Effect of remimazolam on postoperative delirium and cognitive function in adults undergoing general anesthesia or procedural sedation: a meta-analysis of randomized controlled trials

Eligible criteria

The following population, intervention, comparison, outcomes, and study design (PICOS) were used to define the inclusion criteria: (1) participants undergoing surgeries or procedures, (2) performing general anesthesia sedation using remimazolam and propofol, (3) studies that compared the incidence of postoperative delirium or cognitive function, and (4) randomized controlled trials (RCTs) reporting at least one of the outcomes of interest. Exclusion criteria were retrospective cohort studies, meta-analyses, review articles, editorials, letters to the editor, case reports, observational studies, study protocols, conference abstracts, and animal studies.

Search strategy

To identify suitable RCTs for this study, we searched electronic databases, including PubMed, CENTRAL, Embase, Scopus and Web of Science, from their inception to July 04, 2024, without any restrictions on language, region, publication year, or journal. The search terms included ‘remimazolam,’ ‘Byfavo,’ ‘Anerem,’ ‘Aptimyda,’ ‘Ruima,’ ‘delirium,’ ‘cognitive dysfunction,’ ‘cognitive impairment,’ and ‘cognition.’ The detailed literature search strategies are listed in Supplementary Table 1.

Study selection and data extraction

Two independent reviewers (JIP and HJS) selected appropriate RCTs for data synthesis based on predetermined inclusion and exclusion criteria. The non-English article [10] was translated using Google Translate [11] and ChatGPT-4 [12], both of which have established accuracy. Relevant studies from each electronic database were selected by reviewing titles and abstracts, followed by a full-text review, to determine the final studies to be included in the analysis. Any disagreements were resolved by a third reviewer (HSN).

Following the evaluation of the final RCTs by two independent reviewers (JIP and HJS), key variables were extracted and documented in a data collection form for synthesis. The collected data included information such as authors, publication year, sample size, mean age of participants, American Society of Anesthesiologists Physical Status (ASA-PS), and details regarding remimazolam and propofol (including induction and maintenance doses). Additionally, data on the occurrence of postoperative delirium and hypotension were recorded. Postoperative cognitive function status (Mini-Mental State Examination [MMSE] score) was also assessed. Median values with interquartile ranges were converted to means with standard deviations (SDs) using Wan’s formula [13]. For graphically represented data, values were extracted using WebPlotDigitizer (version 4.8; Ankit Rohatgi; Available at https://automeris.io/WebPlotDigitizer).

Assessment of risk of bias and evidence quality

Two independent reviewers assessed the risk of bias for each outcome of the included studies using the revised Cochrane risk-of-bias tool for randomized trials (RoB 2) [14]. This evaluation addressed five domains: 1) bias arising from the randomization process, 2) bias due to deviations from intended interventions, 3) bias due to missing outcome data, 4) bias in outcome measurement, and 5) bias in the selection of reported results. The risk of bias was categorized as ‘low risk,’ ‘some concerns,’ or ‘high risk.’

The certainty of the evidence for each outcome was evaluated using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) system that considers five domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias [15].

Outcome measures

The incidence of postoperative delirium or cognitive function was considered as the primary outcome. Secondary outcomes included the incidence of hypotension, as it is known to be related to changes in cognitive function after surgery [16].

Statistical analysis

The effect size for each trial was calculated using Stata SE^TM (version 17.0; Stata Corp.) to describe the effects of the interventions. Odds ratios (OR) with 95% CI for dichotomous variables and mean differences (MD) with 95% CI for continuous variables were adopted as effect sizes. A random-effects model was adopted for all statistical analyses to provide a robust estimation of the effect size considering the variability between studies. Statistical significance was set at P < 0.05.

