Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Jul 6.
Published in final edited form as: Crit Care Med. 2015 Mar;43(3):549–556. doi: 10.1097/CCM.0000000000000721

Reversible Increase in Maximal Cortisol Secretion Rate in Septic Shock*

Richard I Dorin 1,2, Clifford R Qualls 3, David J Torpy 4, Ronald M Schrader 3, Frank K Urban III 5
PMCID: PMC8259018  NIHMSID: NIHMS1694047  PMID: 25365720

Abstract

Objective:

Cortisol clearance is reduced in sepsis and may contribute to the development of impaired adrenocortical function that is thought to contribute to the pathophysiology of critical illness–related corticosteroid insufficiency. We sought to assess adrenocortical function using computer-assisted numerical modeling methodology to characterize and compare maximal cortisol secretion rate and free cortisol half-life in septic shock, sepsis, and healthy control subjects.

Design:

Post hoc analysis of previously published total cortisol, free cortisol, corticosteroid-binding globulin, and albumin concentration data.

Setting:

Single academic medical center.

Patients:

Subjects included septic shock (n = 45), sepsis (n = 25), and healthy controls (n = 10).

Interventions:

IV cosyntropin (250 μg).

Measurements and Main Results:

Solutions for maximal cortisol secretion rate and free cortisol half-life were obtained by least squares solution of simultaneous, nonlinear differential equations that account for free cortisol appearance and elimination as well as reversible binding to corticosteroid-binding globulin and albumin. Maximal cortisol secretion rate was significantly greater in septic shock (0.83 nM/s [0.44, 1.58 nM/s] reported as median [lower quartile, upper quartile]) compared with sepsis (0.51 nM/s [0.36, 0.62 nM/s]; p = 0.007) and controls (0.49 nM/s [0.42, 0.62 nM/s]; p = 0.04). The variance of maximal cortisol secretion rate in septic shock was also greater than that of sepsis or control groups (F test, p < 0.001). Free cortisol half-life was significantly increased in septic shock (4.6 min [2.2, 6.3 min]) and sepsis (3.0 min [2.3, 4.8 min] when compared with controls (2.0 min [1.2, 2.6 min]) (both p < 0.004).

Conclusions:

Results obtained by numerical modeling are consistent with comparable measures obtained by the gold standard stable isotope dilution method. Septic shock is associated with generally not only higher levels but also greater variance of maximal cortisol secretion rate when compared with control and sepsis groups. Additional studies would be needed to determine whether assessment of cortisol kinetic parameters such as maximal cortisol secretion rate and free cortisol half-life is useful in the diagnosis or management of critical illness–related corticosteroid insufficiency.

Keywords: adrenal cortex, computer-assisted, cosyntropin, hydrocortisone, metabolic clearance rate, numerical analysis, shock, septic, sepsis

Adrenocortical hypofunction related to decreased production of adrenocorticotropin and cortisol have been proposed to contribute to the pathophysiology of critical illness–related corticosteroid insufficiency (CIRCI) (1). In addition, a decreased rate of cortisol clearance and decreased concentrations of adrenocorticotropin in critically ill subjects, including those with systemic inflammatory response syndrome (SIRS), was recently reported by Boonen et al (2, 3). These and related observations have prompted consideration of the hypothesis that reduced cortisol clearance may lead to suppression of adrenocorticotropin concentrations through negative feedback and consequent impairment of adrenocortical function (2, 3).

Reports of improved 28-day survival in septic shock (SS) patients receiving physiological corticosteroid replacement strengthen the supposition that adrenocortical hypofunction plays a significant and reversible role in the pathophysiology of SS (4, 5). However, in the multicenter, randomized, placebo-controlled The Corticosteroid Therapy of Septic Shock trial, the benefit of corticosteroid administration on survival was not statistically significant (6). A contemporary challenge is the development of validated clinical and/or laboratory criteria that would identify the subset of SS patients who are more likely to benefit from exogenous corticosteroid replacement.

The development of such criteria has been difficult. For example, the increment of total cortisol (Δ-cortisol) from baseline to 60 minutes after 250 μg cosyntropin has not proved useful when applied prospectively to independent study populations (1, 4, 6–9). The laboratory diagnosis of CIRCI is made more complicated by marked variation in the plasma concentration of cortisol-binding proteins and difficulty in measuring or estimating free cortisol concentrations. Furthermore, there may be resistance to the biological actions of corticosteroids, and consequently there is uncertainty regarding the optimal concentrations of total and free cortisol during replacement therapy (1, 10). An additional obstacle is that measured cortisol concentrations may reflect disparate and sometimes countervailing physiological forces. For example, Boonen et al (2) reported that cortisol clearance rate is decreased, whereas baseline cortisol production rate (CPRbase) is increased in critically ill subjects with SIRS. These considerations highlight the limitations of assessment of adrenocortical function using unadjusted total cortisol concentrations. Accordingly, in the present study, we sought to apply computer-assisted numerical modeling methods to derive parameters of free cortisol appearance and elimination rates, including maximal cortisol secretion rate (CSRmax) and free cortisol half-life (11) in SS, sepsis, and control populations.

