Raw and adjusted relative risks of developing adrenal insufficiency for etomidate exposure.
Raw and adjusted relative risks of developing adrenal insufficiency for etomidate exposure by propensity score.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Cotton BA, Guillamondegui OD, Fleming SB, et al. Increased Risk of Adrenal Insufficiency Following Etomidate Exposure in Critically Injured Patients. Arch Surg. 2008;143(1):62–67. doi:10.1001/archsurg.143.1.62
Timely diagnosis and treatment of adrenal insufficiency (AI) dramatically reduces mortality in trauma patients. We sought to identify risk factors and populations with a high risk of developing AI.
Retrospective registry study.
Academic level I trauma center.
All trauma patients in the intensive care unit who underwent cosyntropin stimulation testing (CST) for presumed AI from January 1, 2002, through December 31, 2004.
Cosyntropin stimulation testing, in which response was defined as an increase of 9 μg/dL (248 nmol/L) or more in cortisol level.
Main Outcome Measures
Risk factors for developing AI in critically ill trauma patients.
In 137 patients, CST was performed; 83 (60.6%) were nonresponders and 54 (39.4%) were responders. Age, sex, race, trauma mechanism, Injury Severity Score, and Revised Trauma Score were not statistically different between the groups. Rates of sepsis/septic shock, mechanical ventilation, and mortality were also similar between the 2 groups. However, rates of hemorrhagic shock on admission (45 [54%] vs 16 [30%]), requirement of vasopressor support (65 [78%] vs 28 [52%]), and etomidate exposure (59 [71%] vs 28 [52%]) were all significantly higher in the nonresponder group (P < .01). The increased risk of AI remained after controlling for potential confounding covariates (age, mechanism, Injury Severity Score, and Revised Trauma Score).
Exposure to etomidate is a modifiable risk factor for the development of AI in this sample of critically injured patients. The use of etomidate for procedural sedation and rapid-sequence intubation in this patient population should be reevaluated.
In 2002, Annane and colleagues1 published their findings on the effect of low-dose hydrocortisone and fludrocortisone acetate replacement in patients with septic shock. The authors demonstrated that the incidence of adrenal insufficiency (AI) was substantially underreported and, when left untreated in the septic shock population, was associated with significant mortality. Treatment of these critically ill patients with hydrocortisone and fludrocortisone resulted in a significant reduction in mortality. Numerous subsequent studies attempting to support or refute these findings have resulted in the inclusion of intravenous corticosteroids for patients with AI as a recommendation (albeit a grade C recommendation) in the Surviving Sepsis Campaign guidelines.2-5 The remaining debate now centers on what test is most appropriate to diagnose AI and application of these recommendations to other appropriate populations.6-8
Activation and disturbance of the hypothalamic-pituitary-adrenal (HPA) axis is an almost universal phenomenon after trauma and surgical stress. Although there has been much debate about the prognostic significance of cortisol level in these patients, initial cortisol levels appear to correlate with Injury Severity Score (ISS) in multitrauma patients.9-11 Before the study by Annane and colleagues, both Rivers et al12 and Barquist and Kirton13 demonstrated that a significant number of critically ill surgical patients who required vasopressor support had concurrent AI. Others authors have identified patients with severe traumatic brain injury and early hemorrhagic shock to be at even higher risk.9,14 Noting the similarities between the pathophysiologic features of septic shock and vasopressor-dependent systemic inflammatory response syndrome (SIRS), some institutions have begun to use intravenous corticosteroid replacement in patients who are nonresponders on cosyntropin stimulation.15,16 Recently, our group demonstrated that timely diagnosis and treatment of AI dramatically reduce mortality in trauma patients.16
In addition to hemorrhagic and septic shock, numerous causes of AI have been proposed. One well-known pharmacologic cause of AI is etomidate, the use of which is almost standard in rapid-sequence intubation protocols.17 The adrenal-suppressive effects of etomidate are reported to be reversible and last less than 24 hours (< 6 hours in most cases). Regardless, several authors question whether this drug should be eliminated from use in intensive care or at least be used more selectively in critically ill multitrauma patients.17-19 Low-dose corticosteroid replacement for AI has dramatically reduced mortality among critically ill patients. However, little attention and effort have been focused on identifying potential modifiable risk factors. To this end, we sought to identify risk factors among populations with a high risk of developing AI and to investigate whether initial rapid-sequence intubation doses of etomidate contribute to the nonresponse state in trauma patients.
This study was approved by the Vanderbilt University School of Medicine's institutional review board. We reviewed the trauma registry and trauma patient care cost center at an academic level I trauma center that provides trauma care for approximately 65 000 square miles of the southeastern United States. The trauma center admits approximately 3000 acutely injured patients annually, with more than 900 being admitted to the trauma intensive care unit (ICU). Approximately 750 of these patients require mechanical ventilation for longer than 24 hours. The 14-bed trauma ICU is located within a 31-bed trauma unit. The non-ICU beds include a 7-bed acute admission area and a 10-bed subacute care unit.
We queried the cost-center database for all trauma ICU patients who underwent cosyntropin stimulation testing (CST) for presumed AI from January 1, 2002, through December 31, 2004. These were matched with all trauma patients admitted during the same period through a query of the institution's Trauma Registry of the American College of Surgeons. Individuals missing from the trauma registry were excluded. Patients in whom only a single cortisol value was obtained or who received treatment without undergoing CST were excluded. In addition, patients with a history of AI or long-term corticosteroid use (> 5 mg/d prednisone equivalent) were excluded. Patients underwent CST on the basis of clinical suspicion of AI in the presence of (1) septic shock, (2) SIRS with vasopressor dependence, or (3) unexplained hypotension requiring ongoing resuscitation. The CST was performed only if the patient had completed the initial resuscitation process and was exhibiting evidence of unexplained and persistent hypotension/shock.
Septic shock and SIRS were defined in accordance with the guidelines of the American College of Chest Physicians and the Society of Critical Care Medicine.20 Hemorrhagic shock was defined according to the presence of class III or IV hemorrhagic shock documented on admission by the attending trauma surgeon. Mechanical ventilation requirement was defined as the need for ventilatory support lasting longer than 24 hours. Vasopressor dependence was defined as the requirement of vasopressor agents for 24 hours or more to sustain mean arterial pressure greater than 60 mm Hg. Etomidate exposure was defined as having received etomidate 24 hours or more before performance of the CST. Patients who did not receive etomidate or who received the drug within 24 hours of CST were defined as having nonexposure. Adrenal trauma was defined as the diagnosis of adrenal hematoma or hemorrhage according to the attending radiologist's final reading. Coagulopathy was defined as the presence of clinical bleeding combined with an abnormal laboratory coagulation profile. Human immunodeficiency virus status was determined by confirmed laboratory values using enzyme-linked immunosorbent assay testing and Western blot confirmation.
Baseline cortisol levels were obtained in patients in whom AI was suspected on the basis of the foregoing clinical indicators. Those with cortisol levels less than 9 μg/dL (to convert to nanomoles per liter, multiply by 27.588) or greater than 34 μg/dL did not undergo CST and were excluded. Those with cortisol levels between 9 and 34 μg/dL underwent CST, with a 250-μg dose of cosyntropin intravenously. Cortisol levels were then obtained 1 hour after cosyntropin infusion. Responders were defined as those with a 9-μg/dL or greater rise in their CST values. Nonresponders were defined as those with a poststimulation rise in cortisol level less than 9 μg/dL.
We evaluated trauma registry data, including age, sex, race, and mechanism of injury. Injury scores, including initial Glasgow Coma Scale, Revised Trauma Score, and ISS, were evaluated as well. The weighted Revised Trauma Score incorporates the Glasgow Coma Scale, systolic blood pressure, and respiratory rate, using coded and weighted values that range from 4 (normal) to 0 (poor) for each of the physiological variables (yielding a high of 7.841 and a low of 0). The Abbreviated Injury Scale is an anatomic injury scoring system that quantifies injuries in various body regions from a score of 1 (minor injury) to 6 (nonsurvivable). The ISS is calculated by summing the squares of the 3 highest Abbreviated Injury Scores in 3 different body regions (values range from 1-75).
The incidence of several outcomes was recorded and evaluated as to concurrent incidence of SIRS, septic shock, hemorrhagic shock, etomidate exposure, vasopressor dependence, adrenal trauma, human immunodeficiency virus status, and coagulopathy. Secondary outcomes included hospital days, ICU days, ventilator days, and ventilator-free days (days alive and not receiving ventilatory support).
