Association between renal function and diabetic retinopathy: a mediation analysis of geriatric nutritional risk index

Abstract

Background

Diabetes retinopathy (DR) is a prevalent microvascular complication of type 2 diabetes mellitus (T2DM). This study investigated the correlation between renal function and DR, as well as the potential mediating role of the geriatric nutritional risk index (GNRI).

Method

We classified 1122 adults with T2DM aged ≥ 40 years from the National Health and Nutrition Examination Survey database (2005–2008) into 2 groups: those with DR and those without DR. We used multivariate logistic regression analysis and restricted cubic spline (RCS) model to explore the relationship between renal function indicators and DR. Additionally, we analyzed the mediating impact of GNRI on renal function and DR.

Result

After accounting for all covariates, the weighted multivariate analysis revealed significant associations between renal function markers and DR. Specifically, creatinine, albumin, blood urea nitrogen, and serum uric acid to creatinine ratio (SUACr) were significantly correlated with DR in serum examination, while creatinine was the only marker correlated with DR in urine. GNRI was negatively correlated with DR (odds ratio 0.94, 95% CI 0.92–0.99). Weighted linear regression showed a negative association between SUACr and GNRI (β = 0.37; 95% CI 0.12–0.62). The RCS analysis showed a nonlinear association between serum creatinine and DR (P_non-linear = 0.013). GNRI mediated 14.4% of the relationship between SUACr and DR.

Conclusion

Our study adds to previous research by analyzing the associations between renal function indicators and DR. Furthermore, we highlight the mediating effect of GNRI, suggesting its potential utility as a predictive and treatment index for assessing renal function and DR.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13098-025-01658-z.

Introduction

Worldwide, the prevalence of diabetes is rising, with type 2 diabetes mellitus (T2DM) accounting for approximately 90% of all cases. T2DM is a chronic metabolic disease resulting from the interaction of genetic and environmental factors [1]. Many individuals with diabetes also experience hyperlipidemia and obesity, with dietary patterns being a potential contributing factor [2]. Diabetic retinopathy (DR) is one of the most common late complications of T2DM, affecting approximately one-third of people with diabetes [3]. In the later stages of diabetes, risk factors of DR include inadequate glycemic control, elevated blood pressure, and poor lipid levels that might be associated with endothelial dysfunction [4]. DR is also a leading cause of severe visual impairment in developed countries, with 4.4–8.2% of the American population experiencing severe visual impairment due to the condition [5]. DR is particularly prevalent among middle-aged and elderly individuals [6]. Moreover, the presence of DR can indicate the potential existence of life-threatening macrovascular disease [7].

Patients with T2DM also tend to experience renal function decline, which might progress to diabetic nephropathy (DN) and renal failure [8]. With the increasing incidence and mortality of diabetes, the prevalence of DN is also rising rapidly, making it one of the leading causes of chronic renal failure [8]. Various markers are used to assess renal function, and there is growing attention to indicators of renal dysfunction. Muscle metabolism produces creatinine which is filtered through the glomeruli of the kidneys and discarded in the urine, and creatinine levels in blood and urine are reliable indicators of kidney function. As albumin is filtered through the glomerulus and reabsorbed in the proximal tubule in the kidney, the amount of albumin reflects changes in renal function. BUN is a metabolic waste product of protein catabolism in the urinary cycle that is removed from the blood by the kidneys and is associated with kidney impairment [9]. Uric acid is a degradation product of purines, and hyperuricemia is common in DM patients [10]. Several renal function markers, including uric acid, albumin, and creatinine in both serum and urine, have been reported to be associated with diabetes and its complications, including DR [11–13]. However, the nature of this association varies across different studies. Several studies reported uric acid was an independent risk factor for DR in Chinese T2DM patients [14–16], while previous study considered there is no correlation between uric acid and DR [17]. Another research even found low uric acid was associated with DR in diabetic men [18]. Thus, the correlation between renal function and DR remains controversial [19].

The Geriatric Nutrition Risk Index (GNRI) is commonly used as a simple and efficient tool to assess nutritional status in individuals with chronic diseases [20]. The index is measured as serum albumin and the ratio of ideal weight to actual weight, and is an appropriate instrument for nutritional screening of kidney diseases patients [21]. It is as well tightly associated with renal function as a factor in renal disease progression [22, 23]. In addition, it is used in conjunction with creatinine levels to assess the prognosis of patients undergoing hemodialysis [24, 25]. The role of dietary patterns in the progression of DR remains a prominent research focus. Studies suggest that certain diets, including those rich in fruits, vegetables, and fish, might have a protective effect against the exacerbation of DR [26]. However, these research on correlation between GNRI and DR were few and mainly focused on Asian population. There was a negative association between GNRI scores and prevalence of DR among T2DM patients [27, 28]. Therefore, GNRI might have an important tool to predict DR risk in T2DM patients. Collectively, we hypothesized that renal function may influence DR risk by promoting GNRI based on the association of GNRI for renal function and DR respectively.

In our study, we conducted a comprehensive examination to assess the association between various renal function markers and DR in participants with T2DM in the United States while also analyzing the mediating effect of GNRI on this relationship.

Methods

Data source

The overall study design is shown in Fig. 1. We selected data from two consecutive National Health and Nutrition Examination Surveys (NHANES) cycles: 2005–2006 and 2007–2008. NHANES is a cross-sectional survey that assesses the health and national status of the US population. Data were obtained from the public NHANES website. The NHANES protocol is approved by the National Center for Health Statistics Research Ethics Review Board, and all participants provided written informed consent. The NHANES 2005–2008 cycle included a total of 20,497 participants. We excluded those without fundus photography and complete retinal imaging status. Overall, 7081 participants aged ≥ 40 years were enrolled in the study. Among these, 317 had missing data on GNRI, and 1108 lacked data with definite DR examination results. Finally, we selected participants with confirmed DR and those without DR for this study.

Fig. 1.

The flowchart of the study population

Renal function

Kidney function was assessed using several key parameters categorized into two categories based on their source. Serum markers included blood urea nitrogen (BUN), creatinine, albumin, serum uric acid (SUA), and serum uric acid to serum creatinine (SUACr). Urinary biomarkers included creatinine, albumin, urinary albumin to creatinine ratio (UACR), and estimated glomerular filtration rate (eGFR). The eGFR score calculations adhered to the guidelines provided by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [29].

Diabetic retinopathy

Patients with T2DM were identified based on the diagnostic criteria established by the American Diabetes Association (ADA). These criteria included: (1) a self-reported diabetes diagnosis by a healthcare provider or a documented history of treatment with oral hypoglycemic agents or insulin, (2) a glycosylated hemoglobin (HbA1c) level of ≥ 6.5% (48 mmol/mol), (3) a blood glucose level of ≥ 11.1 mmol/L after a 2-h oral glucose tolerance test (OGTT), and (4) a fasting blood glucose level of ≥ 260 mg/dL. Participants who were healthy, had undiagnosed diabetes, or were classified as prediabetic were excluded from our study. DR was assessed using nonmydriatic retinal photographs of T2DM patients and graded according to the Early Treatment Diabetic Retinopathy Study (ETDRS) scale. The severity of retinopathy was assessed according to the ETDRS grading scale: grade ≥ 14 was defined as DR. The eye with the most severe retinopathy was selected in our study [30]. All individuals diagnosed with DR were confirmed to have both T2DM and DR.

Mediator

The GNRI is a widely used and validated tool for assessing the risk of malnutrition in middle-aged and older individuals. In our study, we used the GNRI to evaluate the nutritional status of participants, with the index calculated based on albumin levels and weight measurements. The GNRI formula is defined as follows:

$$\text{GNRI}=\left(1.489\times \text{albumin}\left(\text{g}/\text{L}\right)\right)+\left(41.7\times \left(\text{measured}\text{body}\text{weight}\left(\text{kg}\right)/\text{ideal}\text{body}\text{weight}\left(\text{kg}\right)\right)\right)$$

Ideal body weight was determined using the Lorentz formula, which calculates ideal weight as (height (cm) − 100) − (height (cm) − 150)/4 for men and (height (cm) − 100) − (height (cm) − 150)/2 for women [31].

