Study design and oversight
This was a single-center, randomized, prospective, double-blind, parallel-arm trial that was designed to evaluate the safety and efficacy of nocturnal dexmedetomidine administration to patients undergoing cardiac surgery as a means of preventing PICS. The Ethics Committee of Shandong Provincial Hospital approved this study (No. 2018-201), which was registered online prior to recruitment at chictr.org.cn (ChiCTR1800014314). This study was the second stage of the ChiCTR1800014314 protocol. In the first stage of the protocol, the primary endpoint was to delineate the relationship among dexmedetomidine, anxiety, and new-onset postoperative atrial fibrillation (POAF) [13]. We found that anxiety incidence was significantly reduced in the dexmedetomidine group compared to the controlled group (51.6% vs. 67.2%). By collecting cognitive and physical status of patients during follow-up, we also unexpectedly found that dexmedetomidine improved the cognitive and physical impairments of study subjects. As such, we further regenerated the ChiCTR1800014314 protocol, with the new primary outcome having been changed to focus on PICS. Based on the previous results of the first-stage and PICS incidence in other studies [15], we recalculated the necessary sample size for this study. The modified protocol has been registered and filed with the ethics committee of our hospital. Therefore, the current study was an improved version of ChiCTR1800014314. All patients or their legal representatives provided written informed consent to participate in this study, and the safety data associated with this trial were regularly reviewed by an independent Data and Safety Monitoring Board.
Patient recruitment
Consecutive patients scheduled to undergo elective cardiac surgery associated with a planned ICU stay of over 2 days who were receiving continuous or intermittent sedatives to facilitate safety and comfort were recruited for this study. For full details regarding study inclusion and exclusion criteria, see Additional file 1: Table S1.
Randomization and masking
Patients were assigned to either a dexmedetomidine or a placebo group at a 1:1 ratio following ICU admission. An independent team that was not involved in study recruitment, assessment, or intervention conducted patient randomization based on a computerized random number generator. The study drug was prepared by the pharmacy or an otherwise uninvolved research associate so that investigators and clinicians were fully blinded to allocation. Selection bias was mitigated by blinding all study personnel to group assignments throughout the 6-month study period until data were unlocked. Identical rehabilitation protocols were conducted in both groups aside from the nocturnal administration of dexmedetomidine, as shown in Additional file 1: Table S2.
Procedures
Sedation goals
This trial sought to achieve light sedation at night unless such sedation was contraindicated or considered to be unsafe by the attending clinician. Such light sedation was defined by a Richmond Agitation and Sedation Scale (RASS) of − 1 to 0 regardless of ventilation status such that patients achieved 6–8 h of sleep at night following surgery.
Dexmedetomidine or placebo administration
Dexmedetomidine or placebo (normal saline) were administered at night to patients in the appropriate treatment group as the sole or primary sedating agent utilized to achieve light sedation. Dexmedetomidine or a similar volume of saline was administered without a loading dose at 0.5 μg/kg/h i.v. from 10:00 PM until 6:00 AM the following day (halved at 5:00 am and discontinued at 6:00 am.), with the infusion rate being increased every 30 min when RASS scores were > 0 up to a maximum rate of 1.2 μg/kg/h until the target RASS was achieved because the majority of the safety concerns associated with this drug are dose-related. When RASS scores were ≤ − 2, clinicians were allowed to decrease administration of the study medication by a unit of 0.2 μg/kg/h every 30 min if clinically indicated. RASS scores were assessed at least once every 2 h based upon responses to voices, stimulation, self-control, and vital signs. If a subject was expected to require sustained treatment with sedatives or agitation occurred in treated patients (RASS ≥ 1), the administration of other sedatives such as propofol, opioids, and benzodiazepines was permitted when the effects of dexmedetomidine or placebo were insufficient. When patients were transferred out of the ICU, dexmedetomidine or placebo were administered as required, and were recommended for those suffering from insomnia. Study medication usage extended for at least 3 nights. While the concomitant use of antipsychotic agents to promote sleep induction was discouraged, the use of these agents was permitted at the clinician’s discretion to manage any delirium that occurred, as shown in Additional file 1: Fig. S1 and S2.
Evaluation of PICS problems
On day 2 after admission, day 7 after surgery, and at 3 and 6 months post-discharge, study personnel blinded to patient group assignments evaluated patient cognitive, physical, and mental health with the MMSE [16], Barthel index [17], Zung Self-Rating Anxiety Scale (SAS) [18], and Zung Self-Rating Depression Scale (SDS) [19] tools, respectively. Assessments were conducted in-person during hospitalization, while they were conducted via an online smartphone-based videoconference during follow-up [20]. Anxiety and depression were selected as representative traits for analyses of mental health challenges, given that anxiety and depression are the most common such manifestations [21], and are five times more common than posttraumatic stress disorder (PTSD) in survivors of critical illness [5]. In light of similar studies and clinical practice, we established that a BI score of < 80 was considered indicative of mild dysfunction and physical impairment [22].
PICS-related impairments were defined using standard threshold values for these respective measurement scales, with an MMSE score ≤ 26 indicating cognitive impairment, a Barthel index score of ≤ 80 indicating disability, and an SAS score and/or SDS > 50 indicating psychological impairment.
Outcome assessments
PICS incidence at 6 months post-discharge served as the composite primary endpoint for this study. Patients exhibiting at least one of these criteria were considered to have met the primary study endpoint.
