Validated Criteria for the Interpretation of a Single Measurement of Serum Cortisol in the Investigation of Suspected Adrenal Insufficiency

Scott D. Mackenzie; Robert M. Gifford; Luke D. Boyle; Mike S. Crane; Mark W. J. Strachan; Fraser W. Gibb

Disclosures

Clin Endocrinol. 2019;91(5):608-615. 

In This Article

Results

Characteristics of the cohorts are summarised in Table 1. Supplementary data include details on clinical indications (Table S1) and basal serum cortisol vs time (Figure S1). Serum cortisol demonstrated an inverse correlation with time across all cohorts (derivation cohort r 2 = .131, P < .001; outpatient validation cohort r 2 = .038, P < .001; inpatient validation cohort r 2 = .014, P = .001; Figure S1).

Supplementary Figure 1.

Basal serum cortisol vs. time. Blue circle = intact adrenal function (serum cortisol ≥430 nmol/L 30 minutes after Synacthen 250μg IM), red circle = adrenal insufficiency (AI; stimulated serum cortisol <430 nmol/L). Tests performed in outpatients (a) and (b) and inpatients (c) Regression lines refer to serum cortisol at baseline vs. time of sample for individuals without AI.

Derivation Cohort

To achieve 95% sensitivity for the prediction of subnormal-stimulated cortisol, a threshold of <275 nmol/L was selected for morning cortisol and <250 nmol/L for afternoon cortisol (Figures 1 and 2). These achieved sensitivities of 96.2 and 96.1%, respectively (P > .99 for difference; Table 2).

Figure 1.

Baseline vs post synacthen serum cortisol. Serum cortisol at baseline vs 30 min post synacthen 250 μg IM in: outpatient derivation cohort (A) (B); outpatient validation cohort (C) (D) and inpatient cohort (E) (F). Baseline samples taken 06:00–12:00 (A) (C) (E) or 12:00–18:00 (B) (D) (F). Horizontal solid line refers to cut-off for post synacthen serum cortisol (430 nmol/L) that defines a normal response in the short synacthen test. Vertical solid lines refer to cut-offs for basal serum cortisol which detect a subnormal-stimulated cortisol with sensitivity > 95% in morning (<275 nmol/L) and afternoon (<250 nmol/L) samples

Figure 2.

Performance of baseline serum cortisol as predictor of a subnormal-stimulated cortisol in all outpatients (A) and inpatients (B). Receiver operating characteristics and baseline cortisol vs sensitivity and specificity for the detection of a subnormal-stimulated serum cortisol (<430 nmol/L at 30 min following synacthen 250 µg IM). Solid line = baseline samples taken 06:00–12:00; Dashed line = baseline samples taken 12:00–18:00

The criteria provided a NPV 98.4% and 98.9% in morning vs afternoon samples, respectively (P = .74 for difference). Overall performance of the test, as assessed by AUC (ROC) was not significantly different in morning vs afternoon samples, although specificity was significantly poorer in the afternoon (am 55.1% vs pm 37.7%, P < .0001), accompanied by a poorer positive predictive value (am 37.7% vs pm 14.0%, P < .0001).

Baseline serum cortisol was <50 nmol/L in 37 samples. Of these, 36 (95%) had a stimulated serum cortisol of <430 nmol/L.

If it were assumed that confirmatory SSTs were only required for individuals whose baseline serum cortisol lay between the upper and lower cut-offs, the criteria would result in a reduction in the need for a SST in 48.2% of individuals tested in the morning and 36.2% in the afternoon.

3.2 Validation Cohorts

Outpatients. Application of the cut-offs identified in the derivation cohort to the outpatient validation cohort retained sensitivities of >95% for both morning (98.6%) and afternoon (100%) samples (Table 2). Sensitivities, negative predictive values and AUC (ROC) were not significantly different in comparison with the relevant time periods in the derivation cohort. For afternoon samples, specificity was greater in comparison with afternoon samples in the derivation cohort (48.9% vs 37.7%, respectively, P = .0001).

All 38 individuals with a baseline serum cortisol <50 nmol/L had a stimulated serum cortisol of <430 nmol/L. The proposed the criteria would result in a reduction in the need for a SST in 52.2% of individuals tested in the morning and 39.7% in the afternoon.

Inpatients. For inpatients, a sensitivity of 95% was achieved for morning samples (97.4%) but not afternoon samples (89.2%), and therefore, afternoon baseline serum cortisol in inpatients did not meet the primary criterion for diagnostic utility.

For morning samples, sensitivity, negative predictive values and AUC (ROC) were not significantly different in comparison with the derivation cohort. Specificity was greater in comparison with morning samples in the derivation cohort (80.0% vs 37.7% respectively, P < .0001). Baseline serum cortisol was <50 nmol/L in 7 individuals (1 of whom passed the SST; see 'false positives'). If it was assumed that confirmatory SSTs were only required for individuals whose baseline serum cortisol lay between the upper and lower cut-offs, the criteria would result in a reduction in the need for a SST in 73.8% of inpatients tested in the morning.

False Negatives

Across all 3301 tests, use of the baseline serum cortisol for the detection of a subnormal-stimulated cortisol would have resulted in a total of 16 false negatives, where stimulated cortisol was subnormal despite a baseline serum cortisol above the proposed thresholds. Excluding those inpatients tested in the afternoon, where sensitivity did not meet the prespecified threshold of 95%, the overall false negative rate was 0.4% (12 patients).

Characteristics of these individuals are detailed in Table S2. Follow-up data are available for all but one patient and upon repeat testing the majority were subsequently found to have a normal SST. In one, the response was borderline, with a stimulated cortisol 430 nmol/L, and it was felt likely that this patient had Addison's disease on the basis of the clinical background and elevated adrenal antibodies. Only one individual failed the Synacthen test upon repeat testing, and one individual was lost to follow-up having commenced glucocorticoid replacement.

False Positives

A basal serum cortisol below 50 nmol/L predicted a subnormal-stimulated cortisol with 98% certainty. Of 88 individuals whose basal cortisol lay below this threshold, 2 were found to have a stimulated cortisol greater than 430 nmol/L and would be designated as 'false positives' under the proposed criteria: one with opioid-induced hypogonadism and one with suspected hypophysitis (see Table S3).

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....