Introduction

Etomidate has a favourable haemodynamic profile and is frequently used as a single bolus for rapid sequence intubation [1, 2]. However, the use of etomidate remains controversial in critically ill patients [3]. Etomidate inhibits 11β-hydroxylase and induces transient corticosteroid insufficiency [4]. Authors of clinical trials have frequently reported an increased rate of adrenal insufficiency and an increased risk of death with etomidate use [5, 6]. In a study involving a general population of intensive care unit (ICU) patients, etomidate did not alter specific outcomes (organ failures, duration of mechanical ventilation) or mortality compared with ketamine [7]. In a recently published systematic review, the authors confirmed an increased mortality rate with the use of etomidate in a sub-group of patients with sepsis [8]. However, hydrocortisone treatment after etomidate use did not reduce the mortality of septic patients [9] and did not reduce organ failures in critically ill patients [10].

Trauma patients are highly susceptible to nosocomial infection [11], and the impact of etomidate is poorly studied in this population. It has been demonstrated that up to 70 % of trauma patients developed a critical illness-related corticosteroid insufficiency (CIRCI) [12–14] and that this could increase the risk of hospital-acquired pneumonia (HAP) [15, 16]. We have recently demonstrated that low-dose hydrocortisone decreases the rate of HAP in multiple trauma patients with CIRCI [17]. Etomidate may increase the rate of post-traumatic complications, but little data are available to date on this specific population of patients [18]. To the best of our knowledge, the ability of hydrocortisone to overcome the effects of etomidate has not yet been delineated in trauma patients.

For the study reported here, we used data from the HYPOLYTE trial, a prospective multi-centre study of low-dose hydrocortisone treatment in intubated trauma patients [17]. We hypothesized that a single dose of etomidate increases the proportion of CIRCI and the rate of HAP. We also tested the effects of the hydrocortisone treatment on the proportion of HAP in patients receiving etomidate.

Materials and methods

Study design

This study is a sub-study of the HYPOLYTE trial, a multi-centre, randomized, double-blind, placebo-controlled trial of hydrocortisone in severe trauma patients [17]. Following approval by the Institutional Review Board of Angers (France), patients in seven French ICUs were enrolled in the HYPOLYTE trial from 15 November 2006 to 4 August 2009. Prior to enrollment, written informed consent was obtained from a next-of-kin. Retrospective consent, if possible, was obtained from patients.

Patients

All of the 149 patients randomized in the HYPOLYTE trial were included in the analysis for this sub-study. In the HYPOLYTE trial, the inclusion criteria were multiple trauma, age of >15 years 3 months and the expectation that mechanical ventilation would be required for >48 h. Exclusion criteria were previous adrenal insufficiency, previous immunosuppression (See [17] for definitions), treatment with corticosteroids within the last 6 months and pregnancy. Etomidate use was not controlled for in the study.

Hydrocortisone therapy

For the purpose of the HYPOLYTE trial, treatment with hydrocortisone or placebo were started in the first 36 h following the trauma. Hydrocortisone (Upjohn Serb®, Paris, France) or placebo were administered for 7 days in patients with CIRCI (continuous intravenous infusion, 200 mg/day for 5 days, 100 mg on day 6 and 50 mg on day 7) and were stopped during the first 2 days that patients were CIRCI-free (after the results of the corticotropin test had been received).

Critical illness-related corticosteroid insufficiency definition

Immediately before beginning the treatment, but at least 8 h after a bolus injection of etomidate, a short corticotropin test was performed in all 149 patients. Cortisolemia was assessed before and 30 and 60 min after an intravenous bolus of 0.25 mg of corticotropin (Novartis, Rueil-Malmaison, France). The delta of cortisolemia was calculated as [cortisolemia (30 or 60 min as stated) basal cortisolemia]. CIRCI was defined as a basal cortisolemia level of <15 μg/dL (413 nmol/L) or a delta of cortisolemia of <9 μg/dL (248 nmol/L).

