Association of remnant cholesterol with long-term outcomes in kidney transplant recipients: a retrospective cohort study Authors: Jingru Chen, Xinyi Gao, Xinxuan Meng, Xiaowei Lou, Jianghua Chen Published: 2026 ABSTRACT AbstractBackgroundPost-transplant dyslipidemia is frequently observed and has been associated with graft survival. Traditional lipid parameters show inconsistent associations with graft outcomes, whereas remnant cholesterol (RC) may provide additional prognostic value; however, its role remains unclear in kidney transplant recipients.MethodsIn this retrospective cohort of 1,943 living donor kidney transplant recipients, a 365-day landmark design was applied. Lipid exposure was defined as mean levels within the first year after transplantation, including RC, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), non-HDL-C, total cholesterol (TC), and triglycerides (TG). Multivariable Cox proportional hazards models were used to estimate adjusted hazard ratios, and restricted cubic spline analyses were used to explore potential dose–response patterns. Lipid parameters were additionally analyzed as categorical variables. Subgroup analyses assessed potential effect modification, and incremental value analyses evaluated model improvement.ResultsElevated RC (HR 2.38, 95% CI: 1.26–4.50, P = 0.008) and TG (HR 1.38, 95% CI: 1.06–1.79, P = 0.015) were independently associated with higher allograft loss risk and showed dose–response patterns. Similar results were observed in categorical analyses. Non-HDL-C was not significantly associated with allograft loss in the primary analysis (HR 1.64, 95% CI: 1.00–2.70, P = 0.051), and LDL-C, HDL-C, and TC showed no consistent associations with graft outcomes. No lipid parameter was significantly associated with all-cause mortality. The TG–allograft loss association was stronger among recipients with fewer human leukocyte antigen mismatches (P for interaction = 0.012). Adding RC to the baseline risk model modestly improved model performance (ΔC-index = 0.008, P = 0.007; ΔAIC = − 4.78).ConclusionsHigher RC and TG levels were independently associated with an increased risk of kidney allograft loss in living donor recipients. RC showed the most consistent association with graft outcomes and modest improvement in model performance, supporting further evaluation of RC as an adjunctive marker in post-transplant lipid assessment and long-term graft care.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12944-026-02996-1.Keywords: Remnant cholesterol, Kidney transplantation, Graft survival, Dyslipidemias, Triglycerides Introduction End-stage renal disease (ESRD) is defined clinically by an estimated glomerular filtration rate (eGFR) below 15 mL/min/1.73 m2 and often necessitates maintenance dialysis or kidney transplantation, placing a substantial burden on global health systems [1]. Kidney transplantation remains the preferred therapeutic option for ESRD, providing more durable kidney function and better quality of life than dialysis [2]. Despite advances in operative techniques and immunosuppressive therapy, kidney allograft loss rates remain unacceptably high, with approximately 20% of recipients experiencing graft loss within five years and up to 50% within ten years [3]. Addressing the challenges posed by the increasing ESRD burden and high allograft failure rates requires identifying independent, modifiable risk factors. Dyslipidemia is highly prevalent after kidney transplantation, affecting 75%−95% of recipients, and typically presents with increased total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), and triglyceride (TG) levels, together with reduced high-density lipoprotein cholesterol (HDL-C) levels [4, 5]. Dyslipidemia in this population is associated not only with cardiovascular disease (CVD) risk but also with allograft injury and late allograft loss [6, 7]. However, evidence regarding the associations of conventional lipid parameters, including LDL-C and TC, with long-term post-transplant outcomes remains inconsistent, and their relationship with allograft loss has not been fully clarified [8, 9]. In recent years, the potential relevance of non-traditional lipid parameters in kidney transplant recipients has gained increasing attention. Remnant cholesterol (RC), which reflects the cholesterol content of triglyceride-rich remnant lipoproteins [10], has been linked to a higher risk of incident cardiovascular disease (CVD) events and all-cause mortality in this population [11]. Beyond transplantation, evidence from studies of early kidney injury in individuals with normal renal function, cardiometabolic risk in hypertensive patients, and non-alcoholic fatty liver disease indicates that RC is associated with adverse outcomes independently of conventional lipid parameters, and in some settings may show stronger associations [12–14]. Non-high-density lipoprotein cholesterol (non-HDL-C) represents the total burden of atherogenic lipoproteins, excluding HDL-C, and has been associated with cardiovascular risk in both general and dialysis populations [15, 16]. Among kidney transplant recipients, higher non-HDL-C levels have been linked to adverse cardiovascular and cerebrovascular events and all-cause mortality, and may also be related to allograft survival [17]. Nevertheless, studies specifically investigating RC and non-HDL-C in kidney transplant recipients remain limited, and their associations with long-term allograft outcomes have not been well characterized. Therefore, higher post-transplant RC levels were hypothesized to show an independent association with kidney allograft loss and to improve risk stratification beyond conventional lipid parameters. To address this hypothesis, this study examined the relationships between post-transplant lipid parameters (RC, LDL-C, HDL-C, non-HDL-C, TC, and TG) and kidney allograft loss as well as all-cause mortality in living donor kidney transplant recipients. Using a landmark design, this study