discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/14813592
Article in JAMA The Journal of the American Medical Association May 1993
DOI: 10.1001/jama.269.17.2232 Source: PubMed
CITATIONS READS
187 79
4 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Jack A Yanovski on 02 December 2015.
Objective.\p=m-\The biochemical and phenotypic presentation of mild hypercortiso- MANY of the laboratory abnormalities
lism in Cushing's syndrome is often indistinguishable from that seen in pseudo\x=req-\ that are characteristic of Cushing's syn
Cushing's states such as depression. Both dexamethasone suppression and drome,1 such as elevated urine free Cor
tisol and 17-hydroxycorticosteroid ex
corticotropin-releasing hormone (CRH) stimulation tests have been used individu-
ally to distinguish these conditions, but neither approach has achieved a diagnostic cretion, disruptions in the normal diur
nal pattern of plasma cortisol secretion,
accuracy greater than 85%. Therefore, we sought to develop a combined and lack of suppression of plasma cor
dexamethasone-CRH test that would take advantage of the altered sensitivity of tisol after administration of 1 mg of dex-
patients with Cushing's syndrome to both dexamethasone and CRH and would amethasone, may be seen in other con
achieve greater accuracy in the diagnosis of Cushing's syndrome. ditions. Situations that cause these Pseu
Design.\p=m-\Prospectivecohort study. do-Cushing's states include long-term,
Setting.\p=m-\Tertiarycare research hospital. active alcoholism and withdrawal from
Patients.\p=m-\Atotal of 58 adults referred for evaluation of mild hypercortisolism ethanol intoxication,23 stressful condi
(urine free cortisol level <1000 nmol/d). The diagnosis of Cushing's syndrome was tions such as surgery or severe illness,4
confirmed at surgery in 39 patients. The diagnosis of a pseudo-Cushing's state was renal failure,5 anorexia and bulimia ner
made in 19 patients on the basis of extended follow-up (mean, 28 months) without vosa,6"8 the depressed phase of affective
progression of cushingoid features. disorders,910 primary glucocorticoid re
Intervention.\p=m-\Thelow-dose dexamethasone suppression test, the CRH stim- ceptor resistance,11 and severe obesi
ulation test, and the CRH stimulation test started 2 hours after completion of low\x=req-\ ty.912 While the context and physical
examination may provide important
dose dexamethasone suppression (the dexamethasone-CRH test) were per- clues that a particular hypercortisolem-
formed in all patients. ic patient does not have Cushing's syn
Main Outcome Measures.\p=m-\Sensitivity,specificity, and accuracy of the three drome, definitive diagnosis may be quite
procedures for diagnosis of Cushing's syndrome were calculated from plasma cor- difficult in obese patients with mild hy-
ticotropin, plasma cortisol, urine free cortisol, and urine 17-hydroxycorticosteroid percortisolism, hirsutism, hypertension,
values. or depression. Thus, such patients may
visit their internist, family physician,
Result.\p=ml-The ow-dose dexamethasone suppression test had 74% specificity,
69% sensitivity, and 71% diagnostic accuracy, using the standard criterion gynecologist, psychiatrist, or other sub-
specialist, complaining of symptoms and
(17-hydroxycorticosteroid excretion level >11.0 \g=m\mol/d on the second day of dex- exhibiting signs consistent with Cush
amethasone administration). With a urine free cortisol criterion for Cushing's syn-
ing's syndrome. Further evaluation of
drome of greater than 100 nmol/d, the low-dose dexamethasone suppression test these patients may then reveal 24-hour
had 100% specificity, 56% sensitivity, and 71% diagnostic accuracy. The CRH urine excretion of free cortisol or 17-
stimulation test without dexamethasone pretreatment had 100% specificity, 64% hydroxycorticosteroid compatible with
sensitivity, and 76% diagnostic accuracy. The diagnostic accuracy of the the diagnosis of mild Cushing's syndrome
dexamethasone-CRH test for Cushing's syndrome was significantly greater than or a pseudo-Cushing's state.
the accuracy of either the low-dose dexamethasone test or the CRH test alone The hypercortisolism of pseudo-Cush
(P<.01). A plasma cortisol concentration greater than 38 nmol/L measured 15 min- ing's states is believed to be mediated
utes after the administration of CRH correctly identified all cases of Cushing's syn- through increased hypothalamic secre
tion of corticotropin-releasing hormone
drome and all cases of pseudo-Cushing's states (100% specificity, sensitivity, and
diagnostic accuracy). From the Developmental Endocrinology Branch,
Conclusion.\p=m-\Thedexamethasone-CRH test is a more accurate test to distin- National Institute of Child Health and Human Develop-
guish Cushing's syndrome from pseudo-Cushing's states in patients with mild ment, National Institutes of Health, Bethesda, Md.
Reprint requests to Bldg 10, Room 10N262, Nation-
hypercortisolism. al Institutes of Health, Bethesda, MD 20892 (Dr
(JAMA. 1993;269:2232-2238) Yanovski).
Fig 2.Results of low-dose dexamethasone suppression test In 58 patients with hypercortisolism. Criteria shown have 100% specificity. Left, Urine free cortisol
on fourth day of low-dose dexamethasone test. Right, 17-hydroxycorticosteroid excretion on fourth day of low-dose dexamethasone test. Definition of symbols as
In Fig 1. Horizontal lines indicate criterion levels.
tion at a dose of 1 g/kg of body weight. 25% for adrenocorticotropin, and 6% and Bonferroni adjustment for multiple
Plasma samples were assayed for cor 13% for cortisol. Each cortisol sample comparisons.
tisol and adrenocorticotropin by Hazel- was also measured in a second cortisol The receiver operating characteris
ton Laboratories, Vienna, Va, at four assay (Abbott TDX kit, Abbott Labo tics26 of each test for the diagnosis of
basal time points (-15, -10, -5, and 0 ratories, North Chicago, 111). Results Cushing's syndrome were then deter
minutes) and then at 5,15,30,45, and 60 were quite similar, and each cut point mined using the RuleMaker program
minutes after CRH administration. The identified using the Hazelton assay had (Digital Medicine Ine, Hanover, NH).
first CRH test was performed while the the same diagnostic accuracy for Ab Sensitivity, specificity, positive and neg
patient took no glucocorticoids. A sec bott kit cortisol determinations. There ative predictive values, and diagnostic
ond CRH stimulation test (the dexa fore, only results from the Hazelton cor accuracy27 were calculated for each test
methasone-CRH test) was performed 2 tisol assay were reported. Each admin variable at multiple criteria. For each
hours after the patient had completed a istration of dexamethasone was sepa test, the variables and criteria with the
course of 0.5 mg of dexamethasone oral rated by at least 2 days to allow for highest sensitivity, given 100% speci
ly every 6 hours for eight doses, start recovery between tests. ficity for the diagnosis of Cushing's syn
ing at noon 2 days before the dexa Data Analysis
drome, were identified and used for be
methasone-CRH test. Thus, the last dose tween-test comparisons. For the low-
of dexamethasone was administered at Data were analyzed on a Macintosh dose dexamethasone suppression test,
6 AM, 2 hours before the scheduled start Ilfx with Superanova and StatView II criteria were determined for urine free
of the dexamethasone-CRH test. Plas (Abacus Concepts Ine, Berkeley, Calif). cortisol and 17-hydroxycorticosteroid
ma cortisol and adrenocorticotropin lev Analysis of variance with repeated levels obtained on day 2 (no dexametha
els were obtained at the aforementioned measures (with logarithmic data trans sone) and day 4 (during dexamethasone),
times. The sensitivity for the adreno formation, where appropriate) was and for the percentage of suppression
corticotropin assay ranged from 0.9 to performed to detect between-group by dexamethasone on day 4 compared
2.2 pmol/L and for cortisol from 5.5 to 22 and time-related differences. Post hoe with day 2. The 17-hydroxycorticoste
nmol/L. The intra-assay and interassay tests were performed, where appro roid excretion was also corrected for cre
variabilities were 7% to 12% and 12% to priate, and interpreted with the atinine excretion: but since results were
Positive Negative
Specificity, Predictive
Sensitivity, Predictive Diagnostic
Test Variable Criterion_%_ %
Value, % Value, % Accuracy, %
Low-Dose Dexamethasone Test
17-0CHSday4 >11.0 / 74_ 69 84
17-OCHSday4 >14.6 /d 100 54 100 51 69
Urine free cortisol
day 4_ >100 nmol/d 100 56 100 53
CRH Test Without Dexamethasone Pretreatment
Cortisol sum of
post-CRH levels >3450nmol/L 100 64 100 76
- -1- -
CRH test peak 0 10 20 30 40 50 60
corticotropin >35.0 pmol/L 100 13 100
Dexamethasone-CRH Test
Basal cortisol
(before CRH) >38.0 nmol/L 100 90 100 84
Cortisol 15 min
after CRH >38.0 nmol/L 100 100t* 100 100tt 100
Corticotropin 30 min
after CRH >3.5 pmol/L 100 74 100 66 83
"Criteria for each test variable were derived from receiver operating characteristic analysis.26 Results are given
for the standard low-dose dexamethasone criterion of 17-hydroxycorticosteroid (17-OCHS) excretion levels greater
than 11 /d, and for criteria chosen to yield 100% specificity for the diagnosis of Cushing's syndrome. When the
tests are compared under these conditions, the dexamethasone-corticotropin-releasing hormone (CRH) test, 15-
minute cortisol value has a significantly greater sensitivity and diagnostic accuracy than any of the other tests.
tP<01, dexamethasone-CRH test 15-minute cortisol vs low-dose dexamethasone suppression test, or CRH test
performed without dexamethasone pretreatment. - -1-1-1-r
tP<.05, dexamethasone-CRH 15-minute cortisol vs dexamethasone-CRH test basal cortisol or dexamethasone- 0 10 20 30 40 50 60
CRH corticotropin. Time, min
not improved by this addition, these data responses between the two groups.
are not presented. For the CRH stim The low-dose dexamethasone test had Fig 3.Results of corticotropin-releasing hormone
stimulation test, without low-dose dexamethasone
ulation tests, criteria were established 74% specificity, 69% sensitivity, and 71%
pretreatment. Top, Plasma adrenocorticotropin val
for adrenocorticotropin and cortisol val diagnostic accuracy for the diagnosis of ues. Bottom, Plasma cortisol values. Means and
ues at each individual time point and for Cushing's syndrome when the standard SEMs are given. Definition of symbols as in Fig 1.
the peak response. Criteria were also criterion, 17-hydroxycorticosteroid ex
determined for the sum of all values cretion level greater than 11 /d on
during CRH tests, for all combinations the second day of dexamethasone ad during the CRH test, which yielded 13%
of the sums of the 5-, 15-, 30-, 45-, and ministration, was used (Fig 2 and Table sensitivity and 41% diagnostic accuracy
60-minute post-CRH time points, and 2). To achieve 100% specificity, a crite (Table 2). The best sensitivity (64%) with
100% specificity was obtained using the
for peak excursion of the response from rion for Cushing's syndrome based on
criterion of a sum of post-CRH cortisol
the baseline. Correlated 2x2 tables were 17-hydroxycorticosteroid level greater levels greater than 3450 nmol/L (Fig 4),
constructed for comparison of test cri than 14 /d was required; this yield
which was significantly better than the
teria, and Cochran Q statistics were in ed 54% sensitivity and 69% diagnostic
terpreted with the Bonferroni adjust accuracy (Table 2).
sensitivity based on the adrenocorti
ment for multiple comparisons. A diagnostic accuracy of 71% was cotropin criterion (P<.02).
achieved using a criterion for Cushing's Attempts to find criteria with im
RESULTS syndrome based on urine free cortisol proved diagnostic accuracy by combin
Low-Dose Dexamethasone level greater than 100 nmol/d (36 g/24 ing results from the low-dose dexa
methasone suppression test with those
Suppression Test h) on the second day of dexamethasone from the CRH test without dexametha
administration (Fig 2 and Table 2). With
Urine 17-hydroxycorticosteroid, free sone pretreatment did not improve di
this criterion, urine free cortisol had
cortisol, and creatinine measurements 100% specificity, but only 56% sensitiv agnostic accuracy over that found with
were available before and during dex the low-dose dexamethasone test alone.
amethasone administration from all 58 ity (P=.48 vs 17-hydroxycorticosteroid).
