Original article 1125
Higher free serum cortisol is associated with worse survival in acute variceal bleeding because of cirrhosis: a prospective study Christos K. Triantosa, Maria Kalafatelia, Dimitrios Samonakise, Konstantinos Zisimopoulosa, Nikolaos Papiamonise, Apostolos Sapountzisa, Marina Michalakib, Georgios Theocharisa, Konstantinos Thomopoulosa, Chryssoula Labropoulou-Karatzac, Venetsanea Kyriazopouloub, Eleni Jelastopulud, Elias A. Kouroumalise, Vasiliki Nikolopouloua and Andrew K. Burroughsf Background and aims Critical illness-related corticosteroid insufficiency has been reported in acute variceal bleeding (AVB). In cirrhosis, free serum cortisol (FC) is considered optimal to assess adrenal function. Salivary cortisol (SC) is considered a surrogate for FC. We evaluated FC and its prognostic role in AVB. Methods Total serum cortisol, SC, cortisol-binding globulin, and FC (Coolens’ formula) were evaluated in AVB (n = 38) and in stable cirrhosis (CC) (n = 31). A Cox proportional hazards model was evaluated for 6-week survival. Results In AVB, the median FC and SC levels were higher with worse liver dysfunction [Child–Pugh (CP) A/B/C: 1.59/2.62/3.26 μg/dl, P = 0.019; CPA/B/C: 0.48/0.897/1.81 μg/ml, P < 0.001, respectively]. In AVB compared with CC, median total serum cortisol: 24.3 versus 11.6 μg/dl (P < 0.001), SC: 0.86 versus 0.407 μg/ml (P < 0.001); FC 2.4 versus 0.57 μg/dl (P < 0.001). In AVB, 5-day rebleeding was 10.5%, and 6-week and total mortality were 21.1 and 23.7%, respectively. Independent associations with 6-week mortality in AVB were FC at least 3.2 μg/dl (P < 0.001), hepatocellular carcinoma (P < 0.001), CPC (P < 0.001), and early rebleeding (P < 0.001). Among patients with normal cortisol-binding globulin (n = 14) and albumin (n = 31), the
Introduction Relative adrenal insufficiency (RAI) is defined as an inadequate glucocorticoid activity relative to the severity of illness. RAI has been reported in patients with cirrhosis and sepsis [1] or septic shock [2] and also in patients with acute liver failure [3,4] with no apparent sepsis. Treatment with hydrocortisone has been assessed in acute liver failure [4,5] and in cirrhosis with septic shock [2,6], and is associated with shock reversal and reduction in vasopressor doses [2,4–6]; however, no
All supplementary digital content is available directly from the corresponding author. 0954-691X © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins
factors were hepatocellular carcinoma (P = 0.003), CP (P = 0.003), and FC (P = 0.036). SC was also found to be an independent predictor of 6-week mortality (P < 0.001). Area under the curve of FC for predicting 6-week mortality was 0.79. Conclusion Higher FC is present in cirrhosis with AVB compared with CC and is associated independently with bleeding-related mortality. However, whether high FC solely indicates the severity of illness or whether there is significant adrenal insufficiency cannot be discerned. Eur J Gastroenterol Hepatol 26:1125–1132 © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins. European Journal of Gastroenterology & Hepatology 2014, 26:1125–1132 Keywords: acute variceal bleeding, adrenal insufficiency, free cortisol, liver cirrhosis, salivary cortisol Departments of aGastroenterology, bEndocrinology, cInternal Medicine, University Hospital of Patras, dDepartment of Public Health, Medical School, University of Patras, Patras, eDepartment of Gastroenterology, University Hospital of Heraklion, Heraklion, Greece and fThe Royal Free Sheila Sherlock Liver Centre, Royal Free Hospital, London, UK Correspondence to Christos K. Triantos, MD, D. Stamatopoulou 4, Rio, 26504 Patras, Greece Tel: + 30 697 289 4651; fax: + 30 261 062 5382; e-mail: [email protected] Received 26 March 2014 Accepted 11 June 2014
improvement in mortality has been shown [2,5,6]. Adrenal impairment has also been observed in patients with stable cirrhosis (CC) [7–14] and with decompensation, other than sepsis, such as ascites [15] or variceal bleeding [16,17]. It has been suggested [18] that adrenal dysfunction could be a feature specific to liver disease, which becomes manifest with precipitating factors such as infection, bleeding, or acute illness, but the underlying mechanism remains unknown. However, assessment of adrenal function in cirrhosis is difficult and the results are difficult to interpret [18,19]. More than 90% of circulating cortisol is bound to albumin and cortisol-binding protein (CBG) [20–22], whereas the DOI: 10.1097/MEG.0000000000000158
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1126 European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 10
remaining unbound (free) fragment is the biologically active substance. In cirrhosis, the reduced concentrations of albumin and CBG [7] may lead to an overestimation of adrenal insufficiency (AI) when total serum cortisol is used to assess adrenal function [23], especially when albumin concentrations are lower than 2.5 g/dl [24]. Plasma free cortisol measurement can be assessed indirectly using Coolens’ formula [25] or using surrogate markers such as salivary cortisol [9,23]. Furthermore, there is no consensus to define the most sensitive adrenocorticotropic hormone (ACTH) stimulation test for the diagnosis of AI in patients with liver disease, and the cutoff values of total, salivary or free cortisol have not been standardized as yet [18]. Acute variceal bleeding in patients with cirrhosis is a lifethreatening complication of portal hypertension, with a current 6-week mortality rate of ∼ 20% [26,27], although survival has improved over recent decades. It is a critical condition with many similarities to sepsis and is frequently associated with bacterial infections [28,29]. Predictive factors of survival include severity of liver disease, shock at admission, active bleeding at endoscopy, and severity of portal hypertension (measured by hepatic venous pressure gradient) [30]. We have recently shown the presence of an inadequate glucocorticoid response with respect to the severity of disease [critical illness-related corticosteroid insufficiency (CIRCI)] in this setting [16]; in another study, RAI has been associated with therapeutic failure and death within 6 weeks in patients with severe variceal bleeding (systolic blood pressure < 100 mmHg or/and hemoglobin < 8 g/dl) [17]. However, in both studies, adrenal function was assessed by measuring serum total cortisol after the standard short synacthen test (SST) or the low-dose short synacthen test (LDSST), and this may overestimate the prevalence of AI, as mentioned before. Evidence of adrenal dysfunction is needed on the basis of free cortisol estimations, and preliminary evidence of a possible association with adverse outcomes is required. Therefore, we carried out this prospective study in acute variceal bleeding to evaluate baseline free cortisol concentrations (assessed by Coolens’ formula), as well as salivary cortisol and evaluated associations with survival.