further compared conventional and non-traditional lipid parameters to assess the potential role of RC in long-term post-transplant risk assessment. Methods Study design and population This retrospective cohort study included kidney transplant recipients. To reduce potential bias related to early post-transplant events, a landmark design was applied, with subsequent analyses focusing on patients who survived with a functioning graft beyond the early period after transplantation. Overall, 2,633 patients who received their first living donor kidney transplantation at the Kidney Disease Center, First Affiliated Hospital, Zhejiang University School of Medicine, between January 2011 and August 2024 were included. Patients were excluded if they had missing survival or graft survival data (n = 9), were younger than 14 years (n = 10), or had missing lipid data (n = 330). The initial (pre-landmark) cohort included 2,284 recipients. Ethics approval was obtained from the Clinical Research Ethics Committee of the First Affiliated Hospital, Zhejiang University School of Medicine, with a waiver of informed consent (IIT20251541A). Data collection Baseline demographic, clinical, transplant-related, and laboratory data were extracted from the hospital information system (HIS). Recipient characteristics included age, sex, body mass index (BMI), primary renal diagnosis, and dialysis history. Donor characteristics included age, sex, and BMI. Transplant-related variables included year of surgery, ABO compatibility, number of human leukocyte antigen (HLA) mismatches, panel reactive antibody (PRA) positivity, cold and warm ischemia times, and induction regimen. Primary renal diagnosis, pre-transplant dialysis history, and induction regimen were categorized into predefined clinical groups, as shown in Table 1 and Supplementary Table S1. Post-transplant complications included delayed graft function (DGF) and acute rejection (AR). DGF was defined as the requirement for at least one dialysis session during the first post-transplant week [18]. AR was biopsy-defined [19]. Creatinine, eGFR, C-reactive protein, blood glucose, hemoglobin, LDL-C, HDL-C, TC, and TG were analyzed using routine clinical laboratory methods. Per institutional practice, blood samples for lipid measurements were typically obtained after fasting for at least 8 h. Table 1Characteristics of landmark cohort according to remnant cholesterol quartiles (selected)VariableOverallRemnant cholesterol, mmol/L(mg/dL)PQ1 (< 0.56 [< 22])Q2 (0.56–0.69[22–27])Q3 (0.69–0.84[27–33])Q4 (≥ 0.84 [≥ 33])n1943486486485486Female568 (29.23)173 (35.60)144 (29.63)128 (26.39)123 (25.31)0.002Recipient age, years32.44 (8.41)31.34 (7.90)32.18 (8.46)32.75 (8.67)33.49 (8.46)0.001BMI, kg/m221.36 (3.35)20.52 (2.95)21.25 (3.29)21.49 (3.39)22.21 (3.52)< 0.001Surgery year2017.88 (3.39)2017.78 (3.31)2018.18 (3.16)2018.01 (3.32)2017.56 (3.71)0.027Dialysis history0.149 Hemodialysis1198 (61.66)287 (59.05)284 (58.44)317 (65.36)310 (63.79) Peritoneal dialysis467 (24.03)117 (24.07)126 (25.93)110 (22.68)114 (23.46) None278 (14.31)82 (16.87)76 (15.64)58 (11.96)62 (12.76)Donor age, years53.43 (7.39)52.90 (7.11)53.68 (7.06)53.57 (7.39)53.56 (7.97)0.331Donor BMI, kg/m223.29 (2.84)23.02 (2.71)23.36 (2.93)23.47 (2.86)23.29 (2.85)0.082ABO incompatible125 (6.43)25 (5.14)33 (6.79)27 (5.57)40 (8.23)0.199HLA mismatch2.82 (1.47)2.70 (1.44)2.86 (1.43)2.81 (1.52)2.91 (1.50)0.157PRA positivity133 (6.85)45 (9.26)28 (5.76)33 (6.80)27 (5.56)0.086Cold ischemia time, hours2.00 [2.00, 3.00]2.00 [2.00, 3.00]2.00 [2.00, 3.00]2.00 [2.00, 3.00]2.00 [2.00, 3.00]0.103Warm ischemia time, minutes2.00 [2.00, 3.00]2.00 [2.00, 3.00]2.00 [2.00, 3.00]3.00 [2.00, 3.00]3.00 [2.00, 3.00]0.033eGFR, mL/min/1.73m264.03 (16.40)67.63 (16.47)64.33 (15.96)62.01 (16.40)62.12 (16.19)< 0.001C-reactive protein, mg/L6.24 [3.40, 12.21]5.80 [3.26, 10.80]5.94 [3.31, 11.88]6.60 [3.40, 13.00]6.76 [3.56, 13.05]0.054Blood glucose, mmol/L5.17 [4.90, 5.48]5.11 [4.87, 5.35]5.14 [4.87, 5.43]5.22 [4.92, 5.56]5.25 [4.97, 5.69]< 0.001Hemoglobin, g/L121.01 (14.35)119.80 (13.83)121.88 (14.57)120.85 (14.61)121.50 (14.31)0.118RC, mmol/L0.72 (0.24)0.48 (0.06)0.63 (0.04)0.76 (0.05)1.03 (0.23)< 0.001LDL-C, mmol/L2.41 (0.57)2.06 (0.45)2.29 (0.46)2.54 (0.55)2.74 (0.58)< 0.001HDL-C, mmol/L1.29 (0.27)1.42 (0.28)1.32 (0.27)1.25 (0.25)1.16 (0.22)< 0.001non-HDL-C, mmol/L3.15 (0.72)2.56 (0.49)2.94 (0.48)3.31 (0.57)3.79 (0.67)< 0.001TC, mmol/L4.44 (0.75)3.98 (0.62)4.26 (0.60)4.56 (0.69)4.94 (0.73)< 0.001TG, mmol/L1.67 (0.60)1.21 (0.26)1.46 (0.25)1.71 (0.31)2.29 (0.74)< 0.001DGF37 (1.90)9 (1.85)8 (1.65)8 (1.65)12 (2.47)0.758AR94 (4.84)18 (3.70)23 (4.73)26 (5.36)27 (5.56)0.532All-cause mortality36 (1.85)8 (1.65)8 (1.65)8 (1.65)12 (2.47)0.716Kidney allograft loss156 (8.03)25 (5.14)27 (5.56)43 (8.87)61 (12.55)< 0.001Death-censored graft failure141 (7.26)21 (4.32)24 (4.94)39 (8.04)57 (11.73)< 0.001Continuous variables are presented as mean (SD) if normally distributed, or median [Q1, Q3] if not. Categorical variables are presented as n (%)HLA Human leukocyte antigen, PRA Panel reactive antibody, eGFR estimated glomerular filtration rate, RC Remnant cholesterol, LDL-C Low-density lipoprotein cholesterol, HDL-C High-density lipoprotein cholesterol, non-HDL-C non-high-density lipoprotein cholesterol, TC Total cholesterol, TG Triglycerides, DGF Delayed graft function, AR Acute rejectionA complete version of this table is provided in Supplementary Table S1 Characteristics of landmark cohort according to remnant cholesterol quartiles (selected) Continuous variables are presented as mean (SD) if normally distributed, or median [Q1, Q3] if not. Categorical variables are presented as n (%) HLA Human leukocyte antigen, PRA Panel reactive antibody, eGFR estimated glomerular filtration rate, RC Remnant cholesterol, LDL-C Low-density lipoprotein cholesterol, HDL-C High-density lipoprotein cholesterol, non-HDL-C non-high-density lipoprotein cholesterol, TC Total cholesterol, TG Triglycerides, DGF