Assessment of adrenal function in patients with acute hepatitis using serum free and total cortisol

Digestive and Liver Disease 47 (2015) 783–789

Contents lists available at ScienceDirect

Digestive and Liver Disease journal homepage: www.elsevier.com/locate/dld

Liver, Pancreas and Biliary Tract

Assessment of adrenal function in patients with acute hepatitis using serum free and total cortisol Thibault Degand a,b , Elisabeth Monnet a,b , Franc¸ois Durand c , Emilie Grandclement d , Philippe Ichai e , Sophie Borot f , Clifford R. Qualls g , Arnaud Agin h , Alexandre Louvet i , Jérôme Dumortier j , Claire Francoz c , Gilles Dumoulin d , Vincent Di Martino a,b , Richard Dorin k , Thierry Thevenot a,b,∗ a

Hepatology and Digestive Intensive Care Unit, University Hospital of Besanc¸on, France EA UPRES 3186 « Agents Pathogènes et Inflammation » of Franche-Comté University, France c Hepatology Unit, Beaujon Hospital, Clichy cedex, France d Laboratory for Endocrinology and Metabolism, University Hospital of Besanc¸on, France e Hepatobiliary Unit and Liver Intensive Care, Paul Brousse University Hospital AP-HP, Villejuif cedex, France f Department of Endocrinology, University Hospital of Besanc¸on, France g Clinical Translational Science Center, University of New Mexico Health Science Center, USA h ICube, UMR 7357, University of Strasbourg and CNRS, FMTS, Strasbourg, France i Department of Hepatogastroenterology, University Hospital of Lille, France j Department of Hepatogastroenterology, University Hospital Edouard Herriot of Lyon, France k Department of Medicine, New Mexico VA Medical Center and University of New Mexico Health Science Center, USA b

a r t i c l e

i n f o

Article history: Received 2 February 2015 Accepted 16 May 2015 Available online 22 May 2015 Keywords: Acute liver failure Hypothalamic–pituitary–adrenal axis

a b s t r a c t Background: Adrenal dysfunction is frequently reported in severe acute hepatitis using serum total cortisol. Aims: Because 90% of serum cortisol is bound to proteins that are altered during stress, we investigated the effect of decreased cortisol-binding proteins on serum total and free cortisol in severe acute hepatitis. Methods: 43 severe and 31 non-severe acute hepatitis and 29 healthy controls were enrolled consecutively and studied prospectively. Baseline (T0 ) and cosyntropin-stimulated (T60 ) serum total and free cortisol concentrations were measured. Results: T0 and T60 serum total cortisol did not differ significantly between severe, non-severe hepatitis and healthy controls. Conversely, serum free cortisol (T0 p = 0.012; T60 p < 0.001) concentrations increased from healthy controls to severe hepatitis, accompanied by a decrease in corticosteroid-binding globulin and albumin (all p < 0.001). In acute hepatitis (n = 74), patients with “low” corticosteroid-binding globulin (<28 mg/L) had higher T0 serum free cortisol than others (103.1 [61.2–157] vs. 56.6 [43.6–81.9] nmol/L, p = 0.0024). Analysis of covariance showed that at equal concentration of total cortisol, the free cortisol concentration was significantly higher in severe than in non-severe hepatitis (p < 0.001) or healthy controls (p < 0.001). Conclusions: In severe hepatitis, the decrease in cortisol-binding proteins impairs correct diagnosis of adrenal dysfunction. This could be corrected by measuring or estimating free cortisol. © 2015 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.

1. Introduction Acute liver failure (ALF) is a rare critical illness characterized by massive hepatic necrosis caused mainly by drug-induced liver

∗ Corresponding author at: Service d’Hépatologie et de Soins Intensifs Digestifs, Hôpital Jean Minjoz, 25000 Besanc¸on, France. Tel.: +33 3 81 66 85 94; fax: +33 3 81 66 84 18. E-mail address: [email protected] (T. Thevenot).

injury [1]. The nonspecific clinical presentation of ALF may be easily confused with symptoms of acute adrenal dysfunction (AD), another life-threatening condition that could exacerbate ALF. AD typically arises in critical conditions such as severe infections, trauma or after aggressive surgery [2–4]. The excessive production of pro-inflammatory cytokines (IL-6 and TNF-␣) found in ALF may account for the high mortality rate reported in this setting, and may also contribute to the occurrence of AD [5]. Indeed, proinflammatory cytokines may cause a shift from cortisol synthesis towards androgen production, or induce the dominant-negative

http://dx.doi.org/10.1016/j.dld.2015.05.016 1590-8658/© 2015 Editrice Gastroenterologica Italiana S.r.l. Published by Elsevier Ltd. All rights reserved.

