Serum ceramides predict all-cause and cardiovascular mortality in Chinese incident hemodialysis patients: a retrospective cohort study

Article information

Korean J Nephrol. 2026;.j.krcp.25.227

Publication date (electronic) : 2026 May 22

doi : https://doi.org/10.23876/j.krcp.25.227

Shuang Zhang^1,2,*, Yu-Lin Wu^1,2,*, Zhe-Yuan Li^1,2, Zhi-Hong Wang^1,2, Shu-Xin Liu^,1,2

¹Department of Nephrology, Central Hospital of Dalian University of Technology (Dalian Municipal Central Hospital), Dalian, China

²Dalian Key Laboratory of Intelligent Blood Purification, Central Hospital of Dalian University of Technology (Dalian Municipal Central Hospital), Dalian, China

Correspondence: Shu-Xin Liu Department of Nephrology, Central Hospital of Dalian University of Technology (Dalian Municipal Central Hospital), No. 826, Xinan Road, Dalian, Liaoning 116033, P. R. China. E-mail: root8848@sina.com

*Shuang Zhang and Yu-Lin Wu contributed equally to this study as co-first authors.

Received 2025 July 15; Revised 2026 February 9; Accepted 2026 February 23.

Abstract

Introduction

End-stage renal disease (ESRD) represents a tremendous and escalating burden on global health. Approximately 650,000 people were afflicted with ESRD in 2010 in the United States, with projections anticipating an increase to two million cases by the year 2030 [1]. Although hemodialysis (HD) is the most effective renal replacement therapy, the annual mortality rate of patients on HD is several times higher than that of the general population [2]. Cardiovascular disease (CVD) serves as the predominant contributor to mortality of patients undergoing dialysis, with the risk of death being approximately 10–20 times higher than that of the general population [3]. According to Ahmadmehrabi and Tang [4], multiple factors elevate CVD risk in patients with ESRD. These factors include fluid overload, altered lipid metabolism, anemia, uremic cardiomyopathy, accumulation of uremic toxins derived from gut microbiota, and secondary hyperparathyroidism. Additionally, conventional HD itself affects uremic patients with impaired cardiovascular systems by mediating myocardial stress and injury [4]. In addition to the traditional cardio-renal risk factors, the accumulation of sphingolipids, such as ceramides, may also contribute to CVD [5]. Moreover, several studies have noted correlations between a lower estimated glomerular filtration rate and greater proteinuria and plasma concentrations of various sphingolipids [6,7]. Therefore, the discovery of novel biomarkers is essential to aid HD patients, allowing early intervention for the purpose of averting or delaying death.

Ceramides are synthesized through an omnipresent biosynthetic pathway starting from the generation of a sphingoid backbone through the condensation of an amino acid (generally, serine) together with palmitoyl-CoA. This scaffold acquires an additional variable fatty acid to become a ceramide, and ceramides serve as the fundamental building blocks of complex sphingolipids [5]. Structurally, ceramides consist of a long-chain base backbone linked to a fatty acyl chain of variable length. Accumulating evidence consistently demonstrates the ability of the acyl chain to have a significant influence on the biological function of ceramides. It has been proven that high-level long-chain ceramides (N-palmitoyl-sphingosine [Cer (16:0)] and N-stearoyl-sphingosine [Cer (18:0)]) in the plasma are related to elevated cardiovascular risk, while very long-chain ceramides (N-Behenoyl-ᴅ-erythro-sphingosine [Cer (22:0)] and N-lignoceroyl-sphingosine [Cer (24:0)]) at high levels probably have a “protective” effect. An important study utilizing data reported in the Ludwigshafen Risk and Cardiovascular Health study demonstrated a notable correlation between clinical cardiac events and ceramides—potentially via the glycosphingolipid pathway—irrespective of conventional risk factors for CVD [8]. In addition, Mitsnefes and colleagues [9] carried out a longitudinal and cross-sectional study revealing that plasma glucosylceramides independently predict the morbidity and mortality of CVD in HD patients. This may be attributed to the role of ceramide signaling in myocardial cell death during ischemia/reperfusion [10]. Furthermore, Savira et al. [11] demonstrated that the inhibition of ceramide synthesis mitigates myocardial injury induced by protein-bound uremic toxins, thereby providing indirect evidence for the cardiovascular impact of ceramides in uremic patients. Similarly, our previous study demonstrated that serum ceramide levels in patients with ESRD significantly decreased before and after roxadustat administration [12]. Therefore, we hypothesized that higher ceramide levels may be correlated with the elevated CVD risk and overall mortality in a cohort of HD patients.