CRH Test With Dexamethasone
patients (Figs 1 and 2). Daily creatinine CRH Stimulation Test Without Pretreatment
measurements varied by no more than
17%, and the range of individual values Dexamethasone Pretreatment All patients completed the dexametha
was between 8.9 and 25.2 mmol/d (1.0 to Data were available from all patients sone-CRH test (Figs 5 and 6). Compared
2.8 g/24 h). Both urine 17-hydroxycor for the CRH test without dexametha with those who had pseudo-Cushing's
ticosteroid and free cortisol excretion sone (Figs 3 and 4). Although both states, the patients with Cushing's syn
were significantly greater in the group adrenocorticotropin and cortisol levels drome had significantly elevated corti
with Cushing's syndrome, whether or were significantly (P<.005) higher in sol (P<.001) and adrenocorticotropin
not dexamethasone was administered patients with Cushing's syndrome at all (P<.01) values at all time points, both
(P<.0001). Although the dexametha- time points, the overlap in individual before and after CRH administration.
sone-suppressed values ofboth urine 17- values was too great to afford diagnos The criterion of a basal plasma cor
hydroxycorticosteroid and free cortisol tic utility at any single point. The best tisol level greater than 38 nmol/L (1.4
excretion were significantly lower in pa criterion with 100% specificity using g/dL) during the dexamethasone-CRH
tients with pseudo-Cushing's states adrenocorticotropin was a value great test had 100% specificity, 90% sensitiv
(P<.001), there was a large overlap in er than 35 pmol/L at any time point ity, and 91% diagnostic accuracy (Table
s.
76% Diagnostic Accuracy 1000 100% Diagnostic Accuracy
100% Positive Predictive Value 100% Positive Predictive value
58% Negative Predictive Value 100% Negative Predictive Value
100
10l
1000
Cushing's Pseudo-Cushing's Cushing's Pseudo-Cushing's
Syndrome States Syndrome States
Fig 4.Criteria with the best diagnostic accuracy for corticotropin-releasing hor Fig 5.Criteria for best-diagnostic accuracy for dexamethasone-corticotropin-
mone (CRH) test sum of post-CRH plasma cortisol levels. Definition of symbols releasing hormone (CRH) test; plasma cortisol levels obtained 15 minutes af
as in Fig 1. Horizontal rule indicates criterion level. ter administration of CRH. Definition of symbols as in Fig 1. Horizontal rule in
dicates criterion level.
2). After administration of CRH, those nostic accuracy, 83% sensitivity, and In contrast to patients with cortisol-
patients with Cushing's disease whose 100% specificity. Thus, regardless of producing adrenal disorders or patients
cortisol levels were less than 38 nmol/L whether patients with adrenal disease with ectopie adrenocorticotropin-produc-
had increases in cortisol level, whereas were excluded, the dexamethasone- ing tumors, most individuals with Cush
those with pseudo-Cushing's states gen CRH test, 15-minute cortisol criterion ing's disease show some suppression of
erally did not. A single plasma cortisol had significantly greater sensitivity, neg adrenocorticotropin and cortisol by dex
level greater than 38 nmol/L, determined ative predictive value, and diagnostic amethasone and stimulation of adreno
15 minutes after administration of CRH, accuracy when compared with the best corticotropin and cortisol by CRH.
separated all cases of Cushing's syn dexamethasone-CRH adrenocorticotro However, cortisol production decreases
drome from all cases of pseudo-Cush pin criterion (P<.01). greatly in some patients with Cushing's
ing's states and yielded 100% sensitiv disease during low-dose dexamethasone
ity, specificity, and diagnostic accuracy Test Comparison administration, possibly because of di
(Fig 5). The 15-minute, post-CRH cor Criteria with 100% specificity and minished dexamethasone clearance.2829
tisol value had significantly greater di maximal sensitivity were used to com Such patients would be misclassified as
agnostic accuracy than the dexametha pare the tests (Table 2). The cortisol having pseudo-Cushing's states if eval
sone-CRH basal cortisol before CRH value of the dexamethasone-CRH test uated only by suppression of cortisol
administration (P<.05). At time points obtained 15 minutes after CRH admin production after low-dose dexametha
more than 15 minutes after administra istration had significantly greater sen sone administration. Thus, the dexa
tion of CRH, cortisol values exceeded sitivity, negative predictive value, and methasone-CRH test basal plasma cor
38 nmol/L in some patients with pseudo- diagnostic accuracy when compared ei tisol measurement greater than 38
Cushing's states and overlapped with ther to the low-dose dexamethasone test nmol/L correctly identified only 91% of
the responses seen in Cushing's syn (P<.01) or to the CRH stimulation test the patients in this series. By the addi
drome. However, excellent discrimina performed without dexamethasone ad tion ofCRH to stimulate greater adreno
tion was still found for the 30- and 45- ministration (P<.01). corticotropin and cortisol secretion (ei
minute cortisol samples (99% and 98% ther by pituitary or adrenal30 mecha
COMMENT nisms) in patients with Cushing's dis
diagnostic accuracy, respectively). The
highest cortisol level recorded for any The dexamethasone-CRH test, a new ease who are CRH-sensitive, those
patient in the pseudo-Cushing's group test for the differential diagnosis of hy patients with unusual sensitivity to
during the dexamethasone-CRH test was percortisolism, distinguished patients dexamethasone suppression achieved
96 nmol/L (3.5 g/dL) at 60 minutes. with mild Cushing's syndrome from adrenocorticotropin and cortisol levels
For the adrenocorticotropin levels those with pseudo-Cushing's states with above those of patients with pseudo-
during the dexamethasone-CRH test, a greater accuracy than was possible us Cushing's states. Thus, the dexametha
plasma adrenocorticotropin level that ex ing either the low-dose dexamethasone sone-CRH test, 15-minute cortisol test
ceeded 3.5 pmol/L at 30 minutes after test or the CRH test alone. The im correctly identified more of the patients
CRH administration offered the high provement in diagnostic accuracy of the with Cushing's syndrome than the dex
est diagnostic accuracy (83%) and had combined dexamethasone-CRH test amethasone-CRH test basal plasma cor
74% sensitivity and 100% specificity (Ta may be explicable by the pathophysiol- tisol, or the low-dose dexamethasone
ble 2). The plasma adrenocorticotropin ogy of Cushing's syndrome. Patients suppression test, either in our series or
value during the dexamethasone-CRH with primary adrenal pathology and in others.918 Administration of CRH fol
test was also examined when patients most patients with the syndrome of ec lowing dexamethasone suppression al
with adrenal causes of Cushing's syn topie adrenocorticotropin generally show lowed the criterion for the diagnosis of
drome were excluded (because their little feedback regulation of cortisol pro Cushing's syndrome to be 100% specific
adrenocorticotropin values are expect duction by glucocorticoid and little re with a much higher sensitivity than was
ed to be very low). Under these condi sponse to CRH.19 These patients are dis possible with the dexamethasone test
tions, a dexamethasone-CRH plasma criminated from patients with pseudo- alone (100% vs 71%, respectively).
adrenocorticotropin level greater than Cushing's states by virtue of their lack The dexamethasone-CRH test should
3.83 pmol/L obtained 30 minutes after of cortisol suppression by low-dose dex be reserved for those patients who the
administration of CRH had 89% diag- amethasone. clinician suspects have Cushing's syn-
DownloadedViewFrom:
publicationhttp://jama.jamanetwork.com/
stats by a National Institutes of Health User on 12/02/2015