Introduction

Cortisol is released during stress following activation of the hypothalamo-pituitary adrenal axis (HPAA) [1]. A non-functioning HPAA during stress is associated with a high mortality [2]. Although similar data in humans do not exist, it is widely accepted that the activity of the HPAA is of vital importance in critical illness; however, reliably estimating activity of the HPAA has been difficult for the following reasons: (a) limitations of a random plasma cortisol [3]; (b) the demonstration of a wide range of plasma cortisol (460–1400 nmol/l) in stressed ICU patients [1, 35] resulting in a number of different basal total plasma cortisol concentrations (415 mol/l, 500 and 550 nmol/l) to be proposed as thresholds to indicate adrenal insufficiency [68]; (c) a reproducible correlation between total plasma cortisol and outcome has been difficult to establish [1, 3, 5, 912]; (d) a lack of consensus on what constitutes a “normal plasma cortisol” response to corticotropin in critical illness [5]; and (e) a lack of clarity on the optimum dose of corticotropin to be used [13].

Recent focus on plasma-free cortisol, the bio-active fraction, is clearly a step forward [1416]; however, it is the cortisol concentrations in target tissues which are the most appropriate pharmacokinetic index for the rational therapy of steroids, yet probably the area about which we know least in critically ill patients.

The concentration of cortisol at the receptor site is a balance between synthesis and metabolism. Although a number of enzymes play a role in the metabolism of cortisol (11β-hydroxysteroid dehydrogenase, 5 alpha and beta reductases and 6 beta hydroxylase), quantitatively, the activity of 11β-hydroxysteroid dehydrogenase (11β-HSD) is the most important pathway. The 11β-HSD modulates the selectivity, specificity, and intensity of glucocorticoid-dependent processes and regulates intracellular concentrations of cortisone (inactive) and active cortisol. Two isoforms of this enzyme have been identified: type 1 (11β-HSD1); and type 2 (11β-HSD2) [17, 18]. The 11β-HSD1 appears to be primarily reductase, converting inactive cortisone to active cortisol. Conversely, 11β-HSD2, a dehydrogenase, has its major site of action in the kidney, where it functions to inactivate cortisol to cortisone prior to its binding to the mineralocorticoid receptor. The ratio of plasma cortisol to cortisone [F:E ratio] reflects the activity of this enzymatic process [19, 20, 21]. Elevations in F:E ratios have been demonstrated in hypertension, obesity, and the metabolic syndrome [22, 23]. Preliminary data from post-operative cardiac surgical patients suggest an increase in F:E ratios indicative of altered 11β-HSD activity [26]; however, these patients represent a group of stable post-operative critically ill cohort. Other common groups of critically ill patients include those suffering from sepsis, trauma, and burns, who have varying degrees of sickness severity. There are no published data on 11β-HSD activity in these groups of patients.

We therefore undertook a prospective observational study to (a) identify the extent of activity of 11β-hydroxysteroid dehydrogenase system, assessed by serum F:E ratio, in three cohorts of critically ill patients, and (b) define serial changes in F:E ratio over the course of the illness.

Materials and methods

The study was approved by the Royal Brisbane Hospital Ethics Committee and informed consent obtained from the patients or their next of kin. From August 2002 to December 2003 we screened patients admitted to our intensive care unit with diagnoses of sepsis, burns, or trauma.

The inclusion criteria were as follows:

  1. 1. 

    Severe sepsis or septic shock. Severe sepsis was defined as sepsis associated with organ dysfunction, hypoperfusion, or hypotension. Sepsis is defined as the presence of two or more of the following conditions as a result of infection; temperature > 38 °C or < 36 °C; heart rate > 90 beats/min; respiratory rate > 20 breaths/min or a PaCO2 < 32 torr; WBC > 12,000 cells/mm3 or > 10% band forms. Septic Shock was defined as sepsis with hypotension despite adequate fluid resuscitation requiring vasopressors.

  2. 2. 

    Burns. Severe burns affecting > 30% of body surface area.

  3. 3. 

    Multiple trauma. Blunt or penetrating trauma of at least two body regions requiring resuscitation and admission to the critical care unit.

