Glucocorticoid-Induced Adrenal Insufficiency

DR KRASNER: Our patient is a 36-year-old woman. Around 15 years ago, she began suffering recurrent episodes of acute sinusitis, followed by asthma attacks. She was treated repeatedly with bronchodilators, antibiotics, and short courses of glucocorticoids. During this time, she underwent 5 operations for hyperplastic sinus polyps. Two years ago, after a particularly severe asthma attack, her physician prescribed an indefinite course of glucocorticoid therapy. She began with prednisone, 25 mg/d for 2 months, followed by 20 mg/d for 2 more months. Then her medication was changed to triamcinolone, starting at 4 mg/d and the dosage slowly tapered to 2 mg/d.

Can you tell us how you felt when your dose was reduced to 2 mg/d?

PATIENT: I began feeling extremely weak, tired, and mildly dizzy. It wasn't anything I could pinpoint, and I didn't understand it.

DR KRASNER: Your dosage of triamcinolone was then discontinued. How did you feel when you saw your physician 2 weeks later?

PATIENT: I felt even worse. I was so weak that I could barely function. I had stopped the triamcinolone as I had been instructed, so I assumed that wasn't the problem. I tried to explain the symptoms to my doctor, and he ordered blood tests. The next day, I received an emergency phone call from the doctor, saying that my cortisol level was zero, and I had been walking around with no steroids whatsoever for 2 weeks.

DR KRASNER: You were urgently referred to an endocrinologist, who prescribed cortisone acetate, 30 mg, for you to take once a day, in the morning. Please describe what happened after you took your first dose.

PATIENT: Late that afternoon, I again started feeling very, very weak. I could hardly move. I described it to my husband as feeling like I had been hit by a truck. I went to bed at 9 o'clock and started to shake uncontrollably. I became nauseated and very dizzy. I wasn't sure what to do. While I was waiting to hear back from the endocrinologist, I took 25 mg of my old prednisone and felt better almost immediately. I felt that I had saved my own life. When I spoke with the endocrinologist , he told me to stop taking the cortisone and to continue the prednisone, every other day.

DR KRASNER: After this, multiple attempts were made to reduce and stop your dosage of prednisone. Five months ago, when you came to see me for a second opinion, your prednisone dosage had been reduced to 15 mg every other day. I saw you on a nondose day, and you complained of severe weakness. Since then, we have largely controlled your symptoms by giving you a physiologic replacement prednisone dose of 5 mg each morning, while we wait for your pituitary and adrenal glands to reactivate.

One of the adverse effects of long-term glucocorticoid therapy in supraphysiologic doses is suppression of the hypothalamic-pituitary-adrenal (HPA) axis. This suppression can render the adrenal glands unable to generate sufficient cortisol after glucocorticoid treatment is stopped. Most physicians consider adrenal insufficiency such an unusual complication of glucocorticoid therapy that few test for it at the completion of treatment unless the patient becomes ill.

Although the true prevalence of this complication is not known, I suspect that the reason physicians rarely see it is that they have been taught to stop high-dose glucocorticoids gradually. In theory, slow glucocorticoid dosage reduction not only forestalls relapse of the disease being treated, but may also allow gradual recovery of the HPA axis when it has been suppressed. Treatment patterns vary and tend to be guided by empiricism and oral tradition. Most physicians who regularly prescribe glucocorticoids have developed their own slow dosage reduction programs that serve them well.^{1 - 2} No prospective studies have defined optimal methods of dosage reduction.

Unfortunately, today's patient and many others do not seem to tolerate glucocorticoid withdrawal. When patients become ill after a dosage reduction, the physician should consider a differential diagnosis that includes adrenal insufficiency, a flare-up of the disease being treated, and steroid withdrawal syndrome. All 3 conditions resolve after patients are restarted on the glucocorticoid regimen that had previously controlled their symptoms.

The most useful definition of steroid withdrawal syndrome may be the original 1960 description by Amatruda et al³ : a symptom complex resembling true adrenal insufficiency, with nonspecific symptoms like weakness, nausea, and arthralgias, occurring in patients who have finished a dosage reduction of glucocorticoid therapy and who respond normally to HPA axis testing. Little is known about the physiological basis for the syndrome except that it does not predispose patients to adrenal crisis. Elevated circulating concentrations of interleukin 6 may cause many of the symptoms of glucocorticoid withdrawal.⁴ Steroid withdrawal symptoms can develop even if the reduced dosage remains supraphysiologic.⁵

If the dosage is being tapered to below physiologic replacement, direct testing of HPA axis function is required to distinguish among the diagnostic possibilities. This should certainly be done in all symptomatic patients. Patients with normal HPA axis function and steroid withdrawal symptoms need very slow, small dosage reductions to tolerate glucocorticoid withdrawal. They may also be helped by nonsteroidal anti-inflammatory drugs.

Identifying patients with true adrenal insufficiency is crucial because the condition can cause life-threatening hypotension ("adrenal crisis") when the patient is subjected to a major stress like illness, injury, or surgery. Adrenal crisis is entirely preventable if patients are given supplemental glucocorticoids before or early in the course of the stress.

Glucocorticoid treatment may not suppress the HPA axis at all, or it may cause central suppression or complete adrenal gland atrophy (Table 1).

The medical literature has little to say about glucocorticoid-treated patients who do not develop detectable HPA axis suppression. Jasani and colleagues⁶ identified such people in a study of 19 glucocorticoid-treated patients who were subjected to complete adrenal axis testing. Eight of the patients had no evidence of HPA–axis suppression even though they had received an average cumulative prednisone dose of 4.36 g for an average of 3.6 years.

Some glucocorticoid-treated patients develop isolated central adrenal insufficiency—prolonged suppression of the hypothalamic-pituitary unit but normally functioning adrenal glands. Supraphysiologic glucocorticoid doses suppress both corticotropin-releasing hormone (CRH) production in the hypothalamus and corticotropin (ACTH) production in the pituitary gland. This suppression can outlast the duration of the glucocorticoid exposure. Five of the 19 patients in the study by Jasani et al⁶ fell into this category.

The most severely affected glucocorticoid-treated patients have complete HPA axis suppression, characterized by functional adrenal gland atrophy. Cortisol production seems to depend on a tonic steady-state exposure to circulating ACTH. In states of profound or prolonged ACTH deficiency, the adrenal glands may not be able to generate cortisol promptly, even if directly stimulated by exogenous ACTH. In the study by Jasani and colleagues,⁶ 6 of the 19 patients were found to have functional adrenal gland atrophy. Today's patient also progressed to this most advanced state of glucocorticoid-induced HPA axis suppression. This conclusion was reached through biochemical tests.

Because all the symptoms and signs of adrenal insufficiency are nonspecific, confirming the diagnosis requires documentation of either subnormal cortisol levels or cortisol levels that remain inappropriately low despite provocation. It is important to measure cortisol levels when the patient is not receiving exogenous glucocorticoid so that the measurement reflects only endogenous adrenal steroid production. All of the commonly used glucocorticoid preparations except dexamethasone cross-react to some degree in the cortisol assay. Fortunately, the chemical half-lives of all these compounds are on the order of 0.5 to 4.0 hours.⁷ Therefore, endogenous cortisol levels can be measured with confidence 24 to 48 hours after the patient's last glucocorticoid dose. I typically measure cortisol or perform stimulation testing just before the patient's daily prednisone dose or first daily hydrocortisone dose.

A single cortisol measurement can be diagnostic of glucocorticoid-induced adrenal insufficiency if the level is very low (<83 nmol/L [<3 µg/dL]).⁸ This assumes that the blood sample is drawn before 9 AM, just after cortisol secretion has normally peaked for the day. Adrenal insufficiency can be ruled out with confidence if a level taken at any time during the day exceeds 550 nmol/L (20 µg/dL). Unfortunately, cortisol secretion is quite variable and single measurements are often in the nondiagnostic "normal" range even in patients who cannot respond to stimulation with adequate cortisol output.⁸

Most patients require stimulatory tests of HPA axis function to document adrenal insufficiency. A natural provocation to the HPA axis is any severe injury or medical illness, especially if it causes fever and hypotension. If a single cortisol level drawn during such acute stress is below 550 nmol/L (20 µg/dL), the patient is adrenally insufficient. For stable ambulatory patients in whom a single cortisol measurement has not been diagnostic, there are 5 well-established stimulatory tests of HPA axis function.

This is the most commonly used test to diagnose adrenal insufficiency. This test is a direct assessment of adrenal gland responsiveness to exogenous ACTH. Although the test can be done in several ways, the protocol in widespread use measures cortisol concentrations before and 30 to 60 minutes after the rapid injection of a highly supraphysiologic dose (250 µg) of a synthetic ACTH analog. Normal adrenal glands should be able to generate a poststimulation cortisol concentration of 550 nmol/L (20 µg/dL).⁹

Our patient's HPA axis progressed to the most advanced phase of suppression, functional adrenal gland atrophy (Table 1), as evidenced by a subnormal 250-µg ACTH stimulation test result. However, some patients with isolated central HPA axis suppression have not yet developed adrenal atrophy and can still respond normally to this test.¹⁰ If our markedly symptomatic patient had had a normal ACTH stimulation test result, it would still have been important to rule out central adrenal axis insufficiency. The 4 remaining stimulatory tests are more sensitive than 250-µg ACTH stimulation because they depend on a completely intact HPA axis for a normal result.