Patients and methods Patients
This prospective, multicenter, observational study was carried out at the University Hospital of Patras (Patras, Greece) and the University Hospital of Heraklion (Heraklion, Greece), and planned at the Royal Free Sheila Sherlock Liver Centre (London, UK); it was approved by the local hospital ethics committee. Written consent was obtained from all patients. We finally evaluated 38 consecutive patients with cirrhosis and variceal bleeding (two were excluded from the study because of disturbance in consciousness level and thus, consent form
could not be obtained) and 31 consecutive patients with cirrhosis attending the hepatology outpatient clinics, included from March 2010 until December of 2011. Patients were followed up until December 2013. Patients with a history of hypothalamic-pituitary or adrenal disease, treatment with corticosteroids within the past 12 months, infection with HIV, and hepatocellular carcinoma (HCC) beyond the Milan Criteria were excluded from the protocol. The severity of liver disease was assessed by the Child–Pugh (CP) class and model for end-stage liver disease at the time of adrenal function assessment. Patients with variceal bleeding were treated with packed red blood cells and plasma expanders to maintain hemodynamic stability and hemoglobin levels at ∼ 8 g/dl [31,32]. Management also included intravenous somatostatin for 5 days, antibiotics (intravenous ceftriaxone or ciprofloxacin), lactulose (when encephalopathy occurred), and endoscopic treatment, with variceal band ligation performed within 12 h of admission [32]. None of the patients received intravenous terlipressin or albumin infusions. The presence of infection was investigated with blood and urine cultures, paracentesis and culture of ascites, and chest radiography. Serum and salivary cortisol were measured the morning after admission (between 8:00 and 9:00 a.m.). Salivary samples were not collected in patients with blood contamination or fungal infection of the oral cavity, presence of a nasogastric tube, endotracheal intubation, or if the time of measurement was close to hematemesis or endoscopic procedure. In CC, serum and salivary cortisol were also measured between 8:00 and 9:00 a.m. at the outpatient clinics and saliva was collected at least 30 min after tooth-brushing and patients were advised to avoid smoking or drinking before saliva collection. Saliva was collected using dedicated kits (Plain Salivette; Sarstedt, Newton, North Carolina, USA). The cotton provided was chewed for 2–3 min and then collected in a plastic tube. Saliva was then collected by centrifugation. Salivary, serum total cortisol (normal range: 6.2–19.4 μg/dl), and plasma ACTH levels were measured using an electrochemiluminescence immunoassay (Roche Diagnostics Ltd, Rotkreuz, Switzerland). Serum CBG concentrations (normal range: 35–50 μg/ml) were assessed by radioimmunoassay (DIAsource Immunoassays S.A., Nivelles, Belgium). Salivary cortisol was available from 23 of 38 patients with variceal bleeding (10 patients had oral blood contamination, three had a nasogastric tube, and two had inadequate amount of collected saliva). Twentyseven CC had available salivary cortisol (four had insufficient salivary samples collected). Calculated plasma free cortisol
Calculated free cortisol was derived using Coolens’ formula [25]: U2K (1 + N) + U [1 + N + K (T − C)] − C = 0, where U represents the molar concentration of unbound
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Free cortisol and variceal bleeding Triantos et al. 1127
cortisol, C the molar concentration of total cortisol, and T the concentration of CBG. K corresponds to the affinity of CBG for cortisol at 37°C (equal to 3 × 10–7/mol/l) and N to the proportion of albumin-bound to unbound cortisol (1.74 is the value used conventionally). Definitions
Cirrhosis was diagnosed on the basis of histological findings, clinical evaluation, laboratory data, or imaging findings compatible with a diagnosis of cirrhosis. Variceal bleeding was diagnosed endoscopically on the basis of the presence of hematemesis or melena with either a bleeding varix (active bleeding or clot adherent to varix or variceal ulceration) or when there was no other visible lesion. Severe bleeding was defined as arterial hypotension (systolic blood pressure < 100 mmHg) or/and hemoglobin less than 8 g/dl at admission. Rebleeding was defined, according to the Baveno V criteria [32], as recurrence of hemorrhage after the first 24 h and within 5 days of admission for the bleeding episode, following initial control of bleeding, that is during the period of somatostatin infusion [33]. Bleeding-related mortality was defined as death within 6 weeks from the bleeding episode [32]. Groups
In this study, there were 69 patients with liver cirrhosis: 38 with variceal bleeding (group A) and 31 with CC (group B), median follow-up: 11 months (0.2–22). Statistical analysis
Numerical data were expressed as median, SD, and range (minimum to maximum) and categorical data as counts and percentages. All variables were tested for normal distribution using the Kolmogorov–Smirnov test. Categorical variables were tested using the χ2 and Fisher’s exact test. Continuous variables with and without a normal distribution were compared using Student’s t-test or the Mann–Whitney U-test, respectively. The discriminative ability of calculated serum free cortisol to predict 6-week mortality was assessed by receiver operating characteristic analysis. The value of free cortisol that maximized the sum of sensitivity and specificity was used as the best cut-off. Univariate analysis was used to determine potential independent predictors of 6-week survival in both groups. The statistical significance of differences in outcome among the different groups was assessed using the log-rank test. Variables significant at the 0.1% level were included in the age-adjusted multivariate analysis using binary logistic regression analysis. All statistical tests were two-sided and significance level was set at 0.05 or less. The results are presented as hazard ratios with 95% confidence intervals in multivariate analyses. The SPSS statistical package (version 19.0 for Windows; SPSS Inc., Chicago, Illinois, USA) was used.
Results Patients’ characteristics
The baseline clinical and laboratory characteristics for each group of patients are shown in Table 1. Both groups were similar at baseline in terms of age, sex, and etiology of liver disease. Thirteen (34.2%) patients with variceal bleeding had active bleeding at endoscopy and 17 (42.1%) had severe bleeding. The median packed red blood cells transfused were 3 (range 0–7), whereas a median of 2 (range 0–5) were administered within the first 24 h following admission. The results of adrenal function parameters assessed are shown in Table 2. Adrenal function in patients with variceal bleeding versus stable patients with cirrhosis
In patients with variceal bleeding, the median values of baseline serum free cortisol and salivary cortisol were significantly higher in CPC compared with CPB and CPA: free cortisol-CPA: 1.59 ± 0.53 versus CPB: 2.46 ± 2.61 versus CPC: 3.33 ± 6.5 μg/dl (Fig. 1); salivary cortisol-CPA: 0.48 ± 0.25 versus CPB: 0.897 ± 0.69 versus CPC: 1.81 ± 1.03 μg/ml (P < 0.001). Serum total cortisol and CBG were not significantly associated with CP class. Patients with severe variceal bleeding (n = 17) showed no differences in adrenal function parameters compared with patients without severe bleeding (total serum cortisol, P = 0.419; salivary cortisol, P = 0.944, free cortisol, P = 0.466; CBG, P = 0.643). In CC, no statistically significant difference was observed in total serum cortisol, salivary cortisol, CBG, and serum free cortisol in relation to CP stage. Comparing variceal bleeders (VB) with CC, serum total, salivary, and free cortisols were significantly higher in variceal bleeding: total cortisol: 24.3 ± 14.8 versus 11.6 ± 4.3 μg/dl, P < 0.001; salivary cortisol: 0.86 ± 0.85 versus 0.407 ± 0.25 μg/ml, P < 0.001; free cortisol: 2.4 ± 4.05 versus 0.57 ± 0.32 μg/dl, P < 0.001. Among the subgroup of VB (n = 14) and CC (n = 21) with baseline serum albumin greater than 2.5 mg/dl and CBG greater than 35 μg/ml levels, total, salivary, and free cortisol levels were also significantly higher in VB compared with CC: total cortisol: 23.68 ± 12.98 versus 12.2 ± 3.27 μg/dl, P < 0.001; salivary cortisol: 0.788 ± 0.709 versus 0.426 ± 0.22 μg/ml, P = 0.005; and free cortisol: 1.75 ± 4.04 versus 0.57 ± 0.3 μg/dl, P < 0.001, respectively. Free cortisol levels were also significantly higher in VB compared with control patients following stratification by CP stage (VB vs. CC CPA, P < 0.001; VB vs. CC CPB, P < 0.001; VB vs. CC CPC, P < 0.001) (Fig. 2). Adrenal function in patients with and without ascites
Patients with variceal bleeding and ascites had a higher CP score (P = 0.021), were more likely to have active bleeding at endoscopy (P = 0.027), and had higher baseline levels of salivary cortisol (P = 0.04).
