A cohort, observational study was performed in an adult ICU of a university hospital from June 2006 until December 2009. Data were extracted from prospective studies conducted in our ICU and previous databases 1251526. The study was approved by the local ethics committee (Comité de Gestion et d'Organisation de l'Anesthésie Réanimation, Montpellier University Hospital) and, in accordance with French law, informed consent was waived. We adhered to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines 29.

Etomidate blocks cortisol synthesis primarily by inhibiting the activity of 11β-hydroxylase for at least 24 to 48 hours 1030. Therefore, to describe the potential impact of etomidate on early and late outcome in septic shock patients, consecutive patients were eligible if they had received an induction agent for endotracheal intubation within the first 48 hours of septic shock onset. Patients were treated according to international guidelines for management of severe sepsis and septic shock 31 but all received hydrocortisone 15. Exclusion criteria included pregnancy, age < 18 years, moribund patients, immunosuppression, and long-term or short-term corticosteroid treatment within the past 4 weeks. A cosyntropin stimulation test with 250 μg cosyntropin was performed in all septic shock patients. A 50 mg intravenous bolus of hydrocortisone was then administered every 6 hours, beginning within the first 12 hours of septic shock, for at least 5 days, tapered and stopped in 5 days according to the reversal of shock. Patients were grouped as those having received etomidate for intubation (etomidate cohort) versus those subjects having received another hypnotic (non-etomidate cohort).

Septic shock was defined by evidence of infection and a systemic response to infection, in addition to systolic blood pressure < 90 mmHg, despite adequate fluid replacement, or a need for vasopressors for at least 1 hour, according to the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference Committee criteria 32. Nonresponse to the cosyntropin stimulation test using an immunoradiology assay (SP2100; Beckmancoulter SAS, Roissy, France) was defined by a delta cortisol (60 minutes after 250 μg cosyntropin) < 9 μg/dl 152628. CIRCI was defined by a delta cortisol (60 minutes after 250 μg cosyntropin) < 9 μg/dl or a baseline plasma cortisol level < 10 μg/l 12.

A standardized data collection instrument and guidance tool was developed for data collection. Record review and data extraction were performed by a single investigator (NC) and regular meetings were conducted to address any problems encountered during the data collection phase according to the recommendations that have been published to minimize validity threats in chart review studies 33. Upon ICU admission, the baseline characteristics and the main variables obtained before intubation were recorded either by a nurse (from June 2006 to Jan 2009) or by computer-driven software plugged to the monitor, which recorded automatically all the variables.

At the time of intubation, clinical data including reason for intubation, interventions including sedative agent used, need for and doses of vasopressors were recorded. During the intubation procedure, drug administration and the difficulty to intubate rate (defined by three or more attempts at laryngoscopy to place the endotracheal tube into the trachea and/or > 10 minutes using conventional laryngoscopy and/or the need for another operator) 5 were documented. Within the first hour after intubation we recorded the short-term life-threatening complications that occurred, defined as previously reported 25: cardiac arrest, severe cardiovascular collapse (defined as systolic blood pressure < 65 mmHg recorded at least once and/or < 90 mmHg that lasted 30 minutes despite 500 to 1,000 ml fluid loading and/or requiring introduction of vasoactive support) and severe hypoxia (defined as a decrease in SpO₂ level < 80% during attempts). Patients who already presented a cardiovascular collapse after fluid loading or who were severely hypoxemic (SpO₂ < 80%) after preoxygenation by noninvasive positive-pressure ventilation were not considered to have had an intubation-related complication, but rather to have presented a life-threatening condition requiring an emergency endotracheal intubation.

During the ICU stay, we documented the results for basal plasma cortisol and that after the cosyntropin test, as well as total amounts and durations of hydrocortisone and vasopressor treatments from day 0 to day 5. Outcome data include the duration of shock, length of mechanical ventilation, nosocomial infection incidence, ICU and hospital lengths of stay, and day-28 mortality.

We had sufficient resources to review 102 patients in total. Descriptive data of quantitative variables were summarized as the mean ± standard deviation or median with interquartile range, according to the normality of the distribution, assessed with the Shapiro-Wilk test and compared with the Mann-Whitney or t test. Categorical data were expressed as the number and percentage and were compared with a chi-square analysis.

Using two statistical methods, we assessed the occurrence of short-term life-threatening complications and the long-term outcomes according to the administration of etomidate versus another hypnotic drug. First, unadjusted differences between patients receiving etomidate or not were compared using logistic regression after calibration with the Hosmer-Lemeshow wellness-of-fit test. Furthermore, long-term survival was assessed by a Cox regression in which we included all variables associated with P < 0.20 in the univariate analysis. A stepwise procedure then allowed the final multivariate model to be obtained.