Patients less than 16 years of age, those with a previous history of adrenal or pituitary disease, prolonged use of oral or inhaled glucocorticoids or current therapy with any such agents were excluded.

All patients had continuous ECG and invasive arterial pressure monitoring as part of standard clinical practice. Patients were sedated with midazolam and fentanyl. Endotracheal intubation and positive pressure ventilation were performed when clinically indicated. No patients received etomidate or ketoconazole.

Data collection

Demographic data, Acute Physiology and Chronic Health Evaluation Score (APACHE) II, Simplified Acute Physiology Score (SAPS) II and Injury Severity Score (ISS) estimations were collected on the day of study.

Blood samples were collected for analysis daily for the first 5 days of admission. Samples were analysed for total plasma cortisol, cortisone, and C-reactive protein (CRP). Subsequent testing was performed on days 7, 10, 15 and 28 where possible. Sampling time was 05:00 each day as part of the routine investigations for the unit.

Biochemical assays

Cortisol and cortisone were measured by high-pressure liquid chromatography (HPLC) as detailed by McWhinney et al. [24]. The coefficients of variation (CV) at cortisol levels of 23 nmol/l and 1006 nmol/l were 6.5 and 2.4%, respectively. The CV for cortisone at a concentration of 110 nmol/l was 10.9% and at 1026 nmol/l was 9.2%.

The CRP was assayed using an immunoturbimetric assay with Roche Modular Clinical Chemistry instrumentation (Roche Diagnostics, Sydney Australia). Interassay CV at 13 mg/L was 4.5%.

Statistical analysis

The means and standard deviations for normals were obtained from population values from our laboratory (as noted in the Results section) and these were then used to compare with the patient populations. Continuous, normally distributed variables were summarized as mean ± standard deviation (SD). Gender interactions were tested and not identified. Differences between groups at baseline were assessed using independent t-test. The degree of association between variables (such as cortisol, cortisone and cortisol/cortisone ratio) and skewed or ordinal outcome measures (such as APACHE and SAPS) was assessed using Spearman's correlation coefficient (r s). One-way ANOVA was used to assess linear trend of parameters across time for each patient group (e.g. sepsis, trauma or burns). The effect of time and group on cortisol/cortisone ratio was assessed using an unweighted means analysis (two-way ANOVA). Statistical significance was taken at a level of 5%. For correlative analysis, and statistical significance accounted for multiple comparisons and was taken at the level of 1% (p < 0.01). All analyses were carried out using SPSS software version 12.0 (SPSS, Chicago, Ill.).

Results

Fifty-two patients were enrolled: 13 in the sepsis cohort (5 with severe sepsis and 8 with septic shock); 20 in the trauma cohort; and 19 in the burns cohort. The demographic data, sites of infection and range of burn surface area are presented in Table 1.

Table 1 Demographic data; LOS length of stay in ICU; ISS Injury Severity Score; n/a not applicable

Cortisol and cortisone values are reported in nanomoles per liter; to convert to micrograms per deciliter, 1 μg/dl = 27.6 nmol/l. (As molecular weights of cortisol and cortisone are nearly the same (362 vs 360), the same conversion factor can be used.)

Cortisol, cortisone, F:E ratio and CRP values over the duration of the study are presented in Table 2. Due to the small numbers of patients reaching day 28, only the data to day 10 are presented. The normal reference range for plasma cortisol, cortisone and F:E ratios in healthy volunteers are 190–750 (reference range from Queensland Health Pathology Services), 22–74 and 3–12 nmol/l, respectively [25]. The normal reference range for CRP in our institution is 0–5 mg/l.

Table 2 Total plasma cortisol, cortisone, F:E ratio and CRP values over the duration of the study

Day 1: Baseline results

Total plasma cortisol, cortisone, and cortisol-cortisone (F:E) ratio

The mean total plasma cortisol was significantly elevated on day 1 in the sepsis (844 ± 520 nmol/l, p < 0.0001) and trauma groups (548 ± 178 nmol/l, p < 0.0001) as compared with the normal reference range in healthy volunteers. In contrast, in the burns cohort, the mean total plasma cortisol concentration on day 1 was not significantly elevated above this range (424 ± 340 nmol/l, p = 0.64).

