View larger version (24K): [in this window] [in a new window]   Figure 1. A diagnostic approach to confirm the existence of the Cushing syndrome and to determine its cause. *See reference 32.

Three decades ago, two noninvasive tests for the differential diagnosis of the Cushing syndrome were introduced by Liddle [5, 6]: the high-dose dexamethasone suppression test and the metyrapone test. New diagnostic criteria for the dexamethasone test have recently been introduced to improve its accuracy [7]. However, even with these new criteria, 17% of patients with the pituitary Cushing syndrome suppressed excretion of 17-hydroxycorticosteroid and urine free cortisol to a lesser extent than did patients with the ectopic ACTH syndrome [7]. The metyrapone test has been less widely used than the dexamethasone suppression test and has been done and interpreted less consistently [8-15]. Additionally, its utility for determining the cause of the ACTH-dependent Cushing syndrome has been questioned [9, 16, 17].

Because of the continuing need for improved noninvasive means of distinguishing pituitary from ectopic ACTH-producing tumors, we reexamined the usefulness of the metyrapone test in a large series of patients with the ACTH-dependent Cushing syndrome. Our objectives were to determine optimal criteria for the diagnosis of the Cushing syndrome by the standard metyrapone test, to determine the relative usefulness of plasma 11-deoxycortisol and 24-hour urine 17-hydroxysteroid (17-OHS) levels as end points for the test, to compare the diagnostic accuracy of the metyrapone test with that of the dexamethasone suppression test, and to evaluate whether combining the results of the two tests would further improve sensitivity and diagnostic accuracy.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

The participants of this pilot study were selected from among patients referred to the National Institutes of Health (NIH). Nearly all were found to have hypercortisolism before referral. In clinical practice, only a few patients screened for suspected Cushing syndrome have hypercortisolism. To qualify for the study, patients had to have had at the NIH the dexamethasone suppression test, the metyrapone test, and a surgical procedure to remove the source of excessive ACTH.

Metyrapone and Dexamethasone Tests

For the metyrapone test, 24-hour urine specimens (starting at 0600 h) and venous blood samples (5 mL starting at 0800 h) were obtained for 4 consecutive days. On days 1 and 2 of the test, patients received no metyrapone; starting at 0800 h of day 3 and ending at 0400 h of day 4, patients received 6 doses of metyrapone at 4-hour intervals. Seven hundred and fifty milligrams per dose was administered to all patients except for an 8-year old boy in whom the dose was adjusted for body surface area to 500 mg per dose [18].

For the dexamethasone suppression test, 24-hour urine specimens (starting at 0600 h) were obtained for 6 consecutive days. On days 1 and 2 of the test, patients received no dexamethasone; starting at 0600 h of day 3 and ending at 2400 h of day 4, all patients received eight 0.5-mg doses of dexamethasone orally every 6 hours (low-dose); starting at 0600 h of day 5 and ending at 2400 h of day 6, all patients received eight 2.0-mg doses of dexamethasone every 6 hours (high-dose), with the exception of an 8-year old boy in whom the dose was adjusted for weight to 0.25 mg for days 3 to 4 and 1.0 mg per dose for days 5 to 6 [18].

The 24-hour urine specimens were refrigerated and sent to the NIH clinical chemistry laboratory. After October 1988, an aliquot was also sent to Hazleton Laboratories (Vienna, Virginia). Urine volume and creatinine levels were measured to confirm the completeness of collection. The 0800 h blood samples obtained during the metyrapone test were refrigerated and separated, and the plasma was sent to Hazleton Laboratories for measurement of plasma 11-deoxycortisol levels.