Delayed graft function, AR Acute rejection A complete version of this table is provided in Supplementary Table S1 Lipid parameters RC was calculated as TC minus LDL-C minus HDL-C [20], and non-HDL-C as TC minus HDL-C [16]. The primary exposures were the mean levels of RC, LDL-C, HDL-C, non-HDL-C, TC, and TG within the first year after transplantation. Other laboratory biomarkers were similarly summarized as mean values over the same period. This time window was chosen to capture the peak period of dyslipidemia while allowing for stabilization of metabolic and immunosuppressive conditions, providing a representative measure of early post-transplant lipid status [21]. Outcomes The primary outcome was kidney allograft loss, defined as a composite of patient death, return to maintenance dialysis, or retransplantation [9]. The secondary outcome was all-cause mortality, defined as death from any cause. Outcome data were obtained from the HIS and the Zhejiang Provincial Public Security Department. Additional outcome information was obtained through telephone follow-up. Vital status was available through December 31, 2024. Data processing and multiple imputation Outliers in continuous variables were coded as missing to reduce bias, and variables with more than 30% missingness were excluded. Multiple imputation by chained equations (MICE) was performed for the remaining variables using the R package mice, with methods selected according to variable type and fixed variables excluded from imputation. Regression estimates from each imputed dataset were pooled using Rubin’s rules. For visualization purposes, analyses were performed using a single representative imputed dataset. Landmark To establish a clear temporal sequence between exposure variables and outcome events and reduce the potential for time-related bias [22], a landmark analysis was performed at 365 days post-transplant. The landmark time point was used as the new baseline, and follow-up time was calculated from day 365 until event occurrence or censoring. Among the initial cohort, 341 recipients died, experienced allograft loss within the first year after transplantation, or had less than 365 days of follow-up and were therefore excluded from the landmark analysis. The landmark cohort included 1,943 recipients. All subsequent analyses were conducted during the post-landmark follow-up period. A detailed flowchart of participant selection and landmark analysis is presented in Fig. 1. Study flowchart Statistical analyses Participant characteristics were described for the landmark cohort overall and by RC quartile. Continuous and categorical variables were presented using appropriate descriptive statistics and compared across RC quartiles with appropriate parametric or nonparametric tests according to variable type and distribution. Cox proportional hazards models were fitted to estimate adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for the associations between continuous lipid parameters and outcomes. Due to multicollinearity among lipid parameters, each lipid parameter was analyzed in a separate model. Restricted cubic splines (RCS) with three prespecified knots were used to evaluate potential nonlinear relationships between lipid parameters and outcomes. Covariates were selected according to clinical relevance [23], and model complexity was limited to avoid overfitting [24]. Models for allograft loss were adjusted for recipient age, sex, BMI, surgery year, dialysis history, donor age, donor BMI, ABO compatibility, HLA mismatch count, PRA positivity, cold ischemia time, blood glucose, and AR. Given the limited number of deaths, models for all-cause mortality were adjusted for recipient age, sex, and AR. Lipid parameters were grouped into quartiles according to their distributions in the study population. In sensitivity analyses, lipid parameters were also categorized according to clinically relevant thresholds derived from published guidelines and prior studies [25, 26]. Cumulative event curves were estimated with the Kaplan–Meier method and compared using the log-rank test. Associations between lipid categories and outcomes were further evaluated using Cox models adjusted for the same covariates. Subgroup analyses were conducted only for RC and TG, which were associated with allograft loss in the primary analysis. Stratified Cox models were adjusted for recipient age, sex, BMI, surgery year, donor age, donor BMI, HLA mismatch count, cold ischemia time, blood glucose, and AR, with stratification variables specified in the corresponding figure. To evaluate whether each lipid parameter provided incremental value beyond established covariates in the allograft loss model, each parameter was entered separately into the base allograft loss model. Model improvement was assessed using discrimination, likelihood ratio (LR) test, and information criteria. A two-sided P < 0.05 was considered statistically significant. All statistical analyses were conducted using R software, version 4.5.0. Sensitivity analyses A series of sensitivity analyses were performed. First, death-censored graft failure (DCGF) was analyzed, defined as return to maintenance dialysis or retransplantation. Cases in which allograft loss and death occurred on the same day could not be reliably classified and were therefore excluded. Second, the mean lipid levels during the entire follow-up period were used as a supplementary cumulative-exposure metric to approximate long-term lipid burden, analogous to cumulative or time-weighted average lipid exposure metrics used in longitudinal cardiometabolic studies [27, 28]. This analysis is exploratory and may still be susceptible to immortal time bias, reverse causation, and time-varying confounding [29]. Third, analyses were repeated in participants with complete data. Results The landmark cohort comprised 1,943 kidney transplant recipients, among whom 156 experienced kidney allograft loss and 36 died during follow-up (Table 1). Recipients had a mean age of 32.44 ± 8.41 years, 29.23% were female, and 61.66% had undergone hemodialysis before transplantation. The mean RC was 0.72 ± 0.24 mmol/L. Higher RC quartiles were characterized by a greater proportion of male recipients, older age, higher BMI, and earlier surgery years. Regarding laboratory parameters, eGFR decreased across increasing RC quartiles, and blood glucose levels were higher in higher RC groups. The incidence of kidney allograft loss and DCGF increased progressively with higher RC quartiles, with event rates in Q4 approximately double those observed in Q1. Continuous lipid parameters In Cox models adjusted for covariates, higher first-year mean RC (HR 2.38, 95% CI: 1.26–4.50, P = 0.008) and TG (HR 1.38, 95% CI: 1.06–1.79, P = 0.015) levels were significantly associated with kidney allograft loss (Table 2). Non-HDL-C was not significantly associated with this outcome (HR 1.23, 95% CI: 0.97–1.56, P = 0.089), and no significant associations were found for LDL-C, HDL-C, or TC. Table 2Association between continuous lipid levels and kidney allograft loss and all-cause mortality in the landmark cohortVariablesKidney allograft lossaAll-cause mortalitybHR95% CIPHR95% CIPRC, mmol/L2.381.26–4.500.0081.640.51–5.280.409LDL-C, mmol/L1.200.89–1.620.2330.890.47–1.690.731HDL-C, mmol/L0.790.40–1.590.5170.390.10–1.520.174non-HDL-C, mmol/L1.230.97–1.560.0890.920.56–1.530.761TC, mmol/L1.170.94–1.460.1680.820.51–1.320.414TG, mmol/L1.381.06–1.790.0151.270.78–2.070.337HR Hazard ratio, CI Confidence interval, RC Remnant cholesterol, LDL-C Low-density lipoprotein cholesterol, HDL-C High-density lipoprotein cholesterol, non-HDL-C non-high-density lipoprotein cholesterol, TC Total cholesterol, TG TriglycerideaMultivariate Cox regression models were additionally adjusted for recipient age, sex, BMI, surgery year, dialysis history, donor age, donor BMI, ABO compatibility, HLA mismatch count, PRA positivity, cold ischemia time, blood glucose, and acute rejectionbMultivariate Cox regression models were additionally adjusted for recipient age, sex, and acute rejectionEach model contained only 1 lipid parameter at a time. HRs for continuous variables represent the risk increase per 1 mmol/L increment Association between continuous lipid levels and kidney allograft loss and all-cause mortality in the landmark cohort HR Hazard ratio, CI Confidence interval, RC Remnant cholesterol, LDL-C Low-density lipoprotein cholesterol, HDL-C High-density lipoprotein cholesterol, non-HDL-C non-high-density lipoprotein cholesterol, TC Total cholesterol, TG Triglyceride aMultivariate Cox regression models were additionally adjusted for recipient age, sex, BMI, surgery year, dialysis history, donor age, donor BMI, ABO compatibility, HLA mismatch count, PRA positivity, cold ischemia time, blood glucose, and acute rejection bMultivariate Cox regression models were additionally adjusted for recipient age, sex, and acute rejection Each model contained only 1 lipid parameter at a time. HRs for continuous variables represent the risk increase per 1 mmol/L increment In RCS analyses, RC and TG showed statistically significant positive associations with kidney allograft loss (P for overall = 0.025 and 0.049, respectively; Fig. 2). Non-HDL-C showed no statistically significant association, although the spline curve showed some variation across the range of values. No clear associations were observed for HDL-C or TC. Tests for nonlinearity were not statistically significant for any lipid parameter (all P for nonlinear > 0.05). Nonlinear Relationships of Lipid Parameters with Kidney Allograft Loss. Solid lines represent hazard ratios (HRs), and the shaded areas denote the 95% confidence intervals (CIs). The dashed line indicates the reference line where HR = 1. Multivariable Cox regression models were adjusted for recipient age, sex, BMI, surgery year, dialysis history, donor age, donor BMI, ABO compatibility, HLA mismatch count, PRA positivity, cold ischemia time, blood glucose, and acute rejection. The P value for the overall association test (P overall) and the P value for the test of nonlinearity (P nonlinear) are presented However, none of the lipid parameters were significantly associated with all-cause mortality in either Cox or RCS analyses (Supplementary Figure S1). Lipid groups For kidney allograft loss, cumulative event rates differed significantly across quartiles of RC, TG, and HDL-C (P < 0.001, P = 0.004, and P = 0.007, respectively; Fig. 3 and Supplementary Figure S2). Event rates were higher in the upper RC and TG quartiles and lower in the upper HDL-C quartiles. No statistically significant differences were observed across LDL-C, TC, or non-HDL-C quartiles. For all-cause mortality, none of the lipid parameters showed significant differences across quartiles. Cumulative Event Curves of Kidney Allograft Loss and All-Cause Mortality According to Quartile Groups of RC, non-HDL-C, and TG. The cumulative event curves of kidney allograft loss and all-cause mortality according to quartile groups (Q1–Q4) of remnant cholesterol (RC), non-high-density lipoprotein cholesterol (non-HDL-C), and triglycerides (TG). Outcomes were tracked until December 31, 2024. The P values were calculated using the log-rank test. The number of events indicated in the figure represents the total number of outcome events that occurred during the follow-up period. No., number In multivariable Cox models, recipients in the highest RC quartile (≥ 0.84 mmol/L) had a higher risk of kidney allograft loss than those in the lowest quartile (HR 1.91, 95% CI: 1.18–3.09, P = 0.008). The highest TG quartile (≥ 1.89 mmol/L) was similarly associated with increased risk (HR 1.65, 95% CI: 1.05–2.60, P = 0.031). For non-HDL-C, the highest quartile showed a non-significant trend toward higher risk (HR 1.64, 95% CI: 1.00–2.70, P = 0.051). No significant associations were found across quartiles of LDL-C, HDL-C, or TC (Table 3). Table 3Association between