Adrenal insufficiency (AI) caused by intrinsic adrenal disorders (i.e., primary AI) is associated with decreased CSRmax; a reversible decrease in CSRmax is also associated with adrenocorticotropin deficiency (i.e., secondary AI) (12–14). To the extent that adrenal hypofunction may be present in sepsis and SS (1– 3), we hypothesized that CSRmax would be decreased in these conditions. In addition, we hypothesized that results obtained by numerical modeling in sepsis would replicate those obtained by the isotope dilution method among a similar group of subjects (i.e., critically ill subjects [including SIRS]) (2).

MATERIALS AND METHODS

Study Data Source

The present report is a post hoc analysis of data obtained and previously reported by Ho et al (15). For clinical definitions, inclusion and exclusion criteria, clinical characteristics, and additional laboratory data for SS, sepsis, and controls, see Ho et al (15).

Subjects

The present study analyzed data obtained in SS (n = 45), sepsis (n = 25), and control (n = 10) groups. The age of subjects (mean ± SD) was 63.7 ± 16.2, 54.1 ± 19.1, and 37.5 ± 12.3 years for SS, sepsis, and control groups, respectively. Overall, 59% of subjects were men. Acute Physiology and Chronic Health Evaluation (APACHE) II scores in the SS group were 24.8 ± 8.2, and 28-day hospital mortality was 41% (15). Cosyntropin testing at visit 1 was performed within 24 hours of the time of diagnosis of sepsis or SS and occurred 5.3 ± 7.7 (mean ± SD) and 3.4 ± 7.3 days after hospital admission for SS and sepsis groups, respectively. None of the study subjects received exogenous corticosteroids prior to visit 1 evaluation. Available survivors of SS and sepsis had follow-up cosyntropin testing performed at the time of hospital discharge (visit 2), a subset of surviving SS and sepsis subjects had a third, outpatient cosyntropin stimulation test, which was performed 6–12 weeks after hospital discharge (visit 3). The study was approved by the Royal Adelaide Hospital Institutional Review Board and the University of New Mexico Human Research Review Committee; all subjects gave written informed consent prior to study participation.

Experimental Protocol

Baseline blood samples for plasma total and free cortisol, corticosteroid-binding globulin (CBG), albumin, and adrenocorticotropin were obtained immediately prior to IV administration of adrenocorticotropin-(1–24) (250 μg). Additional blood samples for total and free cortisol and CBG were obtained 30 and 60 minutes after adrenocorticotropin-(1–24) administration (15).

Assay Methods

Plasma total cortisol was measured using an enzyme-linked fluorescent assay (AxSYM Cortisol assay kit; Abbott Laboratories, Abbott Park, IL), with intra- and interassay coefficients of variation (CV) of 5% and 6.5%, respectively. Plasma free cortisol, CBG, and albumin assays are described previously (15).

Estimation of Maximal CSR and Free Cortisol Half-Life

Total cortisol concentration was modeled by summing CBGbound, albumin-bound, and free cortisol compartments computed using the three simultaneous nonlinear, differential equations as described by Dorin et al (11) (Appendix 1, Supplemental Digital Content 1, http://links.lww.com/CCM/B108 and Supplemental Fig. 1, Supplemental Digital Content 2, http://links.lww.com/CCM/B109). The off-rate constants (k–1) for cortisol binding to CBG and albumin were based on previously reported values (16, 17); on-rates were derived using equilibrium dissociation constant (KD) of 33 nM for CBG (18) and 137,800 nM for albumin (19). Mathematical solutions for CSR were performed by integration of the equations within a Levenberg–Marquardt least squares loop implemented in the Matlab programming environment (Appendix 1, Supplemental Digital Content 1, http://links.lww.com/CCM/B108). We fit the cortisol time series and obtained a unique, best-fit CSRmax and free cortisol half-life by minimizing the difference between computed and measured total cortisol concentrations in a least squares sense (11). Our modeling approach requires that cortisol concentrations increase following cosyntropin (250 μg). For purposes of this study, we define a priori a valid increase in cortisol to be greater than 12% (1.75× the CV) from baseline to 60 minutes post cosyntropin.

Estimation of Unstimulated (Baseline) CSR

Cortisol pulsatility has been reported to be reduced in SS and critical illness (3, 20–22). The relative constancy of cortisol concentrations in this clinical setting approximates steady-state conditions and provides a rationale for exploratory analysis of baseline CSR (CSRbase) in SS and sepsis. CSRbase at visit 1 was assessed using steady-state assumptions as described in Equation 4 in Appendix 1, Supplemental Digital Content 1, http://links.lww.com/CCM/B108). Solutions for CSRbase were individualized for measured concentrations of CBG and albumin, as well as the individual free cortisol half-life estimates obtained during cosyntropin stimulation. Because cortisol concentrations do not approximate steady-state conditions in controls (3), CSRbase was not evaluable in control subjects.

Statistical Analysis

Descriptive statistics include median (lower quartile, upper quartile) and box plots. Simple relationships were assessed by Spearman rank correlation test (ρ). Clinical groups were compared by the nonparametric Kruskal–Wallis or Mann– Whitney U tests because the distributions are skewed (SAS 9.3; SAS Institute, Cary, NC). Because the control group was significantly (p < 0.05) younger than SS and sepsis groups, we also performed a secondary analysis for visit 1 adjusting CSRmax for age based on the CSRmax–age relationship reported by Keenan et al (23). Mixed effects models on a log-scale (PROC MIXED, SAS) were used for longitudinal comparisons; post hoc comparisons were done by paired t tests or Wilcoxon signed rank test as appropriate. Variance comparisons were made using F test after log transformation. The accuracy of the model fit was assessed by root mean squared error (RMS).

