Prognostic Impact of Remnant-like Particle Cholesterol in Patients with Differing Glucose Metabolic Status: an Observational Cohort Study from China

Background: It is uncertain whether remnant-like particle cholesterol (RLP-C) could predict residual risk in patients under different glucose metabolic status. This study aimed to evaluate the relationship between RLP-C and adverse prognosis in patients with non-ST-segment elevation acute coronary syndrome (NSTE-ACS) undergoing percutaneous coronary intervention (PCI) and identify the potential impact of glucose metabolism on the predictive value of RLP-C. Methods: The study enrolled 2419 patients with NSTE-ACS who underwent PCI at Beijing Anzhen Hospital from January to December 2015. RLP-C was calculated as follows: total cholesterol (TC) minus low-density lipoprotein cholesterol (LDL-C) and high-density lipoprotein cholesterol (HDL-C). The primary endpoint was a composite of events as follows: all-cause death, non-fatal myocardial infarction (MI), and ischemia-driven revascularization. Results: RLP-C was signicantly associated with adverse prognosis in the total population [hazard ratio (HR) 1.291 per 1-SD increase of RLP-C, 95% condence interval (CI) 1.119-1.490, P < 0.001], independent of confounding risk factors. However, subgroup analysis showed that increasing RLP-C was shown to be associated with a higher risk of adverse events in the diabetic population only [HR 1.385 per 1-SD increase of RLP-C, 95% CI 1.183-1.620, P < 0.001]. RLP-C failed to be a signicant determinant of adverse prognosis in the pre-diabetic and non-diabetic population. The addition of RLP-C to the baseline model signicantly enhanced the predictive value for adverse events both in the total and diabetic populations. Conclusions: Higher RLP-C level is a signicant and independent predictor of adverse prognosis in diabetic patients with NSTE-ACS who underwent PCI.


Background
As the most serious manifestation of atherosclerotic cardiovascular disease (ASCVD), acute coronary syndrome (ACS) leads to a consistently higher risk of recurrence of cardiovascular events despite the use of evidence-based secondary prevention therapies [1,2]. Low-density lipoprotein cholesterol (LDL-C) has been extensively recognized as one of the important risk factors for ASCVD and reduction of serum LDL-C levels with statins is an effective therapy to reduce cardiovascular risks [3]. Despite regulating LDL-C with statins, residual risk for recurrence of cardiovascular events remains in patients with ACS [4][5][6][7], which indicates that there are factors other than LDL-C that determine risk. Identi cation of the residual risk factors is important in tailoring risk reduction strategies that match individual risk levels and developing new therapeutic targets.
Studies have reported that the residual risk can be partly ascribed to an increased level of remnant lipoproteins [2,4,8,9]. Remnant lipoproteins are lipoproteins that are rich in triglycerides (TGs), components of which include very-low-density lipoprotein (VLDL), intermediate-density lipoprotein (IDL), and chylomicron [10]. The cholesterol content of remnant lipoproteins is de ned as remnant-like particle cholesterol (RLP-C). Nowadays, the pattern of targeting LDL-C alone has changed, with recent guidelines highlighting the important role of non-high-density lipoprotein cholesterol (non-HDL-C), which includes RLP-C, on the pathogenesis of atherosclerosis; and thus its availability as an additional therapeutic target [11]. As a component of non-HDL-C, it is of great signi cance to further clarify the role of RLP-C in the development of coronary atherosclerosis.
The relationship between RLP-C and adverse prognosis in the speci c cohort with non-ST-segment elevation acute coronary syndrome (NSTE-ACS) undergoing percutaneous coronary intervention (PCI) were not fully investigated. Results from previous studies have revealed that the impact of RLP-C seems to be more prominent in high-risk patient groups such as patients with metabolic syndrome or type 2 diabetes [12][13][14][15][16]. It is worth exploring whether the predictive value of RLP-C for adverse outcomes varies among populations with different glucose metabolic states. Therefore, the present study was designed to investigate the prognostic impact of RLP-C in patients with NSTE-ACS undergoing PCI and the potential impact of glucose metabolic status on the predictive value of RLP-C.

