Cushing syndrome (CS) is a rare disorder characterized by excessive cortisol secretion from the adrenal glands [1]. This condition occurs in two distinct forms: adrenocorticotropic hormone (ACTH)-dependent CS and ACTH-independent CS, which are traditionally identified using a plasma ACTH assay. Low levels of plasma ACTH are indicative of ACTH-independent CS because excess autonomously released cortisol suppresses pituitary ACTH secretion [23]. In many cases, however, the distinction between disease subtypes remains unclear. Previous studies have reported high interassay variability and low reproducibility of plasma ACTH assays [45]. In one study, only 85% of all measurements were interpreted correctly, and this rate dropped to 60% in patients with suspected ACTH-independent CS [6]. Plasma ACTH values must be interpreted with caution because the wide variation in patient results often leads to difficulty in differentiating between ACTH-dependent and ACTH-independent CS.

The high-dose dexamethasone suppression test (HDST) has been a mainstay of biochemical testing for distinguishing Cushing' disease (CD) from ACTH-producing ectopic CS [7]. However, this test is rarely performed to distinguish ACTH-dependent CS from adrenal CS. Moreover, the sensitivity and specificity of the HDST for diagnosing CD are 65% to 100% and 60% to 100%, respectively [891011].

Serum dehydroepiandrosterone sulfate (DHEA-S) is the most abundant adrenal androgen in blood. Serum DHEA-S levels generally remain within the high-normal range in patients with CD, but are low in patients with adrenal CS because of low plasma ACTH [121314]. However, the normal range of serum DHEA-S is wide, making serum DHEA-S a poor choice as a first-line test for the differential diagnosis of CS.

Here, we investigated the diagnostic utility of plasma ACTH levels for distinguishing between adrenal CS and CD. We also compared the discriminatory role of the plasma ACTH level with that of the HDST results and serum DHEA-S level.

Adrenal CS was present in 57 patients (62.0%) and CD in 35 patients (38.0%). No significant differences in age, sex, serum cortisol level, or 24-hour UFC level were observed between the two groups (Table 1). Compared with patients with adrenal CS, those with CD exhibited higher plasma ACTH (median, 22.2 pmol/L vs. 3.1 pmol/L; P<0.001) (Fig. 1A) and serum DHEA-S levels (median, 6,321 nmol/L vs. 677 nmol/L; P=0.001) (Fig. 1B); however, some overlap in the plasma ACTH levels was detected between patients with adrenal CS and CD in the range of 3.5 to 15.6 pmol/L (Fig. 1A). The median percentage suppression of serum cortisol after the HDST was 88% in patients with CD. Serum cortisol suppression of >50% after the HDST occurred in 26 (87%) of 30 patients with CD. The median percentage suppression of UFC after the HDST was 94% in patients with CD. UFC was suppressed by >90% in 22 patients (73%) with CD (Fig. 1C, D). Of the 45 patients with adrenal CS, serum cortisol was suppressed by >50% in only one patient, while UFC was not suppressed by >90% in any patients with adrenal CS (Fig. 1C, D). The percentage suppression of serum cortisol and UFC after the HDST was greater in patients with CD than in those with adrenal CS (P<0.001).

No significant differences in the basal serum cortisol or 24-hour UFC level were detected between patients with a suppressed ACTH level of <2.2 nmol/L and those without plasma ACTH suppression (Table 2). The serum cortisol and UFC levels after the HDST and the percentage suppression after the HDST did not differ between the two groups.

ROC curve analyses were used to assess the diagnostic power of the plasma ACTH level, serum DHEA-S level, and percentage suppression of serum cortisol as well as the UFC level after the HDST in differentiating adrenal CS from CD (Table 3). The area under the curve was highest for UFC after HDST (Fig. 2). The sensitivity for plasma ACTH levels of <2.2 pmol/L was only 33.3%, although the specificity was 100.0%. For the HDST, the sensitivity and specificity of the percentage suppression of serum cortisol at 50% was 97.6% and 86.2%, respectively, and those of suppression of UFC at 90% were 100.0% and 73.4%, respectively (data not shown). According to the Youden index, the optimal cut-off values of plasma ACTH and the percentage suppression of serum cortisol and UFC after the HDST were 5.3 pmol/L, 33.3%, and 61.6%, respectively. Using these cut-off points, the sensitivity and specificity of the HDST were higher than those of plasma ACTH, and the sensitivity and specificity of DHEA-S were lower than those of plasma ACTH. Suppression of UFC after the HDST was the most powerful diagnostic tool for differentiating between adrenal CS and CD among the tests examined.

Of the three patients with CD plus adrenocortical adenoma, one patient exhibited no plasma ACTH suppression, while the other two patients showed plasma ACTH suppression to <2.2 pmol/L. However, the percentage suppression of serum cortisol and UFC were >50% and 90%, respectively, for all patients (Table 4). Of the 57 patients with adrenal CS, nine had pituitary numtumors; of these, seven exhibited plasma ACTH levels of >2.2 pmol/L. No serum cortisol or UFC suppression after the HDST was observed in any of these patients (Table 5).

We also analyzed the diagnostic value of the HDST in patients with CD with pituitary macroadenoma. In this study, only three of the 35 patients with CD also had pituitary macroadenomas. The HDST was performed in only two of these patients; both serum cortisol and UFC were suppressed in all patients. Although the percentage of patients with pituitary macroadenoma was <10%, the HDST was not helpful in patients with pituitary macroadenoma.

We investigated the utility of plasma ACTH levels in the differential diagnosis of CS compared with that of the HDST results and serum DHEA-S levels. Using suppressed plasma ACTH level (<2.2 pmol/L) as the discriminatory parameter, our data show that a substantial number of patients with adrenal CS may remain unidentified. The HDST may therefore be a useful additional for the differential diagnosis of CS in patients with plasma ACTH levels of 3.5 to 15.6 pmol/L.

The plasma ACTH assay is the mainstay in the differential diagnosis of CS [1517]. Its convenience and low cost make it a viable choice for both hospitals and outpatient clinics. Plasma ACTH concentrations are detectable in 60% of patients with adrenal CS, and within the normal range in the remaining patients [18]. Similarly, 33.3% of patients with adrenal CS in our study had plasma ACTH concentrations of <2.2 pmol/L. Accordingly, plasma ACTH was of limited value for the differential diagnosis of CS.

Low interassay reproducibility and high variability of plasma ACTH levels may contribute to the poor diagnostic performance of this assay. Although plasma ACTH levels had a lower diagnostic accuracy than did the HDST in detecting adrenal CS, the sensitivity of plasma ACTH was 100% in patients with plasma ACTH levels of <2.2 pmol/L. Thus, further confirmation of adrenal CS may be unnecessary in patients with plasma ACTH levels below the limit of detection. However, in patients with plasma ACTH levels of 3.5 to 15.6 pmol/L, it is often difficult to definitely diagnose adrenal CS using this assay alone. Among patients with plasma ACTH levels within this range, HDST was able to successfully identify all patients with adrenal CS.

Previous studies have reported poor diagnostic accuracy of the HDST in discriminating adrenal CS. However, the positive and negative predictive values of the serum cortisol and UFC levels after the HDST in our patients with adrenal CS were approximately 95%, significantly higher than that of the plasma ACTH level. Neither the serum cortisol nor UFC level was found to be suppressed after the HDST in any patient with adrenal CS. Moreover, when the plasma ACTH level was within the range of 3.5 to 15.6 pmol/L, the specificity of the percentage suppression of serum cortisol and UFC was 100%. These results suggest that the utility of the HDST as a diagnostic tool may have been underestimated. Furthermore, the HDST was helpful in identifying the source of hormone secretion when both pituitary and adrenal nodules were detected by radiologic studies. In our study, 14% of patients with adrenal CS had pituitary adenomas, and 31% of patients with CD had adrenocortical adenomas. These observations demonstrate a clear ability of the HDST to distinguish between adrenal CS and CD in these situations, whereas the plasma ACTH level alone had little discrimination value.

Unlike the HDST, measurement of the serum DHEA-S level did not improve the ability to diagnose adrenal CS when added to plasma ACTH assays. Because serum DHEA-S is directly regulated by plasma ACTH, it was expected to be lower in patients with adrenal CS than in those with CD. However, in this study, the serum DHEA-S levels demonstrated a wide range of overlap between patients with adrenal CS and those with CD. The poor diagnostic value of the serum DHEA-S level may reflect the complex mechanisms of adrenal androgen secretion [12].

Our study included only 35 of 92 patients (38%) with CD compared with 57 patients (62%) with adrenal CS. Such a distribution is unusual given the normal distribution of cases, with CD typically more common than adrenal CS [19]. However, this finding is not without precedent; similar results have also been reported in a Korean population [20]. In that study, the prevalence of adrenal CS was slightly higher than that of CD, and ectopic CS was very rare. This distribution is thought to have been due to the low detection rate of pituitary tumors and ectopic CS in Korea, although racial differences may also be involved. The unusual distribution of patients our study population may have influenced the outcomes of some of our statistical tests; different cutoff values may be possible in a larger or more normally distributed cohort. Similarly, it is conceivable that the high prevalence of adrenal causes could inflate the positive and negative predictive values of the HDST.

As with all studies, the analyses presented here are not without limitations. First, the biochemical data were retrospectively retrieved from patients' medical records, meaning that the number of biochemical tests for each patient was not equal. Second, plasma ACTH levels may deteriorate in response to handling error after blood collection. Under ideal conditions, blood samples should be collected in a prechilled tube containing EDTA, immediately separated at 4℃, and then stored at -80℃ until needed. Even with modern ACTH assays, ACTH levels may deteriorate quickly in whole blood unless cooled and separated and the plasma quickly frozen until needed. It is therefore possible that ACTH levels in patients with CD may have fallen after the blood samples were collected despite the fact that the study was performed with the same protocol and assay method at each institution. Third, simultaneous pituitary and adrenocortical tumors were identified in some patients, as above. These findings include both CD plus an adrenal incidentaloma as well as adrenal CS plus a pituitary incidentaloma. Last, the corticotropin-releasing hormone (CRH) stimulation test was not performed in our study because of limited reagent availability in Korea. It would be of interest to examine whether stimulated ACTH levels in response to CRH better discriminate between adrenal CS and CD.

In conclusion, our study has demonstrated that the plasma ACTH assay has limited value in the differential diagnosis of CS. The diagnostic accuracy of the HDST is higher than that of the plasma ACTH level, whereas the serum DHEA-S level is less useful than the plasma ACTH level. The HDST is particularly beneficial in identifying patients with adrenal CS when the plasma ACTH level is not suppressed to <2.2 pmol/L. We therefore recommend using the HDST for the differential diagnosis of CS when plasma ACTH levels alone are inconclusive.