Diagnosis and Treatment of Adrenal Insufficiency in the Critically Ill Patient
Kwame Asare, Pharm.D. Pharmacotherapy. 2007;27(11):1512-1528. ©2007 Pharmacotherapy Publications
Abstract and Introduction
The reported incidence of adrenal insufficiency varies greatly depending on the population of critically ill patients studied, the test and cutoff levels used, and the severity of illness. Several studies have shown increased mortality in patients with very low or very high baseline cortisol levels. Manifestations of adrenal insufficiency in the critically ill patient are numerous and nonspecific, so clinicians are urged to have a high index of suspicion and be alert to important diagnostic clues, such as hyponatremia, hyperkalemia, and hypotension, that are refractory to fluids and vasopressors without any clear causation. Multiple tests have been developed to diagnose adrenal insufficiency, but the most commonly used test in the intensive care unit is the adrenocorticotropic hormone (ACTH) stimulation test. The low-dose ACTH stimulation test has been shown to be more sensitive and specific than the high-dose test; however, the high-dose test is preferred since the low-dose test has not been validated. Although diagnosing adrenal insufficiency continues to be difficult in the critically ill patient, administration of highdose corticosteroids, defined as methylprednisolone 30 mg/kg/day or more (or its equivalent), over a short period of time provides no overall benefit and may even be harmful; however, administration of low-dose corticosteroids for a longer duration decreases both the amount of the time that vasopressors are required and mortality at 28 days. Hydrocortisone 200-300 mg/day, administered in divided doses or as a continuous infusion, is the preferred corticosteroid in patients with septic shock and should be started as early as possible. For patients in whom the ACTH stimulation test cannot be given immediately, clinicians are urged to consider using dexamethasone until such time that the test can be administered, since, unlike hydrocortisone, it does not interfere with the cortisol test.
Chronic primary adrenal insufficiency was first recognized by Addison in 1844 and described in 1855 in one of the classic articles in medicine. Subnormal cortisol production is the hallmark of the condition. The deficiency of cortisol secretion by the adrenal cortex has been shown to increase morbidity and mortality in patients with septic shock. Septic shock has been associated with a greater than 50% mortality rate,[3,4] and very little progress has been made.[4,5] Administration of corticosteroids to adrenalectomized animals improved survival, whereas the continued suppression of adrenocortical function increased the mortality rate in critically ill patients. The frequency and diagnosis of adrenal insufficiency in critically ill patients remain controversial despite studies demonstrating beneficial outcomes from treatment. Adrenal insufficiency is estimated to occur at a rate of 0-30% in the critically ill population[8,9] and may be as high as 25-40% in patients with septic shock,[10,11] depending on the specific tests and threshold used to diagnose adrenal insufficiency, underlying disease, and severity of illness.
The two adrenal glands, each weighing about 4 g, lie at the superior poles of the two kidneys. Each gland is made of two distinct parts, the adrenal medulla and adrenal cortex. The adrenal medulla acts in concert with the central nervous system to secrete the hormones epinephrine and norepinephrine in response to sympathetic stimulation.
The adrenal cortex is anatomically separated into three zones: zona glomerulosa, zona fasciculata, and zona reticularis. However, functionally it can be seen as two independent units: the outer zona glomerulosa, which is responsible for secreting the mineralocorticoid aldosterone, and the inner zonae fasciculate and reticularis, responsible for secreting glucocorticoids (e.g., cortisol) as well as androgens. An adrenocorticotropic hormone (ACTH), also called corticotropin or adrenocorticotropin, stimulates the adrenal cortex to synthesize and secrete the two principal adrenocortical hormones: cortisol, which regulates carbohydrate, protein, and lipid metabolism, and aldosterone, which regulates fluid and electrolyte balance through sodium and potassium homeostasis. In addition to cortisol and aldosterone, the adrenal cortex also secretes other steroids with glucocorticoid, mineralo-corticoid, or both activities, but in much smaller quantities.
Cortisol levels respond within minutes to stressful stimuli, protecting the organism from the damaging effects of the stressor. Without this response, humans could not resist physical or mental stress, and thus, any minor illness could result in death. The reasons why elevated glucocorticoid levels protect the organism under stress are not completely understood. The detailed process, known as steroidogenesis, is rather complex and beyond the scope of this review, but Figure 1 (which depicts a rather simplified version) will suffice for the purposes of this discussion.
Three organs-hypothalamus, anterior pituitary, and adrenal cortex, collectively known as the hypothalamic-pituitary-adrenal (HPA) axis-maintain appropriate levels of glucocorticoids. There are three distinct modes of regulation of the HPA axis: diurnal rhythm in basal cortisol secretion, marked increases in steroidogenesis in response to stress, and the negative feedback regulation by adrenal cortisol. Typically, cortisol is released in episodic bursts in a diurnal pattern. This diurnal rhythm is such that ACTH secretion rises during the late hours and peaks in early morning, around 8 A.M., after which the negative feedback regulation by the HPA axis modulates the glucocorticoid levels in the appropriate range (Figure 1).[9,12] When daily sleeping habits are changed, the cycle changes correspondingly. Thus, measurements of blood cortisol levels are meaningful when expressed in terms of the time of a person's sleep cycle at which they were measured. It must be kept in mind that stressful circumstances as well as chronic disease may override these normal negative feedback mechanisms, resulting in marked increases in plasma levels of corticosteroids. Almost any type of physical or mental stress can result in enhanced secretion of ACTH, and consequently cortisol secretion, by as much as 20-fold, a factor that is proportional and positively correlated to the severity of the condition. Examples of stressful stimuli include trauma of almost any type, major surgery, severe infection, pain, hypoglycemia, hypovolemia, hypotension, hypoxemia, bleeding, fear, and intense heat or cold.
Adrenocorticotropic hormone is necessary for aldosterone secretion but has little effect in controlling its rate of secretion. Aldosterone secretion is not regulated by the HPA axis but primarily by the renin-angiotensin system in response to a decrease in blood volume, and by elevated extracellular potassium. In the setting of persistently elevated ACTH levels, mineralocorticoid levels initially increase and then return to normal. This phenomenon is sometimes called the ACTH escape.
Divided into two major classes, glucocorticoids and mineralocorticoids, corticosteroids have several important functions. Glucocorticoids preferentially regulate carbohydrate, protein, and fat metabolism and indirectly lead to insulin secretion to counterbalance glucocorticoidinduced hyperglycemia. Insulin, in turn, allows glucose entry into the cells. Under normal conditions, the rate of glucose production regulates plasma glucose concentrations; however, at higher levels of plasma glucose, increases in glucose clearance mediated by insulin become more important. Glucocorticoids also possess antiinflammatory and immuno-suppressive effects; they maintain cardiovascular integrity and blood glucose level, as well as a host of other functions. Cortisol is the primary glucocorticoid in humans, accounting for about 95% of all glucocorticoid activity. Recent studies show that it is produced at a rate of 10 mg/day (equivalent to 20-30 mg/day of hydrocortisone) in contrast to earlier studies that indicated a daily rate of 20-30 mg. Table 1 shows typical rates of secretion of the two most active corticosteroids in humans.
Mineralocorticoids, on the other hand, preferentially regulate fluids and electrolytes such as sodium, potassium, and hydrogen ions. Their net effect is sodium conservation, potassium and hydrogen ion excretion, and expansion of the vascular space. Aldosterone is the primary mineralocorticoid in humans, accounting for about 90% of all mineralocorticoid activity. Cortisol, which has only 1/400th the mineralocorticoid activity of aldosterone, also provides a significant amount of mineralocorticoid activity because of its high plasma concentration. Its plasma concentration is nearly 1000 times that of aldosterone.
Corticosteroids may also be grouped according to their relative glucocorticoid and mineralocorticoid activities. The estimated potencies and half-lives of several common corticosteroids are listed in Table 2 . As synthetic glucocorticoids become increasingly longer acting and potent, their mineralocorticoid activity decreases. Several glucocorticoids (e.g., dexamethasone, methylprednisolone) do not have sufficient mineralocorticoid activity to replace aldosterone and, when prescribed, need the addition of a potent mineralocorticoid (e.g., fludrocortisone). Aldosterone has potent mineralocorticoid activity but, at normal doses or normal rates of secretion, has insignificant glucocorticoid activity and therefore acts as a pure mineralocorticoid. It is by far the most potent naturally occurring mineralocorticoid. Aldosterone acts on the distal tubules and collecting ducts of the kidney to promote the reabsorption of sodium and to increase the urinary excretion of potassium and hydrogen ions, in effect promoting sodium retention at the expense of potassium and hydrogen ions. Hence, its deficiency in patients who are adrenally insufficient can lead to sodium loss, volume depletion, hypotension, and vascular collapse. Total loss of aldosterone secretion usually results in death within 3-14 days unless the patient receives an exogenous mineralocorticoid.
Approximately 90% of cortisol in plasma is reversibly bound to plasma protein under normal circumstances. Almost all of the steroid-binding capacity is accounted for by two proteins: cortisol-binding globulin (also known as transcortin) and, to a lesser extent, albumin. Cortisol-binding globulin has a high affinity for corticosteroids but relatively low binding capacity, whereas albumin has a low affinity but large binding capacity. Only the fraction of corticosteroid that is unbound or free can produce an effect. Recent evidence suggests that in critically ill patients, there is a decrease in cortisol binding and an increase in free cortisol.[18,19] In patients with septic shock, adrenal insufficiency is believed to result from suppression of the HPA axis by cytokines and other inflammatory mediators. Cytokines also cause resistance to glucocorticoids by decreasing the affinity of glucocorticoids for their receptors. Reduction in global cortisol transport has also been reported in patients with septic shock.
Cortisol, with its high degree of protein binding, is protected from hepatic clearance and therefore has slow elimination and a relatively long half-life (60-90 min). Aldosterone, with its lower protein binding (∼60%), has a relatively short half-life (20 min). Binding of corticosteroids to plasma proteins may serve as a reservoir to decrease the fluctuations in freehormone concentrations, providing a uniform distribution of these hormones to the tissues.
The corticosteroids are metabolized by the liver to form glucuronic acid and sulfates. These metabolites are inactive and have no glucocorticoid or mineralocorticoid activity. They are primarily filtered by the kidneys and excreted in the urine (75%), whereas the remaining portion (25%) is eliminated by the bile and feces. Severe liver dysfunction markedly decreases the metabolism of these hormones, and renal dysfunction decreases their rate of elimination. Elimination of glucocorticoids is enhanced by drugs that increase the hepatic metabolism of cortisol (e.g., phenobarbital, rifampin, phenytoin) and is decreased by drugs that inhibit the metabolism of cortisol (e.g., protease inhibitors). Therefore, patients receiving these drugs may need higher or lower doses of glucocorticoids, respectively.
Adrenal insufficiency can be categorized into two types: primary and secondary. Primary adrenal insufficiency, or Addison's disease, is caused by the inability of the adrenal gland to produce cortisol, aldosterone, or both, with the rest of the HPA axis remaining intact. It is associated with a deficiency of glucocorticoid, mineralocorticoid, sex hormones, and cathecholamines (cortisol is required for synthesis of cathecholamines). Primary adrenal insufficiency is characterized by low cortisol production and high plasma ACTH concentration because of the lack of normal ACTH feedback inhibition. The secretion of high levels of ACTH results in the secretion of other hormones with similar chemical structure. One such hormone is melanocyte-stimulating hormone. This hormone causes melanocytes to form a black pigment, melanin, in the epidermis of the skin, leading to hyperpigmentation of the skin and mucous membranes. The effect of this hyperpigmentation is much greater in people with dark skin than in those with light skin. Primary adrenal insufficiency is characterized by orthostatic hypotension, hyponatremia, hyperkalemia, mild metabolic acidosis, and as just described, hyperpigmentation of the skin. Primary adrenal insufficiency becomes clinically evident when about 90% of the adrenal cortex has been destroyed.
Secondary adrenal insufficiency is caused by dysfunction of the hypothalamus, pituitary gland, or both (with a normal adrenal gland) and is characterized by low cortisol production and low or normal plasma ACTH concentration. The glucocorticoid deficiency and low ACTH concentration may result in hypotension and hyponatremia, with normal potassium and hydrogen ion concentrations. The ACTH-induced hyperpigmentation is absent in secondary adrenal insufficiency, and the release of aldosterone, sex hormones, and cathecholamines is usually normal.
Primary and secondary adrenal insufficiency may be categorized as acute or chronic. The most common cause of acute adrenal insufficiency is functional (referred to as relative adrenal insufficiency), from exogenous glucocorticoid administration suppressing the HPA axis. The degree of suppression depends on the pharmacokinetics, dose, and duration of the steroid administered, with larger doses of agents with longer half-lives and an extended course of therapy prolonging the suppression of the HPA axis. Adrenal insufficiency has been reported to occur in patients receiving prednisone 25 mg twice/day for as short a period as 5 days. It should be anticipated in any patient receiving more than 30 mg/day of hydrocortisone or its equivalent for more than 3 weeks. In addition to oral and parenteral routes, glucocorticoids can produce HPA-axis suppression when administered intranasally or by inhalation at high doses.[21-23] The time to recovery of the HPA axis after the discontinuation of exogenous glucocorticoids is variable and may be as short as 2-5 days or as long as 9 months-1 year.[20,24]
In functional or relative adrenal insufficiency, there is either insufficient cortisol or a cortisol level that may be high in absolute terms but insufficient to respond to the degree of stress. Thus, serum cortisol concentrations that are normal in well patients may be inappropriately low in severely sick patients. This inability to mount the appropriate response increases the risk of death during severe illness. Although absolute adrenal insufficiency is quite rare in critically ill patients, relative adrenal insufficiency is considerably more common. In chronic critical illness, adrenal insufficiency may also result from catecholamine receptor desensitization or downregulation and/or chronic secretion of cytokines and other substances that suppress the HPA axis. This latter mechanism is sometimes called adrenal exhaustion syndrome.
Although the causes of primary and secondary adrenal insufficiency are numerous, it has been postulated that the most common causes in the critically ill patient are sepsis and systemic inflammatory response syndrome. The most likely mechanisms are decreased synthesis and/or decreased release of ACTH, corticotropinreleasing hormone (CRH), and cortisol by cytokines and other inflammatory mediators released during sepsis. In vitro, different cytokines reversibly impair glucocorticoid receptor affinity to cortisol or glucocorticoid response elements. This hypothesis was supported in another study that reported decreased responses to corticotropin in patients with septic shock compared with responses in patients without septic shock, despite similar basal cortisol levels.
In the ambulatory patient, tuberculosis remains the most common cause of primary adrenal insufficiency worldwide, with human immunodeficiency virus and other infections in immunosuppressed patients being a close second. In the West, however, the most common cause is idiopathic adrenal insufficiency, also known as autoimmune adrenal insufficiency. Table 3 lists some of the diseases and conditions that can possibly cause adrenal insufficiency. As stated earlier, the most common cause of secondary adrenal insufficiency is the abrupt discontinuation of glucocorticoids.
Several drugs have been implicated in adrenal insufficiency. Glucocorticoid production is impaired by ketoconazole, megestrol (doses > 160 mg/day),[33,34] medroxyprogesterone, aminoglutethimide, mitotane, metyrapone, etomidate,[36,37] and high-dose fluconazole (doses ≥ 400 mg). In a descriptive case report, a possible temporal association of reversible adrenal insufficiency with the institution and withdrawal of fluconazole in two critically ill patients was reported. The authors recommended that in critically ill patients who are not routinely receiving glucocorticoids and who require fluconazole, corticosteroid responsiveness should be monitored with the ACTH stimulation test. Although a pilot study, their data suggest a need for further investigation of the association between high-dose fluconazole and adrenal insufficiency in critically ill patients.
The use of the intravenous anesthetic, etomidate, has also been associated with adrenal insufficiency and increased mortality. It is a selective inhibitor of adrenal 11β-hydroxylase, the enzyme that converts deoxycortisol to cortisol. One group measured the cortisol and aldosterone responses to ACTH stimulation in five patients receiving etomidate. All five patients were found to have adrenal suppression. One patient received a 20-hour infusion and developed marked adrenocortical suppression that was still evident 4 days after discontinuation of the drug. The authors cautioned clinicians that etomidate inhibits adrenal steroidogenesis and suggested coadministration of corticosteroids in selected patients.
The clinical manifestations, as described by Addison, remain as accurate today as they were in 1855. They are similar and nonspecific for both primary and secondary adrenal insufficiency ( Table 4 ). At presentation, patients may have fatigue, weakness, joint pain, dizziness, depression, and gastrointestinal symptoms such as nausea, vomiting, abdominal cramps, unintentional weight loss, and anorexia nervosa. Both primary and secondary adrenal insufficiency can cause hyponatremia, but the pathophysiology of the two disorders is completely different. In primary adrenal insufficiency, hyponatremia is mainly due to aldosterone deficiency and sodium wasting, whereas in secondary adrenal insufficiency it is due to low cortisol and free water retention mediated by secretion of vasopressin (antidiuretic hormone) as a result of relative volume depletion.
For any given level of morning serum cortisol, patients with primary adrenal insufficiency present with more severe symptoms compared with those of patients with secondary adrenal insufficiency. The most specific signs of primary adrenal insufficiency are hyperpigmentation of the skin, craving for salt, hyperkalemia, and acidosis. Specific signs of secondary adrenal insufficiency are pale skin, loss of axillary and pubic hair, decreased libido, and impotence. Because it remains difficult to recognize adrenal insufficiency in the intensive care unit (ICU), clinicians are encouraged to have a high index of suspicion. They should watch for important diagnostic clues like the presence of hyponatremia, hyperkalemia, hypoglycemia (rare), and hemodynamic instability despite adequate fluid and vasopressor therapy (the most common feature). Bear in mind that hyperkalemia may be absent in patients with secondary adrenal insufficiency because of the intact secretion of mineralocorticoid, but it is usually present in primary adrenal insufficiency.
A comparison of the different diagnostic tests is presented in Table 5 .
The Adrenocorticotropic Hormone Stimulation Test
Measurement of cortisol levels with the ACTH stimulation test is the standard and the most convenient test for diagnosing adrenal insufficiency in the ICU. The test is based on the inability of a diseased adrenal gland to secrete adequate cortisol after injection of corticotropin. The test is quick, simple, insensitive to interferences, reliable, and essentially free of adverse effects aside from occasional anaphylactic reactions. However, its interpretation has generated some controversy. A normal response to the test does not rule out adrenal insufficiency since patients with acute-onset adrenal insufficiency (it may take up to 3 wks for the adrenal cortex to readjust to the reduced level of ACTH secretion) and ACTH resistance may respond normally to the test. Also, since the test stimulates the adrenal gland directly, thereby bypassing the hypothalamus and pituitary gland, patients with defects in these organs may be missed by the test.
Corticotropin is a synthetic peptide that consists of the first 24 amino acids of human ACTH. After drawing a baseline blood sample, it is given any time of the day at a dose of 1 μg (low dose) or at a supraphysiologic dose of 250 μg (high or conventional dose), stimulating adrenocortical steroidogenesis. Blood samples for the measurement of serum cortisol are drawn at time 0 (baseline), 30, 60, and sometimes 90 minutes. Ideally, if there is clinical suspicion, clinicians should perform the ACTH stimulation test at the outset rather than wait for the results of a baseline measurement of cortisol.
High- versus Low-dose Adrenocorticotropic Hormone Stimulation Test. Use of the low-dose test has been proposed by some clinicians.[42,43] It is prepared by diluting the 250-μg vial in 250 ml of normal saline to obtain a concentration of 1 μg/ml and using a 1-ml syringe to accurately draw a dose. The syringes can be kept at 4°C for up to 4 months without any decline in biologic activity. Results from a number of studies indicate that the low-dose test is more sensitive that the high-dose test for diagnosing adrenal insufficiency, but it is not as sensitive as the insulin tolerance test.[42,43,45,46] In a recent study, low-dose and high-dose ACTH stimulation tests were compared in 46 patients with septic shock. The two tests were performed consecutively at an interval of more than 4 hours. In each test, the cortisol level was measured at baseline, 30, 60, and 90 minutes. Responders were defined as those with a maximal increase in cortisol level over baseline of at least 9 μg/dl. The authors concluded that the low-dose test identified a subgroup of patients in septic shock with inadequate adrenal reserve who had poor outcomes and would have been missed by the high-dose test. Although the authors stopped short of suggesting that the low-dose test should replace the high dose-test, they expressed interest in the evaluation of glucocorticoid replacement therapy in patients with normal high-dose tests but abnormal low-dose tests. No published, large, randomized, controlled trials have shown that this patient population will actually benefit from glucocorticoid therapy.
The high-dose test is performed by direct intravenous or intramuscular injection of the 250-μg dose of ACTH. Because high-dose ACTH reaches plasma levels approximately 1000 times the values of maximally stressed healthy individuals, it overrides adrenal resistance to ACTH and results in a falsely high normal response in patients with adrenal insufficiency,[45,46] a mechanism similar to the effect of insulin in patients with type 2 diabetes. In healthy individuals, as little as 5 μg of synthetic ACTH or 10 μg of human ACTH maximally stimulates the adrenal cortex. In one study, several patients were characterized as nonresponders with the low-dose test even though they responded adequately to the high-dose test. Thus, the reliance on the high-dose test as the sole measure of adrenal function may cause the withholding of a lifesaving therapy in a significant number of patients.
Four studies have been published that compared the low-dose and high-dose tests in critically ill patients.[31,46,49,50] All four studies concluded that the low-dose test is more sensitive than the high-dose test. Based on these studies, it seems reasonable that therapeutic decisions can be made by using the low-dose test. In a classic review article, the author proposed that because the low-dose test is simple to perform, potentially more sensitive than the high-dose test, and less expensive (a single 250- μg vial of ACTH can be used to test multiple patients), it should be the routine test used in the evaluation of primary and secondary adrenal insufficiency. In contrast, a meta-analysis of the available data on different patient populations reported that the predictive value of the low-dose test was not superior to that of the high-dose test in diagnosing secondary adrenal insufficiency. However, this meta-analysis did not provide any data on the use of the low-dose test for diagnosing primary adrenal insufficiency or its use in critically ill patients.
In 2003, critical care and infectious disease experts from 11 international organizations developed management guidelines for the use of corticosteroids in patients with sepsis and septic shock. They concluded that, although the lowdose test has been shown to have a higher sensitivity and specificity in testing for adrenal insufficiency, the high-dose test is still preferred. The reason for this is that the lowdose stimulation test has not been well validated in critically ill patients and patients with septic shock,[41,47] since all the diagnostic thresholds for the ACTH stimulation tests were set with use of the high dose. In other words, no specific diagnostic or prognostic threshold has been set with use of the low-dose test.
The comparison of the high- and low-dose ACTH stimulation tests will not be complete without a mention of sampling times. Many studies used the same sampling times for both tests (0, 30, and 60 min), although there are differences in cortisol kinetics of the two tests. The maximum cortisol concentration is reached after 30 minutes for the low-dose test and returns to baseline after 2 hours, whereas the high-dose test peaks in 90 minutes and returns to baseline after 4 hours. To improve the interpretation of the test results, it has been proposed that three samples should be taken at 30, 60, and 90 minutes after the ACTH injection. If a clinician is interested in taking only two samples, the 30- and 60-minute times are preferred for the lowdose test and 60 and 90 minutes for the highdose test. In critically ill patients, if the two tests are to be given, they should be separated by at least 4 hours to allow cortisol concentrations to return to baseline.[51,53] In a very recent European, retrospective, multicenter, cohort study, the investigators found that very few patients are mistakenly classified as nonresponders when using a 60-minute ACTH-stimulated cortisol level (250-μg dose) instead of the maximum value between the 30- and 60-minute levels. They suggested the 30-minute level may not provide any benefit in terms of diagnostic accuracy. Thus, the test can be performed with just the baseline and 60-minute ACTH-stimulated levels.
Random Cortisol Levels
In the ambulatory patient, the initial baseline cortisol level should be checked between 6 and 8 A.M. because the cortisol level peaks in the early morning. For critically ill patients, it is not necessary to obtain cortisol levels at a specific time of the day because most patients lose the diurnal variation in their cortisol levels.[2,33] It has been proposed that random cortisol testing, measured any time of the day, may be a more useful tool for diagnosing adrenal insufficiency. One group of authors studied the HPA axis in a group of surgical patients who were vasopressor dependent and suggested that a random cortisol level of less than 25 μg/dl is associated with steroid-responsive hypotension.