RESULTS

Estimates of CSRmax and Free Cortisol Half-Life at Visit 1

Unique, best-fit solutions for CSRmax, and free cortisol half-life were obtained for each valid cosyntropin test. For a few (7/148) of the cosynropin studies, this condition was not met and these data were excluded from analysis. Figure 1 illustrates solutions for a representative subject at the time of SS diagnosis (visit 1) and at the time of hospital discharge (visit 2).

Figure 1.

Figure 1.

Open in a new tab

Representative predicted total cortisol concentrations, maximal cortisol secretion rate (CSRmax) and free cortisol half-life (FCHL) in an individual subject at visit 1 (solid line) and 2 (dashed line) compared with measured cortisol concentrations at visit 1 (triangles) and 2 (circles). Goodness of fit (root mean squared error [RMS]) values for visits 1 and 2 are also indicated.

CSRmax was significantly different (Kruskal–Wallis p = 0.01) among the three clinical groups (SS, sepsis, and controls), as shown in Figure 2. In post hoc analysis, the SS group had significantly higher CSRmax (0.83 nM/s [0.44, 1.58 nM/s]) when compared with sepsis (0.51 nM/s [0.36, 0.62 nM/s]; p = 0.007) and control (0.49 nM/s [0.42, 0.62 nM/s]; p = 0.04) groups, respectively. There was no significant difference in CSRmax between control and sepsis groups. Secondary analysis adjusting CSRmax for age did not change the direction or significance of betweengroup comparisons shown in Figure 2.

Figure 2.

Figure 2.

Open in a new tab

Box plot showing the distribution of log-transformed maximal cortisol secretion rate (CSRmax) (visit 1) in controls (n = 10), sepsis (n = 25), and septic shock (n = 45). In a box plot, the bold line in the middle of the box is at the median and the lower and upper lines of the box at the first and third quartiles of the data, whiskers are drawn to the most extreme values that are not outliers; outliers are drawn separately, and outliers are defined as any value more than 1.5× interquartile range (length of the box) beyond a quartile. ns = not significant.

As illustrated in Figure 2, the variance of CSRmax in the SS group was also significantly increased when compared with control and sepsis groups (F test, both p < 0.01, test performed on log-scale). In an effort to identify factors that might underlie the increased variance in SS, we examined correlations of CSRmax with other factors, including age, gender, APACHE score, and 28-day mortality; none of these factors were correlated with CSRmax and therefore the increased variance in SS could not be related to any identifiable factors that might account for the broader range of CSRmax or distinguish subpopulations.

Free cortisol half-life was significantly different (Kruskal– Wallis p = 0.01) between SS, sepsis, and control subjects. In post hoc analysis, shown in Figure 3, free cortisol half-life was not statistically different between SS (4.6 min [2.2, 6.3 min]) and sepsis (3.0 min [2.3, 4.8 min]). Free cortisol half-life was longer in sepsis and SS (both p < 0.004) when compared with controls (2.0 min [1.2, 2.6 min]). Within clinical groups, there was no association between age and free cortisol half-life.

Figure 3.

Figure 3.

Open in a new tab

Box plot showing distribution of log-transformed free cortisol half-life (visit 1) in controls (n = 10), sepsis (n = 25), and septic shock (n = 45).

Accuracy of Model Fit (Goodness of Fit)

RMS values (mean ± SD) at visit 1 were 14.3 ± 21.1 nM (SS), 26.6 ± 44.8 nM (sepsis), and 13.0 ± 15.4 nM (controls).

Longitudinal Follow-Up of CSRmax and Free Cortisol Half-Life

Follow-up cosyntropin tests were repeated immediately prior to hospital discharge (Fig. 4, visit 2); a subset of surviving subjects had additional outpatient assessment (Fig. 4, visit 3). As shown in Figure 4A, CSRmax decreased significantly for SS survivors from visit 1 (0.80 nmol/L/s [0.42, 1.34 nmol/L/s] to visit 2 (0.53 nmol/L/s [0.39, 0.65 nmol/L/s]; Wilcoxon signed rank test, p = 0.04). Free cortisol half-life decreased from visit 1 (4.60 min [2.23, 6.32 min]) to visit 2 (3.96 min [2.14, 5.94 min]), which was not statistically significant (p = 0.60) in paired analysis (Fig. 4B). In sepsis subjects, CSRmax (p = 0.30) and free cortisol half-life (p = 0.19) were not significantly changed following clinical recovery (Fig. 4).

Figure 4.

Figure 4.

Open in a new tab

Longitudinal estimates of maximal cortisol secretion rate (CSRmax) (A) and free cortisol half-life (B) for surviving subjects with septic shock (n = 21) or sepsis (n = 16) who had at least one cosyntropin test after clinical recovery. Visits 1, 2, and 3 (x-axis) refer to serial cosyntropin studies done at the time of initial evaluation (within 24 hr of diagnosis of sepsis or septic shock), hospital discharge, and outpatient evaluation, respectively. Individual data points are shown as open circles; visits for each individual subject are connected by lines. Heavy lines represent median values by group and visit.