quartile groups of lipid levels and kidney allograft loss and all-cause mortality in the landmark cohortVariablesKidney allograft lossaAll-cause mortalitybHR95% CIPHR95% CIPRC, mmol/L (mg/dL)< 0.56 (< 22)Reference--Reference--0.56–0.69 (22–27)1.080.62–1.870.7900.950.36–2.540.9190.69–0.84 (27–33)1.590.96–2.630.0730.820.31–2.200.697≥ 0.84 (≥ 33)1.911.18–3.090.0081.150.47–2.830.758LDL-C, mmol/L (mg/dL)< 1.99 (< 77)Reference--Reference--1.99–2.37 (77–92)0.770.48–1.240.2850.910.38–2.210.8432.37–2.78 (92–107)1.100.71–1.700.6840.500.18–1.400.185≥ 2.78 (≥ 107)1.220.76–1.950.4150.890.36–2.230.809HDL-C, mmol/L (mg/dL)< 1.10 (< 43)Reference--Reference--1.10–1.26 (43–49)0.810.53–1.250.3491.330.53–3.320.5461.26–1.45 (49–56)0.910.59–1.410.6671.450.59–3.560.415≥ 1.45 (≥ 56)0.810.49–1.350.4150.470.14–1.620.232non-HDL-C, mmol/L (mg/dL)< 2.64 (< 102)Reference--Reference--2.64–3.10 (102–120)1.250.76–2.040.3831.000.40–2.510.9963.10–3.60 (120–139)1.210.74–1.980.4360.460.16–1.350.158≥ 3.60 (≥ 139)1.641.00–2.700.0510.970.38–2.450.943TC, mmol/L (mg/dL)< 3.89 (< 150)Reference--Reference--3.89–4.39 (150–170)0.980.61–1.560.9240.860.34–2.180.7554.39–4.93 (170–191)1.220.78–1.930.3800.900.36–2.230.825≥ 4.93 (≥ 191)1.390.86–2.220.1750.650.25–1.720.386TG, mmol/L (mg/dL)< 1.28 (< 113)Reference--Reference--1.28–1.55 (113–138)1.190.73–1.930.4910.970.35–2.680.9541.55–1.89 (138–168)1.020.63–1.660.9260.900.34–2.400.833≥ 1.89 (≥ 168)1.651.05–2.600.0311.820.75–4.400.185Q1-Q4, quartile groups based on the distribution of each lipid parameterHR Hazard ratio, CI Confidence interval, RC Remnant cholesterol, LDL-C Low-density lipoprotein cholesterol, HDL-C High-density lipoprotein cholesterol, non-HDL-C non-high-density lipoprotein cholesterol, TC Total cholesterol, TG TriglycerideaMultivariate Cox regression models were additionally adjusted for recipient age, sex, BMI, surgery year, dialysis history, donor age, donor BMI, ABO compatibility, HLA mismatch count, PRA positivity, cold ischemia time, blood glucose, and acute rejectionbMultivariate Cox regression models were additionally adjusted for recipient age, sex, and acute rejectionEach model contained only 1 lipid parameter at a time Association between quartile groups of lipid levels and kidney allograft loss and all-cause mortality in the landmark cohort Q1-Q4, quartile groups based on the distribution of each lipid parameter HR Hazard ratio, CI Confidence interval, RC Remnant cholesterol, LDL-C Low-density lipoprotein cholesterol, HDL-C High-density lipoprotein cholesterol, non-HDL-C non-high-density lipoprotein cholesterol, TC Total cholesterol, TG Triglyceride aMultivariate Cox regression models were additionally adjusted for recipient age, sex, BMI, surgery year, dialysis history, donor age, donor BMI, ABO compatibility, HLA mismatch count, PRA positivity, cold ischemia time, blood glucose, and acute rejection bMultivariate Cox regression models were additionally adjusted for recipient age, sex, and acute rejection Each model contained only 1 lipid parameter at a time In sensitivity analyses using clinically defined thresholds, elevated RC (≥ 0.8 mmol/L; HR 1.64, 95% CI: 1.06–2.54) and non-HDL-C (≥ 3.4 mmol/L; HR 1.62, 95% CI: 1.01–2.60) were associated with kidney allograft loss, whereas TG was not significantly associated with this outcome (HR 1.59, 95% CI: 0.99–2.54) (Supplementary Table S2). Subgroup analysis Across prespecified subgroups, the relationship between RC and kidney allograft loss showed no substantial heterogeneity (Fig. 4). For TG, HLA mismatch showed a significant interaction (P for interaction = 0.012), appearing more pronounced in recipients with fewer mismatches (< 3). A borderline interaction was noted for recipient age (P for interaction = 0.076), with a greater association among younger recipients (< 35 years). Subgroup Analyses of the Associations of RC and TG with Kidney Allograft Loss. The forest plot presents the results of subgroup analyses of the associations of remnant cholesterol (RC) and triglycerides (TG) with kidney allograft loss. RC and TG were entered into the multivariable Cox proportional hazards models as continuous variables, and stratification analyses were performed according to prespecified clinical characteristics. All models were adjusted for recipient age, sex, BMI, surgery year, donor age, donor BMI, HLA mismatch count, cold ischemia time, blood glucose, and acute rejection. Stratification variables were not additionally adjusted within their respective subgroups. The right-side column presents the hazard ratios (HRs) and their 95% confidence intervals (CIs) for each subgroup. The P values for interaction were calculated by adding the lipid parameter, the stratification variable, and their interaction term to the Cox model. The vertical line in the figure indicates HR = 1. BMI, body mass index; HLA, human leukocyte antigen; eGFR, estimated glomerular filtration rate Incremental contribution analysis As shown in Table 4, the baseline model yielded a concordance index (C-index) of 0.680 (95% CI: 0.631–0.730). Addition of RC modestly improved discrimination (ΔC-index = 0.008, P = 0.007) and model fit, change in Akaike information criterion (ΔAIC = − 4.78). TG also improved model fit (LR test P = 0.018; ΔAIC = − 3.61), though the change in C-index was minimal. Non-HDL-C did not significantly improve model discrimination (ΔC-index = 0.004, P = 0.069), although a modest improvement in model fit was observed (ΔAIC = − 0.86). LDL-C, HDL-C, and TC did not meaningfully improve the model. Table 4Incremental contribution of lipid parameters to the model for kidney allograft loss in the landmark cohortVariableC-index (95% CI)ΔC-indexP for ΔC-indexLR Test PWald Test PΔAICΔBICBaseline model0.680 (0.631–0.730)------+ RC0.688 (0.639–0.737)0.0080.0070.0090.007−4.78−1.73+ LDL-C0.683 (0.634–0.732)0.0030.0970.2400.2370.623.67+ HDL-C0.683 (0.634–0.732)0.0020.9780.4790.4821.504.55+ non-HDL-C0.684 (0.635–0.733)0.0040.0690.0910.088−0.862.19+ TC0.682 (0.632–0.731)0.0010.0680.1750.1720.163.21+ TG0.681 (0.632–0.730)0.0000.3910.0180.013−3.61−0.56The baseline multivariate Cox regression model was adjusted for recipient age, sex, BMI, surgery year, dialysis history, donor age, donor BMI, ABO compatibility, HLA mismatch count, PRA positivity, cold ischemia time, blood glucose, and acute rejection. ΔC-index: change in C-index vs. baseline. P for ΔC-index: significance of the change (based on Dxy comparison). Likelihood ratio (LR) test P and Wald test P: significance of each added lipid parameter. ΔAIC/ΔBIC: change in Akaike information criterion (AIC) and Bayesian information criterion (BIC) (negative = better fit). Each lipid was added individuallyHR Hazard ratio, CI Confidence interval, RC Remnant cholesterol, LDL-C Low-density lipoprotein cholesterol, HDL-C High-density lipoprotein cholesterol, non-HDL-C non-high-density lipoprotein cholesterol, TC Total cholesterol, TG Triglyceride Incremental contribution of lipid parameters to the model for kidney allograft loss in the landmark cohort The baseline multivariate Cox regression model was adjusted for recipient age, sex, BMI, surgery year, dialysis history, donor age, donor BMI, ABO compatibility, HLA mismatch count, PRA positivity, cold ischemia time, blood glucose, and acute rejection. ΔC-index: change in C-index vs. baseline. P for ΔC-index: significance of the change (based on Dxy comparison). Likelihood ratio (LR) test P and Wald test P: significance of each added lipid parameter. ΔAIC/ΔBIC: change in Akaike information criterion (AIC) and Bayesian information criterion (BIC) (negative = better fit). Each lipid was added individually HR Hazard ratio, CI Confidence interval, RC Remnant cholesterol, LDL-C Low-density lipoprotein cholesterol, HDL-C High-density lipoprotein cholesterol, non-HDL-C non-high-density lipoprotein cholesterol, TC Total cholesterol, TG Triglyceride Sensitivity analysis In analyses of DCGF, significant associations were observed for RC (HR 2.77, P = 0.003), non-HDL-C (HR 1.31, P = 0.031), TC (HR 1.27, P = 0.044), and TG (HR 1.40, P = 0.016) (Supplementary Table S3). Similarly, RCS analyses showed a positive association between RC and DCGF (P for overall = 0.010; Supplementary Figure S3). The incremental contribution of RC to the DCGF model was consistent with findings in the primary analysis (Supplementary Table S4). In analyses using mean lipid levels during the entire follow-up period, 2,284 recipients with available follow-up lipid measurements were included (Supplementary Table S5). Elevated RC (HR 3.39, P < 0.001), HDL-C (HR 0.48, P = 0.027), and TG (HR 1.41, P < 0.001) were significantly associated with kidney allograft loss (Supplementary Table S6). For all-cause mortality, both HDL-C and TC showed inverse associations (P < 0.001 and P = 0.022, respectively) (Supplementary Table S6). Main results from participants with complete data were consistent with the primary findings (Supplementary Tables S7 and S8). Discussion In this retrospective cohort study of kidney transplant recipients, higher post-transplant RC levels were independently associated with an increased risk of kidney allograft loss. TG also showed a significant association, whereas LDL-C and TC were not associated with graft outcomes. Non-HDL-C did not reach statistical significance in the primary analysis, although significant associations were observed in sensitivity analyses. None of the lipid parameters were significantly associated with all-cause mortality in the primary analysis. The association between TG and allograft loss was greater among recipients with fewer HLA mismatches. Additionally, incorporating RC into standard risk models modestly improved model performance. The main finding of this study was the independent association between higher RC levels and kidney allograft loss. Quartile comparison also suggested that elevated RC may reflect a broader state of metabolic disturbance. This observation is consistent with prior evidence linking RC to adverse clinical outcomes. In kidney transplant recipients, the Folic Acid for Vascular Outcome Reduction in Transplantation (FAVORIT) trial, which included 4,110 participants, reported that the highest RC quartile was associated with higher CVD event risks and all-cause mortality than the lowest quartile. A trend toward increased graft failure (approximately 20%) was also noted, although the association was not statistically significant [11]. Furthermore, elevated RC levels have been linked to declining renal function in other populations. RC has been shown to correlate with urinary albumin-to-creatinine ratio [30]. Among patients with type 2 diabetes (T2D)-related chronic kidney disease (CKD), each standard deviation (SD) increment in RC corresponded to a 24% higher risk of ESRD progression [31]. The underlying pathophysiology may involve several interrelated biological mechanisms. RC is thought to contribute to atherosclerosis through the subendothelial accumulation of triglyceride-rich lipoproteins, which may impair renal hemodynamics [31]. Beyond this direct vascular effect, RC has also been linked to the progression of renal injury through systemic inflammation [32, 33]. RC may also be associated with endothelial dysfunction, potentially through pro-inflammatory activation and foam cell formation, thereby contributing to glomerular and microvascular injury [31]. Dyslipidemia is highly prevalent after kidney transplantation, largely attributable to the use of immunosuppressive medications [5, 34]. In this study, TG levels were associated with increased kidney allograft loss risk. This result is consistent with findings from previous studies linking TG and related indices, such as the TG/HDL-C ratio and the triglyceride-glucose (TyG) index, to renal function decline and graft loss in kidney transplant recipients [35, 36]. Another study found that recipients with two or more metabolic abnormalities (hypertension, hyperglycemia, hypertriglyceridemia, or low HDL-C) post-transplant had a significantly increased risk of allograft dysfunction [37]. Additionally, TG levels have been inversely associated with eGFR [38]. Histological studies have also linked hypertriglyceridemia to greater degrees of renal interstitial fibrosis, with possible underlying mechanisms involving lipid deposition, inflammation, endothelial dysfunction, and immune activation [36, 39]. Non-HDL-C was not significantly associated with kidney allograft loss in the primary analysis; however, associations were observed in sensitivity analyses. In prior transplant cohorts, non-HDL-C was identified as one of five key predictors of post-transplant renal dysfunction [40], and non-HDL-C ≥ 4.2 mmol/L was a significant risk factor for eGFR decline among recipients with stable renal function [41]. Elevated non-HDL-C levels have also been linked to adverse cardiovascular outcomes in this population [17]. These findings suggest that non-HDL-C may capture metabolic risk after transplantation, although its role in graft outcomes requires further clarification. However, associations between continuous LDL-C, HDL-C, and TC levels and long-term outcomes were not consistently observed in the present cohort. Evidence from earlier studies aligns with this observation, although some reports have linked post-transplant lipid levels to graft outcomes [34, 40]. One study found no significant associations between elevated conventional lipid parameters and mortality or allograft loss in kidney transplant recipients [9], and a prospective cohort similarly reported no independent association with renal function decline in patients with CKD [42]. Moreover, substantial residual cardiovascular risk may persist despite intensive LDL-C lowering [16]. Statin therapy is commonly used for lipid management, effectively lowering TC and LDL-C and reducing cardiovascular events and mortality [43]. However, studies examining their effect on graft survival have yielded conflicting results, with some reporting no significant benefit [44, 45] and others suggesting a potential protective effect with early post-transplant initiation [46]. In this study, data on statin use were unavailable. Given that Kidney Disease Improving Global Outcomes (KDIGO) guidelines recommend statins for most recipients [47], the lack of adjustment for lipid-lowering therapy represents an important limitation. Nevertheless, these findings represent the associations observed in a clinical setting where statin therapy was prevalent, although its specific influence could not be assessed. No significant association was found between any lipid parameter and all-cause mortality. Previous studies have reported inconsistent findings in kidney transplant recipients: some found no associations with conventional lipids [9], whereas others suggested that higher HDL-C may confer a survival benefit [48] and that non-traditional parameters such as RC and the TyG index have been linked to mortality [11, 49]. Although dyslipidemia contributes to cardiovascular risk and CVD is the prime cause of death after transplantation [5], cause-specific death data were not available in this study. Deaths from infection, rejection, or malignancy were common [50, 51], which may have attenuated the observed relationship between lipid parameters and all-cause mortality. Additionally, the limited number of death events may have resulted in insufficient statistical power and model instability. Subgroup analyses showed a stronger TG–allograft loss association among recipients with fewer HLA mismatches. HLA mismatches are a primary driver of alloimmune rejection and have been associated with poorer graft survival [52, 53]. Thus, in recipients with greater HLA mismatches, allograft loss may be driven predominantly by immunologic mechanisms, potentially obscuring the contribution of metabolic factors. In contrast, among recipients with fewer HLA mismatches, who typically have lower immunologic risk and require less intensive immunosuppression, metabolic disturbances may have a greater relative influence on long-term graft outcomes. In addition, the TG–allograft loss association was also borderline stronger among younger recipients. In younger recipients, who typically mount stronger immune responses [54], elevated TG levels may interact with immunosuppressive therapy and alloimmune processes, thereby contributing to allograft injury. Older recipients have a higher burden of comorbidities, including T2D, hypertension, and DGF [55, 56]. In the context of this complex comorbidity burden, the specific metabolic contribution of TG to allograft loss may be obscured. Furthermore, older recipients are more susceptible to infections and immunosuppression-related complications [57], which act as competing events and may further attenuate the observed association. The addition of RC was associated with the greatest improvement in model performance among the lipid parameters examined. RC is derived primarily from triglyceride-rich remnant lipoproteins, which can penetrate the vascular endothelium and accumulate within the arterial wall, thereby promoting inflammation, oxidative stress, and atherosclerosis [58, 59]. Consequently, RC may better reflect metabolic abnormalities underlying renal microvascular injury and chronic allograft dysfunction. In individuals with normal renal function, RC has shown stronger associations with early kidney injury progression than conventional lipid parameters [14]. A large population-based study in China indicated that RC demonstrated better discrimination for prediabetes than traditional lipid parameters [60]. Similarly, RC has outperformed conventional lipids in identifying cardiometabolic risk and non-alcoholic fatty liver disease [12, 13]. Notably, even among individuals with LDL-C levels within recommended targets or with normal conventional lipid profiles, higher RC levels have been linked to increased metabolic and CVD risk [13, 61], suggesting that