Study population
This study retrospectively screened patients with NSTE-ACS who received PCI treatment in Beijing Anzhen Hospital (Beijing, China) from January to December 2015. NSTE-ACS consisted of non-ST segment elevation myocardial infarction (NSTEMI) and unstable angina pectoris (UA), whose de nitions were determined by the corresponding guidelines [17]. The exclusion criteria were: (1) missing clinical, laboratory, and/or angiographic data; (2) history of cardiogenic shock, chronic in ammatory disease, or neoplasm; (3) evidence of active infection; (4) chronic renal insu ciency with estimated glomerular ltration rate (eGFR) < 30 mL/(min * 1.73 m 2 ) and severe hepatic disease; (5) other serious diseases; and (6) PCI failure, PCI-related complications, and in-hospital death. Ultimately, 2419 participants who met the inclusion criteria were enrolled.

Data collection and de nitions
The enrolled patients' demographic and clinical characteristics, laboratory investigations, and coronary procedural results were retrieved and collected from the medical record system of Beijing Anzhen Hospital. Body mass index (BMI) was calculated as follows: weight (kilogram)/[height (meter)] 2 .
Participants previously diagnosed with diabetes (treated with diet, insulin, or oral agents) or whose glycosylated hemoglobin A1c (HbA1c) level ≥6.5% were considered to have diabetes. Non-diabetes was de ned as a HbA1c level <5.7% and pre-diabetes was de ned as an HbA1c level between 5.7% and 6.4% [18].
Venous blood samples were taken after overnight fasting before baseline PCI. Laboratory parameters including lipid pro les [TGs, total cholesterol (TC), LDL-C, high-density lipoprotein cholesterol (HDL-C)], high-sensitivity C-reactive protein (hs-CRP), creatinine, uric acid, fasting blood glucose (FBG), HbA1c, and other biomarkers, were measured by standard methods in the central laboratory of the hospital. Concentrations of TC, HDL-C, and TGs were quanti ed by standard enzymatic techniques. LDL-C was determined by the homogeneous direct method. RLP-C levels were determined by subtracting LDL-C and HDL-C from TC, which was recommended by relevant dyslipidemia guidelines [19,20]. The eGFR was calculated as follows: eGFR[mL/(min*1.73m 2 )]=186*serum creatinine(mg/dL) -1.154 *age -0.203 (*0.742 if female) [21]. Left ventricular ejection fraction (LVEF) was evaluated by two-dimensional modi ed Simpson's method using an ultrasonic cardiogram (Philips Company, Eindhoven, The Netherlands).
Coronary angiographic data were analyzed and evaluated by visual measurements, and the results were documented and veri ed by at least two experienced cardiologists. A multi-vessel lesion was de ned as more than two main branches with stenosis ≥50%. A chronic total occlusion lesion was de ned as a total occlusion [thrombolysis in myocardial infarction (TIMI) ow grade 0] and an occlusion time ≥3 months. A diffuse lesion was de ned as a single stenotic lesion with a length ≥20 mm. A bifurcation lesion was de ned as the lesion involving the origin of an important side branch. Coronary procedures were carried out based on the current practice guidelines of China [22], and procedure strategies were selected by experienced interventional cardiologists.

Follow-up and endpoint events
After baseline PCI had been performed, all patients received routine follow-up at 3, 6, and 12 months and then annually until 36 months. The prognostic information was obtained by means of telephone interviews with the patient and/or their family members. Corresponding medical records were referred to verify the authenticity in case that ambiguous information was obtained. Adverse events including allcause death, non-fatal myocardial infarction (MI), and ischemia-driven revascularization were documented to perform the present analysis. The primary endpoint was the composite of adverse events mentioned above, and the secondary endpoint was each component of the primary endpoint. If participants experienced multiple adverse events during the 36-month follow-up, the most severe one (allcause death>non-fatal MI>ischemia-driven revascularization) was applied to the present analysis. And for those who suffered from the same event multiple times, only the rst instance of the event was selected.

