The addition of corticotropin-releasing hormone to 2-day low dose dexamethasone suppression test provides additional case detection

The diagnostic value of adding a Corticotropin-Releasing Hormone (CRH) Stimulation Test to the 2-day Low Dose Dexamethasone Suppression Test (Dex-CRH Test) has been debated in the literature. We identified 65 patients with Cushing disease (CD) and 42 patients in whom a diagnosis of Cushing disease could not be confirmed (NCD) after a minimum follow-up of 14 months who underwent the Dex-CRH test. The female sex ratio, median (range) age, and BMI were similar between the two groups. The follow-up for patients with CD and NCD was 74 (4–233) and 52 (14–146) months, respectively. Among 65 patients with CD, 5 (7.7%) had a cortisol level ≤1.4 µg/dl after LDDST but were appropriately classified as CD with a cortisol level >1.4 µg/dL at 15-min post CRH stimulation. In contrast, 3/42 patients (7.1%) in NCD had an abnormal Dex-CRH test. In only one of three patients, the LDDST was marginally normal (cortisol was 1.4 µg/dL and increased to 3.1 µg/dL 15-min post CRH). A cortisol cutoff value of >1.4 µg/dL during the Dex-CRH test provided a sensitivity of 100%, specificity of 93%, and diagnostic accuracy of 97% to diagnose CD. When patients without a Dex level were excluded (n = 74), the sensitivity did not change, but the specificity and accuracy of the Dex-CRH test increased to 97 and 99%, respectively. The Dex-CRH Test provided additional case detection in 5/65 (7.7%) patients with CD. It resulted in one false-positive case compared to LDDST. Measurement of dexamethasone improved diagnostic accuracy of the test.


Introduction
The workup for Cushing syndrome (CS) has become more challenging over the years due to the increased prevalence of diabetes, obesity, hypertension, and depression [1]. These conditions may present with clinical features similar to CS such as obesity, fatigue, hirsutism, and elevated blood pressure, and may be associated with physiologic overactivity of the hypothalamic-pituitary-adrenal (HPA) axis and are referred to as physiologic non-neoplastic hypercortisolism (formerly pseudo-Cushing's syndrome) [2]. Two-day low dose dexamethasone suppression test (LDDST) has been developed for the workup of CS. The Liddle Test was initially performed by giving dexamethasone (Dex) 0.5 mg every 6 h for two days and then measuring urinary 17-hydroxycorticoid excretion metabolites [3]. It was later found that measuring 8 AM serum cortisol rather than urinary cortisol or its metabolites during the Liddle Test is simpler and provides better discrimination [4][5][6][7].
A combination of 2-day LDDST followed by corticotropin-releasing hormone (CRH) stimulation, the Dex-CRH test, was first described in 1993 by Yanovski et al. [8] at the NIH. The authors proposed that this combined test is superior to LDDST in ruling out physiologic non-neoplastic hypercortisolism.
We report our experience with combined Dex-CRH test at the Cleveland Clinic in evaluating patients with suspected CD. Our main objective of the study was to determine if the addition of CRH provided any advantage over the two-day LDDST. In addition, we report the diagnostic characteristics of the Dex-CRH test and the impact of measurement of serum dexamethasone level on the diagnostic accuracy of this test.

Materials and methods
The study was approved by the institutional review board at the Cleveland Clinic Foundation. A total of 117 patients underwent the Dex-CRH test to evaluate ACTH-dependent Cushing syndrome between 2002 and 2014 in the Endocrinology Department at the Cleveland Clinic Foundation, Cleveland, Ohio. The patients were identified from a prospectively maintained pituitary database. All patients had variable symptoms and signs of CS at the time of presentation or referral. Ten patients were excluded from the analysis: three patients had ectopic ACTH syndrome with very high cortisol levels after both LDDST and Dex-CRH tests, two were lost to follow-up, two had cyclic CD, two were on Dilantin with no available Dex level, and one patient was suspected with noncompliance with Dex intake (Fig. 1). Accordingly, the data on a total of 107 patients were analyzed in this study.
Based on the preference of their endocrinologists, the patients have undergone additional testing, including 24 h UFC, late-night salivary cortisol, and 1 mg DST. However, due to the study's retrospective nature, no unified approach was used among the clinicians. We have included the additional biochemical studies of patients of special interest in Tables 2 and 3.
Cushing disease (CD) was defined and confirmed in 65 patients by either a positive immunohistochemistry for ACTH in the surgical histopathology (n = 50) or by the development of postoperative adrenal insufficiency (cortisol ≤2 µg/dL) after trans-sphenoidal pituitary surgery (n = 13) or were proven to have hypercortisolism with elevated ACTH level and failed trans-sphenoidal pituitary surgery and had to have bilateral adrenalectomy (n = 2). Among the patients with positive staining for ACTH on pathology, 58% had a nadir cortisol ≤2 ug/dL, and another 14.6% had a cortisol level between 3-5 ug/dL postoperatively. The Cushing syndrome could not be confirmed (NCD) in 42 patients with additional testing for 24-h urinary free cortisol, late-night salivary cortisol, and 1 mg DST for a median follow-up period of 52 months (range 14-146).

Study protocol
All patients underwent the combined Dex-CRH test as initially described by the NIH group. They were instructed to take eight doses of 0.5 mg of Dex every six hours orally starting at noon so that the last dose of dexamethasone would be at 6:00 AM. Intravenous ovine CRH injection (1 µg/kg, max dose 100 µg) was given two hours after the last Dex dose at 8:00 AM. Serial blood samples for cortisol were obtained from an indwelling venous catheter 15 min

Hormone assays
Serum cortisol level was measured by Siemens ADVIA Centaur XP Chemiluminescence immunoassay with an inter-assay and intra-assay coefficients of variation (CV) of 5.7-10.7 and 3.8-3.9%, respectively. Dexamethasone level was measured by liquid chromatography tandem mass spectrometry at ARUP laboratories (500 Chipeta Way, Salt Lake City, UT, 84108) with an inter-assay and intra-assay CV of <5.4 and >4.8%, respectively.

Statistical analysis
Data of all the records were entered into Research electronic data capture (Redcap) consortium system. Data is presented as median (range). Sensitivity, specificity, and test accuracy were calculated for all patients (n = 107) using Stata 15.1. Calculations were all repeated again after excluding those without available dexamethasone level (n = 74). A cortisol cutoff of >1.4 µg/dL was used to indicate a positive Dex-CRH test as per the initial description of the test by Yanovski et al. [8]. Cortisol cutoffs of >1.4 µg/dL were both used and compared to indicate a positive LDDST.

Results
Among 107 patients included in the analysis, 90 were females, and 17 were males. The sex ratio, median (range) age, and BMI were not statistically significant between the two groups ( Table 1). The follow-up duration for patients with CD and NCD was 74 (4-233) and 52 (14-146) months, respectively.
Six patients in NCD and two in CD groups were on estrogen-containing hormone therapy; all of them had post-Dex and 15-min post-CRH cortisol <1.4 µg/dL (true negative). Corticosteroid binding globulin (CBG) level was available in only two out of the eight patients on estrogen therapy, and it was normal in both (20 and 26 mg/L, respectively, normal range 19-45 mg/L).
The Median (range) cortisol levels post-Dex were as follows: 0. Among 65 patients with CD, we identified five female patients with a cortisol level of ≤1.4 µg/dL after LDDST but correctly identified when cortisol level increased to >1.4 µg/ dL at 15-min after CRH administration ( Table 2). Four of these patients had pathology-proven CD, and one had recurrent CD and underwent two unsuccessful transsphenoidal pituitary surgeries with ACTH levels ranging between 56-78 pg/mL (reference range 5-50 pg/mL). The   Table 3). One of them had a cortisol level <1.4 µg/dL 15-min post CRH, which suggested a normal result. All four patients were followed for more than 7 years with repeat testing, and no diagnosis of Cushing syndrome was made during follow-up. A summary of the patients' flow chart is provided in Fig. 1. The calculated test performance for the Dex-CRH test when using 15-min cortisol >1.4 µg/dL as the criterion for diagnosis of CD was as follows: sensitivity 100%, specificity 93%, and diagnostic accuracy 97%. The calculated test performance for LDDST when using post-Dex cortisol >1.4 µg/dL as the criterion for diagnosis of CD was as follows: sensitivity 92%, specificity 93%, and diagnostic accuracy 93% (sensitivity 91%, specificity 98%, and diagnostic accuracy 93% when using post-Dex cortisol >1.8 µg/dL as the criterion for LDDST).
The dexamethasone level was measured during the Dex-CRH test in 74 patients in this study. Four patients had a dexamethasone level <220 ng/dL. Among them, one patient had confirmed CD. The other three patients with Dex level <220 ng/dL were in the NCD group, and they all had undetectable cortisol levels, both post-Dex and 15-min post-CRH (true negative).
The performances of the LDDST and Dex-CRH tests when only analyzing patients with available Dex levels are as follows: For Dex-CRH sensitivity 100%, specificity 97% and diagnostic accuracy 99%. For LDDST when using post-Dex cortisol >1.4 µg/dL as the criterion, sensitivity 92%, 1 mg DST was reported as normal on at least two occasions in his medical record without documentation of the exact values specificity 97%, and diagnostic accuracy 95% (sensitivity 92%, specificity 100%, and diagnostic accuracy 96% when using post-Dex cortisol >1.8 µg/dL as the criterion for LDDST). The Receiver operating characteristic (ROC) curve analysis showed a cortisol level of 1.4 ug/dL, as the initial NIH report suggested, as the optimal 15-min post-CRH cortisol cutoff during the Dex-CRH test. This resulted in a ROC area of 0.964 compared to the ROC areas of 0.957 and 0.941 when a cortisol cutoff of 1.8 µg/dL and 1.2 µg/dL were used, respectively.
There were four patients who were on medications that may alter the metabolism of Dex. Among those, three had CD and were taking fluoxetine, oxcarbazepine, and topiramate, respectively. All three were correctly identified by both LDDST and Dex-CRH (Dex level not available on those patients). One patient in NCD group was on ritonavir and had normal LDDST and Dex-CRH (Dex level 1000 ng/dL).
Two patients in the CD group had CKD stage 3 or more and were correctly identified by both LDDST and Dex-CRH tests. One patient in the NCD group had mildly elevated liver enzymes in whom both LDDST and Dex-CRH tests were normal.

Discussion
Differentiating Cushing disease (CD) from non-neoplastic physiologic hypercortisolism (formerly known as pseudo-Cushing syndrome) remains a challenge [2]. The endocrine society guidelines in 2008 included the 2-day LDDST along with 24 h UFC, late-night salivary cortisol, and 1 mg DST as one of the initial tests to evaluate patients suspected to have CS [1]. No single test was recommended over the others except for special situations [1,14]. The same guidelines commented that the addition of CRH to 2-day DST may be considered in patients with equivocal results. A Cushing consensus guideline recommends using the Dex-CRH test in patients suspected of having non-neoplastic hypercortisolism [14].
The original 2-day LDDST was first described by Liddle [3] in 1960, where dexamethasone 0.5 mg is administered every six hours for eight doses, then measurement of 24 h urinary 17-hydroxycorticosteroid is obtained. However, the use of serum cortisol is simpler and provides a higher diagnostic accuracy [6,7]. A serum cortisol cutoff ranging from 1.4-1.8 ug/dL has been recommended as the pass criteria for this test, providing a diagnosis sensitivity of 90-95% [4,8,15]. The Dex-CRH test was first described by Yanovski et al. [8] in 1993, where he followed the original 2-day LDDST with CRH stimulation, and it showed 100% specificity, sensitivity, and diagnostic accuracy to differentiate CD from non-neoplastic physiologic hypercortisolism. The utility of the Dex-CRH test in this setting is based on the assumption that only patients with ACTH-dependent Cushing's syndrome will show a cortisol response to CRH after dexamethasone intake. The same group later revealed that the Dex-CRH test also differentiated patients with mild CD from healthy volunteers using the same criteria [16]. Subsequently, multiple other studies debated the diagnostic accuracy of the Dex-CRH test and showed varying degrees of specificity and sensitivity [9][10][11][12][13]. However, it is important to note that, in these studies, the authors used protocols for the Dex-CRH test that were different than the original description by Yanovski et al. [8]. The differences included timing of Dex administration starting at 09:00 instead of noon [9], using different cutoff levels for cortisol as a criterion 1.8 µg/dL [9,11,13] or 2.5 µg/dL [10] instead of 1.4 µg/dL, using human instead of ovine CRH [9,13] or using ovine CRH but at a fixed dose of 100 µg instead of 1 µg/Kg with max dose 100 µg [11,12]. These variations may have contributed to the differences in results observed [14]. In addition, some of the differences may be related to the cortisol assay used [9][10][11][12][13]. Furthermore, none of the above mentioned studies had dexamethasone measurement.
Our study protocol for Dex-CRH was similar to the initial description by Yanovski et al. [8]. Dex 0.5 mg was administered every six hours with last dose given at 6 AM, then serum cortisol was measured, and ovine CRH (1 µg/kg, max dose 100 µg) administered at 8 AM (2 h after the last dose of Dex) and we used the same serum cortisol cutoff of >1.4 µg/dL as a criterion in our study. Our study showed the following test performance for LDDST: sensitivity 92%, specificity 93%, and diagnostic accuracy 93%. The test performance was improved after adding CRH to the LDDST: sensitivity 100%, specificity 93%, and diagnostic accuracy 97%.
In our study, among 65 patients with Cushing disease (CD), 5 patients (7.7%) had a suppressed cortisol level <1.4 µg/dL after the LDDST but were appropriately classified as Cushing disease with a cortisol level >1.4 µg/dL at 15-min post Dex-CRH test. In contrast, 3/42 patients (7.1%) in the group where Cushing could not be confirmed (NCD) had an abnormal Dex-CRH test. In only one of these three patients, the LDDST was borderline normal (cortisol post-DEX was 1.4 µg/dL and increased to 3.1 µg/dL 15-min post CRH). All patients with available Dex level (n = 74) had levels >167 ng/dL during the Dex-CRH test (range 167-1200 ng/dL). There were four patients who had dex level <220 ng/dL which was suggested by the Endocrine Society [1]. One patient had a cortisol level of 12.9 and 16.1 µg/dL after LDDST and 15-min post Dex-CRH, respectively, and had pathology-proven CD. There were three other patients with dex levels <220 ng/dL. All had low cortisol levels <1.4 µg/dL after both LDDST and Dex-CRH tests.
When patients without Dex level were excluded from the analysis (n = 74), the sensitivity did not change, but the specificity and diagnostic accuracy of the Dex-CRH test further increased to 97 and 99%, respectively. This supports the value of measuring dexamethasone level during the Dex-CRH test. Others have shown an improvement in the diagnostic accuracy of dexamethasone suppression test [17], as false positive results can be seen in different settings related to dexamethasone such as rapid or malabsorption due to increased gut transit, chronic diarrhea or celiac disease, rapid metabolism due to concomitant use of CYP3A4 inducers (eg, phenobarbital, carbamazepine, St. John's wort) [14]. False positive results can also be seen due to increased cortisol binding globulin in the setting of estrogen use, pregnancy, or hepatitis which can increase total cortisol concentrations [14,18].
Our study has limitations inherent to its retrospective design and a lack of dexamethasone measurement in all patients. No standard protocol was used in the diagnostic workup of the patients suspected to have Cushing syndrome and their follow-up. We excluded two patients with cyclic CD. However, It is possible that cyclic Cushing syndrome may have contributed to some of the discrepancies among the tests. One of the two patients who underwent bilateral adrenalectomy after unsuccessful pituitary surgeries did not have ACTH staining on pathology. Accordingly, we can not rule out the possibility of a small ectopic neuroendocrine tumor as the etiology of ACTH-dependent CS in this case. The study's strength includes evaluating the Dex-CRH test in more than 100 patients in a referral center, the long follow-up of at least 14 months in patients in whom Cushing disease could not be confirmed, and using the same protocol that was initially described by the NIH.

Conclusion
Our study suggests that adding CRH stimulation to the 2-day LDDST has added value and provided additional case detection in 7.6% of our cohort. However, it caused one extra (2.4%) false-positive case compared to a negative LDDST. The latter marginal advantage of LDDST would be lost if a cortisol cutoff of <1.8 µg/dL is used as the pass criteria. The Dex-CRH test performed better when dexamethasone was measured with excellent discrimination among patients with Cushing disease and those in whom Cushing disease could not be confirmed.