Mean plasma cortisone was also significantly elevated on day 1 in the sepsis and trauma cohorts (36 ± 17 nmol/l, p = 0.05, 34 ± 19 nmol/l, p = 0.001) as compared with control data. In the burns cohort the mean plasma cortisone on day 1 was not significantly elevated (38 ± 21 nmol/l, p = 0.059).

Similarly, the F:E ratios were significantly elevated in the sepsis and trauma cohorts (22 ± 9, p = 0.01, and 23 ± 19, p = 0.0003, respectively) as compared with control data. In the burns cohort, the F:E ratio was not significantly elevated at day 1 (16 ± 15, p = 0.17).

Day 2–10: Results

Sepsis

Total plasma cortisol, cortisone and F:E ratio all remained significantly elevated over control values for the ten recorded days of the study, and no significant differences were observed within the cohort (p = 0.89, p = 0.73, and p = 0.43, respectively).

Trauma

Plasma cortisone and F:E ratio remained significantly elevated over normal values up to day 10, and no significant intra cohort variations were observed for these variables (p = 0.13, p = 0.18), respectively; however, total plasma cortisol was observed to show a significant decrease on days 3–5 when compared with day 1 (p < 0.05).

Burns

Total plasma cortisol, plasma cortisone and F:E ratio were not observed to show any significant change from control values over the ten recorded days of the study. No significant intra cohort variation was observed (p = 0.3, p = 0.34, p = 0.23, respectively).

CRP values

The CRP values were above the normal range in all cohorts for the duration of the trial. Statistically significant increases over baseline values were observed in the burns and trauma cohorts from day 3 onwards.

Inter-cohort variation

Total plasma cortisol

Total plasma cortisol levels were significantly higher over time in the sepsis group when compared with the trauma (p = 0.009) and burns (p = 0.02) group. There was no significant difference between the trauma and burns group (p = 0.80).

Cortisone

There was a time effect on cortisone with a trend to lower levels by day 10 (p = 0.02); however, this was not statistically significant within groups (sepsis p = 0.73, trauma p = 0.13, burns p = 0.34). There was no group effect on cortisone levels until after day 4 when trauma patients maintained a lower cortisone level compared with patients with burns (p = 0.02). There was no difference between sepsis and trauma (p = 0.19) or sepsis and burns (p = 0.59).

Cortisol: cortisone (F:E) ratio

There appears to be a trend towards an increase in F:E ratio over time in the sepsis and the trauma groups particularly over D5 and D7 (Fig. 1); however, there was no statistically significant difference in F:E ratio over time within groups. This is in part because of the diminishing numbers of patients over time (sepsis D1–13, D5–12, D7–12, trauma D1–20, D5–14, D7–13, burns D1–19, D5–18, D7–16); however, across groups F:E ratios in trauma were significantly elevated over those in burns (p = 0.002). No significant difference was observed between the values for sepsis and burns (p = 0.12) and sepsis and trauma (p = 0.42).

Fig. 1
figure 1

Serial plasma F:E ratios over time in sepsis (filled squares); trauma (open circles) and burns (inverted triangles) cohorts

Impact of renal dysfunction on plasma cortisone and F:E ratios

As the kidney is the predominant source of cortisone in humans, we examined whether the presence of renal dysfunction would impact on serum cortisone and F:E ratios. During the study period, 7 patients developed renal dysfunction (3 in sepsis and 4 in the burns group) as defined by a serum creatinine of > 300 μmol/l or a doubling of serum creatinine over a 24-h period. There was no significant difference between the two groups (renal dysfunction vs normal renal function) with respect to plasma cortisone (34 ± 23 vs 28 ± 17 nmol/l, p = 0.13) or F:E ratio (23 ± 18 vs 20 ± 27, p = 0.33).

Correlation with illness severity and outcome

When all patients were analysed together as a group, there was no significant relationship between cortisol, cortisone, F:E ratio and illness severity, although there was a trend for a positive relationship between day 5 F:E ratio and ISS (p = 0.015, r = 0.81). Patients were then analysed according to disease-specific groups and a consistent relationship between F:E ratios and sickness severity was demonstrable only on day 7 in the burns group. Plasma F:E ratio was significantly correlated with APACHE II (r = 0.77, p = 0.0008) and SAPS II (r = 0.7, p = 0.003) on day 7 in the burns cohort. The correlations between F:E ratio and SAPS score were r = 0.65 and p = 0.03 on day 4 in the sepsis cohort and between F:E ratio and ISS on day 4 (r = 0.59, p = 0.02) in the trauma group. It is noteworthy that no statistically significant correlation was demonstrable between either plasma cortisol with sickness or injury severity on any day in any of the cohorts.

