| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
REVIEW |
Division of Clinical and Molecular Endocrinology, Case Western Reserve University and University Hospitals/Case Medical Center, Cleveland, Ohio 44106
Address all correspondence and requests for reprints to: Baha M. Arafah, M.D., Division of Clinical and Molecular Endocrinology, University Hospitals/Case Medical Center, 11100 Euclid Avenue, Cleveland, Ohio 44106. E-mail: baha.arafah{at}case.edu.
 |
Abstract |
|---|
Objectives: The primary objective was to review some of the modulating factors and limitations of currently used methods of assessing HPA function during critical illness and provide alternative approaches in that setting.
Design: This was a critical review of relevant data from the literature with inclusion of previously published as well as unpublished observations by the author. Data on HPA function during three different forms of critical illnesses were reviewed: experimental endotoxemia in healthy volunteers, the response to major surgical procedures in patients with normal HPA, and the spontaneous acute to subacute critical illnesses observed in patients treated in intensive care units.
Setting: The study was conducted at an academic medical center.
Patients/Participants: Participants were critically ill subjects.
Intervention: There was no intervention.
Main Outcome Measure: The main measure was to provide data on the superiority of measuring serum free cortisol during critical illness as contrasted to those of total cortisol measurements.
Results: Serum free cortisol measurement is the most reliable method to assess adrenal function in critically ill, hypoproteinemic patients. A random serum free cortisol is expected to be 1.8 µg/dl or more in most critically ill patients, irrespective of their serum binding proteins. Because the free cortisol assay is not currently available for routine clinical use, alternative approaches to estimate serum free cortisol can be used. These include calculated free cortisol (Coolens’ method) and determining the free cortisol index (ratio of serum cortisol to transcortin concentrations). Preliminary data suggest that salivary cortisol measurements might be another alternative approach to estimating the free cortisol in the circulation. When serum binding proteins (albumin, transcortin) are near normal, measurements of total serum cortisol continue to provide reliable assessment of adrenal function in critically ill patients, in whom a random serum total cortisol would be expected to be 15 µg/dl or more in most patients. In hypoproteinemic critically ill subjects, a random serum total cortisol level is expected to be 9.5 µg/dl or more in most patients. Data on Cosyntropin-stimulated serum total and free cortisol levels should be interpreted with the understanding that the responses in critically ill subjects are higher than those of healthy ambulatory volunteers. The Cosyntropin-induced increment in serum total cortisol should not be used as a criterion for defining adrenal function, especially in critically ill patients.
Conclusions: The routine use of glucocorticoids during critical illness is not justified except in patients in whom adrenal insufficiency was properly diagnosed or others who are hypotensive, septic, and unresponsive to standard therapy. When glucocorticoids are used, hydrocortisone should be the drug of choice and should be given at the lowest dose and for the shortest duration possible. The hydrocortisone dose (50 mg every 6 h) that is mistakenly labeled as low-dose hydrocortisone leads to excessive elevation in serum cortisol to values severalfold greater than those achieved in patients with documented normal adrenal function. The latter data should call into question the current practice of using such doses of hydrocortisone even in the adrenally insufficient subjects.
|
Introduction |
|---|
Regulations of the components of the stress response system are interdependent and interwined (1, 2, 3, 4, 5). For example, CRH stimulates secretion of norepinephrine through specific receptors in the hypothalamus, whereas norepinephrine stimulates CRH secretion through 
1 noradrenergic receptors (1, 2, 3, 4, 5). Both components of the central stress response system are stimulated by cholinergic and serotonergic neurotransmitters and are inhibited by 
-aminobutyric acid-benzodiazepine and proopiomelanocortin peptides (1, 2, 3, 4, 5). Other areas of the brain differentially activate a subset of vagal and sacral parasympathetic efferents that mediates the gut response to stress (1, 2, 3, 4). Detailed reviews of the components of the autonomic nervous system and their regulation have been discussed extensively in the literature (1, 2, 3, 4, 5).
The essential role of the adrenal for survival was first noted by Addison in 1855. Although epinephrine was the first adrenal hormone to be isolated, its failure to prevent death in adrenalectomized animals suggested that the adrenal cortex was essential for survival (7). Subsequent studies over the years revealed the adrenal cortex to include three zones or regions whereby each region is not only regulated differently but also has a specific set of enzymes and produces a specific class of steroids. Whereas the secretion of mineralocorticoids by the zona glomerulosa is regulated by the renin-angiotensin system, the synthesis and secretion of glucocorticoids by the zona fasciculate and adrenal androgens by zona reticularis are controlled by the HPA system. Destructive or infiltrative disease entities affecting the adrenal gland tend to cause partial or complete loss of the three classes of steroids (mineralocorticoids, androgens, and glucocorticoids) secreted by the three zones of the adrenal and result in what is described as primary adrenal insufficiency. In contrast, partial or complete loss of ACTH secretion causes decreased glucocorticoid and adrenal androgen secretion and is referred to as central adrenal insufficiency.
This review will focus on HPA function in critically ill patients. The review will not address the causes, diagnosis, or treatment of primary or central adrenal insufficiency because detailed studies and reviews on these topics have been published (8, 9, 10, 11, 12). This review will not examine adjustments in glucocorticoid therapy during critical illness in patients previously known to have primary or secondary adrenal insufficiency because these issues have been extensively reviewed (9, 13). Instead, the main focus of this review is to address important limitations in studies examining HPA function during critical illnesses in patients who were not previously known to have adrenal dysfunction.
|
Overview |
|---|
Numerous studies have documented activation of the HPA axis during acute and chronic stressful events such as in patients undergoing surgery (16, 18, 19, 20, 21, 22), those with sepsis (23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37), trauma (31), burns (38), and others with different critical illnesses (39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49). In general, the degree of activation of the axis is proportionate to the stress. Activation of the axis as frequently determined by measurements of random serum total cortisol concentrations indicates that the latter are uniformly elevated in the critically ill. Although the degree of elevation in serum cortisol concentration may not correlate linearly with the illness severity, some studies demonstrated that patients with the highest cortisol levels have the highest mortality as well (35). In some studies, activation of the HPA axis in the critically ill has been demonstrated by measuring Cosyntropin-stimulated serum cortisol levels. As will be discussed later, many of the methods and measures used to determine glucocorticoid secretion have significant limitations that can often lead to data misinterpretation.
Over the past few years, newer data on glucocorticoid secretion and/or therapy during critical illness were published. Most of the published data focused on a subgroup of patients with severe sepsis and/or septic shock. Despite, and perhaps because of, the recent data, many controversies in this field became evident. This article will review the physiology of the normal response to critical illnesses and examine potential limitations of published data on assessment of adrenal function and others using glucocorticoids in treating patients with critical illnesses. The article will also discuss some of the confounding factors limiting these studies, including the methods used to define normal secretion. Examples of these limitations are cited, and approaches to address these confounding factors are discussed. The newly introduced yet poorly defined and characterized concept of relative or functional adrenal insufficiency is a very controversial topic that will also be discussed in this review.
|
I. Normal Physiology of HPA Function |
|---|
Vasopressin is also a known modulator of ACTH secretion. Because vasopressin stimulates the release of ACTH in healthy subjects, it is used as a diagnostic test for conditions associated with ACTH excess or deficiency. However, the stimulatory effects of vasopressin on ACTH secretion require the presence of CRH.
|
II. Alterations in HPA Function during Critical Illness |
|---|
have long been recognized as important modulators of the HPA function during critical illness (5, 20). During an inflammatory process, these cytokines are capable of stimulating and maintaining glucocorticoid production to high levels (5, 50, 51, 52, 53, 54). IL-6 receptors are present on pituitary corticotrophs as well as adrenal cortical cells (51, 52). It is believed that cytokines released from the site of injury or after exposure to endotoxin activate the HPA by stimulating the classical pathway of CRH and ACTH secretion (5, 50, 51, 52, 53). These cytokines act synergistically to augment ACTH secretion beyond that achieved by CRH alone (5, 50).
The HPA axis is highly activated during stressful events as evidenced by elevated plasma ACTH levels, increased cortisol secretion, and elevated serum total and free cortisol levels (37, 45). In addition to increased adrenal production or secretion of glucocorticoids during critical illness, impaired glucocorticoid clearance can contribute to the greatly increased serum cortisol concentrations (55). This would be especially likely in patients with impaired hepatocellular function, hepatic blood flow, or renal or thyroidal function (55). The increased cortisol secretion is best appreciated by the marked elevation in the free fraction of the hormone (37, 45). Cortisol secretion during critical illnesses is not only excessive, reaching levels greater than those achieved in patients with Cushing’s syndrome, but also less suppressible by exogenous glucocorticoid administration such as dexamethasone (56, 57). An example of the poor suppressibility of the HPA axis during stress is illustrated in Fig. 1
. In that example (Fig. 1), plasma ACTH and serum cortisol levels were serially determined in patients with normal HPA function after surgical removal of brain tumors. The levels of both hormones (ACTH, cortisol) were not suppressed for more than 2 d despite the administration of large (24 mg/d) doses of dexamethasone. Furthermore, ACTH and cortisol responsiveness to exogenous CRH is enhanced during critical illnesses (57). Even though ACTH continues to be the dominant factor stimulating cortisol secretion by the adrenal cortex throughout the critical illness, other factors play a significant modulating influence on the axis. Such factors include arginine vasopressin (especially in volume-contracted subjects), endothelin, atrial natriuretic factor (ANF) (58), and a variety of cytokines such as IL-6 (50, 51, 52, 53, 54). Macrophage-migration inhibitory factor (MIF) is another modulator of the HPA function, especially during a severe inflammatory process such as septic shock (59, 60). Although the exact role of MIF during critical illness is not well defined, this factor appears to have both proinflammatory and antiinflammatory effects (59, 60, 61) and is considered to have a role in the homeostatic and physiological actions of glucocorticoids in vivo (61).
|
Recent studies on the actions of glucocorticoids suggest that the effects of these steroid molecules are a continuum from permissive to suppressive effects that are observed over the range of concentrations achieved in vivo. Whereas physiological concentrations are associated with permissive effects, the high concentrations observed during serious illnesses are associated with suppressive or antiinflammatory effects (67, 68). The latter antiinflammatory effects of glucocorticoids are what is most known and appreciated about these steroid molecules. Much less is appreciated about the permissive and, at times, proinflammatory effects of glucocorticoids (67, 68). Glucocorticoids can alter their effects on target tissues through modulation of the density and binding affinity (69) and by altering cytokine receptors in glucocorticoid-dependent cells (70). Glucocorticoids also modulate the production of MIF by macrophages (59, 60, 61) and also by altering the hepatic acute phase response and their known effects of cell apoptosis (71, 72).
Resistance to glucocorticoid action, whether it is caused by defects in the glucocorticoid receptor or postreceptor alterations, has been proposed to occur during some critical illnesses, particularly in severe sepsis and septic shock. Inflammatory cytokines, when produced at lower concentrations, appear to stimulate cortisol secretion and enhance its binding to its own receptor (73). However, when cytokine production is excessive, as is the case during an overwhelming inflammatory response, it leads to decreased numbers as well as binding affinity in glucocorticoid receptors (74) and also other postreceptor alterations. Both of these events can lead to glucocorticoid resistance, particularly at the tissue site of excessive cytokine production (74). It is not clear at this point whether MIF secretion contributes to the state of relative glucocorticoid resistance.
|
III. Short-Term Stresses vs. Protracted Critical Illness |
|---|
In general, the HPA displays a biphasic pattern during the course of a critical illness (58, 75, 76). During the initial phase of an illness, such as surgery, uncomplicated trauma, burn, infection, or sepsis, the HPA axis is primarily activated through increased CRH secretion and cytokine production. Biochemically, the initial phase is characterized by elevated plasma ACTH and cortisol concentrations (58, 76). The hypercortisolism in this setting provides energy and protects the body and is reflected by increased gluconeogenesis, maintenance of intravascular volume, and inhibition of the acute inflammatory reaction. In contrast, studies in protracted critical illness showed a decrease in plasma ACTH concentrations despite persistence of the state of hypercortisolism. These features suggest that cortisol secretion is being regulated and stimulated by alternative pathways other than the classical hypothalamic CRH. As discussed earlier, such factors include ANF, endothelin, substance P, and a variety of cytokines. The persistent hypercortisolism observed in protracted critical illness serves to provide similar benefits related to providing energy, maintaining volume, and minimizing inflammation. However, the persistence of hypercortisolism is also likely to contribute to some of the longer-term complications observed with protracted critical illness such as hyperglycemia, myopathy, poor wound healing, and psychiatric alterations.
As stated earlier, this review will discuss some of the often overlooked, confounding factors and limitations in the biochemical assessment of adrenal function. The review will also address newer approaches to avoid and minimize these limitations. Relevant published data on three different forms or examples of critical illnesses will be reviewed. The first form of stress to be reviewed will be that of normal subjects during experimental endotoxemia. The latter is considered an experimental model for an acute inflammatory process. The second example of a short-term, stressful event to be reviewed is that of patients undergoing different surgical procedures. These two examples represent instances of short-term critical illness where the duration of stress is hours to days. In one of these stresses (experimental endotoxemia), the stress induces a dominant inflammatory response. The last and most extensively studied form of stress is that of patients with a variety of medical and surgical illnesses who are studied in intensive care units (ICUs). Although patients with a variety of diagnoses are included in such publications, many of them have focused on patients with a specific illness such as septic shock. These patients have been studied at different times after the onset of their critical illnesses, and, therefore, glucocorticoid secretion could be regulated differently. Despite these and many other limitations, reviewing such data would help improve our understanding of these events.
|
IV. Three Examples of Stressful Events |
|---|
Experimental endotoxemia in humans is a well-characterized model of acute inflammation (77, 78, 79, 80, 81). The iv administration of Gram-negative bacterial lipopolysaccharide (LPS) endotoxin results in an acute inflammatory process that manifests as fever, tachycardia, leukocytosis, and immune cell activation. This is followed by cytokine secretion (TNF, IL-6), the release of catecholamines, and activation of the HPA axis (77, 78, 79, 80, 81). LPS causes an increase in stress and counterregulatory hormones within 1 h of its administration (77, 78, 79, 80, 81). A rise in plasma ACTH, catecholamines, cortisol, and GH have been demonstrated after LPS injection (77, 78, 79, 80, 81). Serum cortisol levels increase within 2 h of LPS injection from an average of 320 nmol/liter (11. 5 µg/dl) to nearly 800 nmol/liter or 29 µg/dl (77, 78, 79, 80, 81). LPS administration increases the release of the antiinflammatory cytokine (IL-10), which may serve a protective role during sepsis (79). When hydrocortisone was infused directly before the LPS injection, the antiinflammatory cytokine, IL-10 levels were greatly increased (79). Based on these studies, it was concluded that stimulation of IL-10 release may contribute to the antiinflammatory properties of glucocorticoids (79). Experimental endotoxemia has been quite a helpful approach in understanding the body’s response to acute inflammation. It is important to point out that despite the established inflammatory response to LPS, experimental endotoxemia is not considered a good model for sepsis or septic shock. Therefore, it is imperative that data obtained in patients during experimental endotoxemia cannot be extrapolated or applied to others with sepsis or septic shock.
B. HPA function during controlled, short-term, stressful events (e.g. surgery)
The need for glucocorticoids during any stressful event such as a surgical procedure is due, in part, to the known effects of these steroids on several components of the host response to the stress of surgery. In this respect, glucocorticoids help support and stimulate cardiovascular response to stress. Additionally, glucocorticoids support many of the components of the inflammatory response to tissue injury occurring during surgery. Glucocorticoids stimulate the release of IL-10, the major antiinflammatory cytokine (82). Such effects of glucocorticoids can be achieved with minimal to moderate increase in secretion. However, higher quantities of cortisol would be necessary when the inflammatory response to tissue injury is more extensive and/or prolonged, as is the case in patients with septic shock.
Several studies investigated the HPA axis during and after minor as well as major surgical procedures. Published data on adrenal function and activity were reported in patients with normal HPA function who had various surgical procedures (18) including abdominal surgery (20), cholecystectomy (19), open heart surgery, coronary artery bypass graft (CABG) surgery (22), and pituitary adenomectomy (21). The studies generally show that the HPA is activated especially after extubation whereby plasma ACTH levels are increased and are associated with elevated serum cortisol concentrations. Thereafter, plasma levels of ACTH decline rapidly to normal levels, whereas serum cortisol concentrations decrease slowly, reaching high normal values approximately 48–72 h after the procedure. A typical example is illustrated in Fig. 2, demonstrating the changes in ACTH and cortisol levels in the perioperative period in patients with normal pituitary-adrenal function who had surgical adenomectomy of non-ACTH-secreting adenomas. This example is quite similar to others published in literature on patients after major surgical procedures (18). In such instances, mean serum cortisol levels obtained 2–4 h after extubation are approximately 40 µg/dl, whereas plasma ACTH concentrations are also elevated at 100–150 ng/liter.
|
It is important to emphasize that such critically ill patients not only have elevated baseline serum cortisol levels but also enhanced responsiveness to Cosyntropin stimulation (22). The latter study demonstrated that patients undergoing CABG surgery have higher baseline and Cosyntropin-stimulated serum cortisol levels in the immediate postoperative period, compared with their own values preoperatively (22). The latter finding reported in patients who had CABG (22) was quite similar to that previously reported in other critically ill patients with various illnesses (45). An important finding in that study was that the Cosyntropin-induced incremental rise in serum cortisol levels cannot be used to define normality of adrenal function (22). The study investigators noted, just as was previously observed in normal subjects (84), that nearly 40% of these stressed patients who also had normal adrenal function had a blunted (<9 µg/dl) Cosyntropin-induced increment in serum total cortisol concentrations. As will be discussed below, the latter response pattern had been advocated by some investigators (32, 36) to define critically ill patients with relative adrenal insufficiency. As will be emphasized subsequently, such a definition should be questioned and not be used to characterize adrenal dysfunction without considering the actual baseline or stimulated serum cortisol levels.
It became apparent many decades ago that patients with adrenal insufficiency may not survive a surgical procedure without being given glucocorticoid supplementation (13). It was conventional wisdom to use stress doses of hydrocortisone in patients with adrenal insufficiency during surgical procedures (13). A study by Udelsman et al. (15) conducted in monkeys that had adrenalectomies challenged the practice of giving large doses of glucocorticoids for minor or moderate surgical procedures. In that study, Udelsman et al. (15) examined the effects of different doses of glucocorticoid replacement on hemodynamic adaptation during surgical procedures (cholecystectomy) in adrenalectomized monkeys. Their study showed that physiological glucocorticoid replacement was necessary and sufficient to tolerate the surgical procedure (15). Furthermore, the study demonstrated that the hemodynamic and metabolic parameters as well as the postoperative survival in the animals given physiological replacement were similar to those noted in animals treated with supraphysiological doses of glucocorticoid (15). Since then there has been a shift in clinical practice in favor of giving lower doses of glucocorticoids according to the complexity and duration of the procedures (13). There is, however, no uniformly acceptable dose for each particular surgical procedure.
C. HPA function during spontaneous, uncontrolled critical illnesses
Published data on the HPA function during critical illnesses include studies in patients with a variety of medical and surgical illnesses. Whereas some studies included heterogeneous groups of patients with several diagnoses (39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49), many others reported on more homogenous groups of patients such as those with septic shock (23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37), trauma (31), and others with major burns (38). One of the most important limitations and difficulties in reviewing such data are that they represent patients with different illnesses of variable duration and who have different nutritional support. Most of the published data detailing the HPA function during critical illness were obtained in patients with severe sepsis and septic shock (23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37). It is primarily in the latter groups of patients that studies have investigated the value of glucocorticoid administration in various forms and dosages on mortality. In reviewing published data on adrenal function during critical illness, one will encounter an amazingly wide range of values for serum cortisol levels. Importantly, patients with severe sepsis and shock appear to be distinctly different from other critically ill subjects in that they have significant inflammation and predictably have a high morbidity and mortality. There are, however, obvious examples of patients with other illnesses (e.g. after CABG, pancreatitis, etc.) who are as critically ill as those in septic shock.
As will be discussed below, the definition of what constitutes normal adrenal response to critical illness continues to be debated. Consequently, published data have used a variety of biochemical criteria to define abnormalities in adrenal function during critical illness. These issues and limitations will be addressed in some detail in the following sections. As detailed in Tables 1 and 2, the incidence/prevalence of adrenal dysfunction (regardless of how it is defined) is much higher in patients with septic shock (40–65%) than in other ICU patients with other diagnoses (0–25%). The latter group includes data on patients who had CABG, those with ruptured abdominal aortic aneurysm, and others with a variety of other illnesses. The differences in prevalence/incidence of adrenal dysfunction, regardless of how it was defined, among different underlying conditions vary significantly and point to different pathophysiology. A very recent study by Ho et al. (37) showed that cortisol levels in patients with septic shock were distinctly different from those seen in others with sepsis only. Serum total and free cortisol concentrations in the former group of patients were higher than those in the latter group (37).
|
|
Similarly, a study by Annane et al. (36) showed that a blunted response to Cosyntropin of less than 9 µg/dl (250 nmol/liter) was associated with a higher chance of not surviving the critical illness (septic shock). A very recent study by Arlt et al. (85) examined the prognostic value of measuring baseline and Cosyntropin-stimulated serum cortisol as well as adrenal androgen concentration in patients with septic shock (85). The latter study showed that among the patients in severe sepsis, those who did not survive had higher baseline total cortisol concentrations than those who did (85). The authors found that baseline DHEA and DHEA-S levels in survivors and nonsurvivors of septic shock were similar (85). Interestingly, however, the same study reported that nonsurvivors of septic shock had a higher cortisol to DHEA molar ratio than those who did survive (85). Thus, the latter ratio could be yet another prognostic marker in this setting (85).
|
V. Modulating Factors and Limitations of Current Methods Assessing HPA Function during Critical Illness |
|---|
). These variables introduce significant limitations to many of the published studies, particularly those with data on patients with protracted critical illnesses, in which data interpretation can be compromised. Whereas some of these factors have been addressed, others are overlooked. The following is a detailed discussion on eight of these confounding factors.
|
Transcortin has a low capacity and high affinity, whereas albumin has a high capacity and low affinity for binding cortisol (86, 87, 88, 89). In humans and at physiological concentrations, transcortin can bind up to 25 µg/dl of the circulating cortisol. As transcortin becomes saturated, a larger proportion of circulating plasma-bound cortisol will be albumin bound. A steady decrease in the percent bound cortisol was noted when albumin concentrations in isolated albumin solutions or plasma were decreased to less than 2.0 gm/dl (87).
The current consensus is that the free rather than the protein-bound fraction of cortisol is responsible for its physiological function (86, 87, 88, 89). Because more than 90% of circulating cortisol in human serum is bound to proteins (transcortin and albumin), we postulated that alterations in binding proteins would affect measured serum cortisol levels (45) and thus the interpretation of tests assessing adrenal function. The importance of the fall in serum transcortin on measured serum cortisol concentrations was recently recognized in patients with sepsis (31), trauma (31), and others undergoing major surgery (90). In these reports, the authors (31, 90, 91) recommended the use of a calculated correction factor, termed the free cortisol index (defined as serum cortisol divided by transcortin serum concentrations), as a surrogate marker that better defines glucocorticoid secretion. The latter studies did not provide measurements of the actual serum free cortisol concentrations and did not take into account the impact of hypoalbuminemia that often accompanies low serum transcortin levels.
One of the mechanisms introduced to explain the decrease in serum transcortin concentration during critical illness is its cleavage by the elastase secreted by activated neutrophils at the site of inflammation (92). The latter process would result in delivery of free cortisol in target cells and, in this instance, at the site of inflammation (92). It is not known whether this process is modulated by the prevailing systemic concentration of free cortisol in the serum.
In a study involving 66 critically ill patients, we noted that the patients have markedly increased glucocorticoid secretion that is indiscernible when only serum cortisol level is measured. Such patients had a 7- to 10-fold increase in serum free cortisol concentrations (45). Despite normal adrenal function, as determined by appropriately elevated baseline and Cosyntropin-stimulated serum free cortisol levels, approximately 40% of hypoproteinemic patients had subnormal Cosyntropin-stimulated serum total cortisol levels. The study suggested that caution should be exercised in interpreting baseline and Cosyntropin-stimulated serum total cortisol data in critically ill patients with hypoproteinemia. A very recent study conducted in patients with sepsis and others with septic shock showed similar findings, supporting the superiority of serum free cortisol over that of total cortisol in defining the HPA function (37). The authors of the latter study reported (37) that reasonable estimates of the prevailing serum free cortisol concentrations can be calculated using the method of Coolens et al. (89). Additional data, involving a larger number of patients, particularly those with hypoproteinemia, are needed before this method of calculating free cortisol can be confidently applied in critically ill patients.
B. Serum cortisol assays
Commercially available assays for serum cortisol determine the total (free plus protein-bound fractions) hormone concentration. The specificity, sensitivity, coefficient of variation, and performance of these commercially available assays are not uniform because they show wide variations in immunoassay characteristics (93). It is possible that the variations in assay characteristics might be even more significant in critically ill subjects, especially those with septic shock. Some patients have heterophile antibodies in their sera that interfere in several immunoassay systems, including that of cortisol (94, 95). The prevalence of these heterophile antibodies is not known (94, 95). The most specific assay uses mass spectrometry but is not commonly available. It is evident from the degree of variation in assay results that different cortisol immunoassays may over- or underestimate the actual cortisol value.
