DIAGNOSTIC ACCURACY OF BASAL CORTISOL LEVEL TO PREDICT ADRENAL INSUFFICIENCY IN COSYNTROPIN TESTING: RESULTS FROM AN OBSERVATIONAL COHORT STUDY WITH 804 PATIENTS.

OBJECTIVE

Adrenocorticotrophic hormone (ACTH) stimulation testing is the current standard for assessing primary and secondary adrenal insufficiency (AI). We aimed to investigate the value of basal cortisol level for prediction of AI.

METHODS

We retrospectively analyzed 804 consecutive patients who had high-dose (250 μg, HDT) or low-dose (1 μg, LDT) ACTH testing as part of their diagnostic work-up. Site-specific cut-off levels for AI were <550 in and <500 nmol/L in HDT and LDT, respectively.

RESULTS

Overall, 70/400 (17.5%) in the LDT group and 118/404 (29.2%) in the HDT group showed an insufficient increase of cortisol and were categorized as AI. The receiver operating characteristic curve analysis showed an overall area under the curve (AUC) for basal cortisol of 0.88, which was comparable in LDT (area under the curve [AUC] 0.88) and HDT (AUC 0.88). If basal cortisol levels were ≥450 nmol/L (n = 234/804, 29.1%), the negative predictive value to rule out AI was 98.7%. If cortisol was ≤100 nmol/L (n = 69/804, 8.6%) the positive predictive value was 93.2% to rule in AI. There was a minimal additional value of the 30-minute cortisol level in HDT as compared to the 60-minute result, as well as for delta values.

CONCLUSION

Basal cortisol levels ≤100 and ≥450 nmol/L were found in almost half of patients tested for possible AI and had high diagnostic accuracy, abolishing the need for formal ACTH testing. The 30-minute cortisol value in HDT did not increase diagnostic accuracy. These data may help guide clinicians when testing can safely be omitted, thereby reducing expenses and simplifying test protocols.

ABBREVIATIONS

ACTH = adrenocorticotropic hormone AI = adrenal insufficiency AUC = area under the curve CI = confidence interval HDT = high-dose test HPA axis = hypothalamus-pituitary-adrenal axis ITT = insulin tolerance test LDT = low-dose test NR = normal responders LR+/- = positive/negative likelihood ratio NPV = negative predictive value PPV = positive predictive value.