Secondary study endpoints included (a) mortality during hospitalization and within 6 months after discharge; (b) PICS incidence and its components at 3 months post-discharge; (c) postoperative ICU stay duration; (d) postoperative hospital stay duration; (e) tracheal intubation time; (f) acute kidney injury(AKI) [23] and POAF [13]; (g) delirium during hospitalization—delirium was defined using the Confusion Assessment Method for the Intensive Care Unit (positive or negative) when the RASS scores were − 2 or higher; (h) postoperative sedative drug use after study drug administration; (i) PICS co-occurrence at 6 months post-discharge—PICS symptom co-occurrence was defined as the new-onset occurrence of two or more of the three PICS elements; (j) sleep quality assessed using Pittsburgh Sleep Quality Index (PSQI) during hospitalization and follow-up; and (k) safety endpoints included hypotension (mean arterial pressure < 60 mmHg during infusion) and severe sinus bradycardia (heart rate < 50 bpm). Supplementary administration of extra fluids or vasoconstrictors to reverse hypotension were also recorded.
Missing data
The rate of missing data in our study was low. As some missing data were duplicated, 12 patients in the dexmedetomidine group and 14 in the placebo group ultimately exhibited missing data (Fig. 1). The missing measurement data primarily included SAS, SDS, MMSE, and Barthel index during hospitalization and follow-up, as well as length of ICU stay, length of hospital stay, and tracheal intubation time due in cases of death. For every dataset with missing values, we created imputed datasets and used predictive mean matching multiple imputation at the time of regression modeling to account for missing data at hospitalization or at each time point during follow-up [24]. For all missing data replacement values, we set age and sex as the independent variables in the imputation model. In addition, missing categorical data were replaced by negative results.
Sample size
Cognitive impairment has previously been shown to affect 30–80% of ICU survivors, while psychological impairment impacts 25–80% of these patients, often for extended periods of time [25, 26]. Specifically, PICS was present in 64% and 56% of survivors at 3 and 12 months, respectively [15]. However, post-ICU sequela rates vary substantially across patient populations, and no comparable studies of the impact of dexmedetomidine on PICS were available. In our previous pilot study, anxiety was significantly reduced by dexmedetomidine administration (51.6% in the dexmedetomidine group and 67.2% in the control group) [13]. We also utilized the MMSE to assess cognitive impairment and the Barthel index to assess physical impairment at the 6-month follow-up time point. PICS incidence was 50.8% (31/61) in the control group and 35.5% (22/61) in the dexmedetomidine group. Based on the literature and our pilot study, we posited that PICS would likely occur in 50% of patients in the placebo group and 35% of patients administered dexmedetomidine at 6 months post-discharge. To achieve 90% power to detect such a difference at a two-tailed significance level of 0.05, we calculated that 227 patients per group would be required. To account for a 15% attrition rate, we targeted enrollment of 261 patients per group, and we ultimately enrolled 268 patients per group.
Statistical analysis
Primary analyses were conducted on a modified intention-to-treat (MITT) population, shown in Fig. 1. Data are given as means with standard deviations (SDs), median with interquartile ranges (IQRs), or frequencies and proportions, as appropriate. Differences between groups are presented with 95% confidence intervals (CIs). Shapiro-Wilk tests were used to evaluate data normality. Logistic regression analyses were used to assess the impact of treatment on study outcomes using odds ratios (ORs) with 95% CIs. Numeric data were compared via Student’s t-tests or Mann-Whitney U tests as appropriate. Categorical variables were analyzed with the χ2 test. MMSE, SAS, Barthel index, and PSQI were analyzed via repeated-measures analyses of variance (RM-ANOVAs) with time as the within factor (Pw), treatment group as the grouping factor (Pg), and interactions identified in the RM-ANOVA (Pi).
Exploratory subgroup analyses of the primary endpoint data were conducted based upon six variables: age (< 65 or ≥ 65), whether or not cardiopulmonary bypass was used, education level (< 9 or ≥ 9 years), PSQI after admission (< 6 or ≥ 6), SOFA score (greater than or below the median [7.14]), and BMI (< 24 or ≥ 24).
Post hoc analyses of PICS risk were divided into two parts. In the first part, 10-fold cross-validation was conducted on the global data of 508 patients (the training cohort of 458 subjects and the validation cohort of 50 subjects) for PICS prediction (Additional file 1: Fig. S3). Univariate logistic regression analyses were used to identify risk factors linked to PICS risk and its components in the training cohort. All baseline variables, such as age, BMI, sex, education, SOFA Score, APACHE score, postoperative atrial fibrillation, dexmedetomidine group status, and previous medical history of hypertension, diabetes and smoking were initially examined as risk factors. If those variables yielded P < 0.05, they were incorporated into subsequent multivariate-adjusted binary logistic regression analyses in an effort to identify independent predictors. These independent predictors were then used to prepare a predictive nomogram using the rms package. The area under the ROC curve (AUROC) was used to present the discriminative ability of this nomogram. Calibration curves were generated by plotting predicted vs. actual PICS rates to establish the accuracy of this nomogram. Individual predictors were marked with horizontal lines in the final nomogram. In the second part, we employed multivariate logistic regression analyses to determine dose-response relationships between risk factors and PICS at 6-month follow-up. Dexmedetomidine group status, age, education, diabetes, renal failure, smoking, postoperative atrial fibrillation, and SOFA Score were included as independent variables in the model. All statistical tests were two-sided, and P < 0.05 served as the significance threshold. Multiple comparisons testing was not performed, and so all secondary outcome data should be considered exploratory. R for Windows (v.3.42, http://www.r-project.org/) was used for all analyses.