C. Standard tests used in the assessment of HPA function during critical illnesses
Several approaches to evaluate adrenal function in critically ill patients have been adopted by various investigators. Whereas some relied entirely on random serum total cortisol concentrations, a few used the low dose (1 µg), although most have used the standard dose (250 µg) Cosyntropin in defining normal adrenal function (Tables 1
and 2). There is, however, no consensus as to which, if any, of these approaches should be used as the criteria for normal function. Most have used baseline and/or Cosyntropin-stimulated serum cortisol data obtained in healthy subjects as the criteria to define normal adrenal function in the critically ill. Recent data, however, demonstrated one of the limitations of the latter approach (22, 45). In that respect, it was shown that critically ill subjects have a higher baseline as well as Cosyntropin-stimulated serum cortisol levels (22, 45). The reliance of total serum cortisol concentrations in such patients with a high likelihood for being hypoproteinemic introduces another significant limitation of these data (45).
Even with the above limitations, the available data are so variable that it would be difficult to achieve a consensus. Random serum total cortisol levels of anywhere from 15 and up to 34 µg/dl have been advocated as a criterion for normality by some investigators (Tables 1 and 2). Others who use Cosyntropin-stimulated levels have used either an increment above baseline (most use >9 µg/dl) and/or a peak response of 20–25 µg/dl (Tables 1 and 2). Thus, depending on which criterion was used, the incidence of adrenal dysfunction in critically ill subjects ranged from 0 to more than 60%, as shown in Tables 1 and 2. Recent and earlier data indicate the use of the Cosyntropin-stimulated increment in serum total cortisol (commonly taken as >7–9 µg/dl) as a diagnostic criterion in critically ill patients can be misleading because nearly 40% of critically ill patients who have no known adrenal disease and who recovered from their illness without glucocorticoid therapy would have been misdiagnosed (22) with relative adrenal insufficiency. It is evident that the Cosyntropin test is not necessarily the best approach in defining normal adrenal function unless primary adrenal insufficiency is the major consideration. The Cosyntropin test is an imperfect test in patients suspected of having central (hypothalamic or pituitary deficiency) glucocorticoid deficiency. Most patients with central adrenal insufficiency have partial rather than complete ACTH deficiency and might therefore have normal response to Cosyntropin (10, 96). It is important to note that measurements of Cosyntropin-stimulated serum cortisol levels in patients receiving dexamethasone are unreliable because even a short course of the latter glucocorticoid can alter the response.
The incremental rise in serum total cortisol levels has been used by some investigators to define adrenal dysfunction in critically ill subjects, particularly those with septic shock. The report by Rothwell et al. (24) depicted the Cosyntropin-induced increment as an important prognostic feature in patients with sepsis. The authors introduced the term relative adrenal insufficiency to describe the blunted response to Cosyntropin in these patients (24). Since then, many investigators adopted the latter definition and used that approach to explore potential benefit from glucocorticoid administration. Whereas the concept of relative adrenal insufficiency will be discussed in a subsequent section, it is reasonable to consider possible explanations for the blunted (<250 nmol/liter or 9 µg/dl) Cosyntropin-induced increment in serum cortisol levels.
It is important to emphasize that, in patients with normal levels of transcortin, cortisol binding to the latter binding protein is saturated at cortisol concentrations of 22–25 µg/dl (86, 87, 88, 89). An increase in serum cortisol above that level (e.g. with Cosyntropin stimulation) would be mostly reflected by an increase in the albumin bound and free fractions of cortisol in the circulation. Current assays for serum cortisol determine the total serum level (transcortin bound, albumin bound, and free). It is therefore reasonable to suggest that a decrease in serum transcortin level, which is a characteristic feature of many critically ill subjects, especially those with septic shock, would result in an increase in the percent free cortisol as was recently demonstrated (37, 45). Thus, the low or blunted Cosyntropin-induced increment in serum total cortisol in critically ill subjects could very well be a physiological reflection of reduced transcortin levels in critically ill subjects. Our own data on the Cosyntropin-induced increment in serum total cortisol concentrations are consistent with this explanation (Table 4). The percent of patients who had a Cosyntropin-induced increment in serum cortisol levels of less than 9 µg/dl was significantly lower (P = 0.002) among critically ill patients with near-normal serum albumin and transcortin concentrations than those with hypoproteinemia (one third vs. one half), despite the fact that both groups had similar free cortisol levels. Similarly, 17% of healthy volunteers had such a Cosyntropin-induced increment. In the setting of septic shock, available data suggest that a blunted (<250 nmol/liter) Cosyntropin-induced increment in serum cortisol is a poor prognostic feature associated with increased mortality (24, 36). It is not known whether this would be true for other forms of critical illnesses.
|
500 nmol/liter). In contrast, all of the critically ill with near-normal serum proteins had a peak Cosyntropin-stimulated serum cortisol level of more than 20 µg/dl (45). Importantly, the serum free cortisol levels in the latter two groups were similar. Thus, it would be safe to state that in critically ill subjects with an albumin of more than 2.5 gm/dl, the peak Cosyntropin-stimulated serum cortisol would be expected to be more than 20 µg/dl with a median value of 32 µg/dl. A Cosyntropin-stimulated serum cortisol of less than 20 µg/dl in this setting of critical illness (albumin > 2.5 gm/liter) would be highly indicative of adrenal insufficiency. Only limited data are available on the use of low-dose (1 µg) Cosyntropin testing in critically ill patients (97, 98, 99, 100, 101). Whereas some studies noted similar findings between the two testing doses of Cosyntropin, others found that the standard dose provides better assessment and higher increments in serum total cortisol concentrations. At this point, most of the data in the literature have used the standard-dose Cosyntropin stimulation, and the available data on the low-dose test are limited and are not sufficient to make sound recommendations.
