Liver cirrhosis is associated with altered glucocorticoid metabolism related to disease severity Authors: Soyeon Park, Min Jeong Park, Hye Won Yang, Jeayeon Park, Su Jong Yu, Seung Shin Park, Man‐Ho Choi, Eun Ju Cho, Jung Hee Kim Published: 2026 ABSTRACT AbstractBackgroundLiver cirrhosis (LC) is frequently complicated by adrenal dysfunction; however, whether cirrhosis drives the coordinated, pathway‐specific reprogramming of cortisol metabolism remains unclear.ObjectivesWe aimed to characterize axis‐specific alterations in cortisol metabolism in patients with LC, those with adrenal insufficiency (AI), and controls using integrated serum and salivary steroid profiling.MethodsThis prospective study included adults with LC (n = 33) and control participants without cirrhosis who underwent cosyntropin‐stimulation test following transsphenoidal surgery for a nonfunctioning pituitary adenoma (controls, n = 27; AI, n = 10). Serum and saliva samples were collected at baseline, 30 and 60 min after stimulation and profiled for 26 serum and 12 salivary steroids through liquid chromatography–tandem mass spectrometry. Multivariate models were constructed to identify cirrhosis‐associated steroid features, and their associations with disease severity.ResultsLC was characterized by axis‐specific steroid remodeling, including suppressed 11β‐hydroxysteroid dehydrogenase 1 and A‐ring reduction activity with the 20α‐reduction pathway activation. Representative changes included a lower peak cortisol‐to‐cortisone ratio and reduced tetrahydrocortisone‐to‐cortisone ratio, alongside increased 20α‐reduced metabolites (all p < 0.001). This pattern was distinct from that in AI, in which steroid production and downstream metabolites were uniformly reduced. Iterative model refinement identified a parsimonious seven‐steroid signature that discriminated patients with LC from controls. Within the LC cohort, the steroid signature correlated with the Child–Pugh score.ConclusionLC is associated with coordinated, axis‐specific cortisol metabolic reprogramming, which is mechanistically distinct from AI. Integrated steroid profiling may provide mechanistic insight into altered glucocorticoid metabolism in cirrhosis and its relationship to disease severity .Keywords: adrenal insufficiency, cortisol, liver cirrhosis, metabolomics, steroid profiling Abbreviations 11β‐hydroxysteroid dehydrogenase 3α‐hydroxysteroid dehydrogenase adrenocorticotropic hormone adrenal insufficiency area under the curve confidence interval cosyntropin‐stimulation test dehydroepiandrosterone sulfate cortisone cortisol liver cirrhosis liquid chromatography–tandem mass spectrometry model for end‐stage liver disease–sodium principal component analysis partial least squares discriminant analysis tetrahydrocortisone tetrahydrocortisol Introduction Liver cirrhosis (LC) is the end stage of chronic liver disease, characterized by progressive fibrosis, architectural distortion, and impaired hepatic function. Globally, LC affects over 120 million individuals and remains a leading cause of morbidity and mortality, accounting for over 1 million deaths annually [1]. Beyond structural liver damage, LC exerts systemic effects, such as metabolic reprogramming, immune dysregulation, and endocrine disturbances [2, 3]. Among these systemic effects, alterations in glucocorticoid metabolism are increasingly recognized for their role in adrenal insufficiency (AI), inflammation, and disease progression [4]. The liver serves as the central hub for systemic steroid metabolism, regulating glucocorticoid bioavailability through oxidative, reductive, and conjugative pathways [5, 6]. These processes are organized around three principal enzymatic axes: 11β‐hydroxysteroid dehydrogenase (11β‐HSD), A‐ring reduction by 5α/5β‐reductases and 3α‐hydroxysteroid dehydrogenase (3α‐HSD), and 20α‐hydroxysteroid dehydrogenases (20α‐HSDs) [6, 7]. First, 11β‐HSD1 regenerates active cortisol from cortisone in hepatocytes, whereas 11β‐HSD2 inactivates cortisol in mineralocorticoid‐sensitive tissues [8]. Second, A‐ring reduction via 5α/5β‐reductases and 3α‐HSD irreversibly reduces cortisol and cortisone into tetrahydrocortisol (THF), allo‐tetrahydrocortisol (allo‐THF), and tetrahydrocortisone (THE), respectively. Third, 20α‐HSD catalyzes downstream reductions of cortisol, cortisone, THF, allo‐THF, and THE into 20α‐dihydrocortisol (20α‐DHF), 20α‐dihydrocortisone (20α‐DHE), α‐cortol, β‐cortol, and α‐ and β‐cortolone, respectively [9, 10]. These pathways depend on hepatocyte integrity, zonal organization, and redox balance—all of which are disrupted in advanced cirrhosis [6, 7]. Altered cortisol dynamics in LC have primarily been investigated in the context of relative AI (RAI), often defined by blunted cortisol responses to cosyntropin stimulation [11, 12]. Importantly, adrenal dysfunction in cirrhosis has been associated with increased mortality, including mortality independent of MELD score severity [13]. Although such “hepatoadrenal syndrome” has been described, prior studies have largely focused on isolated cortisol measurements rather than coordinated remodeling of steroid metabolic networks [14]. Whether cirrhosis induces structured, axis‐specific reprogramming of steroid metabolism remains unclear, particularly when integrating serum and salivary compartments. Recent advances in steroid profiling based on mass spectometry have enabled the simultaneous quantification of multiple precursors and metabolites, along with their metabolic ratios that reflect enzymatic activity in steroidogenesis, thereby providing an opportunity to investigate axis‐specific metabolic remodeling [15]. In this study, we hypothesized that LC induces reproducible axis‐specific reprogramming of hepatic steroid metabolism rather than uniform global suppression. We performed comprehensive serum and salivary steroid profiling in patients with LC, those with AI, and controls at baseline and after cosyntropin stimulation. Using multivariate modeling and iterative feature refinement, we sought to identify the enzymatic axes most disrupted in cirrhosis and assess whether these alterations are associated with disease severity. Patients and methods Study design and participants We prospectively enrolled adult patients (≥19 years) with LC between November 2023 and August 2025 from Seoul National University