784

T. Degand et al. / Digestive and Liver Disease 47 (2015) 783–789

␤-isoform of the glucocorticoid receptor leading to AD [6,7]. Furthermore, it has been shown that IL-6 inhibited the synthesis of cortisol-binding globulin (CBG) [8] and low serum CBG concentrations have been reported in the early phase of severe burn injury and septic shock, two conditions that present high levels of IL-6 [9,10]. The decrease in the main cortisol-binding proteins (i.e. albumin or CBG, both synthesized in the liver) may lead to overestimation of the prevalence of AD when assessed by serum total cortisol (STC, i.e. bound plus free fractions), as previously reported in a cirrhotic population [10]. The assessment of AD using serum free cortisol (SFC) concentrations has been reported to correlate strongly with severity of illness [11]. Using STC, AD has been reported in 62% of ALF patients and could contribute to haemodynamic instability and mortality in these patients [12]. Corticosteroid replacement could be useful in ALF patients suspected to have AD, although this strategy remains conflicting in patients suffering from a septic shock with [13] or without cirrhosis [4,14]. Moreover, corticosteroids have been reported to be associated with an increased risk of shock relapse in cirrhotic patients with septic shock [13]. Therefore, the diagnosis of AD needs to be properly ascertained before starting any corticosteroid treatment. We postulate that the high prevalence of AD reported in ALF [12] is misleadingly overestimated and should be corrected by measuring the free fraction of cortisol. The present CORT-HEPAT study aimed to assess, in patients with severe acute hepatitis (SAH), (i) variations in SFC and STC concentrations, and in cortisol-binding protein concentrations; (ii) correlations between STC and SFC concentrations; (iii) thresholds of SFC to identify patients with AD; (iv) agreement between measured SFC and estimated SFC using Coolens’ or cubic equations [15,16]. 2. Materials and methods 2.1. Study design and patient characteristics This multicentre, prospective, observational study was conducted from August 2011 to February 2014 in the Hepatology Units of five university teaching hospitals in France (Besanc¸on, Clichy, Villejuif, Lyon and Lille). The study was approved by the local ethics committee and patients provided written informed consent in accordance with the ethical guidelines of the Declaration of Helsinki. Inclusion criteria were consecutive patients aged between 18 and 75 years hospitalized for acute hepatitis defined as an abrupt rise in serum aminotransaminase levels during the previous 15 days (AST or ALT >500 IU/L or >10 times the upper normal value). Acute hepatitis was considered as severe if the prothrombin index was <50%, and as non-severe if the prothrombin index was >50%. We excluded patients with history of hypothalamicpituitary or adrenal disease, chronic liver disease, corticosteroid treatment within the previous 6 months, ketoconazole intake, liver transplant recipients, acute alcoholic hepatitis and night workers. Twenty-nine healthy controls (HC) were also enrolled and were stratified with the SAH group in terms of age, gender and oestrogen intake, since oestrogen therapy was the most common cause of changes in CBG levels. To evaluate the range of SFC concentrations, eight patients with known AD caused by impairment of the hypothalamic–pituitary–adrenal (HPA) axis (n = 5) and adrenal gland (n = 3) followed at the Endocrinology Department of Besancon were also studied. 2.2. Laboratory measurements STC and SFC concentrations were measured blindly before (T0 between 8 am and 9 am) and 60 min after (T60 ) intravenous

injection of 250 ␮g of tetracosactrin (Synacthen® , Sigma-Tau laboratory, Issy-les-Moulineaux, France). Serum CBG, albumin and adrenocorticotropic hormone (ACTH) were also measured. Further details are provided in Appendix A. 2.3. Adrenal dysfunction assessment We defined AD as a STC concentration <83 nmol/L (3 ␮g/dL) at T0 or <550 nmol/L (20 ␮g/dL) at T60 [17]. In stressed SAH patients, we used the same criteria as those proposed by Harry et al., i.e. T0 STC <250 nmol/L, T60 STC <500 nmol/L, or a delta STC level (i.e. the difference between cortisol values at T60 and T0 ) <250 nmol/L [12]. Since the normal range of SFC concentrations in SAH patients remains uncertain, we used the range from 5th to the 10th percentile of the distribution of SFC to define abnormal cortisol values, assuming that none of SAH patients studied had AD. To determine the range of SFC values that may be used for AD diagnosis in unstressed patients, we used the maximum value of patients with known AD and the minimum value of HC. 2.4. Calculated serum free cortisol We compared measured SFC concentrations using (1) Coolens’ equation [15], and (2) the cubic solution [16] using the Bland and Altman method (see details in Appendix B). 2.5. Statistical analysis Numerical variables are presented as mean ± standard deviation (SD) or as median and interquartile range [IQR] and categorical variables as number (percentage). The primary outcome was to compare STC, SFC and cortisol-binding proteins (CBG and albumin) concentrations between the three groups namely SAH, NSAH and HC. Continuous variables were compared using the Mann–Whitney test or the Student t test as appropriate. The Kruskal–Wallis test with Bonferroni’s correction was used for within-group multiple comparisons and the Spearman correlation coefficient was calculated. We also compared SFC and STC concentrations in patients with “low” and “high” serum albumin or CBG concentrations. The cut-off values defining “high” and “low” concentrations of albumin and CBG were arbitrarily set at the 25th percentile. The strength of agreement between measured and calculated SFC was analyzed using the Bland–Altman method. An adequate free cortisol index (FCI) was defined as a ratio STC/CBG >12 after ACTH stimulation [18]. Finally, linear and quadratic regression models were fitted in order to study the relationship between STC and SFC within the three groups. Regression lines were compared between groups using covariance analysis, assuming that the hypothesis of homogeneous variance between groups was not rejected. Stratification variables (i.e. age, sex and oestrogen intake) were included in the multivariate model. A p-value <0.05 was considered statistically significant. All statistical analyses were performed using SAS 9.3 (SAS Institute Inc., Cary, NC, USA). 3. Results 3.1. Patients characteristics Seventy-five consecutive patients with acute hepatitis were enrolled, but one SAH patient was later excluded because he had received corticosteroids; 74 patients were included in the final analysis. Per protocol, the three groups (SAH = 43, NSAH = 31 and HC = 29) were proportionally distributed at baseline in terms of age, gender and oral contraception intake (details are reported in Table 1). In SAH patient, 8 patients (18.6%) had encephalopathy

T. Degand et al. / Digestive and Liver Disease 47 (2015) 783–789

785

Table 1 Baseline characteristics of patients with severe and non-severe acute hepatitis, healthy controls and patients with known adrenal dysfunction. Characteristics

Severe acute hepatitis (n = 43)

Age (years) 37.5 ± 13.2 Male gender, n (%) 26 (60.5) Oral contraception, n (%) 3 (7.9) 23.1 ± 3.9 Body mass index (kg/m2 ) 91.1 ± 13.7 Mean arterial pressure (mmHg) 84.2 ± 19.1 Heart rate (beats/min) 31.8 ± 11.4 Prothrombin index (%) INR 3.5 ± 2.2 Total bilirubin (␮mol/L) 82 [36–176] Serum albumin (g/L) 29.9 ± 5.6 32.2 ± 15.4 Serum CBG (mg/L) 61.0 [49.0–78.0] Serum creatinine (␮mol/L) Aetiology of severe acute hepatitis, n (%) Drug-induced 31 (72.1) 27 (62.8) Acetaminophen† 9 (20.9) Viral hepatitis 2 (4.7) Autoimmune hepatitis Other causes‡ 1 (2.3)

Non-severe acute hepatitis (n = 31)

Healthy controls (n = 29)

Patients with known adrenal dysfunction (n = 8)

42.2 ± 16.3 18 (58.1) 2 (6.5) 23.8 ± 5.7 87.0 ± 12.7 75.8 ± 17.1 74.0 ± 16.8 1.3 ± 0.2 45 [25–138] 32.4 ± 4.3 53.3 ± 20.5 63.0 [58.6–78.7]

37.8 ± 10.8 17 (58.6) 3 (10.3) 23.8 ± 3.2 97.6 ± 11.3 72.2 ± 12.2 91.3 ± 7.4 1.1 ± 0.1 ND 41.4 ± 2.2 48.3 ± 17.0 ND

42.0 ± 15.7 5 (62.5) ND 27.5 ± 2.3 92.4 ± 9.1 68.4 ± 11.1 91.6 ± 8.6 1.1 ± 0.1 ND 40.7 ± 3.0 56.1 ± 10.0 ND

10 (32.2) 9 (29.0) 14 (45.2) 2 (6.5) 5 (16.1)

Quantitative values are expressed as means ± SD, except for skewed variables expressed as medians with interquartile ranges. † In cases of multiple drug intake, only the main drug was considered. ‡ No cause was identified in 5 patients, and 1 non-severe acute hepatitis patient had heart failure. CBG, corticosteroid-binding globulin; INR, International Normalized Ratio; ND, not determined.

(one grade I and seven had grade II) according to Trey and Davidson classification [19]. None of SAH patients were eligible for liver transplant listing at initial evaluation [20]. 3.2. Effect of the severity of acute hepatitis on adrenal function The median values of T0 and T60 STC concentrations did not differ significantly between SAH, NSAH and HC (Fig. 1 and Supplementary Table S1). Conversely, percent-SFC at T0 (HC: 8.2 ± 2.0% vs. SAH: 18.6 ± 9.0%; p < 0.001) and T60 (HC: 12.8 ± 2.3% vs. SAH: 28.0 ± 6.0%; p < 0.001) increased significantly from HC to SAH, together with a decrease in albumin and CBG concentrations (Fig. 1 and Supplementary Table S1). Using the 25th percentile of CBG and albumin distribution to distinguish between high and low values, the thresholds were respectively 28 mg/L (“low CBG” <28 mg/L) and 28 g/L (“low albumin” < 28 g/L) in the 74 patients with acute hepatitis (Table 2). According to the cortisol-binding proteins, the main differences in cortisol variations were mainly highlighted using CBG as dichotomous variable (Table 2). There was no significant difference in STC at T0 or T60 between patients with low and high CBG, but SFC at both T0 and T60 was significantly higher in those with low CBG. SFC concentrations in SAH patients were correlated with INR at T0 (r = 0.27, p = 0.082) and at T60 (r = 0.40, p = 0.008). The correlations between T0 and T60 STC and INR were not significant. CBG concentrations were closely correlated with INR in SAH (r = −0.63, p < 0.0001) and in NSAH (r = −0.62, p = 0.0003) but not in HC (r = 0.07, p = 0.73). In SAH patients, CBG and albumin concentrations were not correlated with T0 STC (CBG: r = 0.05, p = 0.77; albumin: r = −0.23, p = 0.16) but were negatively correlated with T0 SFC (CBG: r = −0.37, p = 0.02; albumin: r = −0.40, p = 0.01). Comparisons of STC, SFC, CBG and INR in patients with acetaminophen-related (n = 74) or non-acetaminophen-related acute hepatitis are reported in Appendix C. 3.3. Relationship between STC and SFC In all three groups there was a moderate linear correlation between STC and SFC concentrations at T0 (Spearman correlation, R2 = 0.36, p < 0.0001) and at T60 (R2 = 0.47, p < 0.0001). Considering only T0 cortisol values, the quadratic model did not fit the data

better than the linear model. Analysis of covariance showed that the regression lines differed significantly between the three groups at T0 (p < 0.0001, Fig. 2). This model fitted well to our data set (coefficient of determination, R2 = 0.70). At equal concentrations of T0 STC, the T0 SFC concentration was significantly higher in SAH than in NSAH (p < 0.001) or in HC (p < 0.001). Similar results were observed at T60 and the model fitted our data set adequately (R2 = 0.88).

3.4. Prevalence of adrenal dysfunction We identified only 4 (9.3%) SAH patients suspected to have AD, although these patients had a well preserved haemodynamic profile during hospitalization and an adequate FCI [18,21]. None of these four SAH patients received corticosteroids; two received a liver transplant (one for fulminant hepatitis B and one for unknown causes) and two were alive at 6 months without any symptoms suggestive of AD. Another patient suffering from fulminant hepatitis B underwent liver transplantation but had no AD according to Harry’s criteria [12]. We observed only one death in a SAH patient three months after admission, secondary to acute respiratory distress syndrome complicating acute pancreatitis; this patient had no AD at the time of acute hepatitis. No AD was observed in the HC group according to the STC thresholds described above [17].

3.5. Threshold of serum free cortisol Considering the range of T0 SFC values in patients with known AD (2.8–17.8 nmol/L) and in HC (19.7–89.3 nmol/L), we propose a lower normal limit of T0 SFC in the range of 18–20 nmol/L for the diagnosis of AD in healthy persons. Similarly, we purport that the lower normal limit of T60 SFC could be in the range of 60–85 nmol/L (Supplementary Fig. S1). In patients with SAH, the 5th and 10th percentiles of T0 SFC were respectively 31.5 and 34.2 nmol/L, and 181.4 and 215.6 nmol/L for T60 SFC. We assumed that the lower normal limit (specificity of 90–95%) might be in the range of 32–34 nmol/L for T0 SFC and of 180–210 nmol/L for T60 SFC in SAH patients. To clarify our findings, an explanatory figure is reported in Supplementary Fig. S2. By using the 5th percentile of T0 SFC (<31.5 nmol/L) and T60 SFC (<181.4 nmol/L), five patients with SAH were considered to have adrenal dysfunction, two and three having a low T0 SFC and T60

786

T. Degand et al. / Digestive and Liver Disease 47 (2015) 783–789

Fig. 1. Comparison of serum free cortisol and serum total cortisol concentrations in patients with severe and non-severe acute hepatitis and in healthy controls. Box plots represent comparison of serum free and total cortisol in severe acute hepatitis, non-severe acute hepatitis and healthy controls before (T0 ) and after (T60 ) intravenous injection of 250 ␮g of tetracosactrin. The serum total cortisol concentrations at T0 and T60 did not vary significantly between these three groups, unlike serum free cortisol, which increased significantly from healthy controls to severe acute hepatitis. Boxes show interquartile ranges, and error bars represent the lowest and highest observed values within 1.5 times the length of the box. Horizontal lines denote median values. Data points outside this range are shown individually. HC, healthy controls; NSAH, non-severe acute hepatitis; SAH, severe acute hepatitis; SFC, serum free cortisol; STC, serum total cortisol.

Table 2 Acute hepatitis according to serum albumin and cortisol-binding globulin concentrations. Acute hepatitis Albumin <28 g/L n = 19 SAH, n (%) 15 (79) Serum total cortisol (nmol/L) 520.0 [327.0–621.0] T0 195–1222 Range 1064.0 [886.0–1264.0] T60 515–2255 Range Serum free cortisol (nmol/L) 77.0 [53.1–156.0] T0 18.3–336.0 Range 268.0 [208.0–360.0] T60 121–775 Range

p-Value

Acute hepatitis

p-Value

Albumin ≥28 g/L n = 55

CBG <28 mg/L n = 18

CBG ≥28 mg/L n = 56

28 (51)

16 (89)

27 (48)

416.0 [298.0–621.0] 195.0–1222.0 1083.5 [964.0–1365.0] 515.0–2255.0

495.0 [342.5–642.5] 197.0–2244.0 1086.5 [954.0–1297.5] 333.0–2262.0

NS

103.1 [61.2–157.0] 33.1–336.0 333.0 [295.0–404.0] 121.0–775.0

56.6 [43.6–81.9] 18.2–546.0 245.5 [196.5–305.5] 41.2–665.0

0.0024

461.0 [441.0–668.0] 231–2244 1096.0 [966.0–1352.0] 333–2662

NS

59.2 [44.3–85.2] 18.2–546 266.0 [211.0–325.0] 41.2–665

NS

NS

NS

Values are medians with interquartile ranges. Comparisons with Mann–Whitney test show significant difference if p < 0.05. CBG, cortisol-binding globulin; NS, non-significant; SAH, severe acute hepatitis.

NS

0.0024

T. Degand et al. / Digestive and Liver Disease 47 (2015) 783–789

Fig. 2. Scatter plots and regression lines of serum free and total cortisol in patients with severe and non-severe acute hepatitis, and in healthy controls. The 103 points plotted on the figure represent 43 patients with severe acute hepatitis (solid circles), 31 patients with non-severe acute hepatitis (stars) and 29 healthy controls (open circles). By analysis of covariance, the regression lines in the three groups differed significantly (p < 0.0001). For each value of T0 serum total cortisol, the value of T0 serum free cortisol was higher in severe acute hepatitis patients than in non-severe acute hepatitis patients or in healthy controls. HC, healthy controls; SAH, severe acute hepatitis; NSAH, non-severe acute hepatitis.

SFC, respectively. Only one of these five patients had also a low T0 STC (<250 nmol/L). 3.6. Estimation of calculated free cortisol The Bland–Altman plots comparing measured and calculated SFC by Coolens’ formula (Fig. 3) and by the cubic equation (Fig. 4) showed a systematic bias of −44.1% (95% range of percent error

787

Fig. 4. Bland–Altman plots showing percent error of free cortisol estimated using the cubic solution compared to measured free cortisol. Bland and Altman plot of percent difference between free cortisol estimated by the cubic solution and measured free cortisol (y-axis) vs. the average of estimated and measured free cortisol concentration (x-axis); symbols as in Fig. 2. The solid line (bias = +13.7%) indicates that the cubic solution overestimates measured free cortisol by an average of 13.7%, and the dashed lines indicate the range of percent error (bias ± 1.96 standard deviation with 1.96* standard deviation = 64.3%), which includes approximately 95% of the errors.

−118.8 to 30.6%) for Coolens’ formula, and of +13.7% (95% range of percent error −50.6 to 78.0%) for the cubic solutions. As shown in Figs. 3 and 4, the bias observed with the cubic solution was significantly lower than that observed with Coolens’ solution (paired t-test, p < 0.001). Moreover, variability was statistically reduced with the cubic solution (SD = 32.8%) compared to Coolens’ solution (SD = 38.1%; Wilks’ test, p = 0.006). 4. Discussion

Fig. 3. Bland–Altman plots showing percent error of free cortisol estimated using Coolens quadratic solution compared to measured free cortisol. Bland and Altman plot of percent difference between free cortisol estimated by Coolens’ solution and measured free cortisol (y-axis) vs. the average of estimated and measured free cortisol concentration (x-axis). Severe acute hepatitis (solid symbols, n = 43) and non-severe acute hepatitis and healthy controls (open symbols, n = 60) are shown for T0 (circle symbols) and T60 post-cosyntropin (triangle symbols). The solid line (mean bias = −44.1%) indicates that Coolens’ equation underestimates measured free cortisol by an average of −44.1% and the dashed lines indicate the range of percent error (bias ±1.96 standard deviation with 1.96* standard deviation = 74.7%).

The main findings of our study are, firstly, that SFC concentrations increased significantly in SAH patients as compared to NSAH and HC, whereas STC concentrations were not significantly different between these three groups. Secondly, the cortisol-binding proteins (CBG and albumin) decreased markedly in the more severe patients with acute hepatitis. This decrease concerned mainly CBG, resulting in higher SFC and lower STC concentrations in patients with low CBG as compared to those with high CBG. Thirdly, we suggest, for the first time, lower normal ranges of T0 SFC (32–34 nmol/L) and T60 SFC (78–90 nmol/L) in SAH patients. As previously shown in cirrhotic [10,21,22] and critically ill patients [11,23,24], high SFC concentrations better reflect the severity of acute hepatitis than STC concentrations. This is because the decrease in CBG or albumin concentrations is expected to alter the measurement of STC, leading consequently to misleading overestimation of AD. Compared to HC, CBG and albumin concentrations in SAH decreased by 28% and 24%, respectively. This reduction in these proteins was expected since these two cortisol transporters can decrease by up to 50% within 24 h in severe infections and trauma, two other conditions that, like in SAH, harbour high levels of pro-inflammatory cytokines [9,25]. Although we did not correlate concentrations of pro-inflammatory cytokines with SFC, it would be interesting to investigate this potential relation in future studies. The high concentration of SFC usually reflects stress-induced activation of the hypothalamic–pituitary–adrenal axis. Surprisingly, in SAH patients with high levels of SFC, ACTH concentrations were as high as in HC. This paradoxical dissociation

788

T. Degand et al. / Digestive and Liver Disease 47 (2015) 783–789

between SFC and ACTH concentrations has been already reported in other stress situations and has been interpreted as non-ACTHdriven cortisol production, in which cytokines could play a role [26,27]. We may also hypothesize that such paradoxical dissociation may be related to suppressed expression and activity of cortisol-metabolizing enzymes, contributing to sustained hypercortisolemia with enhanced negative-feedback inhibition of ACTH [28]. However, it remains possible that we missed the expected initial rise of ACTH consecutive to the stress event. In the setting of acute stress like SAH, it is unlikely that low serum ACTH levels lead to adrenal atrophy, because the acute phase of SAH is rarely prolonged. In our study, the baseline values of SFC were closely correlated with INR, a marker of severity of acute hepatitis, whereas there was no correlation between STC concentrations and INR. Moreover; comparison of regression lines using covariance analysis (Fig. 2) showed that; at equal levels of STC, SFC concentrations were constantly higher for SAH than for NSAH or HC. The lack of correlation between INR and STC concentrations may be explained by the decrease in CBG due to pro-inflammatory cytokines in SAH patients. Indeed, STC concentrations did not vary significantly between patients with “low” and “high” CBG, unlike SFC concentrations. Our study confirms that CBG is a major cortisol-binding protein, because a decrease in CBG affects SFC concentrations, unlike hypoalbuminemia. The lower CBG concentrations observed in SAH patients could be due to reduced liver synthesis and proteolytic cleavage by neutrophil elastase, in turn releasing free cortisol but also degrading CBG. Furthermore, CBG is characterized by a oneto-one molar binding ratio, with high affinity, and rapid saturation due to its low capacity [16]. All these mechanisms may explain the substantial increase in SFC concentrations during SAH. In our study, the prevalence of AD assessed by the measurement of STC and the 250 tetracosactrin test in SAH patients was only 9.3%, a figure that is far lower than the 62% prevalence previously observed [12]. The 1-␮g low-dose test has been suggested to unmask subtle AD, not shown with the 250 ␮g dose, in persons with no or moderate stress; this 1-␮g low-dose test, not performed in the present study, would probably not improve AD diagnosis in highly stressed SAH patients [29]. We acknowledge that the severity of acute hepatitis in the patients enrolled in our study was less than that reported in Harry’s study [12]. Indeed, none and 7% of our patients required respectively vasopressor support and emergency liver transplantation (vs. 51% and 22% in [12]). Such important differences surely account for the lower prevalence of suspected AD in our SAH patients. However, the four SAH patients in our study suspected to have AD using STC all displayed lower CBG than the SAH patients with no AD (data not shown). In any case, all SAH patients had an adequate free cortisol index, which has been correlated with SFC in several studies [18,21]. Notably, we observed that the diagnosis of patients suspected of having AD using STC vs. SFC differed, except for one patient diagnosed by both methods. This highlights that current cortisol testing remains controversial for the interpreting adrenal gland function. Further longitudinal studies with serial dosages of cortisol are warranted to explore this issue in greater depth. Our findings indicate that STC concentration may be misleading in patients with acute hepatitis, while SFC concentrations provide a more meaningful reflection of disease severity and adrenocortical function. A similar conclusion has been reached in other conditions associated with decreased CBG and albumin concentrations, including cirrhosis, critical illness, sepsis, and septic shock [16,24,29]. However, current techniques for the direct measurement of SFC are complex and costly, which limit the current use of direct measurement of SFC in the clinical evaluation of adrenocortical function. An alternative approach having superior clinical utility

