McKenzie R, O'Fallon A, Dale J, Demitrack M, Sharma G, Deloria M, Garcia-Borreguero D, Blackwelder W, Straus SE. Low-Dose Hydrocortisone for Treatment of Chronic Fatigue SyndromeA Randomized Controlled Trial. JAMA. 1998;280(12):1061-1066. doi:10.1001/jama.280.12.1061
From the Laboratory of Clinical Investigation (Drs McKenzie, Sharma, and Straus and Mss O'Fallon and Dale) and Division of Microbiology and Infectious Diseases (Ms Deloria and Dr Blackwelder), National Institute of Allergy and Infectious Diseases, and Clinical Psychobiology Branch, National Institute of Mental Health (Dr Garcia-Borreguero), National Institutes of Health, Bethesda, Md; and Department of Psychiatry, University of Michigan, Ann Arbor (Dr Demitrack).
Context.— Chronic fatigue syndrome (CFS) is associated with a dysregulated hypothalamic–pituitary
adrenal axis and hypocortisolemia.
Objective.— To evaluate the efficacy and safety of low-dose oral hydrocortisone
as a treatment for CFS.
Design.— A randomized, placebo-controlled, double-blind therapeutic trial, conducted
between 1992 and 1996.
Setting.— A single-center study in a tertiary care research institution.
Patients.— A total of 56 women and 14 men aged 18 to 55 years who met the 1988
Centers for Disease Control and Prevention case criteria for CFS and who withheld
concomitant treatment with other medications.
Intervention.— Oral hydrocortisone, 13 mg/m2 of body surface area every
morning and 3 mg/m2 every afternoon, or placebo, for approximately
Main Outcome Measures.— A global Wellness scale and other self-rating instruments were completed
repeatedly before and during treatment. Resting and cosyntropin-stimulated
cortisol levels were obtained before and at the end of treatment. Patients
recorded adverse effects on a checklist.
Results.— The number of patients showing improvement on the Wellness scale was
19 (54.3%) of 35 placebo recipients vs 20 (66.7%) of 30 hydrocortisone recipients
(P=.31). Hydrocortisone recipients had a greater
improvement in mean Wellness score (6.3 vs 1.7 points; P=.06), a greater percentage (53% vs 29%; P=.04)
recording an improvement of 5 or more points in Wellness score, and a higher
average improvement in Wellness score on more days than did placebo recipients
(P<.001). Statistical evidence of improvement
was not seen with other self-rating scales. Although adverse symptoms reported
by patients taking hydrocortisone were mild, suppression of adrenal glucocorticoid
responsiveness was documented in 12 patients who received it vs none in the
placebo group (P<.001).
Conclusions.— Although hydrocortisone treatment was associated with some improvement
in symptoms of CFS, the degree of adrenal suppression precludes its practical
use for CFS.
CHRONIC FATIGUE syndrome (CFS) has no established cause or proven treatment.1 It is a symptom complex that has attracted considerable
attention in recent years but, on reflection, is obviously not new. In 1988,
the Centers for Disease Control and Prevention (CDC) proposed the term chronic fatigue syndrome and a case definition requiring
debilitating fatigue and 8 or more of 11 signs and symptoms for at least 6
months.2 This definition was refined and simplified
in 1994 by an international working group.3
It now requires severe, unexplained fatigue for more than 6 months that is
of a new or definite onset, not due to continuing exertion, not resolved by
rest, functionally impairing, and accompanied by 4 or more of the following
8 new symptoms: memory or concentration complaints, sore throat, tender lymph
nodes, muscle pain, multijoint pain, a new pattern of headaches, unrefreshing
sleep, and postexertional malaise lasting more than 24 hours.
Depending on the case definition applied, and how the data are acquired,
the point prevalence of CFS ranges from under 0.1% to 2.6% in the primary
care setting.4- 6
Most authorities urge a limited workup for CFS, judicious use of medication
to ameliorate symptoms, graded exercise, and other rehabilitative measures.7,8
Efforts to identify causative factors in CFS are evolving as hypotheses
are generated and tested. In the early 1980s it was speculated that persisting
Epstein-Barr virus infection sustains the symptoms of CFS.9
That notion was rejected after careful epidemiologic and virologic studies
and the negative outcome of a placebo-controlled trial of acyclovir.10,11 While a viral origin of CFS now seems
remote, other hypotheses continue to be examined. A substantial body of work,
for example, indicates subtle alterations in immune function affecting the
numbers and activity of natural killer cells and various T-cell populations.12- 14
For several years, we have pursued the hypothesis that CFS arises or
is maintained by a derangement of the hypothalamic–pituitary adrenal
(HPA) axis. We were led to consider this possibility because depression is
a prominent feature of CFS and patients with major depression are reported
to demonstrate central nervous system–mediated activation of the HPA
We were surprised, however, to find evidence of HPA axis inactivation in CFS
patients.15 While plasma cortisol levels were
in the normal range, responses to cosyntropin and to ovine corticotropin-releasing
hormone were blunted and the CFS patients excreted, on average, about 30%
less cortisol in 24-hour urine collections than healthy, matched controls.
Similar findings were made subsequently in other patient groups with atypical
depressive features, fatigue, and somatic complaints such as fibromyalgia.19- 21 It seemed appropriate,
then, to determine whether CFS symptoms could be ameliorated through cautious
hormonal supplementation to approximate normal levels and diurnal changes
in cortisol levels. These are the results of our randomized, double-blind,
placebo-controlled trial of low-dose hydrocortisone therapy for CFS.
Men and women aged 18 to 55 years who met the CDC 1988 criteria for
CFS,2 all of whom would also meet the more
liberal 1994 criteria,3 were eligible for enrollment
in this trial, with the added provisos that their illness began over a period
of 6 weeks or less, and that they had no contraindications to systemic steroids,
such as a history of peptic ulcer disease, gastritis, hypertension, glaucoma,
diabetes, or evidence of untreated tuberculosis. Patients could not have any
other acute or chronic medical or psychiatric condition that required ongoing
or intermittent medication. In addition, for 2 to 6 weeks before enrollment
(depending on the drug) and for the duration of the study, most over-the-counter
and prescription drugs except acetaminophen were proscribed. Women needed
to practice effective means of birth control and to have a negative pregnancy
test at enrollment.
The diagnosis of CFS was ascertained by patient history, routine physical
examination, and laboratory tests to exclude other relevant diagnoses, as
recommended.3,7 The computerized
Diagnostic Interview Schedule was used for every patient to determine lifetime
and current psychiatric diagnoses based on Diagnostic and
Statistical Manual of Mental Disorders, Revised Third Edition (DSM-III-R) criteria.22 Active depression that
was of such severity as to warrant treatment precluded enrollment.
This was a randomized, placebo-controlled, double-blind trial of oral
low-dose hydrocortisone. All subjects provided written, informed consent to
participate in a research study whose design was approved by the Institutional
Review Board of the National Institute of Allergy and Infectious Diseases.
Patients were admitted to the National Institutes of Health Warren Grant Magnuson
Clinical Center twice, once for confirmation of study eligibility and initiation
of treatment and again on completion of treatment. They were instructed to
take placebo or hydrocortisone pills, equivalent to about 16 mg/m2
of body surface area per day, 20 to 30 mg every morning at about 8 AM, and
5 mg every day at about 2 PM, for 12 weeks. This dosage was designed to approximate
normal daily cortisol levels and their diurnal variation. Based on our prior
controlled studies, this represented about a 30% net daily increase in CFS
patient exposure to cortisol.15 Patient symptoms
were monitored using packets of self-rating forms.
Patients maintained records weekly for 12 weeks of 21 listed adverse
symptoms that might be attributable to corticosteroid treatment. Biochemical
evidence of steroid effects was also sought. Before administration of the
study drug and at the end of treatment, tests were done to evaluate function
of the HPA axis, including morning serum cortisol levels, and measurement
of serum cortisol levels 1 hour after injection of cosyntropin.
Patients were instructed to request stress doses of corticosteroids
should an emergency arise, and all subjects were provided Medic alert bracelets
indicating this fact. Moreover, because of the theoretical possibility that
12 weeks of low-dose hydrocortisone could suppress adrenal responsiveness,
any patient whose posttreatment stimulated cortisol level was lower than 276
nmol/L (<10 µg/dL) was provided (without breaking the study code)
a 6-week tapering regimen of open-label hydrocortisone treatment, after which
the stimulation test was repeated.
Every day, beginning 2 weeks before treatment and for the duration of
treatment, the patients were instructed to record their current Wellness score,
a single-item global health scale, ranging from zero, representing the worst
they had ever felt, to 100, representing the best they had ever felt.10 Once per week, the patients completed the Profile
of Mood States questionnaire, a standardized instrument for the quantitative
measure of anger, anxiety, confusion, depression, fatigue, and vigor.10,23 Twice before treatment and every
3 weeks thereafter, the patients completed additional standardized self-rating
instruments: the Symptom Checklist–90-R,24
the Sickness Impact Profile,25 the Beck Depression
Inventory,26 and a 10-point Activity scale
that we developed. All patients were also interviewed by a psychiatric specialist
who administered the Hamilton Depression Rating Scale at entry to the study
and again at the completion of treatment. This scale is a 17-item, observer-rated
instrument that assesses the severity of symptoms commonly present in a depressive
Evaluation of efficacy was based primarily on changes in the Wellness
score, using as end points improvement by any amount, by 5, 10, and 15 points,
and the change in mean score during treatment. An individual was considered
improved if he or she completed at least 10 weeks of treatment, did not discontinue
treatment because of worsening symptoms or adverse reactions, and had a mean
Wellness score during treatment that was higher than the mean score for a
1-week pretreatment period. All other subjects were considered not improved.
The amount of improvement in a particular scale was estimated by the
difference between the treatment and pretreatment mean scores. For the Wellness
score only, the amount of improvement was estimated under the assumption that
2 individuals who did not complete the study and were missing either all pretreatment
or all treatment scores had the same change as the average for others who
did not complete the study. Pretreatment Wellness scores were not available
for 5 hydrocortisone recipients, so that improvement could not be assessed
for those patients.
Seventy patients were to be enrolled, leaving 62 patients if the drop-out
rate was about 10%. For an underlying proportion of improvement on placebo
of 50%, a study of 62 patients in 2 equal-sized groups would have 80% power
to detect a treatment effect at a 2-sided .05 significance level, if the rate
of improvement in hydrocortisone recipients were 83%. Randomization was done
in blocks of 10 subjects.
Secondary study end points included improvement and amount of change
in the other self-rating scales, adverse symptoms on treatment, and alterations
in basal and cosyntropin-stimulated cortisol levels.
To measure the association between the Wellness and other scores, Spearman
rank correlation coefficients were calculated for the change in Wellness scores
and the changes in each of the other scores.
A χ2 test or Fisher exact test was used to compare rates
of improvement for the various instruments, baseline demographics, and adverse
effects. Quantitative variables were compared using a 2-tailed Wilcoxon rank
sum test with continuity correction. P≤.05 was
considered statistically significant. No adjustments were made for multiple
comparisons. Randomization was done in blocks of 10 subjects.
A total of 638 patients with chronic fatigue were screened for the study
(Figure 1). Of these, 182 would
not withhold proscribed medications, mostly antidepressants and nonsteroidal
analgesics, and could not be enrolled—none of them were taking systemic
steroids. An additional 151 patients were excluded because they did not meet
the 1988 CDC CFS case definition,2 109 eventually
declined enrollment for unstated personal reasons, 61 had complicating medical
conditions that precluded their study, 30 experienced onset of illness that
was too gradual, 15 individuals did not meet the age criteria, 14 had contraindications
to steroid use, and 2 were pregnant. In all, 74 patients were admitted for
final evaluation. Four proved too depressed to justify discontinuing their
use of antidepressant medications.
Seventy patients were enrolled in the trial, 35 in each treatment arm.
Their demographic and historical features, baseline self-rating scores, and
adrenal-hormone status were well distributed between the 2 study arms. However,
trends toward a shorter duration of illness, higher mean scores on the Beck
Depression Inventory26 and the Profile of Mood
States depression score,23 and more individuals
with no current DSM-III-R diagnoses22
were noted in those randomized to hydrocortisone (Table 1).
The 70 patients began treatment between May 1992 and December 1995.
Seven patients terminated treatment prematurely, after 3 to 75 days. Five
(2 hydrocortisone, 3 placebo) of them terminated due to their perceived exacerbation
of symptoms of CFS. One additional placebo recipient developed a rash, and
1 hydrocortisone recipient decided to pursue a different (nonstudy) medication
after 3 days. Two women who became pregnant during treatment were considered
to have completed the study after 55 and 67 days. The remaining 61 patients
received 76 to 96 days of therapy.
The 2 groups did not differ significantly in the percentage of subjects
who recorded improvement in mean Wellness scores while receiving treatment:
19 (54.3%) of 35 placebo recipients and 20 (66.7%) of 30 hydrocortisone recipients
(P=.31). There was, however, evidence that hydrocortisone
treatment led to some improvement in patient status. First, the percentages
of patients recording improvement of at least 5, 10, or 15 points on the Wellness
scale were greater for hydrocortisone than placebo recipients (5 points, 53%
vs 29%, P=.04; 10 points, 33% vs 14%, P=.07; and 15 points, 20% vs 6%, P=.08). Second,
a plot of the difference between the average amount of improvement in Wellness
score for each day (daily score minus pretreatment mean) showed that the hydrocortisone
recipients improved quickly and sustained a greater increase in Wellness scores
than the placebo recipients for every day of the 12 treatment weeks (Figure 2). Comparison of the numbers of days
on which average improvement was greater for the 2 groups reflects the consistently
higher average improvement for hydrocortisone (P<.001
by Sign test). The mean improvement on treatment in the hydrocortisone group
was 6.3 points, being higher than the mean improvement of 1.7 points for the
placebo recipients (P=.06; Table 2).
While there were higher rates of improvement (data not shown) and mean
changes from the pretreatment scores on most of the other rating scales and
subscales for the hydrocortisone recipients, none of these differences was
statistically significant (Table 2).
For disorders with largely subjective features like CFS, one must use
self-rating instruments to quantify patient performance and change. Here,
we included a series of self-rating instruments, most of which have been validated
individually in multiple other populations.22- 26
The amount of improvement in Wellness score, which is defined by a remarkably
simple, single-item global health scale, was significantly correlated with
improvement in many of the other more complex and well-characterized scales.
The Spearman correlations were generally similar for all treated subjects
combined and for subjects receiving either hydrocortisone or placebo alone
(data not shown). The significant correlations were positive for scales for
which, like the Wellness score, positive changes indicated improvement, and
negative otherwise. For example, changes in the Wellness score correlated
negatively with changes in the Profile of Mood States Fatigue scale (r=−0.56, P<.001), Confusion
scale (r=−0.42, P<.001),
the Beck Depression Inventory (r=−0.42, P<.001), and the Sickness Impact Profile score (r=−0.45, P<.001) and
positively with changes in the Activity scale (r=0.37, P=.003) and the Profile of Mood States Vigor scale (r=0.53, P<.001).
Patients maintained daily records of adverse reactions. Some of these
reactions included symptoms inherent in CFS itself such as fatigue, depressed
mood, and difficulty with concentration.2,3
Others, such as weight gain and acne, are well-recognized reactions to high
doses of corticosteroids.28Table 3 shows that of 21 adverse reactions elicited, 3 occurred
significantly more frequently in hydrocortisone recipients: increased appetite,
weight gain, and difficulty in sleeping. Actual patient weights confirmed
their reports. The mean (SD) weight gain among hydrocortisone recipients during
treatment was 1.5 (1.9) kg, while that of the placebo recipients was 0.5 (1.9)
kg, a small but statistically significant (P=.02)
The morning basal and cosyntropin-stimulated cortisol levels at entry
were well matched between the study arms (Table 1). There was no significant correlation between the pretreatment
basal or stimulated cortisol levels, or the mean change in level on stimulation,
with any of the self-rating scores or their changes with treatment. For example,
the Spearman rank correlation coefficients for changes in Wellness score with
hydrocortisone treatment were 0.05 for both baseline serum cortisol levels
(P=.81) and baseline stimulated cortisol responses
(P=.79). The patients with the lowest cortisol levels
and adrenal reserve were not the most symptomatic, nor were they more likely
to respond to hydrocortisone treatment.
The placebo recipients showed no obvious changes in their basal and
stimulated cortisol levels during the treatment period. The patients who received
hydrocortisone, however, experienced significant adrenal suppression. Specifically,
mean (SD) basal morning serum cortisol levels at the completion of the study
(408  vs 279  nmol/L [14.8 [5.5] vs 10.1 [6.5] µg/dL for placebo
and hydrocortisone recipients, respectively; P=.002)
and the mean level achieved in response to cosyntropin (869  vs 577 
nmol/L; P<.001 [31.5 (3.7) vs 20.9 (9.0) µg/dL])
were reduced by steroid treatment. During posttreatment testing, 5 patients
had depressed cortisol responses to cosyntropin (<276 nmol/L [<10 µg/dL])
and began on a tapering regimen of open hydrocortisone. Upon unblinding the
study, all 5 were found to have been randomized to the steroid treatment arm.