Hospital. Cirrhosis was diagnosed based on clinical or radiological evidence using the following criteria: coarse hepatic echotexture and nodular surface on ultrasonography, thrombocytopenia (platelet count <150 × 103/mm3), or clinical signs of portal hypertension (e.g., ascites, splenomegaly, and varices). During the same period, patients without cirrhosis undergoing surgery for nonfunctioning pituitary adenomas were recruited and scheduled for a cosyntropin‐stimulation test (CST) approximately 3 months postoperatively as part of a routine endocrine evaluation. In the LC group, individuals with a prior diagnosis of AI, a history of pituitary surgery or adrenalectomy, a diagnosis of pituitary or adrenal tumors, or systemic glucocorticoid use within the preceding 3 months were excluded. In all groups, participants who had used medications affecting steroid synthesis or corticosteroid‐binding globulin levels within 3 months, had received albumin or plasma infusion within 24 h, had used vasopressin analogs, or had acute illness (e.g., active bleeding, sepsis, and infection) were also excluded. Participants were enrolled in the outpatient setting or during elective hospitalization for evaluation or planned procedures and were clinically stable at the time of assessment. All patients with HBV‐associated cirrhosis were receiving antiviral therapy with viral suppression at enrollment. Patients with HCV‐associated cirrhosis had no evidence of active viremia at the time of study participation. Patients with alcohol‐associated cirrhosis were enrolled only when clinically stable and abstinent from alcohol at the time of study participation. Among participants without cirrhosis, six with incomplete salivary sampling and three with dexamethasone detected by liquid chromatography–mass spectrometry (LC–MS) were excluded from the final analyses. After the eligibility criteria were applied, 33 patients with LC and 37 without cirrhosis were included in the study. In patients with LC group, salivary steroid profiling was unavailable for six patients with LC owing to insufficient saliva volume (Fig. S1). Based on CST results, participants without cirrhosis with a peak serum cortisol level of <15 µg/dL (LC–MS) were classified as the AI group (n = 10), and the remainder served as controls with normal adrenal function (n = 27) [16]. This cutoff was based on recent LC–MS/MS specific studies suggesting lower cortisol thresholds compared to conventional 18 µg/dL cutoff. As this study was designed as an exploratory pilot investigation aimed at characterizing steroid profiles in LC, a formal sample size calculation was not performed. The study protocol was approved by the Institutional Review Board of Seoul National University Hospital (IRB No. 2311‐009‐1480) and registered at ClinicalTrials.gov (NCT06206655). This study was conducted in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all the participants. Clinical and biochemical parameters Demographic variables, including age, sex, body mass index, and comorbidities, such as hypertension, diabetes mellitus, and dyslipidemia, were recorded. Vital signs, including systolic and diastolic blood pressures and heart rate, were measured at baseline. Laboratory parameters included hemoglobin, platelet count, aspartate aminotransferase, alanine aminotransferase, total bilirubin, total cholesterol, low‐density lipoprotein cholesterol, high‐density lipoprotein [HDL] cholesterol, triglycerides, total protein, albumin, blood urea nitrogen, creatinine, estimated glomerular filtration rate (eGFR‐EPI), and serum sodium levels. Liver disease severity was assessed using the Child–Pugh classification and the model for end‐stage liver disease–sodium (MELD‐Na) score [17, 18]. Hepatic decompensation was defined as the development or recurrence of ascites, variceal bleeding, hepatic encephalopathy, or jaundice [19]. Basal plasma adrenocorticotropic hormone (ACTH) and serum cortisol levels were measured at baseline and 30 and 60 min after cosyntropin stimulation, with peak cortisol defined as the highest value obtained post‐stimulation. Quantitative profiling of serum and salivary steroid Serum and salivary samples were collected between 8:00 and 9:00 a.m. after overnight fasting at baseline and 30 and 60 min after administration of 250 µg cosyntropin, which is a standard dose of short Synacthen test. Samples were obtained and processed individually for each participant (1:1, not pooled). Salivary samples were collected by unstimulated passive drool (without gustatory, visual, or olfactory stimulation), as salivary cortisol enters saliva by passive diffusion of the free fraction and its concentration is independent of salivary flow rate [20, 21]. The target sample volume was 1.0–1.5 mL of saliva, and samples with a collected volume of less than 0.4 mL were considered inadequate for analysis. Both poststimulation time points were collected to capture the true peak, as the stimulated cortisol peak may occur at either time, and reduced cortisol clearance in cirrhosis may prolong the response [11, 22, 23]; the peak was taken as the higher of the two values. All samples were stored at −70°C until analysis. Steroids were quantified using an LC–MS method based on previous reports [24, 25]. A total of 26 serum and 12 salivary steroids were measured at each time point. Peak levels were defined as the highest values obtained after stimulation. The metabolic ratio between each precursor and its metabolite was calculated to reflect enzyme activity (e.g., F/E). Variables labeled with “‐ratio” represent the peak‐to‐basal ratio of the corresponding steroid following stimulation [e.g., dehydroepiandrosterone sulfate (DHEA‐S)‐ratio = peak/basal DHEA‐S]. Multivariate modeling for the identification of an LC‐specific steroid signature All serum and salivary steroid variables were log 10‐transformed (after minimal positive shifting, when necessary) and autoscaled to unit variance. Values below the limit of detection were imputed as half of the minimum detected value for each variable, and zero‐variance features were excluded. Unsupervised principal component analysis (PCA) was performed to visualize global steroid patterns among the three groups. Supervised partial least squares–discriminant analysis (PLS‐DA) with two components was used for classification. Model performance was