Statistical analysis
Continuous variables were presented as mean±standard deviation (SD) or the median (25th and 75th percentiles: P25, P75), and differences between groups were examined by the independent-sample t-test or Mann-Whitney U test as appropriate. Nominal variables were expressed as counts (percentages) and compared by the Chi-square test (χ2 test) or Fisher's exact test when appropriate. The correlations between RLP-C and other variables were assessed by Pearson correlation test or Spearman's rank correlation test as appropriate. The Kaplan-Meier survival analysis was performed to evaluate the incidence of adverse events in groups strati ed by the median of the RLP-C level. Then, survival curves were plotted accordingly, and differences between groups were examined by the log-rank test. The univariate Cox proportional hazards analyses were primarily conducted to con rm the signi cant predictors of adverse events. The variables with statistical signi cance (P<0.05) in univariate analysis were analyzed with multivariate analysis to investigate the independent determinants of adverse events. The results of Cox proportional hazards analysis were expressed in terms of hazard ratio (HR) and 95% con dence intervals (CI). The HR was examined by 1-SD change in continuous variables except for age, heart rate, systolic blood pressure (SBP), and number of stents.
C-statistics that consisted of receiver-operating characteristic (ROC) curve analysis was applied to assess the additional discriminative performance of RLP-C on the basis of the baseline model that included traditional risk factors. Differences between the area under the ROC curve (AUC) of various models were compared by DeLong's test. Moreover, the incremental reclassi cation and discrimination ability of RLP-C on the basis of the baseline model for predicting adverse events was further determined by category-free net reclassi cation improvement (NRI) and integrated discrimination improvement (IDI).
The population was divided into three subgroups according to glycometabolic status: diabetic, prediabetic, and non-diabetic groups. Similar statistical analyses were performed for each subgroup. Statistical analyses were conducted by SPSS 23.0 (SPSS Inc., Chicago, Illinois, USA), MedCalc version 19.1 (MedCalc Software, Belgium), and The R Programming Language (version 3.5.1). A two-tailed P value of 0.05 was applied to assess statistical signi cance.

Baseline characteristics
The baseline characteristics of each groups were shown in Table 1. The RLP-C levels in participants with an adverse event were signi cantly higher than that in patients without (0.90±0.61 vs. 0.65±0.35, P<0.001). Patients with an adverse event exhibited higher age, BMI, heart rate, SBP, and higher prevalence of previous history of MI, PCI, stroke, and peripheral arterial disease (PAD). In terms of laboratory indicators, participants that developed adverse events had higher levels of TGs, TC, hs-CRP, creatinine, FBG, and HbA1c, but lower levels of HDL-C, eGFR, and LVEF. As for the angiographic ndings, those with an adverse event showed higher proportions of left main artery disease, multi-vessel disease, and other characteristics of complex coronary artery lesion. RLP-C levels were signi cantly higher in patients with diabetes than pre-diabetes (0.74±0.51 vs 0.68±0.36, P=0.003) and non-diabetes (0.74±0.51 vs 0.66±0.37, P<0.001). However, there was no signi cant disparity in RLP-C levels between pre-diabetic and non-diabetic populations (0.68±0.36 vs 0.66±0.37, P=0.339) ( Figure 1). RLP-C levels were positively correlated with TGs (r=0.853, P<0.001), TC (r=0.455, P<0.001), and LDL-C (r=0.112, P < 0.001), while they were negatively correlated with HDL-C (r=-0.173, P<0.001).

Predictive value of RLP-C in total population
The study population was strati ed into two groups according to the median of the RLP-C level. Kaplan-Meier curves for the incidence of the composite and each component of endpoint events according to the median of RLP-C were shown in Figure 2. Compared with patients with a lower median of RLP-C, those with a higher median of RLP-C presented with a signi cantly higher incidence of composite endpoint events ( Figure 2A, Log-rank P<0.001). The difference was mainly driven by the increased incidence of non-fatal MI ( Figure 2C, Log-rank P=0.002) and ischemia-driven revascularization ( Figure 2D, Log-rank P<0.001). Kaplan-Meier curves for all-cause death between the lower and higher RLP-C group failed to reach statistical signi cance ( Figure 2B, Log-rank P=0.260).
Multivariate Cox proportional hazard analysis that was adjusted for variables that were statistically signi cant (P<0.05, details shown in Table S1) were performed to assess the predictive value of RLP-C for the composite and each component of the endpoint events. After adjustment of the confounding factors, increased RLP-C levels were consistently observed to be an independent risk indicator of composite endpoint events, non-fatal MI, and ischemia-driven revascularization, despite regarding RLP-C as a nominal or continuous variable ( Table 2).
The addition of RLP-C signi cantly enhanced the AUC obtained from the baseline model adjusted for traditional risk factors including age, sex (female), smoking, hypertension, prior MI, prior PCI, eGFR, HbA1c, TC, HDL-C, LVEF, left main disease, and multi-vessel disease (AUC: baseline model, 0.798 vs. baseline model+RLP-C, 0.811, P for comparison <0.001) ( Table 3). Moreover, adding RLP-C to the baseline model signi cantly promoted the discriminative performance for prediction of adverse events with a category-free NRI of 0.084 and an IDI of 0.017 (both P<0.05) ( Table 3).

