Bile Acid Profiles Define Disease-Specific Cholestatic States in Primary Sclerosing and Primary Biliary Cholangitis Authors: Brian D. Juran, Bryan M. McCauley, Chang Hu, Ahmad H. Ali, Erik M. Schlicht, Jackie K. Bianchi, Marjana J. Schell, Elizabeth J. Atkinson, Johannes R. Hov, Mette Vesterhus, William Rosenberg, Nicholas F. LaRusso, Gregory J. Gores, Tom H. Karlsen, Konstantinos N. Lazaridis Published: 2026 ABSTRACT Background and AimsPrimary sclerosing cholangitis (PSC) and primary biliary cholangitis (PBC) are chronic cholestatic liver diseases with overlapping features but distinct biology. While bile acids (BAs) are central to cholestatic pathophysiology, the extent to which circulating BAs define disease-specific states remains unclear.MethodsPlasma BA profiling and clinical laboratory tests were performed in patients with PSC (n = 269), PBC (n = 262), and matched controls (n = 269/262), respectively. BA composition, conjugation patterns, and pool characteristics were compared within and across diseases. Associations with hepatic decompensation were evaluated using univariable and exploratory multivariable analyses, including comparison with established prognostic scores.ResultsBoth PSC and PBC exhibited elevated total BA levels compared with controls; however, BA composition and conjugation patterns differed between diseases. PSC was characterized by increased BA conjugation fraction and reduced glycine-to-taurine ratios, whereas PBC showed preserved overall conjugation with increased glycine predominance. In PSC, BA composition was further modulated by inflammatory bowel disease and colectomy, with depletion of secondary BA species. Across both diseases, advancing liver disease stage was associated with shifts toward primary and conjugated BA dominance. BA variables and clinical laboratory tests were associated with hepatic decompensation, particularly in PSC. Exploratory multivariable models showed comparable performance of BA-based and clinical laboratory–based models, and their combination did not materially improve prediction.ConclusionPlasma BA profiles define distinct cholestatic states in PSC and PBC that reflect disease type, intestinal modifiers, and liver disease stage. While BA variables are associated with clinical outcomes, their principal value lies in complementing established clinical laboratory–based assessments rather than improving predictive accuracy. Introduction Primary sclerosing cholangitis (PSC) and Primary biliary cholangitis (PBC) are chronic cholestatic liver diseases characterized by immune-mediated biliary damage that can progress to cirrhosis and end-stage liver disease. PSC, more commonly affecting men, is frequently associated with inflammatory bowel disease (IBD) and carries an increased risk of cholangiocarcinoma, with biliary injury affecting medium- and large-sized intrahepatic and extrahepatic ducts.1 In contrast, PBC predominantly affects women and is commonly associated with antimitochondrial antibodies, with biliary damage primarily affecting small intrahepatic bile ducts.2 Genome-wide association studies have highlighted immune-mediated pathways as contributors to disease susceptibility in both conditions, although limited overlap in genetic risk loci suggests important biological differences in disease pathogenesis.3,4 Impaired bile flow in cholestatic liver diseases leads to hepatic and systemic accumulation of bile acids (BAs), which are increasingly recognized as active signaling molecules rather than passive detergents.5 BAs influence hepatic metabolism, immune responses, and gut microbial homeostasis, and alterations in BA composition have been implicated in liver injury and disease progression.6 Primary BAs, cholic acid (CA), and chenodeoxycholic acid (CDCA), are synthesized in the liver and conjugated with glycine or taurine before secretion into bile. In the intestine, BAs undergo deconjugation and microbial transformation into secondary BAs, which are efficiently reabsorbed through enterohepatic circulation.7 Disruption of this tightly regulated system can lead to accumulation of hydrophobic BA species with hepatotoxic potential, and therapeutic strategies in cholestatic liver diseases have therefore focused on modulating BA metabolism and signaling.8 Advances in clinical BA profiling now allow detailed characterization of circulating BA composition and conjugation patterns. Beyond reflecting cholestasis severity, BA profiles may capture biologically meaningful features of disease state related to intestinal metabolism, enterohepatic circulation, and therapeutic exposure. Such information may complement standard clinical laboratory tests (CLTs), which primarily reflect liver injury and synthetic function but do not directly assess BA composition. We previously developed the primary sclerosing cholangitis bile acid profile (PSC-BAP) score, demonstrating that selected BA variables are associated with hepatic decompensation (HD) in PSC.9 In the present study, we extend this work by performing comprehensive plasma BA profiling in an independent PSC cohort and in patients with PBC, with the goal of defining disease-specific BA states and evaluating how these profiles relate to intestinal modifiers, liver disease stage, and established clinical risk measures. By integrating BA composition with clinical laboratory data, we aim to clarify the