assessed using R 2 Y, fivefold cross‐validated Q 2, and permutation testing (n = 100). A two‐group PLS‐DA model (control vs. LC) was constructed to identify LC‐specific discriminative steroid variables. Variable importance in projection (VIP) scores were extracted, and steroid variables with VIP ≥ 1.5 were retained. Highly collinear variables (|ρ| ≥ 0.80) were reduced according to predefined criteria that prioritize metabolic ratio‐derived and peak variables. A parsimonious model was derived using an iterative VIP‐based reduction with leave‐one‐out cross‐validation. Feature selection and VIP‐based reductions were performed within each cross‐validation loop to avoid information leakage. The number of variables was sequentially reduced according to the VIP ranking. Heatmaps were generated using standardized log‐transformed features, grouped into three predefined enzymatic axes (11β‐HSD, A‐ring reduction, and 20α‐HSD). Group differences were assessed using the Kruskal–Wallis test with Dunn's post hoc comparisons. Statistical analysis Continuous variables were expressed as medians [interquartile range] and compared using the Kruskal–Wallis test or Mann–Whitney U test with Dunn's post hoc comparisons, whereas categorical variables were expressed as frequencies (percentages) and compared using the chi‐square test. Associations between the selected seven steroid variables and clinical severity indices in the LC group were assessed using Spearman rank correlation analysis, with correlation coefficients (ρ) and corresponding p‐values reported. All statistical analyses were performed using Python (version 3.12.12). Statistical significance was defined as a two‐sided p‐value of <0.05. Results Baseline characteristics and global steroid profiling The baseline characteristics of the study population are summarized in Table 1. The median age of the LC group was 60.0 years, which was comparable to that of the non‐cirrhotic control (53.0 years) and AI (63.5 years) groups. The proportion of females was significantly lower in the LC group (27.3%) than in the non‐cirrhotic control group (70.4%, p = 0.001). The LC group exhibited significantly lower albumin levels, platelet counts, and total cholesterol compared to the control group (all p < 0.001). Basal ACTH levels were higher (49.4 vs. 24.0 pg/mL), whereas peak serum cortisol levels following stimulation were significantly blunted in the LC group compared with that in the control group s (20.8 vs. 31.5 µg/dL, all p < 0.001). Unsupervised PCA revealed a distinct clustering of the LC group along the PC2 axis (explaining 15.6% of the variance), indicating a global shift in the steroid profile (Fig. 1a). Supervised PLS‐DA further confirmed this separation with significant predictive power (R 2 Y = 0.618, Q 2 = 0.499, permutation p = 0.010) (Fig. S2a). Table 1Baseline characteristics of the study participants.VariableControls (n = 27)Liver cirrhosis (n = 33)Adrenal insufficiency (n = 10) p‐valueAge (years)53.0 [31.5–65.0]60.0 [52.0–66.0]63.5 [56.5–66.0]0.107Female19 (70.4%)9 (27.3%) * 2 (20.0%)0.001Body mass index (kg/m2)24.0 [22.5–29.0]25.0 [23.0–27.0]25.5 [22.2–26.0]0.774Systolic BP (mmHg)129.0 [109.5–137.5]117.0 [109.0–133.0]126.5 [118.8–135.0]0.566Diastolic BP (mmHg)75.0 [68.5–85.5]73.0 [64.0–80.0]78.5 [73.2–79.0]0.369Heart rate (bpm)75.0 [67.5–82.0]70.0 [63.0–84.0]68.0 [63.5–89.2]0.730Hypertension2 (7.4%)9 (27.3%)4 (40.0%)0.053Diabetes mellitus3 (11.1%)13 (39.4%) * 3 (30.0%)0.048Dyslipidemia7 (25.9%)9 (27.3%)4 (40.0%)0.109Etiology of liver disease—HBV11 (33.3%)HCV3 (9.1%)Alcohol12 (36.4%)Others7 (21.1%) Laboratory findings Hemoglobin (g/dL)13.0 [12.0–14.0]11.0 [11.0–13.0] * 13.0 [12.0–13.0]0.004Platelet count (×109/L)250.5 [214.5–295.0]78.0 [60.0–103.0] * 210.5 [177.5–232.8]<0.001AST (U/L)22.5 [13.2–26.0]41.0 [30.0–61.0] * 20.0 [19.0–27.0]<0.001ALT (U/L)19.0 [13.2–28.8]32.0 [22.0–39.0] * 26.0 [11.0–28.0]0.011Total bilirubin (mg/dL)1.0 [1.0–1.0]1.2 [0.9–2.0] * 1.0 [1.0–1.0]0.172BUN (mg/dL)12.0 [11.2–16.5]13.0 [10.0–18.0]15.0 [13.0–16.0]0.916Creatinine (mg/dL)1.0 [1.0–1.0]0.8 [0.7–0.9] * 1.0 [1.0–1.0]<0.001eGFR‐EPI (mL/min/1.73 m2)99.5 [93.0–115.2]94.8 [80.9–103.5]87.0 [80.0–101.0]0.069Total cholesterol (mg/dL)169.0 [141.0–205.0]137.0 [111.0–159.0] * 196.0 [185.0–248.5]<0.001LDL cholesterol (mg/dL)92.5 [64.2–131.8]79.0 [67.0–98.0]128.0 [114.0–180.0]0.009HDL cholesterol (mg/dL)52.5 [41.8–59.2]37.0 [25.0–45.0] * 41.0 [37.5–49.0]0.015Triglycerides (mg/dL)102.5 [71.8–154.0]83.0 [68.0–94.0]147.0 [135.5–235.0]0.002Albumin (g/dL)4.0 [4.0–4.0]3.3 [3.0–3.7] * 4.0 [4.0–4.0]<0.001Serum sodium (mmol/L)140.0 [140.0–141.0]139.0 [135.0–140.0] * 140.0 [140.0–143.0]0.003 Adrenal function Basal plasma ACTH (pg/mL)24.0 [16.5–36.0]49.4 [35.5–66.5] * 18.5 [14.5–25.5]<0.001Basal serum cortisol (µg/dL)12.1 [9.6–14.8]9.3 [5.1–10.9] * 4.4 [1.9–6.3]<0.001Peak serum cortisol (µg/dL)31.5 [27.7–38.3]20.8 [16.3–24.5] * 11.9 [6.1–18.7]<0.001 Note: Values are presented as median [interquartile range] or n (%). p‐values were determined using the Kruskal–Wallis test across the three groups or by the chi‐square test, where appropriate. Statistical significance was set at p‐value < 0.05. The etiology was assessed only in the liver cirrhosis group; therefore, p‐values were not applicable for these variables. To convert to SI units, multiply by the following factors: hemoglobin and albumin (g/dL to g/L) ×10; total bilirubin (mg/dL to µmol/L) ×17.1; blood urea nitrogen (mg/dL to mmol/L) ×0.357; creatinine (mg/dL to µmol/L) ×88.4; total, LDL, and HDL cholesterol (mg/dL to mmol/L) ×0.0259; triglycerides (mg/dL to mmol/L) ×0.0113; ACTH (pg/mL to pmol/L) ×0.220; and cortisol (µg/dL to nmol/L) ×27.59. Platelet count, AST, ALT, sodium, and eGFR are already expressed in SI units.Abbreviations: ACTH, adrenocorticotropic hormone; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BP, blood pressure; BUN, blood urea nitrogen; eGFR‐EPI, estimated glomerular filtration rate calculated using the Chronic Kidney Disease Epidemiology Collaboration equation; Hb, hemoglobin; HBV, hepatitis B virus; HCV, hepatitis C virus; HDL, high‐density lipoprotein; LDL,; low‐density lipoprotein.