Relationship between CRP, cortisol, cortisone, F:E ratios, and mortality

In the total cohort, day 10 CRP was significantly lower in those who survived hospital admission compared with those who died (197 ± 119 vs 347 ± 80, p = 0.03); however, no statistically significant correlation was observed between plasma cortisol, cortisone, or F:E ratio and mortality in illness-specific cohorts.

Discussion

We believe the data presented in this study to be the first reported investigation into 11β-HSD activity in a heterogeneous population of the critically ill. The findings of this study clearly demonstrate marked alteration of 11β-HSD activity in critical illness as shown by elevations in total plasma F:E ratio. The elevations in septic and trauma patients were statistically significant. This increase in F:E ratio persisted throughout the first 10 days of the study. Interestingly, the observed changes were not due solely to an increase in total plasma cortisol levels; the cortisone levels were also significantly increased, suggesting a shift in total body 11β-HSD activity. The changes could be explained by differential increases in 11β-HSD1 and 11β-HSD2 activity, with a proportionately greater increase in the former.

Mechanisms behind activation of 11β-HSD

Trauma, sepsis and burns are characterized by activation of the pro-inflammatory cytokine cascade as part of the stress response [26]. Pro-inflammatory cytokines have been demonstrated in vitro to upregulate 11β-HSD1 activity. TNF-alpha enhances cortisol availability to the cell by enhancing the activity of 11β-HSD1 and suppressing that of 11β-HSD2 [27, 28]. Such effects have been demonstrated in smooth muscle [29], lung epithelium [30] and adipose tissue [31]. In addition, there is evidence of increased 11β-HSD1 activity both in a general hospital population, and a post-operative cardiac surgical population [32, 33], supporting the hypothesis that 11β-HSD activity is upregulated by stress. Elevations in plasma CRP in our study cohort would support the triggering of acute phase response as a contributor to the upregulation of 11β-HSD1 activity. It is of interest, however, that the burns cohort alone failed to show a significant elevation of F:E ratio, total plasma cortisol or cortisone. This finding is somewhat surprising given the extent of the inflammatory response associated with large burns and the observed increase in total plasma cortisol documented in previous studies [34]. Small sample size or variations in burns surface area might account for this finding; however, the burns cohort contained comparable numbers of patients to the others (19 vs 13 and 20 in the sepsis and trauma cohorts, respectively), and only 5 patients had burn areas under 50%. An alternative explanation may lie in the directionality of 11β-HSD1. Although reductase in intact cells catalyses the formation of cortisol from cortisone, in disrupted cells the enzyme appears to act as a dehydrogenase, catalysing the reaction in the opposite direction [35]. It is possible that the extensive destruction of adipose and skin cells seen in severe burns may produce a sufficient shift in the equilibrium of the 11β-HSD1 enzyme to account for our observations. Significant reductions in cortisol binding globulin in burns [36] could potentially reduce total plasma cortisol and impact on the F:E ratio.

Other potential mechanisms for altered 11β-HSD1 activity include changes in redox potential within the cell. The oxo-reductase activity seen in intact cells requires NADPH (whose concentrations are determined by cytosolic redox) and leads to the activation of glucocorticoids [37]. Septic patients frequently demonstrate alterations in cellular redox potential mediated largely by endotoxin [38]. Patients with trauma and burns may also demonstrate changes in redox from altered tissue perfusion [39].

Our observations of statistically significant elevated ratios in trauma and sepsis over 10 days suggest that the adaptation of the organism to severe stress may be more prolonged than previously thought. These data appear to be in broad agreement with those presented by Vogeser et al., in which the post-operative F:E ratio was persistently elevated in cardiac surgical patients, despite a declining total serum cortisol concentration [33].

F:E ratio as a marker of stress response?

In our study, a significant correlation between plasma F:E ratio with illness severity was noted in only the burns cohort and only on day 7. When all three cohorts were examined together as a heterogeneous group of critically ill patients, no statistically significant relationship was noted between F:E ratios and illness severity. This might merely reflect the small sample size and further studies may be required to investigate the intriguing possibility that 11β-HSD activity may serve as a valid marker for stress response. As noted, published data in post-operative patients support this notion [33].

Significance of the findings

Does an increase in F:E ratio signify increased tissue availability of cortisol? Although no direct evidence exists to support this, published data suggest that this might be the case. Rauz et al. in patients with glaucoma were able to demonstrate increased HSD-1 activity in the epithelial cells of the ciliary body. In keeping with this increased activity, they also noted that aqueous humour F:E ratios (14:1) were greater than those of circulating plasma (3:1) consistent with local increases in HSD-1 activity [40]. Tomlinson et al. have demonstrated that inhibition of 11β-HSD 1 activity is associated with limited glucocorticoid availability in the adipose tissue with functional consequences manifest by a decrease in glycerol release indicative of inhibition of glucocorticoid mediated lipolysis [41].

The upregulation of 11β-HSD activity in critical illness may confer survival benefit to the organism. It is tempting to speculate that the potential for greater availability of cortisol might lead to more hemodynamic stability and preserve tissue perfusion and minimise the need for inotropes or pressor use in shock states [42]. The availability of additional glucocorticoids might also protect the cell from the deleterious effects of excess pro-inflammatory cytokines; however, these putative protective effects of glucocorticoids will need to be balanced against their hyperglycaemic effects, which are associated with excess mortality in critically ill patients [43]. Published data clearly show that 11β-HSD1 knockout mice show attenuated glucocorticoid-inducible responses and resist hyperglycaemia of obesity or stress [44].

Limitations of the study

In this preliminary study we were able to demonstrate activation of this enzyme system in the various cohorts of critically ill patients. We recognise the possibility of type-2 error when analysing small data sets; however, we attempted to overcome this problem by setting the significance value to p < 0.01 for correlative analysis. As only three cohorts of patients were examined, the results cannot be extrapolated to the general intensive care population; however, these three groups represent common diagnostic categories seen in intensive care (e.g. severe sepsis or septic shock, major trauma (as defined by an ISS > 16, the mean ISS in our study group being 28) and major burns. Our previous study showed marked hourly fluctuations in plasma cortisol profile [3], and it could be argued that the F:E ratio might also demonstrate similar variations and therefore a single daily measurement may not be reflective of the 24-h profile. In healthy volunteers and patients with chronic fatigue syndrome, respectively, both plasma and salivary F:E ratios have been shown to be stable and free of any circadian variation [45, 46]. Nevertheless, further work is needed in critically ill patients to determine its variability. We have used historical controls; however, these are from healthy volunteers whose endocrine profile has been analysed in our laboratory. We used plasma F:E ratio, which is an indirect indicator of global 11β-HSD activity. Although the urinary ratio of cortisol [F] to cortisone [E] is traditionally used to assess the activity of the 11β-HSD system, plasma ratios have also been shown to reflect the activity of this enzymatic process. Morita et al. measured plasma F:E ratios (n = 98) in the peripheral blood from the femoral vein and compared with those obtained from selective renal, adrenal, and hepatic vein samples [19]. They concluded that peripheral plasma F&E concentrations were determined by the relative strength of activities of the isoenzymes in the whole body. Morineau et al. have also shown a close concordance between plasma, salivary and urinary F:E ratios in healthy volunteers [20]. Tissue-specific analyses of HSD activity using microdialytic or tissue biopsy studies may provide more insight into changes in tissue-specific enzyme activity during critical illness. Finally, it is also important to recognise that other aspects of critical illness might influence cortisol metabolism such as hepatic dysfunction, although that was not the subject of investigation in this study.

Conclusion

In conclusion, the data presented in this preliminary study suggest that in a heterogeneous cohort of critically ill patients there is altered cortisol metabolism due to a shift in the overall set point of 11β-HSD activity, and that this is persistent over several days. Further studies with larger sample sizes specifically designed to examine altered tissue 11β-HSD activity and its clinical significance and correlation with outcome are warranted.