role of BA profiling as a complementary tool for biological disease stratification in chronic cholestatic liver disease. Methods Patients Patients and controls were selected from the PSC Scientific Community Resource, which includes individuals with PSC, PBC, and healthy controls.10 Clinical characteristics were abstracted from the Mayo Clinic Electronic Health Record and diagnoses of PSC and PBC were confirmed by a hepatologist according to the American Association for the Study of Liver Diseases guidelines.11,12 To minimize selection bias, inclusion was based on sample availability for this and other omics-scale studies, blinded to clinical phenotypes and outcomes. Controls were age- and sex-matched and had no evidence of liver disease. Liver cirrhosis was defined by Ludwig stage IV histology,13 imaging or elastography findings,14 or evidence of portal hypertension. HD was defined as ascites, variceal hemorrhage, or hepatic encephalopathy. Ursodeoxycholic acid (UDCA) use was defined using the BA profile as previously described9 (Supplementary Figure 1). All participants provided written informed consent, and the study was approved by the Mayo Clinic Institutional Review Board in accordance with the 1964 Declaration of Helsinki. Clinical Laboratory Tests and Enhanced Liver Fibrosis Score The Mayo Clinic Central Clinical Laboratories (Rochester, MN) performed CLTs, including serum alanine aminotransferase, aspartate aminotransferase, total bilirubin, sodium, albumin, alkaline phosphatase, and complete blood count with differential. Results were normalized to the upper limit of normal for each test. The enhanced liver fibrosis (ELF) score was determined as previously described.15 Bile Acid Profiling Plasma was collected in EDTA tubes under fasting conditions and stored at −80 °C until use. BA profiling was performed in the Mayo Clinic Clinical Biochemical Genetics Laboratory using the bile acid profile, serum assay. The assay quantified primary BA: CA and CDCA; secondary BA: deoxycholic acid (DCA), lithocholic acid, UDCA, and hyodeoxycholic acid; and their glyco- and tauro-conjugates. The total bile acid (TBA) value was the sum of all measured BA. BA “family” concentrations were calculated by adding the values of unconjugated and G- or T-conjugated forms (eg, CA + GCA + TCA). The conjugated fraction was defined as the sum of the conjugated forms divided by the total (eg, (GCA + TCA)/(CA + GCA + TCA)). Glycine-to-taurine (G:T) ratio was calculated by dividing the glycine conjugated form by the taurine conjugated form (eg, GCA/TCA). Ratios of CA:CDCA, CA:DCA, and CDCA:(lithocholic acid + hyodeoxycholic acid + UDCA) were also calculated. BA ratios were not calculated when denominator values were zero. Analytical performance characteristics of the BA profiling assay, including limits of quantification, are maintained by the Mayo Clinic Clinical Biochemical Genetics Laboratory as part of routine clinical validation. Statistical Analyses Continuous variables were expressed as medians (interquartile range) and compared using the Kruskal-Wallis test. Categorical variables were compared using Pearson chi-squared test. Cox proportional hazard models evaluated the univariable ability to predict 5-year HD risk. Censoring was at orthotopic liver transplantation, cholangiocarcinoma diagnosis, last clinical encounter, or death. Hazard ratios of BA variables were calculated using interquartile range–normalized values, reflecting the risk of HD for individuals at the 75th relative to the 25th percentile. Harrell concordance statistic (c-stat) measured discrimination in univariate analysis. Correlations between BA and CLT values were assessed using Spearman rank coefficient. Gradient boosting machine (GBM) models were used to predict 5-year HD, with discrimination assessed using the c-stat and 95% confidence intervals generated by bootstrapping. Calibration was evaluated by comparing observed and expected outcomes across tertiles of model-predicted risk. Results Patient Characteristics Clinical characteristics of patients with PSC, PBC, and matched controls are summarized in Table 1. Patients and controls were well matched for age and sex within each disease group. As expected, PBC patients were predominantly female, whereas PSC patients included a higher proportion of males and frequently had concomitant IBD.Table 1Characteristics of Patients and ControlsCharacteristicPSCControlP valuePBCControlP valuen269269-262262-Age (yrs), Median (IQR)54.9 (41.3–65.6)54.9 (41.0–65.8).935266.2 (60.4–71.4)66.2 (60.7–71.5).9963Female sex, n (%)121 (45.0)121 (45.0)1.0000234 (89.3)234 (89.3)1.0000White race, n (%)249 (92.6)258 (95.9).0959251 (95.8)254 (96.9).4833Age diagnosis (yrs), Median (IQR)46.6 (33.9–56.0)--56.2 (48.8–62.6)--Disease durationa (yrs), Median (IQR)5.8 (1.9–11.7)--8.4 (2.4–16.4)--Cirrhotic at sample---- Compensated, n (%)53 (19.7)35 (13.4) Decompensated, n (%)26 (9.7)18 (6.9)Cirrhosis in follow-up---- Compensated, n (%)12 (4.5)10 (3.8) Decompensated, n (%)46 (17.1)13 (5.0)OLT in follow-up, n (%)40 (14.9)--8 (3.1)--UDCA treatment Yes, n (%)89 (33.1)--241 (92.0)-- No, n (%)162 (60.2)--14 (5.3)-- Undefined, n (%)18 (6.7)--7 (2.7)--AMA positive PBC, n (%)---217 (85.8)--Small duct PSC, n (%)14 (5.2)-----PSC with IBD, n (%)215 (79.9)-----Colectomy at sample, n (%)66 (24.5)-----Colectomy in follow-up, n (%)10 (3.7)-----CCA