RC may reflect residual risk not captured by traditional lipid parameters and could serve as a marker for evaluating novel lipid-lowering therapies. Strengths and limitations This study systematically examined the associations of multiple lipid parameters with long-term outcomes after kidney transplantation. A methodological strength was the use of a Landmark analysis at 365 days post-transplant, which established a clear temporal sequence between lipid exposure and outcome events and reduced the potential for time-related bias. In addition, using first-year mean lipid values, rather than single baseline measurements, may reduce the influence of short-term fluctuations and yield a more stable estimate of early post-transplant lipid exposure. The use of complementary performance metrics allowed for a comprehensive assessment of model performance. The incremental contribution of RC was consistently observed across these metrics, underscoring its potential relevance in this population. To enhance clinical applicability, analyses were also performed using clinically defined lipid categories, which may facilitate interpretation in routine practice. Furthermore, the relatively large sample size contributed to greater statistical stability. The findings of this study should be interpreted in light of several limitations. First, its retrospective design may be subject to information bias and cannot establish causality. Although adjustments were made for multiple recipient- and donor-related characteristics, residual confounding cannot be entirely excluded, given the observational nature of the study. Second, information on lipid-lowering therapy was unavailable; therefore, the analyses could not account for the potential influence of such therapies. Third, the single-center design and relatively racially homogeneous population may limit generalizability; thus, extrapolation to other populations should be undertaken with caution. Fourth, the small number of deaths may have limited statistical power for mortality analyses. In addition, for recipients with long follow-up, the ascertainment of the specific causes of death and allograft loss was challenging; therefore, cause-specific analyses were not performed. Fifth, lipid levels were assessed only within the first year after transplantation, and longitudinal changes in lipid levels were not fully captured. Finally, RC was calculated rather than directly measured. However, calculated RC has been shown to correlate closely with measured values and offers a practical, cost-free approach for clinical application [62]. Conclusions In conclusion, elevated RC and TG levels were independently associated with kidney allograft loss in living donor kidney transplant recipients. RC showed the most consistent association with graft outcomes and provided modest incremental predictive value beyond established risk factors. As a marker derived from routine lipid profiles, RC may serve as an adjunctive marker for identifying recipients at higher metabolic and allograft risk. These findings support further evaluation of RC as part of post-transplant lipid assessment to improve long-term post-transplant care and graft preservation. Further studies are needed to clarify whether incorporating RC into post-transplant risk assessment or targeting remnant lipoprotein metabolism could improve long-term graft outcomes. Supplementary Information Supplementary Material 1. Acknowledgements Not applicable. Declaration of generative AI and AI-assisted technologies During the preparation of this work, the authors used ChatGPT and DeepSeek to improve language. After using this tool, the authors reviewed and edited the content as needed and took full responsibility for the content of the publication. Abbreviations Akaike Information Criterion Acute Rejection Body mass index Confidence Intervals Concordance Index Chronic Kidney Disease Cardiovascular Disease Death-censored graft failure Delayed Graft Function Estimated Glomerular Filtration Rate End-Stage Renal Disease Folic Acid for Vascular Outcome Reduction in Transplantation High-Density Lipoprotein Cholesterol Hospital Information System Human Leukocyte Antigen Hazard Ratio Kidney Disease Improving Global Outcomes Low-Density Lipoprotein Cholesterol Likelihood Ratio Multiple Imputation by Chained Equations Non-High-Density Lipoprotein Cholesterol Panel Reactive Antibody Remnant Cholesterol Restricted Cubic Splines Standard Deviation Total Cholesterol Triglycerides Triglyceride-Glucose Authors’ contribution Jingru Chen and Xinyi Gao are the co-first authors. Jingru Chen contributed to the conceptualization, data curation, formal analysis, methodology, visualization, and draft writing. Xinyi Gao contributed to the data curation, methodology, visualization, and reviewing and editing of the manuscript. Xinxuan Meng contributed to the methodology and data curation. Xiaowei Lou contributed to the methodology. Jianghua Chen contributed to the conceptualization, methodology, funding acquisition, supervision, and reviewing and editing of the manuscript. All authors contributed to the critical revision of the final manuscript and approved the final version of the manuscript. Funding Not applicable. Data availability The datasets generated and/or analysed during the current study are not publicly available due to privacy and ethical concerns but are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate The Clinical Research Ethics Committee of the First Affiliated Hospital, Zhejiang University School of Medicine, approved the study with a waiver of informed consent. All data were deidentified. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Jingru Chen and Xinyi Gao are the co-first authors. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The datasets generated and/or analysed during the current study are not publicly available due to privacy and ethical concerns but are available from the corresponding author on reasonable request.