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1128 European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 10
Table 1
Baseline characteristics of the study groups Patients with variceal bleeding (n = 38)
Sex (M/F) [n (%)] Age [median (range)] (years) Etiology of liver cirrhosis (viral/alcoholic/other) [n (%)] Previous variceal bleeding [n (%)] HCC at baseline [n (%)] Hemoglobin (g/dl) [median (range)] WBC count (×109/l) [median (range)] Platelet count (×109/l) [median (range)] INR [median (range)] Total bilirubin (mg/dl) [median (range)] Albumin (g/dl) [median (range)] Albumin ≤ 2.5 g/dl (%) Creatinine (mg/dl) [median (range)] Serum sodium (mmol/l) [median (range)] Ascites [n (%)] Encephalopathy [n (%)] Child–Pugh score [median (range)] Child–Pugh stage (A/B/C) [n (%)] MELD score [median (range)]
Stable patients with cirrhosis (n = 31)
30/8 (78.9/21.1) 58.5 (33–87) 12/22/4 (31.6/57.9/10.6) 16 (42.1) 8 (21.1) 9.4 (5.7–12.5) 8.9 (3.8–56) 111 (37–491) 1.45 (1.15–2.26) 2.07 (0.5–11.2) 3 (2.2–4.2) 15.2 0.85 (0.6–2.2) 137.5 (120.8–147) 24 (63.2) 9 (23.7) 8 (5–12) 7/22/7 (19.4/61.1/19.4) 15 (9–23)
P value
23/8 (74.2/25.8) 61 (30–88) 7/19/5 (22.6/61.3/16.1) 12 (38.7) 2 (6.5) 12.05 (9.3–15.7) 5.1 (1.8–11.5) 120 (42–326) 1.23 (0.9–2.4) 1.17 (0.51–24.45) 3.7 (0.9–4.6) 10 0.8 (0.5–1.9) 138 (129–144) 12 (38.7) 3 (9.7) 6 (5–12) 17/8/6 (53.3/26.7/20) 10 (6–25)
NS NS NS NS NS < 0.001 < 0.001 NS 0.001 0.005 0.001 NS NS NS 0.043 NS 0.026 0.006 NS
F, female; HCC, hepatocellular carcinoma; INR, international normalized ratio; M, male; MELD, model for end-stage liver disease; NS, not significant; WBC, white blood cells.
Table 2
Adrenal function parameters among the study groups Patients with variceal bleeding (n = 38)
TC (μg/ml) [median (range)] SC (μg/ml) [median (range)] CBG (μg/ml) [median (range)] CBG ≤ 35 μg/ml (%) [median (range)] cFC (μg/dl) [median (range)]
24.3 0.86 32.5 23 2.4
Stable patients with cirrhosis (n = 31)
(9.1–63.4) (0.14–3.31) (15.4–55) (60.5) (0.29–16.62)
11.6 0.407 40.5 8 0.57
P value < 0.001 < 0.001 0.002 0.003 < 0.001
(3.76–20.1) (0.06–1.07) (19.1–111.9) (25.8) (0.21–1.45)
CBG, cortisol-binding globulin; cFC, calculated free cortisol; SC, salivary cortisol; TC, total cortisol.
Fig. 1
Fig. 2
20.00 Median free cortisol (μg/dl)
15.00
10.00
5.00
0.00
5.00
0.00
5.00
PC ho sis
sis ho tro on
C
on
tro
lc
irr
lc irr
irr lc tro on C
C
PA C
le ed ric ea
lb
In the group of CC, ascites was correlated significantly with higher total bilirubin levels (P = 0.017), lower albumin levels (P = 0.01), and more severe liver disease (CP score, P < 0.001) (Supplementary Table 1).
Va
Box-plots of free cortisol levels in relation to Child–Pugh (CP) stage in patients with variceal bleeding: CPA versus CPB, P = 0.018; CPB versus CPC, P = 0.654; CPA versus CPC, P = 0.019.
C
C
Child−Pugh stage
ho sis
B
s
A
C PB
0.00 er
Free cortisol (μg/dl)
20.00
Box-plots of free cortisol levels between variceal bleeders (VB) and control patients with cirrhosis (CC), stratified according to Child–Pugh (CP) stage: VB versus CC CPA, 2.4 ± 4.05 versus 0.51 ± 0.27, P < 0.001; VB versus CC CPB, 2.4 ± 4.05 versus 0.54 ± 0.35, P < 0.001; VB versus CC CPC, 2.4 ± 4.05 versus 0.52 ± 0.45, P < 0.001.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Free cortisol and variceal bleeding Triantos et al. 1129
Cox regression analysis (adjusted for age) of predictive factors of 6-week mortality in patients with variceal bleeding
Table 3
OR [95% CI (lower–upper)] HCC Child–Pugh stage C cFC < 3.2 μg/dl Early rebleeding (< 5 days)
0.324 0.028 8.37 0.019
(0.21–0.51) (0.02–0.04) (5.44–12.88) (0.01–0.03)
P value < 0.001 < 0.001 < 0.001 < 0.001
cFC, calculated free cortisol; CI, confidence intervals; HCC, hepatocellular carcinoma; OR, odds ratio.
Clinical outcomes of patients with stable cirrhosis and variceal bleeding
(1) Overall survival in CC: During follow-up, three patients (9.7%) died (one patient because of variceal bleeding after 4 months of follow-up and two from liver failure after 4 and 18 months, respectively). No difference was observed in baseline total, salivary, or free cortisol levels among patients who did or did not die during follow-up (P = 0.274, 0.164, and 0.565, respectively). (2) Rebleeding: Four patients (10.5%) rebled within the first 5 days of hospital admission. The only factor predictive for 5-day rebleeding was albumin levels lower than 2.5 g/dl. (3) Bleeding-related mortality. Eight patients (21.1%) died in the first 6 weeks following admission. The causes of death were as follows: rebleeding (n = 4), liver failure (n = 3), and sepsis (n = 1). In the univariate analysis, predictors of 6-week mortality were age (P < 0.001), CP stage (P = 0.036), HCC (P < 0.001), international normalized ratio (P < 0.001), total serum bilirubin (P < 0.001), serum albumin (P < 0.001), serum creatinine (P = 0.001), serum sodium (P < 0.001), serum free cortisol (P < 0.001), and early rebleeding (P = 0.005). The area under the curve was 0.79 (95% confidence interval, 0.64–0.94) for evaluating free cortisol with respect to predicting 6-week mortality following variceal bleeding. An optimized cutoff of 3.2 μg/dl was calculated for the 6-week time point (sensitivity 85%, specificity 70%, positive predictive value 40%, negative predictive value 95.5%, positive likelihood ratio 2.86, negative likelihood ratio 0.2). Predictive factors of 6-week mortality in the multivariate analysis (adjusted for age) were HCC, CP stage C, free cortisol at least 3.2 μg/dl, and early rebleeding (Table 3). In the subgroup of patients with serum albumin greater than 2.5 mg/dl and CBG greater than 35 μg/ml levels, three of 14 died within 6 weeks following variceal bleeding; predictive factors of 6-week mortality were HCC (P = 0.003), high CP stage (P = 0.003), and higher free cortisol (P = 0.036). In the subgroup of patients (n = 23) with available salivary cortisol, predictive factors of 6-week survival in the univariate analysis were age (P < 0.001), CP stage (P = 0.019), model for end-stage liver disease (P = 0.003), presence of encephalopathy (P = 0.022), HCC (P < 0.001), international normalized
ratio (P = 0.04), total serum bilirubin (P = 0.037), serum creatinine (P = 0.002), salivary cortisol (P = 0.04), and early rebleeding (P < 0.001). Multivariate analysis for this subgroup could not be carried out because of the small number of events. The characteristics of patients with variceal bleeding according to the free cortisol levels using the cut-off of 3.2 μg/dl are shown in Supplementary Table 2. Patients with free cortisol higher than or equal to 3.2 μg/dl had lower serum albumin and serum sodium levels, higher platelets and white blood cells, more severe liver disease (assessed by CP score), and higher rate of encephalopathy and 6-week mortality.
Discussion This is the first study to evaluate baseline adrenal function with respect to survival in acute variceal bleeding. Several prognostic models have been proposed to evaluate failure to control bleeding and 6-week mortality [30]. In our study, well-known factors predictive of 6-week mortality such as early rebleeding [27,34,35], advanced CP [35,36], and HCC [37–40] were found but independently added to by high free cortisol serum level or salivary cortisol concentrations. Our cohort had the ‘expected’ early rebleeding rate of 10% and a 6-week mortality rate of 20%, consistent with recent reports of variceal bleeding that have shown progressive improvement over time [27,30,34,41]. We have shown previously [16] that patients with cirrhosis and variceal bleeding have CIRCI using an assessment of adrenal function with LDSST (1 μg ACTH), and standard SST (250 μg ACTH) in 20 patients with variceal bleeding, 60 CC, and 14 healthy volunteers. Using LDSST, AI was higher in VB compared with CC (60 vs. 48%, P = 0.01). Consistent with the present study, patients with variceal bleeding had higher baseline concentrations of cortisol than CC, but in the bleeding group, there were no significant increases in delta cortisol, indicating an inadequate response to stress. Another finding of this study is that VB had increasing levels of free and salivary cortisol from CPA to CPC, probably reflecting adrenal response to more severe illness. Similarly, Thevenot et al. [12], who evaluated basal and postsynacthen serum total, free, and salivary cortisol in 95 nonseptic and 30 septic patients with liver cirrhosis and CP score higher than 8, found an increasing but nonsignificant trend from CPA to CPC stages. In a recent study [11], adrenal function was evaluated using LDSST in 101 patients with CC without infection or hemodynamic instability. AI was present in 38 patients (38%) and was related to the severity of liver disease. The median basal total cortisol concentration decreased progressively with worsening severity of liver disease probably because of decreasing CBG level. In our study, no association was observed between CP stage and total cortisol in the
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1130
European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 10
control group with CC, but this may be because of a type 2 error. Comparing VB with CC, total, salivary, and free cortisol were significantly higher in VB, and this difference was independent of CBG or/and albumin levels. This indicates that during variceal bleeding, there is a hyperactivation of the hypothalamus-pituitary-adrenal (HPA) axis. Our results are consistent with those of the study of Thevenot et al. [12], in which basal total cortisol was significantly lower in the nonseptic patient group (11.3 ± 6.6 vs. 16.05 ± 6.57 μg/dl, P = 0.0013), as well as with our previous data (baseline cortisol levels among patients with variceal bleeding and controls, 15.4 vs. 8.7 μg/dl, P = 0.001) [16]. In a recent study, Fede et al. [42] found that patients with ascites and CC had lower serum albumin and CBG levels, more severe liver disease, and significantly higher free cortisol concentrations, probably attributable to the lower albumin and CBG levels in patients with ascites. In our study, both VB and CC with ascites had more severe liver disease (assessed by CP score). Serum albumin and CBG levels were also lower in patients with ascites in both groups, although a statistically significant difference was observed only for albumin levels in CC. Moreover, there were higher free cortisol levels in patients with ascites in both groups, although not significant, except for salivary cortisol levels in the group of VB. The observed differences in CBG, albumin, and free cortisol levels between patients with and without ascites probably failed to achieve statistical significance because of a type 2 error. The findings of both studies imply that advanced liver disease (reflected by the presence of ascites) is accompanied by higher FC levels. The pathophysiology of adrenal impairment in liver dysfunction is not clear [18] and it is still uncertain whether adrenal dysfunction is the reason for or the consequence of the severity of liver disease. Impaired cholesterol synthesis (a significant precursor for steroid synthesis) because of liver failure could contribute toward AI [43]. In addition, increased levels of endotoxin and proinflammatory cytokines that are observed in patients with liver disease could lead to impairment in the HPA axis [44]. The presence of AI could contribute toward hemodynamic instability and decreased response to vasoactive drugs, as in other critical conditions [1], and in variceal bleeding, possibly a greater probability of rebleeding or failure to control bleeding and, therefore, to reduced survival. As we have shown previously [16], during variceal bleeding, there is a hyperactivation of the HPA axis and a large percentage of patients have inadequate steroid response relative to the severity of the illness (CIRCI). This is probably related to a tissue resistance to glucocorticoids and not to decreased steroid synthesis [45–47]; thus, novel studies assessing the tissue action of cortisol are needed. Similarly, in the present study, the high levels of free cortisol might reflect an adrenocortical response to stress that, however, might be
inappropriate for the degree of severity of stress [16] and consequently could be related to worse outcomes. One limitation of this study is that we measured free cortisol indirectly using Coolens’ formula, which includes measurement of CBG for the calculation of the free fraction that may be reduced in cirrhosis. Although Coolens’ method has been validated in septic shock [48], it remains to be validated in advanced liver disease. However, there is enough evidence that calculated free cortisol is highly comparable with directly measured free cortisol [48], although discrepant results have also been published [10]. Moreover, free cortisol levels can be determined reliably using total cortisol and CBG levels in critically ill patients with sepsis and septic shock, who also have reduced CBG and albumin levels [48]. Direct measurement of free cortisol is currently unsuitable for routine clinical practice because of the need for complex nonautomated [49] and nonstandardized methods [10,50, 51]. Although we have not measured free cortisol, we obtained similar results in patients with normal CBG and albumin levels. Furthermore, we measured salivary cortisol, which has been considered a surrogate marker for free cortisol in noncirrhotic patients [52] and in stable patients with alcoholic cirrhosis [9], and found it to be a significant predictive factor for 6-week mortality. Galbois et al. [9] reported that the observed correlation between basal salivary cortisol and free cortisol (calculated using Coolens’ formula) was excellent (Spearman coefficient = 0.91, P
Higher free serum cortisol is associated with worse survival in acute variceal bleeding because of cirrhosis: a prospective study Christos K. Triantosa, Maria Kalafatelia, Dimitrios Samonakise, Konstantinos Zisimopoulosa, Nikolaos Papiamonise, Apostolos Sapountzisa, Marina Michalakib, Georgios Theocharisa, Konstantinos Thomopoulosa, Chryssoula Labropoulou-Karatzac, Venetsanea Kyriazopouloub, Eleni Jelastopulud, Elias A. Kouroumalise, Vasiliki Nikolopouloua and Andrew K. Burroughsf Background and aims Critical illness-related corticosteroid insufficiency has been reported in acute variceal bleeding (AVB). In cirrhosis, free serum cortisol (FC) is considered optimal to assess adrenal function. Salivary cortisol (SC) is considered a surrogate for FC. We evaluated FC and its prognostic role in AVB. Methods Total serum cortisol, SC, cortisol-binding globulin, and FC (Coolens’ formula) were evaluated in AVB (n = 38) and in stable cirrhosis (CC) (n = 31). A Cox proportional hazards model was evaluated for 6-week survival. Results In AVB, the median FC and SC levels were higher with worse liver dysfunction [Child–Pugh (CP) A/B/C: 1.59/2.62/3.26 μg/dl, P = 0.019; CPA/B/C: 0.48/0.897/1.81 μg/ml, P < 0.001, respectively]. In AVB compared with CC, median total serum cortisol: 24.3 versus 11.6 μg/dl (P < 0.001), SC: 0.86 versus 0.407 μg/ml (P < 0.001); FC 2.4 versus 0.57 μg/dl (P < 0.001). In AVB, 5-day rebleeding was 10.5%, and 6-week and total mortality were 21.1 and 23.7%, respectively. Independent associations with 6-week mortality in AVB were FC at least 3.2 μg/dl (P < 0.001), hepatocellular carcinoma (P < 0.001), CPC (P < 0.001), and early rebleeding (P < 0.001). Among patients with normal cortisol-binding globulin (n = 14) and albumin (n = 31), the
Introduction Relative adrenal insufficiency (RAI) is defined as an inadequate glucocorticoid activity relative to the severity of illness. RAI has been reported in patients with cirrhosis and sepsis [1] or septic shock [2] and also in patients with acute liver failure [3,4] with no apparent sepsis. Treatment with hydrocortisone has been assessed in acute liver failure [4,5] and in cirrhosis with septic shock [2,6], and is associated with shock reversal and reduction in vasopressor doses [2,4–6]; however, no
All supplementary digital content is available directly from the corresponding author. 0954-691X © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins
factors were hepatocellular carcinoma (P = 0.003), CP (P = 0.003), and FC (P = 0.036). SC was also found to be an independent predictor of 6-week mortality (P < 0.001). Area under the curve of FC for predicting 6-week mortality was 0.79. Conclusion Higher FC is present in cirrhosis with AVB compared with CC and is associated independently with bleeding-related mortality. However, whether high FC solely indicates the severity of illness or whether there is significant adrenal insufficiency cannot be discerned. Eur J Gastroenterol Hepatol 26:1125–1132 © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins. European Journal of Gastroenterology & Hepatology 2014, 26:1125–1132 Keywords: acute variceal bleeding, adrenal insufficiency, free cortisol, liver cirrhosis, salivary cortisol Departments of aGastroenterology, bEndocrinology, cInternal Medicine, University Hospital of Patras, dDepartment of Public Health, Medical School, University of Patras, Patras, eDepartment of Gastroenterology, University Hospital of Heraklion, Heraklion, Greece and fThe Royal Free Sheila Sherlock Liver Centre, Royal Free Hospital, London, UK Correspondence to Christos K. Triantos, MD, D. Stamatopoulou 4, Rio, 26504 Patras, Greece Tel: + 30 697 289 4651; fax: + 30 261 062 5382; e-mail: [email protected] Received 26 March 2014 Accepted 11 June 2014
improvement in mortality has been shown [2,5,6]. Adrenal impairment has also been observed in patients with stable cirrhosis (CC) [7–14] and with decompensation, other than sepsis, such as ascites [15] or variceal bleeding [16,17]. It has been suggested [18] that adrenal dysfunction could be a feature specific to liver disease, which becomes manifest with precipitating factors such as infection, bleeding, or acute illness, but the underlying mechanism remains unknown. However, assessment of adrenal function in cirrhosis is difficult and the results are difficult to interpret [18,19]. More than 90% of circulating cortisol is bound to albumin and cortisol-binding protein (CBG) [20–22], whereas the DOI: 10.1097/MEG.0000000000000158
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1126 European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 10
remaining unbound (free) fragment is the biologically active substance. In cirrhosis, the reduced concentrations of albumin and CBG [7] may lead to an overestimation of adrenal insufficiency (AI) when total serum cortisol is used to assess adrenal function [23], especially when albumin concentrations are lower than 2.5 g/dl [24]. Plasma free cortisol measurement can be assessed indirectly using Coolens’ formula [25] or using surrogate markers such as salivary cortisol [9,23]. Furthermore, there is no consensus to define the most sensitive adrenocorticotropic hormone (ACTH) stimulation test for the diagnosis of AI in patients with liver disease, and the cutoff values of total, salivary or free cortisol have not been standardized as yet [18]. Acute variceal bleeding in patients with cirrhosis is a lifethreatening complication of portal hypertension, with a current 6-week mortality rate of ∼ 20% [26,27], although survival has improved over recent decades. It is a critical condition with many similarities to sepsis and is frequently associated with bacterial infections [28,29]. Predictive factors of survival include severity of liver disease, shock at admission, active bleeding at endoscopy, and severity of portal hypertension (measured by hepatic venous pressure gradient) [30]. We have recently shown the presence of an inadequate glucocorticoid response with respect to the severity of disease [critical illness-related corticosteroid insufficiency (CIRCI)] in this setting [16]; in another study, RAI has been associated with therapeutic failure and death within 6 weeks in patients with severe variceal bleeding (systolic blood pressure < 100 mmHg or/and hemoglobin < 8 g/dl) [17]. However, in both studies, adrenal function was assessed by measuring serum total cortisol after the standard short synacthen test (SST) or the low-dose short synacthen test (LDSST), and this may overestimate the prevalence of AI, as mentioned before. Evidence of adrenal dysfunction is needed on the basis of free cortisol estimations, and preliminary evidence of a possible association with adverse outcomes is required. Therefore, we carried out this prospective study in acute variceal bleeding to evaluate baseline free cortisol concentrations (assessed by Coolens’ formula), as well as salivary cortisol and evaluated associations with survival.
