36 years. Among them, 4,502 were in the non-DKD group and 2,689 were in the DKD group, accounting for 37.40% of the diabetic population. Overall, 52.72% of the participants were male and 48.28% were female. Compared with the non-DKD group, participants in the DKD group had significantly higher levels of SBP, WBC, RDW, RAR, HbA1c, FPG, OGTT, TG, and UACR (all p < 0.05), whereas DBP, RBC, serum albumin, LDL-C, and eGFR were significantly lower (all p < 0.05). In addition, significant differences were observed between the two groups in terms of education level, marital status, PIR, smoking, alcohol consumption, CHD, and hypertension (all p < 0.05). Detailed baseline characteristics are presented in Table 1.

Table 1.

Baseline characteristics of participants

Characteristic	Overall N = 7,191	NON-DKD N = 4,502	DKD N = 2,689	p-value
Age(years)	59.36 ± 13.94	56.53 ± 13.39	64.88 ± 13.33	< 0.001
BMI(kg/m²)	32.94 ± 7.57	33.00 ± 7.61	32.81 ± 7.49	0.898
SBP(mmHg)	128.98 ± 18.81	126.48 ± 16.95	134.09 ± 21.24	< 0.001
DBP(mmHg)	70.36 ± 13.54	71.03 ± 12.31	69.00 ± 15.66	< 0.001
WBC(1000 cells/uL)	7.79 ± 2.36	7.70 ± 2.26	7.97 ± 2.54	0.003
RBC(million cells/uL)	4.68 ± 0.53	4.74 ± 0.47	4.56 ± 0.60	< 0.001
RDW(%)	13.63 ± 1.38	13.53 ± 1.31	13.84 ± 1.51	< 0.001
Serum albumin(g/dL)	4.12 ± 0.36	4.15 ± 0.35	4.05 ± 0.37	< 0.001
RAR	3.35 ± 0.53	3.29 ± 0.49	3.45 ± 0.59	< 0.001
HbA1c(%)	7.13 ± 1.64	7.01 ± 1.57	7.35 ± 1.74	< 0.001
FPG(mmol/L)	8.40 ± 3.26	8.26 ± 3.09	8.71 ± 3.58	0.003
OGTT(mmol/L)	12.42 ± 4.13	12.16 ± 4.08	13.20 ± 4.18	0.002
TG(mmol/L)	1.82 ± 1.51	1.76 ± 1.44	1.96 ± 1.63	0.036
HDL-C(mmol/L)	1.24 ± 0.38	1.24 ± 0.39	1.23 ± 0.38	0.801
LDL-C(mmol/L)	2.71 ± 0.99	2.77 ± 0.98	2.57 ± 0.99	< 0.001
Urinary albumin(g/dL)	116.91 ± 663.84	11.85 ± 11.49	296.62 ± 1,069.26	< 0.001
Creatinine(mg/dL)	116.37 ± 70.81	118.91 ± 72.55	112.03 ± 67.53	0.024
UACR(mg/g)	118.70 ± 598.45	10.16 ± 6.54	304.36 ± 957.21	< 0.001
eGFR	83.37 ± 24.92	91.94 ± 17.54	66.68 ± 28.48	< 0.001
Sex, n(%)				0.708
Male	3,735 (51.72%)	2,303 (51.96%)	1,432 (51.25%)
Female	3,456 (48.28%)	2,199 (48.04%)	1,257 (48.75%)
Race, n(%)				0.082
Mexican American	1,284 (9.89%)	866 (10.21%)	418 (9.28%)
Other Hispanic	779 (6.21%)	547 (6.77%)	232 (5.11%)
Non-Hispanic White	2,521 (61.17%)	1,464 (60.80%)	1,057 (61.90%)
Non-Hispanic Black	1,835 (13.98%)	1,125 (13.57%)	710 (14.76%)
Other Race	772 (8.75%)	500 (8.65%)	272 (8.94%)
Education, n(%)				< 0.001
Less than high school	2,434 (22.78%)	1,433 (20.44%)	1,001 (27.34%)
High school or GED	1,699 (26.51%)	1,048 (25.32%)	651 (28.82%)
College or above	3,058 (50.72%)	2,021 (54.25%)	1,037 (43.83%)
Marital status, n(%)				< 0.001
Never married	693 (9.20%)	473 (9.93%)	220 (7.77%)
Married or cohabit	4,265 (63.60%)	2,824 (67.15%)	1,441 (56.68%)
Widowed or divorced	2,233 (27.20%)	1,205 (22.91%)	1,028 (35.55%)
PIR, n(%)				< 0.001
< 1.30	2,316 (23.24%)	1,396 (21.46%)	920 (26.72%)
1.30–3.49	3,319 (44.25%)	2,043 (42.80%)	1,276 (47.08%)
≥ 3.50	1,556 (32.51%)	1,063 (35.75%)	493 (26.19%)
BMI, n(%)				0.098
< 25	948 (11.37%)	557 (10.68%)	391 (12.72%)
25–30	2,208 (28.42%)	1,407 (29.02%)	801 (27.24%)
≥ 30	4,035 (60.21%)	2,538 (60.29%)	1,497 (60.04%)
Drink status, n(%)				0.028
No	6,237 (87.49%)	3,925 (88.40%)	2,312 (85.72%)
Yes	954 (12.51%)	577 (11.60%)	377 (14.28%)
Hypertension, n(%)				< 0.001
No	2,518 (36.24%)	1,845 (42.00%)	673 (25.02%)
Yes	4,673 (63.76%)	2,657 (58.00%)	2,016 (74.98%)
Smoke status, n(%)				0.016
No	3,667 (49.95%)	2,387 (51.49%)	1,280 (46.95%)
Yes	3,524 (50.05%)	2,115 (48.51%)	1,409 (53.05%)
CHD, n(%)				< 0.001
No	6,474 (89.07%)	4,177 (91.41%)	2,297 (84.50%)
Yes	717 (10.93%)	325 (8.59%)	392 (15.50%)

Categorical variables are presented as unweighted frequencies and weighted percentages, and group comparisons are performed using weighted chi-square tests. Continuous variables are presented as weighted means ± standard deviations, and group comparisons are performed using weighted t-tests or weighted Mann-Whitney tests

Quantitative relationship between baseline characteristics and RAR quartiles

To explore the dose-response relationship between RAR and baseline characteristics, participants were categorized into quartiles based on their RAR levels (Q1–Q4). Compared with those in the lower quartiles, participants in the higher RAR quartiles exhibited significantly elevated levels of BMI, WBC, RDW, RAR, HbA1c, urinary albumin, and UACR (all p < 0.001), while levels of RBC, serum albumin, TG, TC, and eGFR were significantly reduced (all p < 0.001). Additionally, the distributions of sex, race/ethnicity, marital status, PIR, BMI, CHD, and hypertension varied significantly across the quartile groups (p < 0.05). Notably, the prevalence of DKD increased progressively with higher RAR quartiles (25.20% vs. 30.17% vs. 39.34% vs. 43.33%, p < 0.001). Detailed results are presented in Table 2.

Table 2.

Baseline Characteristics by RAR Quartiles

Characteristic	Q1(≤ 3.023) N = 1,797	Q2(3.023-3.023,279) N = 1,797	Q3(3.282-3.282,618) N = 1,797	Q4(≥ 3.618) N = 1,800	p-value
Age(years)	56.98 ± 13.32	59.43 ± 13.90	61.00 ± 13.80	60.49 ± 14.49	< 0.001
BMI(kg/m²)	30.31 ± 5.42	32.13 ± 6.76	34.29 ± 7.65	35.82 ± 9.29	< 0.001
SBP(mmHg)	128.22 ± 17.75	128.62 ± 18.22	129.89 ± 18.84	129.41 ± 20.68	0.304
DBP(mmHg)	71.06 ± 13.35	70.21 ± 13.31	70.48 ± 13.15	69.51 ± 14.40	0.012
WBC (1000 cells/uL)	7.41 ± 2.10	7.64 ± 2.29	7.99 ± 2.42	8.23 ± 2.60	< 0.001
RBC(million cells/uL)	4.77 ± 0.48	4.71 ± 0.48	4.67 ± 0.52	4.54 ± 0.60	< 0.001
RDW(%)	12.62 ± 0.62	13.21 ± 0.63	13.77 ± 0.72	15.25 ± 1.73	< 0.001
Serum albumin(g/dL)	4.45 ± 0.23	4.19 ± 0.19	4.00 ± 0.21	3.73 ± 0.33	< 0.001
RAR	2.84 ± 0.14	3.15 ± 0.07	3.44 ± 0.10	4.11 ± 0.53	< 0.001
HbA1c(%)	6.99 ± 1.60	7.10 ± 1.63	7.28 ± 1.65	7.17 ± 1.65	< 0.001
FPG(mmol/L)	8.21 ± 3.09	8.51 ± 3.24	8.66 ± 3.36	8.20 ± 3.37	0.002
OGTT(mmol/L)	12.52 ± 4.29	11.95 ± 3.88	12.62 ± 3.88	12.93 ± 4.44	0.235
TC(mmol/L)	4.93 ± 1.27	4.79 ± 1.20	4.76 ± 1.22	4.57 ± 1.17	< 0.001
TG(mmol/L)	2.02 ± 1.72	1.90 ± 1.63	1.72 ± 1.35	1.57 ± 1.12	< 0.001
HDL-C(mg/dL)	1.23 ± 0.37	1.23 ± 0.36	1.24 ± 0.43	1.25 ± 0.38	0.463