Estimation of CSRbase

As shown in Figure 5A, CSRbase in SS was highly variable (0.31 nM/s [0.22, 0.93 nM/s]) and was significantly (p < 0.001) increased when compared with sepsis (0.13 nM/s [0.06, 0.19 nM/s]). We also evaluated CSRbase as a percentage of CSRmax. As shown in Figure 5B, CSRbase/CSRmax ratio was 54.4% [32.7%, 69.5%] in SS, which was significantly higher (p < 0.001) than sepsis subjects (27.3% [15.2%, 34.0%]).

Figure 5.

Figure 5.

Open in a new tab

A, Box plots showing distribution of log-transformed baseline cortisol secretion rate (CSRbase) at visit 1 for septic shock (n = 45) and sepsis (n = 25). B, Box plots for CSRbase/maximal cortisol secretion rate (CSRmax) for septic shock (SS) and sepsis subjects (visit 1). CSRbase is the estimated cortisol secretion rate required to achieve initial total cortisol concentration under steady-state conditions and using individually measured corticosteroid-binding globulin and albumin concentrations as well as individually estimated free cortisol half-life.

Correlations

Among SS subjects, CSRmax was correlated with total cortisol measured both 30 minutes (Spearman ρ = 0.52; p < 0.001) and 60 minutes (ρ = 0.43; p = 0.003) after cosyntropin. There was no correlation between CSRmax and either total or free cortisol at baseline.

CSRbase was positively correlated with measured free cortisol at baseline (ρ = 0.71; p < 0.001) in SS. In addition, CSRbase was inversely correlated with free cortisol half-life in SS (ρ = –0.59; p < 0.001) and sepsis (ρ = –0.43; p = 0.03). These correlations were maintained after adjusting for free cortisol concentration (partial ρ = –0.85; p < 0.001 for SS and partial ρ = –0.53; p = 0.03 for sepsis), indicating that the inverse relationship between CSRbase and free cortisol half-life was not driven by differences in free cortisol.

CSRbase was positively correlated with CSRmax in SS (ρ = 0.77; p < 0.001) and in sepsis (ρ = 0.62; p = 0.001), suggesting that common factor(s) may mediate the increase in CSRbase and in CSRmax. Among SS subjects, adrenocorticotropin concentration was positively correlated with CSRbase/CSRmax ratio (ρ = 0.54; p < 0.001) and with APACHE scores (ρ = 0.38; p = 0.03). The positive correlation between adrenocorticotropin concentration and CSRbase/CSRmax ratio was not driven by differences in APACHE score because partial correlation adjusting for APACHE scores indicates no change in the relationship between adrenocorticotropin and CSRbase/CSRmax ratio (partial ρ = 0.47; p = 0.006).

Sensitivity Analysis

To assess the relationship between error in laboratory assay measurement and computed parameters, we performed sensitivity analysis. As shown in Appendix 2 (Supplemental Digital Content 3, http://links.lww.com/CCM/B110), ± 1 sd error in laboratory measurements for all three cortisol concentrations, as well as CBG and albumin concentrations, results in sds of 25.2–36.7 for percentage error in CSRmax (Appendix 2, Supplemental Digital Content 3, http://links.lww.com/CCM/B110). These results indicate that measurement error does substantially propagate through computed parameters. However, because the computed parameters reported herein already include this variability, the betweengroup differences observed in this study remain valid.

Subgroup Analysis of SS Subjects Having Δ-Cortisol Less Than or Equal To or Greater Than 248 nM

A cut score of total cortisol increment (Δ-cortisol) of 248 nM after cosyntropin stimulation has been useful in predicting response to corticosteroids in some but not in all studies (1, 4, 6–9). In our group of SS subjects, CSRbase (p = 0.05) and CSRbase/CSRmax (p < 0.001) were significantly higher among subjects having Δ-cortisol less than or equal to 248 nM (n = 12) when compared with those with Δ-cortisol greater than 248 nM (n = 33) (Supplemental Fig. 2, Supplemental Digital Content 4, http://links.lww.com/CCM/B111; Supplemental Fig. 3, Supplemental Digital Content 5, http://links.lww.com/CCM/B112). There was no difference in CSRmax between subgroups defined by Δ-cortisol less than or equal to or greater than 248 nM, whereas a trend (p = 0.07) for shorter free cortisol half-life was observed for SS subjects with Δ-cortisol less than or equal to 248 nM.

DISCUSSION

The stable isotope dilution method using 2H-labeled cortisol infusion is considered a gold standard methodology for estimation of CPR. However, procedures such as 2H-labeled cortisol infusion are strictly limited to research applications. By contrast, the numerical modeling methods used to obtain parameters of free cortisol secretion and half-life in the present study are well suited to clinical translation.

We found general consistency between comparable measures obtained by stable isotope dilution methodology in the report of Boonen et al (2) and those obtained by numerical modeling in the present study. For example, Boonen et al (2) reported that the rate of cortisol clearance was decreased in critical illness, including SIRS, to a level of approximately 50% of control subjects. This result is similar to the approximately 60% reduction in the free cortisol elimination rate we observed in sepsis when compared with controls.

Boonen et al (2) also reported that CPRbase among critically ill subjects without SIRS (1.8 ± 1.1 mg/hr) was similar to controls (1.5 ± 1.5 mg/hr), whereas CPRbase among SIRS subjects was significantly increased (3.4 ± 1.1 mg/hr) relative to controls (2). To facilitate comparison of results obtained using numerical modeling and stable isotope dilution methods, we applied nominal estimates of cortisol volume of distribution to convert numerical estimates of CSR (The term CSR is used for numerical modeling methods and solved in the concentration space; CSR is typically expressed in units of mass/volume/ time (e.g., μmol/L/s). As reported by Keenan et al [23], 24-hr CPR can be estimated from CSR by applying nominal mean cortisol volume of distribution, which has been reported to be approximately 8.2 L/m2 [25–27].) into units of CPR, as previously described (11, 23). Adjusting for cortisol volume of distribution, the resulting CPRbase in our study was of 2.3 mg/hr [1.09, 3.5 mg/hr] for sepsis. This result is similar to the CPRbase of 3.4 ± 1.1 mg/hr obtained by Boonen et al (2) for a comparable clinical group (critically ill subjects with SIRS).

Our hypothesis is that CSRmax would be subnormal in a significant subset of sepsis and SS subjects was not supported by the present analysis. To the contrary, we observed that on average CSRmax was significantly increased in SS. The comparison with control subjects in this study was complicated by the relatively small number of controls (n = 10) and significantly younger age of the control group relative to sepsis and SS. In a larger cohort of control subjects (n = 21) using a more intensive cortisol sampling regimen (11), CSRmax and free cortisol half-life were normally distributed with a reference range (0.44 ± 0.13 nM/s for CSRmax and 2.2 ± 1.1 min for free cortisol half-life) that was similar to results obtained for controls in the present study (CSRmax, 0.49 nM/s [0.42, 0.62 nM/s] and free cortisol half-life, 2.0 min [1.2, 2.6 min]). Adjusting for age using data of Keenan et al (23) did not alter our finding that CSRmax was significantly greater in SS when compared with control and sepsis groups. Our finding that CSRmax was similar between control and sepsis groups is also consistent with the report of Boonen et al (3), who found no difference in adrenocorticotropin-cortisol dose–response relationships between controls and critically ill subjects including SIRS.

The observed increase in CSRmax in SS is consistent with the previous report of Ulrich-Lai et al (28), demonstrating increased CSRmax and adrenal weight in a rodent model of chronic stress. In the present study, we found that CSRmax was already elevated within 24 hours of the time of diagnosis of SS. This observation suggests that in a significant proportion of patients with SS, the increase in CSRmax occurs quite early in the clinical course. Jung et al (8) found that total adrenal gland volume (TAGV) assessed by CT scan also increases early in the course of SS. These findings, in conjunction with the association between adrenal weight and CSRmax observed in an animal model of chronic stress (28), raise the possibility that increased TAGV and CSRmax in clinical SS are correlated. TAGV was not assessed in the present study and so the relationship between TAGV and CSR was not evaluable in our dataset. Additional studies involving more rigorously defined comparison groups would be needed to determine which factors within clinically defined groups contribute to the observed differences in CSRmax and related parameters.

The variance of CSRmax in the SS group was also significantly greater than that of control or sepsis groups. Accordingly, there was a very broad range of CSRmax among SS subjects, ranging from the low end of the normal range to levels that far exceed the upper limit of control and or sepsis subjects. The greater variance in SS was not because of greater degree of propagation of laboratory error in CSRmax estimates because the magnitude of laboratory error propagation was similar between groups (Appendix 2, Supplemental Digital Content 3, http://links.lww.com/CCM/B110). Therefore, the greater variability in CSRmax among the SS subjects observed in this study likely reflects biological variation. The greater variance among SS subjects was not related to clinical differences, including age, gender, APACHE score, and 28-day mortality. However, other clinical information, such as use of vasopressors, days of intubation, measures of hepatic and renal function, concentration of inflammatory markers such as interleukin (IL)-6 and IL-10, and corticosteroid administration after visit 1 were unavailable for analysis. Therefore, we cannot exclude the potential contribution of these factors to the increased variance observed among SS subjects.

Boonen et al (2) reported that SIRS is associated with an approximately two-fold increase in CPRbase when compared with controls. Our data indicate that SS is associated with a far greater increase in CSRbase when compared with sepsis (Fig. 5A). The increase in CSRbase and CSRbase/CSRmax parallels the clinical extremity of these clinical conditions in our study populations (28-day mortality of 0% and 41% in sepsis and SS, respectively). In a significant proportion of SS subjects, CSRbase exceeded the upper limit of maximal cortisol secretion rate for healthy controls. This observation highlights the potential importance of increased CSRmax as a compensatory or adaptive mechanism.