Patients and methods Patients
This prospective, multicenter, observational study was carried out at the University Hospital of Patras (Patras, Greece) and the University Hospital of Heraklion (Heraklion, Greece), and planned at the Royal Free Sheila Sherlock Liver Centre (London, UK); it was approved by the local hospital ethics committee. Written consent was obtained from all patients. We finally evaluated 38 consecutive patients with cirrhosis and variceal bleeding (two were excluded from the study because of disturbance in consciousness level and thus, consent form
could not be obtained) and 31 consecutive patients with cirrhosis attending the hepatology outpatient clinics, included from March 2010 until December of 2011. Patients were followed up until December 2013. Patients with a history of hypothalamic-pituitary or adrenal disease, treatment with corticosteroids within the past 12 months, infection with HIV, and hepatocellular carcinoma (HCC) beyond the Milan Criteria were excluded from the protocol. The severity of liver disease was assessed by the Child–Pugh (CP) class and model for end-stage liver disease at the time of adrenal function assessment. Patients with variceal bleeding were treated with packed red blood cells and plasma expanders to maintain hemodynamic stability and hemoglobin levels at ∼ 8 g/dl [31,32]. Management also included intravenous somatostatin for 5 days, antibiotics (intravenous ceftriaxone or ciprofloxacin), lactulose (when encephalopathy occurred), and endoscopic treatment, with variceal band ligation performed within 12 h of admission [32]. None of the patients received intravenous terlipressin or albumin infusions. The presence of infection was investigated with blood and urine cultures, paracentesis and culture of ascites, and chest radiography. Serum and salivary cortisol were measured the morning after admission (between 8:00 and 9:00 a.m.). Salivary samples were not collected in patients with blood contamination or fungal infection of the oral cavity, presence of a nasogastric tube, endotracheal intubation, or if the time of measurement was close to hematemesis or endoscopic procedure. In CC, serum and salivary cortisol were also measured between 8:00 and 9:00 a.m. at the outpatient clinics and saliva was collected at least 30 min after tooth-brushing and patients were advised to avoid smoking or drinking before saliva collection. Saliva was collected using dedicated kits (Plain Salivette; Sarstedt, Newton, North Carolina, USA). The cotton provided was chewed for 2–3 min and then collected in a plastic tube. Saliva was then collected by centrifugation. Salivary, serum total cortisol (normal range: 6.2–19.4 μg/dl), and plasma ACTH levels were measured using an electrochemiluminescence immunoassay (Roche Diagnostics Ltd, Rotkreuz, Switzerland). Serum CBG concentrations (normal range: 35–50 μg/ml) were assessed by radioimmunoassay (DIAsource Immunoassays S.A., Nivelles, Belgium). Salivary cortisol was available from 23 of 38 patients with variceal bleeding (10 patients had oral blood contamination, three had a nasogastric tube, and two had inadequate amount of collected saliva). Twentyseven CC had available salivary cortisol (four had insufficient salivary samples collected). Calculated plasma free cortisol
Calculated free cortisol was derived using Coolens’ formula [25]: U2K (1 + N) + U [1 + N + K (T − C)] − C = 0, where U represents the molar concentration of unbound
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Free cortisol and variceal bleeding Triantos et al. 1127
cortisol, C the molar concentration of total cortisol, and T the concentration of CBG. K corresponds to the affinity of CBG for cortisol at 37°C (equal to 3 × 10–7/mol/l) and N to the proportion of albumin-bound to unbound cortisol (1.74 is the value used conventionally). Definitions
Cirrhosis was diagnosed on the basis of histological findings, clinical evaluation, laboratory data, or imaging findings compatible with a diagnosis of cirrhosis. Variceal bleeding was diagnosed endoscopically on the basis of the presence of hematemesis or melena with either a bleeding varix (active bleeding or clot adherent to varix or variceal ulceration) or when there was no other visible lesion. Severe bleeding was defined as arterial hypotension (systolic blood pressure < 100 mmHg) or/and hemoglobin less than 8 g/dl at admission. Rebleeding was defined, according to the Baveno V criteria [32], as recurrence of hemorrhage after the first 24 h and within 5 days of admission for the bleeding episode, following initial control of bleeding, that is during the period of somatostatin infusion [33]. Bleeding-related mortality was defined as death within 6 weeks from the bleeding episode [32]. Groups
In this study, there were 69 patients with liver cirrhosis: 38 with variceal bleeding (group A) and 31 with CC (group B), median follow-up: 11 months (0.2–22). Statistical analysis
Numerical data were expressed as median, SD, and range (minimum to maximum) and categorical data as counts and percentages. All variables were tested for normal distribution using the Kolmogorov–Smirnov test. Categorical variables were tested using the χ2 and Fisher’s exact test. Continuous variables with and without a normal distribution were compared using Student’s t-test or the Mann–Whitney U-test, respectively. The discriminative ability of calculated serum free cortisol to predict 6-week mortality was assessed by receiver operating characteristic analysis. The value of free cortisol that maximized the sum of sensitivity and specificity was used as the best cut-off. Univariate analysis was used to determine potential independent predictors of 6-week survival in both groups. The statistical significance of differences in outcome among the different groups was assessed using the log-rank test. Variables significant at the 0.1% level were included in the age-adjusted multivariate analysis using binary logistic regression analysis. All statistical tests were two-sided and significance level was set at 0.05 or less. The results are presented as hazard ratios with 95% confidence intervals in multivariate analyses. The SPSS statistical package (version 19.0 for Windows; SPSS Inc., Chicago, Illinois, USA) was used.
Results Patients’ characteristics
The baseline clinical and laboratory characteristics for each group of patients are shown in Table 1. Both groups were similar at baseline in terms of age, sex, and etiology of liver disease. Thirteen (34.2%) patients with variceal bleeding had active bleeding at endoscopy and 17 (42.1%) had severe bleeding. The median packed red blood cells transfused were 3 (range 0–7), whereas a median of 2 (range 0–5) were administered within the first 24 h following admission. The results of adrenal function parameters assessed are shown in Table 2. Adrenal function in patients with variceal bleeding versus stable patients with cirrhosis
In patients with variceal bleeding, the median values of baseline serum free cortisol and salivary cortisol were significantly higher in CPC compared with CPB and CPA: free cortisol-CPA: 1.59 ± 0.53 versus CPB: 2.46 ± 2.61 versus CPC: 3.33 ± 6.5 μg/dl (Fig. 1); salivary cortisol-CPA: 0.48 ± 0.25 versus CPB: 0.897 ± 0.69 versus CPC: 1.81 ± 1.03 μg/ml (P < 0.001). Serum total cortisol and CBG were not significantly associated with CP class. Patients with severe variceal bleeding (n = 17) showed no differences in adrenal function parameters compared with patients without severe bleeding (total serum cortisol, P = 0.419; salivary cortisol, P = 0.944, free cortisol, P = 0.466; CBG, P = 0.643). In CC, no statistically significant difference was observed in total serum cortisol, salivary cortisol, CBG, and serum free cortisol in relation to CP stage. Comparing variceal bleeders (VB) with CC, serum total, salivary, and free cortisols were significantly higher in variceal bleeding: total cortisol: 24.3 ± 14.8 versus 11.6 ± 4.3 μg/dl, P < 0.001; salivary cortisol: 0.86 ± 0.85 versus 0.407 ± 0.25 μg/ml, P < 0.001; free cortisol: 2.4 ± 4.05 versus 0.57 ± 0.32 μg/dl, P < 0.001. Among the subgroup of VB (n = 14) and CC (n = 21) with baseline serum albumin greater than 2.5 mg/dl and CBG greater than 35 μg/ml levels, total, salivary, and free cortisol levels were also significantly higher in VB compared with CC: total cortisol: 23.68 ± 12.98 versus 12.2 ± 3.27 μg/dl, P < 0.001; salivary cortisol: 0.788 ± 0.709 versus 0.426 ± 0.22 μg/ml, P = 0.005; and free cortisol: 1.75 ± 4.04 versus 0.57 ± 0.3 μg/dl, P < 0.001, respectively. Free cortisol levels were also significantly higher in VB compared with control patients following stratification by CP stage (VB vs. CC CPA, P < 0.001; VB vs. CC CPB, P < 0.001; VB vs. CC CPC, P < 0.001) (Fig. 2). Adrenal function in patients with and without ascites
Patients with variceal bleeding and ascites had a higher CP score (P = 0.021), were more likely to have active bleeding at endoscopy (P = 0.027), and had higher baseline levels of salivary cortisol (P = 0.04).
