Non-High-Density Lipoprotein-to-High-Density Lipoprotein Cholesterol Ratio as a Predictive Biomarker for Diabetic Retinopathy Risk: A Population-Based Analysis of US Adults Open Access

Diabetic retinopathy (DR) is one of the most common microvascular complications of diabetes and a leading cause of blindness in adults [1]. With the global prevalence of diabetes on the rise, the incidence of DR is increasing rapidly, presenting a significant public health challenge worldwide [2]. According to the International Diabetes Federation (IDF), the global population of individuals with diabetes is projected to increase from 537 million in 2021 to 783 million by 2045 [3]. Therefore, early prevention and effective management of DR are crucial to reducing the burden of vision loss.

Numerous studies have demonstrated that lipid metabolism disorders are closely associated with diabetes and its complications [4]. High-density lipoprotein cholesterol (HDL-C) is generally considered to have anti-atherosclerotic properties, while non-high-density lipoprotein cholesterol (non-HDL-C) includes components such as low-density lipoprotein cholesterol (LDL-C) and very-low-density lipoprotein, both of which are associated with atherosclerosis [5]. Research has shown a strong correlation between non-HDL-C and the risk of DR, while the protective role of HDL-C cannot be overlooked [5]. In recent years, the ratio of non-HDL-C to HDL-C (NHHR) has emerged as a comprehensive marker of atherosclerotic risk and is increasingly recognized as a predictor of cardiovascular diseases and diabetes complications [6, 7].

Although the prognostic value of NHHR in cardiovascular diseases has been widely acknowledged [8], its association with DR remains underexplored. Previous studies have mostly focused on individual lipid markers, such as LDL-C or HDL-C, in relation to diabetes complications [9, 10]. However, few studies have systematically evaluated the predictive value of NHHR as a composite index in DR. Therefore, investigating the relationship between NHHR and DR could provide more precise risk assessments for patients with diabetes and offer a theoretical basis for developing clinical interventions.

This study utilizes population data from the National Health and Nutrition Examination Survey (NHANES) from 2005 to 2018 to explore the relationship between NHHR and the risk of developing DR. We hypothesize that NHHR is significantly associated with the occurrence of DR and hope to reveal its potential value in DR risk assessment through this research.

The NHANES is an ongoing comprehensive national survey whose core objective is to assess the health and nutritional status of the US population. The study was approved by the National Center for Health Statistics (NCHS) Research Ethics Committee and ensured that all participants signed a written consent form prior to enrollment [11]. The complete dataset from seven data collection cycles (2005 to 2018) was used for this analysis. A total of 6,783 people with diabetes participated in the NHLBI Health and Nutrition Survey. Of these, we excluded participants younger than 18 years of age, those who were pregnant at the time of the survey, and those who lacked data from the NHHR (1,357), as well as those who failed to provide information on retinopathy (491). Ultimately, we selected 3,742 participants for analysis. The participant screening process is shown in Figure 1.

Participants were classified as having diabetes if they met one of the following criteria: (1) a glycosylated hemoglobin level ≥6.5% or a fasting plasma glucose level ≥126 mg/dL; or (2) a self-reported diagnosis of diabetes (“Has your doctor told you you have diabetes?”) or current insulin use (“Are you using insulin?”) [12]. Non-blurred, two-dimensional retinal images were captured using the Canon CR6-45NM and CR4-45NM fundus cameras under dim lighting, with two images per eye (focused on the optic nerve and macula). Image quality was validated by trained graders at the University of Wisconsin’s Department of Ophthalmology. Retinopathy Case Definition: In the NHANES protocol, DR is defined using the Early Treatment Diabetic Retinopathy Study (ETDRS) scale. According to the NHANES criteria, any participant with an ETDRS severity level ≥14 (including levels 14 and above) is classified as having DR. This binary classification (present/absent) does not subdivide DR into mild, moderate, or severe stages for the purposes of the NHANES database. For transparency, we clarify that ETDRS level ≥14 includes a spectrum of lesions, ranging from mild non-proliferative retinopathy (e.g., microaneurysms) to proliferative retinopathy (e.g., neovascularization). The NHANES dataset does not further stratify cases beyond this threshold (e.g., levels 14–51 vs. ≥53) in its publicly available coding [13].

Based on the existing literature, we examined several potential covariates, including age, gender, BMI, ethnicity, educational attainment, deprivation income index, smoking and drinking status, eGFR, hypertension, and hyperlipidemia [14]. All data were obtained through standardized questionnaires, physical examinations, and laboratory tests and were collected by qualified medical staff. Detailed measurement procedures for these variables are available at www.cdc.gov/nchs/nhanes/.

All analyses were conducted using R (version 4.2) and EmpowerStats (version 4.1). Appropriate NHANES sample weights were used in the statistical analyses. In the baseline characteristic table, continuous data were expressed as means ± standard deviation, whereas categorical variables were expressed as proportions (%). To compare the differences between the diabetic combined retinopathy group and the diabetic normal group, we used weighted t tests and weighted chi-square tests for continuous and categorical variables, respectively. For the correlation between NHHR and DR, we performed a multivariate logistic regression analysis and used three models to assess this relationship. The unadjusted model included only NHHR as the predictor: model I adjusted for age, gender, and race/ethnicity, and model II further adjusted for education level, smoking status, alcohol consumption, BMI, hyperlipidemia, and hypertension [15]. To explore the potential nonlinear relationship between NHHR and DR, we performed restricted cubic spline (RCS) analysis. The RCS model was constructed by dividing the NHHR values into intervals using knots, which were placed at the 5th, 35th, 65th, and 95th percentiles to ensure flexibility in capturing nonlinear patterns. The number of knots and their positions were determined based on the distribution of NHHR and prior methodological recommendations. The RCS model was fully adjusted for all covariates included in model II. The risk estimates (odds ratios [ORs]) and their 95% confidence intervals (CIs) were derived from the RCS regression analysis. The significance of the nonlinear relationship was assessed using a likelihood ratio test comparing the RCS model to a linear model. A statistically significant nonlinear relationship was defined as a p value <0.05 for the likelihood ratio test. Subgroup analyses were also performed, stratified by key demographic and clinical variables, to examine potential effect modification. The threshold for statistical significance for all analyses was a two-sided p value of less than 0.05.

A total of 4,935 participants were included in this study, of which 3,742 had no DR, while 1,193 were diagnosed with DR. The mean age of the overall population was 58.04 ± 14.76 years. Participants in the DR group had a significantly lower mean age (56.30 ± 15.81 years) compared to those without DR (60.79 ± 16.92 years) (p < 0.001). The gender distribution revealed that the proportion of females was higher in the DR group than in the non-DR group (59.27% vs. 54.94%, p = 0.027). Significant differences were also observed between the two groups in terms of BMI, race, education level, and smoking status (see Table 1).

The NHHR was found to be significantly positively associated with DR (see Table 2). In the adjusted model, each one-unit increase in NHHR was associated with a 19% higher risk of DR (OR = 1.19, 95% CI: 1.07–1.31, p < 0.05). When analyzed by quartiles, participants in the highest NHHR quartile had a 98% higher risk of DR compared to those in the lowest quartile (OR = 1.84, 95% CI: 1.62–2.06, p < 0.001).

Supplementary Table 1 (for all online suppl. material, see https://doi.org/10.1159/000545816) demonstrates that non-HDL-C remains independently associated with DR even after adjusting for individual lipid parameters (triglycerides, HDL-C, TC, or LDL-C). For each 10 mg/dL increment in non-HDL-C, the risk of DR increased by 4% to 8% across models (OR range: 1.04–1.08, all p values <0.05). In contrast, the added lipid parameters (e.g., HDL-C: OR = 0.98, p = 0.12; TG: OR = 1.02, p = 0.22) did not attain statistical significance, supporting that non-HDL-C is the dominant lipid component driving the association, which aligns with the robustness of NHHR as a composite NHHR.

To gain a deeper understanding of the association between NHHR and DR, a subgroup analysis was conducted to explore whether this relationship was influenced by various demographic characteristics, such as age, gender, BMI, race, smoking, and alcohol consumption status. The results indicated that the positive association between NHHR and DR remained consistent across most subgroups, with no significant interaction effects observed (p > 0.05). Notably, the association between NHHR and DR remained significant across different BMI levels and smoking statuses, suggesting that NHHR is a robust predictor of DR risk, regardless of whether patients are of normal weight, overweight, or obese. Similarly, smoking status did not significantly modify the effect of NHHR on DR risk. These findings indicate that NHHR is a broadly applicable predictor of DR risk across diverse populations (see Fig. 2).

To further explore the potential complex relationship between NHHR and DR, we employed a RCS model to assess nonlinearity. The results indicated a significant linear association between NHHR and DR risk (p for nonlinearity = 0.843). This linear relationship underscores the utility of NHHR not only in screening high-risk individuals but also in dynamically assessing disease progression risk in clinical practice (see Fig. 3).

This study systematically investigated the association between the NHHR and DR, demonstrating that an elevated NHHR significantly increases the risk of DR. After adjusting for multiple potential confounders, it was found that each unit increase in NHHR corresponded to a higher risk of DR. Notably, individuals in the highest NHHR quartile had nearly double the risk of DR compared to those in the lowest quartile. Subgroup analysis further revealed that the positive association between NHHR and DR was consistent across different age groups, genders, BMIs, and smoking statuses, with no significant interaction effects observed. This indicates that NHHR has strong predictive value for DR risk across a broad patient population. These findings highlight the clinical significance of NHHR as a potential predictor of DR, suggesting that excessive accumulation of non-HDL cholesterol and a decrease in HDL cholesterol may exacerbate microvascular damage, thereby increasing the risk of retinopathy as metabolic dysregulation worsens.

This study revealed a significant association between NHHR and DR, which can be explained by several pathophysiological mechanisms. First, an elevated NHHR reflects a relative increase in non-HDL cholesterol and a relative decrease in HDL cholesterol. Non-HDL cholesterol (including LDL-C, very low-density lipoprotein, and intermediate-density lipoprotein) is a major contributor to atherosclerosis. In diabetic patients, the accumulation of non-HDL cholesterol is closely linked to microvascular complications [16]. This accumulation promotes endothelial inflammation and oxidative stress, leading to endothelial dysfunction, which subsequently damages microvessels, particularly the retinal microcirculation [17, 18]. Additionally, oxidized low-density lipoprotein is thought to exacerbate local inflammation by activating the NF-κB signaling pathway, resulting in vascular leakage, microaneurysm formation, and neovascularization [19], all key pathological features of DR.

Furthermore, HDL cholesterol plays a protective role against atherosclerosis and microvascular damage through its anti-inflammatory, antioxidant, and reverse cholesterol transport functions [20]. HDL removes cholesterol from tissues and transports it back to the liver, reducing lipid accumulation in tissues [21]. However, decreased HDL levels weaken this protective mechanism, making retinal vessels more susceptible to inflammation and oxidative stress [22]. The observed linear relationship between NHHR and DR suggests that as NHHR increases, DR risk also increases proportionally, without any significant plateau effect at higher NHHR levels. Thus, NHHR, as an indicator of the balance between harmful and protective lipids, can not only help predict the occurrence of DR but also reflect the risk of disease progression, highlighting its important clinical utility.

The finding of a significant positive correlation between NHHR and DR in this study is consistent with the results of several previous studies. First, our study demonstrates that elevated NHHR significantly increases the risk of DR, a finding aligned with previous research on NHHR and diabetes-related complications. For example, Sheng et al. [23] found that an increase in NHHR was significantly associated with the incidence of cardiovascular events in diabetic patients, suggesting that NHHR, as a composite lipid marker, plays a crucial role in predicting the risk of various diabetes-related complications. Additionally, NHHR, by reflecting the relative increase in non-HDL cholesterol, underscores the close relationship between lipid dysregulation and microvascular complications, which is consistent with the study of Li et al. [24] on lipid abnormalities and microvascular diseases.

Moreover, the subgroup analysis in this study showed that the positive association between NHHR and DR remained consistent across different BMI, age, and gender groups. This further supports the potential value of NHHR as a universal predictor. Similarly, previous studies on NHHR and other metabolic diseases have reached similar conclusions. Yu et al. [25] reported that elevated NHHR was significantly associated with cardiovascular disease and all-cause mortality, with consistent results across various subgroups. These studies collectively suggest that NHHR may serve not only as a predictor of DR but also as a risk assessment tool broadly applicable to multiple metabolic-related diseases.

However, despite the strong predictive capability of NHHR demonstrated in several studies, our research also revealed a nonlinear association between NHHR and DR, particularly with the risk increase slowing down after NHHR reaches a certain level. This finding aligns with the study of Han et al. [22], which noted that in certain high-risk patients, despite elevated NHHR, its correlation with complication risk tends to plateau. This may reflect a saturation effect of lipid burden on microvascular damage, where the impact on disease progression diminishes after reaching a certain lipid threshold. The discovery of this nonlinear relationship suggests that NHHR can be used not only to screen high-risk individuals but also to dynamically assess disease progression, offering a basis for personalized treatment in future clinical practice. Nevertheless, some studies have raised differing views on HDL’s role in diabetes complications. For instance, extremely high levels of HDL have been associated with poor outcomes in some studies, suggesting a dual role for HDL [26]. These discrepancies may be due to characteristics of the study populations and other confounding factors. Therefore, future research should further explore the role of HDL and its metabolism in different populations, and consider a broader range of metabolic biomarkers for comprehensive analysis.

Our study indicates a significant association between the NHHR and DR. Lipid metabolism abnormalities have been confirmed as a crucial factor in the progression of DR, and fibrates, as a class of commonly used lipid-lowering drugs, have gained increasing attention in recent years. Studies have shown that fibrates can effectively improve lipid profiles and may slow the progression of DR by lowering triglycerides and increasing HDL-C [27]. For example, Mozetic et al. [28] found that fibrates significantly reduced the need for laser treatment in patients with DR, particularly for diabetic macular edema and proliferative DR. Additionally, Stewart and Lois [27] pointed out that fibrates may improve retinal vascular permeability and reduce damage and apoptosis in retinal endothelial cells, further affecting the pathological process of DR. However, as our current study does not include specific medication usage data, we are unable to directly assess the role of fibrates in the lipid-DR association. Future research, particularly clinical studies with longitudinal follow-up and detailed medication information, will help better evaluate the potential of fibrates in DR.

Despite the significant association between NHHR and DR found in this study, there are several limitations. First, as a cross-sectional study, it cannot establish a causal relationship between NHHR and DR. While a significant correlation was found, we cannot conclude whether elevated NHHR directly causes retinopathy. Longitudinal studies are needed to confirm this causal relationship and explore how dynamic changes in NHHR affect long-term outcomes in diabetes patients. Second, the sample for this study is from the US NHANES database, which may not fully represent populations from other regions. Differences in race, cultural backgrounds, and access to healthcare may affect the generalizability of the results, and future studies should validate these findings in other countries. Additionally, NHANES defines DR in a binary manner (present/absent based on ETDRS ≥14) and does not further stratify severity (e.g., mild, moderate, proliferative DR). Future research should consider more detailed DR grading (e.g., ETDRS levels 14–53 for non-proliferative vs. ≥53 for proliferative) to assess the relationship with NHHR across different severity stages. Another limitation is that we did not distinguish between type 1 and type 2 diabetes. Given their different underlying mechanisms and treatment strategies, the relationship between NHHR and DR may differ between the two types. Future research should explore this relationship separately in both populations to better understand the applicability of NHHR as a predictive biomarker across both types. Finally, although we adjusted for several confounders, unmeasured confounders, such as dietary habits, physical activity, and medication use, may still influence the relationship between NHHR and DR. Future studies should consider additional confounders, particularly lifestyle and treatment interventions.

Our study confirms that higher NHHR is associated with an increased risk of DR among US adults. However, due to the limitations of this study, further research is needed to validate and confirm these findings.

We would like to thank all participants in this study.

The Institutional Review Board of the First Hospital Affiliated to Heilongjiang University of Chinese Medicine, China, determined that the study did not need approval because it used publicly available data.

The authors declare that there are no competing interests.

No funding was received for this study.

Zhirui Zhang and Changxing Liu: writing papers and statistical analysis of results; Lingying Zhao, Xufang Tan, Ximing Yu, and Jiadi Wang: collecting and processing data; Jing Yao: supervision and reviewing writing.

Zhirui Zhang and Changxing Liu contributed equally and are regarded as co-first authors.

The publicly available datasets presented in this study can be found in online repositories. These data can be found here: https://www.cdc.gov/nchs/nhanes/.