Covariates

Potential confounders in this study included demographic data, lifestyle behaviors, and chronic diseases. We selected potentially confounding variables as covariates, each of which was associated with renal function as well as GNRI and was associated with outcome DR. Furthermore, no covariance existed between these confounding factors. The sociodemographic variables considered included sex, age, race, education status, marital status, and poverty status. Poverty status was categorized into two levels based on the poverty income ratio (PIR): poverty (PIR < 1) and not poverty (PIR ≥ 1). Lifestyle behaviors of interest were body mass index (BMI), smoking, and drinking, while chronic diseases associated with DR included cardiovascular disease (CVD), hyperlipidemia, and hypertension. Hypertension was operationally defined as meeting criteria for either a self-reported physician diagnosis, an average of 3 systolic blood pressure (SBP) readings ≥ 140 mmHg, an average of 3 diastolic blood pressure (DBP) readings ≥ 90 mmHg, or the use of antihypertensive medications for primary hypertension. Participants were classified as having hyperlipidemia if they met any of the following criteria: total cholesterol 200 mg/dL, triglycerides 150 mg/dL, high-density lipoprotein (HDL) 40 mg/dL in males and 50 mg/dL in females, or low-density lipoprotein (LDL) 130 mg/dL [32]. CVD was defined based on self-reported history and included conditions such as congestive heart failure, coronary heart disease, angina, or myocardial infarction.

Statistical analysis

Continuous variables are presented as weighted mean ± standard deviation (SD), and categorical variables are presented as weighted percentages and numbers. ANOVA and Chi-square test were performed separately on individuals with and without DR. The correlation between renal function and DR, as well as GNRI and DR, was assessed using weighted logistic regression model. Weighted linear regression models were used to explore the relationship between renal function and GNRI. Albumin and related ratios were excluded from GNRI calculations due to considerations related to albumin. All regression analyses included three models. A crude model was unadjusted, and model 1 that partially adjusted (sex, age, and race). Model 2 that fully adjusted by adding marital status, smoking, drinking, poverty status, education, BMI, CVD, hyperlipidemia, and hypertension. The crude model demonstrated a direct relationship between exposure and outcome, with Model 1 capturing results after eliminating the biasing effects of the covariates sex, age, and race on the relationship. In contrast, Model 2 results were stable after controlling for the effects of all selected covariates on the association. The nonlinear relationship was analyzed using weighted restricted cubic splines (RCS) while adjusting for all variables. Mediation analysis was constructed using a two-step approach, with the first step involving a fully adjusted model examining the association between the renal function indicators and GNRI, as well as GNRI and DR. The second step investigated the potential mediating role of GNRI. The mediation analysis was performed to assess the total effect, indirect effect, and direct effect of renal function on DR mediated by GNRI. The mediated proportion was estimated with 95% confidence intervals (CI) using nonparametric bootstrapping with 1000 iterations. Subgroup analyses were conducted and presented as a forest plot. All statistical analyses and graphical representations were performed using R software (version 4.3.1). Statistical significance was defined as a two-sided P-value < 0.05.

Results

Clinical characteristics

From 2005 to 2008, a total of 1122 participants with diabetes were enrolled in the study, with a prevalence of DR at 25.07% (Table 1). The weighted number of participants with DR was 8,634,166. Among the demographic characteristics, males and older individuals were found to have a higher likelihood of developing DR. Non-Hispanic White individuals also exhibited a higher prevalence of DR compared to other racial groups. Approximately one-third of individuals with CVD experienced DR, which was a significant difference compared to those without CVD. In addition, all renal function markers, except for SUA and eGFR, showed significant differences between the two groups. This pattern was consistently observed throughout the study period.

Table 1.

Characteristics and clinical features of participants from 2005 to 2008

Characteristic	OverallN = 34,443,298	No DRN = 25,809,131	DRN = 8,634,166	P-value
Age (years)	60.00 ± 11.37	60.00 ± 11.31	62.00 ± 11.41	0.040
Sex				0.004
Female	16,972,448 (49%)	13,249,748 (51%)	3,722,700 (43%)
Male	17,470,850 (51%)	12,559,384 (49%)	4,911,466 (57%)
Race				0.019
Mexican American	2,506,813 (7.3%)	1,819,624 (7.1%)	687,189 (8.0%)
Non-Hispanic Black	4,912,405 (14%)	3,187,062 (12%)	1,725,342 (20%)
Non-Hispanic White	24,270,388 (70%)	18,579,653 (72%)	5,690,735 (66%)
Other Hispanic	1,456,273 (4.2%)	1,073,625 (4.2%)	382,647 (4.4%)
Other Race-Including Multi-Racial	1,297,419 (3.8%)	1,149,167 (4.5%)	148,252 (1.7%)
Marital status				0.528
Divorced	4,599,025 (13%)	3,286,709 (13%)	1,312,316 (15%)
Living with partner	1,091,283 (3.2%)	887,300 (3.4%)	203,983 (2.4%)
Married	22,124,715 (64%)	16,720,353 (65%)	5,404,362 (63%)
Never married	1,797,473 (5.2%)	1,451,786 (5.6%)	345,687 (4.0%)
Separated	659,762 (1.9%)	415,002 (1.6%)	244,760 (2.8%)
Widowed	4,171,040 (12%)	3,047,980 (12%)	1,123,059 (13%)
Education status				0.174
College graduate/AA degree	15,169,334 (44%)	11,607,803 (45%)	3,561,530 (41%)
High School Grad/GED or equivalent	10,565,077 (31%)	8,179,665 (32%)	2,385,412 (28%)
Less than 12th grade	8,708,887 (25%)	6,021,662 (23%)	2,687,224 (31%)
Poverty				0.679
No	30,257,951 (88%)	22,725,130 (88%)	7,532,822 (87%)
Yes	4,185,346 (12%)	3,084,002 (12%)	1,101,345 (13%)
BMI	31.48 ± 7.23	31.60 ± 7.41	31.00 ± 6.66	0.332
Smoke				0.115
Former	12,234,651 (36%)	9,528,968 (37%)	2,705,683 (31%)
Never	16,427,668 (48%)	11,737,710 (45%)	4,689,958 (54%)
Now	5,780,978 (17%)	4,542,453 (18%)	1,238,525 (14%)
Alcohol				0.161
1–10 drinks/month	20,639,215 (60%)	16,010,658 (62%)	4,628,557 (54%)
10 + drinks/month	974,875 (2.8%)	671,005 (2.6%)	303,869 (3.5%)
Non-drinker	12,829,208 (37%)	9,127,468 (35%)	3,701,740 (43%)
CVD				0.002
No	25,665,502 (75%)	20,023,417 (78%)	5,642,085 (65%)
Yes	8,777,796 (25%)	5,785,715 (22%)	2,992,081 (35%)
GNRI	125 ± 15	126 ± 15	124 ± 14	0.004
Hypertension				0.895
No	9,692,540 (28%)	7,297,144 (28%)	2,395,397 (28%)
Yes	24,750,757 (72%)	18,511,987 (72%)	6,238,770 (72%)
Hyperlipidemia				0.586
No	3,819,902 (11%)	2,760,596 (11%)	1,059,306 (12%)
Yes	30,623,395 (89%)	23,048,535 (89%)	7,574,860 (88%)
Renal function
Serum creatinine (mg/dL)	0.90 (0.77, 1.10)	0.90 (0.77, 1.05)	1.00 (0.76, 1.20)	0.014
Uric creatinine (mg/dL)	107 (69, 152)	111 (74, 154)	98 (59, 148)	0.026
SUACr	6.23 ± 1.85	6.41 ± 1.80	5.64 ± 1.93	< 0.001
SUA	5.60 ± 1.58	5.70 ± 1.53	5.50 ± 1.72	0.151
Uric albumin (mg/L)	12 (6, 32)	11 (6, 29)	14 (8, 54)	0.016
Serum albumin (g/dL)	4.10 ± 0.33	4.20 ± 0.31	4.10 ± 0.36	0.013
eGFR	84 ± 22	86 ± 20	78 ± 26	0.069
BUN (mg/dL)	14 ± 7	14 ± 7	15 ± 9	0.008
UACR	0.12 (0.06, 0.29)	0.10 (0.06, 0.24)	0.17 (0.07, 0.61)	< 0.001

BMI body mass index, SUACr serum uric acid to creatinine ratio, SUA serum uric acid, BUN blood urea nitrogen, UACR urinary albumin to creatinine ratio, eGFR estimated glomerular filtration rate, GNRI geriatric nutritional risk index, CVD cardiovascular disease

Association between renal function and DR

Among the nine renal function indexes examined, five indicators significantly correlated with DR in both the unadjusted and fully adjusted models (Table 2). Specifically, serum creatinine exhibited a positive correlation with DR, with a 50.7% increase in the likelihood of DR for each unit increase in the serum creatinine (OR 1.51, 95% CI 1.03–2.20; P = 0.036). Similarly, BUN was positively related to DR, with an OR of 1.04 for DR (95% CI 1.02–1.06; P = 0.003). Conversely, serum albumin, SUACr, and urinary creatinine were found to have a negative relationship with DR. The RCS analysis showed a linear association for all indicators except for serum creatinine (P_non-linear = 0.013), which exhibited a lower value interval of 0.86 mg/dL. The correlation between serum creatinine and DR exhibited a U-shaped pattern, and a negative association was observed when serum creatinine levels were below 0.86 mg/dL and a positive association was observed at higher values (Fig. 2).

Table 2.

Logistic regression of renal function and diabetic retinopathy

Category	Characteristic	Crude Model, OR	95% CI	P-value	Model 1, OR	95% CI	P-value	Model 2, OR	95% CI	P-value
Serum	Creatinine	1.873	1.279, 2.739	0.002	1.547	1.125, 2.124	0.009	1.509	1.039, 2.189	0.036
	Albumin	0.453	0.291, 0.704	< 0.001	0.411	0.251, 0.672	0.001	0.358	0.209, 0.611	0.002
	BUN	1.045	1.022, 1.068	< 0.001	1.039	1.019, 1.061	< 0.001	1.040	1.017, 1.064	0.003
	SUACr	0.821	0.746, 0.902	< 0.001	0.853	0.777, 0.937	0.002	0.852	0.758, 0.959	0.015
	SUA	1.049	0.851, 1.067	0.395	0.909	0.802, 1.031	0.131	0.886	0.769, 1.022	0.087
Urinary	Creatinine	0.997	0.994, 1.000	0.048	0.995	0.992, 0.999	0.010	0.995	0.991, 0.999	0.027
	Albumin	1.000	0.999, 1.002	0.102	1.001	1.000, 1.002	0.115	1.001	0.999, 1.002	0.176
	UACR	1.081	0.992, 1.177	0.073	1.073	0.993, 1.160	0.071	1.080	0.978, 1.192	0.116
	eGFR	0.989	0.980, 0.999	0.026	0.991	0.981, 1.002	0.091	0.992	0.980, 1.005	0.194

Crude model, unadjusted. Model 1, adjusted for age, sex, race. Model 2, adjusted for age, sex, race, marital status, education status, poverty status, smoking, alcohol, BMI, hypertension, hyperlipidemia, and cardiovascular disease. Significant results with P-values are marked in bold

SUACr serum uric acid to creatinine ratio, SUA serum uric acid, BUN blood urea nitrogen, UACR urinary albumin to creatinine ratio, eGFR estimated glomerular filtration rate, OR odds ratio, CI confidence interval

Fig. 2.

The RCS models the linear relationships between renal function markers and DR. A Serum creatinine. B Serum albumin. C BUN. D Urinary creatinine. E SUACr. F eGFR

Association between GNRI and DR

Table 3 illustrates the inverse relationship between the GNRI and DR in individuals with T2DM. The GNRI score exhibited a significant negative association with DR in the unadjusted model (OR 0.98, 95% CI 0.97–0.99; P = 0.006), which remained statistically significant in both partially adjusted (OR 0.99, 95% CI 0.98–0.99; P = 0.048) and fully adjusted models (OR 0.94, 95% CI 0.92–0.99; P = 0.005). A reduction of one unit in the GNRI score was linked to a decrease of 0.058 units in DR after controlling for all covariates.

Table 3.

Logistic regression of GNRI and diabetic retinopathy

Model	OR	95% CI	P-value
Crude Model	0.985	0.974, 0.995	0.006
Model 1	0.990	0.978, 0.998	0.048
Model 2	0.941	0.918, 0.986	0.005

Crude model, unadjusted. Model 1, adjusted for age, sex, race. Model 2, adjusted for age, sex, race, marital status, education status, poverty status, smoking, alcohol, BMI, hypertension, hyperlipidemia, and cardiovascular disease

GNRI geriatric nutritional risk index, OR odds ratio, CI confidence interval

Association between renal function and GNRI

While the renal function indicators demonstrated a significant correlation with the GNRI score through the weighted linear regression in the crude model except BUN and urinary creatinine (Table 4), only the inverse correlation between SUACr and GNRI remained statistically significant in the fully adjusted model (β = 0.94, 95% CI 0.98–1.02; P = 0.005).

Table 4.

Linear regression of renal function makers and GNRI

Category	Characteristic	Crude Model, β	95% CI	P-value	Model1, β	95% CI	P-value	Model 2, β	95% CI	P-value
Blood	SUA	1.324	0.582, 2.065	0.001	2.021	1.264, 2.780	< 0.001	0.264	− 0.091, 0.618	0.127
	Creatinine	− 3.201	− 4.663, − 1.739	< 0.001	− 0.528	− 1.806, 0.751	0.402	− 0.590	− 1.205, 0.024	0.071
	BUN	− 0.089	− 0.211, 0.032	0.143	0.152	0.059, 0.245	0.003	− 0.019	− 0.068, 0.030	0.328
	SUACr	2.299	1.804, 2.808	< 0.001	1.498	0.929, 2.096	< 0.001	0.368	0.121, 0.618	0.009
Urinary	Creatinine	0.011	− 0.003, 0.026	0.122	0.019	0.002, 0.036	0.031	0.002	− 0.005, 0.009	0.487
	eGFR	0.076	0.021, 0.118	< 0.001	− 0.030	− 0.079, 0.018	0.211	0.016	− 0.005, 0.037	0.109

Crude model, unadjusted. Model 1, adjusted for age, sex, race. Model 2, adjusted for age, sex, race, marital status, education status, poverty status, smoke, alcohol, BMI, hypertension, hyperlipidemia and cardiovascular disease. Significant results with P-values are marked in bold

SUACr serum uric acid to creatinine ratio, BUN blood urea nitrogen, SUA serum uric acid, OR odds ratio, CI confidence interval

Mediation effect of the GNRI on SUACr and DR

The mediation analysis was carried out based on the results from regression analysis (Fig. 3; Table 5). Upon adjusting for all the covariates, a significant correlation was found between SUACr and DR significantly, with a total effect of − 0.026 (P = 0.006). The GNRI score was observed to significantly mediate the relationship between the SUACr and DR, accounting for 14.4% of the effect (indirect effect: − 0.023, P = 0.020). Additionally, the direct effect of the SUACr on DR remained significant (direct effect: − 0.004, P = 0.012).

Fig. 3.

The mediation analysis for the association between serum uric acid to serum creatinine ratio and diabetic retinopathy with the GNRI level as a potential mediator

Table 5.

Bootstrap analysis of the mediation effect of the model

Effect type	Effect value	95% CI	P-value
Direct effect	− 0.004	− 0.007, − 0.001	0.012
Indirect effect	− 0.023	− 0.041, − 0.004	0.020
Total effect	− 0.026	− 0.045, − 0.009	0.006

CI confidence interval

Subgroup analysis between SUACr and DR

Based on the potential influencing factors on DR, subgroup analysis and interaction tests were conducted (Supplementary Table 1). The association between SUACr and DR remained significant in age ≤ 63 year old, male gender, and participants without hyperlipidemia (P < 0.01) (Fig. 4). Furthermore, there was an interaction between age and SUACr (P for interaction = 0.011). No interaction effect was observed in other characteristics, indicating the robustness of our results.