at sample, n (%)17 (6.3)-----CCA in follow-up, n (%)7 (2.6)-----AMA, antimitochondrial antibody; CCA, cholangiocarcinoma; IBD, inflammatory bowel disease; IQR, interquartile range; OLT, orthotopic liver transplantation; UDCA, ursodeoxycholic acid.aDisease duration of PSC or PBC at time of sample collection. Categorical variables compared using Fisher’s exact test, continuous variables compared using Kruskal-Wallis test. Table 1Characteristics of Patients and ControlsCharacteristicPSCControlP valuePBCControlP valuen269269-262262-Age (yrs), Median (IQR)54.9 (41.3–65.6)54.9 (41.0–65.8).935266.2 (60.4–71.4)66.2 (60.7–71.5).9963Female sex, n (%)121 (45.0)121 (45.0)1.0000234 (89.3)234 (89.3)1.0000White race, n (%)249 (92.6)258 (95.9).0959251 (95.8)254 (96.9).4833Age diagnosis (yrs), Median (IQR)46.6 (33.9–56.0)--56.2 (48.8–62.6)--Disease durationa (yrs), Median (IQR)5.8 (1.9–11.7)--8.4 (2.4–16.4)--Cirrhotic at sample---- Compensated, n (%)53 (19.7)35 (13.4) Decompensated, n (%)26 (9.7)18 (6.9)Cirrhosis in follow-up---- Compensated, n (%)12 (4.5)10 (3.8) Decompensated, n (%)46 (17.1)13 (5.0)OLT in follow-up, n (%)40 (14.9)--8 (3.1)--UDCA treatment Yes, n (%)89 (33.1)--241 (92.0)-- No, n (%)162 (60.2)--14 (5.3)-- Undefined, n (%)18 (6.7)--7 (2.7)--AMA positive PBC, n (%)---217 (85.8)--Small duct PSC, n (%)14 (5.2)-----PSC with IBD, n (%)215 (79.9)-----Colectomy at sample, n (%)66 (24.5)-----Colectomy in follow-up, n (%)10 (3.7)-----CCA at sample, n (%)17 (6.3)-----CCA in follow-up, n (%)7 (2.6)-----AMA, antimitochondrial antibody; CCA, cholangiocarcinoma; IBD, inflammatory bowel disease; IQR, interquartile range; OLT, orthotopic liver transplantation; UDCA, ursodeoxycholic acid.aDisease duration of PSC or PBC at time of sample collection. Categorical variables compared using Fisher’s exact test, continuous variables compared using Kruskal-Wallis test. Characteristics of Patients and Controls AMA, antimitochondrial antibody; CCA, cholangiocarcinoma; IBD, inflammatory bowel disease; IQR, interquartile range; OLT, orthotopic liver transplantation; UDCA, ursodeoxycholic acid. Disease duration of PSC or PBC at time of sample collection. Categorical variables compared using Fisher’s exact test, continuous variables compared using Kruskal-Wallis test. At the time of plasma sample collection, approximately one-fifth of PBC patients and nearly one-third of PSC patients had established cirrhosis, with both compensated and decompensated disease represented. PSC patients exhibited substantial clinical heterogeneity, including variability in disease duration, presence of IBD, and history of colectomy. The majority of PBC patients were positive for antimitochondrial antibodies, consistent with established diagnostic criteria. Follow-up data were available for assessment of HD and liver transplantation. During follow-up, HD events occurred more frequently in PSC than in PBC, reflecting differences in disease course and event rates between the 2 conditions. These outcomes were considered in subsequent analyses evaluating associations between BA profiles, CLTs, and disease progression. Standard CLTs demonstrated expected differences between patients and controls and between disease groups, consistent with chronic cholestatic liver disease (Supplementary Tables 1 and 2). As anticipated, markers of hepatic synthetic function, cholestasis, and portal hypertension differed by disease status and severity. These clinical laboratory measures were considered alongside bile acid variables in subsequent association and multivariable analyses. Plasma Bile Acid Profiles Define Disease-Specific Cholestatic States in Primary Sclerosing Cholangitis and Primary Biliary Cholangitis We first compared plasma BA concentrations and compositional features between patients with PSC or PBC and their respective age- and sex-matched controls (Figure 1; Supplementary Tables 1 and 2). Both disease groups exhibited significantly elevated TBA concentrations relative to controls, consistent with impaired bile flow in chronic cholestatic liver disease. In both PSC and PBC, increases were most pronounced for the primary BAs CA and CDCA, as well as for UDCA, reflecting both disease-related cholestasis and therapeutic exposure.Figure 1Plasma bile acid profiles define disease-specific cholestatic states in PSC and PBC. Plasma bile acid (BA) concentrations, conjugation fraction, and glycine-to-taurine (G:T) conjugation ratios are shown for patients with primary sclerosing cholangitis (PSC), primary biliary cholangitis (PBC), and their respective matched controls. Both diseases demonstrate elevated total bile acid (TBA) levels compared with controls; however, BA composition and conjugation patterns differ markedly between PSC and PBC. PSC is characterized by a globally increased conjugated fraction and reduced G:T ratios across BA families, whereas PBC shows preserved overall conjugation with increased glycine predominance. Statistical comparisons were performed using the Kruskal–Wallis test; exact P values are shown. Plasma bile acid profiles define disease-specific cholestatic states in PSC and PBC. Plasma bile acid (BA) concentrations, conjugation fraction, and glycine-to-taurine (G:T) conjugation ratios are shown for patients with primary sclerosing cholangitis (PSC), primary biliary cholangitis (PBC), and their respective matched controls. Both diseases demonstrate elevated total bile acid (TBA) levels compared with controls; however, BA composition and conjugation patterns differ markedly between PSC and PBC. PSC is characterized by a globally increased conjugated fraction and reduced G:T ratios across BA families, whereas PBC shows preserved overall conjugation with increased glycine predominance. Statistical comparisons were performed using the Kruskal–Wallis test; exact P values are shown. Beyond absolute BA concentrations, marked disease-specific differences were observed in BA conjugation and composition. PSC patients demonstrated a significantly higher conjugated fraction of TBA compared with controls, driven by increased conjugation across all major BA families (Figure 1). In PBC the overall conjugated fraction of TBA did not differ significantly from controls, although individual BA families showed heterogeneous changes in conjugation. This generalized increase in conjugation distinguished PSC from both PBC and healthy controls and was not explained by differences in total BA levels alone. Striking differences between diseases were also evident in G:T conjugation ratios. PBC patients exhibited significantly increased G:T ratios for TBA and multiple BA families compared with controls, whereas PSC patients showed significantly reduced G:T ratios (Figure 1). These opposing shifts indicate fundamentally different patterns of BA conjugation in the 2 diseases and suggest that BA pool composition captures disease-specific biological states rather than reflecting cholestasis severity alone. Consistent with these findings, the relative composition of the plasma BA pool differed between PSC and PBC. PSC patients exhibited increased proportions of CA and UDCA with reduced secondary BAs, particularly DCA, compared with controls whereas, PBC patients demonstrated an increased proportion of UDCA with relative reductions in other BA species. When UDCA was excluded from compositional analyses to account for therapeutic exposure, the remaining BA pool in PBC more closely resembled that observed in PSC, characterized by relative enrichment of primary BA species (Figure 2).Figure 2BA pool composition highlights disease-specific enrichment of primary BA species in PSC and relative UDCA enrichment in PBC, shown with and without UDCA included. These features extend beyond total bile acid (BA) levels and define distinct cholestatic states. BA pool composition highlights disease-specific enrichment of primary BA species in PSC and relative UDCA enrichment in PBC, shown with and without UDCA included. These features extend beyond total bile acid (BA) levels and define distinct cholestatic states. Together, these results demonstrate that plasma BA profiles encode distinct, disease-specific cholestatic states in PSC and PBC. Differences in conjugation fraction, G:T ratios, and BA pool composition extend beyond changes in total BA levels and are not captured by routine CLTs, supporting the use of BA profiling as a complementary approach for biological disease stratification in chronic cholestatic liver diseases. Altered Bile Acid Composition in Primary Sclerosing Cholangitis Is Modulated by Intestinal Disease and Colectomy Given the close physiological coupling between the intestine and BA metabolism, we next examined whether concurrent IBD and colectomy modify plasma BA composition in PSC (Figure 3; Supplementary Tables 3 and 4). Among PSC patients, those with IBD were less likely to be receiving UDCA compared with those without IBD, although TBA concentrations did not differ significantly between groups.Figure 3Intestinal disease and colectomy modulate bile acid composition in PSC. Plasma bile acid (BA) composition in PSC patients stratified by inflammatory bowel disease (IBD) status and colectomy. (A) Ratio of the primary BA cholic acid (CA) to its secondary BA deoxycholic acid (DCA) in PSC patients with and without IBD. (B) CA:DCA ratio in PSC patients with prior colectomy and PSC-IBD patients without colectomy. (C) BA pool composition illustrates enrichment of primary BA species associated with intestinal disease and surgical disruption of enterohepatic circulation. These findings suggest that intestinal modifiers act on an underlying PSC-associated BA state rather than defining independent phenotypes. Statistical comparisons were performed using the Kruskal–Wallis test; exact P values are shown. Intestinal disease and colectomy modulate bile acid composition in PSC. Plasma bile acid (BA) composition in PSC patients stratified by inflammatory bowel disease (IBD) status and colectomy. (A) Ratio of the primary BA cholic acid (CA) to its secondary BA deoxycholic acid (DCA) in PSC patients with and without IBD. (B) CA:DCA ratio in PSC patients with prior colectomy and PSC-IBD patients without colectomy. (C) BA pool composition illustrates enrichment of primary BA species associated with intestinal disease and surgical disruption of enterohepatic circulation. These findings suggest that intestinal modifiers act on an underlying PSC-associated BA state rather than defining independent phenotypes. Statistical comparisons were performed using the Kruskal–Wallis test; exact P values are shown. Despite similar TBA levels, PSC patients with IBD exhibited alterations in BA composition consistent with impaired conversion of primary to secondary BA. Specifically, the primary-to-secondary BA ratio, summarized by the CA:DCA ratio, was significantly higher in PSC patients with IBD compared with those without IBD (Figure 3A). This shift was reflected in the plasma BA pool, which showed relative depletion of secondary BA species and enrichment of primary BA species in the presence of intestinal disease (Figure 3C). We next assessed whether colectomy, a major anatomical disruption of enterohepatic circulation, exerted a stronger influence on BA composition within PSC patients. Compared with PSC-IBD patients without colectomy, those with a prior colectomy demonstrated a markedly higher CA:DCA ratio (Figure 3B), indicating a more pronounced shift toward primary BA dominance. BA pool composition further confirmed substantial depletion of secondary BA species in PSC patients with colectomy (Figure 3C). Importantly, comparison of the magnitude of these effects indicates that colectomy is associated with a substantially greater alteration in BA composition than IBD alone. Whereas PSC patients with IBD but intact colonic anatomy exhibited modest shifts in primary-to-secondary BA balance, those with colectomy showed the most pronounced disruption of the circulating BA pool. Collectively, these analyses demonstrate that intestinal disease and surgical alteration of the gut modulate the BA-defined cholestatic state in PSC, with colectomy exerting the strongest effect. The observed shifts toward primary BA dominance support the concept that plasma BA profiles capture biologically meaningful variation related to the gut–liver axis. These modifiers act on an underlying PSC-associated BA state rather than defining independent disease phenotypes. Bile Acid Composition Shifts With Advancing Liver Disease Stage in Primary Sclerosing Cholangitis and Primary Biliary Cholangitis To evaluate whether BA profiles reflect disease stage in chronic cholestatic liver disease, we compared plasma BA composition among patients with no cirrhosis, compensated cirrhosis, and decompensated cirrhosis in both PSC and PBC (Figure 4; Supplementary Tables 5-8). Cirrhosis status was determined using established histologic, imaging, and clinical criteria and was defined independently of BA measurements.Figure 4Bile acid composition shifts with advancing liver disease stage in PSC and PBC. Plasma bile acid (BA) composition across liver disease stages in PBC and PSC, stratified as no cirrhosis, compensated cirrhosis, and decompensated cirrhosis. (A) In both diseases, advancing stage is associated with increased ratio of the primary BA cholic acid (CA) to its secondary BA deoxycholic acid (DCA). (B) BA pool composition demonstrates that compositional shifts occur with the development of cirrhosis and are largely maintained with further disease progression. (C) In PSC, established prognostic scores (PREsTo, PSC-BAP, and ELF) increase across cirrhosis stages, paralleling but not fully mirroring BA compositional changes. Statistical comparisons were performed using the Kruskal–Wallis test; exact P values are shown. Bile acid composition shifts with advancing liver disease stage in PSC and PBC. Plasma bile acid (BA) composition across liver disease stages in PBC and PSC, stratified as no cirrhosis, compensated cirrhosis, and decompensated cirrhosis. (A) In both diseases, advancing stage is associated with increased ratio of the primary BA cholic acid (CA) to its secondary BA deoxycholic acid (DCA). (B) BA pool composition demonstrates that compositional shifts occur with the development of cirrhosis and are largely maintained with further disease progression. (C) In PSC, established prognostic scores (PREsTo, PSC-BAP, and ELF) increase across cirrhosis stages, paralleling but not fully mirroring BA compositional changes. Statistical comparisons were performed using the Kruskal–Wallis test; exact P values are shown. In both PSC and PBC, the transition from no cirrhosis to compensated cirrhosis was associated with marked alterations in BA composition. Patients with compensated cirrhosis exhibited significantly higher total BA concentrations and a shift toward primary BA dominance, summarized by an increased CA:DCA ratio (Figure 4A). These changes were accompanied by broader compositional remodeling of the plasma BA pool, as visualized by relative enrichment of primary BA species and depletion of secondary BA species (Figure 4B). Across both diseases, additional BA features—including increased conjugation and altered G:T ratios—also differed between noncirrhotic and cirrhotic patients (Supplementary Tables 5-8), indicating that multiple dimensions of BA metabolism change with advancing disease stage. Notably, differences between compensated and decompensated cirrhosis were more modest, suggesting that major BA compositional shifts occur early in the cirrhotic disease course and are largely maintained thereafter. In PSC, established prognostic scores, including the Primary Sclerosing Cholangitis Risk Estimate Tool (PREsTo),16 the PSC-BAP score,9 and the ELF score,15 increased progressively across cirrhosis stages (Figure 4C). BA compositional changes paralleled these clinical severity measures but did not fully mirror them, supporting the concept that BA profiles capture biologically meaningful aspects of disease state that are related to, yet partially independent of, laboratory-based prognostic markers. Together, these findings demonstrate that plasma BA composition evolves with advancing liver disease stage in both PSC and PBC. Shifts toward primary BA dominance and altered conjugation occur early in cirrhosis and are preserved with further disease progression, reinforcing the interpretation of BA profiles as indicators of biologically meaningful disease states rather than simple markers of decompensation. Associations Between Bile Acid Variables, Clinical Laboratory Tests, and Hepatic Decompensation To contextualize the biological relevance of BA profiles with respect to clinical outcomes, we evaluated associations between representative BA variables, standard CLTs, and time to HD in PSC and PBC (Figure 5; Supplementary Tables 9 and 10). These analyses were performed to assess whether BA compositional features associated with disease state and stage were also linked to subsequent clinical events.Figure 5Associations between bile acid variables, clinical laboratory tests, and hepatic decompensation. Univariable associations between selected (A) clinical laboratory tests (CLTs) and (B) bile acid (BA) variables and time to hepatic decompensation in PSC and PBC. (C) Published prognostic scores (PREsTo, PSC-BAP, and ELF) demonstrate strong associations but show limited correlation with one another, indicating partially distinct disease information. In PBC, associations are shown for completeness but should be interpreted as exploratory due to the small number of events. Points represent hazard ratios scaled to the interquartile range, with whiskers indicating 95% confidence intervals. Associations between bile acid variables, clinical laboratory tests, and hepatic decompensation. Univariable associations between selected (A) clinical laboratory tests (CLTs) and (B) bile acid (BA) variables and time to hepatic decompensation in PSC and PBC. (C) Published prognostic scores (PREsTo, PSC-BAP, and ELF) demonstrate strong associations but show limited correlation with one another, indicating partially distinct disease information. In PBC, associations are shown for completeness but should be interpreted as exploratory due to the small number of events. Points represent hazard ratios scaled to the interquartile range, with whiskers indicating 95% confidence intervals. In PSC, 34 patients experienced HD during follow-up, with a median follow-up time of 4.57 years. CLTs reflecting liver disease severity, including alkaline phosphatase, aspartate aminotransferase, platelet count, and total bilirubin, were significantly associated with HD risk (Figure 5A). In parallel, representative BA variables, including total BAs and primary BA species, were also associated with increased risk of decompensation (Figure 5B), indicating that BA composition relates to clinically meaningful outcomes in PSC. Notably, the BA variables shown in Figure 5 represent components of broader BA compositional patterns described earlier in the article and overlap conceptually with those contributing to the PSC-BAP score. At the same time, correlations among established prognostic scores, including PREsTo, PSC-BAP, and the ELF score, were moderate rather than strong (Figure 5C), suggesting that these measures capture overlapping but nonidentical dimensions of disease progression. In PBC, only 9 HD events occurred during follow-up, limiting the interpretability of outcome associations. Although selected BA variables and CLTs demonstrated nominal associations with HD, these findings should be interpreted as exploratory and hypothesis-generating. Collectively, these analyses demonstrate that BA compositional features associated with disease state and stage are also linked to clinically relevant outcomes, particularly in PSC. However, the overlap between BA-based measures, CLTs, and established prognostic scores indicates that BA profiles provide complementary biological information rather than independent prognostic signals. These observations motivate subsequent multivariable modeling to further assess the relationship between BA-defined disease states and established clinical risk measures. Multivariable Models Demonstrate Complementary, Not Additive, and Prognostic Information From Bile Acids To further assess the relationship between BA profiles and established clinical laboratory–based prognostic measures, we evaluated the performance of published risk scores and developed exploratory multivariable models using BA variables, CLTs, and their combination (Figure 6; Supplementary Figures 2 and 3). These analyses were performed to determine whether BA-defined disease states provide prognostic information distinct from, or redundant with, standard clinical markers.Figure 6Multivariable models demonstrate complementary prognostic information from bile acid profiles and clinical laboratory tests. Performance of exploratory multivariable models and published risk scores for prediction of hepatic decompensation. In PSC, the Primary Sclerosing Cholangitis Risk Estimate Tool (PREsTo) and PSC-BAP scores demonstrate comparable discrimination. Exploratory gradient boosting machine (GBM) models derived using bile acid (BA) variables alone or clinical laboratory tests (CLTs) alone show similar performance, while combined BA/CLT models do not materially improve discrimination. In PBC, model results are exploratory due to limited events. Model discrimination is summarized by concordance statistics (c-stat). Multivariable models demonstrate complementary prognostic information from bile acid profiles and clinical laboratory tests. Performance of exploratory multivariable models and published risk scores for prediction of hepatic decompensation. In PSC, the Primary Sclerosing Cholangitis Risk Estimate Tool (PREsTo) and PSC-BAP scores demonstrate comparable discrimination. Exploratory gradient boosting machine (GBM) models derived using bile acid (BA) variables alone or clinical laboratory tests (CLTs) alone show similar performance, while combined BA/CLT models do not materially improve discrimination. In PBC, model results are exploratory due to limited events. Model discrimination is summarized by concordance statistics (c-stat). In PSC, validation of PREsTo and the PSC-BAP score in this independent cohort demonstrated good discrimination for HD, with concordance statistics of approximately 0.85 for both scores. Calibration of these scores was acceptable in PSC, supporting their applicability in this cohort. We next developed exploratory GBM models using BA variables alone. These BA-based models demonstrated good discrimination for HD in PSC (c-statistic ≈0.92) and, when evaluated in an independent PSC cohort used for development of the original PSC-BAP score, retained moderate discriminatory performance. Importantly, the BA variables contributing most strongly to these models overlapped with those defining the PSC-BAP score, reinforcing the internal consistency of BA-based disease phenotyping. Parallel GBM models constructed using CLT variables alone also demonstrated comparable discrimination in PSC (c-statistic ≈0.93) and included multiple variables incorporated in PREsTo, as well as additional hematologic measures. When BA and CLT variables were combined into a single multivariable model, discrimination was like that of CLT-only models and only modestly higher than BA-only models. These findings indicate that BA and CLT variables capture overlapping dimensions of disease severity, and that combining them does not materially improve overall predictive performance. Consistent with this interpretation, PREsTo and PSC-BAP scores showed limited correlation with one another in PSC (Figure 5C), suggesting that each reflects partially distinct aspects of disease biology and progression despite comparable predictive performance. Together, these observations support the view that BA-based models and CLT-based models are complementary rather than competitive, capturing different facets of the same underlying disease process. In PBC, GBM models derived using BA variables, CLT variables, or their combination showed high discrimination in derivation. However, given the small number of HD events and the absence of an independent validation cohort, these models should be considered exploratory and hypothesis-generating. Notably, published PSC-derived scores also demonstrated good discrimination but suboptimal calibration in PBC, further underscoring differences in disease context and event structure. Overall, multivariable modeling demonstrates that BA profiles encode prognostically relevant biological information comparable to standard clinical laboratory–based measures, without conferring substantial additive predictive value when combined. These findings support the use of BA profiling as a complementary tool for disease stratification and biological characterization in chronic cholestatic liver diseases, rather than as a replacement for established clinical risk models. Discussion In this study, we performed a comprehensive analysis of plasma BA composition in patients with PSC and PBC and evaluated how BA-defined disease states relate to intestinal disease, liver disease stage, and clinical outcomes. Our findings demonstrate that BA profiles encode distinct, disease-specific cholestatic states in PSC and PBC that extend beyond total BA levels and conventional CLTs. While BA variables associate with HD, their principal value lies not in outperforming established clinical risk models but in providing complementary biological information relevant to disease stratification and translational research. A central observation of this study is the marked divergence in BA conjugation and composition between PSC and PBC. Notably, patterns of BA conjugation appear to encode both disease identity and disease stage. In the case–control setting, PSC and PBC exhibited opposing shifts in G:T conjugation ratios, suggesting fundamentally different baseline BA handling in the 2 diseases. However, with advancing liver disease stage, both conditions converged toward reduced G:T ratios, suggesting that cirrhosis exerts a dominant effect on BA conjugation that supersedes disease-specific signatures. This convergence suggests that BA conjugation patterns may reflect hierarchical biological influences, with disease-specific mechanisms shaping early BA states and progressive hepatic dysfunction imposing a common late-stage metabolic constraint. While mechanistic studies will be required to delineate the drivers of these shifts, these observations highlight the potential value of BA conjugation metrics as sensitive indicators of evolving disease biology in PSC and PBC. Alterations in BA composition were further modulated by intestinal disease and anatomy in PSC. Patients with concomitant IBD, and particularly those with prior colectomy, exhibited depletion of secondary BA species and enrichment of primary BA species. These findings are consistent with the known dependence of BA transformation on intestinal microbial activity and enterohepatic circulation.17 While our data do not directly assess microbial composition or function, the observed BA shifts support the concept that perturbations of the gut–liver axis modify the circulating BA pool in PSC.16 Notably, these intestinal modifiers acted upon an underlying PSC-associated BA state rather than defining independent phenotypes, reinforcing the interpretation of BA profiles as state descriptors. BA composition also evolved with advancing liver disease stage in both PSC and PBC. Across diseases, cirrhosis was associated with increased total BA concentrations, higher conjugation fractions, and a shift toward primary BA dominance. These changes were most pronounced when comparing noncirrhotic to cirrhotic patients, with more modest differences between compensated and decompensated cirrhosis. This pattern suggests that major alterations in BA metabolism occur early in the course of cirrhotic disease and are subsequently maintained. Importantly, these compositional shifts were observed despite heterogeneity in UDCA use, supporting the notion that BA profiles reflect disease biology beyond therapeutic exposure. Consistent with prior work, both BA variables and standard CLTs were associated with HD, particularly in PSC. Taurine-conjugated BA species featured prominently among BA variables associated with outcomes, overlapping with components of the previously described PSC-BAP score.9 However, substantial correlation between BA variables and CLTs indicates that BA outcome associations track with established markers of disease severity. At the same time, limited correlation between PREsTo,16 PSC-BAP,9 and ELF15 scores suggests that these models capture partially distinct aspects of disease progression. Multivariable modeling further clarified the relationship between BA profiles and clinical prognostication. Exploratory BA-based models demonstrated discrimination comparable to clinical laboratory–based models in PSC but combining BA and clinical variables did not materially improve predictive performance. These findings underscore that BA profiles are complementary rather than additive to standard clinical risk measures. In PBC, outcome modeling was limited by event counts and should be interpreted as exploratory; nevertheless, similar patterns of overlap between BA and clinical predictors were observed. Together, these results argue against viewing BA profiling as a replacement for established prognostic tools and instead support its use as a biological stratification approach that captures disease-relevant information not otherwise measured. The implications of these findings are primarily translational. In this context, BA profiling may be particularly valuable for enriching and interpreting clinical trials targeting BA metabolism, gut–liver signaling, or microbial pathways. Standard CLTs primarily reflect liver injury and synthetic function, whereas BA composition provides a direct readout of enterohepatic metabolic state. Integrating these complementary dimensions may improve biological resolution in early-phase trials, help distinguish pharmacodynamic effects from background disease progression, and support patient stratification based on metabolic phenotype rather than clinical severity alone. More broadly, the definition of BA-based cholestatic states provides a framework for integrative multiomic studies aimed at linking circulating metabolites with downstream transcriptional, immunologic, and microbial signatures. Such approaches may clarify how BA dysregulation interfaces with immune-mediated injury and intestinal perturbations in chronic cholestatic liver disease. Several limitations merit consideration. This was a retrospective, cross-sectional analysis, precluding assessment of longitudinal changes in BA composition or direct inference regarding causality. UDCA exposure was inferred from plasma BA measurements rather than medication records, introducing potential misclassification, although this approach allowed objective assessment of BA pool composition. The near-universal use of UDCA in PBC limited analyses of untreated patients, and the small number of HD events constrained outcome modeling in this disease. In addition, we did not assess serum markers of BA synthesis, such as 7α-hydroxy-4-cholesten-3-one or directly measure gut microbial composition or activity. Future studies integrating longitudinal sampling, microbiome profiling, and multiomic analyses will be required to delineate the mechanisms underlying BA-defined disease states more fully. Conclusion Plasma BA profiles define distinct and biologically meaningful cholestatic states in PSC and PBC that reflect disease type, intestinal modifiers, and liver disease stage. While BA variables associate with clinical outcomes, their principal value lies in complementing established clinical laboratory–based assessments rather than improving predictive accuracy. These findings position BA profiling as a foundational tool for biological disease stratification and provide a framework for future integrative studies of the gut–liver axis in chronic cholestatic liver diseases.