Predictive value of RLP-C in subgroups with various glycometabolic status
The predictive value of RLP-C was further evaluated in subgroups with various glycometabolic status [non-diabetic population (n=926), pre-diabetic population (n=645), diabetic population (n=848)]. Kaplan-Meier curves for the incidence of the composite and each component of the endpoint events according to the median of RLP-C in various subgroups were summarized in Figure 3. In patients with diabetes, the incidence of composite endpoint events, non-fatal MI, and ischemia-driven revascularization in the higher RLP-C group was signi cantly higher than that in the lower RLP-C group [ Figure 3 In multivariate Cox proportional hazard analysis, increasing RLP-C levels were shown to be signi cantly correlated to a higher risk of adverse events in the diabetic population. However, RLP-C failed to be a signi cant determinant of adverse prognosis in the pre-diabetic and non-diabetic populations ( Table 4).
The increased AUC resulting from the addition of RLP-C to the baseline model (AUC: baseline model, 0.788 vs. baseline model+RLP-C, 0.836, P for comparison <0.001) was signi cant in the diabetic population. By contrast, the addition of RLP-C did not show an incremental effect on AUC in the prediabetic and non-diabetic populations (Table 5, Figure 4). Furthermore, adding RLP-C to the baseline model prominently promoted the reclassi cation and discrimination ability for predicting adverse events in the diabetic population with a category-free NRI of 0.155 and an IDI of 0.040 (both P<0.05), but the additional effect was not found in the pre-diabetic and non-diabetic populations ( Table 5).

Discussion
The present study demonstrated a strong and independent relationship between fasting RLP-C levels and adverse prognosis in patients with NSTE-ACS treated with PCI. Further subgroup analyses elucidated that RLP-C showed a better predictive value in the diabetic population. However, RLP-C failed to be a signi cant determinant of adverse prognosis in the pre-diabetic and non-diabetic populations. The addition of the RLP-C level had a signi cant incremental effect on the predictive value for adverse events.
It has been widely demonstrated that LDL-C is one of the most signi cant risk indicators for ASCVD, and reduction of serum LDL-C levels with statins is a well-established therapy to reduce the ASCVD risk. However, many patients whose LDL-C levels are well controlled by statins continue to suffer recurrent cardiovascular events [3][4][5][6][7]. In recent years, factors related to obesity and metabolic syndrome, such as triglycerides rich lipoproteins (TRLs), have been considered as potential metabolism-related risk factors for cardiovascular diseases and a possible cause of residual risks other than LDL-C. As the cholesterol component of the subset of TRLs, RLP-C has been demonstrated to be a causal risk factor for ischemic heart disease (IHD) [23][24][25]. Clinical studies also revealed that higher RLP-C levels showed favorable predictive value for the risk of recurrent cardiovascular events in patients with either stable coronary artery disease (SCAD) or ACS, regardless of the baseline treatment of statins and level of LDL-C [12,[26][27][28][29]. The current analyses extend these ndings to a cohort of patients with NSTE-ACS treated with PCI and indicate that elevated RLP-C is signi cantly associated with adverse prognosis.
Previous studies have also demonstrated the signi cant association of RLP-C with plaque characteristics of the coronary arteries, such as plaque burden, composition, and vulnerability. Lina et al. revealed that RLP-C levels were signi cantly related to coronary atherosclerotic burden evaluated by computed tomography coronary angiography (CTCA), even in patients with optimal LDL-C levels [30]. Puri et al. demonstrated that non-HDL-C levels were closely correlated with the progression and regression of atherosclerotic plaque burden assessed by intravascular ultrasound (IVUS), independent of LDL-C levels [31]. Matsuo et al. found that in statin-treated patients, RLP-C levels, as opposed to LDL-C levels, were strongly associated with the proportion of plaque necrosis (a marker of plaque vulnerability) evaluated by IVUS [32]. These ndings provide important con rmation and interpretation of results from previous clinical studies, suggesting that a high RLP-C level is one of the risk factors for cardiovascular events. Additionally, this correlation between RLP-C and plaque characteristics was observed in the statin-treated and optimal LDL-C level group, indicating that high RLP-C levels may be a residual risk factor in the statin-treated population.
In this study, the LDL-C level did not show predictive value for poor prognosis, which was consistent with previous studies [5,13,29]. The underlying causes can be complex. Firstly, most participants that were enrolled in the present study underwent statin therapy, whose lipid-lowering effects in conjunction with other effects may have potential impacts on the association of LDL-C levels with adverse events. Moreover, patients with complex coronary lesions or clinical conditions may be inclined to receive more intensive lipid-lowering therapy. Such treatment selection bias or so-called "confounding by indication" may have a certain in uence on the predictive ability of LDL-C. Additionally this may lead to a paradox phenomenon, such as the phenomenon that the use of angiotensin-converting enzyme inhibitors (ACEI) could predict adverse events, which was also present in our study. The present study revealed that RLP-C levels remained a predictor of adverse prognosis despite the probable in uence of statin treatment on RLP-C levels, which indicated that RLP-C may have greater atherogenicity than other serum lipid parameters. TGs, TC, and HDL-C lost their predictability in the multivariate Cox proportional hazard analysis using covariates, including RLP-C, in the present study; which can partly be attributed to the strong correlation between them and RLP-C levels.
Results from previous studies have revealed that the impact of RLP-C seems to be more prominent in high-risk patients, such as those with metabolic syndromes or type 2 diabetes [12][13][14][15][16]. Our study also shows that RLP-C has predictive value for poor outcomes only in patients with diabetes, which indicates that there is signi cant interaction between glycometabolic status and RLP-C level on the risk of an adverse prognosis. Diabetic patients have more complex lipid metabolism disorders than non-diabetic patients characterized by increased TGs levels and decreased HDL-C levels [33]. Therefore, in addition to LDL-C, other lipid-metabolic indicators may also have a certain impact on the cardiovascular risk of diabetic patients. Previous studies have proven that hypertriglyceridemia and high TRLs play an important role in the development of coronary artery disease (CAD) [2,4,9]. TGs is predominantly carried by TRLs, which binds to arterial endothelium, where lipoprotein lipase initiates TGs hydrolysis, nally leading to the production of remnant lipoproteins. Thus, the concentrations of TGs are closely related to the cholesterol content of remnant lipoproteins, that is, RLP-C [34, 35]. The association of RLP-C with the TGs level was also veri ed in the present study. Studies have also shown that RLP-C levels increased in patients with diabetes compared with non-diabetic patients [12,26,35], which was consistent with our study. These phenomena may magnify the predictive value of RLP-C for adverse prognosis in patients with recognized diabetes.
Several pathophysiologic mechanisms may account for the association between high RLP-C levels and the increased prevalence of recurrent adverse events which was observed in the current study. These include: (1) RLP-C can upregulate the expression of intercellular adhesion molecule-1 (ICAM-1) and vascular cell adhesion molecule-1 (VCAM-1) in endothelial cells, which further induces the migration of monocytes into the arterial wall [36]; (2) RLP-C increases the generation of tissue factors (TF), which is essential for the formation of thrombus in vessels [36]; (3) There is evidence that RLP-C can enhance the aggregation of platelets [37]; (4) RLP-C promotes the propagation of smooth muscle cells that is independent from the impact of oxidative stress [38]; (5) RLP-C is causally related to low-grade in ammation, with a nearly three-fold increase in CRP level for each 1 mmol/L increase in RLP-C [39]; (6) RLP-C was demonstrated to be a risk indicator for endothelial vasomotor dysfunction [16,40]; and (7) High concentrations of RLP-C were proven to be correlated to in ammation in the arterial wall in cases of endothelial injury [41]. The pro-in ammatory and pro-atherothrombotic roles of RLP-C listed above may be the explanation for the relationship between high RLP-C levels and future adverse prognosis observed in the current study.
Studies have shown that less than a quarter of patients exhibited an LDL-C level below the guidelinerecommended target, despite remaining on statin therapy during the secondary prevention period [28,42]. This so-called "treatment gap" between the target value and clinical practice is common in the real world. In this context, while regarding LDL-C as the primary target, the exploration of residual risk factors can also provide complementary therapeutic strategies for reducing cardiovascular risk. The relationship between high RLP-C levels and increased incidence of recurrent adverse events in diabetic patients with NSTE-ACS treated with PCI demonstrated by the present study shows that RLP-C may be a complementary risk predictor and therapeutic target.
Previous reports showed that lipid-lowering agents, such as brates, ezetimibe, and statins, as well as diet adaptation, proper aerobic exercise, and obesity reduction, may effectively decrease RLP-C levels to varying degrees [26,43,44], thus enabling RLP-C as a therapeutic target. However, in addition to statin treatment for LDL-C, it is uncertain whether RLP-C should be a therapeutic target in recognized CAD patients. Clinical trials of non-statin, lipid-lowering treatments have shown signi cant bene t in reducing residual risk, but none have speci cally targeted RLP-C. Newer agents, such as potent omega-3 fatty acid derivatives [45] or antisense oligonucleotide to apolipoprotein C-III [46], were proven to have the potential to reduce TRLs signi cantly and may provide useful tools for answering this question. In JELIS (Japan EPA Lipid Intervention Study), eicosapentaenoic acid (an omega-3 fatty acid derivative) combined with low-dose statins reduced triglycerides by about 5% and coronary events by 19% compared to low-dose statins alone [47]. Novel inhibitors of apolipoprotein C-III, a key regulator of remnant metabolism, have also shown promising results [48]. Furthermore, antibodies to PCSK9, although primarily intended to lower LDL-C concentrations, was also proven to reduce the cholesterol contained in TRLs to some extent [49].
Nowadays, the pattern of targeting LDL-C alone has changed, with recent guidelines highlighting the important role of non-high-density lipoprotein cholesterol (non-HDL-C), which includes RLP-C, on the pathogenesis of atherosclerosis and thus its availability as an additional therapeutic target [11]. Therefore, it is necessary to develop new therapies targeting RLP-C and conduct randomized trials evaluating whether lowering the RLP-C level can regulate plaque morphology and reduce the residual risk of substantial cardiovascular events.
The major strengths of present study were the long-term follow-up period and the large number of the enrolled subjects. This observational cohort study also expanded the relationship between RLP-C and poor outcomes to a speci c cohort of patients with NSTE-ACS undergoing PCI. Additionally, the prognosis impact of RLP-C was evaluated in patients with differing glucose metabolic status. However, there are some limitations to our study: (1) Remnant lipoproteins mainly contain VLDL and chylomicron remnants. In the fasting state of the present study, VLDL remnants are the major constituent of circulating remnants, so that the contribution of chylomicron remnants to atherosclerosis and plaque burden may have been underestimated [50]. (2) Although potentially not as accurate as direct measurement, calculated remnant cholesterol as used in our study can be easily performed on a standard lipid pro le without any additional cost. (3) Although evidence-based statin treatment was administrated, no speci c statin agent or dose was speci ed. (4) Finally, although sequential surveillance may provide more information, only baseline lipid pro les before PCI were obtained in our study.