Statistical analysis was performed with Stata 8.2 (StataCorp, College Station, Texas). The χ2 and Fisher exact tests were performed for evaluation of the binary variables. Two-sample t tests and the Wilcoxon rank sum test were used for continuous variables. Raw relative risks were calculated for the following potential risk factors: etomidate exposure, need for mechanical ventilation, need for vasopressors, hemorrhagic shock, septic shock, coagulopathy, and adrenal hematoma. Poisson regression was performed for each potential risk factor, controlling for age (categorized by quartiles), sex, race, mechanism of injury, ISS, and Revised Trauma Score by means of the robust standard error adjustment. A large number of covariates were included within this simple regression model relative to the number of events encountered within this sample. This ratio of covariates to events of interest increases the risk of aberrant regression results. Therefore, we performed propensity score analysis as a conservative check of our regression work. A comprehensive propensity score was generated for each patient to account for the potential impact of each covariate in predicting etomidate exposure. Poisson regression was repeated adjusting for this propensity score (first as a raw score and then categorized as quintiles) as a confirmatory/secondary regression analysis. A final regression model was used to consider the potential for confounding by the other risk factors (ie, Poisson regression incorporating quintiled propensity score, need for mechanical ventilation, need for vasopressors, hemorrhagic shock, septic shock, coagulopathy, and adrenal hematoma).
A total of 198 patients met initial search criteria through the cost center database. Twenty-nine patients were eliminated because of their absence from the trauma registry database. Three patients were eliminated because of documented preinjury history of AI, and 5 were eliminated because of long-term corticosteroid use. An additional 24 patients were found to have had only a single cortisol value determined and to have received treatment without undergoing CST. Thus, 137 patients were deemed eligible for final analysis.
Among the remaining sample, 83 patients (60.6%) were noted to be nonresponders and 54 patients (39.4%) were classified as responders. Table 1 and Table 2 provide demographic breakdowns for categorical and continuous variables, respectively. Mean age, sex, race, mechanism of injury, Revised Trauma Score, and mortality rate for nonresponders were not statistically different from those of responders. There was a statistically significant difference between responders and nonresponders in hospital days, ICU days, and total ventilator days. The difference in ISS between the responders and nonresponders approached statistical significance.
Breakdown by responder status for each potential risk factor is provided in Table 3. Statistically significant differences were noted between the 2 groups for the following potential risk factors: hemorrhagic shock, need for vasopressors, etomidate exposure, and coagulopathy. Of the 87 patients who were exposed to etomidate, only 9 had exposure less than 48 hours before CST. The remaining 78 patients underwent CST 48 hours or more after etomidate exposure. When these 9 patients were eliminated from the analysis, the risk of developing AI after etomidate exposure remained statistically significant.
Table 4 details the results of base Poisson regression for each potential risk factor. Hemorrhagic shock, need for vasopressors, etomidate exposure, and coagulopathy had relative risks of developing AI that were statistically significant after adjustment for age (categorized by quartiles), sex, race, mechanism of injury, ISS, and Revised Trauma Score. Table 5 includes results of confirmatory propensity analysis by etomidate exposure. The adjusted relative risk for etomidate exposure remained statistically significant throughout this confirmatory analysis. Figure 1 provides graphical comparison of the relative risk for each risk factor before and after regression analysis. Figure 2 depicts the effect of propensity score analysis on the relative risk of developing AI associated with etomidate exposure.
Activation of the HPA axis after trauma and surgical stress is an adaptive and protective mechanism, the integrity of which determines, in part, the response of the patient to injury and stress.14,21 Recent evidence demonstrates that an “uncoupled” or maladaptive systemic inflammatory response state precipitates development of multisystem organ failure and leads to higher mortality rates.22,23 Although “classic” AI is rare among trauma patients, occult AI has been noted in up to 60% of severely injured patients. Occult AI is associated with persistent systemic inflammation, a hyperdynamic cardiovascular state, and vasopressor-dependent shock.24,25 This state frequently goes undiagnosed and untreated. This failure in therapy is likely a result of the many similarities between occult AI and an exaggerated SIRS presentation. Recent data suggest that, like patients with septic shock, trauma patients exhibiting a catecholamine-dependent SIRS state should undergo HPA axis evaluation.14,16,25 On the basis of the findings of Annane et al1 and others,2-4,16 if AI is demonstrated by CST, these patients should receive a short course of low-dose corticosteroid replacement therapy.
Several studies have identified potential risk factors for development of AI. Many of these factors have been directly implicated in impairment of adrenal function, while others may simply be characteristics of high-risk populations. Human immunodeficiency virus infection, coagulopathy, adrenal injury, prolonged mechanical ventilation, and traumatic brain injury have been associated with the development of AI.14,16,26-28 In the present study, however, coagulopathy was the only factor from this list associated with an increased risk of development of AI in the severely injured patient. Conditions associated with splanchnic hypoperfusion, such as septic shock, hemorrhagic shock, and vasopressor dependence, have also been associated with the development of AI in the critically ill patient. The presence of hemorrhagic shock and the need for vasopressor support were each associated with a significantly increased risk of AI. Considering that corticosteroids control catecholamine release from sympathetic cells, the association of AI and vasopressor requirements is not surprising and likely explains the restoration of sympathetic tone with hydrocortisone administration.29,30 As to the presence of septic shock, there was a trend toward the development of AI, but this did not reach statistical significance.
Several pharmacologic agents frequently used in the severely injured and critically ill populations have been shown to impair adrenal function and steroidogenesis. Propofol impairs adrenal steroidogenesis, and other agents, such as midazolam hydrochloride, morphine sulfate, and fentanyl, have been noted to blunt the HPA axis and/or interfere with corticosteroid metabolism.31-33 Etomidate is an imidazole derivative frequently used as an induction agent for rapid-sequence intubation. In the trauma and critically ill patient, this drug has gained popularity for its rapid onset of action, cardiovascular stability, and limited respiratory depression. Its well-established adrenal suppression (via inhibition of 11β-hydroxylase), has, however, brought the use of etomidate into question.17,34 In the present study, we identified exposure to etomidate as a significant risk factor for the development of AI. In fact, of all those factors evaluated in the present study, etomidate use appears to be the only modifiable risk factor identified.
Following a report by Ledingham and Watt35 documenting an increase in AI and mortality among multitrauma patients receiving etomidate infusions, the use of this agent has been restricted to single-bolus administration.34 Using etomidate in this single-dose manner has been associated with less risk of AI and observation of only a transient effect (4-12 hours).35 In fact, recent data by Cohan and colleagues14 (which identified etomidate exposure in 70% of nonresponders to CST) noted that the effects were not observed after the first postinjury day. However, even after eliminating patients receiving etomidate within 48 hours of CST, we found that exposure to this induction agent was associated with a significantly increased risk of developing AI. Our findings are consistent with those of other investigators who have reported that the effects of etomidate on the HPA axis exceed the previously observed 24-hour window.36,37
Limitations of the present study include the relatively small sample size, which may have limited our ability to detect differences in certain high-risk populations such as those with adrenal trauma, septic shock, and traumatic brain injury. The present study was performed retrospectively by means of a trauma registry database. As such, the decision to perform CST was determined by the attending physician assigned to the trauma service at that time and his or her clinical suspicion of AI in a given patient. In addition, we did not evaluate other pharmacologic agents such as benzodiazepines, opiates, and anticonvulsants, which have been noted to interfere with the HPA axis and corticosteroid metabolism. The almost universal use of these agents within this population and their weaker association with adrenal dysfunction made such evaluation less likely to yield significant findings. Although there was no randomization among these patients, most received the standard protocol dose of etomidate (0.3 mg/kg) during performance of rapid-sequence intubation, and the time to clinical presentation of AI appeared to vary considerably. Other agents that were used for induction purposes in our patient population include propofol, midazolam, ketamine hydrochloride, and thiopental sodium. None of these agents was evaluated for the association with the development of AI. This lends further support to the need to validate these findings in a controlled, prospective, and randomized fashion.
Etomidate, a frequently used induction agent for intubation, has been shown to cause AI in the severely injured patient. Although sustained hemorrhagic shock and vasopressor dependence are risk factors we identified among the trauma patients who developed AI, etomidate was the only modifiable risk factor for development of AI identified in the present study. Moreover, this risk appears to remain significant regardless of time between exposure to etomidate and CST diagnosis of AI. Until this increased risk of developing impaired adrenal function has been refuted by a randomized, controlled trial, the routine use of etomidate in the critically injured patient should be reevaluated and alternative agents (midazolam, thiopental, phenobarbital, pentobarbital, and ketamine) should be considered.
Correspondence: Bryan A. Cotton, MD, Division of Trauma and Surgical Critical Care, Department of Surgery, Vanderbilt University Medical Center, 1211 21st Ave S, 404 Medical Arts Bldg, Nashville, TN 37212 (email@example.com).
Accepted for Publication: August 14, 2006.
Author Contributions: Dr Cotton had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Cotton and Guillamondegui. Acquisition of data: Cotton, Guillamondegui, Fleming, and Patel. Analysis and interpretation of data: Cotton, Guillamondegui, Carpenter, Morris, and Arbogast. Drafting of the manuscript: Cotton, Guillamondegui, and Carpenter. Critical revision of the manuscript for important intellectual content: Cotton, Guillamondegui, Fleming, Carpenter, Patel, Morris, and Arbogast. Statistical analysis: Carpenter and Arbogast. Administrative, technical, and material support: Cotton, Guillamondegui, Fleming, Patel, and Morris. Study supervision: Cotton.
Financial Disclosure: None reported.