LDL-C(mmol/L)	2.76 ± 0.96	2.74 ± 0.97	2.68 ± 0.99	2.63 ± 1.05	0.116
Urinary albumin(g/dL)	49.81 ± 176.23	61.94 ± 269.88	97.24 ± 325.65	295.72 ± 1,333.34	< 0.001
Creatinine(mg/dL)	113.48 ± 69.89	114.27 ± 69.42	120.73 ± 70.98	117.76 ± 73.29	0.054
UACR(mg/g)	50.07 ± 207.18	63.63 ± 280.83	87.66 ± 276.38	312.69 ± 1,180.65	< 0.001
eGFR	87.75 ± 20.81	85.16 ± 22.97	81.85 ± 25.75	77.44 ± 29.28	< 0.001
Sex, n(%)					< 0.001
Male	1,137 (64.61%)	980 (54.36%)	876 (46.55%)	742 (38.05%)
Female	660 (35.39%)	817 (45.64%)	921 (53.45%)	1,058 (61.95%)
Race, n(%)					< 0.001
Mexican American	418 (11.42%)	338 (9.83%)	292 (8.88%)	236 (9.15%)
Other Hispanic	197 (5.97%)	220 (6.62%)	209 (6.70%)	153 (5.48%)
Non-Hispanic White	679 (64.58%)	686 (65.85%)	614 (59.84%)	542 (52.82%)
Non-Hispanic Black	239 (6.61%)	385 (11.32%)	506 (16.12%)	705 (24.02%)
Other Race	264 (11.42%)	168 (6.38%)	176 (8.47%)	164 (8.53%)
Education, n(%)					0.032
Less than high school	638 (22.51%)	594 (20.59%)	600 (23.74%)	602 (24.66%)
High school or GED	404 (25.08%)	425 (26.35%)	425 (25.43%)	445 (29.64%)
College or above	755 (52.42%)	778 (53.06%)	772 (50.83%)	753 (45.70%)
Marital status, n(%)					< 0.001
Never married	149 (9.11%)	160 (7.62%)	180 (9.69%)	204 (10.65%)
Married or cohabit	1,197 (68.00%)	1,096 (67.54%)	1,035 (60.75%)	937 (56.53%)
Widowed or divorced	451 (22.88%)	541 (24.84%)	582 (29.56%)	659 (32.82%)
PIR, n(%)					< 0.001
< 1.30	523 (19.14%)	568 (21.38%)	577 (24.83%)	648 (28.86%)
1.30–3.49	822 (40.70%)	817 (44.28%)	845 (45.95%)	835 (46.80%)
≥ 3.50	452 (40.15%)	412 (34.34%)	375 (29.22%)	317 (24.34%)
BMI, n(%)					< 0.001
< 25	339 (15.78%)	256 (11.27%)	171 (8.21%)	182 (9.41%)
25–30	677 (34.99%)	616 (30.11%)	481 (23.86%)	434 (23.12%)
≥ 30	781 (49.23%)	925 (58.61%)	1,145 (67.93%)	1,184 (67.47%)
Drink status, n(%)					0.902
No	1,558 (88.06%)	1,552 (87.37%)	1,546 (87.03%)	1,581 (87.41%)
Yes	239 (11.94%)	245 (12.63%)	251 (12.97%)	219 (12.59%)
Hypertension, n(%)					< 0.001
No	771 (43.43%)	667 (37.25%)	596 (34.37%)	484 (28.09%)
Yes	1,026 (56.57%)	1,130 (62.75%)	1,201 (65.63%)	1,316 (71.91%)
Smoke status, n(%)					0.791
No	910 (49.25%)	929 (50.25%)	934 (51.34%)	894 (48.95%)
Yes	887 (50.75%)	868 (49.75%)	863 (48.66%)	906 (51.05%)
CHD, n(%)					< 0.001
No	1,667 (92.48%)	1,618 (89.46%)	1,602 (87.82%)	1,587 (85.70%)
Yes	130 (7.52%)	179 (10.54%)	195 (12.18%)	213 (14.30%)
DKD, n(%)					< 0.001
No	1,294 (74.80%)	1,211 (69.83%)	1,057 (60.66%)	940 (56.67%)
Yes	503 (25.20%)	586 (30.17%)	740 (39.34%)	860 (43.33%)