Second, since patients were not randomly assigned to etomidate or other hypnotic in this observational study, we developed a propensity score using all variables associated with P < 0.20 in the univariate analysis. The propensity score is defined as a subject's probability of receiving a specific treatment (for example, etomidate) conditional on the observed covariates, and thus controls for selection bias in observational studies 34. For the coupling process, optimal one-to-one nearest neighbor matching was used. When needed, patients already matched were replaced by the closest one in the in the propensity score. P < 0.05 was considered significant. Statistical analysis was performed by an independent statistician (NM), with R software (version 2.10.1).

During the study period, among 1,632 patients admitted to the ICU, 331 presented septic shock during their stay. Among these 331 patients, 229 either developed septic shock > 48 hours after intubation, did not have a cosyntropin test or data could not be extracted from the charts. Thus, 102 patients meeting the inclusion criteria were analyzed; 60 in the etomidate cohort and 42 in the non-etomidate cohort. The hypnotics used to induce anesthesia for intubation in the non-etomidate cohort were ketamine (n = 18), propofol (n = 10), thiopental (n = 13) or none (n = 1). The nonabdominal source of sepsis, higher Simplified Acute Physiology Score II 35 and Sequential Organ Failure Assessment 36 severity scores were more frequently observed in the etomidate cohort (Table 1).

We first evaluated the association of hypnotics, intubation-related life-threatening complications and outcome in unmatched cohorts.

Intubation was indicated mainly for urgent surgery (42%) or acute respiratory failure (35%). Myorelaxants were used in nearly all of the procedures without any complications related to their use (Table 1). Intubation was difficult in 10 cases (10%), independent of the administered hypnotic. Short-term life-threatening complications within 1 hour of intubation occurred in 37 (36%) of the 102 studied patients (Figure 1). In univariate analysis, the Simplified Acute Physiology Score II was associated with a higher risk of complications and both the administration of norepinephrine prior to intubation and the use of a drug other than etomidate to facilitate intubation were associated with a lower risk of complications (Table 2). In multivariate analysis, the administration of norepinephrine prior to intubation was the sole independent protective factor for life-threatening complications occurring after intubation (Table 2).

Patients were compared according to the hypnotic they received to facilitate intubation. The cosyntropin test was performed within 24 hours after intubation in 85% of the patients, and after the first 24 hours in 15% of the population but always before the first dose of hydrocortisone. Hydrocortisone treatment was started 540 (300 to 1,125) minutes after intubation. The basal plasma cortisol concentration was significantly lower (19 (14 to 35) μg/dl versus 31 (17 to 45) μg/dl; P = 0.04) and the percentage of nonresponders to the cosyntropin stimulation test was significantly higher (79% vs. 52%; P = 0.01) in the etomidate cohort compared with the non-etomidate cohort. CIRCI was also significantly more frequently observed in the etomidate cohort compared with the non-etomidate cohort (79% vs. 59%; P = 0.04). In the etomidate cohort, the cumulative hydrocortisone dose was significantly higher (1,250 (650 to 1,650) mg vs. 750 (350 to 1,150) mg; P = 0.02) and the duration of treatment was significantly longer (168 (96 to 216) hours vs. 96 (48 to 162) hours; P = 0.01) than in the non-etomidate cohort.

Norepinephrine was administered within 12 hours after intubation in 100% of the patients without significant difference between cohorts. Patients in the etomidate cohort needed a higher cumulative dose of norepinephrine during their ICU stay compared with patients anesthetized with another hypnotic (95 (39 to 203) mg vs. 58 (30 to 97) mg from day 0 to day 5; P = 0.02). The duration of norepinephrine treatment was not different between cohorts (58 (37 to 94) hours in the etomidate cohort vs. 48 (25 to 81) hours in the non-etomidate cohort; P = 0.20).

The incidence of nosocomial infections, length of mechanical ventilation, and lengths of ICU and hospital stay did not significantly differ between cohorts (Table 3).

Although the crude day-28 mortality was not different according to the drug used to facilitate intubation, the Cox regression model yielded a hazard ratio for death at day 28 in the etomidate cohort, as compared with the non-etomidate cohort, of 0.33 (0.12 to 0.90; P = 0.03) (Table 4). The Hosmer-Lemeshow test showed that the model fits to predict mortality, with 82% of well-classed patients and P = 0.16. Second, we evaluated the association of hypnotics and both intubation-related life-threatening complications and outcome in matched cohorts. Propensity score matching resulted in a cohort of 56 patients, with 28 patients who received etomidate and 28 patients who did not receive etomidate. In this cohort of 56 patients, matching was based on etomidate use, the Simplified Acute Physiology Score II score without counting age and the basal plasma cortisol level. The occurrence of intubation-related life-threatening complications was similar in both the etomidate and the non-etomidate cohorts. The Kaplan-Meier estimator for 28-day mortality using propensity score matching was significantly lower in the etomidate cohort than in the non-etomidate cohort and showed a hazard ratio for death in the ICU in the etomidate cohort, as compared with the non-etomidate cohort, of 0.33 (0.112 to 0.988) (Figure 2). The c-statistic for the propensity score was 0.7794.