X/97/$03.00/0 Vol. 82, No. 6 Journal of Clinical Endocrinology and Metabolism Copyright 1997 by The Endocrine Society

Transcription

1 X/97/$03.00/0 Vol. 82, No. 6 Journal of Clinical Endocrinology and Metabolism Printed in U.S.A. Copyright 1997 by The Endocrine Society Effectiveness Versus Efficacy: The Limited Value in Clinical Practice of High Dose Dexamethasone Suppression Testing in the Differential Diagnosis of Adrenocorticotropin-Dependent Cushing s Syndrome DAVID C. ARON, HERSHEL RAFF, AND JAMES W. FINDLING Division of Clinical and Molecular Endocrinology and Health Care Research Section, Department of Veterans Affairs Medical Center, and Case Western Reserve University School of Medicine (D.C.A.), Cleveland, Ohio 44106; and the Endocrine-Diabetes Center, St. Luke s Medical Center and Medical College of Wisconsin (H.R., J.W.F.), Milwaukee, Wisconsin ABSTRACT High dose dexamethasone suppression testing has been widely employed in the differentiation between pituitary ACTH-dependent hypercortisolism [Cushing s disease (CD)] and the ectopic ACTH syndrome. We hypothesized that the high dose dexamethasone suppression test as it is performed in practice does not improve the ability to differentiate between these two types of ACTH-dependent Cushing s syndrome. Cases were drawn from 112 consecutive patients with ACTH-dependent Cushing s syndrome, who were then classified based upon results of inferior petrosal sinus sampling for ACTH levels. Analysis of test characteristics of high dose dexamethasone suppression testing was performed in the 73 patients for whom results are available. Statistical modeling was performed using the 68 cases with complete data on all assessed variables. Logistic regression models were used to predict the probability of pituitary-dependent Cushing s syndrome (CD) given the results of high dose dexamethasone suppression testing before and after adjustment for the contribution of a series of potential covariates. Of the 112 patients with ACTH-dependent Cushing s syndrome, 15.2% had the ectopic ACTH syndrome, and the remainder had pituitary-dependent Cushing s syndrome (CD). Patients with the ectopic ACTH syndrome were significantly older (mean, 51.9 vs. 40.2), were more likely to be male (58.8% vs. 27.4%), had shorter duration of clinical findings (mean, 11.6 vs months), were more likely to have hypokalemia (50% vs. 8.6%), had higher baseline 24-h urinary free cortisol [mean, 8317 vs nmol/day (3015 vs. 422 g)] and plasma ACTH levels [mean, 47 vs. 17 pmol/l (210 vs. 78 pg/ml)] and were less likely to suppress urinary free cortisol or plasma cortisol with high dose dexamethasone using the standard criterion of 50% or more suppression compared with patients with pituitary-dependent THE MAJOR challenge in the differential diagnosis of ACTH-dependent hypercortisolism is identifying the ACTH-secreting tumor (1 4). The majority of these patients have a pituitary microadenoma [Cushing s disease (CD)], whereas the others harbor a nonpituitary tumor [the ectopic ACTH syndrome (EAS)]. Many of these tumors, both pituitary and ectopic, are small, making their radiological localization difficult. As pituitary microsurgery is the treatment Received October 28, Revision received January 10, Rerevision received February 24, Accepted February 28, Address all correspondence and requests for reprints to: David C. Aron, M.D., M.S., Medical Service 111(W), Veterans Administration Medical Center, East Boulevard, Cleveland, Ohio aron.david@cleveland.va.gov. Cushing s syndrome. Based upon the standard criterion, the sensitivity and specificity of the high dose dexamethasone suppression test for the diagnosis of pituitary-dependent Cushing s syndrome were 81.0% and 66.7%, respectively. Although the mean percent suppression was significantly greater for patients with CD than for those with the ectopic ACTH syndrome (72.2% vs. 41.3%), the range of suppression was 0 99% for each diagnosis. The area under the receiver operating characteristic curve was (95% confidence interval, ). Logistic regression models were used to evaluate the probability of CD given the responsiveness to high dose dexamethasone suppression testing before and after adjustment for the potential contributions of other factors. A model including all of the variables (age, sex, duration, presence of hypokalemia, urinary free cortisol, and plasma ACTH) had a diagnostic accuracy of 92.7%. A model including all of these variables plus a binary variable indicating whether the patient met the criterion of suppression by 50% or more resulted in 95.6% accuracy, whereas substitution of this binary variable by percent suppression resulted in a model with 94.1% accuracy. There were no statistically significant differences among these models; their values for the c statistic, which is equivalent to the area under the curve in a receiver operating characteristic analysis, were all greater than 0.9. Logistic regression models indicate that the results of the dexamethasone suppression test add little to the differential diagnosis of ACTH-dependent Cushing s syndrome, especially after taking other clinical information into account. In our patient population, the sensitivity and specificity of the dexamethasone suppression test were less than those reported by others. However, because 20 33% of cases of ectopic ACTH syndrome are misdiagnosed with these logistic regression models, other techniques are necessary to achieve greater diagnostic accuracy. (J Clin Endocrinol Metab 82: , 1997) of choice in patients with CD, accurate diagnosis is critical. High dose dexamethasone suppression testing has been a mainstay of biochemical differential diagnosis, either with measurement of basal and suppressed urinary 17-hydroxycorticosteroids or urinary free or plasma cortisol before or after the administration of 2 mg dexamethasone every 6 h for 2 days or as an overnight test with measurement of plasma cortisol before and after a single 8-mg dose. This test is based on the observation that in pituitary-mediated disease, ACTH secretion tends to retain some degree of responsiveness to both hypophysiotropic factors and glucocorticoid negative feedback, whereas tumors responsible for ectopic production of ACTH tend not to do so. The sensitivity of the high dose dexamethasone suppression test has been reported to range 1780

2 DEXAMETHASONE TESTING IN CUSHING S SYNDROME 1781 from % and from 59 92% for the 2-day and overnight tests, respectively; similarly, the specificity has been reported to range from % and from % (1 3, 5 12). There are three major concerns about use of the published results in clinical practice. First, in most series, patients with EAS account for 10 20% of patients with ACTH-dependent Cushing s syndrome. Therefore, as the pretest probability of a pituitary etiology is 80 90%, test performance must be extremely good to be useful. Second, recommendations have not taken into account sufficiently the distinction between efficacy and effectiveness. As applied to diagnostic testing, efficacy refers to the degree to which the test has been shown scientifically to accomplish the desired outcome. In contrast, effectiveness refers to the degree to which the test achieves this outcome in actual clinical practice. Most large studies have been performed in research venues and, thus, are efficacy studies, whereas the effectiveness of tests in practice has not been extensively evaluated. We sought to determine the incremental value of high dose dexamethasone suppression testing as it is performed in clinical practice. Materials and Methods Study design and population Cases were drawn from 112 consecutive patients with ACTHdependent Cushing s syndrome who were referred to a single clinician (J.W.F.) from 1982 to All patients underwent inferior petrosal sinus sampling for ACTH levels, for which informed consent was obtained. This procedure was used as the gold standard to classify patients because of its established high accuracy (1, 2, 13) and the variable results of pituitary surgery (14). Diagnosis of CD was established by a petrosal sinus:peripheral ACTH gradient of more than 2 either at baseline or after iv injection of CRH (13). Analysis of test characteristics of high dose dexamethasone suppression testing was performed in the 73 patients in whom that test was performed. Statistical modeling was performed using the 68 cases with complete data for all assessed variables (see below). Separate analysis of patients with pathologically proven diagnosis was performed. Logistic regression models were used to predict the probability of pituitarydependent Cushing s syndrome (CD) given the results of high dose dexamethasone suppression testing before and after adjustment for the contribution of a series of potential covariates. Biochemical and radiological testing All patients were examined by one clinician (J.W.F.), who estimated the duration of disease from the patients history. Most biochemical test results were provided by the referring physician(s). Hypokalemia was defined as a serum K below 3.5 meq/l. Measurement of urinary free cortisol used a variety of methods, and no adjustments were made in the models; all but four patients had values above the upper limit of normal of all the urinary free cortisol assays. Of those four patients, only the one who had complete data was included in logistic models. High dose dexamethasone suppression testing was performed by the referring physician or the consultant using either the overnight (34 patients) or 2-day test (39 patients). The results from these two tests were pooled; no special efforts were made to assess the validity of the data, e.g. correctness of test performance. All patients had measurements of plasma ACTH in a single laboratory using a two-site immunoradiometric assay (15). Interpretation of magnetic resonance imaging (MRI) or high resolution computed tomographic (CT) scanning of the sella was made independently of the results of petrosal sinus sampling. Statistical analysis In addition to descriptive, parametric, and nonparametric statistics, we used contingency tables and receiver operating characteristic (ROC) curves. Logistic regression models were used to predict the probability of CD (pituitary ACTH-dependent Cushing s syndrome) given the results of high dose dexamethasone suppression testing before and after adjustment for the contribution of potential covariates. These models take the following mathematical form: ln{[p(cd)]/[1 P(CD)]} 0 1 (variable) 2 (variable)..., where P(CD) is the probability of CD, and the variables include age, sex, ACTH, etc. Differences between models were assessed with likelihood ratio tests and the c statistic, which is equivalent to the area under the curve in a ROC curve analysis (16 19). Analyses were performed using SPSS 6.12 (Chicago, IL) and ROC Analyzer (20). P 0.05 was assumed to be statistically significant; no correction was made for multiple statistical tests (21). Results Patient characteristics The patient characteristics for the entire series, stratified by diagnosis (n 112), are shown in Table 1 and stratified by whether dexamethasone suppression testing was performed (n 73) or not (n 39) in Table 2. The frequency of the EAS in the entire series was 15.2% (Table 1). As a whole, patients with EAS were significantly older (mean, 51.9 vs yr), were more likely to be male (58.8% vs. 27.4%), had a shorter duration of clinical findings (mean, 11.6 vs months), and were more likely to have hypokalemia (50% vs. 8.6%) than patients with CD. Patients with EAS also had significantly higher baseline 24-h urinary free cortisol [mean, 8317 vs nmol/day (3015 vs. 422 g)] and plasma ACTH levels [mean, 47 vs. 17 pmol/l (210 vs. 78 pg/ml)] and were less likely to suppress with high dose dexamethasone using the standard criterion of 50% or greater suppression (Fig. 1). However, one third of the patients with EAS met this criterion for suppression (see below). Imaging of the sella turcica (MRI or high resolution CT scanning) was performed in 108 patients. There was no significant difference in the frequency of abnormal pituitary imaging between those with CD and those with EAS (18.5% vs. 12.5%). The positive predictive value of an abnormal pituitary imaging study (true positives/all positives) was 89.5%, which compares with the frequency of CD of 85.2% among those who underwent pituitary imaging. Among the patients who had undergone high dose dexamethasone suppression, the positive predictive value of pituitary imaging was 77.9%, which compares with the frequency of CD of 80.0% in this subgroup. Follow-up information on pathological diagnosis was available for 74 patients diagnosed as having CD who underwent transsphenoidal surgical exploration. Confirmation of the diagnosis was based on the presence of a pituitary tumor or cure after partial or total hypophysectomy in the absence of an identifiable pituitary lesion. Two patients had corticotroph cell hyperplasia. Thus, pituitary-dependent disease was confirmed in 84%. Of the 17 patients diagnosed with EAS, a pathological diagnosis was made in 13 (71%), most of which were bronchial carcinoid tumors. Test characteristics of high dose dexamethasone suppression testing Results were available for 73 patients; EAS occurred in 20.5% of this subgroup (Table 2). Based upon the standard criterion, i.e. suppression by 50% or more of the baseline, the sensitivity and specificity of the test were 81.0% and 66.7%, respectively. The mean 1 sd percent suppression was sig-

3 1782 ARON, RAFF, AND FINDLING JCE&M 1997 Vol 82 No 6 TABLE 1. Characteristics of patients with Cushing s disease and the ectopic ACTH syndrome Cushing s disease Ectopic ACTH syndrome P n Age Mean 1 SD a Range % Female b Duration in months Mean 1 SD a Range % with hypokalemia b 24-h urinary free cortisol (nmol/day) Mean 1 SD , a Range ,829 Plasma ACTH (pmol/l) Mean 1 SD a Range % with suppression 50% 81.0 (n 58) 33.3 (n 15) c % Suppression range % with abnormal pituitary MRI or CT scan c Numbers in parentheses refer to cases who underwent dexamethasone suppression testing. Conversion factors: urinary free cortisol: g/24 h nmol/day; plasma ACTH: pg/ml 4.5 pmol/l. a By Mann-Whitney U test. b By Pearson 2 test. c By Fisher s exact test. TABLE 2. Characteristics of patients stratified by whether high dose dexamethasone suppression (HDD) testing was performed Patients who underwent HDD Patients who did not undergo HDD P n a Age Sex (% female) b Duration (months) a % with hypokalemia b 24-h urinary free cortisol (nmol/day) a Plasma ACTH (pmol/l) a % with ectopic ACTH syndrome b Conversion factors: urinary free cortisol: g/24 h nmol/day; Plasma ACTH: pg/ml 4.5 pmol/l. a By Mann-Whitney U test. b By Pearson 2 test. nificantly greater for patients with CD than for those with EAS ( % vs %; P 0.001, by unpaired t test). The range of suppression was 0 99% for each diagnosis. There was no cut-off point that yielded 100% specificity. The ROC curve for this test is shown in Fig. 2. The area under the curve is 0.710, which was significantly greater than that occurring by chance (95% confidence interval, ). Logistic regression modeling Logistic regression models were used to evaluate the probability of CD given responsiveness to high dose dexamethasone suppression testing before and after adjustment for the potential contributions of other factors. These studies were based on a subset of 73 patients who underwent high dose dexamethasone suppression testing and for whom complete data were available. There were 68 such patients in this subset, and 77.9% of them had CD. The diagnostic accuracy of the models is shown in Table 3. Model 1 included the variables age, sex, duration, presence of hypokalemia, urinary free cortisol, and plasma ACTH plus suppression by 50% or more (a binary variable). Model 2 included these variables, except for percent suppression (a continuous variable), which was substituted for suppression by 50% or more. These models were more accurate than models 5 and 6, which included only the response to dexamethasone. Similar results were observed using criteria for suppression of 60%, 70%, 80%, and 90% or more. Model 3 included the variables age, sex, duration, presence of hypokalemia, urinary free cortisol, and plasma ACTH, but not the results of high dose dexamethasone suppression testing. Model 4 was developed using a stepwise technique (forward likelihood ratio). There were no statistically significant differences among models 1, 2, 3, and 4 (by likelihood ratio test, P 0.1). These models had values for the c statistic, which is equivalent to the area under the curve in a ROC analysis, of 0.946, 0.936, 0.937, and 0.912, respectively. These models correctly diagnosed 53, 52, 52, and 52 of the 53 cases of CD and 12, 12, 11, and 10 of the 15 cases of EAS, respectively. Evaluation of these models in patients with pathologically confirmed diagnoses yielded similar results (data not shown). To assess some of the logistic regression models, we then analyzed those patients who did not undergo high dose dexamethasone suppression. The characteristics of these patients are shown in Table 2. There were no significant differences between those patients who underwent dexameth-

4 DEXAMETHASONE TESTING IN CUSHING S SYNDROME 1783 FIG. 1. Biochemical characteristics of patients with CD and EAS. Left panel, Twenty-four-hour urinary free cortisol (n 99 CD; n 15 EAS); middle panel, plasma ACTH (n 102 CD; n 17 EAS); right panel, percent suppression with high dose dexamethasone (n 61 CD; n 15 EAS). but only 7 of 15 with EAS. Model 4 had a diagnostic accuracy of 90.2%. We also evaluated models that included the results of pituitary imaging in the subgroup of patients who had also undergone high dose dexamethasone suppression. Models 1 and 2 had diagnostic accuracies of 95.4% and 93.9%, respectively. Addition of pituitary imaging (binary variable) to either of these models resulted in a diagnostic accuracy of 93.9%, which was the same as the model analogous to model 3 with the addition of pituitary imaging results. There were insufficient cases of the EAS who had pituitary tumors detected by MRI or CT to evaluate a model in the subset of patients with positive results on pituitary imaging. However, of the two patients with EAS who had abnormal pituitary imaging, one showed no suppression in response to high dose dexamethasone, and the other suppressed by 82%. FIG. 2. ROC curve for high dose dexamethasone suppression testing. The dotted line represents the results equivalent to chance alone. The area under the curve is 0.71, which was significantly greater than that occurring by chance (95% confidence interval, ) asone suppression testing and those who did not, except for the lower frequency of EAS in the latter group. Differences in the results of the logistic regression models for these two populations would suggest selection bias (in who undergoes dexamethasone suppression testing) or other problems, e.g. overfitting. However, there were insufficient numbers of patients with EAS who did not undergo high dose dexamethasone suppression testing. Therefore, we analyzed logistic regression models for the entire population who had complete data (n 102). Model 3 had a diagnostic accuracy of 91.2% and correctly identified 86 of 87 patients with CD, Discussion This study demonstrates that dexamethasone suppression testing as performed in practice is not only inaccurate in the differential diagnosis of Cushing s syndrome, but also provides no incremental value over clinical observations in a simple logistic regression model. Although patients with EAS differed as a group from those with CD, none of our models with or without dexamethasone suppression testing was able to accurately characterize 20 35% of patients with nonpituitary ACTH-secreting tumors, emphasizing the need for a more accurate diagnostic study. Traditionally, high dose dexamethasone suppression testing has been used in the differential diagnosis of Cushing s syndrome. The diagnostic accuracy of dexamethasone suppression testing may be compromised by incomplete urine collections, daily fluctuation in basal steroid excretion, improper timing of plasma cortisol collection, as well as possible inaccuracy in measurement of urinary steroid levels caused by noncompliance, medications, renal and hepatic disease, or poor laboratory performance. The major problem with high dose dexamethasone suppression tests is their diagnostic inaccuracy, which is readily apparent in series

5 1784 ARON, RAFF, AND FINDLING JCE&M 1997 Vol 82 No 6 TABLE 3. Logistic regression modeling of probability of Cushing s disease Model no. Variables Patients who underwent high dose dexamethasone test Sensitivity Specificity Diagnostic accuracy Population with complete data on indicated variables (n 102) Sensitivity Specificity Diagnostic accuracy 1 Age, sex, duration, hypokalemia, urinary free cortisol, plasma ACTH, suppression by 50% 2 Age, sex, duration, hypokalemia, urinary free cortisol, plasma ACTH, % suppression 3 Age, sex, duration, hypokalemia, urine free cortisol, plasma ACTH 4 Duration, hypokalemia, plasma ACTH Suppression by 50% % Suppression with a substantial number of patients with EAS (1, 2). At least 20 30% of patients with EAS will suppress plasma and urinary steroids to less than 50% of baseline values during dexamethasone suppression testing. In addition, as many as 20 30% of patients with CD fail to suppress steroid levels to less than 50%. Consequently, the diagnostic accuracy of high dose dexamethasone suppression testing is only 70 80%. This must be compared with a pretest probability of CD in patients with ACTH-dependent CD of about 85 90%, which exceeds the sensitivity, specificity, and diagnostic accuracy of dexamethasone suppression testing. A recent study analyzed new criteria for the standard low and high dose dexamethasone suppression to identify better specificity, sensitivity, and accuracy. Flack et al. (12) evaluated 118 patients with surgically confirmed causes of CD (94 with CD, 14 with primary adrenal disease, and 10 with EAS). Their study confirmed the very poor sensitivity, specificity, and accuracy of the high dose dexamethasone suppression test when the 50% suppression criterion was used. However, a decrease in urinary free cortisol of more than 90% and a decrease in 17-hydroxycorticosteroid secretion of more than 64% had 100% diagnostic specificity for CD and excluded EAS. Some patients with CD did not suppress to these levels, and the overall accuracy using these new criteria is 85%, which is about the same as the pretest probability of the disease. Moreover, in subsequent publications, the cut-off point to achieve 100% specificity had to be revised (11). Our data indicate that there is no such cut-off point. Analysis of the ROC curve generated from our data confirm the poor performance of high dose dexamethasone suppression testing. The area under the curve, although significantly better than chance (0.5), was still only This value is less than those reported by Dicheck for the standard test (0.903) and the overnight test (0.867), although statistical comparisons cannot be made using the published data. We studied the effectiveness of high dose dexamethasone suppression testing, not its efficacy. Our analyses relied upon the results supplied by the referring physicians. Although no special effort was made to assess the correctness of the test performance, these are the results that are used in actual practice. Moreover, this means that a variety of clinical laboratories and methods were used to measure plasma and urinary free cortisol. This approach is not likely to provide the precision found in efficacy studies performed in research centers using a single methodology and batch assays. Similarly, it is not likely that dexamethasone suppression testing in a real world setting can match that achieved in the research setting. Therefore, our data are more representative of the results likely to be obtained by practicing physicians. A limitation of the study is the possibility of selection bias, especially because disease prevalence is important in generalizing clinical prediction rules (22). Ideally, evaluation of dexamethasone suppression testing should be performed on a population-based, rather than referral-based, sample. Naturally occurring Cushing s syndrome is a rare disease. Consequently, all large series of patients reflect referral patterns. Bias may result from the referral of difficult cases for specialized expertise, such as inferior petrosal sinus sampling. These difficult cases may be those with atypical laboratory results, e.g. the low frequency of abnormal pituitary imaging in the patients with CD (18.5%) (23, 24). Interestingly, the frequency of abnormal pituitary imaging in patients with the EAS (12.5%) was similar to that reported in normal subjects (25). However, the frequency of EAS in our series was similar to that in other reports. Moreover, 39 patients were referred without having undergone dexamethasone suppression testing, and the models performed similarly when these patients were included. Our study used inferior petrosal sinus sampling as the gold standard, raising the possibility that patients may have been misclassified. Surgical confirmation of a pituitary source of ACTH was achieved in 84% of cases diagnosed as CD. Confirmation of the ectopic source of ACTH was achieved in 71% of cases diagnosed as EAS. In addition to the possibility of misclassification, these results may reflect variation in surgical results as well as the occult nature of many of these ACTH-secreting tumors. However, similar results for the statistical models were observed when the analyses were limited to patients with pathologically confirmed diagnoses. Estimation of duration of disease is subject to both ascertainment and recall bias. Finally, validation of the models in other populations is needed, recognizing that application of clinical prediction rules to individual patients is problematic (26 28). Determining cost-effective diagnostic strategies requires careful evaluation not only of a test in isolation, but also in the context of the other information available and the likelihood of disease. Consideration must be given to the ques-

6 DEXAMETHASONE TESTING IN CUSHING S SYNDROME 1785 tion of the value added by a test or procedure. Flagle wrote that the value of information is equal to the enhanced value of outcomes based on that information (29). We conclude that high dose dexamethasone suppression testing has limited incremental value in the differential diagnosis of EAS. Clinical prediction rules notwithstanding, given a male patient with a rapid course and high plasma ACTH and urinary free cortisol levels, the probability of EAS is sufficiently high that the results of dexamethasone suppression testing should not influence the decision about management. Furthermore, given the range of suppression from 0 99% for both CD and EAS, it is clear that, all other things being equal, the pretest probability of CD (usually 90%) will exceed the diagnostic accuracy of dexamethasone suppression testing. On this basis alone, we can recommend that the high dose dexamethasone suppression test be abandoned. We used clinical prediction rules to confirm the probability of CD, and diagnostic accuracy using simple clinical measures was high. However, even if our clinical prediction rules were validated using other series, their utility in individual patients may be limited. We found that 20 33% of cases of EAS are misdiagnosed with these models. Neither clinical features nor routine biochemical tests alone or in combination can establish the diagnosis with sufficient diagnostic accuracy to ensure that appropriate therapy is given to all patients. Although in individual cases, the diagnosis may be clear, in most, if not all, other techniques, such as petrosal sinus sampling, are necessary to achieve the requisite diagnostic accuracy. References 1. Findling JW, Doppman JL Biochemical and radiologic diagnosis of Cushing s syndrome. Endocrinol Metab Clin North Am. 23: Kaye TB, Crapo L The Cushing syndrome: an update on diagnostic tests. Ann Intern Med. 112: Orth DN Cushing s syndrome. N Engl J Med. 332: Orth DN The Cushing syndrome: quest for the holy grail. Ann Intern Med. 121: Liddle GW Tests of pituitary-adrenal suppressibililty in the diagnosis of Cushing s syndrome. J Clin Endocrinol. 20: Crapo L Cushing s syndrome: a review of diagnostic tests. Metabolism. 9: Bruno OD, Rossi MA, Contreras LN, et al Nocturnal high-dose dexamethasone suppression test in the aetiological diagnosis of Cushing s syndrome. Acta Endocrinol (Copenh). 109: Tyrrell JB, Findling JW, Aron DC, et al An overnight high dose dexamethasone suppression test for rapid differential diagnosis of Cushing s syndrome. Ann Intern Med 104: Grossman AB, Howlett TA, Perry L, et al CRF in the differential diagnosis of Cushing s syndrome: a comparison with the dexamethasone suppression test. Clin Endocrinol (Oxf). 29: Hermus AR, Pesman GJ, Benraad T, Pieters GF, Smals AG, Kloppenborg PW The corticotropin-releasing hormone test versus the high-dose dexamethasone test in the differential diagnosis of Cushing s syndrome. Lancet. 2: Dichek HL, Nieman LK, Oldfield EH, Pass HI, Malley JD, Cutler Jr GB A comparison of the standard high dose dexamethasone suppression test, and the overnight 8-mg dexamethasone suppression test for the differential diagnosis of adrenocorticotropin-dependent Cushing s syndrome. J Clin Endocrinol Metab. 78: Flack MR, Oldfield EH, Cutler Jr GB, et al Urine free cortisol in the high-dose dexamethasone suppression test for the differential diagnosis of Cushing syndrome. Ann Intern Med. 116: Findling JW, Kehoe ME, Shaker JL, Raff H Routine inferior petrosal sinus sampling in the differential diagnosis of adrenocorticotropic (ACTH)- dependent Cushing s syndrome: early recognition of the occult ectopic ACTH syndrome. J Clin Endocrinol Metab. 73: Tyrrell JB, Wilson CB Cushing s disease. therapy of pituitary adenomas. Endocrinol Metab Clin North Am. 23: Raff H, Findling JW A new immunoradiometric assay for corticotropin evaluated in normal subjects, and patients with Cushing s syndrome. Clin Chem. 35: Hanley JA, McNeil BJ The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 143: Centor RM, Schwartz JS An evaluation of methods for estimating the area under the receiver operating characteristic (ROC) curve. Med Decision Making. 5: Hilden J The area under the ROC curve and its competitors. Med Decision Making. 11: Ash A, Schwartz M. Evaluating the performance of risk adjustment methods: dichotomous measures. In: Iezzoni LI, ed. Risk adjustment for measuring health outcomes Ann Arbor: Health Administration Press; Centor RM A Visicalc program for estimating the area under receiver operating characteristic (ROC) curve. Med Decision Making. 5: Savitz DA, Olshan AF Multiple comparisons and related issues in the interpretation of epidemiologic data. Am J Epidemiol. 142: Poses RM, Cebul RD, Collins M, Fager SS The importance of disease prevalence in transporting clinical prediction rules. Ann Intern Med. 105: Dwyer AJ, Frank JA, Doppman JL, et al Pituitary adenomas in patients with Cushing s disease: initial experience with Gd-DTPA enhanced MR imaging. Radiology 63: Escourolle H, Abecassis JP, Bertagna X, et al Comparison of computerized tomography and magnetic resonance imaging for the examination of the pituitary gland in patients with Cushing s disease. Clin Endocrinol (Oxf). 39: Hall WA, Luciano MG, Doppman JL, Patronas NJ, Oldfield EH Pituitary magnetic resonance imaging in normal human volunteers: occult adenomas in the general population. Ann Intern Med. 120: Wasson JH, Sox HC, Neff RK, Goldman L Clinical prediction rules. Applications and methodological standards. N Engl J Med. 313: Allison JJ, Centor RM Why models predicting bacteremia in general medical patients do not work. J Gen Intern Med. 11: Braitman LE, Davidoff F Predicting clinical states in individual patients. Ann Intern Med. 125: Flagle CD The value of information for decision making. In: Yaffee R, Zalkind D, eds. Evaluation in health services delivery: proceedings of an Engineering Foundation conference. New York: Engineering Foundation;