Another group compared the sensitivity and specificity of the low-dose ACTH stimulation test, the high-dose ACTH stimulation test, and random cortisol concentrations in 59 patients with septic shock. After measurement of a baseline cortisol level within 48 hours of admission, patients received ACTH 1 μg (lowdose test), followed 60 minutes later by a 249-μg dose (high-dose test). The authors defined relative adrenal insufficiency as a random cortisol level of less than 25 μg/dl. They concluded that the sensitivity (the positivity or ability of a test to identify those patients who will respond to steroids) for random cortisol, low-dose, and high-dose tests were 96%, 54%, and 22%, respectively. The specificity (the negativity or ability of a test to correctly identify those patients who will not respond to steroid therapy) was found to be 57%, 97%, and 100%, respectively. In other words, 96%, 54%, and 22% of the patients who fail the random, low-dose, or highdose test, respectively, will respond to steroids. Because of its high sensitivity, the authors concluded that random cortisol testing is a useful diagnostic threshold for the diagnosis of adrenal insufficiency. It has even been suggested by others that, in critically ill patients, ACTH testing is not necessary to diagnose adrenal insufficiency since patients are severely stressed and therefore the HPA axis is already maximally activated with maximum cortisol secretion. These patients may be unable to further increase their cortisol secretion after the ACTH stimulation test. In these patients, a random cortisol level provides significant information on the function of the entire HPA axis.
Interpretation of the Cortisol Tests
It has been firmly established that both low (< 15 μg/dl) and high (> 34 μg/dl) cortisol levels are associated with poor prognosis. One group of authors measured the admission cortisol levels of 260 patients in an ICU and found the mean cortisol level to be 27 μg/dl in survivors compared with 47 μg/dl in nonsurvivors. They concluded that the serum cortisol level is an independent predictor of outcome. Despite this, the threshold for diagnosing adrenal insufficiency is surrounded by controversy.
Different threshold levels have been proposed by different studies during critical illness, but none is entirely acceptable or satisfactory. Some authors have proposed a basal plasma cortisol concentration (measured at early morning) of less than 15 μg/dl as indicative of adrenal insufficiency,[41,60] whereas others suggest that because patients are maximally stressed during septic shock, a plasma cortisol concentration (measured at any time of the day) less than 20-25 μg/dl is more appropriate.[2,31,56,61] Other approaches have used maximum cortisol levels (irrespective of the baseline cortisol) of less than 18, 20, and 25-30 μg/dl, measured after an ACTH stimulation test,[61-65] whereas others have used a cutoff cortisol increment of less than 9 μg/dl after ACTH administration.[10,57,62,65,66] This incremental response of 9 μg/dl approach has been criticized by some investigators because critically ill patients are maximally stressed with already maximum cortisol release, so any further increase is expected to be a measure of adrenal reserve and not adrenal insufficiency. For example, a critically ill patient with a baseline cortisol level of 40 μg/dl that increases to 45 μg/dl lacks sufficient adrenal reserve but is not adrenally insufficient, according to this school of thought. In other words, it is the absolute level that is important and not the change in cortisol level.
One group suggested that adrenal insufficiency is unlikely when a random cortisol measurement is greater than 34 μg/dl and likely if the measurement is less than 15 μg/dl during critical illness. For cortisol levels between 15 and 34 μg/dl, they suggest performing the ACTH stimulation test; an incremental response less than 9 μg/dl is indicative of adrenal insufficiency (Figure 2), and the administration of a low-dose corticosteroid may be helpful. The author of an editorial added to the controversy by recommending the use of a peak cortisol concentration of less than 20 μg/dl or an incremental response of less than 9 μg/dl after the high-dose ACTH stimulation test to define adrenal insufficiency in critically ill patients. The diagnosis of adrenal insufficiency using the ACTH stimulation test is less problematic in the absence of acute illness.
In a landmark study, the authors identified three groups of patient prognoses by using the ACTH stimulation test. They determined that patients with septic shock with a baseline cortisol level of greater than 34 μg/dl and an increase of 9 μg/dl or less after the ACTH stimulation test are associated with the highest mortality, followed by those with baseline levels greater than 34 μg/dl with an increase of greater than 9 μg/dl. The lowest mortality risk was found in those with baseline cortisol levels of 34 μg/dl or lower and an increase of greater than 9 μg/dl. In other words, the higher the basal plasma cortisol level and weaker the response to the ACTH stimulation test, the higher the mortality rate ( Table 6 ). In another study, the authors reported that the breakpoint for a 50% mortality rate in patients with septic shock was a change in cortisol of 13.4 μg/dl, and they concluded that the lower the response, the greater the mortality rate. In a third study, there was 100% mortality in patients with a cortisol response of 9.1 μg/dl or less and only 32% mortality if the response was greater than 9.1 μg/dl. These studies support the hypothesis that the magnitude of change after the ACTH stimulation test may have a predictive value with regard to mortality.
It is the opinion of this author, as well as many others, that the strongest evidence supports a basal cortisol level of less than 15 μg/dl with an increase of less than 9 μg/dl after administration of the ACTH stimulation test[57,63,64] or a random cortisol level of less than 25 μg/dl[25,31,56] as the best thresholds for diagnosing adrenal insufficiency, whereas the basal cortisol level of 34 μg/dl and an incremental response of 9 μg/dl is the best cutoff to discriminate between survivors and nonsurvivors of septic shock.
Serum Free-Cortisol Levels
Because more than 90% of circulating cortisol is bound to protein, changes in protein binding can affect total serum cortisol concentrations without affecting free-cortisol concentrations. The effect of a decrease in serum proteins (corticosteroid-binding globulin and albumin) on measured total cortisol has been realized by studies in patients with trauma or sepsis and those undergoing major surgery. These studies recommend the use of a correction factor known as the free-cortisol index (defined as the total cortisol concentration divided by corticosteroidbinding globulin concentration) as a surrogate marker for serum cortisol concentration. In critically ill patients, some evidence suggests that there is a decrease in cortisol binding and an increase in free cortisol.[18,19]
In a recent study, baseline serum cortisol, ACTH-stimulated total cortisol, aldosterone, and free-cortisol concentrations were measured in 66 critically ill patients and 33 healthy subjects. The critically ill patients were divided into two groups, with the healthy subjects serving as the control group. Group 1 was made up of patients with albumin levels of 2.5 g/dl or less, and group 2 patients had albumin levels above 2.5 g/dl. The two study groups had similar characteristics except for their serum albumin, total protein, and corticosteroid-binding globulin concentrations. The high-dose ACTH stimulation tests were performed in the two study groups and the control group. Both the baseline and ACTH-stimulated total cortisol concentrations were found to be lower in group 1 (low-albumin group) compared with group 2 (normal-albumin group) and the control group. The free-cortisol level, however, in groups 1 and 2 was much higher than that in the control group. Baseline and ACTH-stimulated serum aldosterone concentrations were similar in all groups.
