Effects of dexamethasone on postoperative cognitive dysfunction and delirium in adults following general anesthesia: a meta-analysis of randomized controlled trials

Background Several studies have investigated the effects of dexamethasone on postoperative cognitive dysfunction (POCD) or postoperative delirium (POD); however, their conclusions have not been consistent. So we conducted a meta-analysis to determine the effects of dexamethasone on POCD/POD in adults following general anesthesia. Methods Cochrane Central Register of Controlled Trials (CENTRAL, 2018, Issue 11 of 12) in the Cochrane Library (searched November 17, 2018); MEDLINE OvidSP (1946 to November 16, 2018); and Embase OvidSP (1974 to November 16, 2018) were searched for randomized controlled trials that evaluated the incidence of POCD/POD following dexamethasone administration, in adults (age ