Introduction

Analgo-sedation for critically ill mechanically ventilated patients provides an opportunity to reduce anxiety, discomfort, ventilator intolerance and dyssynchrony. Appropriate sedation can improve patient cooperation and communication while facilitating clinical examination and procedural tasks [1, 2]. However, it is not entirely a benign entity—existing literature in medical and surgical intensive care units demonstrates that sedative medications can be associated with significant over-sedation and delirium [3]. Drug accumulation in the critically ill can lead to adverse events, including hemodynamic disturbances, respiratory difficulties, prolonged time to extubation, and residual post-traumatic stress following discharge [4, 5].

These challenges faced by the intensivist can often be amplified when taking care of the neurologically ill patient. In the neurocritical care population, there is an accentuated importance on the optimization of cerebral and global hemodynamics including intracranial pressure (ICP), cerebral perfusion pressure (CPP), and mean arterial pressure (MAP). These patient populations often require frequent neurological assessment and management of aggressive delirium, thus emphasizing the need for an agent that allows for convenient and reliable neurological examination and minimal hemodynamic fluctuations [6].

Alpha-2 agonists in particular have become increasingly popular for use in the neurocritical care population due to their proposed effectiveness in facilitating examinations and procedures as well as reducing the need for adjunctive sedative or analgesic agents [7]. Dexmedetomidine, in particular, offers mild analgesic properties and can reduce opioid requirements. Within the medical and surgical intensive care unit (ICU) population, a recent Cochrane review has shown that dexmedetomidine shortens the time to extubation and discharge compared to traditional agents such as propofol [2]. Specific to patients with neurocritical diagnoses, a recent systematic review was conducted by Tsaousi and colleagues on the use of dexmedetomidine for sedation [8]. This review evaluated a primary outcome of sedative effect based on achievement of sedation scores. Hemodynamic adverse events were studied as secondary outcomes and pooled in a meta-analysis and suggested no significant difference in combined hemodynamic effect (hypotension or bradycardia) between dexmedetomidine and controls. The authors acknowledge the use of sedative and hemodynamics adjuncts in various studies, but nonetheless pooled their results. Concurrent and variable use of adjuncts such as dopamine would certainly introduce a significant degree of heterogeneity and confounding bias, thus making it inappropriate to pool results [8]. In addition, the systematic review does not address the cerebral physiologic effects of dexmedetomidine, a consideration that is central to the care of the neurocritically ill patient. It would therefore be premature to make any confident conclusions regarding the hemodynamic safety of dexmedetomidine.

A rigorous evaluation of the current literature with consideration of these issues is needed. The objective of this review is to assess the efficacy and safety of alpha-2 agonists for non-procedural analgo-sedation in mechanically ventilated critically ill adult patients with primary neurological diagnoses.

Methods

This systematic review was conducted based on a previously published protocol [9] and was registered with the international prospective register of systematic reviews (PROSPERO # CRD42016037045). This report has been prepared according to the preferred reporting items for systematic reviews and meta-analysis (PRISMA) Statement [10].

Eligibility Criteria

Studies meeting the following eligibility criteria as structured by the population intervention comparator outcomes (PICOS) study design framework were of interest for this review:

  • Population Neurocritically ill adult patients (age ≥ 18 years) on invasive mechanical ventilation requiring non-procedural analgo-sedation.

  • Intervention Any alpha-2 agonist agent used for non-procedural sedation (enteral or parenteral).

  • Comparator Any analgo-sedative regimen including but not limited to propofol, benzodiazepines, opioids, or ketamine. Anticipating a paucity of literature on our topic, studies without a comparator were also included.

  • Outcomes The primary outcomes of interest were mean and mean difference (before and after administration) in MAP, ICP, and CPP. In studies that included multiple measurements of pressure after commencing treatment, to ensure consistency, we chose the last measurement taken during which treatment was still being administered. Secondary outcomes were included to evaluate the analgesic and sedative properties of dexmedetomidine. These included any measures of cerebral perfusion, analgesic or sedation effectiveness, adverse events (including but not limited to need for rescue therapies, ICP excursions, and hemodynamic variability), duration of mechanical ventilation, incidence of delirium (confusion assessment method–ICU [11]), quality of sedation (Richmond Agitation–Sedation scale [RASS] [12] or COMFORT score [13]), ICU, and length of stay. Narrative descriptions for measures of cerebral perfusion, analgesic and sedative effectiveness, as well as hemodynamic adverse events were not included as outcomes in our previously published protocol and were added following peer review. Studies not reporting our primary outcomes were still included if secondary outcome data were available.

  • Study design The study design consists of randomized controlled trials, quasi-randomized trials, and retrospective or prospective cohort studies. In-progress studies identified from the clinicaltrials.gov registry and CENTRAL Cochrane were included in a qualitative analysis.

Information Sources and Search Strategy

MEDLINE and Embase were searched from January 1, 1946, to May 30, 2016, using a pre-defined search strategy developed under the guidance of a health information specialist with expertise in clinical research and peer-reviewed by a clinical expert (SE). This search strategy was developed using a combination of subject headings and keywords derived from the terms “neurocritical care” and “alpha-2 agonists,” and is available in our protocol [9]. There were no date or language restrictions.

We hand-searched the reference lists of all primary studies and identified systematic reviews, CENTRAL Cochrane Library, as well as the last three years of conference proceedings for the European Society of Intensive Care Medicine Congress, Neurocritical Care Society Annual Meeting, and Critical Care Canada Forum. We also searched the clinicaltrials.gov registry for in-progress studies of relevance.

Study Selection and Data Collection

Two authors (AT and HB) participated in abstract and full-text screening independently and in duplicate. The study selection process is summarized in a PRISMA flow diagram (see Fig. 1). A pre-designed data extraction form was completed independently and in duplicate. Data items include publication and patient characteristics, intervention and comparator characteristics, outcome data, and study methods to assess risk of bias.

Fig. 1
figure 1

Study selection process

Full size image

Risk of Bias Assessment

Risk of bias for randomized controlled trials was assessed using the Cochrane Risk of Bias Tool [14]. The risk of bias for observational studies was assessed using the Newcastle–Ottawa quality assessment Scale [15]. No studies were excluded based upon their assessed level of risk of bias. Assessments were used strictly to provide a narrative summary.

Data Synthesis

Before formal synthesis, patient and study characteristics were reviewed by the authors to determine clinical heterogeneity to assess suitability for meta-analysis. Results from individual studies were presented as means and mean differences with 95% CIs for continuous outcomes, and event rates for dichotomous outcomes. Due to lack of clinical homogeneity of studies, pooling was deemed inappropriate for this review.

Results

Study Selection

Our search yielded 2398 unique records after removal of duplicates, of which 17 studies were included in the final data synthesis [16–32] (see Fig. 1). Twelve of these studies were full-text articles [16–27], and five studies were published conference abstracts [28–32]. Reasons for exclusion are provided in the flow diagram. A list of the studies reviewed at the full-text screening stage, categorized by inclusion and exclusion status, is found in Online Appendix 1.

Study Characteristics

Types of Study Designs

There were four randomized controlled trials [20, 24, 27, 29] and two randomized crossover studies [21, 22] comparing the use of dexmedetomidine to comparator arms of propofol [20–22, 24], an unspecified placebo [29], normal saline [27], or midazolam [24] (see Table 1). There were four prospective cohort studies evaluating the use of dexmedetomidine as a sedative [18, 19, 23, 25], with only one study including a comparator group of propofol [23]. Five of the studies were retrospective cohorts [16, 17, 26, 30, 32], of which two included propofol as a comparator group [17, 26]. The median sample size was 40 (range 3–190), and the median year of publication was 2013 (range 2002–2015).

Table 1 Study characteristics
Full size table

Types of Participants

This review included a total of 1110 participants (see Table 1). From studies with available subgroup data, 534 participants were in the alpha-2 agonist group and received dexmedetomidine, 286 participants received propofol, 30 participants received midazolam, 75 patients received saline, and 25 patients received a placebo of unknown type. Subgroup data for number of patients receiving dexmedetomidine and propofol in the Pajoumand study were not available as data were split up by drug use on any given day rather than by patient [23]. Mean age for the study populations ranged from 33 to 65 years; however, two studies did not report information on age [21, 29]. All studies were composed of entirely critical care patients, only two of which included patients with non-neurologic diagnoses [22, 24].

Types of Interventions

We found no studies evaluating clonidine for use in non-procedural analgo-sedation (see Table 2). Dexmedetomidine was the only alpha-2 agonist evaluated in included studies. Comparative arms evaluated were propofol [17, 20–24, 26, 31], midazolam [24], normal saline [27], and a placebo of unknown type [29]. All sedatives in this review were administered intravenously.

Table 2 Hemodynamics
Full size table

Of 13 studies which reported on dosing, bolus doses were administered in five of the studies [16, 20, 24, 25, 28]. Dexmedetomidine was always evaluated using a continuous sedation protocol with infusion rates ranging from 0.2 to 1.4 mcg/kg/hr.

Risk of Bias in Included Studies

Bias for randomized studies

Six studies were evaluated for risk of bias using the Cochrane Risk of Bias Tool (Online Appendix 2) [20–22, 24, 27, 29]. The risk of selection bias was generally unclear in the randomized trials, as many of the studies were abstracts with limited information. Three of the studies reported randomization of their patients, but did not discuss randomization strategy or allocation concealment [21, 22, 29]. Two studies had low risk of bias for sequence generation [24, 27], but only one of those studies reported concealing of the allocation [27]. Finally, one study was deemed to have high risk of selection bias for both random sequence generation and allocation concealment [20]. Performance bias was identified as unclear in one abstract with limited information [29], and as high in two studies where unblinded medical professionals were responsible for altering the sedation levels of the drugs in order to maintain the desired level of sedation [21, 24]. Detection and attrition bias were low in all studies with the exception of one abstract with limited information, which was identified as unclear [29]. Reporting bias was mostly low (3 of 6 studies), with the exception of two abstracts with limited information, and one study with high risk of selective reporting when published outcomes were compared to the clinicaltrials.gov protocol [22].

Bias for Observational Studies

Eleven studies were evaluated using the NOS [16–19, 23, 25, 26, 28, 30–32]. Overall, quality assessment varied considerably, ranging from 4 to 9 stars out of a possible total of nine (see Online Appendix 3). All studies were judged to be somewhat or truly representative of the exposed cohort. All studies identified that the outcome of interest was not present at the start. Two out of the four studies with comparator groups received two out of two stars for comparability [17, 31]; the other two studies did not include matching in the design or adjustment for confounders [23, 26]. The majority of full-text articles scored a full three out of three stars for outcomes with the exception of Pajoumand et al., who did not report on their method of outcome assessment [23]. All studies had follow-ups of sufficient length of time, and no patients were lost to follow-up.

In-Progress Studies

From the clinicaltrials.gov registry, we identified two studies with ongoing recruitment [Lacarin, NCT02252523 & Takala, NCT01664520] and two studies that were withdrawn or terminated [Stein, NCT01007773 & Keegan, NCT0156559]. Further details on study design and primary outcome are given in Online Appendix 4.

Data Synthesis

Few studies reported on our outcomes of interest. As demonstrated in Tables 2 and 3, there was a great deal of clinical heterogeneity across studies related to study drug administration as well as administration and reporting of sedative and blood pressure adjuncts by study. Therefore, pooling of results was judged inappropriate for any of the pre-specified outcomes, and narrative summaries of study findings were prepared.