Assays and Calculations

An aliquot of each urine collection was used for measurement of excretion of 17-OHS by a modification of the Porter-Silber method [19-21]. The intra-assay and interassay coefficients of variation were 5.9% and 7% to 14%, respectively. The normal range for urine excretion of 17-OHS is 2 to 6 mg/d. Another aliquot of urine was frozen and sent either to SmithKline Bioscience Laboratories (King of Prussia, Pennsylvania) or, after October 1988, to Hazleton Laboratories for measurement of cortisol by radioimmunoassay. The Premix kit (Diagnostic Products Corporation, Los Angeles, California) or the Quanticoat kit (Kallestad Laboratories Incorporated, Austin, Texas) were used by these laboratories, respectively. The intra-assay and interassay coefficients of variation were 5.4% and 9.3%, respectively, for the former and 5.8% and 7.5%, respectively, for the latter. The normal range for urine excretion of free cortisol is 20 to 90 µg/d for both laboratories. After extraction and chromatography, the plasma 11-deoxycortisol level was measured by radioimmunoassay [22, 23]. The intra-assay and interassay coefficients of variation were 11.7% and 18.2%, respectively. In most patients, the 11-deoxycortisol levels before and after the administration of metyrapone were measured in the same assay.

We calculated the percentage of stimulation of 17-OHS after metyrapone was administered to each patient using the following formula: ([stimulated level – baseline level] x 100)/baseline. The stimulated 17-OHS values were calculated in five different ways. In the first method, the higher 17-OHS value between days 3 and 4, which is the conventional outcome measure, was used to interpret the test result [6]. To optimize the sensitivity of the test, we also evaluated the day 3 value of 17-OHS, the day 4 value, the lower of the days 3 and 4 values, and the mean of the day 3 and day 4 17-OHS values as alternate measures of the stimulated level. To test the sensitivity of a shortened test, the day 2 values were also used as a baseline in addition to the average of the values of days 1 and 2. Thus, when the five measures of response were combined with the two baseline measures, 10 different measures for interpreting the 17-OHS response to metyrapone were evaluated. We calculated the percentage of suppression of urine free cortisol levels after the administration of metyrapone for each patient using the following formula: ([baseline level -suppressed level] x 100)/baseline. Suppressed levels were the values of day 3 and baseline levels, the average of days 1 and 2.

The 11-deoxycortisol response to metyrapone was expressed either as the stimulated value on day 4 [13, 17] or as the ratio of the stimulated to baseline values as previously described [12, 15]. For the latter measure, the day 4 stimulated level was divided by the average baseline 11-deoxycortisol levels on days 1, 2, and 3 (11-deoxycortisol on day 3 was drawn before metyrapone administration). To test the diagnostic efficiency of a shortened test, the average of the days 2 and 3 11-deoxycortisol levels or the day 3 11-deoxycortisol level alone, were substituted as alternate measures of the baseline 11-deoxycortisol level. Thus, we assessed four different measures of the plasma 11-deoxycortisol response, that is, the absolute level of the day 4 plasma 11-deoxycortisol and the ratio of the day 4 plasma 11-deoxycortisol level to each of three different baseline measures (day 3 alone, the mean of days 2 and 3, and the mean of days 1, 2, and 3).

We calculated the percentage of suppression of 17-OHS or urine free cortisol after the administration of high-dose dexamethasone for each patient using the following formula: ([baseline level – suppressed level] x 100)/baseline. Suppressed levels were the values on day 6, and baseline levels were the average of days 1 and 2 as previously described [7].

Statistical Analysis

Estimates of sensitivity (true-positive results/[true-positive results + false-negative results]), specificity (true-negative results/[true-negative results + false-positive results]), and diagnostic accuracy ([true-positive + true-negative results]/total number of patients) were determined for each steroid at various suppression levels based on the diagnosis obtained at surgery as the gold standard. For each analysis, the presence of pituitary disease was considered to be a positive response ("diseased") and the presence of ectopic ACTH secretion, a negative response ("non-diseased"). For the purpose of determining the specificity and diagnostic accuracy relative to the gold standard of pituitary surgery, the 15 patients who lacked a confirmed diagnosis after unsuccessful pituitary surgery were considered "nondiseased." We determined likelihood ratios at each suppression level by dividing the sensitivity by 1 minus the specificity [24].

Because one aim of our study was to develop criteria by which the need for trans-sphenoidal surgery could be established by a noninvasive test, the criteria set for the end points of the metyrapone and dexamethasone tests were chosen so as not to misclassify any of the patients with surgically proven ectopic ACTH syndrome as having the pituitary Cushing syndrome. Thus, the criteria were set so that a positive response would indicate the presence of pituitary disease, establishing the need for trans-sphenoidal surgery, whereas a negative response would not necessarily differentiate between pituitary and ectopic ACTH secretion and would indicate the need for further evaluation.

