The effect of dexmedetomidine on vasopressor requirements in patients with septic shock: a subgroup analysis of the Sedation Practice in Intensive Care Evaluation [SPICE III] Trial

The present study is an exploratory, post hoc, retrospective subgroup analysis of patients with septic shock who were enrolled in the SPICE trial III. The subgroup analysis was performed at two of the participating centers, the Austin Hospital, Melbourne, Australia, and the University Hospital of Bern, Switzerland, and was approved by the ethics committees at both sites (approval number LNR/15/Austin/391 and KEK-ID2018-00746).

In addition to the abovementioned inclusion criteria of the SPICE III trial, patients had to meet all the following criteria to be eligible for this subgroup analysis: documented (or strong suspicion of) infection with at least 2 clinical signs of inflammation (temperature > 38 °C or < 36 °C, heart rate > 90/min, respiratory rate > 20/min, or PaCO₂ < 32 mmHg, white cell count > 12 × 10⁹/l or < 4 × 10⁹/l or > 10% immature neutrophils), and administration of vasopressors or inotropes prior to randomization and for a cumulative duration of ≥ 4 h to maintain blood pressure targets set by the treating clinician.

In addition to the exclusion criteria of the SPICE III trial, we excluded patients meeting our inclusion criteria more than 24 h before randomization. A detailed list of all inclusion and exclusion criteria is given in the supplementary appendix. To reflect the effects of the intervention (dexmedetomidine or usual care) unaffected by protocol deviations or non-adherence, we performed a per-protocol analysis excluding patients who did not receive the allocated sedation regimen and patients whose treatment goal was changed to end-of-life care within the first 48 h after randomization.

The primary outcome was the median vasopressor dose in the first 48 h after randomization, expressed as noradrenaline equivalent dose (NEq). We calculated NEq using the method described by Khanna and co-workers [13] as

to account for the different vasopressors used at the two study sites. Secondary outcomes included cumulative and peak NEq dose, change in NEq dose from baseline to peak dose, and the ratio of NEq to MAP (NEq/MAP), to account for different blood pressure targets set by the treating physician. Exploratory outcomes included cumulative duration of vasopressor support, ICU and hospital mortality, duration of mechanical ventilation, and ICU length of stay.

Data collection was performed using existing ICU-based electronic databases (Cerner Electronic Health Record, North Kansas City, MO, USA, and Centricity Critical Care, Clinisoft, GE Healthcare Europe, Helsinki, Finland) and medical record review. Whenever possible, data were downloaded from electronic medical records. Continuous measurements, recorded at minimum once every 2 min, were resampled to obtain hourly median values of hemodynamic data, sedative and vasoactive drug doses for the first 48 h after randomization. Arterial blood pressure was monitored invasively in all patients. Where manual data collection was necessary, it was performed using double data entry or cross-checked by a second investigator. Demographic data and patient-centered outcomes were obtained from the SPICE III database.

Data was initially assessed for normality and log-transformed where appropriate. Group comparisons were performed using chi-square tests for equal proportion, Student’s t tests for normally distributed data, and Wilcoxon rank-sum tests otherwise, with results presented as numbers (%), means (standard deviations), or medians (interquartile range), respectively. Longitudinal analysis was performed using mixed linear modeling fitting main effects for treatment, time, and an interaction between the two to determine if groups behaved differently over time. Multivariable longitudinal sensitivity analyses were performed firstly adjusting for baseline imbalance and known covariates (admission diagnosis, hospital site, ratio of noradrenaline equivalent over mean arterial pressure at baseline, continuous renal replacement therapy, age, administration of hydrocortisone, and presence of liver cirrhosis) and then secondly adjusting for the same covariates but with treatment group replaced by actual dexmedetomidine dosage. Comparison of proportions for secondary outcomes was determined using logistic regression with results presented as odds ratios (95%CI). Differences for continuous secondary outcomes were determined using quantile regression with results presented as difference of medians (95%CI). To account for the competing risk of death, durations of vasopressor support and ventilation are presented as cumulative incidence functions with censoring for death and comparison using Grey’s test. Time to death was displayed using Kaplan-Meier curves with comparison using a log-rank test. All data were analyzed using SAS software, version 9.4 (SAS Institute Inc., Cary, NC, USA), and a two-sided p value of 0.05 was used to indicate statistical significance. No adjustments were made for multiple comparisons.

Between November 2013 and February 2018, 417 patients were recruited in the SPICE III trial at both sites (196 patients at the Austin Hospital and 221 patients at the University Hospital of Bern). Among those who provided written informed consent, 87 patients fulfilled our inclusion criteria of septic shock. After excluding two patients from each group who did not receive the assigned sedation, 83 patients were included in the analysis. Of these, 44 (53%) had been assigned to receive dexmedetomidine (DEX group) and 39 (47%) to usual care. Fifty-seven patients (68.7%) were male, with a mean age of 65.4 years, and a mean pre-randomization APACHE II score of 25.1. Patient characteristics and treatment at randomization were similar in the two groups, except for a higher creatinine level and higher coagulation component of the SOFA score in the usual care group (Table 1).

In the first 48 h after randomization, there was no significant difference in median NEq dose between the DEX group and the usual care group (Table 2 and Fig. 1). Similarly, we found no significant difference in cumulative NEq dose, peak NEq dose, and relative change in NEq from baseline to peak dose between the two groups.

Over the first 48 h, patients in the DEX group had a numerically higher MAP (Fig. 2), although the differences did not reach statistical significance. However, vasopressor requirements to maintain the target MAP (expressed by the NEq/MAP ratio) were lower in the DEX group compared to the usual care group (ratio of difference in geometric means 1.74 [1.02, 2.95], p = 0.04). This result remained significant when adjusted for admission diagnosis, hospital site, baseline NEq/MAP ratio, continuous renal replacement therapy, age, administration of hydrocortisone, and liver cirrhosis (ratio of adjusted difference in geometric means 1.44 [1.07–1.95], p = 0.02) (Fig. 3).

On multivariable sensitivity analysis adjusting for actual dexmedetomidine dosage, the results were consistent with the above findings: Dexmedetomidine dose was associated with a reduction in the NEq/MAP ratio (parameter estimate of − 0.17 +/− 0.07; p = 0.02), indicating lower vasopressor requirements to maintain the target MAP. The NEq/MAP ratio was significantly greater in patients treated in Australia, in patients not on continuous renal replacement therapy, without liver disease, without portal hypertension, or with a renal or cardiovascular APACHE admission category (Table 3).

There was no significant difference in ICU and hospital length of stay, number of patients alive and vasopressor-free at 48 h, time to vasopressor weaning, duration of invasive ventilation, and mortality between the two groups (Table 2 and Figs. 4, 5, and 6). Unlike the SPICE III trial, in the present sub-study, we found no significant difference in outcomes when comparing age groups above and below the median age (63.7 years) (Table S4).

Nineteen patients (43.2%) in the DEX group and 16 patients (41.0%) in the usual care group received hydrocortisone (p = 0.84). Vasoactive drugs were discontinued within 48 h of randomization in 21 (47.7%) patients in the DEX group and 21 (53.8%) patients in the usual care group (p = 0.58). Patients in the DEX group received significantly lower doses of propofol and less midazolam. However, they received similar doses of opioids (Table S1). The median Richmond Agitation-Sedation Scale (RASS) score on study day 1 was − 4 [− 4, − 2] in the DEX group and − 4 [− 4, − 3] in the usual care group (p = 0.42). There was no significant difference regarding treatment with vasopressors, inotropes, and hydrocortisone between patients treated before and after the publication of the 2016 sepsis guidelines (Table S5).

Adverse events were prospectively collected during the original SPICE III trial. Due to the un-blinded study design, events were reported by site investigators but not systematically collected in both groups. Bradycardia was defined as heart rate < 55 beats per minute requiring intervention (e.g., pacing, pharmacological support, or modification of dexmedetomidine or other medication doses). Hypotension was defined as hypotension which is clinically significant in the Principal Investigator’s opinion. In our cohort, hypotension was reported in seven (15.9%) patients in the DEX group and one (2.6%) patient in the usual care group (p = 0.04). Bradycardia occurred in five patients (11.4%), all were from the DEX group (p = 0.06). Serious adverse events occurred in two patients in the DEX group (hypotension) and one patient in the usual care group (cardiac arrest, survived) (Table S2).

Our study has several strengths. First, we studied patients in septic shock randomly assigned to one of two treatment strategies (early DEX sedation vs. usual care) and analyzed the hemodynamic effects of dexmedetomidine compared to usual care sedation. Second, we obtained granular and detailed data from two academic ICUs on hemodynamic parameters, level of sedation, dose, and duration of vasoactive and sedative drugs. Third, we used data from the original SPICE III database, electronic patient health records (whenever possible), and double data entry for manually collected data, to minimize the risk of ascertainment bias.

Our study is a retrospective exploratory subgroup analysis, not prespecified in the SPICE III study protocol and, therefore, comes with inevitable limitations. All findings must be interpreted strictly as hypothesis generating. However, due to the existing clinical equipoise regarding benefits and risks of using dexmedetomidine in patients with septic shock, our results have incremental value in adding to our understanding of the hemodynamic effects of dexmedetomidine in such patients. Because the goal of this investigation was to explore the effects of dexmedetomidine on hemodynamic parameters and vasopressor use, we excluded two patients in each group who did not receive the assigned treatment and opted for a per-protocol instead of an intention-to-treat analysis. However, baseline characteristics of patients included were similar, and it is unlikely that including these four patients would have materially altered our results. Hemodynamic management was not dictated by protocol and, as management of vasodilatory shock differs with respect to utilization of steroids and vasopressors [33], we cannot exclude a bias related to the open-label design of the SPICE III study or due to clinical preferences. However, by calculating the NEq/MAP ratio we sought to account for the different types of vasopressors used and for the individual MAP targets prescribed by the treating clinicians; the proportion of patients treated with hydrocortisone was similar in both groups, and we included hydrocortisone in our multivariable adjusted analysis. Finally, DEX was associated with more episodes of bradycardia and hypotension, as already reported in the SPICE trial (11). Thus, although the overall effect of using DEX as the primary sedating agent in mechanically ventilated patients with septic shock appears to improve the vasopressor dose to MAP ratio in the first 48 h, in some patients it may precipitate unwanted hemodynamic changes. Accordingly, caution should be exercised when starting the infusion and in administering this agent in patients at risk of bradycardia.