* p < 0.05, compared with the control group by post hoc analysis using Dunn's multiple comparison test. Baseline characteristics of the study participants. Note: Values are presented as median [interquartile range] or n (%). p‐values were determined using the Kruskal–Wallis test across the three groups or by the chi‐square test, where appropriate. Statistical significance was set at p‐value < 0.05. The etiology was assessed only in the liver cirrhosis group; therefore, p‐values were not applicable for these variables. To convert to SI units, multiply by the following factors: hemoglobin and albumin (g/dL to g/L) ×10; total bilirubin (mg/dL to µmol/L) ×17.1; blood urea nitrogen (mg/dL to mmol/L) ×0.357; creatinine (mg/dL to µmol/L) ×88.4; total, LDL, and HDL cholesterol (mg/dL to mmol/L) ×0.0259; triglycerides (mg/dL to mmol/L) ×0.0113; ACTH (pg/mL to pmol/L) ×0.220; and cortisol (µg/dL to nmol/L) ×27.59. Platelet count, AST, ALT, sodium, and eGFR are already expressed in SI units. Abbreviations: ACTH, adrenocorticotropic hormone; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BP, blood pressure; BUN, blood urea nitrogen; eGFR‐EPI, estimated glomerular filtration rate calculated using the Chronic Kidney Disease Epidemiology Collaboration equation; Hb, hemoglobin; HBV, hepatitis B virus; HCV, hepatitis C virus; HDL, high‐density lipoprotein; LDL,; low‐density lipoprotein. p < 0.05, compared with the control group by post hoc analysis using Dunn's multiple comparison test. Distinct steroid profiling and discriminant analysis among study groups: (a) principal component analysis (PCA) score plot showing the global steroid metabolic distribution among patients with liver cirrhosis (LC, black circles), non‐cirrhotic controls (control, open squares), and patients with primary adrenal insufficiency (AI, gray crosses); (b) variable importance in projection (VIP) plot from the partial least squares discriminant analysis (PLS‐DA) model, highlighting the top 20 features contributing to group discrimination. 20α‐DHE, 20α‐dihydrocortisone; 20α‐DHF, 20α‐dihydrocortisol; DHEA‐S, dehydroepiandrosterone sulfate; THE, tetrahydrocortisone; THF, tetrahydrocortisol. An LC‐specific steroid signature characterized by axis‐specific reprogramming PLS‐DA was further performed to distinguish patients with LC from controls (permutation p = 0.010) (Fig. S2b). Using a VIP threshold of 1.5, 22 steroid variables were identified as major contributors to the group separation (Fig. 1b, Table S1). These alterations converged on three principal enzymatic axes: suppression of A‐ring reduction (5α/5β‐reductase and 3α‐HSD pathways), disruption of 11β‐HSD activity, and activation of the 20α‐reduction pathway (Table 2). Within the 11β‐HSD axis, the LC group showed significantly lower peak F/E and salivary 20α‐DHF/20α‐DHE ratios. In the 5α/5β‐reductase and 3α‐HSD pathway, basal and peak THE and THF metabolites, as well as their derived ratios (THF/F and [THF + allo‐THF]/F), were markedly reduced in the LC group compared with those in the control group. Conversely, steroid variables reflecting 20α‐reduction activity (20α‐DHE and 20α‐DHF species) were significantly elevated in the LC group. Alterations were also observed in precursor‐ and androgen‐related pathways, including reduced basal P4 and DHEA‐S/DHEA levels and increased P4 and basal 17‐OHP4/P4 ratios. The results for the entire set of steroid variables are summarized in Table S2. Table 2Comparison of steroid metabolites and ratios with high variable importance in projection (VIP) scores (>1.5).Variable (ng/mL)Controls (n = 27)Liver cirrhosis * (n = 33)Adrenal insufficiency (n = 10) p‐value 11β‐HSD axis F/E (peak)8.70 [7.38–10.95]6.12 [5.14–8.05]6.89 [5.24–8.90]<0.001Salivary 20α‐DHF/20α‐DHE (peak)0.29 [0.23–0.36]0.17 [0.13–0.21]0.13 [0.07–0.23]<0.001 5α/5β‐reductase + 3α‐HSD axis THE (basal)32.24 [21.66–39.74]10.15 [5.75–14.50]14.63 [6.79–18.64]<0.001THE (peak)42.08 [32.50–53.53]13.60 [7.08–19.77]19.30 [10.74–23.84]<0.001THF (basal)14.96 [10.82–18.15]6.13 [3.00–9.41]8.02 [4.49–9.04]<0.001THF (peak)28.21 [17.30–36.00]10.73 [5.49–17.43]12.80 [9.07–17.28]<0.001THE/E (basal)1.49 [1.01–2.51]0.45 [0.25–0.75]1.55 [0.95–5.97]<0.001THE/E (peak)1.93 [1.40–2.27]0.47 [0.31–0.66]1.77 [0.80–2.40]<0.001THF/F (basal)0.17 [0.13–0.22]0.09 [0.05–0.12]0.20 [0.11–0.72]<0.001(THF + allo‐THF)/F (basal)0.27 [0.19–0.42]0.12 [0.07–0.21]0.35 [0.20–1.56]<0.001(THF + allo‐THF)/F (peak)0.23 [0.15–0.29]0.09 [0.06–0.14]0.23 [0.19–0.45]<0.001 20α‐HSD axis 20α‐DHE (basal)3.58 [2.92–4.20]8.34 [4.42–11.77]1.47 [0.61–1.87]<0.00120α‐DHE (peak)4.41 [3.15–5.28]10.24 [5.60–12.99]1.88 [1.15–2.43]<0.00120α‐DHF (peak)3.32 [2.81–3.98]4.98 [3.62–6.68]1.50 [0.56–1.79]<0.00120α‐DHF/F (peak)0.02 [0.01–0.02]0.03 [0.02–0.05]0.01 [0.01–0.02]<0.001Salivary 20α‐DHE (peak)2.81 [2.11–3.65]7.70 [4.57–11.06]0.81 [0.36–1.56]<0.001Salivary 20α‐DHF/F (peak)0.12 [0.08–0.14]0.29 [0.23–0.38]0.10 [0.07–0.12]<0.001 Other pathways P4 (basal)0.75 [0.12–1.42]0.09 [0.02–0.12]0.53 [0.28–0.77]<0.001P4 ratio1.84 [1.33–2.78]2.90 [2.00–6.11]1.23 [1.14–1.52]<0.00117‐OHP4/P4 (basal)0.57 [0.22–2.41]6.00 [3.30–9.50]0.46 [0.22–0.78]<0.001DHEA‐S/DHEA (basal)264.98 [205.08–546.50]111.91 [64.50–158.71]694.20 [427.46–973.40]<0.001DHEA‐S ratio1.06 [1.00–1.17]1.25 [1.11–1.38]1.06 [1.00–1.20]<0.001 Note: Values are expressed as median [interquartile range]. p‐values were calculated using the Kruskal–Wallis test. Steroid concentrations are reported in ng/mL. To convert to SI units (nmol/L), multiply by 1000/molecular weight; approximate factors are ×2.76 for cortisol (F), ×2.77 for cortisone (E), ×2.74 for tetrahydrocortisone (THE), ×2.73 for tetrahydrocortisol (THF) and 5α‐tetrahydrocortisol (allo‐THF), ×2.74 for 20α‐dihydrocortisol (20α‐DHF), ×2.76 for 20α‐dihydrocortisone (20α‐DHE), ×3.18 for progesterone (P4), ×3.03 for 17α‐hydroxyprogesterone (17‐OHP4), and ×3.47 for dehydroepiandrosterone (DHEA).Abbreviations: 17‐OHP4, 17α‐hydroxyprogesterone; 20α‐DHE, 20α‐dihydrocortisone; 20α‐DHF, 20α‐dihydrocortisol; allo‐THF, 5α‐tetrahydrocortisol; DHEA, dehydroepiandrosterone; DHEA‐S, dehydroepiandrosterone sulfate; E, cortisone; F, cortisol; HSD, hydroxysteroid dehydrogenase; P4, progesterone; THE, tetrahydrocortisone; THF, tetrahydrocortisol.