Three calculations of free cortisol versus measured values in the critically ill

To investigate the agreement between the calculated free cortisol levels according to widely applied Coolens and adjusted Södergård equations with measured levels in the critically ill.

A prospective study in a mixed intensive care unit. We consecutively included 103 patients with treatment-insensitive hypotension in whom an adrenocorticotropic hormone (ACTH) test (250μg) was performed. Serum total and free cortisol (equilibrium dialysis), corticosteroid-binding globulin and albumin were assessed. Free cortisol was estimated by the Coolens method (C) and two adjusted Södergård (S1 and S2) equations. Bland Altman plots were made.

The bias for absolute (t=0, 30 and 60min after ACTH injection) cortisol levels was 38, -24, 41nmol/L when the C, S1 and S2 equations were used, with 95% limits of agreement between -65-142, -182-135, and -57-139nmol/L and percentage errors of 66, 85, and 64%, respectively. Bias for delta (peak-baseline) cortisol was 14, -31 and 16nmol/L, with 95% limits of agreement between -80-108, -157-95, and -74-105nmol/L, and percentage errors of 107, 114, and 100% for C, S1 and S2 equations, respectively.

Calculated free cortisol levels have too high bias and imprecision to allow for acceptable use in the critically ill.