`Series Editors: Alan Garber, MD, PhD, and Harold Sox, MD
`
`Academia and Clinic
`
`Diagnosis of Adrenal Insufficiency
`
`Richard I. Dorin, MD; Clifford R. Qualls, PhD; and Lawrence M. Crapo, MD, PhD
`
`Background: The cosyntropin stimulation test is the initial en-
`docrine evaluation of suspected primary or secondary adrenal in-
`sufficiency.
`
`Purpose: To critically review the utility of the cosyntropin stim-
`ulation test for evaluating adrenal insufficiency.
`
`Data Sources: The MEDLINE database was searched from 1966
`to 2002 for all English-language papers related to the diagnosis of
`adrenal insufficiency.
`
`Study Selection: Studies with fewer than 5 persons with pri-
`mary or secondary adrenal insufficiency or with fewer than 10
`persons as normal controls were excluded. For secondary adrenal
`insufficiency, only studies that stratified participants by integrated
`tests of adrenal function were included.
`
`Data Extraction: Summary receiver-operating characteristic
`(ROC) curves were generated from all studies that provided sen-
`sitivity and specificity data for 250-g and 1-g cosyntropin tests;
`these curves were then compared by using area under the curve
`(AUC) methods. All estimated values are given with 95% CIs.
`
`Data Synthesis: At a specificity of 95%, sensitivities were 97%,
`
`Adrenal insufficiency is an uncommon clinical disorder
`
`that results from an inadequate basal or stress level of
`plasma cortisol. It is important to diagnose adrenal insuf-
`ficiency because the disorder may be fatal if left unrecog-
`nized or untreated. With diagnosis and appropriate adre-
`nocortical hormone replacement, normal quality of life and
`longevity can be achieved. The presentation of adrenal in-
`sufficiency may be insidious and thus difficult to recognize.
`Once suspected, however, the definitive diagnosis can be
`confirmed by laboratory evaluation of adrenocortical func-
`tion.
`Although many different tests for adrenal insufficiency
`have been developed, few have been adequately studied
`and many are inconvenient for testing in the outpatient
`clinical setting. By contrast, the cosyntropin stimulation
`test is widely used in many different clinical settings and is
`easy to perform. In addition, data on test performance in
`various clinical settings are plentiful. The cosyntropin
`stimulation test has therefore emerged as the initial test
`used to evaluate patients for both primary and secondary
`adrenal insufficiency.
`
`METHODS
`We reviewed all English-language studies in humans
`identified in the MEDLINE database (1966 to 2002)
`through the Ovid search service. Search terms were adrenal
`gland hypofunction restricted to diagnosis. For the normal
`
`194 © 2003 American College of Physicians
`
`57%, and 61% for summary ROC curves in tests for primary
`adrenal insufficiency (250-g cosyntropin test), secondary adrenal
`insufficiency (250-g cosyntropin test), and secondary adrenal
`insufficiency (1-g cosyntropin test), respectively. The area under
`the curve for primary adrenal
`insufficiency was significantly
`greater than the AUC for secondary adrenal insufficiency for the
`high-dose cosyntropin test (P< 0.001), but AUCs for the 250-g
`and 1-g cosyntropin tests did not differ significantly (P> 0.5)
`for secondary adrenal insufficiency. At a specificity of 95%, sum-
`mary ROC analysis for the 250-g cosyntropin test yielded a
`positive likelihood ratio of 11.5 (95% CI, 8.7 to 14.2) and a
`negative likelihood ratio of 0.45 (CI, 0.30 to 0.60) for the diag-
`nosis of secondary adrenal insufficiency.
`
`Conclusions: Cortisol response to cosyntropin varies consider-
`ably among healthy persons. The cosyntropin test performs well in
`patients with primary adrenal insufficiency, but the lower sensi-
`tivity in patients with secondary adrenal insufficiency necessitates
`use of tests involving stimulation of the hypothalamus if the
`pretest probability is sufficiently high. The operating characteris-
`tics of the 250-g and 1-g cosyntropin tests are similar.
`
`Ann Intern Med. 2003;139:194-204.
`For author affiliations, see end of text.
`
`www.annals.org
`
`response to high-dose cosyntropin, we selected studies with
`10 or more participants. For the diagnosis of primary ad-
`renal insufficiency, we selected studies with 5 or more par-
`ticipants. For evaluation of the sensitivity and specificity of
`cosyntropin tests in secondary adrenal insufficiency, we se-
`lected only studies that stratified all participants with sus-
`pected adrenal insufficiency by integrated tests of adrenal
`function (insulin tolerance or metyrapone tests).
`(ROC)
`Summary
`receiver-operating characteristic
`curves were developed from sensitivity and specificity val-
`ues derived from individual studies, as described by Litten-
`berg, Moses, and colleagues (1, 2) (see the Appendix, avail-
`able at www.annals.org, for detailed formulas). Summary
`ROC curves were compared by using area under the curves
`(AUCs), as described by Walter (3). For our data sets, we
`verify the condition (B ⬵ 0; see the Appendix, available at
`www.annals.org) that yields explicit formulas for AUC and
`its CI for the summary ROC curves. The slope parameter
`(B) did not differ significantly from 0 for all data sets used
`to generate summary ROC curves.
`We compared ROC curves for data paired by individ-
`ual participants using likelihood methods with a program
`(ROCKIT 0.9B) developed by Metz and colleagues (4)
`(available at www-radiology.uchicago.edu/cgi-bin/software
`.cgi).
`The funding source had no role in the design, con-
`duct, or reporting of the study or in the decision to submit
`the manuscript for publication.
`
`Downloaded From: http://annals.org/ by a UCLA Digital Collection Svs User on 08/17/2016
`
`
`
`
`JANSSEN EXHIBIT 2051
`Mylan v. Janssen IPR2016-01332
`
`
`
`Diagnosis of Adrenal Insufficiency
`
`Academia and Clinic
`
`DATA SYNTHESIS
`High-Dose Cosyntropin Stimulation Test
`The standard cosyntropin test is performed by admin-
`istering one ampule (250 g) of cosyntropin intramuscu-
`larly or intravenously and measuring serum or plasma cor-
`tisol
`levels 30 to 60 minutes
`later. With a normal
`(negative) test result, the serum cortisol level after cosyn-
`tropin stimulation is generally greater than 500 nmol/L. A
`subnormal cortisol response (⬍500 nmol/L) is defined as a
`positive test result and indicates an increased probability of
`either primary or secondary adrenal insufficiency. The co-
`syntropin test may be performed at any time of the day. In
`patients with suspected adrenal
`insufficiency, a basal
`plasma cortisol level is not usually necessary because nei-
`ther the absolute nor the percentage change from the basal
`level is useful as a diagnostic criterion for the cosyntropin
`test (5). However, in the absence of corticosteroid-binding
`globulin deficiency, an unstimulated serum cortisol level,
`determined between 6:00 and 8:00 a.m., may be helpful
`because a level less than 80 nmol/L strongly suggests adre-
`nal insufficiency (5).
`
`Normal Response to the High-Dose Cosyntropin Test
`In healthy persons without evidence of adrenal insuf-
`ficiency, serum cortisol response 30 or 60 minutes after
`250 g of cosyntropin is administered intramuscularly or
`intravenously has been studied extensively (6 –22). The re-
`sponses to intramuscular and intravenous injections are
`similar, and the responses among normal persons vary. In
`10 studies that included a total of 288 participants and that
`reported the entire range of postcosyntropin serum cortisol
`levels, the levels ranged from 415 to 2200 nmol/L (9, 10,
`12–15, 17, 19 –21). The broad range of normal responses
`to cosyntropin stimulation reflects various factors, includ-
`ing differences in hypothalamic–pituitary–adrenal axis set
`point, serum corticosteroid-binding globulin level, stress
`level, body composition, time of testing, and performance
`characteristics of the cortisol assay used.
`In one detailed study of 100 healthy persons, the dis-
`tribution curves of serum cortisol levels obtained 30 and
`60 minutes after a 250-g intramuscular injection of co-
`syntropin displayed a non-Gaussian configuration for each
`of
`four separate cortisol assays, with the distribution
`skewed to the right toward higher cortisol levels (22). The
`5th percentile lower cortisol cutoff limit for these four as-
`says ranged from 510 to 615 nmol/L at 30 minutes and
`from 620 to 675 nmol/L at 60 minutes. Other studies also
`show increases in the cortisol response at 60 minutes com-
`pared with 30 minutes (16, 18, 20). In 11 studies involv-
`ing 340 healthy participants, the data presented as the
`mean minus 2 SDs show lower limits ranging from 390 to
`620 nmol/L at 30 minutes (6 –10, 16 –20) and from 500
`to 725 nmol/L at 60 minutes (11, 16, 18, 20). Because the
`distribution curve is non-Gaussian, no conclusion can be
`drawn from these studies about the percentage of healthy
`
`www.annals.org
`
`persons with serum cortisol levels less than the lower cutoff
`limit.
`The studies described show that an appreciable num-
`ber of normal persons will have a postcosyntropin cortisol
`level less than a cutoff limit of 500 nmol/L. However, none
`of the 288 participants in the 10 studies described earlier
`(in which the entire range of cortisol responses was re-
`ported) had a cortisol level less than 415 nmol/L.
`Diagnosis of Primary Adrenal Insufficiency
`Primary adrenal
`insufficiency (often called Addison
`disease) is an uncommon disorder that often presents with
`a slowly progressive increase in nonspecific symptoms. The
`prevalence of this disorder in the community is approxi-
`mately 100 cases per 1 million people (23–26); the inci-
`dence is 5 cases per year per 1 million people (26). The
`prevalence of primary adrenal insufficiency is higher (al-
`though not precisely known) in persons with HIV disease,
`family histories of adrenoleukodystrophy, autoimmune en-
`docrine disorders, metastatic cancer, and granulomatous
`disease.
`The prevalence among persons with nonspecific symp-
`toms, such as tiredness,
`fatigue, weakness,
`listlessness,
`weight loss, nausea, and anorexia, is not known. More spe-
`cific symptoms, such as unexplained darkening of the skin,
`orthostatic dizziness, and salt-craving, may not be among
`presenting symptoms.
`
`Cosyntropin Stimulation Tests in Primary Adrenal Insufficiency
`Table 1 summarizes the results of 8 studies in which
`122 patients with primary adrenal insufficiency and con-
`trols were given 250 g of cosyntropin intravenously or
`intramuscularly and the serum cortisol levels were mea-
`sured 30 or 60 minutes later. None of the patients in these
`studies underwent consecutive prospective evaluation for
`adrenal insufficiency; rather, they were selected for study
`either because previous evaluation showed that they had
`typical Addison disease (13, 14, 20, 27–29) or because
`their cosyntropin tests were compared with historical con-
`trols in retrospective surveys (23, 30). Controls in these
`studies varied from healthy volunteers (13, 14, 23) to par-
`ticipants with nonendocrine illness (14, 27) or suspected
`adrenal insufficiency (29). Thus, case-patients and controls
`were not recruited from the same setting. In general, the
`case-patients with Addison disease in these studies were
`selected on the basis of typical clinical and nonendocrine
`laboratory criteria, such as hyperkalemia, supplemented in
`many cases with elevated plasma adrenocorticotropic hor-
`mone (ACTH) levels and low urine steroid responses to
`intravenous ACTH infusions. In several retrospective anal-
`yses using historical controls, cosyntropin tests may have
`contributed to the diagnosis of Addison disease, but several
`patients with Addison disease in each of these surveys had
`normal cosyntropin test results. None of the studies indi-
`cated that patients with borderline cosyntropin test results
`were selectively excluded. However, it is clear that the cases
`
`5 August 2003 Annals of Internal Medicine Volume 139 • Number 3 195
`
`Downloaded From: http://annals.org/ by a UCLA Digital Collection Svs User on 08/17/2016
`
`
`
`Academia and Clinic Diagnosis of Adrenal Insufficiency
`
`Table 1. The 250-g Cosyntropin Stimulation Test in Patients with Primary Adrenal Insufficiency*
`
`Study (Reference)†
`
`Cosyntropin
`Route and Time
`after Injection‡
`
`Serum Cortisol
`Cutoff Level
`
`Sensitivity§
`
`Specificity§
`
`Positive
`Likelihood
`Ratio㥋
`
`Negative
`Likelihood
`Ratio㥋
`
`Speckart et al. (27)
`Nelson and Tindall (14)
`Oelkers et al. (28)
`Fiad et al. (29)
`Kong and Jeffcoate (23)
`Gonzalez-Gonzalez et al. (20)
`Soule (30)
`Speckart et al. (27)
`Dluhy et al. (13)
`Oelkers et al. (28)
`Kong and Jeffcoate (23)
`Gonzalez-Gonzalez et al. (20)
`
`min
`
`IV, 60
`IV, 60
`IM, 60
`IV, 60
`IV, 60
`IV, 60
`IV, 60
`IV, 30
`IM, 30
`IM, 30
`IV, 30
`IV, 30
`
`nmol/L
`
`% (n/n)
`
`415
`415
`415
`415
`415
`415
`415
`415
`415
`415
`415
`415
`
`100 (6/6)
`100 (7/7)
`100 (41/41)
`100 (12/12)
`75 (6/8)
`82 (9/11)
`95 (35/37)
`100 (6/6)
`100 (5/5)
`100 (41/41)
`89 (16/18)
`82 (9/11)
`
`100 (9/9)
`100 (69/69)
`–
`100 (55/55)
`–
`100 (46/46)
`–
`
`88 (7/8)
`100 (12/12)
`–
`–
`100 (46/46)
`
`⬎100
`⬎100
`–
`⬎100
`–
`⬎100
`–
`
`8.3
`⬎100
`–
`–
`⬎100
`
`0
`0
`–
`0
`–
`0.18
`–
`0
`0
`–
`–
`0.18
`
`* IM ⫽ intramuscular; IV ⫽ intravenous.