Table 3 Sedation and analgesia
Full size table

Effects of Interventions

Primary and secondary outcomes are described in Tables 2 and 3. Ten of the studies did not report on any of our primary outcomes of interest [16, 17, 19, 22, 23, 26–28, 30, 32].

Mean Intracranial Pressure

Three studies reported on ICP [18, 21, 29]. Grille et al. evaluated the cerebral hemodynamic changes during dexmedetomidine infusion without a comparator group in 12 patients with severe neurologic injury (Glasgow Coma Scale [GCS] ≤ 8) [18]. The authors found mean ICP to be 16 mmHg, with an increase in ICP of 2 mmHg 3 h after drug administration. Dexmedetomidine was further evaluated against propofol in a randomized crossover study of patients with severe brain injury (GCS ≤ 8) by James et al., who found no significant difference in mean ICP between the groups during active drug infusion (11.6 vs. 14.3 mmHg, P = 0.240) [21]. Additionally, the authors did not find any difference in mean ICP during breaks in sedation for neurological examination (12.3 vs. 13.0 mmHg, P = 0.745) or during the first hour of resuming the infusion following examination (12.6 vs. 10.9 mmHg, P = 0.448) [21]. In their randomized study of 59 patients with acute brain injury requiring ICP monitoring, Hendry and colleagues did find a significant difference in ICP when comparing dexmedetomidine to a placebo (11.5 vs. 12.3 mmHg, P = 0.024) [29]. We note that these were listed as CPP values in the identified published abstract, but were clarified with the study authors to represent ICP (direct author correspondence).

Mean Cerebral Perfusion Pressure

Two studies reported on CPP, with varying results [18, 21]. Grille et al. evaluated dexmedetomidine without a comparator and reported a mean CPP of 87 mmHg, with a change of −4 mmHg from baseline [18]. James et al. evaluated dexmedetomidine against propofol; they found no significant difference in mean CPP when compared to propofol CPP (80.1 vs. 81.6 mmHg, P = 0.828) during active drug infusion, during breaks in sedation for neurological examination (85.5 vs. 77.2 mmHg, P = 0.184) or during the first hour of resuming the infusion (79.4 vs. 78.6, P = 0.869) [21].

Mean Arterial Pressure

MAP was reported in six studies and ranged from 64.7 to 104 mmHg in the dexmedetomidine group [18, 20, 24, 25, 29, 31]. Four studies additionally reported change in MAP from baseline to intervention, with results ranging from −2 to −12.5 mmHg [18, 20, 24, 25].

Several studies compared the effects of dexmedetomidine on MAP to alternative therapies, with mixed results. Hao et al. found no significant difference in mean MAP between dexmedetomidine and propofol (79.9 vs 88.1 mmHg) [20]. Conversely, both Srivastava and Hendry found significant differences in mean MAP when evaluated against propofol (98.1 vs. 92.2 mmHg, P < 0.001 and 88.4 vs. 94.4 mmHg, P = 0.023, respectively) [24, 29], though Srivastava did not find differences when evaluated against midazolam (98.1 vs 98.3 mmHg, P = 0.937) [24]. While Tarabrin et al. also compared mean MAP in dexmedetomidine to propofol (64.7 vs. 71.0 mmHg), they did not report on the significance of their findings [31]. Only two studies reported on the concurrent use of hemodynamic adjuncts (dopamine) [24, 25].

Additional Measures of Cerebral Perfusion

Three studies described additional measures of cerebral perfusion. Grille et al. found no difference in mixed venous oxygenation or cerebral extraction of oxygen for patients during dexmedetomidine infusion [18]. Wang et al. also noted no difference in cerebral blood flow, cerebral metabolic rate equivalent, jugular venous oxygen saturation, or oxygen extraction ratio [25]. James et al. demonstrated no difference in brain tissue oxygenation or lactate-to-pyruvate ratio when compared to propofol [21].

