0021-972X/00/$03.00/0 The Journal of Clinical Endocrinology & Metabolism Copyright © 2000 by The Endocrine Society

Vol. 85, No. 1 Printed in U.S.A.

Cortisol Production Rates in Subjects with Suspected Cushing’s Syndrome: Assessment by Stable Isotope Dilution Methodology and Comparison to Other Diagnostic Methods* M. H. SAMUELS, D. D. BRANDON, L. M. ISABELLE, D. M. COOK, K. E. GRAHAM, J. Q. PURNELL, AND D. L. LORIAUX Division of Endocrinology, Diabetes, and Clinical Nutrition (M.H.S., D.D.B., D.M.C., K.E.G., D.L.L.), Oregon Health Sciences University, Portland, Oregon 97201; Department of Chemical and Biological Sciences (L.M.I.), Oregon Graduate Institute of Sciences and Technology, Beaverton, Oregon 97006; and Division of Metabolism, Endocrinology, and Nutrition (J.Q.P.), University of Washington School of Medicine, Seattle, Washington 98195 ABSTRACT It can be difficult to establish the diagnosis of Cushing’s Syndrome (CS) in patients with mild or nonspecific clinical and biochemical findings, because available diagnostic tests have limited predictive values. We hypothesized that measurement of 24-h cortisol production rates (CPRs) might be a more sensitive indicator of CS in such patients. We measured CPRs in 28 patients with suspected CS (but equivocal biochemical findings) and in 22 healthy control subjects, by infusing tracer amounts of deuterated cortisol, with simultaneous measurements of 24-h urine free cortisol (UFC) levels; and we frequently sampled serum cortisol levels. CPRs were calculated from the ratio of isotopic enrichment to isotopic dilution of cortisol measured by gas chromatography-negative ion chemical ionization mass spectrometry. Nine of the patients proved to have CS by surgery (CS-Yes),

whereas 19 patients were determined not to have CS by biochemical testing (CS-No). Mean 24-h UFCs, nocturnal serum cortisol levels, and CPRs were higher in CS-Yes, compared with CS-No and normal subjects. However, one CS-Yes patient had a normal 24-h UFC, two had normal nocturnal serum cortisol levels, and two had normal 24-h CPRs. There was extensive overlap in all of the biochemical parameters between the CS-Yes and the CS-No groups. Thus, measurement of CPR does not seem to offer any diagnostic advantage over available tests for the diagnosis of CS. Patients with proven CS can have normal UFC levels, normal CPRs, or normal nocturnal cortisol levels, whereas patients not thought to have CS may have elevated levels of any one or more these parameters. (J Clin Endocrinol Metab 85: 22–28, 2000)

M

ANY OF THE clinical manifestations of Cushing’s Syndrome (CS), including weight gain, hypertension, glucose intolerance, and menstrual irregularities, are nonspecific; and most patients with these findings do not have CS (1, 2). For this reason, a number of biochemical screening tests have been developed to distinguish patients with true CS from the great majority of patients with nonspecific symptoms who do not have the disease (3– 6). A widely used initial screening test for the presence of CS is measurement of 24-h urine free cortisol (UFC); markedly elevated UFC values are diagnostic of CS in the absence of severe stress (2). However, many CS patients have mildlyto-moderately elevated UFC values, and further biochemical testing may be necessary in this group. In addition, some subjects with proven CS have normal UFC levels (2). Biochemical tests used to confirm the presence of CS include

2-day low-dose dexamethasone suppression, ovine CRH (oCRH) testing, and the combined 2-day dexamethasone/ oCRH test (7–11). However, these tests have limitations, in terms of suboptimal predictive values or lack of validation in large patient groups that include patients with mild CS and/or those with pseudo-CS (the patient groups where accurate biochemical tests are most needed). More recently, measurement of serum cortisol levels at midnight has been proposed as a screening test to distinguish CS from pseudo-CS or normal (9, 11). This test seems to be more accurate than previous tests but has not been widely applied because of logistical issues, and its performance in large groups of subjects with mild CS or with pseudo-CS has not been reported. Recent developments in stable isotope technology have led to advances in the measurement of cortisol production rates (CPRs) by infusion of tracer amounts of nonradioactive cortisol isotopes (12, 13). This method has been used in healthy adults and children and a few subjects with CS. However, the few subjects with CS studied with this technique had markedly elevated CPRs (13), and the range of CPRs in CS is unclear. In the current study, we measured 24-h UFCs, frequently sampled serum cortisol levels over 24 h, and 24-h CPRs in a group of patients with suspected CS, to answer the following

Received May 4, 1999. Revision received August 30, 1999. Accepted September 16, 1999. Address correspondence and requests for reprints to: Mary H. Samuels, Division of Endocrinology, Diabetes, and Clinical Nutrition, Oregon Health Sciences University, 3181Southwest Sam Jackson Park Road, Mail Code CR 107, Portland, Oregon 97201. E-mail: samuelsm@ ohsu.edu. * This study was supported, in part, by NIH PO1-HD-30236 (to D.L.L.), NIH MO1-RR-00334, NIH MO1-RR-00037 (to J.Q.P.), and NIH R29-DK-48366 (to M.H.S.).

22 Downloaded from jcem.endojournals.org on April 19, 2006

CORTISOL PRODUCTION RATES IN CUSHING’S

questions: 1) Is measurement of CPR by isotope dilution a more sensitive test than UFC for the diagnosis of CS in patients with mild hypercortisolemia? 2) Is there a correlation between 24-h UFC, circadian variation in serum cortisol, and/or CPR in CS patients? 3) Can CS occur in the presence of a normal CPR? 4) If so, is the presence of an abnormal circadian rhythm of cortisol secretion alone sufficient to cause CS? Subjects and Methods Subjects Patients with suspected CS. Twenty-eight subjects with suspected CS were referred for further evaluation to the Endocrinology Clinic at Oregon Health Sciences University (OHSU). There were 26 women and 2 men, and the age range was 21– 65 yr. CS was suspected by the referring physician because of a variety of symptoms and signs, which included nonspecific findings, as well as clinical findings more specific for CS. Each patient was evaluated by a staff endocrinologist, who confirmed the clinical suspicion of CS and recommended further evaluation. Some, but not all, of the subjects had elevated 24-h UFCs and/or nonsuppressed serum cortisol levels after overnight dexamethasone testing, done before referral. Patients were included in this study if they had 24-h UFC values less than three-fold above the upper limit of normal (210 mg/24 h), because subjects with markedly elevated UFCs and classic clinical findings were not thought to require further confirmatory tests for the presence of CS. Thus, this patient group was quite heterogeneous, and the clinical suspicion for the presence of CS varied from high to low. However, in our experience, these patients are representative of the spectrum of patients with suspected CS referred to specialty centers for further evaluation. Upon referral, each patient underwent further biochemical testing at the OHSU General Clinical Research Center (GCRC), in an attempt to confirm or exclude CS. This included measurement of serum cortisol levels every 30 min, over 24h (0800 h to 0800 h), and simultaneous measurement of 24-h UFC (see below for details). Subjects were not required to be asleep when samples were drawn. In general, the diagnosis of CS was made based on examination of serum cortisol levels between midnight and 0200 h; levels above 7.5 mg/dL were considered consistent with CS, whereas levels below 7.5 mg/dL were determined to rule out the presence of CS, based on published NIH criteria using a similar experimental design (9). However, in two of the cases reported in the current series, where the clinical suspicion for CS was high, patients were considered to have the disease despite lower nocturnal serum cortisol levels (see Results for description of these patients). In addition, after the published description of the combined 2-day lowdose dexamethasone suppression/oCRH stimulation test (9), some of the subjects underwent this test in an attempt to clarify the diagnosis. Based on this evaluation, 9 of the 28 subjects were diagnosed with CS and proceeded to further testing to determine the cause of the disease. In all 9 subjects, CS was confirmed by surgical pathology of the pituitary gland or adrenal glands and/or by biochemical cure of the disease by pituitary surgery (see Results). The other 19 subjects were thought not to have CS and were referred back to their primary physicians for continuing care. Based on telephone contact with their physicians, in 1–3 yr of follow-up, none of these subjects has had progressive symptoms of CS. For the purposes of this report, the 28 subjects have been divided into those with definite CS (CS-Yes, n 5 9) and those who probably do not have CS (CS-No, n 5 19), with the realization that a few subjects in the CS-No group may, in fact, have mild, early, or slowly progressive CS that cannot be confirmed by currently available biochemical techniques. CPR values were not used in clinical decision-making in the subjects. Normal subjects. As a control group, 23 healthy subjects were recruited at the OHSU and University of Washington GCRCs. There were 21 women and 2 men, and the age range was 21– 60 yr (chosen to match the patient population). The subjects had no known acute or chronic illnesses, were receiving no medications (except for some of the postmenopausal women, who received estrogen replacement therapy), and had no abnormalities on physical examination. Attempts were made to

23

match control subjects to patients with suspected CS, by weight, to avoid confounding effects of weight on any of the biochemical measurements. Each healthy subject had blood drawn every 30 min, over 24 h, during infusion of deuterated cortisol, for determination of 24-h CPRs, as detailed below. A subset of 9 healthy subjects had serum cortisol levels measured in the 30-min samples; for financial reasons, serum cortisol levels were not measured in all healthy subjects. These 9 subjects did not differ from the other control subjects, in terms of CPR values. UFC levels were not routinely measured in the healthy subjects. Informed consent was obtained before study in all suspected CS and normal control subjects, and these investigations were approved by the University of Washington and Oregon Health Sciences University Institutional Review Boards.

Experimental design; measurement of CPR Each patient with suspected CS and each normal subject was admitted to the GCRC the evening before the start of blood sampling. Two separate peripheral iv lines were started: one for infusion of the deuterium-labeled cortisol isotope, and the other for blood drawing. Starting at 0200 h, a continuous infusion of deuterium-labeled cortisol was administered at a constant rate over the next 30 h (until 0800 h of the final study day). This provided a 6-h period of equilibration (0200 h to 0800 h) before blood sampling began at 0800 h, with continued infusion during the entire blood sampling period. The infusion rate (IR) of deuterated cortisol was 20 mg/h, for a total amount infused (over 24 h) of 480 mg, calculated to be less than 5% of a normal endogenous 24-h CPR. This protocol is similar to that previously published for measurements of CPR in healthy subjects (13). Two milliliters of blood were withdrawn every 30 min, for 24 h, from the second iv line, during the infusion of deuterated cortisol (0800 h to 0800 h). Serum was separated, and 100 mL from each sample were combined into a 24-h pooled sample for mass spectrophometric analysis. The individual serum samples were then analyzed for cortisol concentrations in all 28 patients with suspected CS and in 9 of the control subjects. In the 28 patients with suspected CS, urine was collected from 0800 h on the first day of blood sampling until 0800h on the final day for measurement of UFC.

Laboratory methods The measurement of CPR by isotopic dilution was adapted from previous studies (11, 12), as summarized below: Preparation of deuterated cortisol. Deuterated cortisol (d3-cortisol; 9,12,122 H3-cortisol, 99.8 atom %) was prepared in a single batch (Cambridge Isotope Labs, Andover, MA) and used without further purification. The d3-cortisol was dissolved in 95% ethanol (1.0 mg/mL), filtered through a 0.22-micron Durapore filter (Millipore, Bedford, MA), and placed in sterilized depyrogenated glass vials. Extraction of steroids from serum. Serum (0.5 mL) from the pooled sample was hydrolyzed with 5 mL 0.5 mol/L phosphoric acid (pH 1.5) at 22 C for 15–30 min. After hydrolysis, 300 mL 10 mol/L NaOH was added to adjust the pH (final pH 5 6.5). The hydrolysate was passed over a solid-phase extraction column [3 mL Bakerbond SPE octadecyl (C18) extraction column (JT Baker, Phillipsburg, NJ)] using a vacuum manifold (,10 mm Hg, Baker spe-24G column processor). Extraction columns were conditioned before use by washing with 4 mL methanol and 4 mL deionized sterile H2O. After sample application, columns were washed with 2 mL sterile deionized H2O and 3 mL 30% methanol to remove impurities. Columns were dried for 15 min under vacuum (,20 mm Hg). Steroids were eluted using 3 mL of 100% methanol. The methanol was removed using a vortex evaporator (Buchler Instruments, Fort Lee, NJ). Dried residue was dissolved in 300 ml benzene and dried under a nitrogen stream immediately preceding derivatization. Derivatization. To enhance detection by gas chromatography-negative ion chemical ionization mass spectrometry (GC-NICIMS), samples were derivatized to form fluoroacyl derivatives of cortisol by addition of 50 mL anhydrous acetonitrile (Aldrich, Milwaukee, WS) and 50 mL pentafluoropropionic anhydride (Pierce Chemical Co., Rockford, IL). The reactants were heated at 65 C for 2 h. The sample was then brought to

Downloaded from jcem.endojournals.org on April 19, 2006

24

JCE & M • 2000 Vol 85 • No 1

SAMUELS ET AL.