Categorical variables are presented as unweighted frequencies and weighted percentages, and group comparisons are performed using weighted chi-square tests. Continuous variables are presented as weighted means ± standard deviations, and group comparisons are performed using weighted analysis of variance (ANOVA) or weighted Kruskal-Wallis tests

Association between RAR quartiles and DKD risk

To further investigate the association between RAR and DKD, multivariable logistic regression analyses were performed. As shown in Table 3, in the unadjusted model (Model 1), RAR as a continuous variable was significantly associated with the risk of DKD. Specifically, each one-unit increase in RAR was associated with a 76.4% higher risk of DKD (OR = 1.764, 95% CI: 1.541–2.017, p < 0.001).

Table 3.

Association Between RAR Quartiles and DKD Risk

Outcome	Model 1		Model 2		Model 3
	OR (95%CI)	P	OR (95%CI)	P	OR (95%CI)	P
RAR(Continuous)	1.764 (1.541,2.017)	< 0.001	1.701(1.492,1.939)	< 0.001	1.338 (1.023,1.75)	0.023
RAR (Quartile)
Q1	1(reference)		1(reference)		1(reference)
Q2	1.282 (1.024,1.605)	0.031	1.172(0.921,1.489)	0.196	1.038(0.809,1.333)	0.767
Q3	1.925(1.588,2.332)	< 0.001	1.685(1.359,2.089)	< 0.001	1.305(1.013,1.681)	0.039
Q4	2.269 (1.809,2.846)	< 0.001	2.054(1.632,2.584)	< 0.001	1.36(1.029,1.891)	0.037
P for trend		< 0.001		< 0.001		0.028

