Context. Severe systemic infection leads to hypercortisolism. Reduced cortisol binding proteins may accentuate the free cortisol elevations seen in systemic infection. Recently, low total cortisol increments after tetracosactrin have been associated with increased mortality and hemodynamic responsiveness to exogenous hydrocortisone in septic shock, a phenomenon termed by some investigators as "relative adrenal insufficiency".

Hypothesis. Free plasma cortisol may correspond more closely to illness severity than total cortisol, comparing septic shock and sepsis.

Design. Prospective study

Setting. Tertiary teaching hospital

Patients. Patients with septic shock (SS, n = 45), sepsis (S, n = 19) and controls (HC, n = 10)

Aims. To compare total with free cortisol, measured directly and estimated by Coolens' method, corticosteroid-binding globulin (CBG) and albumin in patients with septic shock (with and without RAI) and sepsis during acute illness, recovery and convalescence.

Results. Comparing SS, S and HC subjects, free cortisol levels reflected illness severity more closely than total cortisol (basal free cortisol: SS 186 vs. S 29 vs. HC 13 nmol/L, P<0.001 compared with basal total cortisol: SS 880 vs. S 417 vs. HC 352 nmol/L, P < 0.001). Stimulated free cortisol increments varied greatly with illness category; SS 192 vs. S 115 vs. HC 59 nmol/L, P = 0.004, whereas total cortisol increments did not; SS 474 vs. S 576 vs. HC 524 nmol/L, P = 0.013. The lack of increase in total cortisol with illness severity is due to lower CBG and albumin. One-third of patients with septic shock (15/45), but no sepsis patients met a recently described criterion for RAI (total cortisol increment after tetracosactrin 248 nmol/L). RAI patients had higher basal total cortisol (1157 vs. 756 nmol/L, P = 0.028) and basal free cortisol (287 vs. 140 nmol/L, P = 0.017) than non-RAI patients. Mean cortisol increments in RAI were lower (total: 99 vs. 648 nmol/L, P < 0.001, free: 59 vs. 252 nmol/L, P < 0.001). These differences were not due to altered CBG or albumin levels. Free cortisol levels normalized more promptly than total cortisol in convalescence. Calculated free cortisol by Coolens' method compared closely to measured free cortisol.

Conclusions. Free cortisol is likely to be a better guide to cortisolemia in systemic infection as it corresponds more closely to illness severity. The attenuated cortisol increment after tetracosactrin in RAI is not due to low cortisol-binding proteins. Free cortisol levels can be determined reliably using total cortisol and CBG levels.