The authors concluded that in critically ill patients with hypoalbuminemia, although the total serum cortisol level may be low, their freecortisol level (the biologically active part) may be normal, and hence, their adrenal function may be intact. In other words, the interpretation of baseline and total cortisol levels after ACTH stimulation should be done with respect to a patient's protein status since hypoproteinemia results in lower concentrations of total cortisol than expected in critically ill patients. These data suggest that measurement of a serum free-cortisol level is a better indicator of adrenal function and may prevent the unnecessary use of glucocorticoid therapy.
The authors go further to recommend a threshold of less than 2 μg/dl at baseline and less than 3.1 μg/dl after the ACTH stimulation test for identifying patients at risk for adrenal insufficiency. This study has been criticized because, although 18 of the 66 patients had sepsis, none had septic shock or multiple organ failure, a rather significant oversight. Also, the technique used for measuring free cortisol is technically difficult, likely to be expensive, and not widely available. Furthermore, there are no published, randomized, controlled studies that have determined normal free-cortisol concentrations in critically ill patients. Finally, the authors did not correlate the measured serum free-cortisol value with the free-cortisol index to determine the compatibility of the two measurements.
Several experts, including the authors of the above study,[19,68] have proposed using total cortisol concentrations until more data and the measurement of free cortisol become widely available.
Other Diagnostic Tests
Insulin Tolerance Test. The insulin tolerance test measures a patient's response to insulin-induced hypoglycemia. It is considered by many to be the gold standard because it tests the integrity of the entire HPAaxis system.[27,41] Insulin is administered to the patient until the serum glucose level decreases to below 40 mg/dl, causing severe hypoglycemia. Hypoglycemia, a potent stressor, then causes the release of ACTH and, hence, cortisol. It has been proposed by others that glucose levels below 30 mg/dl improve the specificity of the test. Usually the dose of insulin administered is 0.1 U/kg. In obese patients with insulin resistance, the dose should be increased to 0.15 U/kg. Cortisol levels are checked at baseline, as well as after insulin administration every 15 minutes over the subsequent 2 hours. A peak plasma cortisol level of less than 18 μg/dl or less than 20 μg/dl at any time during the test is indicative of adrenal insufficiency. The use of the higher cutoff point is preferred because it minimizes underdiagnosis of adrenal insufficiency. An abnormal result implies a problem that can be occurring anywhere between the hypothalamus and the adrenal gland. The insulin tolerance test is labor intensive, requires constant medical supervision, and lacks a clear cutoff level. In addition, it is contraindicated in patients older than 60 years and in those with ischemic heart disease, seizures, or severe cortisol deficiency (cortisol level < 7 μg/dl at 9 A.M.). Because of the small but significant risk of seizures, coma, or precipitation of angina, use of the insulin tolerance test is not safe in the unconscious critically ill patient; thus, it is not commonly used clinically in the ICU.
This is another well-established test for evaluating the HPA axis. Metyrapone, a drug that interferes with cortisol synthesis, is administered to the patient. The reduction of cortisol concentration will provoke an increase in cortisol levels in patients with normal adrenal function. Patients with adrenal insufficiency will not have the compensatory increase in cortisol. In other words, the test measures the ability of the HPA axis to respond to an acute reduction in serum cortisol. Metyrapone 30 mg/kg (maximum 3000 mg) is administered at midnight, and blood samples are drawn in the early morning for cortisol and its precursor, 11-deoxycortisol. In response to metyrapone administration, cortisol should decrease to less than 5 μg/dl and 11- deoxycortisol should increase to greater than 7 μg/dl in patients with adrenal insufficiency. It has been proposed that the sum of cortisol and 11-deoxycortisol in the early morning should be greater than 16.5 μg/dl. Comparison of the metyrapone test, insulin tolerance test, and ACTH stimulation test in patients with suspected HPA-axis suppression (resulting from previous use of corticosteroid or pituitary surgery) showed that the metyrapone test was better in detecting subtle abnormalities of the HPA axis. However, because of the time involved, the risk of further reducing potentially low cortisol levels, and the intermittent availability of metyrapone, this test is rarely used clinically in the ICU.
Corticotropin-Releasing Hormone Stimulation Test
This is a relatively new test. It measures the ability of the pituitary gland to secrete ACTH in response to exogenous CRH, with a subsequent increase in cortisol secretion by the adrenal gland. Impaired response to CRH is associated with increased mortality.[11,74] Although this test bypasses the hypothalamus, it does test the integrity of the pituitary and adrenal glands. Serum cortisol levels are measured at baseline, and after injection of 100 μg or 1 μg/kg of intravenous CRH, serum cortisol and ACTH levels are measured at 15, 30, 45, and 60 minutes. The serum ACTH level usually peaks at 15 or 30 minutes, whereas the cortisol level peaks at 30 or 45 minutes. The proposed cutoff for this test is a cortisol level of 18.5 or 20 μg/dl with a specificity as low as 33%. Some authors have defined a normal response as an increase in cortisol value of 20% or greater from baseline. This test can also be used to distinguish between ACTH deficiency and deficiency in CRH. The CRH stimulation test has been shown to be comparable to the insulin tolerance test in diagnosing adrenal insufficiency. The limitations of the CRH stimulation test are the lack of studies and standardization in critically ill patients and its high cost.
The use of high-dose glucocorticoids was generally acceptable in the 1970s and early 1980s. With regard to sepsis or shock, high doses of glucocorticoids were defined as methylprednisolone 30 mg/kg/day or more (or its equivalent) for a short duration (1-2 days).[79,80] Recently, there has been a consensus that such doses are not beneficial and may even be harmful.[80,81] Two meta-analyses[79,80] showed that use of glucocorticoids provides no overall benefit in the treatment of septic shock, with one of the meta-analyses suggesting that it may be harmful. The studies included in these metaanalyses used high-dose glucocorticoids for a short period of time. In fact, one meta-analysis showed no overall beneficial effect and no difference between low-dose and high-dose glucocorticoids. The authors defined low-dose glucocorticoids as less than 20 g of hydrocortisone or its equivalent. High-dose glucocorticoids have been associated with increased secondary infections, increased mortality, and increased occurrence of renal and hepatic dysfunction in several studies.
