An impaired pituitary–adrenal signalling axis in stable cirrhosis is linked to worse prognosis

Background & Aims Inadequate adrenal function has been described in patients with cirrhosis. We investigated (i) the pituitary–adrenal axis at different clinical stages and (ii) the clinical impact of decreased serum cortisol levels in stable patients with advanced chronic liver disease (ACLD). Methods We included 137 outpatients with ACLD undergoing hepatic venous pressure gradient (HVPG) measurement in the prospective VICIS study (NCT03267615). Patients were stratified into six clinical stages: S0: subclinical portal hypertension (PH) (HVPG 6–9 mmHg), S1: clinically significant PH (HVPG ≥10 mmHg) without varices, S2: presence of varices, S3: previous variceal bleeding, S4: previous non-bleeding decompensation, and S5: further decompensation. Results Fifty-one patients had compensated ACLD (S0: n = 13; S1: n = 12; S2: n = 26), whereas 86 patients had decompensated ACLD (S3: n = 7; S4: n = 46; S5: n = 33). Serum total cortisol (t-Cort) showed a strong correlation with estimated serum free cortisol (f-Cort; Spearman’s ρ: 0.889). With progressive clinical stage, median ACTH levels (from S0: 44.0 pg/ml to S5: 20.0 pg/ml; p = 0.006), t-Cort (from S0: 13.9 μg/dl to S5: 9.2 μg/dl; p = 0.091), and cortisol binding globulin (from S0: 49.3 μg/ml to S5: 38.9 μg/ml; p <0.001) decreased, whereas f-Cort (p = 0.474) remained unchanged. Lower t-Cort levels independently predicted bacterial infections (asHR: 1.11; 95% CI: 1.04–1.19; p = 0.002), further decompensation (asHR: 1.08; 95% CI: 1.02–1.12; p = 0.008), acute-on-chronic liver failure (ACLF; asHR: 1.11; 95% CI: 1.04–1.19; p = 0.002), and liver-related death (asHR: 1.09; 95% CI: 1.01-1.18; p = 0.045). Conclusions The pituitary–ACTH–adrenal–cortisol axis is progressively suppressed with increasing severity of cirrhosis. Lower t-Cort is an independent risk factor for bacterial infections, further decompensation of ACLF, and liver-related mortality—even in stable outpatients with cirrhosis. Clinical trial number Vienna Cirrhosis Study (VICIS; NCT: NCT03267615). Impact and Implications In a cohort of stable outpatients, we observed progressive suppression of the pituitary–adrenal axis with increasing clinical stage of advanced chronic liver disease (ACLD). Increased levels of bile acids and systemic inflammation (assessed by interleukin-6 levels) could be involved in this suppression. Serum total cortisol (t-Cort) was strongly correlated with serum free cortisol (f-Cort) and lower t-Cort levels were independently associated with a higher risk of adverse clinical outcomes, including bacterial infections, further decompensation, acute-on-chronic liver failure, and liver-related death.

Introduction and is frequently triggered by precipitating events, mostly alcoholic hepatitis and bacterial infections. 12 Generally, patients with ACLD are at increased risk of serious infections. 13 Although different factors are involved, including bacterial translocation, 14 intestinal dysbiosis and the development of cirrhosis-associated immune deficiency, 15 dysfunction of the pituitary-adrenal axis may represent another predisposing factor.
Altered cortisol metabolism in cirrhosis has already been described in 1967, 16 linking chronic liver disease to adrenocortical dysfunction. One recent study investigated adrenal function in patients with ACLD hospitalised for acute decompensation, finding that relative adrenal insufficiency was frequent in these patients and had similar impact on the prognosis as did other organ failures as defined by ACLF. 17 Accordingly, another study indicated decreased adrenal function by exploring decreased maximal cortisol secretion rates in patients with ACLD and particularly in patients with Child-Turcotte-Pugh (CTP) stages B and C. 18 A small study indicated that this adrenal insufficiency may be a result of dysregulated pituitary signalling, as direct adrenal stimulation with adrenocorticotropic releasing hormone (ACTH) yielded significantly better results than indirect stimulation. 19,20 Mechanistically, elevated bile acid levels 21,22 and chronic inflammation 23 may impair the hypothalamus-pituitary-adrenal signalling axis.
Nonetheless, abnormalities of the pituitary-adrenal axis have not been systematically investigated in the different substages of cirrhosis. Thus, this study aimed (i) to investigate the pituitaryadrenal axis throughout individual clinical stages of ACLD 6 and (ii) to determine the impact of serum total cortisol (t-Cort) and estimated serum free cortisol (f-Cort) levels on clinical outcomes of ACLD.

