Beruflich Dokumente
Kultur Dokumente
and Behavior
Preliminary Findings
Owen M. Wolkowitz, MD; David Rubinow, MD; Allen R. Doran, MD; Alan Breier, MD; Wade H. Berrettini, MD, PhD;
Mitchel A. Kling, MD; David Pickar, MD
\s=b\ To evaluate the neurochemical, neuroendocrine, and behav- not only in the hypothalamus and pituitary gland, but also in
ioral effects of exogenous corticosteroids in humans, we admin- the hippocampus, septum, and amygdala,1118 areas ofthe brain
istered prednisone (80 mg/d orally for 5 days) in a double-blind believed to be intimately involved in behavior, mood, and
manner to 12 medically healthy volunteers. Behavioral measures memory. Many central nervous system effects of corticoste¬
were assessed before, during, and after prednisone administra- roids are mediated via genomically related events (ie, tran¬
tion in all 12 subjects, and cerebrospinal fluid biochemistry was scription). Other effects may be related to more rapid changes
assessed before and during prednisone administration in 9 of the in brain excitability and synaptic transmission that are pro¬
subjects. Prednisone administration was associated with de- duced by altering cyclic nucleotide metabolism,19 by interact¬
creases in cerebrospinal fluid levels of corticotropin, norepi- ing directly with neuronal membranes,20 by altering specific
nephrine, \g=b\-endorphin,\g=b\-lipotropin,and somatostatinlike immu- ionic conductances,19 and by altering central carbohydrate,
noreactivity. No significant changes were noted in cerebrospinal protein, and lipid metabolism.19 Prior studies have also sug¬
fluid levels of \g=a\-melanocyte-stimulatinghormone, corticotropin- gested that corticosteroids alter the activities of monoamine
releasing hormone, 3-methoxy-4-hydroxyphenylglycol, homo- neurotransmitters1112,16 and of proopiomelanocortin and other
vanillic acid, or 5-hydroxyindoleacetic acid. No consistent or neuropeptides.21"23 Few studies, however, have prospectively
significant group mean changes were observed in structured assessed corticosteroid effects on human neurochemistry and
behavioral ratings, although 9 (75%) of the volunteers studied neuroendocrinology, and few studies have concurrently as¬
reported mild behavioral changes while receiving prednisone. sessed the behavioral and neurochemical effects of corticoste¬
Correlations between the neurochemical and behavioral roids to examine which of these reflections of central nervous
changes are discussed. system activity are altered in concert.
{Arch Gen Psychiatry. 1990;47:963-968) In this study, we employed a prospective, double-blind
method and assessed behavioral and concomitant biochemical
effects in an attempt to delineate the biologic correlates of
administration of corticosteroids is fre¬ specific corticosteroid-induced behavioral changes. Because
Pharmacol
For
quently ogie
associated with behavioral alterations in humans.
example, medically
in ill patients treated systemically
important differences are likely to exist between the central
nervous system effects of endogenous and exogenous cortico¬
with exogenous corticosteroids, a fluctuating array of symp¬ steroids,11 our study was designed to investigate possible
toms may develop, such as mood lability, depression, irritabil¬ mechanisms for exogenous corticosteroid-induced central ner¬
ity, insomnia, poor concentration, and even psychosis.1'10 The vous system changes.