* p < 0.05 for all variables compared with the control group by post hoc analysis using Dunn's multiple comparison test. Comparison of steroid metabolites and ratios with high variable importance in projection (VIP) scores (>1.5). Note: Values are expressed as median [interquartile range]. p‐values were calculated using the Kruskal–Wallis test. Steroid concentrations are reported in ng/mL. To convert to SI units (nmol/L), multiply by 1000/molecular weight; approximate factors are ×2.76 for cortisol (F), ×2.77 for cortisone (E), ×2.74 for tetrahydrocortisone (THE), ×2.73 for tetrahydrocortisol (THF) and 5α‐tetrahydrocortisol (allo‐THF), ×2.74 for 20α‐dihydrocortisol (20α‐DHF), ×2.76 for 20α‐dihydrocortisone (20α‐DHE), ×3.18 for progesterone (P4), ×3.03 for 17α‐hydroxyprogesterone (17‐OHP4), and ×3.47 for dehydroepiandrosterone (DHEA). Abbreviations: 17‐OHP4, 17α‐hydroxyprogesterone; 20α‐DHE, 20α‐dihydrocortisone; 20α‐DHF, 20α‐dihydrocortisol; allo‐THF, 5α‐tetrahydrocortisol; DHEA, dehydroepiandrosterone; DHEA‐S, dehydroepiandrosterone sulfate; E, cortisone; F, cortisol; HSD, hydroxysteroid dehydrogenase; P4, progesterone; THE, tetrahydrocortisone; THF, tetrahydrocortisol. p < 0.05 for all variables compared with the control group by post hoc analysis using Dunn's multiple comparison test. To minimize redundancy, the 22 candidates were refined into a panel of 14 representative steroid variables by selecting one from each highly intercorrelated variables (|ρ| ≥ 0.80) (Table S3). Among the 14 refined steroid variables, 9 representing the three principal enzymatic axes were selected for visualization to illustrate axis‐specific alterations across groups (Figs. 2 and 3). Markers reflecting 11β‐HSD activity were significantly reduced in the LC group compared with those in the control group (p < 0.001), whereas the AI group exhibited intermediate or variable patterns. Similarly, markers of A‐ring reduction catalyzed by 5α/5β‐reductase followed by 3α‐HSD were consistently lower in the LC group than in the control group (all p < 0.001). In contrast, the AI group showed partially preserved or heterogeneous profiles. In contrast to the above axes, 20α‐reduction markers were markedly elevated in the LC group. Notably, the AI group demonstrated the lowest levels across most of these 20α‐reduction markers (p < 0.001). Axis‐specific cortisol metabolic reprogramming in liver cirrhosis: (a) heatmap of 11β‐HSD axis markers (e.g., serum F/E and salivary 20α‐dihydrocortisol [20α‐DHF]/20α‐dihydrocortisone [20α‐DHE] ratios) showing suppression in the liver cirrhosis (LC) group compared with those in the control group and a distinct pattern from the adrenal insufficiency (AI) group. (b) Heatmap of A‐ring reduction markers (tetrahydrocortisol [THF]/F, tetrahydrocortisone [THE]/E, THF), reflecting 5α/5β‐reductase and 3α‐hydroxysteroid dehydrogenase (HSD) activity. LC demonstrates coordinated downregulation of A‐ring reduction pathways. (c) Heatmap of 20α‐HSD–mediated metabolites (serum and salivary 20α‐DHE and 20α‐DHF), showing marked activation in LC but suppression in AI. Values represent group‐level mean z‐scores. Red indicates relative upregulation, and blue indicates downregulation. Group differences were assessed using the Kruskal–Wallis test, followed by post hoc pairwise comparisons with Dunn's test. Statistical significance was annotated as follows: *p < 0.05, **p < 0.01, and ***p < 0.001. (d) Schematic summary of cortisol metabolic remodeling in liver cirrhosis relative to non‐cirrhotic controls. Cholesterol substrate (via hepatic high‐density lipoprotein [HDL]/lecithin–cholesterol acyltransferase, LCAT) and adrenocorticotropic hormone (ACTH) inputs are shown, together with the major enzymatic steps generating each metabolite. Colored steps denote the direction of change in cirrhosis compared with controls (blue, reduced; red, enhanced), highlighting reduced 11β‐HSD1 activity, suppressed A‐ring (5α/5β) reduction, and enhanced 20α‐reduction. DHEA‐S, dehydroepiandrosterone sulfate. Representative steroid variables distinguishing liver cirrhosis from controls: (a) peak serum 20α‐dihydrocortisone (20α‐DHE) levels demonstrating significant elevation in liver cirrhosis (LC); (b) peak serum tetrahydrocortisone (THE)/E ratio showing suppression of A‐ring reduction in LC; (c) peak salivary 20α‐dihydrocortisol (20α‐DHF)/20α‐DHE ratio reflecting altered 11β‐HSD–related interconversion in LC. Data are presented as box‐and‐whisker plots (median, interquartile range, and range). Then, iterative model refinement identified a parsimonious seven‐steroid signature beyond which the predictive performance did not improve (Fig. S3). Accordingly, a seven‐steroid model with high VIP scores comprising the peak THE/E ratio, peak 20α‐DHE, peak 20α‐DHF/F ratio, peak salivary 20α‐DHE, peak salivary 20α‐DHF/20α‐DHE ratio, DHEA‐S ratio, and P4 ratio was identified as the final steroid signature of LC. These alterations were driven primarily by LC status and were independent of sex (Table S4). Steroid reprogramming is associated with disease severity, but not with cholesterol Among patients in the LC group, those with Child–Pugh class B/C exhibited significantly worse liver function indices than those with Class A (all p < 0.05; Table S5). Although basal and peak cortisol concentrations did not differ according to severity, steroid reprogramming became more pronounced with worsening liver function (Tables S6 and S7). Compared with the Class A group, the Class B/C group showed significantly lower peak salivary 20α‐DHF/20α‐DHE ratios, reflecting altered 11β‐HSD activity, and reduced basal and peak THE levels, indicative of suppressed A‐ring reduction (p < 0.05). Conversely, the peak 20α‐DHF/F ratios, indicating 20α‐HSD activation, in serum and saliva were significantly elevated in the more severe group (p = 0.020). Spearman correlation analyses demonstrated that the Child–Pugh score was negatively correlated with the peak THE/E ratio (ρ = −0.41, p = 0.035) and positively correlated with peak 20α‐DHF/F (ρ = 0.42, p = 0.030) and peak salivary 20α‐DHE (ρ = 0.58, p = 0.002). MELD‐Na showed a concordant pattern, correlating with peak 20α‐DHF/F (ρ = 0.52, p = 0.005) and peak salivary 20α‐DHE (ρ = 0.41, p = 0.033), and was strongly correlated with the Child–Pugh score (ρ = 0.70, p < 0.001) (Fig. 4). These relationships were independent of sex and spironolactone use (Table S8). Association of seven key steroid variables with disease severity and clinical outcomes. Spearman correlation heatmap showing relationships between selected steroid variables and clinical severity indices (Child–Pugh score and model for end‐stage liver disease–sodium [MELD‐Na]). Correlation coefficients (ρ) are displayed within each cell; darker shading indicates stronger correlations. 20α‐DHE, 20α‐dihydrocortisone; 20α‐DHF, 20α‐dihydrocortisol; DHEA‐S, dehydroepiandrosterone sulfate; THE, tetrahydrocortisone; THF, tetrahydrocortisol. Cholesterol availability did not account for these changes: Within the cirrhosis cohort, none of the seven markers was associated with total, LDL, or HDL cholesterol after FDR correction (Table S9). The lipid–steroid correlations seen in the total cohort were not retained within the cirrhosis group. Discussion In this study, LC is characterized by coordinated, axis‐specific reprogramming of cortisol metabolism rather than isolated alterations in circulating cortisol concentrations. Although basal and stimulated cortisol concentrations were only modestly affected, comprehensive steroid profiling revealed consistent suppression of the 11β‐HSD axis and A‐ring reduction pathways, accompanied by marked activation of the 20α‐reduction pathway. These alterations were identified using multivariate modeling and were internally validated. Importantly, the derived seven‐steroid signature not only discriminated LC from controls but also correlated with disease severity. Taken together, these findings suggest that cortisol metabolic reprogramming represents a systemic and clinically relevant feature of advanced liver diseases. We observed reduced Peak F/E and salivary 20α‐DHF/20α‐DHE ratios, suggesting altered 11β‐HSD–mediated interconversion between cortisol and cortisone in LC. As 11β‐HSD1 activity depends on liver function and redox balance, progressive parenchymal loss and altered NAD+/NADH availability in cirrhosis may attenuate local cortisol regeneration [26]. In contrast to earlier stages of steatotic liver diseases, where 11β‐HSD1 activation has been described as a compensatory mechanism, advanced cirrhosis appears to reflect loss of this adaptive capacity, resulting in reduced systemic cortisol regeneration despite relatively preserved circulating cortisol concentrations [7]. In this study, markers reflecting A‐ring reduction, including THF/F, THE/E, and absolute THF and THE levels, were consistently suppressed in LC. Because 5α/5β‐reductase and 3α‐HSD are highly expressed in the liver and play a central role in cortisol inactivation and clearance [27], their downregulation likely reflects impaired hepatic metabolic capacity. Dynamic shifts in A‐ring reductase activity occur during the progression of metabolic dysfunction‐associated steatotic liver disease. 5α‐reductase is initially activated in steatosis to buffer cortisol; however, it declines in steatohepatitis, where concurrent 11β‐HSD1 upregulation locally increases cortisol levels to limit inflammation [7]. Concurrently, 5β‐reductase expression wanes with advancing disease [28]. Ultimately, in the cirrhotic phase, the combined suppression of 5α/5β‐reductase and 3α‐HSD reflects extensive hepatocellular loss and impaired hepatic perfusion [6]. This observation suggests that disease stage may have a greater influence on cortisol metabolic pathways than the underlying etiology in advanced cirrhosis. Reduced A‐ring reduction may prolong the systemic half‐life of active glucocorticoids, partially compensating for blunted cortisol responses, but ultimately signifies inefficient cortisol clearance in advanced disease [11]. In parallel, metabolites reflecting 20α‐reduction were markedly elevated in LC. This pathway has received limited attention in LC; however, Zumoff et al. described increased α‐cortolone levels in LC, a 20α‐reduction metabolite of THE, as in our study [6]. The selective upregulation of this pathway suggests a compensatory diversion of steroid flux toward alternative inactivation routes when conventional A‐ring reduction is impaired. Unlike A‐ring reduction, 20α‐HSD–mediated metabolism may be relatively preserved or even induced under inflammatory or oxidative stress conditions [6, 11]. Experimental data indicating that inflammatory signaling can induce 20α‐HSD expression provide a plausible mechanistic link, given the persistent systemic inflammation characteristic of advanced cirrhosis [29]. The concordant increase in serum and salivary 20α‐reduced metabolites supports true systemic pathway activation [6, 30, 31]. Moreover, the strong association between 20α‐reduction markers and both Child–Pugh score and decompensation risk suggests that this pathway reflects disease progression rather than a passive consequence of reduced hepatic mass. Our findings fit a multilevel model of cortisol dysregulation in cirrhosis, encompassing central HPA‐axis changes, adrenal output, and peripheral metabolism [32, 33, 34]. Consistent with the preserved free cortisol reported in advanced disease, basal salivary cortisol was maintained in our cohort despite lower total cortisol, and basal ACTH was elevated, suggesting an earlier compensatory phase. Within the cirrhosis cohort, however, the steroid signature tracked disease severity (Table S8) rather than cholesterol (Table S9), indicating that the reprogramming reflects peripheral metabolic remodeling rather than substrate‐limited steroidogenesis. Our data revealed compartment‐specific alterations in cortisol metabolism in LC. Compared with the control group, the LC group exhibited significantly lower serum F/E ratios, suggesting impaired hepatic 11β‐HSD1–mediated cortisol regeneration. In contrast, salivary F/E ratios did not differ, indicating that local 11β‐HSD2 activity is not substantially altered in early‐stage disease (Table S2). However, in the LC cohort, disease severity was associated with lower salivary F/E ratios, whereas serum F/E remained unchanged (Table S6). Because salivary F/E primarily reflects 11β‐HSD2 activity, this pattern suggests enhanced tissue‐level cortisol inactivation in advanced cirrhosis. Thus, LC progression appears to involve a shift from impaired systemic cortisol regeneration to increased peripheral cortisol inactivation, potentially reducing effective glucocorticoid availability, despite relatively stable circulating cortisol levels. Basal ACTH levels were elevated in the LC group, consistent with the compensatory activation of the hypothalamic–pituitary–adrenal axis [26]. Reduced hepatic regeneration and enhanced peripheral inactivation of