We observed a decreased rate of free cortisol elimination, which corresponds to a longer free cortisol half-life, in both sepsis and SS. The free cortisol half-life parameter is obtained through numerical modeling and is inversely proportional to the free cortisol elimination rate constant (see α in Supplemental Fig. 1, Supplemental Digital Content 2, http://links.lww.com/CCM/B109). Free cortisol half-life is to be distinguished from more commonly reported estimates of total cortisol half-life (2, 24, 25, 29). In the numerical model used to obtain free cortisol half-life (11, 23, 25), only the free cortisol compartment is subject to elimination; CBG- and albumin-bound cortisol compartments are considered to be in dynamic, reversible equilibrium with the free cortisol compartment (Supplemental Fig. 1, Supplemental Digital Content 2, http://links.lww.com/CCM/B109). Using free rather than total cortisol for half-life may be more accurate in conditions such as SS or critical illness, since total cortisol half-life is biased by variations in CBG, albumin, and total cortisol concentrations (30), whereas free cortisol half-life is independent of these variables (29). The mechanism for reduced rate of free cortisol elimination in sepsis and SS observed in the present study is likely related to the decreased expression and activity of 11β-hydroxysteroid dehydrogenase type 2 and A-ring reductases previously reported in critical illness and SS (2, 31).

Our finding of longer free cortisol half-life in sepsis and SS is consistent with previous reports of prolonged total cortisol half-life in SS (32–34) and with the report of reduced total cortisol clearance rates in critical illness obtained by stable isotope dilution (2). Our results confirm and extend these findings by showing that free cortisol half-life is prolonged to a similar degree in sepsis and SS. In addition, our findings indicate that the decreased rate of free cortisol elimination persists through the duration of inpatient hospitalization for both sepsis and SS subjects.

Our sensitivity analysis indicates that laboratory error in measured concentrations of cortisol, albumin, and CBG propagates through the computation of numerical modeling parameters of free cortisol appearance and elimination. The level of accuracy of parameter estimates achieved in the present study was useful for analysis of general distributions and groups comparisons. A higher level of accuracy and reproducibility would be desirable for clinical translational applications or assessment of cortisol appearance and elimination rates in individual patients. Such refinements might include more frequent cortisol sampling, more precise cortisol assay, and the use of a greater than usual number of replicates for cortisol concentration measurements.

CSRmax and related parameters provide physiological information that is qualitatively distinct from more traditional concentration-based methods. Therefore, it is at least theoretically possible that determination of cortisol appearance and elimination rates could be usefully applied to clinical applications related to SS. For example, availability of CSRmax and CSRbase/CSRmax estimates could be useful in the laboratory diagnosis of CIRCI. However, as illustrated by results and power considerations of the CORTICUS trial (6), the beneficial and deleterious treatment effects of corticosteroid replacement are relatively small. Accordingly, the question of whether kinetic cortisol parameters are useful in the diagnosis of CIRCI would need to be addressed in future studies that are appropriately powered and use a randomized, placebo-controlled study design. In addition, pretreatment knowledge of free cortisol half-life might be useful in achieving more predictable concentrations of free cortisol during corticosteroid replacement. Given the large variation in total and free cortisol concentrations that is observed during usual clinical care dosing of hydrocortisone (34), it is plausible that greater uniformity in free cortisol concentrations achieved during corticosteroid administration could be useful in preventing undesirable side effects related to exogenous hydrocortisone administration and, in future studies, aid in the experimental evaluation of optimal free cortisol concentration for treatment of CIRCI. Finally, assessment of kinetic parameters of cortisol appearance and elimination might be useful in identification of patients with frank AI. Both primary and secondary AI are associated with durable abnormalities in adrenocortical function and is to be distinguished from CIRCI (1, 12, 35, 36). Although frank AI are uncommon in the setting of SS, it is of clinical importance because the clinical presentation of acute AI can mimic SS (35), while the beneficial treatment effect of corticosteroid replacement in frank AI is far greater than in SS (12, 35, 36). Our finding that CSRmax is either normal or elevated in SS suggests the possibility that patients with frank AI, for example, because of acute adrenal hemorrhage or chronic adrenocorticotropin deficiency, could be identified by the finding of subnormal CSRmax.

CONCLUSIONS

Numerical modeling is useful in obtaining kinetic parameters of cortisol appearance and elimination in sepsis and SS. Results obtained by numerical modeling are consistent with comparable results obtained using the gold standard stable isotope dilution method. CSRmax was similar in control and sepsis subjects. Relative to control and sepsis groups, SS was associated with a reversible increase in CSRmax and increased variance of CSRmax. Free cortisol half-life was increased to a similar degree in both sepsis and SS. We conclude that in spite of a reduce rate of free cortisol elimination in both sepsis and SS groups, subnormal CSRmax is not a characteristic finding in either of these clinically defined groups.

Supplementary Material

Supplemental Fig. 2, Supplemental Digital Content 4, Supplemental Fig. 3, Supplemental Digital Content 5
Appendix 2 (Supplemental Digital Content 3
Supplemental Digital Content (Figure Legends)
SUPPLEMENTAL DIGITAL CONTENT (Methods Appendix)
Appendix 1, Supplemental Digital Content 1,Supplemental Fig. 1
Appendix 1 Equation 4

Acknowledgments

Supported, in part, by University of New Mexico Clinical and Translational Science Center Department of Health and Human Services/National Institutes of Health (NIH)/National Center for Research Resources 1UL1RR03197701 and by VA Research Service and, in part, by the National Center for Research Resources and the National Center for Advancing Translational Sciences of the National Institutes of Health through grant number 8UL1TR000041. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Dr. Dorin received support for article research from the National Institutes of Health (NIH) and disclosed government work. His institution received grant support from the NIH and VA Research Service. Dr. Qualls disclosed government work. The remaining authors have disclosed that they do not have any potential conflicts of interest.