Hospital-acquired pneumonia definition [19]

Pneumonia was considered as diagnosis when at least two signs (body temperature >38 °C; leukocytosis >12,000/mL or leukopenia <4,000/mL; purulent pulmonary secretions) were associated with the appearance of a new infiltrate or changes in an existing infiltrate on the chest X-ray. Diagnosis was confirmed by tests on a respiratory tract sample using a quantitative culture with a predefined positive threshold of 104 colony-forming units per milliliter (CFU/mL) for a bronchoalveolar lavage or nonbronchoscopic sample or 103 CFU/mL for a protected specimen brush. HAP was defined as pneumonia that occurred 48 h after admission that had not been incubating at the time of admission.

Data collection

Overall patient characteristics, including demographics, injury severity score and abbreviated injury score, fluid infusions, vasopressors, antibiotic prophylaxis, etomidate use, surgery, infections, organ failures, length of ventilatory support, ICU hospitalization and death on day 28 of follow-up were recorded.

Endpoints

The primary endpoint of the HYPOLYTE trial was HAP on day 28 of follow-up [17]. In this sub-study, the primary endpoint was the 28-day HAP rate and whether etomidate increased the HAP rate. We compared patients with or without etomidate therapy. Secondary outcomes were the results of the short corticotropin test, duration of mechanical ventilation (MV) and ICU length of stay and mortality.

Statistical analysis

Continuous variables are expressed as the median and interquartile range (IQR) and categorical variables as percentage. Characteristics and outcomes of patients with etomidate and those without etomidate were compared with the χ2 test, Student’s t test or the Wilcoxon rank-sum test were used as appropriate. Multivariate analyses of the occurrence of HAP on day 28 were carried out using logistic regression models adjusted on hydrocortisone therapy. The calibration of the models was tested by the Hosmel–Lemeshow’s test. The effect of etomidate on HAP within 28 days was assessed in the hydrocortisone and placebo groups by means of a Cox multivariate proportional hazards model that included three predefined covariates: centre, traumatic brain injury (TBI, presence or absence), and the injury severity score (≤30 or >30). Corresponding hazard ratios (HR) along with their 95 % confidence intervals (CI) were reported. A sensitivity analysis was performed using propensity score-adjusted models. The propensity score was constructed using logistic regression analyses for etomidate use, including variables recorded before the induction of anaesthesia, such as age, sex, Injury Severity Score (ISS), TBI and hypotension. The caliper matching method was used for matching 54 (100 %) of the patients in the non-etomidate group with 54 patients in the etomidate group (caliper size: one-quarter of the standard deviation of the logit of the propensity score). Discrimination of the propensity-adjusted logistic regression model was assessed using the area under the receiver-operating characteristic curve (c statistic). All statistical tests were two-sided. A p value of <0.05 was considered to be statistically significant. Statistical analyses were performed using the SAS ver. 9.1 (SAS, Cary, NC) software package.

Results

Patients

Of the 149 patients analysed, 95 (64 %) received a single bolus of etomidate prior to inclusion in the study [see Electronic Supplementary Material (ESM) Fig. S1 for flow chart]. No patient received etomidate after study inclusion. There were no differences in the general characteristics of patients who received etomidate (etomidate group) compared with those who had not received etomidate (non-etomidate group) except for the time from trauma to tracheal intubation [60 (IQR 30–120) min and 105 (IQR 30–270) min, respectively; p = 0.014) (Table 1). The median time from induction of anaesthesia to corticotropin test was 22 (IQR 20–28) h in the non-etomidate group and 22 (IQR 17–29) h in the etomidate group (p = 0.826).

Table 1 General characteristics of the patients
Full size table

Etomidate effects on basal cortisolemia and results of the corticotropin test

Of the 95 patients in the etomidate group and 54 patients in the non-etomidate group, 79 (83 %) and 34 (63 %), respectively, satisfied the criteria for CIRCI (p = 0.006). Basal cortisolemia was not different between the etomidate group and the non-etomidate group [20.0 (IQR 12.5–27.9) and 18.5 (IQR 13.4–25.4) μg/dL, respectively; p = 0.73]. The delta of cortisolemia at 60 min was significantly lower in the etomidate group than in the non-etomidate group [7.6 (IQR 2.7–14.0) and 12.9 (IQR 5.2–19.0) μg/dL, respectively; p = 0.007] (Table 2). Time from induction of anaesthesia to corticotropin test was not correlated with basal cortisolemia in the etomidate or the non-etomidate group (Fig. 1a, b). Time from induction of anaesthesia to the cosyntropin test was correlated with delta cortisolemia in the etomidate group (R 2 = 0.19, p = 0.001; Fig. 1c) but not in the non-etomidate group (Fig. 1d).