Effectiveness of Analgesia

Six studies reported on analgesic effectiveness. Five of these studies noted that dexmedetomidine was associated with a reduction in fentanyl or morphine requirements [19, 20, 24, 27, 32], while one study did not find any significant difference [22].

Effectiveness of Sedation

Sedation quality using the RASS was reported by three studies [19, 21, 22]. Target RASS scores ranged from −3 to +3, varying both by study and by patient, as some studies set individual target RASS scores. The percent of participants who reached the target RASS score was reported to be 100% in both the intervention and control groups of two of the studies [21, 22] and 67% in one study without a comparator [19]. Two studies noted 84 and 98% sedation target achievement using Riker Sedation–Agitation Scale as compared to 81 and 93% for alternative agents [20, 27]. Studies by James and Srivastava demonstrated no difference in sedation between dexmedetomidine and propofol as assessed by the Ramsay Sedation Scale or Bispectral Index [21, 24], respectively. In their study evaluating cognitive retention, Mirski et al. note that dexmedetomidine had significantly better Adapted Cognitive Exam scores than propofol. No studies reported on incidence of delirium. One study by Mirski et al. reported on one patient with incident delirium, but the authors did not specify whether this patient was in the intervention or comparator group [22]. Further details are found in Table 3.

Adverse Events

Adverse events were described in 13 of the studies. Adverse events identified included hypotension (6–46% of patients) in eight studies [16, 17, 23, 24, 26, 27, 30, 32], hypertension (0–56%) in four studies [16, 23, 26, 27], bradycardia (5–41%) in five studies [17, 26, 27, 30, 32], tachycardia (3–26%) in two studies [23, 27], and respiratory events (2–5%) in two studies [16, 26]. However, the definitions of adverse events were not always presented and varied in those that did. Where presented, specific definitions of adverse events varied by study from simply haemodynamic changes requiring interventions [22, 24], to exact cutoffs for bradycardia (HR < 40–60 bpm, depending on study) [16, 17, 23, 26, 27], tachycardia (HR > 120) [23, 27], hypotension (SBP < 90 or MAP < 65) [17, 23, 26, 27], and hypertension (SBP > 160–180 mmHg, depending on study) [23, 26, 27].

Studies by Erdman, Mirski, Pajoumand, and Srivastava found no significant differences in adverse events between dexmedetomidine and propofol [17, 22–24], or between dexmedetomidine and midazolam [24]. Yokota et al. described the adverse events between the dexmedetomidine and propofol groups as comparable, with the exception of 20% of patients developing bradycardia in the dexmedetomidine group compared to 0% in the propofol group [26]. In the Zhao study comparing dexmedetomidine to normal saline, it was found that dexmedetomidine was associated with higher rates of bradycardia (5.3 vs. 0%, P = 0.043), but with lower rates of tachycardia (2.7 vs. 18.7%, P = 0.002) [27]. Studies by Yokota and Tarabrin noted that heart rate was significantly slower in the dexmedetomidine group in comparison with propofol [26, 31]. Two studies noted that dexmedetomidine was associated with lower mannitol requirements [17, 30], but there was no significant difference in need for total rescue therapies [30], ICP variability [29], and incidence of ICP excursions [30] when compared to alternative agents.

Duration of Mechanical Ventilation

One study by Srivastava et al. looked at duration of mechanical ventilation and found no significant difference in length between dexmedetomidine, propofol, or midazolam groups (12.0 vs. 12.9 vs. 12.7, P = 0.601) [24].

ICU Length of Stay

ICU length of stay was examined in three of the studies, with results ranging from 1.2 to 31 days [18, 20, 27]. One study by Hao et al. included a comparator group, but did not report on whether the length of stay was significantly different between dexmedetomidine and propofol [20].

Discussion

This systematic review rigorously identified and synthesized the existing data on the use of alpha-2 agonists on mechanically ventilated brain-injured patients according to an a priori defined, previously published protocol. The relatively few eligible studies identified, which reported on several different measures and outcomes, highlight the paucity of existing literature regarding the safe and effective use of alpha-2 agonists in this population. Recognizing the limited data available, dexmedetomidine does not appear to result in severe, uncompensated hemodynamic disturbances (cerebral or systemic). The distinction of “uncompensated” changes is important as while many of the studies did not find significant differences between the hemodynamics effects of dexmedetomidine and comparators, we could not ascertain the degree (if any) of pharmacologic compensation required from blood pressure adjuncts such as vasopressors. Sedation target levels were generally achieved while reducing requirements for adjunctive sedative and analgesic agents, tentatively suggesting the efficacy of dexmedetomidine for analgo-sedation in this population. In addition, the findings of improved cognitive retention by Mirski et al. suggest the potential for valuable secondary properties of dexmedetomidine with regard to patient cognition, cooperation, and facilitation of neurological examination [22]. As noted earlier, the neurologically ill patient presents significant challenges for management of aggression, agitation, and delirium. Unfortunately, no studies described or compared rates of delirium for patients receiving dexmedetomidine. Considering its well-known properties for docile, analgo-sedation in the general critical care population, confirmation of such benefits in this population would be of great value and warrants further study.

The safety of traditional sedative agents in patients with critical brain injury continues to come under scrutiny. The recently released fourth edition guidelines for the management of severe traumatic brain injury [33] reviewed the existing evidence for the safety of propofol in patients with severe traumatic brain injury and noted significant concerns regarding its use. These guidelines therefore advocate for “extreme caution” when using high-dose infusions or exceeding a total usage of 48 h. However, there was no comment or recommendation with regard to alpha-2 agonists as the evidence for its use in this population, as noted in this review, continues to be lacking. Similarly, the most recent clinical practice guidelines for sedation in critically ill patients acknowledge that several studies demonstrate the safety of dexmedetomidine infusions for long-term sedation up to 28 days in the general critical care population [34]. Citing several advantages including ease of rousing, improved interaction, minimal respiratory depression and reduction in opioid requirements, the guidelines recommend the preferred use of dexmedetomidine infusions over benzodiazepine infusions for patients at risk of suffering from delirium. The growing popularity and emerging body of evidence for dexmedetomidine certainly present a convincing case for more rigorous study of its potential application in the neurocritical subpopulation.

To best understand the existing evidence for dexmedetomidine use in the neurocritical population, we adopted a practical and inclusive protocol design for this review to maximize the available data while also maintaining a cautious approach to its interpretation and application. We note a few key limitations to consider. To begin with, only seven studies reported on our primary outcomes of interest and only four studies identified in this review evaluated hemodynamics as a primary outcome. It is therefore not surprising that important contextual variables such as dosing, sedation protocols, hemodynamic targets, use of hemodynamic, and sedation adjuncts were frequently unreported. Isolated numerical reporting of ICP, CPP, and MAP without the appropriate clinical context makes interpretation quite challenging and generalization to other contexts even more so. With regard to the secondary outcomes, we note similar limitations with regard to lack of reporting on the use and dosing of sedative or analgesic adjuncts. As advocated by the Society of Critical Care guidelines, we highlight that analgesia-first sedation is preferred in the mechanically ventilated population due to the high incidence of pain as a cause of agitation [34].

From these review findings, we note that the presence or significance of required co-interventions to maintain these hemodynamic parameters is unclear and is a limitation for interpretation and generalization. Understanding the safety and efficacy of alpha-2 agonists for sedation in the neurocritical population with the primary intention of evaluating their effect on patient hemodynamics is essential and a clear area in need of high-quality research. Clear reporting of dosing strategy, sedation protocol use, co-interventions administered, and a priori defined adverse events are recommended. Given the challenging hemodynamic sensitivities and distinctive sedation needs of the neurocritical care patient, the validation of an effective and safe agent would prove invaluable in the future care of this unique population.