Other tests assessing the integrity of the entire HPA axis have been used in many instances in which evaluation of adrenal function was attempted. Such tests include the use of insulin-induced hypoglycemia, metyrapone testing, and determining the pituitary (ACTH) and adrenal (cortisol) responses to the administration of the hypothalamic hormone, CRH. The limited data available on the latter test (CRH) demonstrated enhanced responses but without any additional diagnostic accuracy (57). Although the insulin-induced hypoglycemia is considered the gold standard in the assessment of adrenal function, it is impractical and unsafe in the setting of critical illness. Similarly, metyrapone is not only hard to obtain, but the test is also very unsafe, and the data are difficult to interpret in the setting of critical illness. Hypotension is a potent stimulus for activation of the HPA axis as well as vasopressin release. In that respect, some investigators have looked at the serum total cortisol levels during hypotension as an assessment tool for these patients (34, 102).
In summary, there are limitations to all of the tests used to define adrenal function in critically ill patients. Despite these limitations, baseline and Cosyntropin-stimulated serum total (and preferably free) cortisol levels remain the most practical approaches. As will be discussed below, additional value is likely to be obtained by measuring the other ACTH-dependent steroids, namely DHEA and its sulfated ester, DHEA-S. Insufficient data are currently available to suggest the alternative use of the low dose (1 µg) test. Acknowledging and appreciating the limitations of the Cosyntropin testing are important in interpreting data in these patients.
D. Medications
Many medications alter the baseline and/or the Cosyntropin-stimulated serum cortisol levels (Table 3). Such medications often influence binding proteins (e.g. the estrogen-induced increase in transcortin), directly interfere with glucocorticoid synthesis (e.g. ketoconazole, etomidate), or have direct inhibitory effects on CRH /ACTH secretion (e.g. oral, dermal, intraarticular, or inhaled glucocorticoids). Some drugs have direct antiglucocorticoid effects (e.g. RU486), whereas others have glucocorticoid-like activities and would therefore suppress the HPA axis just like glucocorticoids would. Examples of the latter class would include progestational agents such as medroxyprogesterone (short-acting and depo forms) and a similar agent, megestrol, used in the treatment of breast and endometrial cancers as well as in the management of anorexia. Such drugs have glucocorticoid-like activity sufficient enough to result in clinical features of Cushing’s syndrome with prolonged therapy (103). Similarly, the use of such drugs before or during critical illness can alter the functional integrity of the HPA and greatly increase the likelihood for true adrenal insufficiency.
As shown in Table 3, etomidate is another drug that can influence adrenal function during critical illness. Etomidate is a carboxylated imidazole that is still used as an anesthetic agent to facilitate endotracheal intubation (104, 105, 106, 107). The drug was shown to cause reversible inhibition of the 11-hydroxylase enzyme and result in decreased cortisol secretion (105). Initial reports showed that its prolonged use caused adrenal insufficiency with the associated increased morbidity and mortality (104). Subsequent reports indicated that when given as a single injection for induction of anesthesia or as an infusion to maintain that etomidate was associated with impaired HPA function (106, 107). Despite its known effects on the HPA function, some reports suggested that etomidate can still be used (108). Recent opinions in the critical care literature concluded that etomidate should be abandoned (109), whereas others suggested that, in light of its other properties, it can still be used as long as its effects on adrenal function are acknowledged and addressed (110). From an endocrine standpoint, it would be best to avoid the use of etomidate if possible. However, if under certain circumstances etomidate is needed for induction of anesthesia, then hydrocortisone therapy should be administered for 24–36 h.
E. Type and severity of illness
During the acute phase of a critical illness, serum cortisol levels are generally proportionate to the degree of stress (16). This was best demonstrated in patients with presumed normal adrenal function who had various surgical procedures of increasing complexity (16). In one study (31), serum cortisol levels after a major trauma were as elevated as those seen in patients with sepsis. Another study by Sam et al. (35) showed that septic patients not only have a wide range of serum cortisol levels but also that those levels do not correlate with the commonly used measure of illness severity, the score of the Acute Physiology, Age and Chronic Health Evaluation. However, the same study found that patients with the highest random total cortisol levels have the highest mortality rate (35).
A recent study examined the serum levels of macrophage-MIF in patients with sepsis and in others with trauma (60). The study demonstrated that serum total cortisol levels were similarly elevated in the two patient groups (60). In contrast, serum levels of macrophage-MIF were markedly higher in patients with sepsis, compared with those who had trauma (60). Higher levels of the same factor were also observed in the subgroup of patients who developed acute respiratory distress syndrome and in those who did not survive (60). As expected, patients with sepsis had much higher levels of serum markers of inflammation such as procalcitonin, C-reactive protein, and LPS-binding protein.
The presence or absence of glucocorticoid resistance during critical illnesses is an important factor influencing adrenal function in general and the effects of cortisol at the tissue and cellular levels. Currently there are no definitive data assessing the impact of glucocorticoid sensitivity in a variety of critical illnesses on adrenal glucocorticoid secretion. It is reasonable to suggest that patients who suffer from critical illnesses associated with glucocorticoid resistance would be expected to have higher glucocorticoid levels than patients with illnesses not associated with resistance.
A recent study investigated the reproducibility of two Cosyntropin tests (1 d apart) in critically ill patients (111). The authors found that, in the critically ill group of patients without sepsis, repeat Cosyntropin testing yielded similar and reproducible results (111). In contrast, there was no correlation between the two consecutive Cosyntropin tests in patients with septic shock such that five of eight subjects who were considered to have relative adrenal insufficiency (increment of <9 µg/dl) on the first test had normal responses on the following day (111). Similarly, six of 12 subjects who had normal responses (>9 µg/dl) had blunted responses to Cosyntropin the following day (111). Similar findings were previously reported by Bouachour et al. (112). The reason(s) for the discordance between the two consecutive Cosyntropin testing results in patients with septic shock are not clear. The findings raise concerns and questions about the validity of one-time testing in this group of patients. It is not known whether fluid administration to these patients during the first day of critical illness contributed to these alterations. Another confounding factor in these hypotensive patients is that they often have volume contraction that is severe enough to stimulate the release of arginine vasopressin, which enhances CRH effects on ACTH secretion by the pituitary. After iv fluids, plasma volume will no longer be as powerful of a stimulus for ACTH release.
F. Chronicity/duration of the critical illnesses
It is commonly believed that glucocorticoid secretion during critical illnesses is generally proportionate to the degree of stress (5, 16). Earlier studies suggested that the secretory activity of the adrenal glands is augmented during the early phase of a critical illness with subsequent diminution in glucocorticoid secretion as the illness progresses into a subacute or chronic phase. However, most of the latter studies demonstrating decreased adrenal glucocorticoid secretion with advancing chronicity of critical illnesses were based on measurements of serum total cortisol levels. The latter measurements have serious limitations, especially during chronic illnesses in which malnutrition and hypoproteinemia are more common. This is supported by the findings that when serum free cortisol levels are measured, they continue to be elevated throughout the critical illness (45).
Similarly, plasma ACTH concentrations are reported to be high during the early phase of any critical illness (58). Such levels have been shown to be increased after major surgical procedures (Fig. 2) as well as experimental endotoxemia (77, 78, 79, 80, 81) and practically after most stressful events. However, over time, plasma ACTH levels decline gradually, even though serum cortisol (total and free) concentrations continue to be elevated (58). As discussed earlier, it is believed that the discordance between plasma ACTH and serum cortisol levels during prolonged critical illnesses raises the question of the presence of other factor(s) that are stimulating and regulating glucocorticoid secretion. Such factors include endothelin, ANF, arginine vasopressin, and other cytokines (58).
G. Effects of intravascular plasma volume/hemodilution
Because cortisol circulates predominantly as a protein-bound steroid, its measured level is greatly influenced by the concentration of its binding proteins, transcortin, and albumin. It is well known that changes in intravascular fluid volume have a major impact on the concentration of many cells, and compounds present in the circulation, such as red blood cells, urea nitrogen, and plasma proteins, to name a few. The influence of the changes in the intravascular fluid volume on serum cortisol levels has been best illustrated in patients undergoing coronary artery open heart bypass surgery (113). In such patients, the hematocrit, albumin, and transcortin values decrease rapidly with the institution of extracorporeal circulation (113). Similarly, serum total cortisol (mostly bound) levels decrease proportionately, whereas the free cortisol concentrations remain elevated and unchanged (113). After the bypass is completed, serum total cortisol as well as transcortin concentrations increase rapidly (113).
Although the degree of acute hemodilution in patients during extracorporeal circulation is unique, it is reasonable to suggest that similar, albeit less impressive, effects occur in other instances during which hypotensive patients who are volume contracted are given large amounts of iv fluids for volume resuscitation. In a recent study by Le Roux et al. (90, 114), a significant decrease in serum levels of transcortin was noted in the postoperative period in patients undergoing elective surgery. The latter study showed that the decrease in serum transcortin levels, which was partly due to large amounts of fluids administered during surgery, led to a decrease in measured total serum cortisol concentrations that would have been mistakenly diagnosed as adrenal insufficiency (90). The authors reported that the latter mistake would be avoided by the use of a calculated free cortisol index, defined as the serum cortisol concentration divided by the serum transcortin levels (90). Similar conclusions were reached by other investigators studying patients with trauma and others with sepsis (31, 91). However, the latter studies did not address volume status as a contributing factor. It is likely that patients in the latter studies received large amounts of fluids because they were hypotensive and that they probably had dilutional decrease in serum albumin as well as transcortin concentrations (31). Thus, it is imperative to consider patients’ volume status in the interpretation of serum cortisol concentrations. An additional confounding factor here is that volume contraction is a powerful stimulus for arginine vasopressin release, which can augment the effects of CRH on ACTH release. Fluid resuscitation can often suppress this stimulus for increased ACTH secretion and result in lower plasma levels. Similarly
Top
Abstract
Introduction
2.5 g/dl) and others with near-normal serum albumin levels (>2.5 g/dl)