Table 2 Etomidate use alters adrenal function in trauma patients
Full size table
Fig. 1
figure 1

Correlations between delta cortisolemia and time from induction of anaesthesia to cosyntropin test in patients who received etomidate (a, c, filled circles) and who did not receive etomidate (b, d open circles). Horizontal dotted line (9 μd/dL for delta maximum and 15 mg/dL for baseline) discriminate between adapted corticosteroid function (upper) and corticosteroid insufficiency (lower). Percentages of patients by quadrant are provided

Full size image

Etomidate was an independent risk factor for HAP

Sixteen (29.6 %) patients in the non-etomidate group and 49 (51.6 %) patients in the etomidate group developed HAP by day 28 (p = 0.009). The median time to HAP was 4 (IQR 3–8) days in the non-etomidate group and 6 (IQR 4–8) days in the etomidate group (p = 0.25; see ESM Fig. S2 for Kaplan–Meier curves). We hypothesized that etomidate may be an independent risk factor for HAP and therefore performed a multivariate analysis for the risk factors of HAP. Twenty-nine variables potentially associated with the occurrence of HAP in trauma patients [19, 20] were tested by univariate analysis (Table 3). In the univariate analysis, HAP was associated (at p < 0.25) with etomidate use, ISS, TBI and time from trauma to tracheal intubation (Table 3). After the logistic regression analysis, HAP remained associated with etomidate use [odds ratio (OR) 2.48, 95 % CI 1.19–5.18; p = 0.016] and TBI (OR 3.05, 95 % CI 1.5–6. 21; p = 0.002) (Table 4). In the sensitivity analysis, after adjustment for the propensity score and the study drug (hydrocortisone), the association between etomidate use and HAP remained significant (adjusted OR 2.52, 95 % CI 1.08–5.87; p = 0.032) (Table 4).

Table 3 Univariate analysis of factors relating to 28-day hospital-acquired pneumonia
Full size table
Table 4 Results of the logistic regression analysis for predictors of hospital-acquired pneumonia
Full size table

Outcomes of etomidate-exposed patients

The duration of MV was 11 (6–19) days in the non-etomidate group and 13 (7–22) days in the etomidate group (p = 0.275), and the ICU length of stay was 15 (9–28) days in the non-etomidate group and 16 (11–27) days in the etomidate group (p = 0.422). Four (7.4 %) patients in the non-etomidate group and six (6 %) patients in the etomidate group died (p = 1).

Hydrocortisone improved the outcomes of etomidate-exposed patients

Etomidate increased the rates of both CIRCI and HAP. We therefore assessed the impact of hydrocortisone in the group of patients receiving etomidate. Of the 95 patients in the etomidate group, 45 (52.6 %) were treated with hydrocortisone and 50 (47.4 %) with placebo. Of the 45 patients receiving hydrocortisone in the etomidate group, 18 (40 %) developed HAP compared with 31 (62 %) of the 50 patients receiving placebo (p = 0.032; Table 5). Of the 20 patients in the etomidate group with early hydrocortisone infusion (≤20 h after trauma), eight (40 %) developed HAP by day 28 compared with ten of the 25 (40 %) patients in the etomidate group with late hydrocortisone infusion (>20 h after trauma; p = 1). The duration of MV was 9 (6–16) days in patients receiving hydrocortisone and 18 (10–25) days in those receiving placebo (p < 0.001), and the ICU length of stay was 13 (8–20) days with hydrocortisone and 21 (14–28) days with placebo (p < 0.001) (Table 5). Three of 45 (6.7 %) patients in the hydrocortisone group and three of 50 (6 %) in the placebo group died (p = 1) (Table 5).