reported incidence of such side effects varies from study to
study, ranging from 1.8% to 57%z'""7; the individual risk factors SUBJECTS AND METHODS
for experiencing them are unclear. Clinical investigations of
these effects have been hampered by méthodologie difficul¬ Subjects
ties, including the study of patients with acute medical illness¬
es, the lack of prospective study designs, and the lack of Twelvemedically healthy volunteers (4 women, 8 men; aged 21 to
baseline (presteroid) assessments. The majority of prior stud¬ 41 years) participated in this institutional review board-approved
ies, therefore, have been unable to separate behavioral re¬ study after granting written informed consent. All volunteers were
screened for absence of medical illness, including endocrinopathies,
sponses directly attributable to corticosteroids from those by thorough history, physical examination, and laboratory examina¬
attributable to the underlying concurrent medical illnesses. tion. Psychiatric history was elicited by structured interview (Sched¬
The biologic mechanisms by which corticosteroids may af¬ ule for Affective Disorders and Schizophrenia).24 The Schedule for
fect behavior are also not fully understood. A large body of Affective Disorders and Schizophrenia interview sheets were re¬
evidence now suggests that corticosteroids have direct and viewed by two research psychiatrists, blind to study outcome, and
indirect effects on the central nervous system in animals and in consensus diagnoses were arrived at according to DSM-IH-Rr and
humans,11"1' and corticosteroid receptors are densely located Research Diagnostic Criteria.26 All volunteers were psychiatrically
healthy when studied and for at least 2 years before the beginning of
Accepted for publication October 18,1989. the study. However, 6 of the subjects had a personal or family
From the Department of Psychiatry, University of California, San Francisco psychiatric history; personal histories were of generally minor sever¬
(Dr Wolkowitz); the National Institute of Mental Health, Bethesda, Md (Drs ity (eg, adjustment disorder; Table 1). The remaining 6 subjects were
Rubinow, Berrettini, Kling, and Pickar); the Department of Psychiatry, Uni- free of such personal or family histories. In particular, all volunteers
versity of California, Davis, Sacramento (Dr Doran); and Maryland Psychiatric met Research Diagnostic Criteria for "currently not mentally ill, " and
Research Center, University of Maryland School of Medicine, Baltimore (Dr 7 met Research Diagnostic Criteria for "never mentally ill." Volun¬
Breier). teers were free of all medications, including birth control pills and
Read before the 26th Annual Meeting of the American College of Neuropsy-
chopharmacology, San Juan, Puerto Rico, December 9,1987. steroid-containing dermatologie preparations, for at least 1 year, and
Reprint requests to Department of Psychiatry, University of California, San were free of significant alcohol or other drug use for at least 21k years
Francisco, 401 Parnassus Ave, San Francisco, CA 94143 (Dr Wolkowitz). before beginning this study. Volunteers were maintained on a caf-
6/M/28 Adjustment Slight elevation of without preservative and frozen at 80°C until assayed for ß-lipotro-
disorder (26) mood, restless pin (BLPH) and cortisol. -
sleep All CSF samples were assayed in duplicate in the same assay batch,
7/M/21 Irritable (WD), and paired samples (ie, baseline and prednisone) were from the same
trouble sequential aliquots. Cerebrospinal fluid corticotropin was assayed in
concentrating extracted CSF by radioimmunoassay (RIA).27 The intra-assay coeffi¬
(WD) cient of variation (CV) equals 4.4%, and the sensitivity was 3 to 5
8/M/21 Irritable, angry, pg/mL. The CSF CRH was assayed in extracted CSF by RIA.28 The
increased energy, intra-assay CV was equal to 6.0%, and the sensitivity was 1.5 to 2.0 pg
anxiety, per tube. The CSF SLI was assayed by RIA29 employing tyrosine-io-
confusion, dine 125-somatostatin synthetic cyclic somatostatin standards, rabbit
depersonalization, antisomatostatin antiserum, and charcoal separation. The intra-assay
racing thoughts, CV was equal to 5.1%, and the sensitivity was less than 1.0 pg per
mildly elated tube. The CSF BE was assayed by RIA.30 The intra-assay CV was less
9/M/41 Irritable, polyuria, than 1.5%, and the sensitivity was 15 pg/mL. The antibody cross-
difficulty reacted with BLPH at the 10% level. The CSF BLPH was assayed by
concentrating RIA.30 The intra-assay CV was less than 1.5%, and the sensitivity was
10/M/34 Polyuria 20 pg/mL. The CSF -MSH was assayed by RIA.30 The intra-assay
11/M/23 Mildly "high," CV was less than 1.5%, and the sensitivity was 5 pg/mL. The CSF for
labile mood, HVA, NE, MHPG, and 5-HIAA was assayed using high-pressure
depersonalization liquid chromatography with electrochemical detection.3,32 Intra-as¬
12/M/32 Alcohol "Buzzed," say CV was equal to 5%, 7%, 4.5%, and 2.0%, respectively. Sensitiv¬
intoxication (28) decreased sleep ities were 200 nmol/mL, 0.17 pmol/mL, 80 nmol/mL, and 100
*Age (years) at which diagnosis was applicable. All subjects were free of nmol/mL, respectively. The CSF for cortisol was assayed by RIA33;
psychiatric illness for at least 2 years before beginning the study. intra-assay CV was equal to 6%, and sensitivity was 0.2 µg/dL.