Model 1: Unadjusted; Model 2: Adjusted for sex, age, and race/ethnicity; Model 3: Further adjusted for age, sex, race/ethnicity, education level, marital status, PIR, BMI, SBP, DBP, TG, TC, HDL-C, LDL-C, WBC, RBC, smoking status, alcohol use, CHD, and hypertension

In addition, when RAR was analyzed as a categorical variable by quartiles, participants in the highest quartile (Q4) had a significantly increased risk of DKD compared with those in the lowest quartile (Q1), with an OR of 2.269 (95% CI: 1.809–2.846, p < 0.001), indicating a 2.27-fold higher risk. This association remained statistically significant after adjusting for sex, age, and race/ethnicity in Model 2, and after further adjustment for all potential confounders in Model 3, suggesting a robust relationship between RAR and DKD risk.

Subgroup analysis

To further evaluate the predictive role of RAR for DKD risk across different populations, stratified subgroup analyses were conducted (Fig. 2). The results showed significant interactions between RAR and both sex and BMI (P for interaction < 0.05), suggesting that the association between RAR and DKD risk may be modified by these factors. In contrast, consistent positive associations were observed across other subgroups, including race/ethnicity, education level, marital status, poverty income ratio (PIR), smoking, alcohol consumption, coronary heart disease (CHD), and hypertension, with no significant interactions detected (P for interaction > 0.05). These findings indicate that while the overall relationship between elevated RAR and increased DKD risk is robust, particular attention should be paid to sex- and BMI-related differences in this association.

Fig. 2.

Subgroup analysis forest plot

Restricted cubic spline analysis

To investigate the potential nonlinear relationship between RAR and DKD risk, a restricted cubic spline (RCS) regression was performed. As shown in Fig. 3, in the unadjusted model (Model 1), a significant nonlinear association was observed between RAR and the risk of DKD (P for nonlinearity < 0.001). However, this nonlinear relationship was no longer significant in the partially adjusted model (Model 2) and the fully adjusted model (Model 3) (P for nonlinearity > 0.05). These findings suggest that although a nonlinear trend may exist in the unadjusted analysis, the association between RAR and DKD risk tends to become linear after controlling for confounding variables, further supporting a stable positive association when RAR is treated as a continuous variable.

Fig. 3.

Restricted cubic spline (RCS) analysis of the association between RAR and DKD risk across different models. Panel A represents the unadjusted Model 1, Panel B represents the partially adjusted Model 2, and Panel C represents the fully adjusted Model 3

Sensitivity and robustness analyses

To assess the robustness of the results, sensitivity analyses were conducted for both the multivariable logistic regression models and the restricted cubic spline (RCS) models. First, in the logistic regression analysis, key covariates—including BMI, TG, TC, WBC, RBC, CHD, and hypertension—were sequentially excluded, and outliers and extreme values of related variables were removed. The results showed that the association between RAR and DKD risk remained statistically significant across different model specifications (P < 0.05), indicating strong robustness of the findings. Second, in the RCS analysis, each covariate was individually added to the model for adjustment. It was found that the previously significant nonlinear relationship between RAR and DKD risk disappeared once any covariate was included (P > 0.05). This further supports the conclusion that the association between RAR and DKD risk is primarily linear after adjustment for confounders.

Discussion

Based on data from NHANES 2005–2020, this study included a total of 7,191 individuals with diabetes to systematically evaluate the association between the red cell distribution width-to-albumin ratio (RAR) and diabetic kidney disease (DKD). The results showed that RAR levels were significantly higher in DKD patients than in non-DKD individuals, and the prevalence and risk of DKD increased progressively across higher RAR quartiles. Subgroup analyses suggested that the association between RAR and DKD risk may vary across specific populations. Furthermore, restricted cubic spline (RCS) analyses indicated a potentially linear relationship between RAR and DKD risk after adjusting for covariates, with no significant evidence of nonlinearity. Overall, this study provides novel epidemiological evidence supporting a positive association between elevated RAR and increased DKD risk in individuals with diabetes, in line with previous findings linking RAR to cardiovascular events in the general population.

