Prevention of post-operative delirium using an overnight infusion of dexmedetomidine in patients undergoing cardiac surgery: a pragmatic, randomized, double-blind, placebo-controlled trial

We conducted a pragmatic, randomized, double-blind, parallel group, placebo-controlled trial. Nine centres in France participated to the study. The study protocol and statistical analysis plan have been extensively described and previously published [20].

Screening for eligibility was performed prior to a planned consultation with an anaesthesiologist a few weeks before surgery. All eligible patients were asked for consent and included in the study. The inclusion criteria were patient aged 65 years or older who underwent cardiac surgery with or without cardiopulmonary bypass. Patients were excluded if they met the following criteria: documented cognitive failure or dementia, patients previously included in a study on sedation or analgesia, predicted length of stay in the ICU < 24 h, alpha-2 agonists allergy or intolerance, emergency surgery for immediate life threatening situation, uncontrolled hypotension, 2nd or 3rd grade atrio ventricular block in the absence of a pacemaker, hepatocellular insufficiency defined by the presence of the diagnosis in medical records, altered hepatic tests defined by abnormal values of laboratory test, acute cerebrovascular disease, patients receiving Clonidine, patients under guardianship or curatorship.

Randomisation was centralized and performed by the independent clinical research unit at the Brest University Hospital. A blocked randomisation with varying block sizes was performed. Randomization was also stratified on centres and planned modality of surgery between ‘on pump’ and ‘off pump’. Before surgery, patients were randomly assigned (1:1), by local investigators, either in the dexmedetomidine group or the placebo group using a dedicated and protected website (CSOnline; Clinsight). All the randomisation process allows a rapid and concealed treatment assignment from patients and site investigators.

The study drugs, dexmedetomidine (100 µg/ml) and placebo (sodium chloride 0.9%), were conditioned in vials of 2 ml. The size and shape of the two vials, the colour and texture of the two treatments are strictly identical. Blinding was performed by the pharmacist at the coordinating centre by erasing the original label of the vials and then labelling them accordingly to study treatment.

For all participants, a continuous infusion of dexmedetomidine or matching placebo was started the day of surgery from 8 pm to 8 am on the next day. Minimal infusion rate was 0.1 µg/kg/h and maximum 1.4 µg/kg/h. Infusion rate was modified by the treating nurse or the clinician by 0.1 µg/kg/h every hours with an objective of a Richmond Agitation and Sedation Scale (RASS) from − 1 to + 1. From the day of surgery, the treatment was administered every night until the patient was discharge from ICU or stopped after 7 days if the patient remained in ICU. Open label use of dexmedetomidine were not authorized. Administration of clonidine was not allowed in both groups.

Other medical interventions, especially sedatives/analgesics were left at the discretion of clinicians. The nine participating centres are high-volume cardiac surgery centres. Local investigators follow the latest French guidelines on enhanced recovery after cardiac surgery endorsed by the French Society of Anaesthesia and Intensive Care medicine and the French Society of Thoracic and Cardiovascular Surgery [17]:

In all participating centre, delirium prevention relied on ABCDE bundle.

The primary outcome was the occurrence of PoD, evaluated by the CAM-ICU, within the 7 days after surgery. Before the beginning of the study, staffs and research personnels were trained to performed CAM-ICU with the French version of the CAM-ICU Training Manual [21]. During the study, the CAM-ICU was measured twice daily during the 7 days following surgery. Measurements were performed in the morning between 8 and 12 am and in the afternoon between 4 and 8 pm. At the time of evaluation, dexmedetomidine or placebo was no longer administered since it was strictly a night infusion. If patients were discharged from ICU before the 7th day after surgery, twice daily PoD evaluation was performed in the surgical ward with the same modalities. A PoD event was considered if at least one of all the CAM-ICU evaluations was positive within 7 days after surgery.

