Can dexmedetomidine be a safe and efficacious sedative agent in post-cardiac surgery patients? a meta-analysis

Two researchers independently carried out a comprehensive literature search. The literature search was conducted in January 2012 using multiple databases including EMBASE, MEDLINE, the Cochrane Library and Science Citation Index, from January 1979 through January 2012. A basic search was performed using keywords: 'dexmedetomidine' AND ('cardiac surgery' OR 'coronary artery bypass grafting' OR 'heart surgery' OR 'heart valve' OR 'cardiopulmonary bypass') AND ('sedation' OR 'sedative'). In addition, we reviewed abstracts from selected major cardiac surgical scientific meetings (American Heart Association, American Association for Thoracic Surgery, European Association for Cardio-Thoracic Surgery, and Asian Society for Cardiovascular Surgery) for unpublished studies, and contacted the authors for detailed information if needed. Our searches were restricted to English language studies for convenience reasons. To be eligible for inclusion in this article, publications met the following four inclusion criteria: (1) original research comparing dexmedetomidine with a placebo or an alternative sedative agent in elective cardiac surgery patients aged over 18 years; (2) study design: randomized controlled trial (RCT), non-randomized controlled trial or cohort study; (3) studies that continued use of dexmedetomidine for postoperative sedation for more than 6 hours, and not used for anesthesia in the operating theater, and (4) availability of full text (detailed information).

The two reviewers who extracted the data were blinded to the journal names and authors. If there were any differences in data abstraction or quality assessment, then the differences were reconciled by a third reviewer. The following information was abstracted and tabulated from each paper: author and year of publication; design; number of patients; surgery type; sedation goal and dose of dexmedetomidine, and placebo or alternative sedative agent. The following outcomes were extracted if reported: duration of mechanical ventilation; length of ICU stay; morphine equivalents; length of hospital stay; mortality at hospital discharge; risk of bradycardia or hypotension requiring interventions; risk of delirium; ventricular tachycardia; atrial fibrillation; hyperglycemia; nausea and vomiting; reintubation within 5 days after extubation, and any postoperative infection. If there was incomplete reporting of clinical outcomes in any of the articles, we attempted to contact the authors to obtain additional information. For example, given that virtually all patients were intubated for surgery, not all studies reported the mean value and standard deviation of the length of mechanical ventilation. We contacted authors [11‐13] for additional detail on the outcomes mentioned above; however, no additional information was added. A ratio of relative risks and difference between treatment agents for categorical and continuous outcomes respectively, were extracted from all publications (if presented). Categorical outcomes are reported as risk ratio (RR) and corresponding 95% confidence interval (95% CI), while continuous outcomes are reported as weighted mean difference (MD). A P-value < 0.05 was used to determine statistical significance.

Quality assessment was undertaken independently by two authors (Table 1). The quality of included studies was evaluated based on a well established, validated Newcastle-Ottawa Scale (http://​www.​ohri.​ca/​programs/​clinical_​epidemiology/​oxford.​htm). Differences were resolved by discussion and consensus, and if disagreement still persisted, the opinions of all members of the research team were sought.

Where outcomes of interest were reported by two or more studies, effect estimates were combined with meta-analyses in Review Manager (Version 5.1., The Cochrane Collaboration, 2011). Statistical heterogeneity between trials was evaluated by the Cochran χ² statistic, and was considered to be significant when the P-value for heterogeneity was ≤ 0.1. When statistical heterogeneity was observed, a random-effect model was used for the analysis. In the absence of statistically significant heterogeneity (P-value for heterogeneity > 0.1), only the fixed-effect model was utilized. Potential publication bias or small study bias was examined by visual inspection of constructed funnel plots.

To further test the robustness of the results, several sensitivity analyses were performed a priori. First, we evaluated whether the model of the statistical method (random-effect vs. fixed-effect model) would change the results; second, we determined whether the quality of publication, high quality (RCT) or low quality studies (retrospective cohort), could influence the results of the meta-analysis. Moreover, subgroup analysis was performed according to different criteria (for example, different sedative agents, high and low dose of dexmedetomidine). Where data were not presented in a way that could be included in the meta-analysis, or where only one study was identified for a given outcome, results of individual studies were presented.

A total of 530 studies were retrieved in the literature search, including 233 articles in MEDLINE, 146 articles in EMBASE, 3 articles in the Cochrane Library, and 148 articles in the Science Citation Index. After a check for duplicates and removal of reviews, 257 publications remained eligible for inclusion in the meta-analysis (Figure 1). Of these, the abstracts were screened for dexmedetomidine, sedation and cardiac surgery, and 204 publications were removed for non-cardiac surgery. Among the remaining 53 eligible studies, 13 publications were removed because the studies did not report on dexmedetomidine and postoperative sedation. Eleven papers were excluded because they included infant or pediatric cardiac surgery, six papers reported only the outcomes of dexmedetomidine and did not compare outcomes with a placebo or an alternative sedative agent, five papers were excluded because they reviewed the use of dexmedetomidine in the ICU but failed to report any outcomes, five were excluded for use of anesthesia in the operating theater, and two were excluded for not being published in English. Thus, 11 papers were included in the meta-analysis.

Table 2 presents details of the included studies. Study designs included three randomized control trials, four prospective cohort studies and four retrospective studies. The smallest study contained 28 cardiac surgery patients (Aziz, et al.), whereas the largest study included 10,352 patients (Dasta, et al.). Eight studies compared dexmedetomidine with alternative hypnotic agents (seven studies with propofol [12‐18], three with midazolam [16, 17, 19], and one with lorazepam [17]). Two studies compared dexmedetomidine with morphine [11, 20]. The remaining study compared three drugs (dexmedetomidine, and propofol plus midazolam) with two drugs (propofol plus midazolam) [21].

Meta-analysis of nine studies [11, 14‐21] revealed that dexmedetomidine significantly reduced the length of mechanical ventilation (MD -2.70, 95% CI -5.05, -0.35, P = 0.02) (Figure 2A). Our results found dexmedetomidine treatment did not appear to reduce the length of ICU stay (MD -3.44, 95% CI -11.40, 4.52, P = 0.40) [15, 16, 18, 20], length of hospital stay (MD -0.28, 95% CI -0.64, 0.07, P = 0.36) [15‐17, 20, 21], or morphine equivalents (MD 0.45, 95% CI -1.86, 2.77, P = 0.70) [13, 15, 16, 18] compared with other sedatives (Additional File 1). There was significant heterogeneity between the pooled studies in the length of mechanical ventilation, length of ICU stay and morphine equivalents. Thus, a random-effect model was used for these three analyses.

When pooled, dexmedetomidine was found to significantly increase the risk of bradycardia (RR 2.08, 95% CI 1.16, 3.74, P = 0.01) (Figure 2B), but not hypotension (RR 1.06, 95% CI 0.72, 1.56, P = 0.60) (Figure 2C). Additionally, dexmedetomidine reduced the incidence of delirium following cardiac surgery (RR 0.36, 95% CI 0.21, 0.64, P = 0.0004) (Figure 3A). Sedation with dexmedetomidine was associated with a lower risk of ventricular tachycardia (RR 0.27, 95% CI 0.08, 0.97, P = 0.04) (Figure 3B) and hyperglycemia (RR 0.78, 95% CI 0.61, 0.99, P = 0.04) (Figure 3C).

Dexmedetomidine was not associated with a significant reduction of atrial fibrillation (RR 0.90, 95% CI 0.62, 1.29, P = 0.56) [14, 18, 20], postoperative nausea and vomiting (RR 1.02, 95% CI 0.72, 1.46, P = 0.91) [11, 14, 20] (Additional File 2). Furthermore, there was no effect of dexmedetomidine on reintubation (RR 1.62, 95% CI 0.51, 5.13, P = 0.41) [12, 15, 20], postoperative infection (RR 0.92, 95% CI 0.65, 1.29, P = 0.62) [14, 18, 20] or hospital mortality (RR 0.89, 95% CI 0.38, 2.12, P = 0.08) [15, 17, 18, 20, 21] (Additional File 3). As to cardiovascular complications, Herr et al. [14] found no difference in the incidence of myocardial infarction (P = 0.371) and cardiac failure (P = 0.723) between dexmedetomidine and propofol, and Yapici et al. [19] and Shehabi et al. [20] reported similar incidence of postoperative low output syndrome (P = 0.093) and cardiac arrest (P = 0.513), respectively.

Using the length of mechanical ventilation as an endpoint, the funnel plot implies the possibility of publication bias (Figure 4). This bias could be explained by the following reasons: Barletta et al. [12] adopted a fast-track recovery model, which was different from other included articles, and it just needed shorter duration of mechanical ventilation; Yapici et al. [19] enrolled patients who had failed extubation and this caused longer-term ventilation support. The sensitivity analysis shows that regardless of which effect model was applied, the outcomes remained similar (Table 3). Further, we excluded the four retrospective studies [12, 13, 17, 21] and Yapici et al.[19], which included the patients already presenting in a delirium state. Table 4 shows that after excluding the five above-mentioned studies, the outcomes still shared similarities with the outcomes when all eleven studies were included.