Traditionally, this has been the definitive way to assess central HPA axis integrity. Insulin is injected into the patient until the serum glucose concentration falls below 2.2 mmol/L (40 mg/dL). Hypoglycemia is among the most potent stimulants of ACTH and hence cortisol production. A hypoglycemic patient with a cortisol concentration below 550 nmol/L (20 µg/dL) has a significant HPA axis defect in either the hypothalamic-pituitary unit or the adrenal glands.

Age older than 60 years and coronary artery disease are relative contraindications to the insulin-induced hypoglycemia test, as is a history of seizures. Because patients become hypoglycemic during the test, they require intensive monitoring. However, the test is safe when performed by experienced personnel. A review of 6581 insulin-induced hypoglycemia tests revealed only 6 adverse events, all of which were quickly reversed with intravenous glucose.¹¹ Although this test is usually considered the "definitive" assessment of the integrity of the HPA axis, it occasionally yields false-negative results.¹² Because it is labor-intensive and entails some risk, physicians and nurses who have experience performing it are becoming few and far between.

This is another well-established method of evaluating the central HPA axis. Metyrapone is a drug that interferes with cortisol synthesis. In people who have a normal HPA axis, metyrapone-induced lowering of cortisol concentrations provokes a compensatory rise in ACTH secretion. Patients with central adrenal insufficiency do not have this rise in ACTH. Some physicians recommend hospitalization for the metyrapone test because it can induce acute adrenal crisis.¹⁰

This is 1 of 2 newer tests. The CRH stimulation test has been shown to be diagnostically comparable to the insulin-induced hypoglycemia test.⁸ Corticotropin-releasing hormone is now commercially available and safe for outpatient evaluations. For these reasons, it is currently my preferred approach for patients who need to be tested for central adrenal insufficiency. The disadvantage of CRH is that it is very expensive.

The other newer test, 1-µg ACTH stimulation, not only measures adrenal gland responsiveness but may be able to detect subtle degrees of adrenal atrophy caused by central adrenal insufficiency.^{12 - 13} Because ACTH is not commercially available in 1-µg doses, it must be diluted before being infused. Although this test is safe, fast, and relatively inexpensive, no lower limit of normal response is universally accepted.¹⁴ Furthermore, not all studies concur that the 1-µg test is more sensitive than the standard 250-µg ACTH stimulation test.^{15 - 16}

Some of the risk factors for HPA axis suppression are clearly defined; others are less certain.¹⁷ The choice of steroid preparation, for example, is known to affect risk for HPA axis suppression. Prednisone, methylprednisolone, and dexamethasone are all synthetic analogs of cortisol (hydrocortisone). Cortisone is a naturally occurring steroid metabolite that, when given systemically, is converted into cortisol by the liver. Although hydrocortisone and cortisone acetate are commonly used for adrenal replacement therapy, the synthetic analogs are much better tolerated as anti-inflammatory agents because they cause significantly less sodium retention at supraphysiologic doses.¹⁸

These agents differ in their anti-inflammatory potency. Glucocorticoid potency is known to correlate with risk for adrenal insufficiency.^{7 ,18} The anti-inflammatory potency and duration of biological effect of glucocorticoid preparations roughly correlate with duration of ACTH suppression after a single dose (Table 2).^{18 ,20} Hydrocortisone (cortisol) and cortisone acetate are the least potent and therefore least suppressive agents. Prednisone, prednisolone, methylprednisolone, and triamcinolone are moderately suppressive. Of commonly used glucocorticoids, dexamethasone suppresses ACTH the longest. When glucocorticoids are given at a frequency that does not allow the HPA axis to recover between doses, patients are at higher risk for HPA axis suppression.

Systemic glucocorticoid therapy is more likely to suppress the HPA axis than are intra-articular, inhalational, or topical glucocorticoids.⁷ Adrenal axis suppression due to high-dose local therapies has been described but is uncommon. There is evidence that patients are at lower risk for adrenal insufficiency if they can take glucocorticoids on alternate days from the outset or if they can convert to alternate-day therapy before the HPA axis is suppressed.²¹

Evening doses of glucocorticoid tend to suppress the normal early-morning surge of ACTH secretion.²² Whenever possible, therefore, patients should take their glucocorticoids all at once just after waking; this is generally after the daily peak of ACTH secretion. Once-a-day dosing is usually feasible for prednisone, triamcinolone, and dexamethasone. Most patients take the short-acting drugs, hydrocortisone and cortisone acetate, twice a day, on waking and at around 5 PM. The second dose is taken at dinnertime rather than bedtime to keep it from suppressing the next morning's ACTH surge. To mimic normal diurnal cortisol rhythms, the morning dose is two thirds, and the dinnertime dose, one third of the total daily dose.

Traditionally, the duration of glucocorticoid therapy and the cumulative dose of glucocorticoid received have been viewed as predictive of the likelihood of HPA axis suppression. Such predictions may not be reliable. In their 1967 study, Jasani and colleagues⁶ compared the lengths of time and cumulative doses of glucocorticoids that their 19 patients had received. The investigators found a striking degree of overlap. The 5 patients who developed isolated central adrenal insufficiency had been taking steroids for anywhere from 2 to 8 years. Six patients with advanced adrenal atrophy had been taking steroids for 1 to 10 years, and 8 patients with no HPA axis suppression at all had been treated for 1 to 7 years. The results of this and a more recent study⁸ suggest that duration of therapy and cumulative dose only roughly predict HPA axis suppression.

Axelrod⁷ suggests that any patient who has received the equivalent of 20 to 30 mg/d of prednisone for longer than 5 days may be at risk for adrenal insufficiency. Although the HPA axis can be suppressed by very brief treatment, clinically evident adrenal insufficiency is probably extremely rare in patients who are treated for 1 week or less.²³ It is also possible that if such short courses do suppress the HPA axis, the suppression lasts for only a few days.²⁴

Perhaps the best predictor of HPA axis suppression is simply the patient's current glucocorticoid dosage. A strong correlation has been found between prednisone maintenance doses above 5 mg/d and a subnormal ACTH-stimulation test result.²⁵

This finding calls into question the traditional prednisone "physiologic replacement dose," which for adults was thought to be 7.5 mg/d. Some adults develop Cushing syndrome when treated with 7.5 mg/d of prednisone or equipotent doses of dexamethasone (0.75 mg/d), hydrocortisone (30 mg/d), or cortisone acetate (37.5 mg/d). Furthermore, it has long been known that prednisone doses of around 5 mg/d are of therapeutic value in inflammatory diseases such as rheumatoid arthritis.²⁶ The association of low-dose glucocorticoid therapy with accelerated bone loss^{27 - 28} also suggests that our traditional "low" doses may be more than patients need simply to replace physiologic glucocorticoid output from the adrenal glands. Endogenous cortisol production normally averages 10 mg/d.²⁹ Endogenous production, however, does not translate directly into exogenous glucocorticoid doses because of differences in bioavailability, protein binding, and the kinetics of glucocorticoid clearance.³⁰

The life-threatening manifestation of adrenal insufficiency is stress-induced hypotensive crisis. The best-studied model of stress is surgery. The risk for perioperative hypotensive crisis related to glucocorticoid treatment was discussed in 37 case reports, 3 of which documented adrenal crisis as the cause.³¹

Glucocorticoid-induced hypotensive crisis can be prevented with empiric exogenous glucocorticoid "coverage," which some physicians give to all patients who are receiving glucocorticoid treatment and others give only to those known to be adrenally suppressed. Universal coverage is by far more common, because intensive short (<48 hours) glucocorticoid courses rarely cause significant complications.¹⁸

Because of evidence that standard "stress doses" are excessive, some authors have suggested individualizing perioperative glucocorticoid management. How much cortisol do patients with normal HPA axes produce during surgical stress? Salem and colleagues³¹ reviewed studies that measured cortisol production rates in response to operations of various intensities. They found that patients who undergo minor operations lasting less than 1 hour probably generate, at most, 50 mg/d of cortisol in response to the surgery. Patients undergoing moderate to major operations produce 75 to 200 mg/d of cortisol, which is still significantly less than the 300 mg/d stress doses that patients traditionally have been given.

Salem and colleagues' article recommends that when planning perioperative glucocorticoid coverage for patients with known adrenal suppression, doses should be based on the intensity of the operation. For minor operations like inguinal hernia repair, the authors suggest giving adrenally suppressed patients 25 mg of hydrocortisone equivalent. Moderate operations, such as nonlaparoscopic cholecystectomy, lower extremity revascularization, total joint replacement, or abdominal hysterectomy, can be covered with 50 to 75 mg of hydrocortisone equivalent for 1 to 2 days. Major operations, such as a Whipple operation, total colectomy, or cardiac surgery with cardiopulmonary bypass, should be covered with 100 to 150 mg of hydrocortisone equivalent for 2 to 3 days. Perioperative coverage should begin on call to the operating room and can be continued in divided doses at regular intervals or as a continuous infusion until the patient is in recovery.

The recommendations of Salem and colleagues are not based on controlled clinical trials. It is possible that even these low doses are excessive. A small recent study showed that simply giving prednisone maintenance doses on the day of surgery appeared adequate for preventing hypotension.³²

The major drawback to empiric perioperative therapy is that many patients are treated unnecessarily. Can we use preoperative endocrine tests to select the patients who need steroid coverage? The study that best examined this issue suggested that a subnormal 250-µg preoperative ACTH stimulation test may best predict surgery-induced adrenal crisis,^{33 - 34} but also showed that this complication is fairly rare, even in patients with functionally atrophic adrenal glands.

Although most glucocorticoid-treated patients do not need perioperative coverage, supplementing all such patients is simpler and probably more feasible and cost-effective than performing preoperative endocrine tests on all patients to find the minority who need supplements. Furthermore, empiric coverage eliminates risk for perioperative adrenal crisis.

Finally, let us return to the issue facing our patient today: How do we withdraw glucocorticoids from people who are known to have adrenal insufficiency? Again, the guidance given in the literature is not based on controlled clinical trials.^{1 - 2}

A plan for stopping glucocorticoid therapy should be discussed with patients when they start treatment. The rate of dosage reduction from supraphysiologic anti-inflammatory doses to physiologic replacement doses (5 mg/d of prednisone or the equivalent) depends primarily on the activity of the disease being treated. Once it is established that the disease is quiescent and patients are receiving stable replacement glucocorticoid dosages, many physicians simply continue to taper the dosage until the drug is stopped. If the dosage reduction is done this way, without HPA axis testing, the physician should observe patients to document that they have no symptoms or signs of adrenal insufficiency. Such patients should be alerted that they may need to take glucocorticoid supplements at times of illness or injury for up to 1 year after stopping treatment.

Patients should undergo endocrine testing if they develop symptoms of glucocorticoid deficiency as their dosage is tapered below the physiologic replacement range. If a morning cortisol level is not diagnostically low, patients need a stimulation test. The choice of tests is debatable, but I believe that it makes the most sense to perform a sensitive test of central HPA axis function such as CRH stimulation, because a symptomatic patient would need this or a similar test even if the 250-µg ACTH stimulation test results were normal.¹⁰

When endocrine testing shows that glucocorticoid-treated patients have adrenal suppression, the same precautions should be taken as for patients with permanent forms of adrenal insufficiency such as naturally occurring Addison disease. Such patients should receive physiologic replacement glucocorticoid doses and be instructed to double or triple the dose if they suffer a moderate illness or injury. A patient who is under severe stress or who cannot take medicines by mouth should receive parenteral glucocorticoids, the first dose of which can be taken at home before the patient is brought to the emergency department. Many patients with adrenal insufficiency are taught to self-inject 100 mg of hydrocortisone or 4 mg of dexamethasone at home in emergency situations. Patients with adrenal insufficiency of any cause should wear an identification tag or bracelet.

Once a patient's dosage has been tapered to a prednisone dosage of 5 mg/d, many physicians successfully finish the tapering process by reducing the dose slowly, for example, by 1 mg per month. An alternative approach is to convert patients from these low prednisone doses to equipotent doses of hydrocortisone.² For example, a regimen of 5 mg/d of prednisone could be converted to hydrocortisone, 15 mg in the morning and 5 mg at dinnertime. This dosage can then be tapered by 2.5 mg/wk, as tolerated, until patients are taking only a single morning dose of 10 mg. Patients can remain taking this dosage until the HPA axis reactivates. Although it has not been proven, hydrocortisone's shorter duration of action may speed recovery of the HPA axis. Small observational studies suggest that once a patient has adrenal suppression from exogenous glucocorticoids, changing to alternate-day therapy does not speed axis recovery.³⁵ There is also no evidence that exogenous ACTH treatment hastens recovery of the HPA axis.³⁶

Graber and colleagues³⁷ showed that spontaneous recovery of the HPA axis from glucocorticoid-induced suppression can take longer than 9 months. A useful way to monitor for HPA axis reactivation is to measure morning cortisol concentrations each month.² The blood must be obtained before patients take their morning drug dose so that the glucocorticoid being measured is endogenous rather than exogenous. When the cortisol concentration exceeds 280 nmol/L (10 µg/dL), the 250-µg ACTH stimulation test should be performed. Because the hypothalamic-pituitary unit reactivates before the adrenal glands themselves, a normal test result implies that the entire axis has recovered and the patient can stop the glucocorticoid replacement regimen.