Patients and methods
Study design This prospective study investigated consecutive patients with ACLD and portal hypertension (i.e. HVPG > − 6 mmHg) who underwent HVPG measurement at the Vienna Hepatic Hemodynamic Laboratory (Division of Gastroenterology and Hepatology, Medical University of Vienna) between June 2018 and March 2020 under standardised conditions. 24 We only included clinically stable outpatients in whom data on t-Cort and ACTH were available.
Patients with occlusive portal vein thrombosis (PVT), a history of vascular liver disease, hepatocellular carcinoma (HCC), liver transplantation (LT), transjugular intrahepatic portosystemic shunt (TIPS), human immunodeficiency virus, or active infection at the time of HVPG measurement were excluded. Additional exclusion criteria included: therapy with corticosteroids or other medications that influence steroidogenesis, ongoing significant 25 self-reported alcohol intake at characterisation, and HVPG measurement while on non-selective beta blocker treatment.
HVPG measurements and central venous blood withdrawals were conducted under fasting conditions to evaluate cortisol axis parameters. Information was obtained on the development of clinically apparent ascites, variceal bleeding, hepatic encephalopathy (HE), ACLF, acute kidney injury (AKI), and HCC, as well as LT and (liver-related) death during clinical follow-up (FU).

Patient cohorts
For the baseline characterisation of the pituitary-adrenal axis in patients with ACLD, patients were stratified according to ACLD severity and clinical stages as proposed by D'Amico et al. in 2018. 5,6 The absence of any previous decompensating event defined cACLD, which was further subdivided into three distinct stages (S): S0: subclinical portal hypertension (PH; HVPG 6-9 mmHg), S1: clinically significant PH (i.e. HVPG > − 10 mmHg) without the presence of varices, and S2: presence of varices. In contrast, patients with a history of at least one decompensating event were classified as patients with dACLD with three substages: S3: history of decompensation from acute variceal bleeding, and S4: non-bleeding decompensation events including ascites and HE, as well as S5: further decompensation defined as history of two different types of decompensation events.

Decompensation events
For patients with cACLD, decompensation was defined as the occurrence of moderate or large ascites, variceal bleeding, or clinically overt hepatic encephalopathy. In patients with dACLD, further decompensation was defined as the occurrence of a second type of hepatic decompensation or worsening of the first hepatic decompensation. Worsening included development of refractory ascites or spontaneous bacterial peritonitis in patients with ascites, first hospital admission with HE in patients with H3, and variceal rebleeding in patients with a history of variceal haemorrhage. Finally, ACLF as defined by the EASL-Chronic Liver Failure (EASL-CLIF) consortium definition 8,26 was regarded as a fourth and distinct type of decompensation event, which could potentially occur in both cACLD, and dACLD patients.
Assessment of HVPG and liver stiffness measurement HVPG measurements were conducted according to a standardised operating procedure, as previously described. 24 Briefly, after local anaesthesia, a catheter introducer sheath was positioned in the right internal jugular vein. Subsequently, a specifically designed balloon catheter 27 was inserted under fluoroscopic guidance into a large hepatic vein. HVPG was calculated by subtracting the free pressure from the wedged hepatic vein pressure. A total of three measurements were performed per patient and the mean HVPG of these measurements was used for further analyses.
As previously described, liver stiffness measurement (LSM) was assessed by vibration-controlled transient elastography (FibroScan ® ; Echosens, Paris, France) 28 under fasting conditions. Research article ml) was calculated from t-Cort and CBG levels as previously described. 30 Bile acids (normal range: 0.0-10.0 lmol/L; consisting of unconjugated acids, taurine-and glycine-conjugated acids) were measured by high-performance liquid chromatography-high-resolution mass spectrometry (HPLC-HRMS; Q Exactive MS/MS, ThermoFisher Scientific, Waltham, MA, USA) as previously described. 31,32 Standard laboratory methods were used for the assessment of routine laboratory parameters.

