Clinical effectiveness of a sedation protocol minimizing benzodiazepine infusions and favoring early dexmedetomidine: a before-after study

The results of the present study demonstrate that implementation of a protocol that reduces benzodiazepine use and targets light sedation can lead to significant improvements in the time spent at moderate to deep levels of sedation, the duration of mechanical ventilation, and the requirement for tracheostomy. These benefits were observed despite increases in the prevalence and duration of ICU delirium. Although the primary change between phases was to favor earlier and more frequent use of dexmedetomidine, it is important to note that the sedation protocol allowed use of other sedatives according to the ICU team’s discretion and incorporated other key elements from recent pain, agitation and delirium guidelines, such as goal-directed analgesia therapy, multidisciplinary involvement, use of validated tools for routine assessment of sedation and delirium, and daily spontaneous breathing trials. This study therefore represents a valuable assessment of the clinical effectiveness of modifying sedation practices in a real-world setting.

The observed shift in sedative use indicates that the protocol was effective in reducing midazolam use in favor of early dexmedetomidine as the primary change in sedation-related practices between phases. In parallel with the changes in sedative use, a reduction in sedative assessments at moderate sedation was observed, as well as a small increase in the percentage of scores above goal. Although there is additional room for improvement, this represents a positive step toward the goal of minimizing time spent at moderate to deep levels of sedation. The occurrences of self-extubation and need for reintubation were infrequent and not statistically different between groups; however, it should be noted that both were increased in the after phase, and this study was not sufficiently powered to assess these differences. Interestingly, although more than 90% of patients in both groups received a fentanyl infusion for pain, the total dose was significantly reduced in the after phase. The most likely explanation is a reduced duration of fentanyl infusion secondary to reduced time on mechanical ventilation, although it is possible that the analgesic properties of dexmedetomidine or perhaps an improved ability to assess pain in the after phase could also account for some of this difference. These results are consistent with early randomized trials of dexmedetomidine in surgical patients, which demonstrated reduced opioid doses [8,9]. However, more recent studies examining long-term dexmedetomidine use in mixed medical-surgical ICU populations have shown either no difference [4,6] or increased opioid needs compared with benzodiazepines [3]. In the latter study, the authors found that the increased opioid needs were primarily in patients targeted for deep sedation, and they proposed that this likely reflected clinicians’ use of fentanyl to achieve deeper sedation.

With respect to potential adverse effects, increased occurrences of hypotension or bradycardia were not observed between groups, and the percentage of patients requiring the addition of a vasopressor while on sedative therapy was actually lower in the after phase. The majority of patients requiring initiation or an increased dose of vasopressor were receiving midazolam at the time of the potential adverse effect. In part, this may have been a result of clinicians choosing to defer the use of propofol and/or dexmedetomidine while patients remained hypotensive. Bradycardia (HR <50 bpm) more commonly occurred while patients were receiving propofol or dexmedetomidine. Dexmedetomidine failure occurred in one in four patients, with the most common reason being hypotension. Although the definitions of hypotension and bradycardia vary among studies, both toxicities have been well described in randomized trials [3-6]. Taken together, these findings suggest that clinicians observed hypotension or bradycardia early and intervened before it reached a concerning level. Failure to achieve RASS goal occurred in 7% of patients, which is slightly lower than the rates of dexmedetomidine discontinuation due to lack of efficacy in the Dexmedetomidine versus Midazolam for Continuous Sedation in the Intensive Care Unit (MIDEX; ClinicalTrials.gov Identifier: NCT00481312) (9%) and Dexmedetomidine Versus Propofol for Continuous Sedation in the Intensive Care Unit (PRODEX; ClinicalTrials.gov Identifier: NCT00479661) (14%) trials [6].

Researchers in multiple studies have reported significant improvements in patient outcomes when sedation is effectively minimized by using various strategies [10-15]. Compared with benzodiazepines, dexmedetomidine has also been shown to improve clinical outcomes, including reduced duration of mechanical ventilation [4,6]. In the present study, a nursing staff-directed sedation protocol targeting light sedation was the standard of care. Daily spontaneous breathing trails were also routinely performed in both the SICU and MICU, and interruption of sedatives was combined with this in the SICU. Similarly to the aforementioned studies comparing continuous infusions of dexmedetomidine and midazolam, we found a significant reduction in time on the ventilator. This finding supports the clinical effectiveness of the protocol and is striking, considering that 27% of patients in the after phase still received midazolam as the initial sedative, which would be expected to reduce between-group differences. One important distinction in our study is that dexmedetomidine was used up front in 39% of patients in the after phase, whereas in randomized controlled trials of dexmedetomidine, the study drug was not initiated until 22 to 41 hours after initiation of mechanical ventilation and prior sedative therapy [3,4]. Delayed initiation of dexmedetomidine in these trials may reduce possible benefits because of residual effects of previous sedatives, particularly prior benzodiazepines. A statistically significant reduction in ICU length of stay was not found, although a trend in favor of the after phase can be appreciated. Individual studies of dexmedetomidine also have not shown a significant reduction in ICU length of stay [3-6], although the authors of one meta-analysis reported this finding [16].

With regard to our hypothesis that reduced midazolam use would lead to reduced prevalence and duration of delirium, we found the opposite. However, this is not entirely inconsistent with the available evidence. In one large study comparing continuous infusions of dexmedetomidine and midazolam, a significant reduction in the prevalence and duration of delirium was observed [4]. When compared with continuous lorazepam, the prevalence of delirium or coma as a combined endpoint was reduced; however, when the individual endpoints were assessed, this difference was found to be primarily a result of reduced coma [3]. In a subsequent subgroup analysis of the same study, the daily risk of delirium was found to be significantly lower in the dexmedetomidine versus the lorazepam group [17]. In a randomized pilot study, Ruokonen and colleagues [5] compared midazolam or propofol to dexmedetomidine and found that the combined endpoint of CAM-ICU and adverse events of delirium and confusion were increased in the dexmedetomidine group (44% versus 25%, P =0.035), whereas the proportion of positive CAM-ICU assessments was not different between groups. Finally, in the MIDEX study, delirium was not routinely assessed with a validated tool while patients were receiving sedative therapy, although the trialists reported no difference in a combined endpoint of neurocognitive adverse events (anxiety, agitation or delirium) between the dexmedetomidine and midazolam groups [6].

Similarly to earlier studies including comparable patients [18-22], we found that the overall prevalence of ICU delirium was very high in the present study. Recent studies shed greater light into the diagnosis of ICU delirium as it relates to both the level of sedation and its potential impact on patient outcomes [22,23]. Patel and colleagues [22] performed a prospective cohort study of 102 adult MICU patients requiring mechanical ventilation and continuous infusion of sedatives in which they performed CAM-ICU assessments before and for up to 2 hours after a daily interruption of sedatives and analgesics. The prevalence of delirium was 89% when assessed prior to daily interruption, as opposed to 77% when assessed after interruption. Patients with rapidly reversible, sedation-related delirium had significantly fewer ventilator days and hospital days than did patients who had delirium that persisted after 2 hours. Given that all patients in our present study required continuous sedatives for at least 24 hours and that CAM-ICU assessments were not performed during an interruption of sedative therapy, it is likely that a percentage of this delirium was sedation-related and potentially rapidly reversible, which would explain our findings of reduced duration of mechanical ventilation despite increased delirium. Similarly, Haenggi and colleagues [23] found that the prevalence of delirium (using both CAM-ICU and Intensive Care Delirium Screening Checklist) was reduced by 22% when patients with a RASS of −2 or −3 were excluded, suggesting that the depth of sedation can have a significant influence on the apparent prevalence of delirium. Notably, we included all assessments for patients at a RASS of −3 and higher, consistent with the practice in which the CAM-ICU was originally validated. Despite the reduction in the median percentage of scores at a RASS of −3, the prevalence of delirium was still increased in the after phase. Accordingly, it would seem unlikely that this is an explanation for the increased delirium we observed in the after phase. However, though we observed a reduction in the percentage of scores at moderate sedation, this does not mean that the level of sedation was lower at the time CAM-ICU assessments were performed, as sedation was assessed much more frequently than delirium, which was assessed only twice daily at fixed time points. In fact, the number of CAM-ICU evaluations deemed unassessable was higher in the after phase, suggesting a higher frequency of deep sedation at the time of CAM-ICU assessments. It may thus still be possible that the level of sedation may have influenced the CAM-ICU assessments and the prevalence of delirium. Finally, because use of any combination of all three sedatives was allowed in both phases, it is not possible in the present before-after open-label trial to truly discern the relative contribution of each sedative to the observed prevalence of delirium.

This study has several important limitations. First, rather than one discrete change in drug therapy, this study assesses the impact of implementing a sedation protocol that merely favors the early use of dexmedetomidine to minimize use of midazolam infusions. Use of alternative sedatives was still allowed in both phases. Additionally, we provided reeducation of sedation-related principles to facilitate implementation of the protocol. Therefore, it is difficult to establish that the observed differences are due solely to the change in sedative therapy. However, this approach provides useful insight into the clinical effectiveness of such a change in an ICU clinical practice, where the use of multiple sedatives is often necessary.

Second, pain was assessed using an institution-specific scale for non-communicative patients that has not been validated, and pain assessments were not collected. However, frequent assessments of pain and appropriate management were emphasized in both ICUs, and patients appeared to be treated aggressively, with over 90% of patients in both phases receiving continuous infusions of fentanyl. Third, although all nurses were required to complete an online education module and one-to-one follow-up by clinical nurse specialists dedicated to the SICU and MICU was provided, CAM-ICU assessments were performed by the bedside nurse and were not validated by other practitioners. Therefore, it is possible that some inaccuracy exists related to interrater reliability and changes in assessments and scoring that may have occurred as the study progressed. Nevertheless, this is representative of how delirium is commonly assessed in clinical practice and thus makes our results more generalizable to typical ICU environments.

With respect to the assessment of sedative failure, we chose to focus on the potential failure of dexmedetomidine because (1) a main goal of the protocol was to increase dexmedetomidine use and (2) less is known about tolerance in a real-world setting. Although we report potential hypotension and bradycardia events for all drugs, we did not assess for midazolam and propofol failure in the same way. Finally, this was a before-after study and was not blinded or randomized, and thus a risk of bias cannot be eliminated.