Share this article on:

Myths in Emergency Medicine: Inconsistencies in Hydrocortisone Trial for Septic Shock Make Findings Suspect

Spiegel, Rory, MD

doi: 10.1097/01.EEM.0000533727.15436.ba
Myths in Emergency Medicine

Dr. Spiegel is a clinical instructor in emergency medicine at the University of Maryland Medical Center. Visit his blog at http://emnerd.com, follow him on Twitter @emnerd_, and read his past articles at http://bit.ly/EMN-MythsinEM.

Figure

Figure

We have existed in a state of ambiguity, torn between two contradictory randomized, controlled trials, when it comes to glucocorticoid therapy for septic shock. The Annane trial suggested a mortality benefit in giving corticosteroids to patients found to have relative adrenal insufficiency. (JAMA 2002;288[7]:862.) The larger CORTICUS study, however, found no benefit in using corticosteroids compared with placebo. (N Engl J Med 2008;358[2]:111.)

Those in favor of administering corticosteroids have argued that patients in the Annane, et al., trial were much sicker, isolating the subset of patients who truly benefit from steroid supplementation. They contended that the negative results in the CORTICUS trial were the consequence of enrolling a much healthier population, washing out any chance of identifying the underlying signal. The contrary opinion is simple: Those against corticosterioids say the results of the Annane trial were due to statistical noise and a small sample size and CORTICUS was nothing more than the expected regression to the mean.

Now there is a third study that could liberate us from this limbo—the ADRENAL trial, a pragmatic, double-blind, parallel-group RCT comparing hydrocortisone with placebo in patients with septic shock. This trial's impressive sample size and methodological rigor overshadow those of the Annane and CORTICUS cohorts, but I fear its results will deliver far less certitude than we had hoped.

Venkatesh, et al., enrolled adults 18 and older with septic shock who had two or more SIRS criteria, who had been on vasoactive agents for at least four hours, and who required mechanical ventilation. (N Engl J Med 2018;378[9)]:797.) Patients were randomized to receive an intravenous infusion of hydrocortisone at a dose of 200 mg per day or matching placebo for seven days. The authors enrolled 3,800 patients in 64 ICUs over four years and across five countries, and 1,898 were assigned to hydrocortisone and 1,902 to placebo.

The authors found no difference in the primary outcome—90-day mortality—between the hydrocortisone and placebo groups (27.9% v. 28.8%; CI 0.82 to 1.10; p=0.5). Nor did they find a difference in 28-day mortality (22.3% v. 24.3%; CI 0.76 to 1.03; p=0.13). The authors noted that patients in the hydrocortisone group had a faster time to resolution of shock (three v. four days), a shorter time to discharge from the ICU (10 v. 12 days), and a shorter duration of initial mechanical ventilation (six v. seven days). These results were statistically significant, but it is important to remember that these are secondary endpoints and are acutely vulnerable to the whimsies of statistical chance.

Time to ICU discharge was statistically shorter in patients randomized to receive hydrocortisone, for example, but the authors found no difference in the number of days alive and outside the ICU or days alive and outside the hospital. And duration of initial mechanical ventilation was shorter in the hydrocortisone group, but no difference was seen in days alive and free of mechanical ventilation nor in the rate of recurrent mechanical ventilation. Such inconsistencies make one wonder if these findings were simply due to multiple measurements and statistical chance.

Back to Top | Article Outline

A Nominal Effect

Even the improvement in time to resolution of shock seems to be of questionable clinical relevance. It did not translate into a mortality benefit, the mean difference in mean arterial pressure between the two groups was only 5.39 mm Hg, and no difference in the daily dose of norepinephrine existed. The authors also observed a 4.7 percent absolute increase in the need for blood transfusion in patients in the placebo group (37% v. 41.7%). Given the secondary nature of this finding and the unclear mechanistic underpinning, it is hypothesis at best.

The authors also noted a statistically significant increase in the rate of adverse events in the hydrocortisone group (1.1% v. 0.3%, p=0.009). Most of these were clinically inconsequential (hyperglycemia, hypernatremia, etc.), but it is important to remember that RCTs are notoriously poor at identifying rare harms and will underestimate the harm of the treatment in question.

I am sure some will find minute chinks in the scientific armor of the ADRENAL trial to discount its findings despite its methodological rigor. Most notably, no minimum dose of vasopressors was required before enrollment. The argument is, like the CORTICUS trial, this cohort represents a group of patients that is unlikely to benefit from hydrocortisone supplementation. But when the authors examined patients with elevated pressor requirements (both >15 mcg/min and >25 mcg/min) and elevated APACHE scores (>25) who met the Sepsis-3 criteria for septic shock, they were unable to identify a signal of benefit in any of these sicker subsets of patients, all of whom represent a far larger sample size than the Annane cohort.

It is difficult to view these results as anything but negative. Some potential signals of benefit are seen in patients who received hydrocortisone, but they were fleeting and ultimately had minimal influence on patient-centered outcomes. Any true benefit, if it exists, is small and would require an impossibly large trial to demonstrate empirically.

In the end, the ADRENAL trial will have a nominal effect on practice. The desire to act in the face of critical illness is strong enough that we will reason away this lack of benefit simply because we have so little else to offer. Many of us faced with a patient in refractory shock will continue to give corticosteroids in the hope that they may help the individual patient before us. We will just do so with an added degree of hand-wringing, knowing that we have read the literature refuting our practice and simply want nothing to do with its conclusions.

Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved.