Delirium, which is an acute disorder of cognition, attention, and perception, is a common, serious, and often potentially preventable clinical syndrome with an increased risk of morbidity and mortality among patients who are 65 yr of age or older.1,2 The prevalence of delirium ranges from 12 to 51% in patients after non-cardiac surgery, and it increases with age.2,3 Postoperative delirium (POD) is distressing for patients and their families, and has been associated with higher morbidity and mortality, prolonged length of hospital stay, and increased healthcare costs.4 At present, primary prevention with multicomponent nonpharmacological approaches has gained widespread acceptance as the most effective strategy for delirium.2,5 These approaches include reorientation, therapeutic activities, reduction of psychoactive medications, early mobilization, promoting sleep, maintaining hydration and nutrition, and providing vision and hearing adaptations. Dexmedetomidine (DEX) is a highly selective alpha-22) adrenoreceptor agonist that provides anxiolysis, sedation, and modest analgesia with minimal respiratory depression.6 One randomized-controlled trial (RCT) and one meta-analysis have shown that DEX reduces the incidence of POD in patients after cardiac surgery.7,8 Recently, another RCT revealed that DEX was associated with a decrease in POD incidence in elderly patients undergoing non-cardiac surgery.3 Nevertheless, the literature is not entirely congruent, with another study concluding that DEX does not decrease the incidence of delirium in elderly patients after non-cardiac surgery.9 A recent meta-analysis of RCTs by Duan et al.10 revealed that DEX reduced the incidence of POD in elderly patients after both cardiac and non-cardiac surgery. A recent meta-analysis assessed the effect of DEX on the incidence of delirium in the intensive care unit (ICU) and concluded that DEX might be more efficacious than placebo or standard sedatives in reducing POD in patients after cardiac and non-cardiac surgery.11 Nevertheless, this meta-analysis did not include some published non-English studies, and, in the meantime, a number of more recent studies evaluating the effect of DEX on delirium in elderly patients after non-cardiac surgery have been published. Therefore, we performed an updated meta-analysis of RCTs to ascertain more precisely the efficacy of DEX on POD incidence in elderly patients after non-cardiac surgery.


This meta-analysis was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.12

Search strategy

We systematically searched relevant articles via PubMed, EMBASE, the Cochrane Library, Web of Science, and the Cumulative Index of Nursing and Allied Health Literature (CINAHL) from inception to February 24, 2019. We retrieved conference proceedings from Conference Proceedings Citation Index-Science (CPCI-S) and CPCI-Social Science & Humanities (CPCI-SSH). In addition, we searched and grey-literature databases (GreyNet International, OpenGrey, and System for Information on Grey Literature in Europe) to identify further eligible studies. Finally, we manually searched the references of our included studies and of previous systematic reviews for further studies that were not identified in our initial search. A senior specialist in evidence-based medicine reviewed the final search result. Our basic search strategy included the following terms: (“dexmedetomidine”[MeSH Terms] OR “dexmedetomidine”[All Fields]) AND (“delirium”[MeSH Terms] OR “delirium”[All Fields]) AND (“aged”[MeSH Terms] OR “aged”[All Fields] OR “elderly”[All Fields]). No language restrictions were applied to the search.

Eligibility criteria

We included studies with the following criteria: 1) subjects were elderly patients (either ≥ 60 or ≥ 65 yr old) undergoing non-cardiac surgery; 2) the study compared the effect of DEX with that of normal saline (NS) or other anesthetic drugs, regardless of starting time, dose, or duration of administration; 3) incidence of POD was reported as the outcome; 4) studies were RCTs; 5) the Confusion Assessment Method (CAM) or CAM-ICU was used to evaluate delirium; and 6) the full text was available.

Exclusion criteria

Non-RCTs, case reports, conference abstracts, comments, and systematic reviews were excluded. Studies involving cardiac surgery, non-elderly patients, the combination regimes of DEX and other drugs, oral DEX, and animal experiments were also excluded.

Study selection and data collection process

Two authors (P.H. and Z.M.Y.) independently performed the eligible study selection and data extraction. Disagreements between these two authors were resolved by discussion with a third author (W.Y.). The extracted data were as follows: first author; year of publication; age; surgery type; number of patients; starting time; dose and duration of DEX; anesthesia depth monitoring; and methods of delirium assessment. Although the primary endpoint was the incidence of delirium, secondary outcomes were also extracted including hypotension, bradycardia, time to extubation, duration of mechanical ventilation (MV), and length of ICU and hospital stays. Hypotension and bradycardia were defined by the authors of individual studies.

Risk of bias in individual studies

We used the Cochrane Collaboration tool for assessing the risk of bias (Cochrane, London, UK) to evaluate the quality of included studies.13 This assessment was completed independently by two authors (P.H. and Z.M.Y.); disagreements were resolved by discussion with a third author (W.Y.) until a consensus was reached.

Statistical analysis

We performed the meta-analyses using Review Manager software version 5.3 (Cochrane). Risk ratios (RR) and their 95% confidence intervals (CI) were calculated for dichotomous outcomes. For continuous outcomes, mean differences (MD) or standardized mean differences (SMD) with 95% CIs were calculated. When median and interquartile range or median and 95% CI were used in included studies, standard deviation was calculated by the formula suggested by the Cochrane Handbook. A random-effects model using the DerSimonian and Laird method was applied in all analyses. Statistical heterogeneity was assessed by the Chi square test and I2 index (where we considered an I2 > 50% to represent substantial heterogeneity).

Subgroup analyses of various DEX strategies (starting time, dose and duration of administration, with or without loading dose) were performed to evaluate the efficacy of these strategies on POD. We also performed subgroup analysis of the strategy for various control drugs to evaluate the effect of DEX on POD and explore any potential source of heterogeneity. We also performed a sensitivity analysis to evaluate the stability of the pooled results. To evaluate potential publication bias, we created a funnel plot, and we assessed its symmetry using Egger’s test using Stata version 15 (StataCorp LLC, College Station, TX, US). P < 0.05 (two-sided) was considered to be statistically significant.


Study selection

A flow chart of study selection is shown in Fig. 1. We initially identified 988 studies by searching databases and references. After screening these articles according to titles and abstracts, we excluded 962 that did not meet eligibility criteria. In total, we preliminarily evaluated 26 full-text articles for eligibility. Ultimately, we included 11 studies3,9,14,15,16,17,18,19,20,21,22 meeting all eligibility criteria in our meta-analysis.

Fig. 1
figure 1

Flow chart of study selection

Study characteristics

Characteristics of the included studies and DEX strategies are presented in Tables 1 and 2, respectively. We acquired POD incidence data from one study21 using Getdata Graph Digitizer software ( Dexmedetomidine starting time and dose varied across the included studies. Patients in four studies3,15,16,22 received DEX during the postoperative period, but those in the study by Deiner et al.9 received DEX during the intraoperative period and two subsequent hours. In six other studies,14,17,18,19,20,21 DEX was administered during the intraoperative period only. In four studies, the loading doses used were 1.0 μg·kg−1,14,20 0.8–1.0 μg·kg−1,19 and 0.5 μg·kg−1,21 but in other studies DEX was administered continuously at different rates (0.1–0.7 μg·kg−1·hr−1) without a loading dose.3,9,15,16,17,18,22 In five studies,14,15,18,19,21 different anesthetic drugs (propofol, midazolam, or ketamine) were used in the control group, whereas eight studies3,9,14,16,17,20,21,22 used NS in the control group. These studies differed in duration of DEX administration. Nevertheless, only two studies3,16 reported the exact data of duration—12 hr16 and 14.95 hr.3 Although one study9 also reported duration of DEX administration (two hours) during the postoperative period, the figure for the intraoperative period was not obtainable. The total mean (standard deviation) dose of DEX was 105 (43) µg in one study,19 but the exact amount of DEX administered was not reported in the other studies. The largest study included 700 patients,3 while the smallest had 90 patients.14,21

Table 1 Characteristics of all included studies
Table 2 DEX strategy

Risk of bias within studies

The results of quality assessment are presented in Fig. 2. One study20 was rated as high risk for “blinding of participants and personnel” bias, while another22 was rated as high risk for “selective reporting” bias.

Fig. 2
figure 2

Risk of bias assessment according to the Cochrane Risk of Bias Methods

Meta-analysis of POD

This meta-analysis of 11 studies including 2,890 patients showed that DEX was associated with a significant decrease in the incidence of POD (RR, 0.47; 95% CI, 0.38 to 0.58; P < 0.001; Fig. 3). Heterogeneity (I2 = 26%; P = 0.20) between included studies was not substantial. We next performed a sensitivity analysis of POD incidence by excluding each study individually. Exclusion of the study by Mei et al.,19 in which patients received regional anesthesia resulted in a similar pooled result (RR, 0.47; 95% CI, 0.37 to 0.60; P < 0.001). Statistical heterogeneity (I2 = 0%; P = 0.92) was absent when the study conducted by Deiner et al.9 was excluded, but the pooled result changed minimally (RR, 0.43; 95% CI, 0.36 to 0.52; P < 0.001). To limit the risk of bias, we performed a sensitivity analysis excluding the two studies with a high risk of bias.20,22 The result (RR, 0.45; 95% CI, 0.33 to 0.60; P < 0.001; I2 = 36%) was consistent with that of primary analysis for all studies. When we conducted another sensitivity analysis excluding the studies that did not monitor anesthesia depth by bispectral index (BIS), the pooled results remained similar (RR, 0.44; 95% CI, 0.36 to 0.55; P < 0.001), and statistical heterogeneity was absent (I2 = 0%; P = 0.66).

Fig. 3
figure 3

Meta-analysis of the incidence of postoperative delirium after non-cardiac surgery

We conducted a subgroup analysis based on starting time of DEX administration (intraoperative, postoperative, or perioperative; Fig. 4). We grouped four studies3,15,16,22 including 1,417 patients into the postoperative period subgroup and six studies14,17,18,19,20,21 including 1,083 patients into the intraoperative period subgroup. The test of subgroup differences showed a statistically significant subgroup effect (P for interaction = 0.008), meaning that starting time of administration modified the effect of DEX on POD incidence compared with control. Pooled results showed that a decrease of POD incidence favoured DEX over control drugs during the intraoperative (RR, 0.44; 95% CI, 0.34 to 0.58; P < 0.001; I2 = 0%) and postoperative periods (RR, 0.42; 95% CI, 0.32 to 0.54; P < 0.001; I2 = 0%). Nevertheless, we included only one study9 in the perioperative (intraoperative and postoperative) period subgroup, so this result might be uncertain and required further confirmation.

Fig. 4
figure 4

Subgroup analysis of the incidence of postoperative delirium by starting time of dexmedetomidine administration (intraoperative, postoperative, or perioperative)

We conducted another subgroup analysis by whether a loading dose was used (Fig. 5). In four studies,14,19,20,21 patients (combined total of 794 patients) received DEX with a loading dose; in the other seven studies,3,9,15,16,17,18,22 (combined total of 2,096 patients) there was no loading dose. The test of subgroup differences indicated no statistically significant subgroup effect (P for interaction = 0.67), suggesting that administration of DEX with or without loading dose did not modify the effect of DEX on POD incidence in comparison with controls. Nevertheless, the DEX-with-loading-dose subgroup comprised only a small number of trials and patients. So this analysis might be underpowered to detect subgroup differences. Pooled subgroup results revealed that POD incidence decreased in the DEX-with-loading-dose subgroup (RR, 0.50; 95% CI, 0.36 to 0.70; P < 0.001; I2 = 0%). Nevertheless, there was significant heterogeneity between the studies included in the DEX-without-loading-dose subgroup (I2 = 52%). There was no certain evidence to confidently confirm this result of the subgroup effect so that further explanation for the heterogeneity was needed.

Fig. 5
figure 5

Subgroup analysis of the incidence of postoperative delirium by whether a loading dose was used

We performed further subgroup analysis for different control drugs (NS, propofol, midazolam, or ketamine; Fig.6). Two studies14,21 included two comparators in addition to DEX (groups 1 and 2). Eight studies with 2,334 patients contributed data to the NS subgroup, and five studies with 616 patients contributed to the anesthetic drug subgroup. The test for subgroup differences suggested that there was no statistically significant subgroup effect (P for interaction = 0.41), meaning that different control drugs did not modify the effect of DEX on POD compared with controls. Nevertheless, a smaller number of trials and patients contributed data to the anesthetic drug subgroup compared with the NS subgroup. This means that the analysis may be underpowered to detect subgroup differences. The results suggested that DEX could significantly reduce POD incidence in the anesthetic drug subgroup (RR, 0.40; 95% CI, 0.26 to 0.60; I2 = 0%), but there was a mild statistical heterogeneity in the NS subgroup (I2 = 45%). Therefore, there was no certain evidence to confidently confirm this result of the subgroup effect.

Fig. 6
figure 6

Subgroup analysis of the incidence of postoperative delirium by different control drugs

Meta-analysis of secondary outcomes

The incidence of hypotension was reported in five studies,3,9,15,16,22 while that of bradycardia was reported in four studies.3,9,15,16 These studies all defined hypotension and bradycardia differently; the definitions are summarized in Table 3. There was a significant increase in the incidences of hypotension (RR, 1.20; 95% CI, 1.04 to 1.39; P = 0.01; I2 = 0%) and bradycardia (RR, 1.33; 95% CI, 1.08 to 1.63; P = 0.007; I2 = 0%) in the DEX group compared with the control groups. Three studies3,18,21 compared the time to extubation between the two groups. Our meta-analysis showed that DEX could significantly shorten the time to extubation (SMD, -0.60; 95% CI, -1.17 to -0.03; P = 0.04; I2 = 88%). Four studies3,9,19,22 provided data for length of hospital stay, and results showed that DEX shortened hospital stays (MD, -0.50 days; 95% CI, -0.97 to -0.03; P = 0.04; I2 = 56%). Only the study by Huang et al.15 reported median duration of MV between the DEX (21.0 hr; 95% CI, 18.6 to 21.4) and control (28.0 hr; 95% CI, 25.6 to 30.4) groups. Two studies compared the length of ICU stay between groups. Therefore, there were insufficient data to perform meta-analyses for duration of MV and length of ICU stay.

Table 3 Definitions of hypotension and bradycardia

Publication bias

We evaluated potential publication bias of included studies using a funnel plot (Fig. 7). Egger’s test showed no significant asymmetry of the funnel plot (P = 0.55), indicating that there was no significant publication bias.

Fig. 7
figure 7

Funnel plot of meta-analysis of the incidence of postoperative delirium


This was an updated meta-analysis, including 11 studies with 2,890 patients, exploring the efficacy of DEX on delirium in elderly patients after non-cardiac surgery. Our meta-analysis, based on that conducted by Duan et al.,10 included an additional five studies involving 1,249 patients, which contributed to a more precise estimate of effect. We also performed subgroup analyses of DEX administered with or without a loading dose and of different control drugs to evaluate the efficacy of DEX on POD incidence. We found that DEX administration reduced the incidence of POD as well as the time to extubation and length of hospital stay, compared with the control group. But we also found that DEX was associated with increased risks of hypotension and bradycardia.

Delirium is a common, serious condition associated with increased morbidity and mortality in older patients.23 The multifactorial model for the etiology of delirium has been well-validated and widely accepted in elderly patients.24 Dexmedetomidine’s ability to reduce POD might be associated with its specific characteristics. Dexmedetomidine has a significant opioid-sparing effect due to high and specific α2-adrenergic receptor selectivity. It also lacks clinically significant anticholinergic effects, attenuates the inflammatory response, and improves quality of sleep in critically ill patients.25,26,27,28 Moreover, at least two theories have been formulated to explain how DEX reduced POD incidence by these criteria in the study by Maldonado et al.29 Therefore, DEX presents a potentially attractive means of preventing POD in elderly patients after non-cardiac surgery. In addition, there was no significant heterogeneity among all studies in our meta-analysis (I2 = 26%; P = 0.20). The pooled result of our meta-analysis was stable in sensitivity analyses. In addition, all of the studies in our meta-analysis used CAM or CAM-ICU to identify delirium. This method is the most widely used because of its high validity and reliability, with sensitivity of 94–100% and 81% and specificity of 90–95% and 96%, respectively.30,31 Furthermore, we included three studies3,9,22 with large double-blinded and parallel-group RCTs in our meta-analysis, as they would have an important impact on the pooled result. Consequently, that DEX reduces the incidence of POD in elderly patients after non-cardiac surgery seems to be a reliable conclusion.

Subgroup analysis revealed that DEX administration reduced POD incidence during the intraoperative and postoperative periods. This conclusion was in accordance with the results of a recent meta-analysis of DEX’s efficacy on POD.10 Our meta-analysis included more studies with elderly patients after non-cardiac surgery, and we conducted analyses of three subgroups representing the intraoperative, postoperative, and perioperative periods. Nevertheless, we included only one study in the perioperative subgroup. Further exploration is needed to clarify the effect of perioperative (intraoperative and postoperative) DEX administration on POD incidence.

We performed further subgroup analysis by whether a loading dose was used, and the result agreed with another meta-analysis that showed a decrease in POD incidence when DEX was administered with or without a loading dose.32 Notably, the studies in the DEX-without-loading-dose subgroup were accompanied by significant heterogeneity (I2 = 52%; P = 0.05), so the result was uncertain. One study in this subgroup, that by Deiner et al.,9 was considered an important source of heterogeneity. This study restricted DEX administration to the intraoperative period and two subsequent hours, rather than infusing it over a prolonged postoperative period. This duration of administration was relatively shorter than those in the other four studies3,15,16,22 in which DEX was administered postoperatively. The DEX infusion rate differed in every study we included. In the study by Deiner et al.,9 DEX was administered at a constant dose of 0.5 μg·kg−1·hr−1 for all participants, but in other studies it was infused at different doses: 0.1 μg·kg−1·hr−1,3,22 0.2–0.7 μg·kg−1·hr−1,15,18 0.2 μg·kg−1·hr−116 and 0.2–0.4 μg·kg−1·hr−1.17 In addition, anesthesia depth was not monitored,9 leading to potentially deeper anesthesia, whereas BIS monitoring was performed in three other studies.16,17,22 Furthermore, a large proportion of patients in this study had presurgical cognitive impairment (63%).9 Deiner et al. used the CAM-ICU instrument in the postanesthesia care unit to assess delirium, meaning that the detection rate for delirium was limited. Also, this study9 was conducted in the U.S., so the characteristics of its participants might differ from those of patients in the studies conducted in China.3,15,16,17,18,22 Because of all these factors, Deiner et al. drew the opposite conclusion from the other studies. In addition, three studies15,16,18 were assessed as having unclear risk of selection, performance, and detection bias, while four others3,9,17,22 were evaluated as having low risk of bias. The types of surgery that patients underwent differed among these studies in this subgroup, and so they also might contribute to this heterogeneity.

Additionally, subgroup analysis for different control drugs (NS or anesthetic drugs) revealed that the DEX group had a significant decrease in POD incidence compared with the control group. Dexmedetomidine is thought to have intrinsic delirium-reducing properties, based on its multiple characteristics.29 Also, propofol, midazolam, and morphine have been shown to increase the risk of POD.25 Previous studies have similarly revealed that DEX was associated with reduced POD incidence when compared with anesthetic drugs or NS.29,32 The study by Deiner et al.9 contributed heterogeneity here as well, mild in this case (I2 = 45%; P = 0.08) for the NS subgroup. Therefore, more high-quality studies with large sample size are needed to confirm this result.

The most commonly reported adverse events of DEX are hypotension and bradycardia, which are associated with an α2 adrenoreceptor agonist mechanism.33,34 It was not surprising that the incidences of hypotension and bradycardia were higher in the DEX group in our meta-analysis. In addition, we found that DEX shortened the time to extubation and the length of hospital stay. Conversely, one meta-analysis32 found no significant difference in length of hospital stay between DEX and control groups. Nevertheless, there were limited data and substantial heterogeneity among studies, which would contribute to unreliability of our pooled results.

There are several overall limitations to our meta-analysis. First, the sample sizes of some studies included in our meta-analysis are small, so there is a high risk of overestimating treatment effects. Second, some excluded studies might have reported adverse events and important clinical results; therefore, we could not acquire complete data to perform the meta-analysis of adverse events and some clinical results. Third, surgery types and DEX regimens vary widely among the included studies. These variables might have produced the clinical heterogeneity that influenced the results, but unfortunately such variables could not be excluded. Fourth, the majority of non-cardiac surgical patients in the included studies did not routinely require postoperative ICU care or long-term maintenance of MV. Current guidelines suggest sedation with DEX for a relatively long time in adult ICU patients on MV.35 The patients in four studies3,15,16,22 admitted to the ICU received DEX for at least 12 hr. Nevertheless, Deiner et al. infused DEX for only two hours in the postanesthesia care unit. Thus, the validity and generalizability of DEX in non-ICU patients that was not explored in our meta-analysis should be further evaluated. Finally, the maintenance dose (0.2–0.7 μg·kg−1·hr−1) in some studies was not fixed and may have been titrated based on hemodynamic changes or sedative scores, while in other studies the rate of infusion was maintained by fixed dose (0.1, 0.2, 0.4 or 0.5 μg·kg−1·hr−1). Because the dose of DEX varied in these studies, we could not draw from them an accurate conclusion about what dose would be optimal.


In summary, the administration of dexmedetomidine during the intraoperative or postoperative periods reduces the incidence of POD in elderly patients after non-cardiac surgery, but this benefit comes at the cost of an increased incidence of hypotension and bradycardia. Further high-quality large randomized clinical trials are needed both to confirm the validity of our results and to determine the optimal dose and timing of dexmedetomidine administration.