Table 5 Impact of hydrocortisone on outcomes of patients treated with etomidate
Full size table

In the HYPOLYTE study [17], after the results of the corticotropin test were known, patients with CIRCI were treated for 7 days, whereas treatment was stopped in patients without CIRCI (see ESM Fig. S1). For the 79 patients with CIRCI in the etomidate group, 14 (38.9 %) of 36 patients receiving hydrocortisone and 25 (58.1 %) of 43 patients receiving placebo developed HAP (p = 0.088; see ESM Table S2). The duration of MV as well as ICU length of stay were reduced in patients receiving hydrocortisone compared with placebo (p < 0.001 for both durations; see ESM Table S2). In the etomidate group, 16 patients (9 patients in the hydrocortisone group and 7 in the placebo group) did not have CIRCI and were treated for 34 (20–49) h. Hydrocortisone did not alter the outcomes of patients without CIRCI in the etomidate group (see ESM Table S1).

Discussion

In this sub-study of the HYPOLYTE trial, basal cortisolemia was not modified by etomidate treatment, whereas the delta of cortisolemia was altered. Etomidate was found to be an independent risk factor for HAP in intubated trauma patients, and hydrocortisone was found to significantly decrease the rate of HAP and the duration of MV for etomidate-exposed patients. These results could have major implications in the care of severe trauma patients.

CIRCI corresponds to an inadequacy between the severity of illness and the cortisolemia [21]. CIRCI is defined by either an insufficient rise in cortisol production [22] and/or a peripheral resistance to glucocorticoids [23]. The combination of a basal cortisolemia of <10 μg/dL and/or a delta cortisolemia of <9 μg/dL provides the best specificity and sensitivity for the diagnosis of CIRCI in severe septic patients [24]. Recent recommendations of the American College of Critical Care Medicine have advocated the use of either basal or delta cortisolemia for the diagnosis of CIRCI in critically ill adult patients [25]. However, etomidate inhibits 11 beta-hydroxylase [4], inducing an iatrogenic corticosteroid insufficiency [26] that may be differentiated from CIRCI. Thus, the diagnosis of CIRCI after etomidate use is challenging and still poorly documented. As previously mentioned in a sub-study of the CORTICUS trial [9], etomidate was able to bias the result of the cosyntropin test up to 36 h but did not alter basal cortisolemia. These results have already been described in the first 6 h after rapid sequence intubation with etomidate in trauma patients [18]. The current results confirm that after etomidate use, basal cortisolemia remains a reliable criterion for the diagnosis of CIRCI.

The controversy regarding the use of a single bolus of etomidate in critically ill patients was raised in the last decade, mainly in septic patients [3]. Etomidate increases both the rate of corticosteroid insufficiency and the rate of death in septic patients [8], but to date no causal link has been demonstrated. Etomidate is widely used in trauma patients as a single bolus for tracheal intubation. There are many advantages to using etomidate in this population, including a short duration of action with few cardiovascular side effects [2]. Despite these attractive characteristics, we observed an increased rate of HAP in etomidate-exposed patients. Moreover, Hildreth et al. [18] reported increases in the duration of MV and the ICU length of stay of trauma patients after etomidate use. These data confirm the morbidity of etomidate previously demonstrated in septic patients.

The equilibrium of the inflammatory response is a major determinant in avoiding post-traumatic infections [27]. Etomidate increases pro-inflammatory cytokine production ex vivo in whole blood cell cultures challenged with lipopolysaccharide [28] and could therefore prolong the systemic inflammatory response syndrome (SIRS) frequently observed in trauma patients. Prolonged SIRS is predictive of nosocomial infection in trauma patients [29] and has also been correlated with corticosteroid dysfunction in critically ill [30] and trauma patients [13]. It may therefore be hypothesized that etomidate exposes patients to HAP, at least in part, through an excessive inflammatory response.