Fig. 4.

Forest plot of subgroup analysis between serum uric acid to serum creatinine ratio and diabetic retinopathy

Discussion

Our research investigated various renal function indices and found significant associations with DR in the US population. The SUACr exhibited a negative correlation with DR. Additionally, we observed that the GNRI had a negative mediating effect on the relationship between SUACr and DR. These findings could provide new strategies for the prevention and treatment of DR in individuals with T2DM.

Similar to DR, diabetic kidney disease (DKD) is a major microvascular complication of DM. DR was shown to be strongly associated with preclinical morphological changes in DKD [33]. However, more than 10% of DR patients remain without renal insufficiency [34]. Therefore, the association between DR and renal function in patients with DM remains uncertain. The decline in renal function in T2DM patients is associated with retinopathy status, and diverse indicators have been used to assess the decline in renal function [35]. SUA levels were closely tied to renal insufficiency and a critical element in the progression of kidney disease. Although our study did not find a significant association between SUA and DR, many previous studies have reported that SUA was a risk factor for DR in different populations [11, 36, 37]. Among the Asian population, serum creatinine was reported to be associated with the severity of DR in individuals with T2DM [12, 38]. Our study found significant relationships indicating that serum creatinine is a risk factor for DR in the American population. Furthermore, the RCS curve showed a nonlinear relationship between serum creatinine and DR. A serum creatinine level of 0.86 mg/dL was the turning plot, with a negative association observed at lower values. The U-shaped relationship may provide a novel direction in exploring serum creatinine with diabetic complications and other chronic diseases. SUACr was calculated using SUA and serum creatinine. It was considered as an endogenous uric acid level more precisely than SUA, which represents renal function-normalized SUA. However, it is unclear whether the index has relevant risks with DR. Thus, we explored the relationship between SUACr and DR and found a negative correlation in the US population. This result suggests that SUACr might have a significant impact on DR. More research should be conducted across diverse populations and locations, as the related studies remain limited. Serum albumin was found to have a significant negative correlation with DR in patients with T2DM, consistent with the findings of Wang et al. from their study on NHANES 2011–2020 [39]. A cross-sectional study of the Chinese population also reached a similar conclusion [40]. BUN is a common laboratory examination that measures the amount of nitrogen in the blood, which originates from the kidney [41]. Du et al. found that BUN levels were related to DR in their study using NHANES 2005–2008 cycle data [42]. In the Chinese population, BUN was found to be a risk factor in DR but influenced by the duration of DR [43]. Our research also confirmed the risk factor in DR.

In our research, we did not identify urinary biomarkers other than urinary creatinine that were associated with DR. However, other studies have made different conclusions about the correlations. Chen et al. identified urinary albumin as an obvious risk factor for DR in Chinese patients [37]. UACR and eGFR are key biomarkers of DKD. Although some previous studies evaluated different populations and found that UACR and eGFR could predict the occurrence of DR in patients with DM [13, 44], our study did not find any such associations. The result for urinary creatinine differed from serum creatinine, as the urinary creatinine was affected differently by different functional and structural kidney damage, such as glomerular filtration and tubular secretion. Interestingly, our research indicated that the association between serum creatinine and DR was the opposite of that observed for urinary creatinine. However, the RCS curve showed this relationship was nonlinear, suggesting that the correlation remained consistent up to a serum creatinine was 0.86 mg/dL. In addition, our subgroup analysis showed a significant interaction between age and DR. Several studies agreed that age was an important predictor for DR development [45, 46]. Hence, age period might be a special focus in evaluating renal function of DR patients.

The prevalence of malnutrition in DR patients with DM has been studied infrequently in clinical trials. GNRI, which assesses the nutritional status of middle-aged and older people, has been identified as a protective factor in chronic diseases [20, 47]. It is associated with the severity of nutritional status-related complications [31] and has been reported as a prognostic marker for foot ulcers and hypoglycemia in the DM population [48, 49]. Two studies conducted on Asian patients have reported that GNRI had an inverse association with the prevalence of DR [27, 28]. We analyzed the US population and found a negative correlation between GNRI and DR, aligning with these previous findings.

GNRI is composed of serum albumin and weight. DR has a complex pathophysiology driven by hyperglycemia and involves multiple mechanisms, including mitochondrial damage, cellular apoptosis, inflammation, lipid peroxidation, and structural and functional alterations [50]. Serum albumin, the most abundant protein in blood plasma across species, including humans [51], has immunomodulatory properties and can inhibit inflammation by affecting neutrophils [52]. Inflammation and immunomodulation are central to the pathogenesis of diabetes mellitus (DM) and its microvascular complications. In patients with DR, inflammatory cytokines such as TNF-α, IL-6, and IL-1β are elevated in serum, vitreous, and aqueous humor [53]. In general, serum albumin plays a role in the inflammatory and immunomodulatory processes of DR. Elevated serum albumin levels might serve as a potential predictor of the risk of DR in patients with T2DM.

Another key component of GNRI is weight. Overweight and obesity were reported as risk factors for DR in previous studies. Obesity was related to some chronic diseases such as hyperlipidemia and hypertension, which affected the prevalence of DR [54]. Obesity is closely associated with chronic inflammation, and there is growing attention to the impact of inflammatory cytokines on obese populations [55]. In addition to its effects on inflammation, obesity can increase oxidative stress throughout the body [56]. Oxidative stress is the basis of many neurodegenerative diseases, including DR [57]. Thus, reducing obesity can control DR by addressing both inflammation and oxidative stress. While BMI has shown either no or a positive correlation with DR in Western populations [5, 58, 59], this association remains ambiguous.

This study showed that GNRI plays a partial mediating role in the relationship between SUACr and DR. We found the relationship between SUACr and DR among the US population as the association remained unknown. SUACr was considered as a new biomarker and was regarded as a superior indicator of net uric acid production, this relationship could provide a new evaluation of renal function in DR patients. GNRI is a simple and readily available tool for nutritional evaluation. Therefore, this marker has potential clinical utility in the prognosis and diagnosis of DR. This means the risk of DR might be modulated through diet changes in people with abnormal renal function. Albumin and weight are the concrete indicators that should be given higher importance in the diet evaluation. Uncovering the mediation effect offers new routes for therapeutic intervention and prevention for DR in patients with T2DM.

However, our study has some limitations. We did not include clinical information from participants outside the cycles we selected in the NHANES database, such as Cystatin C. In addition, we excluded indicators based on albumin as our mediator calculated through serum albumin. Future research should consider other potential mediators beyond albumin in relation to renal function biomarkers to DR. Despite taking into account many critical confounders, we cannot completely eliminate the risk of residual confounding from unmeasured or unidentified confounders. Due to the nature of the cross-sectional study, we cannot establish causative associations between SUACr and DR and the mediation role of GNRI. The strength of the relationships between various renal indicators and DR could not be compared, highlighting the need for future studies to address this limitation.

Conclusions

This research provides preliminary insights into the association between renal function indicators and DR in the T2DM population, highlighting the mediation effect of GNRI between renal function and DR risk. Serum albumin, SUACr, urinary creatinine, and GNRI are potential protective factors against DR, while serum creatinine and BUN had positive correlations. GNRI demonstrated a negative correlation with DR and partially mediated the association between SUACr and DR. These findings suggest that managing renal function and dietary patterns may help prevent DR in individuals with T2DM. Even if renal function is impaired, it may still be possible to control nutritional status.

Author contributions

YC and WQ designed the study and searched the data. XM, YY, and YD collated the data. YC, WQ, and TT performed analysis and wrote the manuscript. JC and QZ reviewed the statistical analysis and manuscript. All authors reviewed and agreed on the final version of the manuscript.

Funding

No funding.

Data availability

All the data of our study is available from https://www.cdc.gov/nchs/nhanes/index.htm.

Declarations

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare no competing financial interests.