Cortisol axis parameters and routine laboratory parameters
Blood withdrawals were conducted in a standardised manner after HVPG measurement under fasting conditions and after resting in the supine position for at least 30 min. All blood samples for assessment of the parameters of the pituitaryadrenal axis were taken between 9 am and 12 noon.

Statistical analysis
For categorical variables, the number (n) and proportion (%) of patients displaying the parameter of interest were reported. The median and IQR were used to present continuous data. The Mann-Whitney U test was implemented for comparison of continuous non-normally distributed variables between two different groups. The Kruskal-Wallis test was performed for comparison of continuous variables in three of more groups. Pearson's Chi squared test and Fisher's exact test were used for group comparisons of categorical variables, as appropriate. Correlations were assessed using Spearman's Rho (q).
Clinical outcomes at the 3-year follow-up were assessed by cumulative incidence calculation. Cumulative incidence comparison was performed using Gray's test, as previously described. 33 Fine and Gray competing risk regression models were calculated using the R package cmprsk 33,34 (R Foundation for Statistical Computing, Vienna, Austria) to evaluate whether t-Cort and f-Cort levels were associated with the risk of clinical events of interest. Apart from t-Cort or f-Cort, well-established risk factors for worse outcomes in ACLD (i.e. age, CTP score, serum creatinine, sodium, HVPG as a marker for portal hypertension, and C-reactive protein (CRP) as a parameter of systemic inflammation) and sex were evaluated by univariate and multivariate analysis. Parameters that conferred prognostic value in univariate regression analysis (as evidenced by p <0.100) were carried forward to multivariate regression analysis. The exception was f-Cort as a key parameter of interest, which was carried forward to test its potential stage-specific prognostic information after adjustment for relevant parameters (particularly liver function, i.e. CTP score). LT and non-liver-related death were considered as competing risks for decompensation/further decompensation and liver-related death, whereas LT and death were considered competing events for bacterial infections and ACLF for both the cumulative incidence comparison and the competing risk regression. Youden's index was used to determine the optimal cut-off values for t-Cort and f-Cort for the prediction of ACLF, as a close link between this critical condition and adrenal function has been previously reported. 17 Statistical analyses were performed using GraphPad Prism v8 (GraphPad Software, La Jolla, CA, USA), IBM SPSS v27.0 statistic software (IBM, Armonk, NY, USA) and R v4.2.1 (R Core Team). A two-sided value of p <0.05 was considered statistically significant.

Ethics
The study was approved by the Ethics Committee of the Medical University of Vienna (No. 1262/2017). It was performed in accordance with the current version of the Declaration of Helsinki. All included patients are part of the Vienna Cirrhosis Study (VICIS; NCT: NCT03267615) and gave their written informed consent before inclusion in the study.

Patient characteristics
In total, 137 patients with ACLD (67.9% males) and a median age of 58.1 years were included in the study. Fig. 1 (Fig. 3). Furthermore, corticosterone levels did not differ across the clinical stages (from S0: 6.0 ng/ml to S5: 6.7 ng/ ml; p = 0.826), whereas aldosterone levels increased, particularly in patients with ACLD with non-bleeding decompensation (from S0: 78.0 pg/ml to S4: 280.0 pg/ml and to S5: 262.0 pg/ml; p <0.001). Table S2 shows the levels of parameters of the pituitaryadrenal axis in cACLD compared with dACLD.