dryness under a stream of filtered nitrogen; the residue was taken up in 50 mL of cyclohexane and immediately analyzed by GC-NICIMS. Gas chromatography. After derivatization, a 5890 gas chromatograph (Hewlett-Packard Co., Palo Alto, CA) was used to separate steroids in serum extracts. Sample (1.0 ml) was injected (injector temperature 5 220 C) onto a 15-m DB-5 capillary column (0.25-mm id with 0.25-micron film thickness; J&W Scientific, Folsom, CA) using alium as the carrier gas. Oven temperature was held at 60 C for 1 min after injection and then increased to 220 C at 50 C/min. The rate of heating was then slowed to 10 C/min until oven temperature reached 260 C. To remove any residual contaminants, the column was then heated to 350 C at 40 C/min, and this temperature was held for 2 min. There was no detectable carryover of cortisol derivatives between individual chromatographic runs. Mass spectrometry and calculations. Selected negative ion mass spectrometry was conducted using a Hewlett-Packard Co. 5988 GC-NICIMS system. Pentafluorotributylamine (negative ion m/e 633) was used for mass calibration. The parent ions of fluoroacyl derivatives of cortisol and d3-cortisol (m/z 782.0 and 785.0) were used for selected ion monitoring. The area under the curves for the total selected ion chromatogram of fluoroacyl derivatives of cortisol (d0, m/z 782) and d3-cortisol (d3, m/z 785) were used to determine the isotopic dilution ratio (d3/d0) in 24-h pooled serum samples. Retention times of the fluoroacyl derivatives were verified by authentic standards. Isotopic contribution at m/z 785 is 1.0%, and this contribution was subtracted from the area under the curve for the total ion chromatogram of the m/z 785 peak. CPR was calculated from the product IR and the ratio of the isotopic enrichment (IE) to isotopic dilution in serum (CPR 5 IR 3[IE/pooled serum samples]), as described (12). CPR values were expressed per square meter of body surface area. Standard, control, and subject samples were run in triplicate to assess intraassay variability, which was less than 10%. Interassay variability averaged 10%. Analysis of serum and urine cortisol concentrations. Cortisol levels were measured in individual serum samples by two-site chemiluminescent immunometric assay (Nichols Institute Diagnostics, San Juan Capistrano, CA). Intraassay coefficients of variation were 3–5%, and interassay coefficients of variation were 6 –10%. Assay sensitivity was 0.8 mg/dL. All samples from a single individual were run in duplicate in the same assay. UFC levels were run by the same immunometric assay after extraction of steroids. Intraassay coefficients of variation for UFCs were 3–5%, interassay coefficients of variation were 12%, and assay sensitivity was 0.8 mg/dL. The reference range for the UFC assay in normal subjects was 8 –70 mg/24 h.

Statistical analysis Twenty-four-hour mean serum cortisol levels were calculated as the average of each subject’s cortisol levels, measured every 30 min, from 0800 h to 0800 h. Nocturnal cortisol levels were calculated as the mean of the three lowest cortisol levels between 2400 h and 0200 h. A mean nocturnal cortisol level below 7.5 mg/dL was considered normal, based on NIH data (9) (we have routinely used this mean level, rather than a single midnight cortisol level, because there is variability in nocturnal cortisol levels in normal subjects, with occasional single midnight values above 7.5 mg/dL; in addition, some of our patients travel from different time zones, and it can be difficult to determine the single best time point to assess nadir serum cortisol levels; analysis of our data using the single midnight cortisol level shows no difference from the mean level calculated here). Differences in demographic data, serum and urine cortisol levels, and CPRs among the three groups (CS-Yes, CS-No, and Normal) were determined by ANOVA. Possible correlations between different biochemical parameters were assessed by regression analysis.

Results Patient and control subject characteristics (Table 1)

Nine patients were determined to have CS, on clinical and biochemical grounds, and they underwent further evaluation to determine the source of the CS. These patients included eight women and one man and ranged in age from 38 – 60 yr (mean 49 6 3 yr). Their mean body surface area was 1.9 6 0.1 m2 (range 1.46 –2.20). All nine patients proved to have CS, either by surgical cure or by surgical pathology. Eight of the nine patients had pituitary-dependent disease, as determined preoperatively by plasma ACTH levels measured peripherally and during bilateral cavernous sinus sampling. These subjects underwent transsphenoidal surgery, with confirmation of a corticotroph adenoma, by immunohistochemistry and/or postoperative biochemical cure of CS, and need for postoperative glucocorticoid treatment. The ninth patient had ACTH-independent CS, as determined by elevated cortisol levels measured after dexamethasone, and low plasma ACTH levels measured at baseline and during

TABLE 1. Demographic and biochemical data in the three subject groups Age (yr)/gender

CS-Yes No. 1 No. 2 No. 3 No. 4 No. 5 No. 6 No. 7 No. 8 No. 9 Mean CSYes Mean CSNo Mean Normal

60/F 59/F 38/F 47/M 54/F 44/F 49/F 45/F 41/F 49 6 3 8F/1M 43 6 3 18F/1M 44 6 2 21F/2M

Diagnosis

Pituitary CS Bilat nodular adrenal hyperplasia Pituitary CS Pituitary CS Pituitary CS Pituitary CS Pituitary CS Pituitary CS Pituitary CS

UFC (ug/24 h)

Mean 24-h cortisol (ug/dL)

Nocturnal cortisol (ug/dL)

CPR (mg/m2 z 24 h)

75 76

15.2 6.4

9.5 4.7

5.5 6.4

133 180 172 48 117 158 149 123 6 16a

14.5 13.9 10.6 11.0 11.9 13.4 18.4 12.8 6 1.1a,b

13.6 14.2 8.0 8.7 5.0 9.0 17.7 10.0 6 1.4a,b

21.6 19.2 15.7 17.9 23.2 18.1 21.0 16.5 6 2.1a,b

67 6 8

7.7 6 0.6

2.7 6 0.3

8.7 6 0.9

ND (normal range 8 –70 ug/24-h)

6.8 6 0.3

1.5 6 0.3

6.4 6 0.4

Individual data are shown for the nine subjects with proven Cushing’s Syndrome; mean 6 SEM data are shown for all three groups. Mean 24-h cortisol levels were calculated as the arithmetic mean of individual serum cortisol levels measured every 30 minutes, over 24 h. a Significantly different than the CS-No group by t test or ANOVA, P , 0.001; 24-h UFCs were not routinely measured in the control subjects; the normal range for the assay is given instead. Mean 24-h and nocturnal serum cortisol levels were measured in 9 of the 23 normal subjects. b Significantly different than the Normal group by ANOVA, P , 0.001.

Downloaded from jcem.endojournals.org on April 19, 2006

CORTISOL PRODUCTION RATES IN CUSHING’S

peripheral oCRH testing. Abdominal CT scan revealed bilateral adrenal nodules, and the patient underwent bilateral adrenalectomy. Pathology confirmed bilateral nodular adrenal hyperplasia. Nineteen patients with suspected CS were determined not to have CS, on clinical and biochemical grounds, and did not undergo further evaluation. These patients included 18 women and 1 man, and they ranged in age from 21– 65 yr (mean 43 6 3 yr). Their mean body surface area was 2.0 6 0.1 m2 (range 1.45–2.64). The 23 control subjects included 21 women and 2 men and ranged in age from 21– 60 yr (mean 44 6 2 yr). Their mean body surface area was 1.9 6 0.1 m2 (range, 1.52–2.41), confirming the adequacy of matching control subjects to suspected CS subjects for weight. Biochemical data

CS-Yes subjects had higher mean 24-h UFC values, 24-h serum cortisol levels, nocturnal serum cortisol levels, and 24-h CPR values, compared with normal and CS-No groups (Table 1). There were no differences in 0800-h serum cortisol levels among the three groups (data not shown). Mean serum cortisol levels, measured every 30 min, over 24 h, in the three groups are shown in Fig. 1, top panel. Individual UFC, nocturnal cortisol, and CPR values for the three groups of subjects are shown in Fig. 2. Although mean UFC levels were higher in the CS-Yes group, compared with the CS-No group (P , 0.001), there was significant overlap in individual values between the two

25

groups (Fig. 2, left panel). One CS-Yes subject had a normal UFC of 48 mg/24 h (no. 6) but had pituitary CS, proven by positive ACTH immunohistochemistry of a pituitary adenoma and postoperative clinical and biochemical cure of her CS. Two CS-Yes subjects had UFC values slightly above the normal range for the assay (no. 1: 75 mg/24 h, and no. 2: 76 mg/24 h); one had an immunohistochemistry-proven ACTH adenoma and surgical cure, whereas the other had bilateral nodular adrenal hyperplasia and ACTH-independent CS. The CS-No subjects had a wide range of UFC levels, which completely overlapped the normal range and the CS-Yes group. Mean nocturnal serum cortisol levels were higher in the CS-Yes group, and there was less overlap in nocturnal cortisol levels between the CS-Yes group and the other two groups (Fig. 2, middle panel). However, this was expected by our criteria for the diagnosis of CS, because an important part of our diagnostic algorithm is the measurement of nocturnal cortisol levels. Even using this criteria, we found two subjects (no. 2 and no. 7) with proven CS who had nocturnal cortisol levels below the 7.5 mg/dL cut-off recommended by the NIH. One of these two subjects (no. 2) had bilateral nodular adrenal hyperplasia (nocturnal cortisol of 4.7 mg/dL), whereas the other (no. 7) had recurrent pituitary-dependent CS with removal of an ACTH-positive adenoma and biochemical cure of CS (nocturnal cortisol of 5.0 mg/dL). Mean CPR levels were higher in the CS-Yes group, compared with the CS-No and normal groups. However, there

FIG. 1. Mean serum cortisol levels, measured every 30 min, over 24 h, in the three subject groups. Top panel, the mean of all nine CS-Yes subjects is indicated with triangles; the mean of the CS-No subjects, with squares; and the mean of the normal subjects, with circles. Bottom, the mean of the two CSYes subjects with normal 24-h CPR values is indicated with diamonds; the mean of the other seven CS-Yes subjects is indicated with triangles; and the mean of the normal subjects, with circles.

Downloaded from jcem.endojournals.org on April 19, 2006

26

SAMUELS ET AL.

JCE & M • 2000 Vol 85 • No 1

FIG. 2. Individual 24-h urine free cortisol (left panel), nocturnal cortisol (middle ), and 24-h CPRs (right) in the three subject groups. Note that UFCs were not measured in the control group; the horizontal lines on the left indicate the normal range for 24-h UFC. Mean levels and SE bars are superimposed on each data set. *, Significantly different than the normal and CS-No groups, by t test or ANOVA, P , 0.001.

was a wide range of CPR values in both the CS-No and the CS-Yes groups (Fig. 2, right panel). Five of the CS-No subjects had 24-h CPR values above the range for the normal subjects (10.7, 10.9, 12.4, 16.1, and 17.5 mg/m2), whereas two of the CS-Yes subjects had 24-h CPR values within the range of the normal subjects (5.5 and 6.4 mg/m2). One of these subjects (no. 1) had pituitary-dependent CS with positive immunohistochemical staining of an ACTH-secreting adenoma and postoperative cure, whereas the other (no. 2) had bilateral nodular adrenal hyperplasia. Fig. 1, lower panel shows the 24-h serum cortisol levels in these two subjects (with normal 24-h CPRs) separated from the other 7 patients with proven CS. Results of regression analysis of 24-h UFC vs. 24-h mean cortisol, nocturnal cortisol, and CPR levels in the CS-Yes and CS-No groups are shown in Fig. 3. In each case, there was a significant correlation between UFC and each of the other variables. However, there was a high degree of scatter among the data, and correlation coefficients were only 0.63– 0.68. Correlations were lower when only CS-Yes subjects were considered (data not shown). Discussion

In the absence of obvious clinical findings and markedly elevated UFC levels, it can be difficult to distinguish patients with true CS from the great majority of patients who do not have the disease. This is because of significant overlap in

clinical and biochemical parameters between patients with mild CS and those with nonspecific symptoms and mildly elevated UFC levels attributable to other causes. A number of biochemical tests have been proposed to attempt to separate these two groups, including assessment of the circadian rhythm of cortisol levels, suppression of cortisol levels by dexamethasone, stimulation of cortisol levels by oCRH, and combined administration of dexamethasone and oCRH (3– 11). However, all of these tests are limited, in terms of diagnostic accuracy and/or validation in large numbers of patients, especially in patients with mild disease and in patients with pseudo-CS. Therefore, there is currently no so-called gold standard test for the diagnosis of mild-tomoderate CS. Using a nonradioactive isotopic dilution method for the measurement of CPRs (12, 13), we compared 24-h CPRs, UFCs, and nocturnal cortisol levels in patients with suspected mild-to-moderate CS. We hypothesized that CPR would be a more sensitive diagnostic test for CS than 24-h UFC or nocturnal serum cortisol levels. However, in the nine patients with proven CS, this was not invariably the case. One of the nine subjects had 24-h UFC values within the normal range, two had nocturnal serum cortisol levels below the cut-off proposed by NIH investigators (9) for the diagnosis of CS, and two had 24-h CPR values within the normal range of our healthy subjects. The two CS-Yes subjects with normal CPRs did tend to have lower nocturnal cortisol se-

Downloaded from jcem.endojournals.org on April 19, 2006

CORTISOL PRODUCTION RATES IN CUSHING’S

FIG. 3. Scatter graphs and regression analysis of 24-h urine free cortisol (x-axis) vs. 24-h mean cortisol levels (top), nocturnal cortisol levels (middle), and 24-h CPRs (bottom). Squares represent the CS-No group; and triangles, the CS-Yes group. First-order regression lines, correlation coefficients, and P values are shown for each analysis.

cretion than the other seven subjects, suggesting that they may have had an early or more mild form of CS. However, in general, there was poor correlation among the three tests in the CS-Yes patients, suggesting that each test, when used individually, has limited ability to diagnose CS in patients with mild-to-moderate disease. These data illustrate the clinical problem familiar to endocrinologists who see significant numbers of patients with suspected CS, because there is no straightforward method for confirming the diagnosis. Though disappointing, these results are not entirely surprising, because no single test to date has proved adequate for the accurate discrimination between CS and pseudo-CS in every patient. Instead, a combination of clinical experience and the use of several biochemical tests is required for accurate diagnosis in many of these patients. Patients with obvious clinical manifestations of CS and markedly elevated UFC levels (.210 mg/24 h) were excluded from this study. In such patients, there might have been a better correlation among CPRs, UFCs, and nocturnal serum cortisol levels. However, such patients rarely require extensive biochemical evaluation to confirm the presence of CS, and results from the current study would have little clinical relevance for them.

27

This study used healthy control subjects who were age- and weight-matched to the subjects with suspected or proven CS. For this reason, some of the healthy subjects were obese (BSA range up to 2.41), which could have increased CPRs. However, most of the subjects and patients were not markedly obese, in keeping with the mild nature of the CS in the patients. In addition, there was no correlation between BSA and either UFC or CPR in any of the groups (data not shown). For these reasons, we do not believe that obesity explains the overlap in our measures of cortisol production among the three groups. From this small series of 9 patients, it seems that CS can occur in the face of a normal 24-h CPR, UFC, or circadian cortisol rhythm. These findings are reminiscent of previous reports of normal biochemical tests (UFCs, dexamethasone suppression, oCRH stimulation) in some patients with proven CS. The more recently proposed measurement of nocturnal cortisol levels (9, 11) seems to have superior sensitivity and specificity, compared with other biochemical tests for the diagnosis of CS, confirming older studies in small numbers of patients (14–17). Given the limitations of other biochemical tests for CS, it is not surprising that nocturnal cortisol levels would also have a defined false-negative rate in patients with mild disease [2 of 9 in the current series; 9 of 234 in the NIH series, which included more severely affected subjects (9)]. Another possible interpretation of the nocturnal cortisol data in our patients is that the suggested NIH cut-off of 7.5 mg/dL is too high, because none of our normal subjects had nocturnal cortisol levels above 3.2 mg/dL. Using this more stringent cut-off, all of our patients with proven CS had nocturnal cortisol levels above the normal range. However, 4 of the CS-No subjects also had nocturnal cortisol levels above 3.2 mg/dL (3.8, 4.3, 5.0, and 5.3 mg/dL). A lower cut-off for normal nocturnal serum cortisol levels would also lead to increased false-positive rates in the larger NIH series (9). A large series from St. Bartholomew’s Hospital in London recently reported that healthy control subjects all had midnight cortisol levels of less than 1.8 mg/dL, while 150 patients with CS all had midnight cortisol levels that ranged between 2.5 and 72 mg/dL (11). The strength of this study was that all subjects were acclimated to the hospital for 48 h and were required to be asleep when blood samples were drawn, thereby reducing potential stress-induced cortisol elevations. This may explain the better discrimination between CS and normal subjects using a much lower cut-off than achieved in our study or the NIH series. However, this study did not include a group of pseudo-CS subjects, or subjects with suspected CS who were eventually thought not to have the disease. It is unclear whether such subjects would have met the rigorous cut-offs for midnight cortisol levels, or whether they would have shown more overlap with the normal and proven CS subjects. The main limitation of our study is the classification of the CS-No subjects. These subjects were determined not to have CS, based on clinical judgment combined with normal 24-h UFC, nocturnal cortisol levels, and (in more recent subjects) the combined dexamethasone/CRH test. However, the limitations of these tests have already been mentioned, and it is possible that a few of these subjects have early or mild CS. Previous studies have used clinical follow-up to confirm the absence of CS in such subjects, with lack of progression of

Downloaded from jcem.endojournals.org on April 19, 2006

28

SAMUELS ET AL.

symptoms and physical findings indicating an accurate diagnosis (9, 10). This issue is especially pertinent in the two CS-No subjects who had 24-h CPR levels identical to those in the CS-Yes group, or almost twice the upper limit of CPR levels in our normal subjects. As in the CS-Yes patients, there was poor correlation among the biochemical tests in the CS-No subjects, which highlights the difficulty in using a single biochemical test to rule out CS in subjects with equivocal clinical findings. In summary, we measured nocturnal serum cortisol levels, 24-h UFC levels, and 24-h CPRs by stable isotope dilution in a series of patients with suspected mild-to-moderate CS. We found that measurement of CPRs does not seem to offer any advantage over these other biochemical techniques in differentiating patients with CS from those who do not have the disease. Patients with proven CS can have normal cortisol production rates, normal UFC levels, or normal circadian cortisol rhythms, at least by the currently accepted standards for these tests. Conversely, patients not thought to have the disease after extensive clinical evaluation can have elevated biochemical parameters, although the accuracy of the diagnosis may be suboptimal in at least some of these patients. Thus, the development of a true so-called gold standard test for the diagnosis of mild-to-moderate CS awaits further insights into the pathophysiology and pathogenesis of the disease.

3. 4. 5. 6. 7.

8. 9. 10.

11. 12.

13. 14.

Acknowledgments

15.

We thank the nursing and laboratory staff of the OHSU and University of Washington General Clinical Research Centers for outstanding patient care and assay performance.

16.

References

17.

1. Yanovski JA, Cutler GB. 1994 Glucocorticoid action and the clinical features of Cushing’s syndrome. Endocrinol Metab Clin North Am. 23:487–509. 2. Danese RD, Aron DC. 1994 Principles of clinical epidemiology and their

JCE & M • 2000 Vol 85 • No 1

application to the diagnosis of Cushing’s syndrome: Rev. Bayes meets Dr. Cushing. Endocrinologist. 4:339 –346. Orth DN. 1995 Cushing’s syndrome. N Engl J Med. 332:791– 803. Trainer PJ, Grossman A. 1991 The diagnosis and differential diagnosis of Cushing’s syndrome. Clin Endocrinol (Oxf). 34:317–330. Kaye TB, Crapo L. 1990 The Cushing’s syndrome: an update on diagnostic tests. Ann Intern Med. 112:434 – 444. Meikle AW. 1993 A diagnostic approach to Cushing’s syndrome. The Endocrinologist. 3:311–320. Nieman LK, Chrousos GP, Oldfield EH, Avgerinos PC, Cutler GB, Loriaux DL. 1986 The ovine corticotropin-releasing hormone stimulation test and the dexamethasone suppression test in the differential diagnosis of Cushing’s syndrome. Ann Intern Med. 105:862– 867. Gold PW, Loriaux DL, Roy A, et al. 1986 Responses to corticotropin-releasing hormone in the hypercortisolism of depression and Cushing’s disease. N Engl J Med. 314:1329 –1335. Papanicolaou DA, Yanovski JA, Cutler GB, Chrousos GP, Nieman LK. 1998 A single midnight serum cortisol measurement distinguishes Cushing’s syndrome from pseudo-Cushing’s states. J Clin Endocrinol Metab. 83:1163–1167. Yanovski JA, Cutler GB, Chrousos GP, Nieman LK. 1993 Corticotropinreleasing hormone stimulation following low-dose dexamethasone administration. A new test to distinguish Cushing’s syndrome from pseudo-Cushing’s states. JAMA. 269:2232–2238. Newell-Price J, Trainer P, Perry L, Wass J, Grossman A, Besser M. 1995 A single sleeping midnight cortisol has 100% sensitivity for the diagnosis of Cushing’s syndrome. Clin Endocrinol (Oxf). 43:545–550. Brandon DD, Isabelle LM, Samuels MH, Kendall JW, Loriaux DL. 1999 Cortisol production rate measurement by stable isotope dilution using gas chromatography-negative ion chemical ionization mass spectrometry. Steroids. 64:372–378. Esteban NV, Loughlin T, Yergey AL, et al. 1991 Daily cortisol production rate in man determined by stable isotope dilution/mass spectrometry. J Clin Endocrinol Metab. 71:39 – 45. Glass AR, Zavadil AP, Halberg F, Cornelissen G, Schaaf M. 1984 Circadian rhythm of serum cortisol in Cushing’s Disease. J Clin Endocrinol Metab. 59:161–165. Refetoff S, van Cauter E, Fang VS, Laderman C, Graybeal ML, Landau RL. 1985 The effect of dexamethasone on the 24-hour profiles of adrenocorticotropin and cortisol in Cushing’s syndrome. J Clin Endocrinol Metab. 60:527–535. Liu JH, Kazer RR, Rasmussen DD. 1987 Characterization of the twenty-four hour secretion patterns of adrenocorticotropin and cortisol in normal women and patients with Cushing’s disease. J Clin Endocrinol Metab. 64:1027–1035. Van der Berg G, Frohlich M, Veldhuis JD, Roelfsema F. 1995 Combined amplification of the pulsatile and basal modes of adrenocorticotropin and cortisol secretion in patients with Cushing’s disease: evidence for decreased responsiveness of the adrenal glands. J Clin Endocrinol Metab. 80:3750 –3757.

Downloaded from jcem.endojournals.org on April 19, 2006