Bile acids as a predictor for all-cause mortality in peritoneal dialysis patients: a retrospective cohort study

Abstract

Background

Bile acids (BAs) are now recognized as endocrine hormones involved in glucose and lipid metabolism regulation and inflammation control. Recent studies have also shown the association of BAs level in chronic kidney disease progression and hemodialysis patients’ prognosis. However, their role in peritoneal dialysis (PD) is unknown. Therefore, based on a longitudinal cohort, this study aims to explore the relationship of BAs and all-cause mortality in PD patients.

Methods

A retrospective cohort study was conducted involving 517 PD patients. Serum BAs level prior to PD catheterization were collected. Restricted cubic spline (RCS) analysis with full adjustment was used to characterize the dose-response relationship between serum BAs and all-cause mortality. Cox proportional hazards regression models were performed to assess the independent association between BAs levels and all-cause mortality.

Results

A total of 517 patients were included. During a median follow-up of 3.65 (1.36, 5.46) years, death occurred in 210 of 517 (40.6%) patients. The relationship between BAs levels and the risk of all-cause mortality was fitted using multivariate RCS adjusted for clinical relevant factors showing that BAs levels were linearly associated with all-cause mortality (p for non-linearity = 0.589). Further Kaplan-Meier survival analysis revealed a significantly higher all-cause mortality risk in the high BAs group (log rank score 13.04, p < 0.001). Cox regression showed that elevated BAs levels were an independent risk factor for all-cause mortality in PD patients (HR = 1.056, 95% CI: 1.005–1.109, p = 0.031) after adjusting for confounding factors.

Conclusion

Elevated pre-dialysis serum BAs levels are independently associated with increased all-cause mortality in PD patients. Serum BAs may represent a potential prognostic indicator, although further validation is required before clinical application.

Introduction

Peritoneal dialysis (PD) accounts for 11% of all dialysis treatments worldwide [1, 2]. Long-term PD use is associated with glucose and lipid metabolic abnormalities, which contribute to adverse clinical outcome [3, 4]. Bile acids (BAs), traditionally known for their role in lipid digestion, serve as sensitive indicators of hepatocyte function and integrity, and are used to assess cholestasis [5]. They are now recognized as endocrine hormones involved in glucose and lipid metabolism regulation and inflammation control [6–8]. BAs could improve insulin sensitivity and promote high density lipoprotein function while reducing very low-density lipoprotein (VLDL) levels. These metabolic benefits support its investigation as a potential plasma biomarker. Recent studies have also shown the association of BAs with Chronic Kidney Disease (CKD) disease progression and clinical prognosis [9–11].

However, the association of altered BAs levels and outcome of dialysis patients is unclear. A recent study in hemodialysis (HD) patients demonstrated that lower BAs levels were associated with lower lipid levels in patients with diabetes mellitus (DM) on HD and could be an independent risk factor for all-cause death in patients with DM on HD [9]. The observed correlation warrants further investigation into the role of BAs in PD outcome. We chose to measure total serum BAs because it represents a routine, widely accessible, and cost-effective clinical biomarker for real-world prognostic evaluation. In contrast, individual BA subtypes require specialized metabolomic platforms that are not routinely available in clinical practice for risk stratification of PD patients.

Therefore, in the present study, based on a longitudinal cohort, this study aims to explore the relationship of BAs and all-cause mortality in PD patients.

Methods

Study population

In this retrospective cohort study, all incident patients who started PD therapy from 1 July 2010 to 12 April 2024 from Peking University Third Hospital were included. Patients were excluded if they had age less than seventeen years or if they had malignant tumor, hepatobiliary diseases (active hepatitis, liver cirrhosis and cholecystolithiasis), or without pre-dialysis BAs values. Patients were followed until death, cessation of PD or end of study as of 25 September 2024.

In our center, pre-dialysis patients need to be hospitalized to department of nephrology for PD catheter implantation. Therefore, patients’ detailed comorbidities, demographics, laboratory parameters were recorded when admitted. During the study period, patients’ hospitalization electronic medical charts were carefully reviewed to obtain the data, Serum BAs level prior to PD catheterization were collected. Serum BAs were measured using a standardized enzymatic colorimetric assay on a fully automated biochemical analyzer. Other assay indictors including hemoglobin, albumin, creatinine, calcium, phosphorus, total cholesterol, triglycerides, high-density lipoprotein cholesterol (HDL-c) and low-density lipoprotein cholesterol (LDL-c) were also obtained. All blood samples were collected after an overnight fasting of at least 8 h to ensure standardization and consistency of BAs levels. The estimated glomerular filtration rate (eGFR) was estimated by Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) creatinine equation [12]. All baseline data, including serum BAs and other laboratory parameters, were collected at admission and prior to PD catheter insertion and PD initiation. Therefore, PD‑related parameters such as Kt/V, peritoneal transport status, peritonitis rate, and residual kidney function during PD treatment were not available at baseline and could not be included in the primary analysis.

The primary outcomes examined were all-cause mortality on PD (censored for loss to follow-up, renal transplantation, recovery of renal function, transferred to HD, transfer to other dialysis centers and end of study).

This study has been approved by the Ethics Committee of Peking University Third Hospital (No. 096 − 03). The research protocol conforms to the ethical standards of the 1964 Declaration of Helsinki and its later amendments.

Statistical analysis

Results were expressed as frequencies and percentages for categorical variables, mean ± standard deviation for continuous normally distributed variables, and median (interquartile range, IQR) for continuous variables that were not normally distributed. Comparison for patients in different groups was performed using chi-square tests, two-tailed unpaired t-tests or Mann-Whitney’s tests, depending on data distribution. Missing values for baseline variables were less than 5% and regarded as negligible. No imputation was performed for missing data, and all statistical analyses were conducted using the available complete case data only.

Restricted cubic spline (RCS) curve with 3 knots was displayed to test nonlinearity when estimating all-cause mortality [13].We constructed multivariable RCS models by a full adjusted model. A likelihood ratio test was performed to compare the model with a linear term to the model with the full spline term. During the likelihood ratio test, if BAs levels were linearly associated with all-cause mortality (p > 0.05), BAs level is incorporated as continuous variables in multivariable Cox regression models, otherwise, they would be utilized as categorical variable in the multivariable Cox regression model. Kaplan-Meier survival curves were used to evaluate the survival curves of all-cause mortality, and the Log-rank test was applied for the comparison. For the multivariable Cox proportional hazards regression model, covariates were selected based on clinical relevance as established prognostic factors in PD patients and variables with p < 0.05 in univariate analysis. The proportional hazards assumption was verified, and multicollinearity was assessed using the variance inflation factor (VIF). A competing risk analysis was performed using the Fine–Gray subdistribution hazards model to evaluate the association between baseline serum BAs and all-cause mortality. Transfer to hemodialysis and renal transplantation were treated as competing events. Subdistribution hazard ratios (SHRs) with 95% confidence intervals (CIs) were estimated.

Statistical analysis was performed using IBM SPSS software, version 22.0 (Armonk, NY) and R (version 4.1.3). A p-value < 0.05 was considered statistically significant.

Results

Baseline clinical characteristics of the study population and all-cause mortality

A total of 517 patients were included in the study (Fig. 1). During a median follow-up of 3.65 (1.36, 5.46) years, death occurred in 210 of 517 (40.6%) PD patients. 38 patients (7.4%) transferred to HD, 46 patients (8.9%) underwent renal transplantation. Loss to follow up occurred in 26 patients (5.0%) and 34 patients (6.6%) transferred to other dialysis centers.

Fig. 1.

Flowchart of this study

All 517 PD patients were stratified into two groups according to the median level of BAs: low BAs group (BAs ≤ 2.20µmol/L) and high BAs group (BAs > 2.20µmol/L). The baseline clinical characteristics across the two BAs subgroups are summarized in Table 1. Compared with the low BAs group, patients in the high BAs group had significantly lower levels of triglycerides (p < 0.001), total cholesterol (p = 0.012), LDL-c (p = 0.012), and serum phosphorus (p = 0.003), along with a notably higher eGFR (p = 0.033). With respect to comorbidity profiles, the high BAs group presented with a higher prevalence of coronary heart disease (CHD) (p = 0.008) yet a lower prevalence of hypertension (p = 0.016) relative to the low BAs group. There were no substantial missing data for key baseline variables, and complete case analysis was applied without data imputation.

Table 1.

Baseline clinical characteristic of all patients at the initiation of PD

Variables	All(n=517)	Low BAs(n=262)	High BAs(n=255)	p -value
Age, years	60.9 (46.7-73.6)	56.9 (45.4-70.2)	64.4 (49.8-75.4)	0.002
Male, n (%)	301 (58.2)	141 (53.8)	160 (62.7)	0.049
BAs, µmol/L	2.20 (1.40-3.60)	1.40 (1.00-1.80)	3.60 (2.85-5.05)	<0.001
BMI, kg/m2	23.51 (21.22-25.94)	23.33 (21.23-26.14)	23.67 (21.22-25.85)	0.933
Hypertension, n (%)	353 (68.3)	191 (72.9)	162 (63.5)	0.016
CHD, n (%)	125 (24.2)	50 (19.1)	75 (29.4)	0.008
CVD, n (%)	75 (14.5)	32 (12.2)	43 (16.9)	0.180
DM, n (%)	287 (55.5)	145 (55.3)	142 (55.7)	1.000
eGFR, ml/min/1.73m2	4.76 (3.82-6.08)	4.64 (3.67-5.92)	4.96 (3.90-6.24)	0.033
Potassium, mmol/L	4.52 (4.04-5.02)	4.56 (4.12-5.10)	4.48 (3.95-4.98)	0.055
Sodium, mmol/L	139.00 (137.00-141.00)	139.00 (137.00-141.00)	139.00 (137.00-141.00)	0.289
Serum albumin, g/L	36.28 ± 5.36	36.11 ± 5.39 (32.60-40.00)	36.45 ± 5.18(32.80-40.00)	0.462
Hemoglobin, g/L	88.00 (75.00-101.00)	87.00 (73.00-100.00)	89.00 (76.00-101.00)	0.442
Triglyceride, mmol/L	1.34 (1.01-1.98)	1.46 (1.11-2.07)	1.23 (0.88-1.75)	<0.001
Total cholesterol, mmol/L	4.04 (3.42-4.98)	4.22 (3.51-5.24)	3.96 (3.32-4.75)	0.012
LDL-c, mmol/L	2.36 (1.80-3.09)	2.41 (1.87-3.25)	2.29 (1.73-2.93)	0.012
HDL-c, mmol/L	0.93 (0.78-1.18)	0.93 (0.78-1.18)	0.94 (0.79-1.18)	0.735
Serum calcium, mmol/L	2.04 (1.86-2.20)	2.04 (1.81-2.20)	2.05 (1.88-2.20)	0.352
Serum phosphorus, mmol/L	1.94 (1.55-2.36)	1.98 (1.65-2.47)	1.88 (1.50-2.24)	0.003
Death, n (%)	210 (40.6)	93 (35.4)	117 (45.9)	0.021

The relationship between BAs levels and risk of all-cause mortality

The relationship between BAs levels and the risk of all-cause mortality was fitted using multivariate RCS analysis adjusted for age, gender, DM, CHD, hypertension, cerebrovascular disease (CVD), eGFR, hemoglobin, blood glucose, serum phosphorus and serum albumin. Consequently, BAs levels were linearly and positively associated with all-cause mortality (p for non-linear = 0.589) (Fig. 2). The risk of all-cause mortality increased progressively with rising serum BAs levels, which supported the use of BAs as a continuous variable in subsequent Cox regression models.

Fig. 2.

The association between bile acids and all-cause mortality risk after full adjustment. Clinically relevant variables were included in multivariate RCS analysis, including age, sex, eGFR, hemoglobin, blood glucose, serum phosphorus, albumin, hypertension, coronary heart disease, cerebrovascular disease, and diabetes mellitus

Kaplan-Meier survival analysis demonstrated that patients in the high BAs group had a significantly lower cumulative survival rate compared with those in the low BAs group. The log-rank test confirmed a marked difference between the two groups (log-rank score = 13.04, P < 0.001) (Fig. 3), indicating that higher baseline BAs levels were associated with worse long-term survival in PD patients.

Fig. 3.

Kaplan-Meier survival curve for all-cause mortality comparing PD patients with different levels of Bile acids (BAs)

Univariate Cox regression analysis identified that BAs levels, age, hypertension, hemoglobin, serum albumin, serum phosphorus, serum potassium levels, CHD, diabetes, and CVD were all significantly associated with all-cause mortality (all p < 0.05). In the multivariate Cox proportional hazards regression model, elevated BAs levels discovered to be an independent factor for all-cause mortality in PD patients (HR = 1.056, 95% CI: 1.005–1.109, p = 0.031) after adjusting for age, serum albumin, CHD, CVD, eGFR, hemoglobin, serum phosphorus, hypertension, and DM. After accounting for competing risks of transferred to hemodialysis and transplantation, higher BA levels demonstrated a significant association with increased all-cause mortality (SHR = 1.08, 95% CI: 1.03–1.12, p < 0.001). Given the narrow physiological range of serum BAs, this per-unit HR appears modest. To improve clinical interpretability, we further stratified patients by the median BAs level (≤ 2.20 vs. > 2.20 µmol/L). Compared with the low BAs group, the high BAs group had a higher crude mortality rate (45.9% vs. 35.4%), with an absolute risk difference (ARD) of 10.5%, indicating a clinically meaningful prognostic impact. Details are presented in Table 2.

Table 2.

Cox regression analysis for all-cause mortality in PD patients

Variables		All (n = 517)	Low BAs (n = 262)		High BAs (n = 255)			p-value
Age, years		60.9 (46.7–73.6)	56.9 (45.4–70.2)			64.4 (49.8–75.4)		0.002
Male, n (%)		301 (58.2)	141 (53.8)			160 (62.7)		0.049
BAs, µmol/L		2.20 (1.40–3.60)	1.40 (1.00-1.80)			3.60 (2.85–5.05)		< 0.001
BMI, kg/m2		23.51 (21.22–25.94)	23.33 (21.23–26.14)			23.67 (21.22–25.85)		0.933
Hypertension, n (%)		353 (68.3)	191 (72.9)			162 (63.5)		0.016
CHD, n (%)		125 (24.2)	50 (19.1)			75 (29.4)		0.008
CVD, n (%)		75 (14.5)	32 (12.2)			43 (16.9)		0.180
DM, n (%)		287 (55.5)	145 (55.3)			142 (55.7)		1.000
eGFR, ml/min/1.73m2		4.76 (3.82–6.08)	4.64 (3.67–5.92)			4.96 (3.90–6.24)		0.033
Potassium, mmol/L		4.52 (4.04–5.02)	4.56 (4.12–5.10)			4.48 (3.95–4.98)		0.055
Sodium, mmol/L		139.00 (137.00-141.00)	139.00 (137.00-141.00)			139.00 (137.00-141.00)		0.289
Serum albumin, g/L		36.28 ± 5.36	36.11 ± 5.39 (32.60–40.00)			36.45 ± 5.18(32.80–40.00)		0.462
Hemoglobin, g/L		88.00 (75.00-101.00)	87.00 (73.00-100.00)			89.00 (76.00-101.00)		0.442
Triglyceride, mmol/L		1.34 (1.01–1.98)	1.46 (1.11–2.07)			1.23 (0.88–1.75)		< 0.001
Total cholesterol, mmol/L		4.04 (3.42–4.98)	4.22 (3.51–5.24)			3.96 (3.32–4.75)		0.012
LDL-c, mmol/L		2.36 (1.80–3.09)	2.41 (1.87–3.25)			2.29 (1.73–2.93)		0.012
HDL-c, mmol/L		0.93 (0.78–1.18)	0.93 (0.78–1.18)			0.94 (0.79–1.18)		0.735
Serum calcium, mmol/L		2.04 (1.86–2.20)	2.04 (1.81–2.20)			2.05 (1.88–2.20)		0.352
Serum phosphorus, mmol/L		1.94 (1.55–2.36)	1.98 (1.65–2.47)			1.88 (1.50–2.24)		0.003
Death, n (%)		210 (40.6)	93 (35.4)			117 (45.9)		0.021

Abbreviations: CI, confidence interval; HR, hazard ratio; BMI, body mass index; BAs, bile acids; LDL-c, low-density lipoprotein cholesterol; HDL-c, high-density lipoprotein cholesterol; CHD, coronary artery disease; CVD, cerebrovascular disease; eGFR, estimated glomerular filtration rate

For the multivariable Cox proportional hazards regression model, covariates were selected based on clinical relevance as established prognostic factors in PD patients and variables with p < 0.05 in univariate analysis. The proportional hazards assumption was verified, and multicollinearity was assessed using the variance inflation factor (VIF)

Discussion

In this retrospective cohort study with a large sample size and long-term follow-up, we provide evidence for an association between serum BAs levels and all-cause mortality in PD patients. Further analyses confirmed a linear relationship between BAs levels and mortality risk. Elevated serum BAs were identified as an independent risk factor for all-cause mortality in PD patients, and this association remained robust after adjustment for established prognostic variables.

In addition to higher all-cause mortality observed in the high BAs group, several baseline characteristics is observed differed between the low and high BAs groups. Patients with elevated BAs levels were older, exhibited a greater prevalence of underlying coronary heart disease, and higher eGFR; conversely, they had a lower prevalence of hypertension, lower total cholesterol, lower LDL-C, and lower serum phosphorus levels compared with those in the low BAs group. Notably, BAs remained an independent predictor of all-cause mortality after multivariable adjustment for lipids, age, comorbidities, and renal function. This indicates that the BAs-mortality association is not explained by lipid differences or other baseline factors. Consistent with existing evidence, BAs may influence prognosis through non-lipid pathways, including inflammation, oxidative stress, and signaling involved in cardiovascular and renal fibrosis.

BAs act not only as lipid emulsifiers but also as pleiotropic signaling molecules that regulate glucose and lipid metabolism via farnesoid X receptor (FXR) and G-protein-coupled bile acid receptor (TGR5) [14–18]. Dysregulated BA signaling disrupts metabolic homeostasis, promoting insulin resistance and metabolic dysfunction [19, 20]. In PD patients, elevated BAs may increase mortality risk through three key mechanisms. First, high BAs alter gut microbiota composition and impair intestinal immunity, exacerbating chronic inflammation [21, 22]. Second, dysregulated BA signaling disrupts glucose and lipid metabolism, increasing cardiovascular risk [23]. Third, excess hydrophobic BAs exert direct cytotoxicity, worsening tissue injury and chronic damage in dialysis patients.

Research investigating the prognostic impact of BAs on PD patients remains remarkably scarce in the existing literature. Only a study employing non-targeted metabolomics indicated that increased levels of cholic acid and adenine, along with decreased isoleucine, were identified as independent risk factors for poor prognosis in PD patients, suggesting that levels of cholic acid, a BAs subtype, are positively associated with the risk of all-cause mortality in PD patients [10]. In the present study, with a specific focus on circulating BAs levels, we revealed that serum BAs levels are positively associated with PD patients’ all-cause mortality. Notably, this biomarker is characterized by its ready measurability in clinical settings, which lends greater feasibility to its potential application as a prognostic indicator for PD patients.

While our findings highlight the prognostic role of BAs in PD patients, studies in other populations, including those with diabetic nephropathy and end‑stage kidney disease (ESKD), have yielded inconsistent results. Notably, our findings differ markedly from those in HD cohorts, where lower BAs were associated with higher mortality [9]. Moreover, several studies have reported that elevated BAs, particularly deoxycholic acid, predict adverse outcomes in ESKD patients [24, 25]. These discrepancies may be explained by key differences in BAs handling between dialysis modalities. BAs are highly protein‑bound; HD relies on diffusion and poorly clears protein‑bound solutes, whereas PD enables continuous, prolonged clearance of such molecules, leading to distinct systemic BAs exposure and biological effects [26]. Additionally, prior HD studies enrolled maintenance patients, while our cohort included incident PD patients with baseline BAs measured pre‑dialysis. These differences in patient population and timing of measurement further contribute to divergent findings. Further research is needed to validate these mechanisms and clarify the differential prognostic impact of BAs across dialysis modalities.

Notably, patients with higher serum BAs had significantly lower triglycerides, total cholesterol, and LDL‑c than those with lower BAs. This inverse association indicates that elevated BAs in PD patients reflect not only cholestasis but also a distinct metabolic phenotype with reduced lipemia. BAs are synthesized from cholesterol and, beyond facilitating fat absorption, act as signaling molecules that suppress hepatic lipogenesis and promote cholesterol excretion via FXR‑mediated pathways [27, 28]. In CKD, elevated BAs may further lower circulating lipids by altering gut microbiota [29], impairing lipid absorption, and exacerbating inflammation‑related hepatic lipoprotein suppression. Importantly, after multivariable adjustment including lipid parameters, elevated BAs remained independently associated with increased all‑cause mortality. This indicates that lipids are unlikely to confound the BAs–mortality relationship, and that BAs may affect prognosis through non‑lipid pathways, such as chronic inflammation, gut dysbiosis, or direct cytotoxicity [30, 31].

This study is subject to several limitations that warrant acknowledgment. First, serum BAs levels were measured only at a single time point prior to PD initiation, which may not fully represent long-term BAs exposure during the entire follow-up period. Second, only total serum BAs were measured, whereas individual BAs subtypes are known to exert distinct, even opposing, biological effects on metabolism, inflammation, and cellular function. Failure to analyze specific BAs subtypes limits mechanistic interpretation and may mask divergent prognostic impacts of individual BAs species. Third, it is a single-center retrospective study, limiting generalizability and leaving potential residual confounding. Fourth, the observational design precludes causal inference. Additionally, key PD-related parameters (Kt/V, peritoneal transport, peritonitis, residual kidney function) were unavailable for adjustment. Future multicenter, prospective studies with serial BAs measurements and detailed PD parameters are needed to validate our findings.

In conclusion, in this retrospective cohort study, elevated pre-dialysis serum BAs levels were independently associated with an increased risk of all-cause mortality in PD patients. These findings suggest that serum BAs may serve as a potential prognostic indicator, although further validation in multicenter and prospective cohorts is required before strong clinical application can be established.

Abbreviations

BAs

Bile acids

CKD

Chronic kidney disease

CKD‑EPI

Chronic Kidney Disease Epidemiology Collaboration

CVD

Cerebrovascular disease

CHD

Coronary heart disease

DM

Diabetes mellitus

ESKD

End‑stage kidney disease

eGFR

Estimated glomerular filtration rate

FXR

Farnesoid X receptor

TGR5

G‑protein‑coupled bile acid receptor

HD

Hemodialysis

HDL‑c

High‑density lipoprotein cholesterol

IQR

Interquartile range

LDL‑c

Low‑density lipoprotein cholesterol

PD

Peritoneal dialysis

RCS

Restricted cubic spline

VIF

Variance inflation factor

VLDL

Very low‑density lipoprotein

Author contributions

Wen Tang conceived the idea of this study and assisted in the data collection. Siyuan Zhao was involved in data analysis, interpretation and manuscript draft. Xiaoxiao Wang assisted in the statistical analysis. Wen Tang and Qiong Bai reviewed the manuscript and provided mentorship. All authors read and approved the final manuscript.

Funding

This work was supported by funds from the National Natural Science Foundation of China (82470774) and grant of Clinical Cohort Construction Program of Peking University Third Hospital (BYSYDL2023004) to Wen Tang and Key Clinical Program of Peking University Third Hospital under Grant No. BYSYZD2021036 to Qiong Bai.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of Peking University Third Hospital (No. 096 − 03). The study procedure was conducted in accordance with the principles laid down by the Declaration of Helsinki. Informed consent was waived by IRB given the study was a retrospective, non-interventional design and using a de-identified dataset.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.