Discussion
In this prospective exploratory study, we investigated the pituitary-adrenal axis in a thoroughly characterised cohort of stable outpatients with ACLD. We observed a decline in ACTH and t-Cort levels with the progressive decline in clinical stage of ACLD. Importantly, low levels of levels of t-Cort and f-Cort were independent risk factors for clinical events when analysed as continuous and dichotomised variables. This suggests that even in stable outpatients with ACLD, decreased levels of cortisol may be of clinical and prognostic relevance.
Dysregulation of cortisol metabolism and homeostasis have been described in patients with ACLD. These include reduced 11b-hydroxysteroid dehydrogenase type 2 activity with consecutively increased circulating glucocorticoids, 35,36 decreased elimination rates of free cortisol, as well as reduced maximal cortisol secretion rates. 18 Previous studies have mostly focused on adrenal function in patients with cirrhosis and acute decompensation and/or in those admitted to intensive care: Inadequate adrenal response, as assessed by a short synacthen test (SST), was often reported in acutely decompensated patients with cirrhosis. Relative adrenal insufficiency was observed in a significant share of patients (between 24% and 60% of cirrhotic patients) 17,37-39, and acutely hospitalised patients with relative adrenal insufficiency showed inferior clinical outcomes with decreased survival, 37 and these patients more often experienced decompensation events, including bleeding and hepatorenal syndrome. 40 However, the presence and clinical relevance of decreased t-Cort and f-Cort levels have not yet been specifically investigated in stable outpatients with ACLD. In this study, we found a decrease in t-Cort levels with the progressive severity of ACLD. It should be noted that 90% of circulating cortisol is bound to CBG and, to a lesser extent, to albumin, which are both produced in the liver and thus decreased in patients with ACLD. 20,41 Accordingly, CBG levels decreased significantly throughout all clinical stages of ACLD in our study, and the resulting median levels of f-Cort were not significantly different between different stages of ACLD, suggesting a maintained homeostasis of cortisol in many patients with ACLD, even in more advanced clinical stages. This is in line with previous data showing an absence of dynamics of the biologically free serum cortisol fraction across patients in different CTP stages. 18,42 Reduced elimination of f-Cort in patients with more severe ACLD 18 may be involved in the largely maintained levels of f-Cort.
However, both low t-Cort and low f-Cort indicated an increased risk of subsequent clinical complications, especially bacterial infections and ACLF, possibly indicating altered cortisol secretion with an impaired stress response in a subgroup of patients with ACLD. Interestingly, low f-Cort and t-Cort are related to conditions of chronic inflammation. 23 Low f-Cort and t-Cort are observed more frequently in patients with severe ACLD. Specifically, f-Cort <4.8 ng/ml was twice as prevalent in patients with ACLD substage S5 as in substage S0 (42.4% vs. 21.4%), whereas comparable findings were obtained for f-Cort <12 lg/dl (Fig. 1).
Notably, since t-Cort and f-Cort showed a strong correlation and similar prognostic utility, the assessment of t-Cort may be the preferable and more straight-forward surrogate marker for adrenal function in stable outpatients with ACLD without current infection or acute decompensation. Patients with t-Cort <12 lg/ dl or f-Cort <4.8 ng/ml appear to be at particularly higher risk of subsequent development of bacterial infections and ACLF. Further studies are required to determine whether these patients may benefit from specific therapeutic interventions, such as antibiotic prophylaxis and/or closer clinical monitoring.
The observed decline in ACTH with progressing substages of ACLD is consistent with a previous study showing that direct stimulation of the adrenal gland through ACTH administration led to increased cortisol responses compared with indirect stimulation, 20 whereas another study found frequent  impairment of the pituitary-adrenal axis through the corticotropin releasing hormone (CRH) stimulation test, particularly in patients with CTP stage B and C. 43 Hence, there is evidence that impaired adrenal function in ACLD is at least partially caused by a dysfunction of stimulating pituitary-adrenal signalling. Bile acid signalling might have been a factor involved in the pathogenesis of the adrenal dysfunction observed in this study.
Bile acids modify cortisol homeostasis by suppressing CRH excretion 21 and influencing cortisol metabolism and catabolism through farnesoid X receptor (FXR) signalling. 44 Furthermore, although acute inflammation is linked to activation of the hypothalamus-pituitary-adrenal axis, chronic inflammation seems to have an opposite effect, suppressing hypothalamus-pituitaryadrenal signalling. 23 Indeed, our data showed an increase of both systemic inflammation (reflected by IL-6), 5 as well as systemic bile acid levels with a more advanced clinical stage. These two factors may interfere with cortisol homeostasis and pituitaryadrenal signalling with increasing severity of ACLD.
Although previous research indicated that decreased levels of total and HDL cholesterol are involved in the impairment of adrenal function in ACLD, 17,45 in our study, total and HDL cholesterol levels were not significantly different in different substages. This might be because, we included stable outpatients with mostly preserved liver function (i.e. primarily CTP stages A and B).
Our study also has limitations. First, we did not perform SST or sequential measurements in each individual patient. Thus, we cannot investigate relative adrenal insufficiency or SST responses in this study population. Second, the adequacy of t-Cort measurements has been debated in patients with cirrhosis for various reasons, including assay dependency 46,47    validated reference values specific to ACLD or hypoproteincorrected reference values. 47 Of note, our study found robust results for t-Cort and t-Cort also correlated significantly with f-Cort. However, we understand that external validation is required for the proposed f-Cort and t-Cort cut-offs at <4.8 ng/ml and <12 lg/dl, respectively. Moreover, the direct measurement of serum free cortisol is costly and complex, limiting its use in clinical practice. In this study, we used Coolen's formula, a wellestablished and widely used formula to calculate f-Cort based on t-Cort and CBG. However, this formula has limitations, particularly in patient groups with expected variations of CBG and albumin. More refined equations adjusted for albumin levels may be more precise in estimating serum-free cortisol 48 in patients with ACLD, but may also be more difficult to use in clinical practise. Notably, we did not include patients with regular night working hours, or individuals with undisclosed co-factors that may have impacted on circadian rhythm. Our patients exhibited a relatively high median mean arterial pressure (MAP). This underlines that this study involved a stable patient population with ACLD without acute decompensation or critical illness and should be interpreted as such. As ALD was the main liver disease aetiology in our patient cohort and because chronic alcohol consumption influences cortisol homeostasis, 35 it should be emphasised that most (i.e. 84.4%) of the included patients with ALD were abstinent at the time of characterisation.
Most patients in stage 4 and 5 ACLD were treated with diuretics, specifically aldosterone receptor antagonists that may have likely affected aldosterone levels. 49,50 Finally, our observational study cannot establish a causative association between decreased levels of cortisol and adverse clinical outcomes. Thus, it is unclear whether decreased levels of cortisol are solely prognostic biomarkers or if they may serve as a therapeutic target.
In conclusion, our study demonstrates that pituitary ACTH and adrenal t-Cort levels decrease with progressive clinical substages of ACLD, suggesting suppression of the hypothalamicpituitary-adrenal axis, particularly in patients with dACLD. Increased levels of systemic inflammation and bile acids might be associated with the suppression of pituitary-adrenal cortisol Table 3. Impact of serum total cortisol on the risk of (i) bacterial infections, (ii) decompensation/further decompensation, (iii) acute-on-chronic liver failure (ACLF) and (iv) liver-related death. Research article signalling. Importantly, low levels of t-Cort and low levels of f-Cort are associated with worse clinical outcomes and indicate a higher risk of subsequent development of bacterial infections, further decompensation, ACLF, and liver-related death.

Parameter of interest
Additional studies are required to determine whether specific therapeutic interventions or closer clinical monitoring are warranted in patients with ACLD and low levels of f-Cort or t-Cort.

Financial support
No specific funding was received for this study.  Please refer to the accompanying ICMJE disclosure forms for further details.

Authors' contributions
All authors contributed either to research design (LH and TR) and/or the acquisition (LH, BSim, MJ, DB, RP, BSch, LB, GS, MM, and TR), analysis (LH, PW, MM, and TR) or interpretation (all authors) of data. LH and TR drafted the manuscript, which was critically revised by all other authors.

Data availability statement
The data derived from this study are available upon reasonable request to the corresponding author.        Table-S3. Impact of serum free cortisol on the risk of (i) bacterial infections, (ii) decompensation/further decompensation, (iii) acute-onchronic liver failure (ACLF) and (iv) liver-related death. Univariate and multivariate multivariate competing risk regression models are shown.

Supplementary data
Liver transplantation and death were considered as competing risks for (i) and (iii), while liver transplantation and non -liver-related death were considered as competing risks for (ii) and (iv). Adjusted subdistribution hazard ratio (asHR) with 95% confidence interval (95%CI) and significance levels of multivariate Fine and Gray competing risk regression models for serum total cortisol: (i) asHR: 1.14 (95%CI 1.02-1.25), p=0.014; (ii) asHR: