Endocrine Abstracts

Objective: During the current pandemic of COVID-19, many therapeutic protocols adopted high dose systemic glucocorticoids (GC) for treatment of moderate to severe respiratory insufficiency. The suppression of the hypothalamic–pituitary–adrenal axis by synthetic GC, even after a short treatment period, cannot be reliably predicted due to pathophysiological changes in cortisol dynamics in critically ill, inter-individual pharmacokinetic differences, and sensitivity variations in glucocorticoid receptors. Moreover, the revealed pathological adrenal changes in Covid-19 patients make prior estimates even less reliable.

Aim: We aimed to estimate percentage of patients with adrenal insufficiency (AI) at the end of acute phase of COVID-19 after tapering GC toward replacement dose of hydrocortisone for at least 1 week and to investigate for the first time the basal cortisol level for prediction of AI in this population.

Methods: We retrospectively analysed 287 consecutive patients (median 65 years, range 26-91 years); 113 (39.4%) female and 174 (60.6%) male) who had low-dose (1 μg) cosyntropin testing. Site-specific cut-off levels for AI were <500 nmol/l. The overall potential of cortisol at baseline to classify patients into adrenal (in)sufficiency was assessed using receiver-operating-characteristic (ROC) curve analysis.

Results: The average cortisol level at baseline was 419 nmol/l (median 429, range 36-953 nmol/l), and at 30’ 617 nmol/l (median 623, range 114-1092 nmol/l); Overall, 65 (22.6%) showed an insufficient increase of cortisol and were categorized as AI. ROC showed an overall area under the curve (AUC) for basal cortisol of 0.84, with 95% confidence interval 0.79-0.89. If basal cortisol level was below 100 nmol/l, which was the case in 7 patients (2.4%), the positive predictive value (PPV) to predict AI was 100%. If basal cortisol was at least 450 nmol/l, which was the case in 125 patients (43.6%), the negative predictive value (NPV) to predict no AI was 96.0%; if it was at least 400 nmol/l (158 patients, 55.1%), NPV was 93.0%; and if it was at least 460 nmol/l (113 patients, 39.4%), NPV was 99.1%. PPV sharply declines after the threshold of 100 nmol/l. On the other hand, PPV increases approximately linearly over the threshold range from 350 to 460 nmol/l, where it practically reaches 100%.

Conclusion: Basal cortisol levels ≤100 and ≥460 nmol/l in patients tested for possible AI were found in 41.2% and had sufficient diagnostic accuracy to safely abolish the need for cosyntropin testing. The data may help guide clinicians when testing for AI can be simplified.