Adrenal Insufficiency Common in Surgical ICU Patients

Abstract & Commentary

Synopsis: Adrenal insufficiency, as diagnosed by an abnormal response to ACTH stimulation, occurred in 34% of the patients older than 55 who required vasopressors following complicated trauma or surgery. Those patients who received therapeutic steroid replacement experienced a lower mortality than those who did not (21% vs 45%).

Source: Rivers EP, et al. Adrenal insufficiency in high-risk surgical ICU patients. Chest. 2001(3);119:889-896.

Occult adrenal insufficiency (ai) has been identified in up to 28% of seriously ill medical patients. The occurrence of AI in other patient populations has not been quantitated, and the implications for patient outcome in this disorder are not clear. Rivers and associates sought to identify the patterns of adrenal function in critically ill surgical patients. They prospectively studied patients following surgery who were older than 55 years and dependent on vasopressors during the first 24 hours of ICU treatment. Patients were considered vasopressor dependent if, after a carefully controlled, pulmonary artery assisted-volume loading procedure, they needed a vasoactive substance to maintain a mean arterial blood pressure of 65-70 mm Hg. Patients with known adrenal dysfunction, those who had receive intra-operative steroids, those who had received etomidate (which is known to inhibit steroid synthesis), or those who had received therapeutic steroids in the previous 3 months, were eliminated.

A total of 104 patients were studied over a 2-year period. Each patient received a 1-hour cosyntropin (ACTH) stimulation test. Immediately following determination of baseline serum cortisol level, 0.25 mg of ACTH was administered, and cortisol was measured again at 30 minutes and 1 hour. Normal results would be all cortisol levels greater than 20 mg/mL and a rise of a greater than 50% over baseline (to at least 30 mg/mL), or an absolute rise of more than 9 mg/mL. Two patterns of AI were defined: patients who had any abnormal cortisol level (< 20 mg/mL on any measurement) and experienced less than a 10 mg/mL rise following stimulation, and those with any value < 30 mg/mL or an inadequate response to stimulus. The 43% of patients entered into the study had one or the other form of AI; 63% of patients were functionally normal.

There was no difference between normals and patients with AI with respect to age, sex, body temperature, leukocyte count, serum electrolytes or glucose, or hemodynamic variables. However, the relative percentage and absolute number of blood eosinophils was higher in the patients with AI (1.5% vs 3%). A total of 42% of all the patients received therapeutic steroids (100 mg hydrocortisone every 8 hours for 3 doses), and those with AI who received treatment were more likely to be weaned from vasopressors within 24 hours (19 out of 23 who received steroids, or 83%; vs 3 out of 11 who did not, or 27%). Normal patients receiving steroids also appeared to wean more quickly, but this effect did not reach statistical significance (10 out of 23 who received steroids, 43%; vs 16 out of 47 who did not, 34%).

Mortality was higher in patients without AI (46%) than in those with AI (29%). However, mortality was higher in those patients with AI who did not receive steroids (45%) than those who did (21%).


While this study has several major problems, it confirms the high incidence of AI in the very ill surgical patients. The study included a prospective collection of data, but no treatment arm was included. Patients were treated or not at the choice of the care team. This resulted in some normals being treated with steroids and some patients with demonstrated AI not being treated. The effect on outcome from treatment with steroids cannot be inferred from this study. How the decision to treat was made has not been evaluated and was not randomized. However, treatment appeared to have a positive effect on weaning from vasopressors and maybe improved survival. One of the effects of steroids is to increase the number of adrenergic receptors. In stressed patients on vasopressors, it is not surprising that pharmacological doses of steroids would improve vasopressor responsiveness.

The effect on survival is more problematic. Those patients with AI, treated with steroids, actually did better than the group as a whole, while those with AI not treated did the same as the group as a whole. This suggests (but does not prove, due to the study design) that the patients with AI were actually less sick than the others. This could be true since the effects of untreated AI would be to need more vasopressors to achieve normal blood pressure. Since hypotension was a study criterion, this might mean that less ill patients with AI would be selected. The fact that untreated patients with AI did no worse than the group as a whole gives little comfort, as these patients might have done better with treatment of their AI.

A major problem with this study is that determination of adrenal function was only performed at one point in time. It is possible that if repeated measurements were performed, many other patients would have demonstrated AI at some point in their illness. Also, some of the initially identified patients with AI might have proved normal on subsequent study. The "normal" variability of adrenal function during critical illness has not been well studied. The duration of treatment needed for AI could be more appropriately determined with such studies; only 24 hours of steroid treatment was used in this group of patients.

The implication for empiric steroid treatment, although not proven, is certainly suggested by this and other studies. With the high percentage of AI demonstrated in vasopressor-dependent patients and the lack of significant side effects of treatment, consideration must be given to empiric steroid replacement in these very ill (> 40% mortality) patients.