However, two small placebo-controlled trials showed that low-dose hydrocortisone (200-300 mg/day) over a prolonged period (≥ 5 days) significantly decreased the time for vasopressor withdrawal in patients with septic shock, with little effect on mortality.[83,84] In one study, the authors compared hydrocortisone therapy and placebo in patients with septic shock, with a primary end point of shock reversal and a secondary end point of all-cause mortality. The authors defined reversal of shock as systolic blood pressure of 90 mm Hg or higher for at least 24 hours without vasopressors or fluid boluses. Twenty-two patients were randomly assigned to receive hydrocortisone treatment and 19 patients received placebo. The treatment group was noted to have greater reversal of shock at 7 days (p=0.007) and at 28 days (p=0.005) after initiation of therapy. Compared with the placebo group, the median time for cessation of vasopressors was dramatically decreased in the hydrocortisone group (4 vs 13 days). There was also a nonsignificant trend in reduction of 28-day all-cause mortality in the treatment group. In a large, randomized, placebo-controlled, double-blind, parallel-group trial in 19 ICUs in France, patients with septic shock were randomly assigned to receive either placebo (149 patients) or hydrocortisone 50 mg intravenously every 6 hours plus fludrocortisone 50 μg once/day through a nasogastric tube for 7 days (151 patients). Therapy was begun within 8 hours after onset of septic shock. The main outcome measure was 28-day survival in patients with adrenal insufficiency, defined as nonresponders to the ACTH stimulation test (an increase of < 9 μg/dl at 30 or 60 min after administration of the test). Vasopressor therapy was withdrawn within 28 days in 46 patients (40% of 115 nonresponders) in the placebo group and 65 patients (57% of 114 nonresponders ) in the corticosteroid group (hazard ratio 1.91, 95% confidence interval 1.29-2.84, p=0.001). There was a 30% decrease in mortality in the corticosteroid-treated patients. The adverse events rates were similar between the two groups. In addition, the study showed that in patients with septic shock who had adrenal insufficiency, one additional life can be saved at day 28 for every seven patients treated with corticosteroids. The authors concluded that low doses of hydrocortisone and fludrocortisone significantly reduced 28-day mortality and the duration of vasopressor administration in all patients with septic shock, particularly those with adrenal insufficiency, with no increase in adverse effects. For the first time, this study showed a significant reduction in 28-day mortality with the use of low-dose hydrocortisone. It must be noted that patients in this study had more severe septic shock (systolic blood pressure < 90 mm Hg while still receiving vasopressor therapy) compared with other studies (systolic blood pressure could be > 90 mm Hg).
The efficacy of low-dose corticosteroids in patients with less severe sepsis, septic shock, or late sepsis has been questioned by the soon to be published Corticosteroid Therapy in Septic Shock (CORTICUS) study. This was an international, multicenter, double-blind, randomized, controlled trial of corticosteroids for the treatment of sepsis with a primary end point of 28-day all-cause mortality in nonresponders (defined as a change of ≤ 9 μg/dl in cortisol level after the high-dose ACTH stimulation test). Secondary end points included mortality in the entire population, organ failure resolution, and safety. The study was designed to enroll 800 patients in order to have enough power to detect a 10% difference in mortality. However, the study was stopped after 500 patients were enrolled because of difficulty recruiting patients. Patients in the corticosteroid arm received hydrocortisone (without fludrocortisone) 50 mg every 6 hours for 5 days with the dosage tapered over the next 6 days. All-cause mortality was similar between the two arms (34% corticosteroids vs 31% placebo). Although not statistically significant, the rates of shock reversal appeared better in the corticosteroid group. This study differs from the above-mentioned study in several ways. First the patients were not as severely ill; second, corticosteroid therapy was started within 72 hours versus 8 hours; and third, fludrocortisone was not administered. Although underpowered, this study has raised questions about the use of corticosteroids in late sepsis and in less severely ill patients. In patients with septic shock, the benefits of corticosteroids, given within 24 or 48 hours, remain to be shown. Needless to say, it might be reasonable to start low-dose corticosteroid therapy as early as possible in patients with severe septic shock, especially when overwhelming inflammation predominates.
Overall, the data suggest that the use of lowdose corticosteroids for a longer period of time shortens the time receiving vasopressor therapy in severely ill patients with septic shock. In addition, none of the newly published clinical trials suggests an increase in mortality in patients with septic shock after low-dose glucocorticoids (i.e., hydrocortisone 200-300 mg/day). It must be emphasized that low-dose corticosteroids should not be used for the treatment of sepsis in the absence of shock, although there is no contraindication to continuing maintenance therapy for other conditions. Low-dose corticosteroid replacement may also be beneficial in patients with trauma, burns, and surgical and medical conditions in which there is evidence of adrenal insufficiency.
Intravenous hydrocortisone, methylprednisolone, and dexamethasone are the three glucocorticoids most commonly administered to critically ill patients with adrenal insufficiency due to stress. To this author's knowledge, no comparative study of the different corticosteroids has been performed in critically ill patients; however, hydrocortisone is usually the preferred agent because it is the synthetic equivalent of cortisol, directly replacing cortisol with no metabolism required. Furthermore, it has both glucocorticoid and mineralocorticoid activities (unlike dexamethasone and methylprednisolone) and has been most extensively studied.[57,83,84,88,89] How essential the addition of a mineralocorticoid to the glucocorticoid remains controversial. Although one can argue that mineralocorticoid activity is necessary for possible absolute adrenal insufficiency, this condition is quite rare in septic shock (reported frequency of 0-3%).[10,90] Others have argued that the dose of hydrocortisone used has enough mineralocorticoid activity, and therefore fludrocortisone is not needed even when the patient has primary adrenal insufficiency.[27,55] It is estimated that 20 mg of hydrocortisone is equivalent to 0.05 mg of fludrocortisone with regard to mineralocorticoid potency, and hence, fludrocortisone is only needed when the daily hydrocortisone dose is less than 50 mg.