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1128 European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 10
Table 1
Baseline characteristics of the study groups Patients with variceal bleeding (n = 38)
Sex (M/F) [n (%)] Age [median (range)] (years) Etiology of liver cirrhosis (viral/alcoholic/other) [n (%)] Previous variceal bleeding [n (%)] HCC at baseline [n (%)] Hemoglobin (g/dl) [median (range)] WBC count (×109/l) [median (range)] Platelet count (×109/l) [median (range)] INR [median (range)] Total bilirubin (mg/dl) [median (range)] Albumin (g/dl) [median (range)] Albumin ≤ 2.5 g/dl (%) Creatinine (mg/dl) [median (range)] Serum sodium (mmol/l) [median (range)] Ascites [n (%)] Encephalopathy [n (%)] Child–Pugh score [median (range)] Child–Pugh stage (A/B/C) [n (%)] MELD score [median (range)]
Stable patients with cirrhosis (n = 31)
30/8 (78.9/21.1) 58.5 (33–87) 12/22/4 (31.6/57.9/10.6) 16 (42.1) 8 (21.1) 9.4 (5.7–12.5) 8.9 (3.8–56) 111 (37–491) 1.45 (1.15–2.26) 2.07 (0.5–11.2) 3 (2.2–4.2) 15.2 0.85 (0.6–2.2) 137.5 (120.8–147) 24 (63.2) 9 (23.7) 8 (5–12) 7/22/7 (19.4/61.1/19.4) 15 (9–23)
P value
23/8 (74.2/25.8) 61 (30–88) 7/19/5 (22.6/61.3/16.1) 12 (38.7) 2 (6.5) 12.05 (9.3–15.7) 5.1 (1.8–11.5) 120 (42–326) 1.23 (0.9–2.4) 1.17 (0.51–24.45) 3.7 (0.9–4.6) 10 0.8 (0.5–1.9) 138 (129–144) 12 (38.7) 3 (9.7) 6 (5–12) 17/8/6 (53.3/26.7/20) 10 (6–25)
NS NS NS NS NS < 0.001 < 0.001 NS 0.001 0.005 0.001 NS NS NS 0.043 NS 0.026 0.006 NS
F, female; HCC, hepatocellular carcinoma; INR, international normalized ratio; M, male; MELD, model for end-stage liver disease; NS, not significant; WBC, white blood cells.
Table 2
Adrenal function parameters among the study groups Patients with variceal bleeding (n = 38)
TC (μg/ml) [median (range)] SC (μg/ml) [median (range)] CBG (μg/ml) [median (range)] CBG ≤ 35 μg/ml (%) [median (range)] cFC (μg/dl) [median (range)]
24.3 0.86 32.5 23 2.4
Stable patients with cirrhosis (n = 31)
(9.1–63.4) (0.14–3.31) (15.4–55) (60.5) (0.29–16.62)
11.6 0.407 40.5 8 0.57
P value < 0.001 < 0.001 0.002 0.003 < 0.001
(3.76–20.1) (0.06–1.07) (19.1–111.9) (25.8) (0.21–1.45)
CBG, cortisol-binding globulin; cFC, calculated free cortisol; SC, salivary cortisol; TC, total cortisol.
Fig. 1
Fig. 2
20.00 Median free cortisol (μg/dl)
15.00
10.00
5.00
0.00
5.00
0.00
5.00
PC ho sis
sis ho tro on
C
on
tro
lc
irr
lc irr
irr lc tro on C
C
PA C
le ed ric ea
lb
In the group of CC, ascites was correlated significantly with higher total bilirubin levels (P = 0.017), lower albumin levels (P = 0.01), and more severe liver disease (CP score, P < 0.001) (Supplementary Table 1).
Va
Box-plots of free cortisol levels in relation to Child–Pugh (CP) stage in patients with variceal bleeding: CPA versus CPB, P = 0.018; CPB versus CPC, P = 0.654; CPA versus CPC, P = 0.019.
C
C
Child−Pugh stage
ho sis
B
s
A
C PB
0.00 er
Free cortisol (μg/dl)
20.00
Box-plots of free cortisol levels between variceal bleeders (VB) and control patients with cirrhosis (CC), stratified according to Child–Pugh (CP) stage: VB versus CC CPA, 2.4 ± 4.05 versus 0.51 ± 0.27, P < 0.001; VB versus CC CPB, 2.4 ± 4.05 versus 0.54 ± 0.35, P < 0.001; VB versus CC CPC, 2.4 ± 4.05 versus 0.52 ± 0.45, P < 0.001.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Free cortisol and variceal bleeding Triantos et al. 1129
Cox regression analysis (adjusted for age) of predictive factors of 6-week mortality in patients with variceal bleeding
Table 3
OR [95% CI (lower–upper)] HCC Child–Pugh stage C cFC < 3.2 μg/dl Early rebleeding (< 5 days)
0.324 0.028 8.37 0.019
(0.21–0.51) (0.02–0.04) (5.44–12.88) (0.01–0.03)
P value < 0.001 < 0.001 < 0.001 < 0.001
cFC, calculated free cortisol; CI, confidence intervals; HCC, hepatocellular carcinoma; OR, odds ratio.
Clinical outcomes of patients with stable cirrhosis and variceal bleeding
(1) Overall survival in CC: During follow-up, three patients (9.7%) died (one patient because of variceal bleeding after 4 months of follow-up and two from liver failure after 4 and 18 months, respectively). No difference was observed in baseline total, salivary, or free cortisol levels among patients who did or did not die during follow-up (P = 0.274, 0.164, and 0.565, respectively). (2) Rebleeding: Four patients (10.5%) rebled within the first 5 days of hospital admission. The only factor predictive for 5-day rebleeding was albumin levels lower than 2.5 g/dl. (3) Bleeding-related mortality. Eight patients (21.1%) died in the first 6 weeks following admission. The causes of death were as follows: rebleeding (n = 4), liver failure (n = 3), and sepsis (n = 1). In the univariate analysis, predictors of 6-week mortality were age (P < 0.001), CP stage (P = 0.036), HCC (P < 0.001), international normalized ratio (P < 0.001), total serum bilirubin (P < 0.001), serum albumin (P < 0.001), serum creatinine (P = 0.001), serum sodium (P < 0.001), serum free cortisol (P < 0.001), and early rebleeding (P = 0.005). The area under the curve was 0.79 (95% confidence interval, 0.64–0.94) for evaluating free cortisol with respect to predicting 6-week mortality following variceal bleeding. An optimized cutoff of 3.2 μg/dl was calculated for the 6-week time point (sensitivity 85%, specificity 70%, positive predictive value 40%, negative predictive value 95.5%, positive likelihood ratio 2.86, negative likelihood ratio 0.2). Predictive factors of 6-week mortality in the multivariate analysis (adjusted for age) were HCC, CP stage C, free cortisol at least 3.2 μg/dl, and early rebleeding (Table 3). In the subgroup of patients with serum albumin greater than 2.5 mg/dl and CBG greater than 35 μg/ml levels, three of 14 died within 6 weeks following variceal bleeding; predictive factors of 6-week mortality were HCC (P = 0.003), high CP stage (P = 0.003), and higher free cortisol (P = 0.036). In the subgroup of patients (n = 23) with available salivary cortisol, predictive factors of 6-week survival in the univariate analysis were age (P < 0.001), CP stage (P = 0.019), model for end-stage liver disease (P = 0.003), presence of encephalopathy (P = 0.022), HCC (P < 0.001), international normalized
ratio (P = 0.04), total serum bilirubin (P = 0.037), serum creatinine (P = 0.002), salivary cortisol (P = 0.04), and early rebleeding (P < 0.001). Multivariate analysis for this subgroup could not be carried out because of the small number of events. The characteristics of patients with variceal bleeding according to the free cortisol levels using the cut-off of 3.2 μg/dl are shown in Supplementary Table 2. Patients with free cortisol higher than or equal to 3.2 μg/dl had lower serum albumin and serum sodium levels, higher platelets and white blood cells, more severe liver disease (assessed by CP score), and higher rate of encephalopathy and 6-week mortality.
Discussion This is the first study to evaluate baseline adrenal function with respect to survival in acute variceal bleeding. Several prognostic models have been proposed to evaluate failure to control bleeding and 6-week mortality [30]. In our study, well-known factors predictive of 6-week mortality such as early rebleeding [27,34,35], advanced CP [35,36], and HCC [37–40] were found but independently added to by high free cortisol serum level or salivary cortisol concentrations. Our cohort had the ‘expected’ early rebleeding rate of 10% and a 6-week mortality rate of 20%, consistent with recent reports of variceal bleeding that have shown progressive improvement over time [27,30,34,41]. We have shown previously [16] that patients with cirrhosis and variceal bleeding have CIRCI using an assessment of adrenal function with LDSST (1 μg ACTH), and standard SST (250 μg ACTH) in 20 patients with variceal bleeding, 60 CC, and 14 healthy volunteers. Using LDSST, AI was higher in VB compared with CC (60 vs. 48%, P = 0.01). Consistent with the present study, patients with variceal bleeding had higher baseline concentrations of cortisol than CC, but in the bleeding group, there were no significant increases in delta cortisol, indicating an inadequate response to stress. Another finding of this study is that VB had increasing levels of free and salivary cortisol from CPA to CPC, probably reflecting adrenal response to more severe illness. Similarly, Thevenot et al. [12], who evaluated basal and postsynacthen serum total, free, and salivary cortisol in 95 nonseptic and 30 septic patients with liver cirrhosis and CP score higher than 8, found an increasing but nonsignificant trend from CPA to CPC stages. In a recent study [11], adrenal function was evaluated using LDSST in 101 patients with CC without infection or hemodynamic instability. AI was present in 38 patients (38%) and was related to the severity of liver disease. The median basal total cortisol concentration decreased progressively with worsening severity of liver disease probably because of decreasing CBG level. In our study, no association was observed between CP stage and total cortisol in the
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1130
European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 10
control group with CC, but this may be because of a type 2 error. Comparing VB with CC, total, salivary, and free cortisol were significantly higher in VB, and this difference was independent of CBG or/and albumin levels. This indicates that during variceal bleeding, there is a hyperactivation of the hypothalamus-pituitary-adrenal (HPA) axis. Our results are consistent with those of the study of Thevenot et al. [12], in which basal total cortisol was significantly lower in the nonseptic patient group (11.3 ± 6.6 vs. 16.05 ± 6.57 μg/dl, P = 0.0013), as well as with our previous data (baseline cortisol levels among patients with variceal bleeding and controls, 15.4 vs. 8.7 μg/dl, P = 0.001) [16]. In a recent study, Fede et al. [42] found that patients with ascites and CC had lower serum albumin and CBG levels, more severe liver disease, and significantly higher free cortisol concentrations, probably attributable to the lower albumin and CBG levels in patients with ascites. In our study, both VB and CC with ascites had more severe liver disease (assessed by CP score). Serum albumin and CBG levels were also lower in patients with ascites in both groups, although a statistically significant difference was observed only for albumin levels in CC. Moreover, there were higher free cortisol levels in patients with ascites in both groups, although not significant, except for salivary cortisol levels in the group of VB. The observed differences in CBG, albumin, and free cortisol levels between patients with and without ascites probably failed to achieve statistical significance because of a type 2 error. The findings of both studies imply that advanced liver disease (reflected by the presence of ascites) is accompanied by higher FC levels. The pathophysiology of adrenal impairment in liver dysfunction is not clear [18] and it is still uncertain whether adrenal dysfunction is the reason for or the consequence of the severity of liver disease. Impaired cholesterol synthesis (a significant precursor for steroid synthesis) because of liver failure could contribute toward AI [43]. In addition, increased levels of endotoxin and proinflammatory cytokines that are observed in patients with liver disease could lead to impairment in the HPA axis [44]. The presence of AI could contribute toward hemodynamic instability and decreased response to vasoactive drugs, as in other critical conditions [1], and in variceal bleeding, possibly a greater probability of rebleeding or failure to control bleeding and, therefore, to reduced survival. As we have shown previously [16], during variceal bleeding, there is a hyperactivation of the HPA axis and a large percentage of patients have inadequate steroid response relative to the severity of the illness (CIRCI). This is probably related to a tissue resistance to glucocorticoids and not to decreased steroid synthesis [45–47]; thus, novel studies assessing the tissue action of cortisol are needed. Similarly, in the present study, the high levels of free cortisol might reflect an adrenocortical response to stress that, however, might be
inappropriate for the degree of severity of stress [16] and consequently could be related to worse outcomes. One limitation of this study is that we measured free cortisol indirectly using Coolens’ formula, which includes measurement of CBG for the calculation of the free fraction that may be reduced in cirrhosis. Although Coolens’ method has been validated in septic shock [48], it remains to be validated in advanced liver disease. However, there is enough evidence that calculated free cortisol is highly comparable with directly measured free cortisol [48], although discrepant results have also been published [10]. Moreover, free cortisol levels can be determined reliably using total cortisol and CBG levels in critically ill patients with sepsis and septic shock, who also have reduced CBG and albumin levels [48]. Direct measurement of free cortisol is currently unsuitable for routine clinical practice because of the need for complex nonautomated [49] and nonstandardized methods [10,50, 51]. Although we have not measured free cortisol, we obtained similar results in patients with normal CBG and albumin levels. Furthermore, we measured salivary cortisol, which has been considered a surrogate marker for free cortisol in noncirrhotic patients [52] and in stable patients with alcoholic cirrhosis [9], and found it to be a significant predictive factor for 6-week mortality. Galbois et al. [9] reported that the observed correlation between basal salivary cortisol and free cortisol (calculated using Coolens’ formula) was excellent (Spearman coefficient = 0.91, P
