High remnant cholesterol level is relevant to diabetic retinopathy in type 2 diabetes mellitus

Diabetic retinopathy (DR) is the primary oculopathy causing blindness in diabetic patients. Currently, there is increasing interest in the role of lipids in the development of diabetic retinopathy, but it remains controversial. Remnant cholesterol (RC) is an inexpensive and easily measurable lipid parameter; however, the relationship between RC and DR in type 2 diabetes mellitus (T2DM) has not been elucidated. This research investigates the relevance between RC levels and DR severity while building a risk prediction model about DR. In this single-centre retrospective cross-sectional study. Each hospitalised T2DM patient had no oral lipid-lowering drugs in the past three months, and coronary angiography showed epicardial coronary artery stenosis of less than 50% and completed seven-field stereo photographs, fluorescein fundus angiography, and optical coherence tomography detection. The RC value is calculated according to the internationally recognised formula. Binary logistic regression was used to correct confounding factors, and the receiver operating characteristic (ROC) analysis was used to identify risk factors and assess the nomogram’s diagnostic efficiency. A total of 456 T2DM patients were included in the study. The RC levels in the DR team was higher [0.74 (0.60–1.12) mmo/l vs 0.54 (0.31–0.83) mmol/l P < 0.001] in the non-DR team. After adjusting for confounding elements, RC levels are still associated with DR risk (OR = 5.623 95%CI: 2.996–10.556 P < 0.001). The ratio of DR in every stage (except mild non-proliferative diabetic retinopathy) and DME in the high RC level team were further increased compared to the low-level team (all P < 0.001). After ROC analysis, the overall risk of DR was predicted by a nomogram constructed for RC, diabetes duration, and the neutrophil-lymphocyte ratio as 0.758 (95%CI 0.714–0.802 P < 0.001). High RC levels may be a potential risk factor for diabetic retinopathy, and the nomogram does better predict DR. Despite these essential findings, the limitation of this study is that it is single-centred and small sample size analysis.


Introduction
As the worldwide prevalence of diabetes mellitus increases, 463 million adults were diagnosed with diabetes in 2019, and according to this trend, there will be 700 million diabetes worldwide by 2045 [1]. Diabetic retinopathy (DR) is a prevalent neurovascular complication of diabetes mellitus and is the leading cause of blindness in the working-age generation [2]. Although glycemic levels, diabetes duration, microalbuminuria, and blood pressure levels are known risk elements for the development and advance of DR, current studies report that there may be other, as yet unknown, risk factors for the disease [3]. The association of dyslipidemia with DR has been thoroughly investigated. However, in many extensive studies, controversy still exists regarding the exact role of dyslipidemia in DR [4,5]. Perhaps, traditional lipid levels alone are inadequate, although a recent study shows that novel serum biomarkers of lipid profiles and inflammatory reactions are potentially related to DR [6].
Remnant cholesterol (RC) is the cholesterol content of triglyceride-rich lipoproteins, which is IDL and VLDL in the fasting state and the non-fasting state chylomicron remnants [7]. Current studies suggest that RC is derived from two pathways, the endogenous pathway is VLDL produced by the liver, and exogenously, chylomicron is made within the intestine. Triglycerides and chylomicrons in VLDL enter the bloodstream and are hydrolysed by lipoprotein lipase (LPL) at high speed, continuously producing smaller, denser residual lipoprotein particles. RC level assessment is inexpensive and convenient, perhaps providing valuable data for clinical management, and can be easily calculated using an established formula [8]. Recent studies have confirmed that RC positively affects a strong association with cardiovascular endpoints [9,10]. Whether remnant cholesterol also contributes to the development of microvascular disease in individuals with diabetes is an interesting question. Up to now, the relationship between RC and DR in T2DM patients has not been researched. The present research explored the connection between RC levels and DR in Chinese T2DM patients.

Study design and participants
This retrospective observational cross-sectional study was undertaken in the First Affiliated Hospital of Harbin Medical University. According to the Declaration of Helsinki, the research was ratified by the Ethics Committee of the FAH in HMU. Consecutively selected hospitalised T2DM patients in the endocrinology and cardiology departments in the FAH for HMU between November 2018 and October 2021 were enrolled according to the following criteria. Inclusion criteria are: (1) the diagnosis of T2DM was per 1999 WHO diagnostic criteria [11]; (2) no gender restrictions on participants over 18 years of age; (3) coronary angiography showed epicardial coronary artery stenosis of less than 50%; (4) completed seven-field stereo photographs, fluorescein fundus angiography (FFA), and Optical Coherence Tomography (OCT) detection, and (5) blood lipid and other biochemical indicators were collected on the second day of admission, and the content was complete. Exclusion criteria are: (1) had taken fenofibrate or other oral lipidlowering drugs during the last three months; (2) definite cerebral infarction or cerebral revascularisation; (3) peripheral arterial occlusive disease; (4) any uncontrolled systemic disease other than sub-optimally controlled diabetes mellitus; (5) other ophthalmic diseases affect DR diagnosis, such as severe cataracts, and (6) any confirmed malignancies.

Assessment of DR and DME
Detection of the fundus was achieved by performing seven-field stereo photographs and FFA. The International Clinical Diabetic Retinopathy Disease Severity Scale was used [16] to grade DR. DR was divided into five stages of non-DR, non-proliferative diabetic retinopathy (mild, moderate, severe), and proliferative diabetic retinopathy (PDR). OCT was used to measure the thickness of macular fovea in eyes with macular oedema and determine the severity of the diabetic macular oedema using the International Clinical Diabetic Macular Oedema Disease Severity Scale.

Statistical analysis
Groups were divided according to the presence or absence of diabetic retinopathy. The statistical software SPSS 26.0 and "R language" statistical packages were used for the statistical analyses. The data's normalcy was assessed using the Shapiro-Wilk test. The continuous normality variables were introduced as the mean ± SD and evaluated using the Student's t-test. The skewed continuous variables were shown as the median (interquartile range, IQR) and were assessed using the Mann-Whitney U test. The Chi-square test was used to compare categorical variables. Spearman correlation and partial correlation analysis were used to analyse the connection between RC and DR stages. Spearman correlation analysis also examined the relationship between RC and other variables. The P-value of P < 0.05 was defined as statistically significant.
The univariate binary logistic regression analyses were used to analyse the presence of DR as the dependent variable and risk elements of DR as independent variables. Those variables which were statistically significant in the univariate analysis but did not mutually interfere were included in the multiple binary regression analysis models. Outcomes of these regression analyses were assessed using ORs, 95%CI, and p-values. Hosmer-Lemeshow goodness of fit implementation was used in the optimal model in multivariate regression analysis to select variables to develop a nomogram of the prediction of DR occurrence. The receiver operating characteristic (ROC) analysis was used to identify risk factors and assess the nomogram's diagnostic efficiency.
The RC level was divided into three groups, according to tertile. The composition ratio of DR in each group, composition ratio at each DR stage, and DME composition ratio were analysed. Chi-square implementation was used to compare teams; P < 0.05 was deemed statistically significant.

Correlation between RC and other variables in the lipid profiles with stages of DR
The correlation between RC and other variables in the lipid profiles (including TC, TG, LDL-C, HDL-C, ApoA, and ApoB) with the DR stages was assessed using Spearman correlation analysis. Table 2 shows that only RC and TC were correlated with the DR stages (r = 0.360 P < 0.001) and (r = 0.142 P = 0.002), respectively. Even after controlling TC by partial correlation analysis, RC (r = 0.298 P < 0.001) was still correlated with DR stages. In the univariate analysis, the risk factors of DR, age, diabetes duration, DNP, oral antihypertensive drugs, systolic blood pressure, TC, RC, mAlb, TyG, and NLR were statistically significant. After adjusting confounding factors using binary logistic regression analysis, RC (OR = 5.623 95%CI: 2.996-10.556 P < 0.001) was still a risk factor for DR, as shown in Tables 3 and S1. Moreover, the variables of NLR(OR = 1.742 95% CI: 1.218-2.493 P = 0.002)and diabetes duration(OR = 1.250 95% CI: 1.064-1.496 P = 0.007)were also statistically significant in the multivariate analysis.
The relationship between RC level and DR, every stage of DR and DME The RC levels were divided into three teams as per tertile; Tertile1 team (< 0.52 mmol/, N = 150), Tertile 2 team (0.52-0.83mmo/l, N = 152) and Tertile3 group (> 0.83 mmol/l, N = 154). In contrast with the Tertile 1 team, the ratio of DR, moderate-NPDR, severe-NPDR, PDR, and DME in the Tertile3 group increased significantly (all P < 0.01), but the ratio of mild-NPDR yielded no statistically significant difference (Fig. 2).

Discussion
The current research findings offer novel evidence that the circulating level of RC is positively correlated with the occurrence and severity of DR (Table 3, Fig. 3). RC may be another risk factor for DR, in addition, to complete cholesterol, low-density lipoprotein, triglycerides, and traditional blood lipid profiles. However, the contribution of high-level RC in mild DR was minor, and the damage of RC to microvascular environments in diabetic patients may be a chronic and long-term process [17]. It is well known that most T2DM patients have relatively high BMIs, insulin resistance, and insulin secretion disorders [18], increasing the number of cholesterol-rich VLDL residual particles [19]. TyGindex is a proxy for insulin resistance and is highly correlated with the gold standard hyperinsulinemiaorthoglycemic clamp [20], which can be a good assessment of insulin resistance in patients. It has been reported that RC was correlated with insulin resistance [21]. In the current study, TyG-index is associated with RC (r = 0.488 P < 0.001). Insulin resistance leads to elevated RC levels, which may be involved in the occurrence of DR [22]. Table 2 The correlation between stages of DR and the following lipid profiles  It is widely believed that NLR can be used as an inflammatory marker for some non-infectious diseases [23,24]. In this study, there is a significant difference between the DR group and the non-DR team [1.6 (1.3-1.93) vs1.69 (1.41-2.76) P < 0.01], which was consistent with the results of Wang et al. [25]. In addition, a low-grade inflammatory response is one of the recognised pathogenesis of DR. It is associated with oxidative stress, high expression of inflammatory factors and the formation of advanced glycation end products (AGEs). Notably, RC can also trigger an inflammatory response, causing damage to the vascular endothelium [8]. Animal experiments have confirmed that RC particles easily enter the intima of the artery due to their particles size and composition characteristics, do not readily diffuse back into the blood, and can be directly ingested by macrophages [26]. Macrophages promote lipid decomposition, and triglyceride-rich lipoproteins (TGRLs) generate core components, especially high levels of fatty acids and saturated fatty acids, which combine with toll-like receptors 4 and 2, induce inflammasome activation, promote interleukin-1β secretion, and amplify inflammatory effects [27]. Varbo's clinical study on the relationship between RC and LDL and ischemic heart disease and lowgrade systemic inflammation reported that only elevated RC levels had a causal relationship with low-grade systemic inflammation [28]. In the current study, RC and NLR were higher in the DR group, but there was no apparent correlation. The relationship between the two needs to be further validated by large sample size studies.
Additionally, there was a dynamic relationship between RC and UA (r = 0.15 P < 0.001). Some studies have reported that a high standard of UA is closely associated with oxidative stress and inflammation [29]. Other studies suggested that UA can protect vascular endothelium, which may be caused by the antioxidant effect of UA concentrations in a particular range [30]. However, inflammatory responses can be stimulated [31]. Notably, a sizeable cross-sectional study of a rural Chinese populace reported that high uric acid levels might be associated with various blood lipids [32]. Do increased UA levels contribute to the RC reduction? Investigating the relationship between the two is worthy of further research.
RC may be a risk factor for both macrovascular and microvascular diseases in diabetes patients. A large cohort study found that residual cholesterol, rather than LDL-C, may be the primary cause of arteriosclerotic cardiovascular disease. The relative risk of major adverse cardiovascular events increased by 21% for every 10 mg/ dL increase in residual cholesterol levels (about 0.26 mmol/L) [10], and a second cross-sectional study in China showed that RC was related to the risk of  [33]. Furthermore, a recent cohort study suggests that RC is associated with visionthreatening DR in type 1 diabetes [34]. The present study selected type 2 diabetic patients without coronary heart disease to avoid patients affected by the relationship between RC and diabetic retinopathy. Finally, in this study, microalbuminuria was related to RC (r = 0.179 P < 0.001), but the relationship between RC and DR was not disturbed after adjusting for the 24-h microalbumin variable using binary logistic regression (Tables 4 and S2).
The two primary visual-threatening complications of diabetic retinopathy are the proliferative phase of retinal neovascularisation and macular oedema, leading to the destruction of the central blood-retinal obstacle. In this study, the incidence of PDR and DME increased significantly in the high RC level group (Fig. 2). Related metaanalysis and cross-sectional studies suggest that the occurrence of DME is closely associated with high triglyceride levels [35], and the current study found that RC correlates with TG (r = 0.447 P < 0.001). RC may be a risk factor masked by TG for DME and PDR. There is evidence of considerable residual risk of cardiovascular events even after LDL-C reduction or HDL-C increase to recommended concentrations with statins and other lipid-lowering agents due, in part, to elevated RC levels [36]. Similarly, the elevation of RC levels is important for preventing and delaying DME and PDR.
The nomogram can predict the risk of disease occurrence and has attracted more attention due to its simple, intuitive, and reference value. Therefore, RC, NLR, and diabetes duration were selected and contributed to the most stable model in multivariate logistic regression analysis and resulted in the predicted risk of DR nomogram. Diagnostic efficiency was 0.758 (95%CI 0.714-0.802 P < 0.001). For communities or hospitals without facilities for fundus examination, physicians can use this predictive model to screen people at higher risk of DR.
Although the relationship between HbA1c and DR is widely recognised [37,38], the current study found no significant difference in HbA1c between the DR group and the non-DR participants [8.5 (7.6-9.8) vs 8.2 (7.1-9.5) P = 0.076]. Those participants that were hospital inpatients often had poorly controlled glycemic levels, which may have caused this discrepancy. Interestingly, a similar phenomenon was also reported in other studies [39,40]. It should be acknowledged that selection bias is challenging to avoid when hospital inpatients are participants in research studies.
The effects of RC on patients with type 2 diabetes may involve the imbalance of oxidative and antioxidant Fig. 2 The relationship between RC levels and DR, stages of DR, and DME. Abbreviations: DR: diabetic retinopathy; NPDR: non-proliferative diabetic retinopathy; PDR: proliferative diabetic retinopathy; DME: diabetic macular edema. Notes: The RC level was divided into three groups by tertile, Tertile1 group (< 0.52 mmol/l, N = 150), Tertile2 group (0.52-0.83mmo/l, N = 152) and Tertile3 group (> 0.83 mmol/l, N = 154). Chi-square test was used for comparison between groups, *P < 0.05, **P < 0. 01 capacity [27], RC is involved in the damage of the vascular endothelium [8] and the activation of the inflammatory response. Also, insulin resistance may be involved in the increase of RC [21]. These effects have been confirmed in previous studies on macrovascular complications in T2DM patients. However, the correlation between RC and DR was confirmed in this study based on the hospitalisation of type 2 diabetes mellitus for the first time.