To our knowledge, no studies have reported on the validity of ceramides as predictors of mortality from CVD plus all causes in Chinese patients undergoing maintenance HD, particularly when considering specific types of ceramide that may influence outcomes. Therefore, we conducted this retrospective cohort study to determine serum ceramide levels and assess their relation to mortality in adults undergoing HD.

Methods

Study design and participants

A retrospective cohort study was carried out at a single hospital to collect demographic, clinical, and lifestyle data from patients undergoing HD. In total, the study recruited 280 patients who underwent HD at the largest HD center located in the three provinces in northeast China, namely, the Dialysis Center of Dalian Municipal Central Hospital (Dalian, China). Recruitment took place from January 2015 to December 2017, and the follow-up period was extended to December 31, 2023.

We excluded HD patients who had missing data for the variables essential to the study (n = 20) and those with unqualified serum sample quality control (n = 7). Consequently, the final analysis was performed for 253 incident HD patients (Fig. 1). The study protocol was formulated in compliance with the principles stated in the Declaration of Helsinki and implemented under approval from the Institutional Medical Ethics Committee of the aforementioned hospital (protocol No.: YN2022-039-21). Written informed consent was obtained from all the participants.

Figure 1.

A flow chart indicates patient enrollment.

Results

The relationships between baseline clinical characteristics and the prognosis in patients with HD are presented in Table 1. Our study enrolled 253 patients in total: 60.87% were men and the median age was 59 years (interquartile range = 22). Over a median follow-up period of 88 months, 122 deaths were reported, with 85 deaths were attributed to CVD. Compared with the surviving group, the deceased group demonstrated significantly older age, elevated systolic blood pressure, increased alkaline phosphatase levels, and higher proportions of diabetic nephropathy, diabetes mellitus, and CVD. Conversely, albumin and creatinine levels were lower, with a lower proportion of glomerulonephritis (all p < 0.05). Besides, apart from the state of drinking alcohol, no statistically significant variables were observed when classified based on the Cer (24:1) level (Supplementary Table 1, available online).

Table 1.

The relationships between baseline clinical characteristics and the prognosis in patients with hemodialysis

The correlations of serum ceramide content with all-cause mortality plus CVD mortality among HD participants are shown in Tables 2 and 3. There was a relationship between higher all-cause mortality and the highest quintile of Cer (24:1)/Cer (24:0) rather than the lowest quintile (HR, 1.88; 95% CI, 1.04–3.40). Similarly, the highest quintile of Cer (24:1)/Cer (24:0), rather than the lowest quintile, was associated with higher CVD mortality (HR, 2.94; 95% CI, 1.37–6.27), with a linear trend (p for trend < 0.05). In addition, each standard deviation increase in Cer (24:1) exhibited a correlation with raised CVD mortality (HR, 1.28; 95% CI, 1.03–1.58). In competing-risk analyses (Fine-Gray), higher Cer (24:1)/Cer (24:0) (sHR, 2.49; 95% CI, 1.17–5.33) and each SD increase in Cer (24:1) (sHR, 1.30; 95% CI, 1.04–1.63) remained associated with CVD mortality (Table 4). In sensitivity analyses, we identified a similar association (Supplementary Tables 2, 3; available online).

Table 2.

Hazard ratios and 95% confidence intervals for all-cause mortality by quintile of different kinds of ceramides (Cer)

Table 3.

Hazard ratios and 95% confidence intervals for cardiovascular mortality by quintile of different kinds of ceramides (Cer)

Table 4.

Subdistribution hazard ratios and 95% confidence intervals for cardiovascular mortality by quintile of different kinds of ceramides (Cer) according to Fine-Gray competing-risk model

For potential nonlinear relationships, Cer (24:1) and Cer (24:1)/Cer (24:0) showed a positive relationship with the mortality of HD patients from all causes, indicating that the likelihood of all-cause mortality increased with the elevation of Cer (24:1) and the Cer (24:1)/Cer (24:0) ratio (all p for nonlinearity > 0.05) (Fig. 2). Similarly, we observed an increased risk of CVD mortality with rising Cer (24:1)/Cer (24:0) ratio (p for nonlinearity > 0.05). It is worth emphasizing that Cer (24:1) demonstrates a U-shaped trend with CVD mortality. Specifically, HD patients demonstrate the lowest CVD mortality risk when Cer (24:1) levels range between 1.27 and 2.50 nmol/mL. However, this non-linear relationship did not reach statistical significance (p for nonlinearity > 0.05) (Fig. 3, available online). In sensitivity analyses, we observed a similar association between Cer (24:1) and CVD mortality (Supplementary Fig. 1, available online). These findings substantiate the reliability of our main results.