Footnotes

This work was performed at New Mexico VA Healthcare System, University of New Mexico School of Medicine, Royal Adelaide Hospital, and Florida International University.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (http://journals.lww.com/ccmjournal).

*

See also p. 702.

REFERENCES

  • 1.Marik PE, Pastores SM, Annane D, et al. ; American College of Critical Care Medicine: Recommendations for the diagnosis and management of corticosteroid insufficiency in critically ill adult patients: Consensus statements from an international task force by the American College of Critical Care Medicine. Crit Care Med 2008; 36:1937–1949 [DOI] [PubMed] [Google Scholar]
  • 2.Boonen E, Vervenne H, Meersseman P, et al. : Reduced cortisol metabolism during critical illness. N Engl J Med 2013; 368:1477–1488 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Boonen E, Meersseman P, Vervenne H, et al. : Reduced nocturnal ACTH-driven cortisol secretion during critical illness. Am J Physiol Endocrinol Metab 2014; 306:E883–E892 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 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. JAMA 2000; 283:1038–1045 [DOI] [PubMed] [Google Scholar]
  • 5.Annane D, Sébille V, Charpentier C, et al. : Effect of treatment with low doses of hydrocortisone and fludrocortisone on mortality in patients with septic shock. JAMA 2002; 288:862–871 [DOI] [PubMed] [Google Scholar]
  • 6.Sprung CL, Annane D, Keh D, et al. ; CORTICUS Study Group: Hydrocortisone therapy for patients with septic shock. N Engl J Med 2008; 358:111–124 [DOI] [PubMed] [Google Scholar]
  • 7.Annane D, Bellissant E: Prognostic value of cortisol response in septic shock. JAMA 2000; 284:308–309 [PubMed] [Google Scholar]
  • 8.Jung B, Nougaret S, Chanques G, et al. : The absence of adrenal gland enlargement during septic shock predicts mortality: A computed tomography study of 239 patients. Anesthesiology 2011; 115:334–343 [DOI] [PubMed] [Google Scholar]
  • 9.Moreno R, Sprung CL, Annane D, et al. : Time course of organ failure in patients with septic shock treated with hydrocortisone: Results of the Corticus study. Intensive Care Med 2011; 37:1765–1772 [DOI] [PubMed] [Google Scholar]
  • 10.Marik PE: Critical illness-related corticosteroid insufficiency. Chest 2009; 135:181–193 [DOI] [PubMed] [Google Scholar]
  • 11.Dorin RI, Qiao Z, Qualls CR, et al. : Estimation of maximal cortisol secretion rate in healthy humans. J Clin Endocrinol Metab 2012; 97:1285–1293 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Dorin RI, Qualls CR, Crapo LM: Diagnosis of adrenal insufficiency. Ann Intern Med 2003; 139:194–204 [DOI] [PubMed] [Google Scholar]
  • 13.Landon J, Wynn V, James VH, et al. : Adrenal response to infused corticotropin in subjects receiving glucocorticoids. J Clin Endocrinol Metab 1965; 25:602–611 [DOI] [PubMed] [Google Scholar]
  • 14.Thorn GW, Goetz FC, Streeten DH, et al. : Use of the intravenous ACTH test in clinical practice. J Clin Endocrinol Metab 1953; 13:604–613. [DOI] [PubMed] [Google Scholar]
  • 15.Ho JT, Al-Musalhi H, Chapman MJ, et al. : Septic shock and sepsis: A comparison of total and free plasma cortisol levels. J Clin Endocrinol Metab 2006; 91:105–114 [DOI] [PubMed] [Google Scholar]
  • 16.Mendel CM, Miller MB, Siiteri PK, et al. : Rates of dissociation of steroid and thyroid hormones from human serum albumin. J Steroid Biochem Mol Biol 1990; 37:245–250 [DOI] [PubMed] [Google Scholar]
  • 17.Stroupe SD, Harding GB, Forsthoefel MW, et al. : Kinetic and equilibrium studies on steroid interaction with human corticosteroid-binding globulin. Biochemistry 1978; 17:177–182 [DOI] [PubMed] [Google Scholar]
  • 18.Westphal U. Steroid-Protein Interactions. New York, Springer-Verlag, 1971 [Google Scholar]
  • 19.Dorin RI, Pai HK, Ho JT, et al. : Validation of a simple method of estimating plasma free cortisol: Role of cortisol binding to albumin. Clin Biochem 2009; 42:64–71 [DOI] [PubMed] [Google Scholar]
  • 20.Arafah BM: Hypothalamic pituitary adrenal function during critical illness: Limitations of current assessment methods. J Clin Endocrinol Metab 2006; 91:3725–3745 [DOI] [PubMed] [Google Scholar]
  • 21.Bornstein SR, Licinio J, Tauchnitz R, et al. : Plasma leptin levels are increased in survivors of acute sepsis: Associated loss of diurnal rhythm, in cortisol and leptin secretion. J Clin Endocrinol Metab 1998; 83:280–283 [DOI] [PubMed] [Google Scholar]
  • 22.Voerman HJ, Strack van Schijndel RJ, Groeneveld AB, et al. : Pulsatile hormone secretion during severe sepsis: Accuracy of different blood sampling regimens. Metabolism 1992; 41:934–940 [DOI] [PubMed] [Google Scholar]
  • 23.Keenan DM, Roelfsema F, Carroll BJ, et al. : Sex defines the age dependence of endogenous ACTH-cortisol dose responsiveness. Am J Physiol Regul Integr Comp Physiol 2009; 297:R515–R523 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Purnell JQ, Brandon DD, Isabelle LM, et al. : Association of 24-hour cortisol production rates, cortisol-binding globulin, and plasma-free cortisol levels with body composition, leptin levels, and aging in adult men and women. J Clin Endocrinol Metab 2004; 89:281–287 [DOI] [PubMed] [Google Scholar]
  • 25.Keenan DM, Roelfsema F, Veldhuis JD: Endogenous ACTH concentration-dependent drive of pulsatile cortisol secretion in the human. Am J Physiol Endocrinol Metab 2004; 287:E652–E661 [DOI] [PubMed] [Google Scholar]
  • 26.Kraan GP, Dullaart RP, Pratt JJ, et al. : Kinetics of intravenously dosed cortisol in four men. Consequences for calculation of the plasma cortisol production rate. J Steroid Biochem Mol Biol 1997; 63:139–146 [DOI] [PubMed] [Google Scholar]
  • 27.Kraan GP, Dullaart RP, Pratt JJ, et al. : The daily cortisol production reinvestigated in healthy men. The serum and urinary cortisol production rates are not significantly different. J Clin Endocrinol Metab 1998; 83:1247–1252 [DOI] [PubMed] [Google Scholar]
  • 28.Ulrich-Lai YM, Figueiredo HF, Ostrander MM, et al. : Chronic stress induces adrenal hyperplasia and hypertrophy in a subregion-specific manner. Am J Physiol Endocrinol Metab 2006; 291:E965–E973 [DOI] [PubMed] [Google Scholar]
  • 29.Picard-Hagen N, Gayrard V, Alvinerie M, et al. : A nonlabeled method to evaluate cortisol production rate by modeling plasma CBG-free cortisol disposition. Am J Physiol Endocrinol Metab 2001; 281:E946–E956 [DOI] [PubMed] [Google Scholar]
  • 30.Bright GM: Corticosteroid-binding globulin influences kinetic parameters of plasma cortisol transport and clearance. J Clin Endocrinol Metab 1995; 80:770–775 [DOI] [PubMed] [Google Scholar]
  • 31.Venkatesh B, Cohen J, Hickman I, et al. : Evidence of altered cortisol metabolism in critically ill patients: A prospective study. Intensive Care Med 2007; 33:1746–1753 [DOI] [PubMed] [Google Scholar]
  • 32.Melby JC, Spink WW: Comparative studies on adrenal cortical function and cortisol metabolism in healthy adults and in patients with shock due to infection. J Clin Invest 1958; 37:1791–1798 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Melby JC, Egdhal RH, Spink WW: Secretion and metabolism of cortisol after injection of endotoxin. J Lab Clin Med 1960;56:50–62 [Google Scholar]
  • 34.Oppert M, Reinicke A, Gräf KJ, et al. : Plasma cortisol levels before and during “low-dose” hydrocortisone therapy and their relationship to hemodynamic improvement in patients with septic shock. Intensive Care Med 2000; 26:1747–1755 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Dorin RI, Kearns PJ: High output circulatory failure in acute adrenal insufficiency. Crit Care Med 1988; 16:296–297 [DOI] [PubMed] [Google Scholar]
  • 36.Werbel SS, Ober KP: Acute adrenal insufficiency. Endocrinol Metab Clin North Am 1993; 22:303–328 [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplemental Fig. 2, Supplemental Digital Content 4, Supplemental Fig. 3, Supplemental Digital Content 5
NIHMS1694047-supplement-Supplemental_Fig__2__Supplemental_Digital_Content_4__Supplemental_Fig__3__Supplemental_Digital_Content_5.pdf (9.1KB, pdf)
Appendix 2 (Supplemental Digital Content 3
NIHMS1694047-supplement-Appendix_2__Supplemental_Digital_Content_3.docx (8KB, docx)
Supplemental Digital Content (Figure Legends)
NIHMS1694047-supplement-Supplemental_Digital_Content__Figure_Legends_.pdf (274.5KB, pdf)
SUPPLEMENTAL DIGITAL CONTENT (Methods Appendix)
NIHMS1694047-supplement-SUPPLEMENTAL_DIGITAL_CONTENT__Methods_Appendix_.docx (20.8KB, docx)
Appendix 1, Supplemental Digital Content 1,Supplemental Fig. 1
NIHMS1694047-supplement-Appendix_1__Supplemental_Digital_Content_1_Supplemental_Fig__1.pdf (204.6KB, pdf)
Appendix 1 Equation 4
NIHMS1694047-supplement-Appendix_1_Equation_4.docx (11.4KB, docx)

RESOURCES