Purpose: To study the value of free versus total cortisol levels in assessing relative adrenal insufficiency during critical illnessrelated corticosteroid insufficiency. Methods: A prospective study in a mixed intensive care unit from 2004 to 2007. We consecutively included 49 septic and 63 non-septic patients with treatment-insensitive hypotension in whom an adrenocorticotropic hormone (ACTH) test (250 lg) was performed. Serum total and free cortisol (equilibrium dialysis), corticosteroid-binding globulin (CBG) and albumin were assessed. Results: Although a low CBG resulted in a high free cortisol level relative to total cortisol, free and total cortisol and their increases were well correlated (r = 0.77-0.79, P \ 0.001). In sepsis, hypoalbuminemia did not affect total and free cortisol, and increases in total cortisol upon ACTH predicted increases in free cortisol regardless of low binding proteins. In non-sepsis, total cortisol was lower with than without hypoalbuminemia; free cortisol did not differ, since hypoalbuminemia concurred with a low CBG. Increases in total cortisol depended less on binding proteins than on raw levels. The areas under the receiver operating characteristic curve for predicting increases in free from total cortisol were 0.93-0.97 in sepsis and 0.79-0.85 in non-sepsis (P = 0.044 or lower for sepsis vs. non-sepsis). Conclusions: Although the biologically active free cortisol fraction depends on binding proteins, total cortisol correlates to free cortisol in treatment-insensitive hypotension during critical illness. In sepsis, albumin is not an important binding molecule. Subnormal increments in total cortisol upon ACTH suffice in assessing relative adrenal insufficiency, particularly in sepsis.