involves estimation of SFC using stoichiometric principles, which incorporate measured concentrations of STC and CBG [15,16]. Several stoichiometric models that vary by degree of complexity have been validated for estimation of SFC [15,16,30]. A limitation of Coolens’ equation in some settings is its assumptions of (i) normal CBG affinity for cortisol, (ii) normal albumin concentration, and (iii) normal albumin affinity for cortisol. In clinical situations such as cirrhosis, critical illness, and septic shock where aspects of cortisol binding to serum proteins may vary significantly from these assumptions [11,21], agreement between Coolen’s and measured SFC may be reduced. In our study, Coolens equation was associated with significant, negative bias (−44%, p < 0.001), such that SFC estimated using Coolens’ equation was consistently lower than measured SFC. Our results differ from those in cirrhotic patients, where positive bias of Coolens’ equation was observed [21]. Differences in CBG affinity for cortisol in patients with cirrhosis compared to patients with acute hepatitis, may account for the observed discrepancy in bias using Coolens’ equation in different study populations. For example, reduced CBG affinity for cortisol would lead to negative bias of Coolens’ compared to measured SFC, whereas increased cortisol binding affinity would lead to positive bias. Factors that reduce CBG affinity for cortisol without affecting CBG-immunoreactivity include increased temperature, decreased glycosylation of CBG, and elastase cleavage of CBG [31]. Therefore, we hypothesize that reduced CBG affinity for cortisol, perhaps related to fever or alterations in CBG glycosylation and/or higher concentrations of elastase-cleaved CBG, may contribute to the negative bias of Coolens’ equation observed in our subjects with acute hepatitis. As shown in Figs. 3 and 4, the cubic solution [16] has better agreement with measured SFC (reduced bias and variability) than Coolens equation. Our results, in addition to other direct comparisons of Coolens’ and cubic solutions, indicate that the cubic solution provides a more accurate estimate of SFC in defined clinical settings, including acute hepatitis in addition to cirrhosis, sepsis, and septic shock [16,29]. To date, no normal ranges for SFC during critical illness (like SAH) have been established. Published data on the normal lower limit of SFC are sparse, and thresholds of 33 nmol/L [32,33] and 55.2 nmol/L [24] have been suggested in healthy individuals and in critically ill patients respectively. We propose normal lower ranges of SFC at T0 (18–20 nmol/L) and T60 (60–85 nmol/L) for healthy individuals. We also suggest normal lower ranges for T0 (32–34 nmol/L) and T60 (180–210 nmol/L) for patients in situation of stress, like those with SAH. However, we acknowledge that the appropriate SFC range required to produce its effect currently remains unknown, partly due to the current inability to assess tissue glucocorticoid resistance which is an important mechanism responsible for AD [29]. Future prognostic studies should clarify which patients should receive corticosteroids at a given SFC concentration. A further limitation is that we did not include a group of patients in whom to validate the proposed normal ranges of SFC. In summary, our study indicates that SFC better reflects the severity of acute hepatitis and adrenal reserve than STC, because cortisol-binding proteins are low in this context, leading to misdiagnosis (over and perhaps infra-diagnosis) of AD when assessed by STC. Stoichiometric methods such as Coolens’ and cubic solutions have superior clinical utility compared to direct measurement of SFC; in patients with acute hepatitis, estimation of SFC using the cubic solution was significantly more accurate than the Coolens’ solution. Finally, the lower normal ranges of SFC proposed herein in healthy subjects and in highly stressed patients such as those with SAH, remain to be validated.

Conflict of interest None declared.

T. Degand et al. / Digestive and Liver Disease 47 (2015) 783–789

Funding Dr. Thévenot received a grant from the French Inter-regional Program for Clinical Research (R/2011/42). Acknowledgments We thank Sigma-Tau France laboratory for logistic support and Fiona Ecarnot (EA3920, University Hospital Besancon, France) for editorial support. We also thank the CORT-HEPAT study group which includes the following members: I Belnard, CE Ber, L Bermont, P Beyne, G Blasco, JP Cervoni, L Chardon, S Dritsas, L Elkrief, V Esnault, C Gombert, O Guillaud, M Héberlé, P Mathurin, K Mouyabi, E Muel, I Ogier, P Pham, C Richou, Rouchet C, D Samuel, F Schillo, D Valla, C Vanlemmens and M Woronoff-Lemsi. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at http://dx.doi.org/10.1016/j.dld.2015.05.016 References [1] Wang D-W. Advances in the management of acute liver failure. World Journal of Gastroenterology 2013;19:7069. [2] Rothwell PM, Udwadia ZF, Lawler PG. Cortisol response to corticotropin and survival in septic shock. Lancet 1991;337:582–3. [3] Tsai M-H, Peng Y-S, Chen Y-C, et al. Adrenal insufficiency in patients with cirrhosis, severe sepsis and septic shock. Hepatology 2006;43: 673–81. [4] Annane D, Sébille V, Troché G, et al. A 3-level prognostic classification in septic shock based on cortisol levels and cortisol response to corticotropin. Journal of the American Medical Association 2000;283:1038–45. [5] Borrelli E, Roux-Lombard P, Grau GE, et al. Plasma concentrations of cytokines, their soluble receptors, and antioxidant vitamins can predict the development of multiple organ failure in patients at risk. Critical Care Medicine 1996;24:392–7. [6] Jäättelä M, Ilvesmäki V, Voutilainen R, et al. Tumor necrosis factor as a potent inhibitor of adrenocorticotropin-induced cortisol production and steroidogenic P450 enzyme gene expression in cultured human fetal adrenal cells. Endocrinology 1991;128:623–9. [7] Webster JC, Oakley RH, Jewell CM, et al. Proinflammatory cytokines regulate human glucocorticoid receptor gene expression and lead to the accumulation of the dominant negative beta isoform: a mechanism for the generation of glucocorticoid resistance. Proceedings of the National Academy of Sciences of the United States of America 2001;98:6865–70. [8] Bartalena L, Hammond GL, Farsetti A, et al. Interleukin-6 inhibits corticosteroid-binding globulin synthesis by human hepatoblastoma-derived (Hep G2) cells. Endocrinology 1993;133:291–6. [9] Beishuizen A, Thijs L, Vermes I. Patterns of corticosteroid-binding globulin and the free cortisol index during septic shock and multitrauma. Intensive Care Medicine 2001;27:1584–91. [10] Thevenot T, Borot S, Remy-Martin A, et al. Assessment of adrenal function in cirrhotic patients using concentration of serum-free and salivary cortisol: adrenal insufficiency and cirrhosis. Liver International 2011;31: 425–33.

