Plasma Steroid Profiles in Subclinical Compared With Overt Adrenal Cushing Syndrome

Jimmy Masjkur; Matthias Gruber; Mirko Peitzsch; Denise Kaden; Guido Di Dalmazi; Martin Bidlingmaier; Stephanie Zopp; Katharina Langton; Julia Fazel; Felix Beuschlein; Stefan Richard Bornstein; Martin Reincke; Graeme Eisenhofer


J Clin Endocrinol Metab. 2019;104(10):4331-4340. 

In This Article


Patient Characteristics and Routine Test Results

Patients with SC were older at presentation compared with the other groups, whereas there was a higher proportion of women in the AC group compared with the other groups (Table 1). Plasma cortisol concentrations after DST in patients with SC were higher, as expected, than in the EX group (P < 0.05) but lower compared with those of patients with AC (P < 0.005). Midnight SFC and UFC levels were higher in the AC and SC groups than in the EX group (P < 0.05). Basal plasma ACTH concentrations were lower in patients with SC than in the EX group (P < 0.01) but higher compared with patients with AC (P < 0.0005). There was minimal overlap of cortisol concentrations after DST between the AC or SC groups and the EX group, in keeping with use of this test as part of the gold standard for classification.

Steroid Profiles

Significances of differences among groups were assessed by models with multivariate analyses to correct for differences in sex and age (Table 2). Among the 15 steroids of the panel, 11-deoxycortisol and 11-deoxycorticosterone concentrations were consistently increased (P < 0.05) in both groups of patients with AC and SC compared with both the reference and EX groups (Fig. 1). Plasma concentrations of cortisol were increased (P < 0.05) in AC and SC groups above the reference but not the EX group. There were no differences in plasma aldosterone concentrations between groups, whereas 18-oxocortisol and 18-hydroxycortisol levels in patients with SC and AC and the EX group were higher (P < 0.05) than concentrations of the reference group.

Figure 1.

Results of steroids of the 15-steroid panel in patients with ACTH-independent Cushing syndrome (AC and SC) and in the EX group, compared with that of the reference population. Values of all steroids are shown as least square means corrected for age and sex and determined from exponents of logarithmically transformed data (i.e., geometric means) with similarly determined positive and negative SEs. *P < 0.05, different from reference; P < 0.05 different from EX; § P < 0.05, different from AC.

DHEA and DHEA-S were the only two steroids for which plasma concentrations were consistently lower (P < 0.05) in patients with AC and SC than both reference and EX groups, with additionally lower concentrations of both these two steroids in the AC group than in the SC group (Fig. 1). Androstenedione levels were lower in the AC group than in the EX and reference groups. Corticosterone levels were higher (P < 0.05) in patients with SC compared with the EX group. There were no differences in plasma concentrations of 21-deoxycortisol, cortisone, and 17-hydroxyprogesterone among the four groups, whereas there were unusually lower concentrations of pregnenolone (P < 0.05) in the EX and SC groups compared with the reference and AC groups. Progesterone concentrations were lower (P < 0.05) in the SC and AC groups compared with the reference group.

Diagnostic Test Performance

A combination of 14 steroids was established using a stepwise analysis to provide optimal diagnostic performance for distinguishing patients with and without SC. With the full profile, the analysis indicated six steroids (11-deoxycortisol, 11-deoxycorticosterone, DHEA-S, DHEA, androstenedione, and progesterone) that provided the highest discriminatory power with respective F-ratios of 41.7, 35.6, 34.8, 28.6, 9.4, and 8.7. With omission of steroids to obtain different steroid combinations, 11-deoxycortisol always maintained the highest rank. With discriminant analysis, this combination provided a misclassification rate of only 14.4%. With exclusion of 18-oxo-cortisol, the combination of the remaining 14 steroids provided a lower misclassification rate of 4.8%. Using this combination, areas under ROC curves (AUC-ROC) showed higher (P < 0.01) diagnostic performance of the steroid panel compared with SFC, basal ACTH, basal serum cortisol, and UFC (Fig. 2; Table 3). The AUC-ROC for the steroid panel did not differ from that of the DST.

Figure 2.

Results of model comparisons of the steroid profile and routine diagnostic tests. ROC curves are shown together in one panel for the five routine screening tests and a selection of 14 steroids of the steroid panel, which exhibited optimal discrimination of patients with ACTH-independent Cushing syndrome from the EX group. Prediction accuracy of each group was measured as AUC in the accompanying table. The groups are distinguished by different color lines.

Classification of Patients With SC, AC, and EX

Routine measurements of UFC, midnight SFC, basal plasma ACTH, and serum cortisol combined with the DST provided optimal discrimination of patients in the AC, EX, and SC groups with areas under ROC curves of 1.000, 0.9930, and 0.9941, respectively (Fig. 3). Most patients with SC (83%) and 100% of patients with AC were correctly classified by the combination of all five routine measurements. Four percent of patients in the EX group were incorrectly classified using the combination of routine tests.

Figure 3.

Results of (A–C) discriminant analyses for use of routine diagnostic tests compared with (D–F) 14 steroids of the steroid panel that provided optimal discrimination of the three patient groups. Triangle, AC group; diamond, EX group; circle, SC group. (B, E) Two-dimensional canonical plots; (A, D) AUC-ROC; (C, F) predicted vs actual groupings according to discriminant analyses. Routine diagnostic tests included SFC and UFC measurements, DST, and basal plasma cortisol and basal plasma ACTH measurements. Steroids for the selected steroid profile included 11-deoxycortisol, 11-deoxycorticosterone, DHEA, DHEA-S, androstenedione, aldosterone, cortisol, corticosterone, cortisone, 18-hydroxycortisol, 21-deoxycortisol, 17-hydroxyprogesterone, progesterone, and pregnenolone. For the analyses of routine clinical tests, complete results for all five tests were only available for 169 of the 208 patients.

With use of the steroid panel, 97% of patients with SC were correctly classified, whereas up to 5% of patients with AC and of those in the EX group were misclassified (Fig. 3). AUC-ROC of the steroid panel were 0.9962, 0.9904, and 0.9901 for the AC, EX, and SC groups, respectively, similar to those observed for the combination of routine tests (Fig. 3). Although both methods could completely distinguish patients with SC from those with AC, there were fewer false-negative identifications of patients among the SC group with use of the steroid profile than with the routine test combination.