To compare individual end points of the two tests at multiple levels of steroid stimulation or suppression without the bias of predetermined criteria, we constructed univariate curves of the receiver operating characteristics (ROC) by plotting the sensitivity against 1 minus the specificity at each stimulation or suppression level using the Rule-Maker program (Digital Medicine, Inc., Hanover, New Hampshire). We calculated the area under each ROC curve using the Rule-Maker program by dividing the curve into trapezoidal sections and calculating the sum of these areas. The area under each curve represents the inherent accuracy of the test endpoint independent of the criteria, with the area of a "perfect" test being 1.00 [25]. The sensitivity of each of the tests was compared at predetermined specificity levels by the chi-square statistic (1 degree of freedom) = {[absolute value of (t – f)] –1}²/(t + f), where t equals the number of cases that were correctly labeled by the rule of one test and incorrectly labeled by the rule of the second test; and f equals the number of cases that were incorrectly labeled by the rule of the first test but were correctly labeled by the rule of the second test [26]. To compare diagnostic accuracies at the predefined criterion of each test, similar chi-square tests were done. To compare the area under the ROC curve for composite rules, we constructed correlated 2 x 2 tables [27]; Cochran Q statistics were interpreted with the Bonferroni adjustment for multiple comparisons [28]. Ninety-five percent confidence intervals were determined using standard approaches [29, 30].

Results

Top Methods Results Discussion Author & Article Info References

Between November 1982 and December 1992, we identified 518 patients with hypercortisolism. Of these, 29 had a pseudo-Cushing state, and 48 had the ACTH-independent Cushing syndrome due to a primary adrenal cause. The evaluation of the other 441 patients diagnosed as having the ACTH-dependent Cushing syndrome included pituitary magnetic resonance imaging, an adrenal computed tomographic scan, inferior petrosal sinus sampling [3], and a corticotropin-releasing hormone test [31].

Of the 441 patients with the ACTH-dependent Cushing syndrome who were potential candidates for this study, 255 were not entered because of referral to a nonparticipating NIH physician (106 patients), inability to do both the metyrapone and dexamethasone tests (85 patients), unavailability of data (30 patients), incomplete urine collections (7 patients), or failure to have a diagnostic surgical procedure (27 patients). Twenty of the latter 27 patients were thought to have pituitary disease and 7 to have an unidentified ectopic source of ACTH.

The decision to do pituitary surgery represented a collective clinical judgment by the endocrinologists and neurosurgeon on the basis of all diagnostic test results. The inferior petrosal sinus sampling results were most influential, followed by the corticotropin-releasing hormone test, the dexamethasone test, and the metyrapone test.

Of the 186 patients entered in the study, 171 received a definitive diagnosis on the basis of surgical pathologic conditions or curative resection, and 15 had "unsuccessful pituitary surgery." Diagnosis by surgical pathologic conditions was based on immunohistochemical staining for ACTH of a pituitary adenoma or an ectopic ACTH-secreting tumor. Curative resection was defined by a mean postoperative morning plasma cortisol level (of all available values) of less than 190 nmol/L (normal, 193 to 690 nmol/L). By these criteria, 156 patients had pituitary disease and 15 had the ectopic ACTH syndrome caused by bronchial carcinoid (10 patients), thymic carcinoid (1 patient), medullary carcinoma of the thyroid (1 patient), oat cell carcinoma of the lung (1 patient), adenocarcinoma of the pancreas (1 patient), and metastatic carcinoma of undetermined origin (1 patient).

The other 15 patients were later classified on the basis of their follow-up over 3 to 5 years. One patient had the ACTH-independent Cushing syndrome that was cured after a hyperfunctioning unilateral adrenal nodule was removed. This patient had a 43% increase in 17-OHS levels and a 730-fold increase in 11-deoxycortisol levels after the administration of metyrapone and a 57% increase in 17-OHS levels with a 38.5% decrease in urine free cortisol after administration of high-dose dexamethasone. The other 14 patients were diagnosed with the pituitary Cushing syndrome on the basis of low plasma corticotropin-releasing hormone levels (which ruled out a source of ectopic, corticotropin-releasing hormone) and surgical pathologic conditions or curative resection at a previous or subsequent pituitary surgery (9 patients), response after pituitary irradiation, such as decreasing ACTH levels and remission from the stigmata of the Cushing syndrome (4 patients), or pituitary tumor growth documented by pituitary scan (1 patient).

Metyrapone Test

We first evaluated the usefulness of the 17-OHS response to metyrapone as an end point for the test. The two best measures were equivalent: the stimulated level from day 3 (the day of metyrapone administration) combined with either of the two baseline measures (data not shown). For simplicity of doing the test, the measure combining the day 2 baseline level and the day 3 stimulated level was selected (Table 1, Figure 2).

View larger version (21K): [in this window] [in a new window]   Figure 2. Stimulation of urine 17-hydroxysteroid (left) and plasma 11-deoxycortisol (right) during the metyrapone test in 186 patients with the Cushing syndrome. Stimulation of urine 17-hydroxysteroid excretion is expressed as the percentage increase on the day of metyrapone administration compared with the baseline value on the previous day (day 3 compared with day 2). Stimulation of 11-deoxycortisol is expressed as the ratio of stimulated value (day 4 value) divided by the baseline value (the mean of the two preceding days). For determining sensitivity and specificity of the metyrapone test, patients were classified as positive or negative for the pituitary Cushing syndrome according to the results of surgical exploration: patients with surgically confirmed pituitary syndrome, those with surgically confirmed ectopic adrenocorticotropic hormone (ACTH) syndrome, and those in whom trans-sphenoidal surgery (TSS) failed to confirm a pituitary adenoma (failed TSS) but who were diagnosed after follow-up as having the pituitary forms of Cushing syndrome [14] and the ACTH-independent Cushing syndrome [1]. The 10 data points at the top of the panels represent values above the scale of the vertical axis.

We next analyzed the usefulness of plasma 11-deoxycortisol levels as an end point for the metyrapone test. The end point that gave the best sensitivity (38%) without misclassifying the patients with the ectopic ACTH syndrome was the ratio of the stimulated value on the morning of day 4 to the mean baseline value of days 2 and 3 (Table 1, Figure 2). A 400-fold increase over the baseline value in the 11-deoxycortisol level was required for the diagnosis of pituitary disease without misclassification of the patients with the ectopic ACTH syndrome. The sensitivity of 38% found with plasma 11-deoxycortisol as an end point was significantly lower than that observed with the optimal criterion for 17-OHS (60%; P = 0.001).

We next asked whether the diagnostic usefulness of the metyrapone test could be improved by combining the urine 17-OHS and the plasma 11-deoxycortisol end points. When a combined criterion of an increase in urine 17-OHS levels of greater than 70% (on the day 3 compared with the day 2 baseline values) or an increase in the plasma 11-deoxycortisol level of more than 400-fold over the baseline value (the ratio of the day 4 level after administration of metyrapone to the average baseline level on days 2 and 3) was used for the diagnosis of pituitary disease, the test sensitivity increased significantly to 72% compared with the 11-deoxycortisol end point (P = 0.001), and the diagnostic accuracy of the test increased significantly compared with either the 11-deoxycortisol end point (P = 0.001) or the 17-OHS end point (P = 0.005).

The percentage of suppression of urine free cortisol after administration of metyrapone was 69% ±25% for the patients with pituitary disease and 83% ±12% for the patients with ectopic ACTH-secreting tumors. The percentage of stimulation of 17-OHS after administration of metyrapone in the patients with pituitary disease did not correlate with the percentage of suppression of urine free cortisol (r = 0.05, P > 0.2). A decrease in urine free cortisol levels of less than 50% gave a sensitivity of 12% for the diagnosis of the pituitary Cushing disease without misclassifying any of the patients with surgically proven ectopic ACTH secretion. Combining this urine free cortisol criterion with the previously stated criteria did not result in a significant improvement in the sensitivity or diagnostic accuracy of the metyrapone test (data not shown).

Including the 14 patients diagnosed with the pituitary Cushing syndrome from information obtained during follow-up gave the previously defined criteria a specificity of 100% for the diagnosis of the pituitary form of Cushing syndrome (Table 2). When evaluated with criteria that had a specificity of 100% for the pituitary Cushing syndrome, both sensitivity and diagnostic accuracy were significantly greater for the test combining both urine 17-OHS and plasma 11-deoxycortisol than those for the urine 17-OHS test (P = 0.01 and P = 0.001 for sensitivity and diagnostic accuracy, respectively) or the plasma 11-deoxycortisol test (P = 0.001 and P = 0.001, respectively).

Dexamethasone Suppression Test

The suppression criteria (for the day 6 value relative to the mean baseline value on days 1 and 2) required to avoid misclassifying patients with proven ectopic ACTH secretion were greater than 69% for 17-OHS (Table 1, Figure 3) and greater than 90% for urine free cortisol (Table 1, Figure 3). Using these criteria, we observed sensitivities of 54% and 59% for the urine 17-OHS and urine free cortisol end points, respectively. Combining the criteria for the two end points (that is, the diagnosis of the pituitary Cushing syndrome if urine 17-OHS suppression was greater than 69% or if urine free cortisol suppression was greater than 90%) yielded an increase in sensitivity to 72% and a similar increase in diagnostic accuracy, significantly greater than that achieved with either end point alone (P = 0.001), without misclassification of any of the patients with surgically proven ectopic ACTH syndrome.

View larger version (17K): [in this window] [in a new window]   Figure 3. Suppression of urine 17-hydroxysteroid (left) and urine free cortisol (right) during a standard 6-day dexamethasone suppression test in 186 patients with the Cushing syndrome. Suppression of urine 17-hydroxysteroid and free cortisol excretion is expressed as the percentage decrease on day 6 (last day of high-dose dexamethasone) compared with the mean baseline values from days 1 and 2. The patient groups are the same as in Figure 2. ACTH =adrenocorticotropic hormone; TSS = trans-sphenoidal surgery.

Combining the Metyrapone and Dexamethasone Tests

When the best criteria for the two tests were combined (an increase of 17-OHS levels >70% or of 11-deoxycortisol levels >400 times the baseline value for the metyrapone test; or a suppression of 17-OHS >69% or of urine free cortisol >90% for the dexamethasone test), the sensitivity for the diagnosis of pituitary disease increased significantly to 88%, without misclassification of patients with proven ectopic ACTH when compared with the sensitivity of 72% when either the dexamethasone or metyrapone tests were used alone (P = 0.01 compared with combination dexamethasone test). Combining the metyrapone and dexamethasone composite end point tests yielded a greater diagnostic accuracy (82%, P = 0.001) than was achieved by either of the tests alone. The optimal outcome measures, the criteria selected (to achieve the highest sensitivity without misclassification of patients with proven ectopic ACTH cases), the sensitivities, specificities, diagnostic accuracies, and likelihood ratios for each of the end points of the metyrapone and dexamethasone suppression tests and their combinations are shown in Table 1.

When criteria for the metyrapone and dexamethasone tests were combined, the specificity was 53% compared with 67% for the individual tests (Table 1). However, this decrease in specificity was entirely attributable to the 15 patients who lacked a surgically confirmed diagnosis and were therefore assigned as having nonpituitary disease on the basis of unsuccessful pituitary surgery. When the patient with the ACTH-independent Cushing syndrome was excluded and the other 14 patients were assigned to the group of patients with the pituitary Cushing syndrome on the basis of their follow-up evaluation Table 2, the test combining both metyrapone and dexamethasone criteria had a specificity of 100%, a sensitivity of 88%, and a diagnostic accuracy of 89% for the diagnosis of the pituitary form of Cushing syndrome. Similarly, when examined against the strict gold standard of the results of pituitary surgery, the likelihood ratios of each of the tests range from 1.44 to 2.21, with the likelihood ratio diminishing as tests are combined. Such results would suggest that combining tests makes them predictively worse. However, the low likelihood ratios are produced because of the indeterminate patients; when these cases are resolved through follow-up, infinite likelihood ratios are produced Table 2, suggesting that the tests may all be useful in distinguishing between the ectopic ACTH and pituitary-dependent Cushing syndrome.

Receiver Operating Characteristics

Both the intratest and intertest accuracies of the two end points of each of the two tests were equivalent when assessed by the area under the ROC curve. Thus, for the metyrapone test, the area under the ROC curve for 11-deoxycortisol did not differ significantly (P > 0.2) from the area under the ROC curve for 17-OHS (Figure 4). Similarly, for the dexamethasone suppression test, the area under the ROC curve for 17-OHS did not differ significantly (P > 0.2) from the area under the ROC curve for urine free cortisol (Figure 4). The ROC curves for the combination criteria are also given in Figure 4, as are the ROC curves that result from the assignment of the 14 patients found on follow-up to have the pituitary Cushing syndrome. Because these patients were the ones that prevented the defined criteria from having a specificity of 100% for the diagnosis of the Cushing syndrome, the curves now extend to the ordinate for each of the tests and have a correspondingly greater area under the curve. In this revised analysis, each of the individual tests was again found to have equivalent areas under the ROC curve. The combined dexamethasone and combined metyrapone tests for both analyses were also found to have areas under the ROC curve that did not differ significantly (P > 0.2).

View larger version (26K): [in this window] [in a new window]   Figure 4. Receiver operating characteristic curves for the diagnosis of the Cushing syndrome in patients with the ACTH-dependent Cushing syndrome. Parts A, B, and C compare the tests against the strict gold standard of response to pituitary surgery or positive surgical pathologic specimens. Parts D, E, and F compare the tests against the diagnoses made either at surgery or at follow-up. Parts A and D. Receiver operating characteristic curves of the metyrapone test for plasma 11-deoxycortisol (gray line) and urine 17-hydroxysteroid (striped line) levels, and a composite rule that combined both endpoints of the metyrapone test (black line). Parts B and E. Receiver operating characteristic curves for the high-dose dexamethasone test for urine free cortisol (gray line), and urine 17-hydroxysteroid (striped line), and a composite rule that combined both end points of the dexamethasone test (black line). Parts C and F. Receiver operating characteristic curves for the composite rules for the high-dose dexamethasone test using end points from both urine free cortisol and 17-hydroxysteroid (gray line), the metyrapone test using end points from both plasma 11-deoxycortisol and 17 hydroxysteroid (striped line), and the combined composite rule using all four end points (black line).

Discussion

Top Methods Results Discussion Author & Article Info References

On the basis of a retrospective pilot analysis of metyrapone testing in 186 patients who had surgery for the ACTH-dependent Cushing syndrome, we have determined criteria to distinguish patients with the pituitary Cushing syndrome from patients with the ectopic ACTH syndrome for the metyrapone test. A 70% increase in urine 17-OHS levels or a 400-fold increase over the baseline value of plasma 11-deoxycortisol levels identified 60% and 38%, respectively, of patients with surgically proven pituitary Cushing syndrome without misclassification of any patient with surgically proven ectopic ACTH syndrome. Thus, in our series, when the results of pituitary surgery were used as the "gold standard," urine 17-OHS proved to be a more useful diagnostic end point than did plasma 11-deoxycortisol. However, combining the 11-deoxycortisol and urine 17-OHS results further increased the sensitivity to 72%, significantly higher than with either endpoint alone. Allowing patients to be diagnosed with the Cushing syndrome on the basis of their follow-up as well as their pituitary surgery changed the results such that these criteria had a specificity of 100% and a sensitivity of 88% for the diagnosis of the Cushing syndrome.

The criteria proposed here for the diagnosis of the Cushing syndrome by the metyrapone test yielded a diagnostic sensitivity and specificity identical to that of the more widely used high-dose dexamethasone suppression test. For example, the best single end point for each test (a 17-OHS stimulation of >70% for the metyrapone test and a urine free cortisol suppression of >90% for the dexamethasone test) yielded sensitivities of 60% and 59%, respectively, for this series of patients, without misclassification of any of the patients with proven ectopic ACTH secretion. Similarly, the best combined end point criteria (a 17-OHS stimulation of >70% or an increase of 11-deoxycortisol of >400-fold over the baseline values for metyrapone, and a 17-OHS suppression of >69% or a urine free cortisol suppression of >90% for dexamethasone) gave sensitivities of 72% for both the metyrapone test and the dexamethasone test. Finally, the ROC analysis showed similar diagnostic efficiency of the two tests across the entire range of specificity. Thus, these data do not support the claim, based on fewer patients and different diagnostic criteria, that the metyrapone test is less accurate than the dexamethasone test for the differential diagnosis of the ACTH-dependent Cushing syndrome [17].

When the conventional outcome of the 17-OHS end point was used in our series (the higher value of day 3 or 4 instead of the value for day 3 alone compared with baseline), there was an unacceptably low sensitivity of 25% that did not lead to misclassification of any patient with proven ectopic ACTH syndrome. When, as suggested by Crapo [11], the criterion of an increase of 17-OHS levels of greater than 0% for the diagnosis of the pituitary Cushing syndrome was examined (using the higher 17-OHS value of day 3 or day 4 as the end point), a sensitivity of 96% and a specificity of 80% was found in our patients compared with a sensitivity of 98% and a specificity of 54% in Crapo's meta-analysis of the metyrapone test from 11 different series [11].

When evaluated against the gold standard of pituitary surgery, the combination test that used results of both the metyrapone and dexamethasone tests improved the test sensitivity from 72% to 88% and the diagnostic accuracy from 71% and 72%, respectively, to 82%, without misclassification of any of the patients with proven ectopic ACTH syndrome. When the diagnosis of the pituitary form of Cushing syndrome was made on the basis of both follow-up and pituitary surgery, the combined dexamethasone and metyrapone criteria yielded a specificity of 100%, a sensitivity of 88%, and a diagnostic accuracy of 89% for the diagnosis of the Cushing syndrome. We hypothesize that the advantage of combining the tests reflects their different physiologic bases: That an ACTH-secreting adenoma fails to stimulate when cortisol levels are lowered does not necessarily imply that it will fail to suppress when glucocorticoid levels are increased and vice versa. In a previous smaller study, we showed that combining the results of a corticotropin-releasing hormone test and the dexamethasone test could also improve sensitivity [32]. These observations emphasize the clinical usefulness of combining the results of several diagnostic tests when an unequivocal diagnosis cannot be reached with a single test.

These diagnostic criteria for the metyrapone test depend only on the results obtained during the days before and of administration of metyrapone. Thus, to improve cost-effectiveness, the test protocol can be shortened to 48 hours without loss of diagnostic accuracy. The revised protocol is as follows: Beginning at 0800 h, 24-hour urine specimens are obtained for 2 days for measurement of 17-OHS levels. Plasma 11-deoxycortisol levels are measured at 0 and 24 hours (before administration of metyrapone) and at 48 hours (after administration). Metyrapone is administered at 750 mg every 4 hours from 0800 h to 0400 h on the second day of the test.

The interpretation of the metyrapone test result is subject to several caveats, which also apply to the dexamethasone suppression test, the corticotropin-releasing hormone test, and the inferior petrosal sinus sampling procedure. First, none of these tests is accurate in patients with a pseudo-Cushing state rather than the true Cushing syndrome. Thus, in cases of mild to moderate hypercortisolism (urine free cortisol levels, 270 to 1000 nmol/day), the diagnosis of a pseudo-Cushing state should be excluded by the combined dexamethasone and corticotropin-releasing hormone test, by diurnal measurements of cortisol, or by the 2-day, low-dose dexamethasone suppression test if corticotropin-releasing hormone is unavailable [2, 33]. Second, the absence of stimulation by metyrapone in patients with the ectopic ACTH syndrome depends on suppression of the normal corticotropin-releasing hormone and ACTH axis by long-standing hypercortisolism. Patients with ectopic ACTH secretion who have the periodic Cushing syndrome or who have been treated recently with adrenal steroidogenesis inhibitors may not have suppression of the normal corticotropin-releasing hormone and ACTH axis and thus may have false-positive test results for the Cushing syndrome. Additionally, patients receiving long-term phenytoin therapy may have an accelerated metabolism of metyrapone. An inhibition of 11a-hydroxylase activity, similar to that of normal patients, can be achieved in such patients by doubling the regular dose of metyrapone to 750 mg every 2 hours [34].

The cut-off points (criteria) described in this pilot study cannot be considered absolute because they are defined by our data set and may not apply to a different patient population. Furthermore, the percentage of stimulation of urine steroids after administration of metyrapone or suppression after administration of dexamethasone in an individual patient can be affected by variations in urine collection, laboratory errors, and daily fluctuations in steroid excretion. Thus, data from patients with widely fluctuating urine excretion of steroids whose percentage of stimulation by metyrapone or suppression by dexamethasone is close to these cut-off points should be interpreted with some caution. Additionally, the number of patients with proven ectopic ACTH syndrome (n = 15) is relatively small. As more patients with the ectopic ACTH syndrome are studied, the cut-off points may have to be increased to avoid misclassifying patients who have proven ectopic ACTH syndrome, with a corresponding decrease in sensitivity. An example of this is provided by the dexamethasone suppression cut-off point, which in a earlier study was reported as a 17-OHS suppression of greater than 64% [7] but was increased in our study to a suppression of greater than 69% to avoid misclassifying any of the patients with proven ectopic ACTH syndrome.

Several potential sources of bias should be considered in evaluating our study. First, the NIH receives referrals of patients with the Cushing syndrome because of its extensive experience in diagnostic testing, particularly diagnostic radiology, inferior petrosal sinus sampling, and surgical treatment, and its case mix may be uniquely biased [35]. Our series has few patients with readily diagnosed causes of ectopic ACTH secretion such as oat cell carcinoma. Second, among the 441 patients with the ACTH-dependent Cushing syndrome referred to the NIH during the period of this study, only 186 received both the dexamethasone and metyrapone tests and had a surgical procedure to remove the source of excess ACTH. Although the selection of these patients was dictated primarily by logistical factors and investigator interest, we cannot exclude the possibility that some of the patients included in this report were studied to help resolve borderline results from the other tests. Thus, this subgroup may have included patients presenting unusual diagnostic difficulty. Third, 27 of the patients who received both dexamethasone and metyrapone testing did not have surgery and were therefore excluded from analysis. Finally, none of the patients had trans-sphenoidal surgery, the procedure determining disease state, without having first had bilateral inferior petrosal sinus sampling. This may have raised both the specificity and sensitivity of the evaluated tests [35].

What do we see as the role of the metyrapone test in the differential diagnosis of the Cushing syndrome? The diagnostic evaluation should begin with a screening test, either the 24-hour urine free cortisol or the overnight 1-mg dexamethasone test [36, 37], that indicates the presence of hypercortisolism. If the hypercortisolism is mild to moderate (270 to 1000 nmol/day) and the signs of the Cushing syndrome are equivocal, the next step is to differentiate the true Cushing syndrome from a pseudo-Cushing state. This can be done with the recently developed dexamethasone and corticotropin-releasing hormone test [33] and can be confirmed by measurement of the diurnal variation in morning and late evening serum cortisol levels. If the patient has the Cushing syndrome, ACTH should be measured to distinguish the ACTH-dependent from the ACTH-independent causes (see Figure 1).

A normal or elevated ACTH level indicates an ACTH-dependent cause, which could be either the pituitary Cushing syndrome or the ectopic ACTH syndrome. Our study establishes the metyrapone test to be as effective as the high-dose dexamethasone suppression test for differentiating these two conditions. By combining the metyrapone and dexamethasone tests (and the corticotropin-releasing hormone test when it becomes more widely available), it should be possible to diagnose the Cushing syndrome noninvasively with high accuracy, even without positive pituitary imaging. This may reduce the need for inferior petrosal sinus sampling in diagnosing the Cushing syndrome, although the verification biases [35] inherent in this study make this speculative. Petrosal sinus sampling will undoubtedly remain necessary for diagnosis in a few patients in whom the noninvasive tests give equivocal or conflicting results. Thus, we see the principal role of the metyrapone test as complementing the available noninvasive tests and thereby reducing the number of patients who must be diagnosed by petrosal sinus sampling.