Secondary efficacy outcomes included agitation-sedation status evaluated by the Richmond Agitation-Sedation Scale (RASS) until day 7, occurrence of agitation-related adverse events until day 7 (defined by an unplanned extubation, a medical device removal, a falling out of bed, an ICU runaway, an immobilisation device removal, a self-aggression and/or an agrgession towards medical staff), sleep quality evaluated with a numerical scale (from 0 to 10) until day 7, quality of sleep with the Leeds Sleep Evaluation Questionnaire (LSEQ) until day 7, ICU length of stay, hospital length of stay and hospital mortality. Long-term functional outcomes were also evaluated at 3 months: quality of life with Short-Form 36 questionnaire (SF-36) [18, 19], cognitive function with the Cognitive Failure Questionnaire (CFQ), occurrence of post-traumatic stress disorder (PTSD) evaluated by the Post Traumatic Stress Disorder Checklist (PCL-5) questionnaire (CFQ, PCL-5 and LSEQ are described in Additional file 1: Appendix 1 to 3). As preplanned ancillary analysis, we also evaluated the effect of dexmedetomidine on supra ventricular arrhythmias [20]. Due to the relevance of this outcome, it was finally decided to include this ancillary analysis in the article.

Secondary safety outcomes included bradycardia, hypotension, arrythmia and renal failure (assessed by the renal sub-part of the SOFA score) within the 7 days after surgery.

According to recent clinical trials, we hypothesized a PoD incidence of 25% in the studied population [2, 14]. We expected a 50% decrease in incidence of delirium in the dexmedetomidine group. Data from previous clinical trials showed that such treatment effect is clinically relevant [20‐22]. We calculated that 332 patients would be needed to detect this difference with a 5% one-tailed type I error and a power of 90%. Considering a lost to follow-up rate of 5%, the final sample size was fixed at 348 patients. Sample size calculation was performed using SAS statistical software (version 9.4).

For continuous variables, baseline characteristics of patients were described with mean and standard deviation for normally distributed variables, or median and interquartile range for the other continuous variables. Number and frequencies were used for categorical variables description.

Primary outcome analysis was performed with Chi-square test in the intention-to-treat population. No interim analysis was performed. If the occurrence of PoD is missing for a patient, we have considered that the patient had no PoD within 7 days. We performed a sensitivity analysis of the primary outcome without missing values for PoD. A post hoc analysis was also performed to evaluate the number of number of days alive without delirium at Day 7 in both groups in the overall population and also in the sub-group of patients with delirium. For this analysis, the number of number of days alive without delirium was defined as the number of days alive without any positive CAM-ICU at Day 7. If a patient had both a positive and a negative CAM-ICU in the same day, we considered this case as a positive for PoD at that day. This post-hoc analysis was performed with a Wilcoxon test.

Secondary outcomes analysis was also performed in the intention-to-treat population, with Student t test for continuous variables and Chi-square test for categorical variables.

For the primary outcome, analysis was also performed in the following subgroups: modality of surgery (by-pass vs. off-pump) and type of cardiac sugery (valvular, coronary or combined surgery).

As an exploratory analysis, we also evaluated PoD and sleep quality (LSEQ and numerical scale), in the as-treated population (i.e., patients who received dexmedetomidine or placebo the night before the evaluation).

Statistical analysis was performed using SAS statistical software (version 9.4), except for figures which were made with R statistical software (version 3.6.1).

In the intention-to-treat population, PoD occurred in 40 patients (12.5%): 20 of 165 patients assigned to the dexmedetomidine group (12.6%) and 20 of 166 patients assigned to the placebo group (12.4%), p = 0.97 (Table 2). Eleven PoD assessments were missing: 6 patients in dexmedetomidine and 5 in the placebo group had no CAM-ICU evaluation within 7 days. Main reason why CAM-ICU assessment was missing from Day 1 to Day 7 are reported in the Additional file 1: Table S3. When considering missing values for PoD without imputation, sensitivity analysis did not find any difference between the two groups: 14 (8.5%) versus 15 (9%), p = 0.97. In the as-treated population (n = 312), PoD occurred in 27 patients (8.9%): 13 of 153 patients assigned to the dexmedetomidine group (8.7%) and 14 of 159 patients assigned to the placebo group (9.1%), p = 0.9 (Table 2). In the post hoc analysis, the median number of delirium free days at Day 7 were similar in both groups: 7 [7] versus 7 [7], p = 0.98. In the sub-group of patients with delirium (n = 40), there was also no difference in the median number of delirium free-days: 6 [5, 6] versus 6 [4.5–6], p = 0.44.

Minimum and maximum RASS score were comparable between the two groups (Table 2). The incidence of agitation-related adverse events was non-significant between the two groups: 15 of 165 patients assigned to the dexmedetomidine (9.1%) and 14 of 166 patients assigned to the placebo group (8.4%), p = 0.83 (Table 2). In the intention-to-treat population, there were no difference in sleep quality evaluation neither on numerical scale (median score: 5.2 vs. 5.3, p = 0.98) nor on LSEQ (median score: − 4 vs. − 4.2, p = 0.32) (Table 2). Differences in daily evaluation of the LSEQ are shown in Additional file 1: Table S4. Details of the LSEQ different sections during the 7 days of observation are shown in Additional file 1: Table S5. Main reasons why sleep quality evaluation was not performed or missing are shown in Additional file 1: Table S3.

There was no significant effect of dexmedetomidine on ICU length of stay, hospital length of stay and 90 days mortality (Table 2). There was no difference between the two groups regarding long-term postoperative functional outcomes (Short-Form 36, cognitive failure and incidence of PTSD). The incidence of supra-ventricular arrhythmias was comparable between the two groups: 74 of 165 patients assigned to the dexmedetomidine (44.8%) and 65 of 166 patients assigned to the placebo group (39.2%), p = 0.29.

In the as-treated population, sleep quality evaluated by numerical scale was significantly improved in the patients receiving dexmedetomidine (5.7 vs 5; p = 0.01) (Table 2). Moreover, the average scores of LSEQ were also significantly improved for patients receiving dexmedetomidine compared to those receiving placebo (− 4.0 vs. − 7.3; p = 0.02) (Table 2). Evolution of the LSEQ in the as-treated population is shown in Fig. 2.

Patients receiving dexmedetomidine had significantly more episode of hypotension compared to the patients in the placebo group (7.3% vs. 0.6%; P < 0.01). The occurrence of bradycardia was higher in the dexmedetomidine group compared to placebo, but the difference did not reach significancy (3.6% vs 0.6%; P = 0.07). These results are summarized in Table 3. Finally, no additional renal failure assessed by the renal component of the SOFA score were associated with the use of dexmedetomidine compared to placebo. The daily assessment of renal component of the SOFA score was displayed in the Additional file 1: Table S6.

There was no difference in PoD in the five preplanned sub-group analysis. Incidence of PoD was similar whenever the modality of surgery or type of cardiac surgery were. The results of subgroup analysis are shown in Additional file 1: Table S7.

Randomized, double-blind, placebo-controlled design remains the main strength of our study. Regarding the number of participating centres, our findings are also generalizable to the other units. Moreover, patients’ characteristics are similar to those include in a recent large multicentre French cohort [34]. As our study was conducted as a pragmatic trial, our findings are also largely applicable in cardiac surgery setting. Finally, our study is the first one which documented a potential effect of dexmedetomidine on long-term cognitive function in cardiac surgery setting.

Our study has also some limitations. First, the incidence of PoD was lower than expected so our study suffers from lack of power. We observed an incidence that is comparable to the incidence found recently in the DECADE trial (12% vs. 17%) [14], but it was higher than the PoD incidence reported in the MINDSS trial (8.5% vs. 2.9%) [15]. These discrepancies between RCTs reinforce the need for further open-access, large, integrated and scalable registries in cardiac surgery to inform trialists of outcome incidences. The effect of the intervention observed in our study is close to nil (between-group difference: 0.2%) and far less important than the effect size chosen for the study sample size calculation. So, our study was underpowered to identify such small effect size. The evaluation of sleep quality was not performed using polysomnography, which is the gold standard technology to assess sleep quality and diagnose sleep disturbance. However, this method cannot be applied to a large number of patients in a typical ICU environment. The LSEQ is a self-rating questionnaire that has been widely used and is considered to be robust and reliable and can be used as a surrogate for polysomnography [35, 36]. Our data show consistency between the results using LSEQ and the sleep quality numerical scale. Sample size calculation was performed using one-sided alpha (5%). It ignores the possibility that dexmedetomidine may increase PoD or increase ICU stay by over-sedating the patients. To our knowledge, such side effects have not been reported in the literature. A two-tailed alpha remains the statistical gold standard, but it this particular case it would not have influence the result of our study. Finally, approximatively one third of the patients received at least once some benzodiazepine during their ICU stay, although this is balanced between the two groups it could have mitigate the effect of the intervention.