`† In six studies (13, 14, 20, 27–29), cases of typical Addison disease (proven by clinical criteria, low urine steroids levels, or high serum adrenocorticotropic hormone levels)
`were selected for cosyntropin testing from outpatient clinics. Two studies (23, 30) are retrospective surveys of patients with suspected Addison disease who had cosyntropin
`testing and were compared with historical controls. Control groups were historical (23, 28, 30), healthy volunteers (13, 14, 20), persons with nonendocrine illness (14, 27),
`or persons with suspected adrenal insufficiency with a normal metyrapone test result (29).
`‡ Time after injection is when the serum cortisol is drawn in minutes after the 250-g cosyntropin injection.
`§ Sensitivity is the percentage calculated from raw data (shown in parentheses) indicating the number of persons with positive cosyntropin test results among true-positive
`persons. Specificity is the percentage calculated from raw data (shown in parentheses), indicating the number of persons with negative cosyntropin test results among
`true-negative persons.
`㛳 Definitions of positive and negative likelihood ratios are shown in equation A2 in the Appendix (available at www.annals.org).
`
`of Addison disease selected in these studies were more ad-
`vanced and easily recognized by well-established clinical
`and laboratory criteria. Thus, in most cases in these studies,
`the diagnosis of Addison disease was based on clinical evi-
`dence supported by serum electrolyte, plasma ACTH, and
`urine steroid levels. Cosyntropin tests were then performed
`in these patients, and the results were interpreted indepen-
`dently of the original diagnostic criteria.
`For the summary ROC curve, which is based on four
`of the studies in Table 1 (14, 20, 27, 29), the point on the
`summary ROC where sensitivity and specificity are equal
`was 96.5% (95% CI, 94.5% to 98.5%) for the diagnosis of
`primary adrenal insufficiency. When specificity is set at
`95%, this summary ROC curve yields a sensitivity of
`97.5% (CI, 95% to 100%), with a corresponding positive
`likelihood ratio of 19.5 (CI, 19.0 to 20.0) and a negative
`likelihood ratio of 0.026 (CI, 0 to 0.6). The AUC for this
`summary ROC curve was 0.99 (CI, 0.985 to 1.000), indi-
`cating excellent test discrimination.
`As a result of the selection bias in these studies toward
`patients with severe Addison disease, the cosyntropin test
`performance characteristics derived from Table 1 are most
`applicable to such patients. Patients with mild Addison
`disease or subclinical Addison disease probably have cosyn-
`tropin test performance characteristics that would be con-
`siderably less robust than those in Table 1. After a positive
`cosyntropin test result, the diagnosis of primary adrenal
`insufficiency may be confirmed by an elevation of plasma
`ACTH level (5, 28), whereas patients with secondary ad-
`renal insufficiency typically have normal or low plasma
`ACTH levels.
`
`196 5 August 2003 Annals of Internal Medicine Volume 139 • Number 3
`
`Problems of Diagnosis in Primary Adrenal Insufficiency
`Diagnosis of Mild Primary Adrenal Insufficiency. One
`difficulty in the diagnosis of primary adrenal insufficiency
`is the nonspecific nature of presentation and the resultant
`lack of clinical suspicion for the disorder. There is a con-
`tinuum of adrenal insufficiency ranging from subclinical
`hypoadrenalism (characterized by a normal cortisol re-
`sponse to cosyntropin and an elevated basal or corticotropin-
`releasing hormone–stimulated plasma ACTH level) to
`overt primary adrenal insufficiency (characterized by a neg-
`ligible cortisol response to cosyntropin and a very high
`plasma ACTH level). Most of the patients with primary
`adrenal insufficiency in Table 1 had cortisol responses to
`cosyntropin substantially less than 275 nmol/L, which
`poses no problem in laboratory diagnosis. However, several
`patients in Table 1 had a normal response to cosyntropin,
`with cortisol levels greater than 550 nmol/L and simulta-
`neously high plasma ACTH levels. These patients clinically
`improved after receiving glucocorticoid therapy. Longitu-
`dinal follow-up of patients with subclinical hypoadrenal-
`ism who were identified among the patients with HIV
`disease, adrenal autoantibodies, or a family history of ad-
`renoleukodystrophy or adrenomyeloneuropathy (32–35)
`demonstrates progression to overt primary adrenal insuffi-
`ciency in some patients (33). Thus, the cortisol response to
`cosyntropin depends on the degree of adrenal gland failure,
`and the sensitivity of the cosyntropin stimulation test de-
`pends on whether patients have mild or severe primary
`adrenal insufficiency.
`Because patients with mild primary adrenal insuffi-
`ciency sometimes have a normal cosyntropin stimulation
`
`www.annals.org
`
`Downloaded From: http://annals.org/ by a UCLA Digital Collection Svs User on 08/17/2016
`
`
`
`Diagnosis of Adrenal Insufficiency
`
`Academia and Clinic
`
`test result (20, 23, 30), other tests, such as the plasma
`ACTH– cortisol ratio or the plasma renin activity–aldoste-
`rone ratio in paired blood samples (28), may be appropri-
`ate. However, few studies of this type have been reported,
`and the renin–aldosterone ratio is elevated in other, more
`common medical conditions.
`Diagnosis of Primary Adrenal Insufficiency in Acute Set-
`tings. The variability of basal serum cortisol and cosyn-
`tropin-stimulated serum cortisol levels is even greater in
`acutely ill persons than in healthy persons; basal
`levels
`range from 140 to 11 000 nmol/L (36 – 63). Measure-
`ments of cortisol levels in critically ill patients in intensive
`care or the emergency department (36 – 48), patients with
`sepsis or septic shock (49 –57), and surgical patients in the
`postoperative period (58 – 63) show a broad range of cor-
`tisol responses to stress and to cosyntropin; therefore, de-
`termining which patients have adrenal insufficiency is not
`straightforward. The problem of diagnosis is particularly
`difficult in patients with well-documented septic shock, as
`illustrated in one study in which almost 20% of the sur-
`viving patients with sepsis had initial basal cortisol levels
`less than 275 nmol/L and cosyntropin-stimulated levels
`
`less than 500 nmol/L (56). Subsequently, all survivors
`demonstrated a normal response to cosyntropin. Thus, the
`diagnosis of adrenal insufficiency in the acute setting is ex-
`ceedingly difficult.
`Postcosyntropin Cortisol Cutoff Level. As will be dis-
`cussed, a higher cortisol cutoff level is required to achieve a
`reasonable level of sensitivity in secondary adrenal insuffi-
`ciency. Therefore, if the diagnostic application of the co-
`syntropin test can be restricted to primary adrenal insuffi-
`ciency on the basis of clinical or nonsteroid laboratory
`findings, it may be useful to use a lower cortisol cutoff
`level, such as 415 nmol/L (Table 1). In clinical practice,
`this distinction is not always possible and a higher cortisol
`cutoff level (500 to 600 nmol/L) should be applied to
`achieve reasonable sensitivity for secondary adrenal insuffi-
`ciency. The risk of a higher cutoff level for primary adrenal
`insufficiency is a false-positive cosyntropin test result, lead-
`ing to potentially lifelong, physiologic corticosteroid re-
`placement therapy for the euadrenal patient. This can be
`avoided by using adjunctive tests, such as the plasma
`ACTH test, to confirm the diagnosis of primary adrenal
`insufficiency.
`
`Table 2. Usefulness of the 250-g Cosyntropin Stimulation Test in Patients Who Are Taking Glucocorticoids or Have
`Pituitary Disease*
`
`Study (Reference)†
`
`Cosyntropin
`Route and Time
`after Injection‡
`
`Serum Cortisol or Deoxycortisol Cutoff
`Level after Stimulation§
`
`Sensitivity¶
`
`Specificity¶
`
`Positive
`Likelihood
`Ratio**
`
`Negative
`Likelihood
`Ratio**
`
`Kehlet et al. (73)
`Lindholm et al. (74)
`Cunningham et al. (75)
`Lindholm and Kehlet (76)
`Stewart et al. (77)
`Hartzband et al. (78)
`Jackson et al. (79)
`Tordjman et al. (80)
`Kane et al. (81)
`Hurel et al. (16)
`Rasmuson et al. (82)
`Ammari et al. (83)
`Orme et al. (84)
`Mukherjee et al. (85)
`Weintrob et al. (19)
`Mayenknecht et al. (18)
`Bangar and Clayton (86)
`Talwar et al. (87)
`Abdu et al. (31)
`Suliman et al. (88)
`
`min
`
`IV, 30
`IV, 30
`IM, 60
`IV, 30
`IM, 30
`IV, peak
`IV, 30
`IV, 30
`IM, 30
`IV, 30
`IV, peak
`IV, 30
`IM, peak
`IM, 30
`IV, peak
`IV, 30
`IV, 30
`IV, peak
`IV, 30
`IV, 30
`
`ITT㥋
`
`MT㥋
`
`Cosyntropin Test
`
`nmol/L
`
`% (n/n)
`
`500
`500
`550
`500
`500
`500
`550
`500
`500
`520
`500
`550
`500
`580
`520
`550
`500
`550
`500
`–
`
`500
`500
`500
`500
`550
`500
`550
`550
`500
`385
`550
`550
`550
`580
`520
`620
`600
`550
`500
`500
`
`175
`
`200
`
`200
`
`200
`
`90 (9/10)
`85 (29/34)
`40 (8/20)
`73 (19/26)
`90 (9/10)
`80 (8/10)
`69 (9/13)
`50 (8/16)
`100 (9/9)
`33 (20/60)
`81 (13/16)
`47 (8/17)
`83 (5/6)
`71 (5/7)
`90 (9/10)
`65 (15/23)
`85 (17/20)
`54 (7/13)
`100 (12/12)
`67 (10/15)
`
`87 (13/15)
`96 (54/56)
`100 (15/15)
`99 (135/136)
`85 (51/60)
`100 (13/13)
`100 (11/11)
`89 (33/37)
`69 (9/13)
`95 (101/106)
`91 (10/11)
`85 (11/13)
`60 (6/10)
`91 (10/11)
`100 (20/20)
`95 (20/21)
`96 (47/49)
`100 (11/11)
`90 (27/30)
`100 (36/36)
`
`6.9
`21.3
`⬎100
`73
`6.0
`⬎100
`⬎100
`4.5
`3.2
`6.6
`9.0
`2.6
`2.1
`6.3
`⬎100
`13.0
`21.2
`⬎100
`10.0
`⬍100
`
`0.11
`0.16
`0.58
`0.27
`0.12
`0.20
`0.31
`0.56
`0.00
`0.71
`0.21
`0.65
`0.28
`0.41
`0.10
`0.37
`0.16
`0.46
`0.00
`0.33
`
`* IM ⫽ intramuscular; ITT ⫽ insulin tolerance test; IV ⫽ intravenous; MT ⫽ overnight metyrapone test.
`† All studies are prospective except two retrospective reviews (16, 86). In five studies, most of the patients with suspected adrenal insufficiency had excessive glucocorticoid
`exposure (75, 78, 81, 87, 88). Otherwise, patients with suspected adrenal insufficiency had known or suspected hypothalamic or pituitary disease. Two studies included
`consecutive patients (76, 83).
`‡ Time after injection is when serum cortisol is drawn after the 250-g cosyntropin injection. Peak denotes the time (usually 60 minutes) at which the serum cortisol level
`is maximal.
`§ All MT values are for deoxycortisol. In one study (75), the MT cutoff level for deoxycortisol is 175 nmol/L, and in three studies (18, 80, 88), it is 200 nmol/L. In one
`study (80), if a postcosyntropin cortisol cutoff level of 500 nmol/L is applied, the sensitivity is only 6% (1/16); from the receiver-operating characteristic curve of Tordjman
`and colleagues (80), we have selected a cutoff level of 550 nmol/L, which yields a sensitivity of 50%.
`㛳 Diagnostic reference standard for secondary adrenal insufficiency.
`¶ Sensitivity is the percentage calculated from raw data (shown in parentheses) indicating the number of persons with positive cosyntropin test results among true-positive
`persons (as defined by a metyrapone or insulin tolerance test). Specificity is the percentage calculated from raw data (shown in parentheses), indicating the number of persons
`with negative cosyntropin test results among true-negative persons.
`** Definitions of positive and negative likelihood ratios are shown in equation A2 in the Appendix (available at www.annals.org).
`
`www.annals.org
`
`5 August 2003 Annals of Internal Medicine Volume 139 • Number 3 197
`
`Downloaded From: http://annals.org/ by a UCLA Digital Collection Svs User on 08/17/2016
`
`
`
`Academia and Clinic Diagnosis of Adrenal Insufficiency
`
`Figure 1. Summary receiver-operating characteristic (SROC)
`curves for high-dose (250-g) and low-dose (1-g)
`cosyntropin tests in secondary adrenal insufficiency.
`
`The SROC curve for the high-dose cosyntropin test was derived from
`SROC analysis of 20 independent studies (Table 2), where each point
`(white circles) represents an individual study. The SROC curve for the
`low-dose cosyntropin test was derived from 9 independent studies (Table
`4), where each point (white squares) represents an individual study.
`
`Establishing the Cause of Primary Adrenal Insufficiency
`It is important to search for the cause of primary ad-
`renal insufficiency after the diagnosis is determined. Of
`particular interest are treatable disorders, such as tubercu-
`losis and other granulomatous diseases, as well as HIV dis-
`ease and its associated infections. In addition to careful
`investigation of family history, medical history, and clinical
`evaluation, it may be useful to perform specific laboratory
`studies, such as determining very-long-chain fatty acid lev-
`els to confirm adrenoleukodystrophy or adrenomyeloneu-
`ropathy or determining antiadrenal antibodies to confirm
`an autoimmune cause. Imaging procedures, such as chest
`radiography, adrenal computed tomography, or magnetic
`resonance imaging, may help establish the cause of adrenal
`insufficiency; adrenal biopsy to establish cause is appropri-
`ate in selected cases.
`Diagnosis of Secondary Adrenal Insufficiency
`The prevalence of secondary adrenal insufficiency is
`much higher than that of primary adrenal insufficiency,
`primarily because of the common use of glucocorticoid
`hormones. In patients who have taken moderate to high
`doses of exogenous glucocorticoid for long periods, the
`prevalence of secondary adrenal
`insufficiency can be as
`high as 50%. Secondary adrenal insufficiency occurs in
`about 30% of patients who have a pituitary macroadenoma
`or who have had a transsphenoidal hypophysectomy or
`pituitary irradiation; secondary adrenal insufficiency always
`occurs after the surgical cure of Cushing syndrome but is
`generally not permanent.
`
`198 5 August 2003 Annals of Internal Medicine Volume 139 • Number 3
`
`Nonprovocative tests, such as measuring morning se-
`rum cortisol levels or an overnight urine-free cortisol incre-
`ment (64), seem to have limited sensitivity for secondary
`adrenal insufficiency. Provocative tests, which use a physi-
`ologic stimulus to cortisol secretion, include both compo-
`nent and integrated tests. Component tests include the
`rapid high-dose or low-dose infusion of cosyntropin, which
`acts directly on the adrenal cortex to stimulate cortisol se-
`cretion, and intravenous infusion of corticotropin-releasing
`hormone, which acts directly on the pituitary to release
`ACTH (5, 65– 69). Integrated tests require contributions
`of all three components of the hypothalamic–pituitary–
`adrenal axis to activate cortisol secretion. Integrated tests
`use a central stimulus, hypoglycemia, in the insulin toler-
`ance test (5,70 –72) and a decrease in serum cortisol in the
`metyrapone test (5, 29, 70) to activate release of hypotha-
`lamic corticotropin-releasing hormone, vasopressin, and
`other ACTH secretagogues. Integrated tests require more
`time and experience to perform and are generally consid-
`ered to be the “gold standard” against which simpler com-
`ponent tests are compared.
`
`High-Dose (250-g) Cosyntropin Test in Secondary
`Adrenal Insufficiency
`The 250-g cosyntropin stimulation test is useful in
`the diagnosis of secondary adrenal insufficiency because the
`adrenal cortex atrophies when ACTH is deficient. The du-
`ration and degree of ACTH deficiency determine the de-
`gree of atrophy.
`Table 2 summarizes 20 studies in which all patients
`with suspected secondary adrenal insufficiency underwent
`both a 250-g cosyntropin stimulation test and an insulin
`tolerance test or metyrapone test (16, 18, 19, 31, 73– 88).
`In general, these studies are better designed than those for
`primary adrenal
`insufficiency because case-patients and
`controls are recruited from the same setting and have a
`continuous range of abnormality. Therefore, these studies
`of secondary adrenal insufficiency do not tend to overesti-
`mate test performance to the degree seen in the studies of
`primary adrenal insufficiency.
`Figure 1 shows summary ROC analysis of the 250-g
`cosyntropin stimulation test in secondary adrenal insuffi-
`ciency. When sensitivity and specificity are equal, the sum-
`mary ROC curve yields an overall sensitivity and specificity
`of 83.5% (CI, 79.6% to 87.4%); the AUC is 0.90 (CI,
`0.87 to 0.94). When specificity is set at 95%, the summary
`ROC curve for the 250-g cosyntropin test yields a sensi-
`tivity of 57% (CI, 44% to 71%), with a corresponding
`positive likelihood ratio of 11.5 (CI, 8.7 to 14.2) and a
`negative likelihood ratio of 0.45 (CI, 0.30 to 0.60).
`Thus, at clinically useful cutoff levels (postcosyntropin
`cortisol level, 500 to 600 nmol/L), where specificity is ap-
`proximately 95%, a positive cosyntropin test result sub-
`stantially increases the likelihood that the patient has sec-
`ondary adrenal
`insufficiency. This is influenced by the
`
`www.annals.org
`
`Downloaded From: http://annals.org/ by a UCLA Digital Collection Svs User on 08/17/2016
`
`
`
`Diagnosis of Adrenal Insufficiency
`
`Academia and Clinic
`
`Table 3. Bayes Theorem in Testing for Secondary Adrenal
`Insufficiency*
`
`Pretest Probability
`of Secondary
`Adrenal
`Insufficiency
`
`Post-Test Probability of
`Secondary Adrenal
`Insufficiency after a
`Normal (Negative)
`Cosyntropin Stimulation
`Test Result (95% CI)
`
`Post-Test Probability of
`Secondary Adrenal
`Insufficiency after an
`Abnormal (Positive)
`Cosyntropin Stimulation
`Test Result (95% CI)
`
`1
`5
`10
`25
`50
`75
`90
`
`0.3 (0.2–0.4)
`1.5 (0.9–2.1)
`3.1 (1.9–4.4)
`8.8 (5.5–12.1)
`22.5 (15.2–29.7)
`46.5 (36.1–56.8)
`72.3 (63.9–80.6)
`
`%
`
`7.0 (6.0–7.9)
`28.0 (25.1–31.0)
`45.1 (41.5–48.7)
`71.1 (68.1–74.1)
`88.1 (86.6–89.6)
`95.7 (95.1–96.3)
`98.5 (98.3–98.7)
`
`* See the Appendix (available at www.annals.org), which describes how to use
`Bayes theorem to calculate the post-test probability of secondary adrenal insuffi-
`ciency based on a likelihood ratio.
`
`pretest probability of disease, as indicated by using Bayes-
`ian analysis (Table 3). Conversely, a negative (normal) test
`result only modestly decreases the likelihood that the pa-
`tient has secondary adrenal insufficiency (Table 3), partic-
`ularly if the pretest probability is high. Thus, the 250-g
`cosyntropin test is helpful for ruling in but not ruling out
`secondary adrenal insufficiency. The data that demonstrate
`limited sensitivity for the high-dose cosyntropin test in sec-
`ondary adrenal insufficiency suggest that when the pretest
`probability of adrenal insufficiency is high and the cosyn-
`tropin test result is normal, additional evaluation using
`tests with better sensitivity should be performed.
`
`Comparison of High-Dose Cosyntropin Test Performance in
`Primary and Secondary Adrenal Insufficiency
`Comparison of the AUC for summary ROC curves for
`the high-dose cosyntropin test in primary (Table 1) and
`secondary (Table 2) adrenal insufficiency showed signifi-
`cantly (P ⬍ 0.001) better performance in the clinical set-
`ting of primary adrenal
`insufficiency (AUC, 0.99 [CI,
`0.985 to 1.000]) than in secondary adrenal insufficiency
`(AUC, 0.90 [CI, 0.76 to 0.97]).
`
`Low-Dose Cosyntropin Tests in Secondary Adrenal Insufficiency
`Dose-response studies in normal persons indicate that
`cosyntropin doses as low as 0.5 to 1 g will give a near-
`maximal cortisol response within 15 to 30 minutes (89 –
`94). The performance characteristics of the low-dose 1-g
`cosyntropin stimulation test could be superior to the con-
`ventional-dose 250-g test for diagnosing secondary adre-
`nal insufficiency because the plasma ACTH level is closer
`to the physiologic range (18, 90, 95–97). However, recent
`reviews comparing these two tests offer conflicting conclu-
`sions (98 –103). Several investigators have performed the
`1-g cosyntropin stimulation test, which requires intrave-
`nous administration and timed blood sampling to obtain
`the peak cortisol response, in patients with suspected sec-
`ondary adrenal insufficiency (18, 19, 31, 80, 82, 87, 88,
`104, 105) (Table 4).
`Receiver-operating characteristic curves, which provide
`an analysis of test performance over a range of cortisol
`cutoff levels, have been developed to directly compare the
`high-dose and low-dose tests. In an analysis by Abdu and
`colleagues (31, 106), the performance characteristics of the
`
`Table 4. Usefulness of 1-g Cosyntropin Stimulation Test in Patients Who Are Taking Glucocorticoids or Have Pituitary Disease*
`
`Study (Reference)
`
`Cosyntropin
`Route and Time
`after Injection†
`
`Serum Cortisol or Deoxycortisol
`C