cortisol may increase the central drive to maintain homeostasis, reflecting adaptive neuroendocrine activation rather than primary adrenal failure. However, our results reflect mainly hepatic 11β‐HSD1‐mediated cortisol regeneration and do not capture the renal compartment. In the kidney, 11β‐HSD2 inactivates cortisol to cortisone, and its activity is reflected by the urinary cortisol to cortisone ratio, which we did not measure. Urinary cortisol/cortisone profiling would help clarify how the hepatic and renal arms of the cortisol–cortisone metabolism are remodeled as cirrhosis progresses. A distinct feature of the cirrhosis profile was the dissociation between DHEA and its sulfoconjugate: Unconjugated DHEA was largely preserved, whereas DHEA‐S was substantially reduced, lowering the DHEA‐S/DHEA ratio relative to controls. Because DHEA‐S is generated by SULT2A1‐mediated sulfonation and circulating concentrations also depend on hepatic steroid handling, this pattern is most consistent with diminished sulfoconjugation in the setting of hepatic dysfunction, in line with the reduced hepatic DHEA‐sulfotransferase content reported in chronic liver disease [35], rather than with a primary failure of adrenal androgen synthesis. Notably, it is opposite to the profile of AI, in which DHEA falls disproportionately and the DHEA‐S/DHEA ratio rises, underscoring that the low DHEA‐S of cirrhosis is not merely a sign of adrenal hypofunction. Consistent with this, reduced DHEA‐S and a low DHEA‐S/cortisol ratio have been associated with greater disease severity, systemic inflammation, and mortality in cirrhosis [36], suggesting that this hepato‐adrenal axis may carry prognostic relevance that merits dedicated study. Although the LC group demonstrated blunted peak cortisol responses compared with controls, their steroid metabolic profiles differed fundamentally from those observed in classical AI. In the AI group, cortisol production and downstream metabolites were uniformly reduced, reflecting impaired adrenal steroidogenesis. In contrast, LC was characterized by coordinated suppression of A‐ring reduction and 11β‐HSD–mediated interconversion, accompanied by marked activation of the 20α‐reduction pathway. Notably, 20α‐reduced metabolites were elevated in LC but profoundly decreased in AI, indicating that the metabolic signature observed in cirrhosis cannot be attributed simply to reduced adrenal output. These findings suggest that functional glucocorticoid vulnerability in LC arises from dysregulated peripheral metabolism in addition to altered adrenal responsiveness. Accordingly, conventional definitions of RAI, primarily based on peak serum cortisol responses to cosyntropin, may not fully capture the metabolically complex alterations in glucocorticoid handling observed in cirrhosis [12]. This study has several strengths. First, comprehensive steroid profiling was applied across multiple enzymatic axes combined with CST, enabling the assessment of dynamic adrenal–hepatic metabolic responses rather than relying solely on static cortisol measurements. This approach allowed for the identification of coordinated axis‐level alterations that were not apparent under basal conditions. Second, the integration of salivary steroid analysis enhanced physiological relevance by capturing the biologically active free fraction of hormones in a population with altered binding proteins. The parallel activation of the 20α‐reduction pathway in serum and saliva supports true systemic metabolic remodeling rather than changes driven solely by protein binding. Third, a parsimonious seven‐steroid signature associated with disease severity was identified, with multivariate modeling indicating potential discriminative value. Finally, by including an AI group as a comparator, we show that steroid alterations in LC are a unique metabolic feature of liver disease, rather than a simple manifestation of secondary adrenal failure. These features support several clinically relevant inferences. Free or salivary cortisol may more accurately define adrenal function in cirrhosis and limit over‐diagnosis of relative AI, and the severity‐linked signature may serve as a prognostic biomarker. Notably, the most discriminating markers were metabolic ratios rather than absolute concentrations, by reflecting enzyme activity and minimizing confounding from binding proteins and hepatic clearance, which may be more robust in this setting. Reduced hepatic 11β‐HSD1 and A‐ring reduction also prolong glucocorticoid exposure, suggesting glucocorticoid dosing may require individualization in advanced disease. This study had several limitations. First, although healthy volunteers were not included, the control participants were evaluated under standardized cosyntropin‐stimulation conditions after postoperative stabilization. This design allowed for the comparison of dynamic adrenal–hepatic responses while minimizing confounding due to acute surgical stress. Their baseline serum steroid concentrations fell within published sex‐specific reference intervals for healthy adults, supporting their use as a physiological reference [37] (Table S10). Nevertheless, the absence of a community‐based healthy cohort remains a limitation. The sample size was modest, and salivary measurements were unavailable in six patients with more advanced diseases, which may have reduced statistical power. Notably, these individuals were among those with a more advanced disease severity (Table S11). Although the principal contrasts were internally validated with large effect sizes (Table S12), the findings remain exploratory, and external validation using a larger cohort is warranted. Second, although the seven‐steroid model showed potential discriminative value, the possibility of overfitting should be considered given the limited sample size, despite cross‐validation and permutation testing. Third, we did not directly measure the enzyme activity or hepatic expression of the implicated pathways. Although many of these enzymes are highly expressed in the liver, extrahepatic tissues, such as immune cells and adipose tissue, may also contribute to systemic steroid remodeling, particularly in the inflammatory milieu of advanced cirrhosis. Fourth, the cohort was relatively compensated, with approximately half classified as Child–Pugh A, which likely biased our estimates