The reader must bear in mind, however, that although most studies showed hemodynamic improvement without adding fludrocortisone,[83,84,88,92] the only study that showed a significant reduction in 28-day mortality used both hydrocortisone and fludrocortisone. Whether the addition of mineralocorticoid accounted for the decrease in mortality is unclear. There has not been a comparative study between hydrocortisone alone and hydrocortisone plus fludrocortisone. Hydrocortisone and prednisone are the preferred glucocorticoids in the pregnant patient with suspected or confirmed adrenal insufficiency since they are readily inactivated by the placenta. In contrast, dexamethasone readily crosses the placenta and suppresses fetal adrenal function.
The recommended dosage of hydrocortisone is 200-300 mg/day in three or four divided doses of 50 mg every 6 hours or 100 mg every 8 hours, or as a 50-100-mg bolus over 30 minutes followed by a 10-mg/hour continuous infusion. Although a comparative study of intermittent versus continuous infusion has not been performed, hemodynamic improvement to a similar extent has been reported with both regimens.[57,84,88,92,94] If needed, the dose of fludrocortisone supported by the literature is 0.5 mg/day, with a recommended range of 0.5-2 mg/day. If the ACTH stimulation test cannot be performed immediately for some reason in a patient with suspected adrenal insufficiency, it is recommended to begin equivalent doses of dexamethasone until the test can be performed. Dexamethasone does not cross-react with the cortisol assay and will not interfere with the measurement of cortisol in response to the ACTH stimulation test,[2,12] and it should be continued until the test can be performed. When the result is available, therapy may be continued in nonresponders and discontinued in responders. One group of authors observed that hemodynamic improvement often occurred within 24 hours of glucocorticoid administration in patients with septic shock. As a patient improves clinically, discontinuation of the corticosteroid should be considered.
An important question is whether glucocorticoids should be tapered after resolution of shock. Although use of low-dose glucocorticoids over 5-7 days is safe, it is recommended, but not required, to taper at the end of therapy by halving the dose every 2-3 days.[30,83] The reader is hereby cautioned that this recommendation is not based on controlled clinical trials. When shock recurs during weaning, reestablishing the original steroid dose might be considered.
While a patient is receiving glucocorticoids, hyperglycemia should be avoided by routinely monitoring blood glucose levels.[95-98] In surgical and medical ICUs, the blood glucose level should probably be maintained below 110 mg/dl[95,96] since it has been shown that tight glycemic control reduces mortality among patients with extended ICU stays.[95,96,98] Concern for hypoglycemic episodes has been the primary reason for the limited use of tight glycemic control (blood glucose levels 80-110 mg/dl) in the critically ill patient. In a recent article, the shortterm consequences of hypoglycemia (blood glucose level < 45 mg/dl) in patients in the ICU were evaluated. The authors found no association between incidental hypoglycemia and mortality. This is consistent with the findings of an earlier study that investigated the association between hypoglycemia and mortality in hospitalized elderly patients. A larger study is needed to confirm the conclusions of these rather small studies. However, the findings of Comparing the Effects of Two Glucose Control Regimens by Insulin in Intensive Care Unit Patients (GLUCONTROL), a recent unpublished study, have regenerated the ongoing debate over the optimal range for glycemic control. This was a multicenter, international study in mixed ICU populations that compared a tight glycemic group (blood glucose level 80-110 mg/dl) with a conventional group (blood glucose level 140-180 mg/dl). The researchers had planned to enroll 3500 patients to determine if a 4% reduction in mortality could be detected. After 1100 patients were recruited, an interim analysis showed no difference in 28-day mortality (16.9% for the tight glycemic group vs 15.2% in the conventional group), neither was there a significant difference in length of stay. However, an alarming increase in hypoglycemic episodes was observed in the tight glycemic group (8.6% for tight glycemic group vs 2.4% for the conventional group). Since the GLUCONTROL study was not completed, it remains to be seen what impact it will have on clinical practice. The ongoing Normoglycemia in Intensive Care Evaluation and Survival Using Glucose Algorithm Regulation (NICE-SUGAR) study, which plans to enroll 6100 patients in 35 ICUs to compare two protocols (81-108 vs 144-180 mg/dl), hopefully will provide some answers to this debate.
Adrenal insufficiency can be life threatening in stressful situations, and its diagnosis is always a challenge, especially in patients with critical illness. Since the condition appears to be common in patients with septic shock, clinicians should have a high index of suspicion for its occurrence in critically ill patients with persistent hypotension despite adequate fluid resuscitation and/or poor hemodynamic response to vasopressors. Careful history taking and physical examination, complemented by other signs such as hyponatremia and hyperkalemia, should help identify most cases. Treatment with physiologic doses of corticosteroids should be started as soon as possible since short-term treatment carries very few risks and has been shown to decrease both morbidity and mortality.
Table 1. Normal Daily Production Rate and Circulation Levels of the Predominant Corticosteroids
Table 2. Relative Potencies and Equivalent Doses of Common Corticosteroids
Table 3. Causes of Adrenal Insufficiency
Table 4. Manifestations of Primary and Secondary Adrenal Insufficiency or Both
Table 5. Summary of Tests for Diagnosing Adrenal Insufficiency
Table 6. Patient Prognosis Based on Cortisol Level at Baseline and Response to ACTH Stimulation Test
The author is deeply indebted to the following individuals from St. Thomas Hospital, Nashville, Tennessee, for their careful review and constructive suggestions in the preparation of the manuscript: Richard Tomicheck, M.D., Kerry L. Butler, Pharm.D., M.B.A., Michael Pepper, Pharm.D., Kim Madewell, Pharm.D., BCPS, and Kimberly Lindsey, Pharm.D. I would also like to thank the following for their help in the preparation of the manuscript: Kathy Anderson, B.S., Jacqueline Kelly, Pharm.D., Jeannie Watson, Pharm.D., and Cathy Turner, Pharm.D.
Kwame Asare, Pharm.D., BCPS, BCNSP, St. Thomas Hospital, Pharmacy Department, 4220 Harding Road, Nashville, TN 37202; e-mail: Kasare@stthomas.org .
Kwame Asare, Pharm.D.
From the Pharmacy Department, St. Thomas Hospital, Nashville, Tennessee.