As a composite indicator that integrates red blood cell distribution variability and nutritional-inflammatory status, the red cell distribution width-to-albumin ratio (RAR) has gained increasing attention for its prognostic value in various chronic diseases in recent years. In the field of kidney diseases, previous studies have confirmed that elevated RAR levels are closely associated with deterioration in renal function. For instance, a cohort study from Japan involving patients with chronic kidney disease (CKD) reported that higher RAR levels significantly predicted the risk of end-stage kidney disease (ESKD), with patients in the highest RAR tertile having nearly a threefold increased risk compared to those in the lowest tertile (HR = 2.92, 95% CI: 1.44–5.94) [29]. As one of the components of RAR, red cell distribution width (RDW) has also been recognized as an independent predictor of renal impairment in early-stage diabetic kidney disease (DKD) [3]. One study found that when RDW levels exceed 14.5%, CKD patients experience significantly faster renal function decline [30]. These epidemiological findings support the clinical predictive significance of RAR in chronic kidney disease, particularly in diabetes-related nephropathy, suggesting that RAR may be involved in the development and progression of DKD.

From a mechanistic perspective, an increase in RAR may be involved in the initiation and progression of DKD via several pathological mechanisms. First, increased RDW indicates greater red blood cell size variability, which is typically linked to chronic inflammation, impaired erythropoiesis, and oxidative stress [31–33]. Under diabetic conditions, hyperglycemia and insulin resistance lead to excessive ROS production, which destabilizes red blood cell membranes, shortens their lifespan, and increases volume heterogeneity [34]. Inflammatory cytokines such as IL-6 and TNF-α can suppress the maturation of erythroid precursor cells and activate JAK/STAT and NF-κB signaling pathways, contributing to glomerular filtration barrier disruption and tubular injury, which are key mechanisms in DKD pathogenesis [35–37]. Moreover, serum albumin—the denominator in RAR—reflects the host’s nutritional status and inflammatory burden. In patients with DKD, persistent proteinuria, inflammation-induced suppression of hepatic albumin synthesis, and poor nutritional intake may all contribute to hypoalbuminemia [38, 39]. Low albumin levels not only impair colloid osmotic pressure maintenance but are also closely associated with endothelial dysfunction and oxidative stress, accelerating renal function deterioration [40]. Furthermore, elevated RAR may also indicate increased blood viscosity and impaired microcirculatory perfusion. Abnormal erythrocyte morphology and hypoalbuminemia together impair hemorheology, promoting chronic renal hypoxia and interstitial fibrosis—core mechanisms in DKD progression [41, 42]. Therefore, elevated RAR likely reflects a physiological state of “high inflammatory load and poor nutritional status” in DKD patients, mechanistically explaining its association with adverse renal outcomes.

Notably, in the restricted cubic spline (RCS) analysis, the unadjusted model showed a significant nonlinear association between RAR and DKD risk; however, this nonlinear relationship disappeared after adjusting for demographic and clinical covariates, revealing a linear trend. This suggests that the initially observed nonlinearity may be influenced by confounding factors such as age, sex, and BMI. After adjustment, the relationship between RAR and DKD risk is more stable and linear, underscoring the importance of controlling confounders in epidemiological studies. Future research could further explore how these confounders modulate the association between RAR and DKD and examine risk differences in specific populations.

RAR, as a composite biomarker integrating information on inflammation and nutritional status, is derived from routine blood tests, making it easily accessible and cost-effective, with promising clinical applicability. In the assessment of DKD risk, RAR demonstrates superior predictive performance and offers greater discriminatory power than individual markers such as RDW or serum albumin alone. Moreover, it exhibits relatively high stability and is less susceptible to confounding factors like sampling time or fluid infusion status, suggesting its potential for early screening and dynamic monitoring of renal impairment in patients with diabetes. Notably, previous studies have also shown that RAR has prognostic value in various chronic diseases, including cardiovascular diseases and cancers [14, 15, 17, 18], implying that its utility may extend beyond kidney diseases and could serve as a supplementary tool for chronic disease risk stratification.