toward the null. As the steroid signature tracked disease severity within the cohort, more advanced disease would likely show more pronounced remodeling, suggesting our findings are conservative and warrant confirmation in decompensated cirrhosis. Nevertheless, coordinated, axis‐consistent alterations across multiple metabolites support structured metabolic remodeling rather than isolated fluctuations. Lastly, although altered protein binding or renal clearance could influence steroid concentrations in patients with cirrhosis, comparable renal function across groups and parallel changes in salivary steroids, reflecting the free fraction, argue against these factors as primary drivers of the observed patterns. In summary, our findings support a model in which LC is characterized by structured axis‐specific reprogramming of cortisol metabolism, involving impaired cortisol regeneration, reduced A‐ring reduction, and activation of the 20α‐reduction pathway. These findings highlight a systemic reprogramming of adrenal–hepatic steroid metabolism rather than a uniform suppression of steroid production. Although a parsimonious steroid signature was associated with disease severity, these observations should be interpreted as exploratory and require validation in larger, independent cohorts. Future studies are warranted to further elucidate the mechanistic basis and clinical implications of these metabolic alterations. Author contributors Conceptualization: Jung Hee Kim and Man‐Ho Choi. Data curation: Soyeon Park, Min Jeong Park, Hye Won Yang, Seung Shin Park, Su Jong Yu, Jeayeon Park, and Eun Ju Cho. Formal analysis: Jung Hee Kim. Funding acquisition: Jung Hee Kim and Man‐Ho Choi. Investigation: Soyeon Park, Min Jeong Park, Hye Won Yang, Su Jong Yu, Jeayeon Park, and Eun Ju Cho. Methodology: Soyeon Park, Min Jeong Park, Jung Hee Kim, Man‐Ho Choi, and Eun Ju Cho. Project administration: Jung Hee Kim, Soyeon Park, Min Jeong Park, Man‐Ho Choi, and Eun Ju Cho. Software: Jung Hee Kim, Soyeon Park, and Man‐Ho Choi. Supervision: Jung Hee Kim, Man‐Ho Choi, and Eun Ju Cho. Validation: Jung Hee Kim and Eun Ju Cho. Visualization: Soyeon Park, Jung Hee Kim, Man‐Ho Choi. Writing—original draft: Jung Hee Kim, Man‐Ho Choi, and Eun Ju Cho. Writing—review and editing: Soyeon Park, Min Jeong Park, Hye Won Yang, Seung Shin Park, Su Jong Yu, Jeayeon Park, Eun Ju Cho, Man‐Ho Choi, and Jung Hee Kim. Conflict of interest statement The authors declare no conflicts of interest. Supporting information Figure S1: joim70149‐sup‐0001‐figureS1.tif Figure S2A: joim70149‐sup‐0002‐figureS2A.tiff Figure S2B: joim70149‐sup‐0003‐figureS2B.tiff Figure S3: joim70149‐sup‐0004‐figureS3.png Table S1: Variable importance in projection (VIP) scores of steroid variables distinguishing LC from controls. Table S2: Comparison of serum and salivary steroid profiling among patients with liver cirrhosis (LC), control subjects, and adrenal insufficiency (AI). Table S3: Correlation matrix of 22 steroid variables. Table S4: Effects of liver cirrhosis (LC) and sex on steroid variables. Table S5: Baseline characteristics of patients with liver cirrhosis according to Child‐Pugh grade. Table S6: Comparison of major steroid metabolites and ratios by Child–Pugh grade in patients with liver cirrhosis. Table S7: Comparison of steroid metabolites and ratios by Child‐Pugh Grade in patients with liver cirrhosis. Table S8: Correlation analysis between steroid metabolites and clinical severity scores (Child‐Pugh, MELD‐Na). Table S9: Associations between the seven‐steroid signature markers and serum cholesterol in patients with liver cirrhosis and in the total cohort. Table S10: Baseline serum steroids in non‐cirrhotic controls, stratified by sex, versus Eisenhofer et al. 2017 sex‐specific LC–MS/MS reference intervals. Table S11: Comparison of baseline characteristics based on the availability of salivary steroid analysis. Table S12: Effect sizes for principal steroid markers, LC (n = 27) versus controls (n = 27). Acknowledgments This study was supported by a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI) and funded by the Ministry of Health and Welfare of the Republic of Korea (Project No. RS‐2024‐00406806 and RS‐2026‐25476606). This study was also supported by a grant from the Korea Institute of Science and Technology Institutional Program (Project no. 2E33742). Contributor Information Eun Ju Cho, Email: creatioex@gmail.com. Jung Hee Kim, Email: jhee1@snu.ac.kr. Data availability statement The data are not publicly available due to privacy and ethical restrictions but are available upon reasonable request to corresponding authors. References Associated Data Supplementary Materials Figure S1: joim70149‐sup‐0001‐figureS1.tif Figure S2A: joim70149‐sup‐0002‐figureS2A.tiff Figure S2B: joim70149‐sup‐0003‐figureS2B.tiff Figure S3: joim70149‐sup‐0004‐figureS3.png Table S1: Variable importance in projection (VIP) scores of steroid variables distinguishing LC from controls. Table S2: Comparison of serum and salivary steroid profiling among patients with liver cirrhosis (LC), control subjects, and adrenal insufficiency (AI). Table S3: Correlation matrix of 22 steroid variables. Table S4: Effects of liver cirrhosis (LC) and sex on steroid variables. Table S5: Baseline characteristics of patients with liver cirrhosis according to Child‐Pugh grade. Table S6: Comparison of major steroid metabolites and ratios by Child–Pugh grade in patients with liver cirrhosis. Table S7: Comparison of steroid metabolites and ratios by Child‐Pugh Grade in patients with liver cirrhosis. Table S8: Correlation analysis between steroid metabolites and clinical severity scores (Child‐Pugh, MELD‐Na). Table S9: Associations between the seven‐steroid signature markers and serum cholesterol in patients with liver cirrhosis and in the total cohort. Table S10: Baseline serum steroids in non‐cirrhotic controls, stratified by sex, versus Eisenhofer et al. 2017 sex‐specific LC–MS/MS reference intervals. Table S11: Comparison of baseline characteristics based on the availability of salivary steroid analysis. Table S12: Effect sizes for principal steroid markers, LC (n = 27) versus controls (n = 27). Data Availability Statement The data are not publicly available due to privacy and ethical restrictions but are available upon reasonable request to corresponding authors.