Conclusions
Increased RLP-C levels was a signi cant and independent predictor of adverse prognosis in diabetic patients with NSTE-ACS undergoing PCI, as opposed to in the subgroup of pre-diabetic and non-diabetic populations. The addition of the RLP-C levels had a signi cant incremental effect on the predictive value for adverse events, especially in diabetic patients. The current study indicated that the measurement of RLP-C may be important, not only for evaluating the risk of adverse prognosis, but also for tailoring treatment to prevent impending cardiovascular events in speci c populations, such as diabetic patients. Further studies investigating whether appropriate therapeutic strategies targeting RLP-C levels can signi cantly improve the prognosis of CAD patients are needed to be proceeded.

Declarations
Ethics approval and consent to participate Given the retrospective nature of the current study, the requirement for informed consent was waived. The study protocol was approved by the Clinical Research Ethics Committee of Beijing Anzhen Hospital, Capital Medical University.

Consent for publication
Not applicable.

Availability of data and materials
The datasets generated and analyzed for this study are available from the corresponding author upon reasonable request.

Competing interests
The authors declare that they have no con icts of interest. Authors' contributions QZ ( rst author) and TYZ made substantial contributions to study design, data collection, data analysis, and manuscript writing. YJZ (corresponding author) made substantial contributions to study design and intellectual direction. They contributed equally to this work. YJC, YM, YKX, JQY made contributions to data collection and analysis. All authors read and approved the nal manuscript.    Table S1).
* The HR was examined regarding the lower median of RLP-C as reference.
** The HR was examined by per 1-SD increase of RLP-C.
* The HR was examined regarding the lower median of RLP-C as reference.
** The HR was examined by per 1-SD increase of RLP-C.
RLP-C, remnant-like particle cholesterol; HR, hazard ratio; CI, con dence interval; MI, myocardial infarction. ROC, receiver operating characteristics; AUC, area under the curve; CI, con dence interval; NRI, net reclassi cation improvement; IDI, integrated discrimination improvement; RLP-C, remnant-like particle cholesterol. Figure 1 RLP-C levels in different glycometabolic status. RLP-C, remnant-like particle cholesterol.  Kaplan-Meier curves for cumulative event rate according to RLP-C levels in various subgroups with different glycometabolic status. Kaplan-Meier curves for cumulative event rate in (a-d) non-diabetic population; (e-h) pre-diabetic population; (i-l) diabetic population. RLP-C, remnant-like particle cholesterol;