Study advantages and limitations
The advantage is that patients with type 2 diabetes with cardiovascular disease were excluded in this study in order to further explore the correlation between the occurrence and severity of RC and DR. However, also some limitations in this research. For example, the study was undertaken in a single-centre, with the observational design based on a relatively cross-sectional analysis with an insufficient sample size. Notably, the prevalence of dyslipidemia and other risk elements in the Asian population was similar to other extensive contemporary tests and real-world registries involving ethnic groups [41,42], which may support the generality of the study outcomes. Secondly, RC is correlated with DR, but the causal relationship between RC and DR cannot be proven. Thirdly, to exclude the influence of lipidlowering drugs on lipid profile metabolism, subjects in this study had not taken any lipid-lowering drugs orally in recent three months. This study cannot differentiate whether lipid-lowering oral medicines impact RC levels. Besides, the DR risk nomogram only has internal validation, and its diagnostic efficacy may be reduced without an external warranty.

Conclusions
In patients with T2DM, high RC levels positively correlate with the occurrence and severity of DR. These striking findings suggested that DR may not only be affected by high blood glucose levels but also influenced by a complex lipid profile. Clinicians and patients could easily predict the occurrence of DR through a nomogram considering RC, NLR, and diabetes duration. Insulin resistance interventions may help reduce RC level and control DR risk, which must be investigated in future multicenter or community-based prospective cohort studies.
Additional file 1: Table S1 The association between RC value and DR. Table S2 Effects of other variables on the relationship between RC and DR.
Additional file 2: Fig. S1 ROC Curve for RC, NLR, DM Duration and Risk of DR Nomogram.

Acknowledgments
I am indebted to my tutor and team members for their efforts and support.

Authors' contributions
Conception and design analysis and data interpretation: XYG, YYS. The manuscript was drafted by: YYS, XYG, QW. Statistical analysis: YTZ, XWT. Data collation: SDP, XYX. Critical revision of the manuscript for important intellectual content: all authors. The authors read and approved the final manuscript.

Funding
The research was supported through grants from the Fund of Scientific Research Innovation of the FAH in HMU (grant number 2020 M27, China).

Availability of data and materials
The authors declare the data related to this study can be available from the corresponding author upon reasonable request.