Published data suggest that adequate cortisol production is critical for normal host defence against infection [15]. In this setting, patients presenting with Addison disease are more susceptible to infections [31]. Cortisol, a natural hormone, increases neutrophil activity [32, 33], preserves monocyte function and increases the homing of dendritic cells [34]. With respect to glucocorticoid therapy, high-dose methylprednisolone and low-dose hydrocortisone display different effects on the immune response [35], and while the former’s regimen is immunosuppressive and induces apoptosis of numerous immune cells [36, 37], hydrocortisone acts as an immunomodulatory drug [34]. In comparison to those receiving placebo, low doses of hydrocortisone were found to enhance neutrophil functions [32] and decrease the overwhelming inflammatory response in septic patients [38]. At the same time, interleukin-12 production was increased, whereas interleukin-10 decreased—with no modification of monocytic human leukocyte antigen DR expression. These data suggest that in septic patients, a low-dose of hydrocortisone may enhance the immune response rather than induce “immunosuppression” [38]. In intubated patients with respiratory failure and CIRCI, hydrocortisone was found to accelerate weaning from ventilatory support [39]. It was also demonstrated that hydrocortisone reduced both the mortality rate and the duration of hospitalization in a population of patients suffering from severe community-acquired pneumonia [40]. Thus, low doses of hydrocortisone should systematically be considered after a single bolus of etomidate in patients at high risk for CIRCI and HAP. In a recent study, Payen et al. [10] failed to demonstrate any improvement in outcomes when treating etomidate-exposed patients with hydrocortisone. The discrepancies between our results and these data merit explanation. Firstly, Payen et al. included a general population of ICU patients, whereas we studied trauma patients and treated all etomidate-exposed patients, although only patients with CIRCI were treated 7 days in the HYPOLYTE study. Secondly, the rate of HAP was not studied. Finally, there was an important trend in favour of hydrocortisone regarding ICU length of stay and ventilator days [4 (IQR 1–10) vs. 8 (IQR 4–17) days and 2 (IQR 1–10) versus 4 (IQR 1–10) days, respectively]. One could hypothesize that the Payen et al. study [10] lacked power for these two outcomes.

This study displays a number of strengths, mainly because the data are from a multi-centre, double-blind study and also because this is the first study that reports the impact of etomidate use in a specific population of trauma patients. However, limitations must be noted. First, the reasons for etomidate use were neither documented nor controlled, although no imbalance in etomidate use was apparent between the centres (data not shown). Also, one could hypothesize that etomidate was used in sicker patients. However, the severity scores did not appear to be different between patients with or without etomidate, and etomidate was associated with HAP in a multivariate analysis. However, the use of other trauma severity scores [41] or of generalist severity scores [42] with better prognostic values may have provided other information. Second, we used a more liberal CIRCI definition than the consensus statement, which was not available at the beginning of the study in 2006 (basal cortisolemia <10 vs. 15 μg/dL) [25]. Previous studies in trauma patients used a basal cortisolemia threshold of ≥15 μg/dL [12–14], and we chose a definition proposed by Annane et al. [43]. Of the 149 patients, 23 (15 %) would not have had CIRCI if the current definition of the consensus statement had been used [25]. Third, the effect of etomidate on adrenal function is transient, but we report more pneumonia (early and late onset) in the etomidate group. No definitive conclusion can be drawn from this descriptive study in terms of the physiological mechanism that links etomidate to pneumonia. In the HYPOLYTE study, the majority of pneumonia was early onset pneumonia [<day 8, 54 early pneumonia (83 %)]. This may explain, at least in part, this surprising result. Finally, etomidate-induced corticosteroid insufficiency (characterized by a decrease in delta cortisolemia with no alteration of basal cortisolemia) is an independent risk factor for HAP, which strengthens the concept of CIRCI based on hyporeactivity to corticotropin [21].

Conclusions

A single bolus of etomidate is an independent risk factor for HAP in intubated trauma patients. A low-dose of hydrocortisone prevented etomidate-induced susceptibility to HAP. Even after etomidate injection, basal cortisolemia remains a reliable criterion for the diagnosis of CIRCI. These data should alert physicians to limit the use of etomidate for the intubation of severe trauma patients unless they use hydrocortisone to counteract the deleterious effects of the drug.