Although this study is based on nationally representative NHANES data, providing generalizability and robust statistical power, several limitations should be acknowledged. First, the cross-sectional design precludes inference of a causal relationship between RAR and DKD; thus, future longitudinal cohort studies are needed to clarify the temporal sequence and predictive validity of RAR in the development and progression of DKD. Second, the study relied primarily on laboratory indices and questionnaire data, lacking inflammatory cytokines and oxidative stress biomarkers. Future investigations incorporating a broader range of biological markers may help elucidate the mechanistic pathways linking RAR to renal injury. Moreover, due to limitations of the NHANES dataset, we were unable to control for medication use such as renin-angiotensin-aldosterone system (RAAS) inhibitors and sodium-glucose cotransporter-2 (SGLT2) inhibitors, which can influence kidney function and inflammatory markers. This may represent a source of residual confounding in our analysis. Finally, despite adjusting for multiple potential confounders, residual confounding—such as medication use and dietary patterns—may still influence the results. Further studies leveraging real-world data or interventional trials are warranted to validate the clinical applicability and robustness of RAR as a risk assessment tool.

Conclusions

This study demonstrates that elevated RAR levels are significantly associated with an increased risk of diabetic kidney disease (DKD), with this association remaining stable across various subgroups. As a simple and cost-effective marker reflecting both inflammatory and nutritional status, RAR may serve as a promising predictive indicator for identifying individuals at high risk of DKD in the diabetic population. Prospective studies are warranted to validate its predictive value, clarify potential causal relationships, and assess its clinical utility in early screening and risk stratification of chronic kidney disease.

Abbreviations

NHANES

National Health and Nutrition Examination Survey

RAR

Red Blood Cell Distribution Width to Albumin Ratio

DKD

Diabetic kidney disease

RDW

Red cell distribution width

WBC

White blood cell count

RBC

Red blood cell count

CHD

Coronary heart disease

PIR

Family income and poverty ratio

BMI

Body mass index

DBP

Diastolic Blood Pressure

SBP

Systolic Blood Pressure

TG

Triglyceride

TC

Total Cholesterol

HDL-C

High-density lipoprotein cholesterol

LDL-C

Low-density lipoprotein cholesterol

HbA1c

Glycohemoglobin

INS

Insulin

FPG

Fasting plasma glucose

eGFR

Estimated glomerular filtration rate

Authors’ contributions

S.Z.: contributed to writing the original draft, reviewing and editing, methodology, data analysis, and visualization. Y.S.: contributed to writing the original draft, reviewing and editing, methodology, data analysis, and visualization. S.S.: participated in reviewing and editing, as well as data analysis. L.J.: Methodology; Validation; Writing-review; supervision. All authors have read and approved the final manuscript for publication.

Funding

This research was supported by the Heilongjiang Provincial Administration of Traditional Chinese Medicine Research Project of Traditional Chinese Medicine (Grant No. ZHY2024-024), the China Rehabilitation Medical Association fund project (No. KFKT-2024-006), the Heilongjiang Provincial Health Commission Research Project (No. 20222121020963), and the Fourth Affiliated Hospital of Heilongjiang University of Traditional Chinese Medicine talent special fund support project (No. FSRC2024ZD02).

Data availability

All datasets provided in this study are derived from the National Health and Nutrition Examination Survey (NHANES) and are accessible on the NHANES official website at [https://wwwn.cdc.gov/nchs/nhanes/Default.aspx].

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki. All data used in this study are publicly available from the National Health and Nutrition Examination Survey (NHANES), and additional ethics approval was not required. The NHANES protocols were approved by the National Center for Health Statistics (NCHS) Research Ethics Review Board, and written informed consent was obtained from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.