The level of heterogeneity among the pooled effect sizes was evaluated using Cochran’s Q test and I² statistics and classified as none (I² < 25%), low (I² = 26%‒49%), moderate (I² = 50%‒74%), or high (I² > 76%) [17]. When the heterogeneity was reported to be high, we performed a meta-regression analysis to explore the potential sources of heterogeneity, including publication year, sample size, age, type of anesthesia, ASA-PS, and dose of remimazolam for induction as covariates. Furthermore, a leave-one-out sensitivity analysis was conducted to detect small study effects to determine whether any single study could influence the robustness of the pooled effect size. Subgroup analyses were conducted according to the type of anesthesia (general anesthesia and procedural sedation).

Considering Cochrane’s guidelines [18], as the number of included studies for each outcome was less than 10, we did not assess publication bias either visually (funnel plot) or statistically (Egger’s test).

In a meta-analysis, the potential non-independence of results from studies with multiple intervention arms was addressed by splitting the shared group into two distinct groups [19]. This decision was guided by the recommendations of the Cochrane Handbook for handling multiple comparisons [18]. Pre-specified criteria were established based on similarities in the intervention type, population characteristics, and outcome measures. Separate effect sizes were calculated for each group, and the robustness of the findings was confirmed through sensitivity analyses.  

Study selection

Initially, 333 articles were identified using electronic databases. After removing 125 duplicates, screening based on titles and abstracts led to the exclusion of 63 and 124 articles, respectively. After excluding two studies for which the full manuscripts could not be retrieved from either the database or the authors, the full texts of 17 eligible studies were thoroughly reviewed, and 11 were included in the final analysis [10,19–28]. The following reasons were identified for excluding studies during the review process: the primary outcome was not measured [29,30]; propofol was used only as an adjuvant agent [31,32]; the study was a protocol rather than a completed study [33]; and the study design was retrospective [34]. A flow diagram is shown in Fig. 1.

Fig. 1.

PRISMA flow diagram of study selection. A total of 333 articles were identified through the electronic databases. After excluding 125 studies for duplication, 63 and 124 articles were removed from the article pool based on the fitness of the title and the abstract, respectively. After removing two studies that could not be retrieved, the full texts of the remaining 17 articles were reviewed. Finally, a total of 11 RCTs were included in the meta-analysis. PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analyses, RCT: randomized controlled trial.

Characteristics of study and participant

The characteristics of the RCTs included in the final analysis are summarized in Table 1. A total of 1,295 adult participants were included, with 664 and 631 assigned to the remimazolam and propofol groups, respectively. In nine studies [10,19,22–28], the participants received general anesthesia, while the remaining two studies [20,21] involved participants undergoing bronchoscopy and hip fracture surgery after sedation. The assessment tools for detecting postoperative delirium include the Confusion Assessment Method (CAM) [20,24,25,27], the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V) [22], and the Postoperative Quality of Recovery Scale (PQRS) [19,28]. The MMSE was used to determine the status of postoperative cognitive function [10,19–21,23,26]. Various regimens were applied for total intravenous anesthesia induction and maintenance with remimazolam and propofol perioperatively, except in one study [25] that added desflurane to remimazolam and propofol for anesthesia maintenance.

Table 1.

Characteristics of Included Studies

Incidence of postoperative delirium

The incidence of postoperative delirium was reported in eight RCTs [19–22,24,25,27,28] with nine comparisons (two different comparisons were made in one study [19]). The incidence of postoperative delirium was 8.0% (44 cases out of 552) in the remimazolam group and 10.4% (54 cases out of 520) in the propofol group. The overall pooled effect size revealed that remimazolam administration did not increase the incidence of postoperative delirium (OR: 0.74, 95% CI [0.39–1.42], P = 0.369, I² = 32%; Fig. 2).

Fig. 2.

Forest plot for the incidence of postoperative delirium. No significant difference was observed in the incidence of postoperative delirium between the remimazolam and the propofol group. In the subgroup analysis, remimazolam lowered the incidence of postoperative delirium compared with propofol when used for sedation. GA: general anesthesia, FMZ: flumazenil.

Subgroup analysis showed that remimazolam significantly reduced the incidence of postoperative delirium compared with propofol after sedation (OR: 0.24, 95% CI [0.08–0.73], P = 0.012, I² = 0%; Fig. 2), whereas no significant differences were observed between remimazolam and propofol after general anesthesia (OR: 1.04, 95% CI [0.65–1.66], P = 0.886, I² = 0%; Fig. 2). Considering that the heterogeneity changed through subgrouping, the low heterogeneity (I² = 32%) in the overall effect size may be due to the type of anesthesia.

Sensitivity analysis showed that the pooled OR did not change significantly after omitting each study, supporting the robustness of our results (Supplementary Fig. 1).

Postoperative cognitive function

Postoperative cognitive function was assessed in six RCTs [10,19–21,23,26] with eight comparisons (two different comparisons were made in two studies [10,19]), using the MMSE as the assessment tool. A significant difference was observed between the two groups (MD: 1.06, 95% CI [0.32–1.80], P = 0.005, I² = 89%; Fig. 3).

Fig. 3.

Forest plots for postoperative cognitive function. Participants who were administered remimazolam showed better cognitive function than those who were administered propofol after surgery, and this was also observed in the general anesthesia subgroup, but not in the sedation subgroup. GA: general anesthesia, FMZ: flumazenil, SD: standard deviation.

In the subgroup analysis, postoperative cognitive function was higher in the remimazolam group than in the propofol group after general anesthesia (MD: 1.30, 95% CI [0.53–2.06], P = 0.001, I² = 90%; Fig. 3), but not after sedation (MD: –0.13, 95% CI [–2.18 to 1.91], P = 0.898, I² = 63%; Fig. 3). Despite subgrouping, the heterogeneity remained high. To identify the potential source of this heterogeneity, additional meta-regression was conducted, including publication year, type of anesthesia, ASA-PS, sample size, induction dose of remimazolam (induction dose of propofol, maintenance dose of remimazolam, and propofol were omitted due to collinearity), and age. No covariates significantly influenced the heterogeneity, suggesting the presence of other hidden factors (Supplementary Table 2).

The leave-one-out sensitivity analysis did not reveal any changes in the pooled effect size of postoperative cognitive function (Supplementary Fig. 2).

Incidence of hypotension

The incidence of hypotension was reported in seven studies [7,8,15–17,20,21,23] with eight comparisons (one study included two different comparisons [19]). The incidence of hypotension was 23.2% in the remimazolam group and 49.9% in the propofol group. The overall pooled effect size indicated that remimazolam significantly lowered the incidence of hypotension in participants undergoing surgery or procedure (OR: 0.28, 95 CI [0.21–0.37], P = 0.000, I² = 0%; Fig. 4).

Fig. 4.

Forest plot for the incidence of hypotension during surgery. Significantly lower incidence of hypotension was observed in the remimazolam group than in the propofol group. GA: general anesthesia, FMZ: flumazenil.

Subgroup analysis demonstrated that remimazolam significantly decreased the incidence of hypotension compared to propofol in both general anesthesia (OR: 0.28, 95 CI [0.20–0.39], P = 0.000, I² = 0%; Fig. 4) and sedation settings (OR: 0.26, 95 CI [0.13–0.52], P = 0.0002, I² = 0%; Fig. 4).

In the leave-one-out sensitivity analysis, no study skewed the effect size, supporting the robustness of our results (Supplementary Fig. 3).

Risk of bias

The overall risk-of-bias assessment results are shown in Supplementary Fig. 4 and Supplementary Fig. 5. The RoB 2 tool assesses studies based on intention-to-treat (ITT) and per-protocol (PP) analyses. The included studies were then evaluated and classified using this approach. Six studies [10,20,24–27] used the ITT method, while five studies [19,21–23,28] used the PP method. Four RCTs [23–25,27] were classified as ‘Low,’ and seven RCTs [7,16,20] had ‘Some concerns’ regarding the randomization process [10,20–22,26,28] or deviations from intended interventions [19,22].

Certainty of the evidence

As shown in Supplementary Table 3, the certainty of the evidence was moderate for the four outcome variables (incidence of postoperative delirium and hypotension). However, the assessment of postoperative cognitive function resulted in a decrease in the evidence level owing to concerns regarding the risk of bias and high heterogeneity.