fSymptoms reported in unstructured subjective logs or nurses' observations Blood was collected via standard forearm venipuncture at 8 AM for
during prednisone treatment, except those followed by WD reported during assay of prednisolone, the active metabolite of prednisone, on the final
postprednisone placebo (withdrawal) period. day of the initial placebo period and on the last 2 days of prednisone
administration. Blood was collected on ice in an edetic acid-containing
tube and centrifuged at 3000 rpm for 10 minutes. Plasma was stored at
20°C until assayed for prednisolone by RIA following extraction and
feine-restricted, low-monoamine diet beginning 3 days before the chromatography.34 The intra-assay CV was equal to 9.4%, and the
-
Health (Bethesda, Md) Clinical Center research ward for 20 days. For
the first 3 days no studies were performed to allow volunteers to Behavioral Measures
acclimate to their surroundings and to begin the low-monoamine diet.
Following that, placebo capsules were administered for 5 days, followed Behavioral measures were obtained daily at the same time of day
by prednisone (80 mg/d orally at 10 AM) for 5 days, followed by placebo throughout the study. Self-ratings, completed between 10 AM and
again for 7 days. All medications were administered in a double-blind noon, consisted of the Symptom Checklist 90 (SCL-90),35 which as¬
manner (ie, subjects, nurses, and raters were unaware of medication sesses a variety of behavioral symptoms, and a group of 10 specially
status or of study design). The dose of prednisone and the length of designed visual analogue scales (VAS) rating sadness, anxiety, rest¬
administration were chosen to match those at which some medically ill lessness, energy, general well-being, confusion, sensory sharpness,
patients being treated with corticosteroids begin to experience behav¬ happiness, irritability, and sexual arousal. Each VAS consisted of a
ioral reactions.2 Previous experience in healthy volunteers with this dose 100-mm horizontal line with polar adjectives as anchor points on either
and regimen in the National Institute of Allergy and Infectious Diseases, end (eg, "most sad ever" and "least sad ever"). Volunteers were
Bethesda, Md, has established that no medical complications were likely instructed to place a perpendicular line through the scale line to
if subjects were screened to exclude those with diabetes mellitus, gastric indicate how they were feeling along that dimension then. The score
ulcer, tuberculosis, or other occult infections (H. Clifford Lane, MD, oral was determined by measuring the distance in millimeters from the left
communication, December 11,1984). end of the scale line to the volunteer's mark. For purposes of analysis,
Placebo Prednisone
(Mean ± SD) (Mean ± SD)
CRH, pg/mL 35.1 ±11.4 31.7±11.4
SLI, pg/mL 37.9±11.1 32.3±13.2f
NE, pmol/mL 0.66 + 0.15 0.52±0.15t
MHPG, pmol/mL 38.0 ±10.8 37.6 ±8.4
HVA, pmol/mL 203.0±83.4 198.2±83.7
5-HIAA, pmol/mt 93.8 ±43.2 94.3 ±41.4
Cortisol, µg/dL 0.67±0.22 0.20±0.0
*CRH indicates corticotropin releasing hormone; SLI, somatostatinlike ¡m-
munoreactivity; NE, norepinephrine; MHPG, 3-methoxy-4-hydroxyphenylgly-
col; HVA, homovanillic acid; and 5-HIAA, 5-hydroxyindoleacetic acid.
fP<.01, paired (test.
<.001, paired?test.
first stage of the CSF biochemical tests, .025 for the second stage, and
.029 for the third stage.
Prednisone-associated changes (mean ± SE) in cerebrospinal fluid Two additional aspects of our data analysis were addressed. Be¬
levels of proopiomelanocortin-related peptides at baseline (open cause half of the subjects studied had personal or family history of
bars) and following 5 days of prednisone administration (shaded psychiatric illness, we separately analyzed the data of the six subjects
bars). BLPH indicates ß-lipotropin; BE, ß-endorphin; -MSH, without such histories to ascertain whether any of our major findings
-melanocyte-stimulating hormone; and asterisk, P<.03 (paired f were altered in this fully normal subgroup. Second, we anticipated
test [corticotropin and BLPH], Wilcoxon Signed-Rank Test [BE]). that examination of group mean changes in specific behavioral vari¬
ables might fail to detect the fluctuating and multidirectional behav¬
the four VAS items energy, happiness, irritability, and sexual arousal ioral changes that we hypothesized would occur, based on previous
were averaged into one "hypomania" scale. In addition to the self- studies.1,3 We therefore also descriptively present behavioral data
ratings, psychiatrists blind to medication status and study design from the unstructured reports of the blinded subjects and ward
rated the volunteers' behavior daily (between noon and 1 pm) with the nurses.
modified 24-item Brief Psychiatric Rating Scale.36 Finally, subjects
were asked to keep unstructured logs (diaries) of their experiences RESULTS
throughout the study, and on completion of the experimental protocol Biochemical Measures
volunteers were specifically asked if they could discern any distinct
behavioral changes during the protocol. Their responses, as well as Prednisone administration was associated with significant de¬
unstructured behavioral reports from blinded nursing staff (gleaned creases in CSF levels of SLI (i[8] 3.40, P<.01),21 NE ( [8] 4.33,
= =
from nursing daily progress notes), were tabulated for descriptive P<.01), corticotropin (i[8] 2.75, P .025), BLPH (i[8] 2.69,
= = =
the Spearman Rank-Order Correlation coefficients as indicated in the [near significant by Bonferroni criteria]; post hoc t tests comparing
text. Because of our interest in steroidal associations with affective baseline with steroid, P<.05; comparing steroid with withdrawal,
states, we were most interested in biologic associations with changes P<.05). There were similarly no significant drug effects on group
in VAS-rated sadness and hypomania and total SCL-90 ratings. How¬ SCL-90 or Brief Psychiatric Rating Scale ratings.
ever, we also present the full matrix of correlations between CSF In contrast to the lack of effect on group mean behavioral measures,
neurochemical changes and changes in the remainder of the behavior¬ 9 of the 12 individual volunteers experienced prednisone-associated
al measures to generate hypotheses for testing in more focused follow- behavioral changes (recorded by unstructured report of the subjects
up studies. or of the blinded nursing staff) with discrete onset during the predni¬
Because of the large number of variables tested in this study, sone administration period (Table 1). In contrast, only 1 subject
significance levels were adjusted with the multistage Bonferroni experienced a behavioral change during the baseline period (mild
correction.37 We thought that in this exploratory study it was appro¬ euphoria) and 3 subjects experienced behavioral changes during the
priate to select a relatively large familywise (ie, for the entire study, withdrawal period. The reported behavioral changes during the ste¬
not individual comparisons) value, .20, to minimize the likelihood of roid period included mild hypomania (eg, irritability, feeling high or
a type II error,38 recognizing that the likelihood of a type I error is speeded or more talkative; 8 subjects), decreased sleep (3 subjects),
thereby increased. In all, we tested 10 CSF biochemical variables, 9 loss of concentration (2 subjects), mild depression (1 subject), and
behavioral variables, and 90 biochemical-behavioral correlations. In¬ depersonalization (2 subjects). The unstructured reports of symptoms
dividual test values were thus determined to be .022 for the behav¬ were substantiated by increases (3=10 mm) in individual subject VAS
ioral tests,. 002 for the biochemical-behavioral correlations,. 02 for the ratings: 7 reported more happiness, energy, irritability, or restless-
Sensory
ASCL-90 Sad Anxious Restless Well-being Confused Sharpness Hypomanic ABPRS
ASLI -60t .68+ .47 -.07 .50 -.25 .14 .68+ .54
75§ .03 .10 .18 .00 -.21 .18 60f .45
ness, 5 more sadness, 4 more tiredness, 3 more anxiety, and 2 more We observed that prednisone administration was associated
confusion during the steroid period compared with the baseline peri¬ with decreases in CSF levels of several biologically and behav-
od. There was no difference in the incidence or type of behavioral iorally active neuropeptides or transmitters. There was, how¬
reactions between the groups with and without psychiatric histories.
ever, a clear absence of significant prednisone effects on group
mean behavioral measures, although a high percentage of our
Biochemical-Behavioral Correlations volunteers (75%) described or exhibited subtle, diverse be¬
A correlation matrix of the prednisone-associated changes in CSF havioral alterations during prednisone administration. Fur¬
biochemistry and changes in behavioral ratings is given in Table 3. thermore, certain behavioral alterations were correlated with
Relatively high correlation coefficients (rs=.60) were observed be¬ specific prednisone-associated neurochemical alterations.
tween changes in CSF levels of SLI, NE, and BE and changes in SCL- These preliminary data, derived from the only prospective,
90 ratings and/or VAS sadness and hypomania ratings. The BE- double-blind study of corticosteroid effects in humans to date,
behavioral correlations were unchanged when BE levels were
corrected for BLPH cross-reactivity. Additionally, changes in CSF represent a first step toward a better understanding of the
effects of exogenous corticosteroids on human biologic and
levels of BLPH, -MSH, CRH, HVA, and 5-HIAA were correlated
with changes in VAS-rated restlessness and confusion, and changes in behavioral function.
CSF levels of corticotropin were correlated with changes in Brief There are a number of important considerations in our
Psychiatric Rating Scale ratings. The relationships between these study, however, that limit the inferences that may be drawn
biochemical and behavioral changes were also reflected in the sub¬ from it: (1) Although all of our volunteers were free of psychi¬
group of subjects without psychiatric histories. When the more strin¬ atric illness for at least 2 years, the presence of personal or
gent Bonferroni approach is used to consider the large number of family psychiatric histories in half of our volunteers may limit
variables tested, only the correlations of changes in CSF levels of BE the generalizability of our findings to individuals without such
and BLPH with changes in SCL-90 ratings and VAS-rated restless¬ histories. The significance of this in the present study is
ness and confusion, respectively, remain significant. Specifically,
lesser suppression of CSF BE levels was associated with increased mitigated somewhat by our use of each subject as his or her
own control and by our demonstration that the major findings
SCL-90 ratings, and greater suppression of CSF BLPH levels was
associated with increased ratings of restlessness and confusion. were unaltered in the subgroup of subjects without personal
or family psychiatric histories. Indeed, heterogeneous sam¬
Prednisolone and Cortisol Levels ples, such as we employed, may more closely approximate the
All volunteers showed an increase in plasma prednisolone levels psychiatrically unselected population that is clinically pre¬
scribed synthetic corticosteroids for medical illnesses. (2) Our
during prednisone administration (mean ± SD baseline level, 1.7 ± 0.9 use of the small sample size increases the risk of a type II
ng/mL, steroid level, 23.9±15.4 ng/mL; <[11]=4.95, P<.001). This error; nonetheless, we were able to demonstrate several sig¬
increase was unaltered when prednisolone levels were corrected for
cortisol cross-reactivity. The volunteers also showed significant pred¬ nificant findings. (3) Our analysis of multiple comparisons
nisone-associated decreases in CSF cortisol levels (¿[8] 6.27, =
poses a further limitation. In exploratory studies such as this,
P<.001) and in plasma cortisol levels (F[2,22] 24.92, P<.0001) with
=
it is frequently desirable to assess a variety of dependent
return of plasma cortisol to baseline levels following prednisone with¬ measures.38 However, definitive conclusions based on our
drawal. Changes in plasma prednisolone levels and in CSF and plasma study must be tempered because the experiment (familywise)
cortisol levels were not significantly correlated with changes in behav¬ error rate based on the Bonferroni correction is 0.20. Our data
ioral ratings or with changes in CSF neurochemistry.
must, therefore, be considered preliminary and as hypothesis-
COMMENT generating for future larger-scale and more focused studies.
(4) Since all of our subjects received lumbar punctures in the
We studied the biologic and behavioral effects of predni¬ same sequence (baseline followed by steroid), it is impossible
sone administration in a heterogeneous group of volunteers to discount the possibility that the observed decreases in
not currently suffering from psychiatric or medical illness. neuropeptide and transmitter levels were secondary to the