In addition, 7 patients had suboptimal posttreatment responses to cosyntropin
(stimulated cortisol levels of 276-497 nmol/L [10-18 µg/dL]) and all
7 had been receiving hydrocortisone treatment.
In a randomized, double-blind, placebo-controlled trial in CFS, low-dose
hydrocortisone treatment was associated with improvement in symptoms as measured
by change in Wellness scores. This is the first such study, to our knowledge,
to demonstrate improvement with a drug treatment of CFS.
Significant therapeutic benefit, however, was not evident by changes
in other self-rating scales. More important, what little improvement might
be attributable to hydrocortisone treatment was achieved at the expense of
significant adrenal suppression. Twelve of 33 patients randomized to hydrocortisone,
but none of 33 placebo recipients, showed suboptimal responses to cosyntropin
challenge at the end of treatment (P<.001 by 2-sided
Fisher exact test). Although steps were taken to avert serious or potentially
life-threatening adrenal insufficiency in the face of emergent stress, the
fact that it could happen with less cautious widespread use precludes the
present regimen of hydrocortisone or comparable doses of other systemic corticosteroids
as acceptable choices for the prolonged treatment of CFS.
Other conclusions can be drawn from the study as well. First, many placebo
recipients demonstrated some improvement on the Wellness score, as chance
alone might predict; however, the magnitude of that improvement was small.
Patients with CFS simply do not improve much through an observation period
of 3 months. In this regard, the present data define thresholds of spontaneous
clinical change that will prove invaluable in projecting sample size and study
design in further trials. Specifically, the percentages of placebo recipients
exhibiting 5-, 10-, and 15-point improvements in Wellness scores were 29%,
14%, and 6%, respectively.
Second, many subjects who were referred for this study proved ineligible
for enrollment. Nearly 25% (151/638) did not meet CDC case criteria for the
diagnosis.2 Clearly, there remains considerable
variability in how the case criteria are applied in the community.7 Another major reason for nonenrollment was the patients'
perceived need for other medications. It is always difficult to know how fastidious
one should be in designing a therapeutic trial like this one. To allow concomitant
medications might better approximate the setting in which a new treatment
will be used in the community, but it will add confounding issues that could
complicate interpretation of adverse reports and treatment outcome. Clearly,
our decision to select subjects who withheld many other medications might
have excluded the sickest subjects. Nevertheless, the results of the present
study should be applicable to the typical CFS patient population because the
degree of disability and clinical characteristics of those enrolled (Table 1) so well emulates that of other
published CFS cohorts.10,29- 31
To improve the homogeneity of the patient sample, enrollment was restricted
to patients whose fatigue arose relatively acutely. As it turned out, less
than 5% of subjects screened reported onset of their illness over a period
of longer than 6 weeks. Moreover, that no patients screened were excluded
for using systemic corticosteroids precludes the possibility that study enrollment
was biased toward those less likely to respond to hydrocortisone.
Third, the present data indicate that prolonged, low-dose daily glucocorticoid
use has risks. There are, to our knowledge, no other placebo-controlled trials
of prolonged daily hydrocortisone treatment involving doses as low as those
used here (25-35 mg, equivalent to approximately 6 mg of prednisone). Low-dose
glucocorticoid replacement, defined as 20 to 40 mg of hydrocortisone in divided
daily doses, was felt to be safe, to cause no symptoms other than occasional
gastric distress, and to benefit patients with chronic fatigue.32
We found that low-dose hydrocortisone treatment has mild side effects (Table 3) and affords minimal therapeutic
benefit for CFS, but significantly suppresses adrenal responsiveness.
Fourth, the data bear importantly on the basic hypothesis under which
this study was undertaken, namely that CFS symptoms are perpetuated through
suboptimal activity of the HPA axis.15 That
the basal and stimulated-cortisol levels did not correlate with illness severity
in this study (data not shown), nor were they predictive of clinical improvement
or response to treatment, argue against the hypothesis. The fact that there
was evidence of symptomatic benefit in hydrocortisone recipients, however,
is concordant with this hypothesis, and yet the limited benefit indicates
that mere supplementation of cortisol is not sufficient. It is possible that
a more salutary effect would derive from a different low-dose regimen, or
from specific supplementation of corticotropin-releasing hormone or pharmacologic
augmentation of its release, were these latter options to become feasible.33,34