789

[11] Ho JT, Al-Musalhi H, Chapman MJ, et al. Septic shock and sepsis: a comparison of total and free plasma cortisol levels. Journal of Clinical Endocrinology and Metabolism 2006;91:105–14. [12] Harry R, Auzinger G, Wendon J. The clinical importance of adrenal insufficiency in acute hepatic dysfunction. Hepatology 2002;36:395–402. [13] Arabi YM, Aljumah A, Dabbagh O, et al. Low-dose hydrocortisone in patients with cirrhosis and septic shock: a randomized controlled trial. Canadian Medical Association Journal 2010;182:1971–7. [14] Sprung CL, Annane D, Keh D, et al. Hydrocortisone therapy for patients with septic shock. New England Journal of Medicine 2008;358:111–24. [15] Coolens JL, Van Baelen H, Heyns W. Clinical use of unbound plasma cortisol as calculated from total cortisol and corticosteroid-binding globulin. Journal of Steroid Biochemistry 1987;26:197–202. [16] Dorin RI, Pai HK, Ho JT, et al. Validation of a simple method of estimating plasma free cortisol: role of cortisol binding to albumin. Clinical Biochemistry 2009;42:64–71. [17] Oelkers W. Adrenal insufficiency. New England Journal of Medicine 1996;335:1206–12. [18] Le Roux CW, Sivakumaran S, Alaghband-Zadeh J, et al. Free cortisol index as a surrogate marker for serum free cortisol. Annals of Clinical Biochemistry 2002;39(Pt 4):406–8. [19] Trey C, Davidson CS. In: Popper H, Shaffner F, editors. Progress in liver diseases: the management of fulminant hepatic failure. Grune and Stratton, 3. New York and London: Heinemann; 1970. p. 282–98. [20] Ichai P, Legeai C, Francoz C, et al. Patients with acute liver failure listed for super-urgent liver transplantation in France: re-evaluation of the clichy-villejuif criteria. Liver Transplantation 2015;21:512–23. [21] Fede G, Spadaro L, Tomaselli T, et al. Comparison of total cortisol, free cortisol, and surrogate markers of free cortisol in diagnosis of adrenal insufficiency in patients with stable cirrhosis. Clinical Gastroenterology and Hepatology 2014;12:504–12. [22] Thevenot T, Dorin R, Monnet E, et al. High serum levels of free cortisol indicate severity of cirrhosis in hemodynamically stable patients. Journal of Gastroenterology and Hepatology 2012;27:1596–601. [23] Poomthavorn P, Lertbunrian R, Preutthipan A, et al. Serum free cortisol index, free cortisol, and total cortisol in critically ill children. Intensive Care Medicine 2009;35:1281–5. [24] Hamrahian AH, Oseni TS, Arafah BM. Measurements of serum free cortisol in critically ill patients. New England Journal of Medicine 2004;350:1629–38. [25] Torpy DJ, Ho JT. Value of free cortisol measurement in systemic infection. Hormone and Metabolic Research 2007;39:439–44. [26] Bornstein SR, Engeland WC, Ehrhart-Bornstein M, et al. Dissociation of ACTH and glucocorticoids. Trends in Endocrinology & Metabolism 2008;19:175–80. [27] Vermes I, Beishuizen A, Hampsink RM, et al. Dissociation of plasma adrenocorticotropin and cortisol levels in critically ill patients: possible role of endothelin and atrial natriuretic hormone. Journal of Clinical Endocrinology and Metabolism 1995;80:1238–42. [28] Boonen E, Vervenne H, Meersseman P, et al. Reduced cortisol metabolism during critical illness. New England Journal of Medicine 2013;368:1477–88. [29] Thevenot T, Weil D, Di Martino V. Surrogate markers of free cortisol in cirrhotic patients: another step has been reached. Clinical Gastroenterology and Hepatology 2014;12:513–5. [30] Nguyen PT, Lewis JG, Sneyd J, et al. Development of a formula for estimating plasma free cortisol concentration from a measured total cortisol concentration when elastase-cleaved and intact corticosteroid binding globulin coexist. Journal of Steroid Biochemistry and Molecular Biology 2014;141:16–25. [31] Chan WL, Carrell RW, Zhou A, et al. How changes in affinity of corticosteroid-binding globulin modulate free cortisol concentration. Journal of Clinical Endocrinology and Metabolism 2013;98:3315–22. [32] Vogeser M, Briegel J, Zachoval R. Dialyzable free cortisol after stimulation with Synacthen® . Clinical Biochemistry 2002;35:539–43. [33] Lewis JG, Bagley CJ, Elder PA, et al. Plasma free cortisol fraction reflects levels of functioning corticosteroid-binding globulin. Clinica Chimica Acta 2005;359:189–94.