Figure 2.

The dose-response curve of the relationship between (A) Cer (24:1) level and (B) Cer (24:1)/Cer (24:0) and all-cause mortality.

The dark blue line and shaded area represent the estimated hazard ratio (HR) and 95% confidence interval (CI).

Cer, ceramide; Cer (24:0), N-lignoceroyl-sphingosine; Cer (24:1), N-nervonoyl-sphingosine.

Figure 3.

The dose-response curve of the relationship between (A) Cer (24:1) level and (B) Cer (24:1)/Cer (24:0) and cardiovascular mortality.

The dark blue line and shaded area represent the estimated hazard ratio (HR) and 95% confidence interval (CI).

Cer, ceramide; Cer (24:0), N-lignoceroyl-sphingosine; Cer (24:1), N-nervonoyl-sphingosine.

Discussion

Our study indicated that, among patients undergoing HD, higher mortality was associated with higher Cer (24:1) and Cer (24:1)/Cer (24:0) ratio, even after adjusting for potential confounders. Additionally, our analysis suggested the presence of a positive linear relationship between Cer (24:1), Cer (24:1)/Cer (24:0) and all-cause mortality as well as Cer (24:1)/Cer (24:0) and CVD mortality in patients on HD.

Current research on the relationship between ceramides and prognosis is mainly focused on healthy populations, those with CVD, and those with type 2 diabetes. For instance, the Ludwigshafen Risk and Cardiovascular Health study [15] was the first to introduce ceramides as biomarkers for assessing long-term cardiovascular events in clinical practice, and included 3,316 participants in Germany (2,583 coronary artery disease-positive patients and 733 healthy controls). The study found that Cer (16:0), Cer (18:0), and Cer (24:1) at higher levels were indicative of an incremental risk of cardiovascular events, independent of the traditional lipid biomarkers HDL and LDL. Similarly, Wang et al. [16] conducted a prospective case cohort study involving 980 participants and found that baseline plasma concentrations of such ceramides as Cer (16:0), Cer (22:0), Cer (24:0), and Cer (24:1) had correlations with the risk of incident CVD and cardiovascular death. Meeusen et al. [17] conducted a study of 495 coronary angiography patients at the Mayo Clinic in the United States and found that Cer (16:0), Cer (18:0), Cer (24:1), and Cer (24:1)/Cer (24:0) were able to forecast the occurrence of primary adverse cardiovascular events within a follow-up period of 4 years. In addition to the aforementioned research on CVD, Vasile et al.’s study [18] validated the relationship between elevated levels of Cer (24:1)/Cer (24:0) and the raised all-cause mortality risk; the findings of this research coincided with these previous studies. In our study, however, a statistically significant association was not found between Cer (16:0), Cer (18:0), and CVD death, was probably caused by differences in the research participants, the small sample size, and limited follow-up time. In an observational study of 423 elderly Chinese patients with chronic heart failure, Yu et al. [19] reported that plasma total ceramide scores were associated with higher all-cause mortality rates. Our results were not consistent with theirs, possibly because of the limited number of participants. In addition, there is currently controversy surrounding research on Cer (24:0). A cohort study of 1,704 Chinese patients with coronary artery disease by Li et al. [20] found that the levels of Cer (24:0) were significantly positively correlated with CVD and all-cause mortality risks. In contrast, Laaksonen et al.’s study [13] considered Cer (24:0) to be a protective factor for cardiovascular events, while Meeusen et al.’s study [17] considered Cer (24:0) to have no clinical significance. Our study did not find any statistical significance for Cer (24:0); therefore, multicenter prospective studies with larger samples are needed to verify these findings.

Although the biological mechanisms underlying the impact of ceramides on CVD and all-cause mortality remain incompletely understood, epidemiological evidence suggests a more significant effect of ceramides on CVD pathogenesis compared with other causes of death. This heightened association is likely due to their involvement in mediating inflammation and vascular damage. Patients on HD have higher levels of inflammation than the general population. Ceramides are integral components of membrane stability and function as secondary messengers in membrane signaling pathways implicated in inflammation [21]. A previous study has confirmed a close relationship between ceramides and inflammation [22]. Cell-based analyses have demonstrated that inhibiting ceramide production can mitigate the inflammatory response [23]. In vitro experiments have shown that alterations in ceramide synthesis are linked to oxidative stress and inflammation. Additionally, ceramides interfere with mitochondrial respiration, causing mitochondrial dysfunction that subsequently impairs cardiac function [24]. The heart’s response to acute ischemia–reperfusion can lead to the production of certain ceramides, resulting in an increase in specific ceramides, which in turn activates mitochondrial autophagy and apoptosis. Studies have shown that ceramides containing long and very-long chains are considered to have the most significant impact on cardiac dysfunction [25], and they exhibit potent cytotoxicity [26]. The accumulation of these ceramides can lead to cardiac remodeling and ultimately result in heart failure. In addition, vascular endothelial cells serve as the first line of defense in the cardiovascular system and play a crucial role in maintaining the integrity and endocrine function of the cardiovascular system. In disease states including uremia, the up-regulation of numerous inflammatory factors can disrupt the normal endocrine function of endothelial cells, leading directly to endothelial cell apoptosis [13].

To eliminate the reverse causal relationship between Cer (24:1) and CVD death, along with potential bias due to early mortality, baseline characteristics across Cer (24:1) level groups demonstrate that patients with low Cer (24:1) did not exhibit more severe clinical conditions at baseline, thus substantially reducing the likelihood of reverse causality. Meanwhile, sensitivity analysis suggests that the association established between Cer (24:1) and CVD death is unlikely to be driven by early mortality, but rather might be attributable to the intrinsic biological effects of Cer (24:1) itself. Notably, Cer (24:1), a pivotal molecule in sphingolipid metabolism exhibits a U-shaped trend with CVD death in HD patients; however, this non-linear relationship fails to achieve statistical significance. Nevertheless, this might also imply the possibility of dual biological effects. As we recognize, ceramides are bioactive lipids that play a dual role in endothelial function. Basal or acutely generated ceramide is necessary for proper nitric oxide (NO) signaling and vasodilation in healthy endothelium. Inhibition of ceramide formation (e.g., via neutral sphingomyelinase blockade) leads to a switch from NO-mediated to hydrogen peroxide-mediated vasodilation, which is less protective and may impair vascular regulation [27]. Conversely, excessive ceramides dysregulate insulin signaling, vascular endothelial function, inflammatory responses, oxidative stress, and lipoprotein aggregation, thereby contributing to atherosclerosis and vascular pathogenesis [28]. Furthermore, ceramides exacerbate myocardial pathology by promoting pathological cardiac remodeling and cardiomyocyte apoptosis [28]. Therefore, suboptimal concentrations of Cer (24:1), whether deficient or excessive, adversely impact cardiovascular prognosis in HD patients. In clinical management, Cer (24:1) levels should be optimally maintained within physiological concentration ranges to preserve cell membrane integrity and ensure fidelity of cellular signal transduction [29].

As we all know, the level of ceramides does not exist in isolation, it is regulated by various physiological and pathophysiological processes. Thus, several factors might influence baseline serum ceramide levels. For instance, the physiological characteristics and living environment differences of various ethnic and regional populations may directly or indirectly alter the baseline levels of serum ceramides. This might be because there are significant differences in gene polymorphisms related to ceramide synthesis (such as the serine palmitoyltransferase genes SPTLC1/2) [30,31] and metabolism (such as the ceramide enzyme gene ASAH1) [32] among different ethnic populations. These genetic differences may lead to disparities in the synthesis rate and degradation efficiency of ceramides in vivo, resulting in race-specific baseline levels of serum ceramides. Besides, due to differences in the composition of causes of death among HD patients of different ethnicities/regions, as well as differences in treatment plans at HD centers in different regions, which also may affect the association between ceramides and the prognosis of HD patients. Additionally, there are significant differences in dietary patterns among populations in different regions. For example, a high saturated fat [33] diet (common in some Western populations) may promote ceramide synthesis by activating the sphingolipid metabolism pathway; whereas a diet rich in Omega-3 unsaturated fatty acids, such as that of some East Asian populations, may have a downregulating effect on ceramide levels [34]. At the same time, differences in physical activity intensity, smoking and drinking rates, and other lifestyle habits among populations in different regions may indirectly alter ceramide levels by affecting lipid metabolism. Furthermore, the prevalence and severity of basic diseases such as diabetes mellitus, hypertension, and hyperlipidemia are different among HD patients of various races/regions. These diseases themselves are significant triggers for elevated levels of ceramides, and their distribution differences may further amplify the disparities in ceramide levels between different populations. Moreover, as described above, HD patients generally have a chronic micro inflammatory state, characterized by elevated levels of circulating inflammatory cytokines (such as interleukin-6 [IL-6], tumor necrosis factor-alpha [TNF-α]) [35]. Several studies have shown that inflammatory signaling pathways (such as activation of sphingomyelinase by TNF-α) can directly promote the hydrolysis of sphingomyelin into ceramides [36,37]. In addition, the prevalence of insulin resistance and diabetes in HD patients is very high, which may lead to an increase in ceramide levels [38]. Dialysis patients often have unique lipid metabolism disorders, such as hypertriglyceridemia and low HDL [39]. As the core of sphingolipid metabolism, the availability of lipid substrates affects the synthesis of ceramides. Meanwhile, the nutritional status of patients may also affect ceramide levels through overall metabolic changes [40]. Furthermore, dialysis related factors such as membrane biocompatibility, dialysis efficiency/dosage, and patient medication may all affect ceramide levels.

Most importantly, the above factors may affect the predictive value of serum ceramides for all-cause mortality and CVD mortality in HD patients through confounding and mediating effects. Therefore, in our study, we have endeavored to control for potential confounding factors to clarify the association between serum ceramide levels and all-cause and CVD mortality in HD patients. However, the persistence of this association following rigorous adjustment for these confounders in our multivariate models suggests that ceramide may function not merely as a passive biomarker but as an active contributor to the pathological processes culminating in mortality. We hypothesize that in HD patients, ceramide integrates multiple metabolic and inflammatory insults, thereby directly triggering cardiotoxic and pro-apoptotic signaling pathways, and thus serves as a key mediator in the ‘maladaptive’ response to the underlying HD condition. Future mechanistic investigations and interventional studies are required to elucidate this causal relationship definitively and to evaluate strategies aimed at lowering ceramide levels as a novel therapeutic approach.

Our study has several strengths. First, the correlation of serum ceramides with mortality of patients undergoing incident HD in China was investigated in this study for the first time. Second, our analyses accounted for established and suspected potential confounding factors. Third, we explored the dose–response relationship between serum Cer (24:1) levels as well as the Cer (24:1)/Cer (24:0) ratio and the mortality of the participants from both all causes and CVD. Fourth, to prevent the influence of lag effects on the outcomes, we excluded patients who died within the initial 18 months of follow-up for sensitivity analysis to explore the association between Cer (24:1) levels and CVD mortality.

However, several important limitations of this study should be highlighted. First, the small sample size of the outcome events prevented the observation of statistically significant associations between serum ceramide levels and mortality, except for Cer (24:1)/Cer (24:0) plus Cer (24:1). In the future, multicenter prospective cohort studies should be conducted to explore these associations in patients with incident HD. Second, given that the study design was a retrospective cohort, it can establish only associations rather than causality. Therefore, large randomized controlled trials are needed in the future to confirm this causal association. Third, as the original data had inherent limitations, despite the adjustments made for a number of potential confounders, several residual confounders or factors that remain unknown, such as information on dietary patterns, socioeconomic status, and physical activity, residual renal function, inflammatory markers (e.g., IL-6, high-sensitivity C-reactive protein), could potentially have an impact on the association of serum ceramides with HD patients’ survival. Therefore, future studies should incorporate these measurements to validate our findings. Fourth, our study did not account for time-varying treatment-related variables such as vascular access type, spKt/V_urea, ultrafiltration rate, interdialytic weight gain, or dialysate composition. Notably, all baseline data were collected during the one-month period preceding the initial dialysis session, specifically to capture the patient’s intrinsic risk profile at the transition to dialysis. We intentionally focused on baseline characteristics to isolate pre-dialysis patient-level risk factors rather than dynamic dialysis-related modifiers, which is why time-varying treatment parameters were not included in the primary analytical framework. Thus, these treatment parameters were not identified as confounders at baseline. Future studies incorporating serial biomarker measurements and time-updated dialysis parameters are warranted to elucidate the independent prognostic significance of ceramides, disentangled from the effects of dialysis-related modifiers. Finally, the patients enrolled in this study were all from a single dialysis center, which may limit their representativeness. Consequently, the findings might not be generalizable to the overall HD population. Therefore, international multicenter research is needed in the future to further validate our findings.

In conclusion, our study found strong associations evident between serum Cer (24:1) content plus the Cer (24:1)/Cer (24:0) ratio and the mortality risk from all causes as well as CVD in the follow-up period with the traditional risk factors controlled. It is imperative to conduct in-depth studies on ceramides regarding their clinical diagnostic value for the identification of HD patients facing a mortality risk.

Supplementary Materials

Article information Continued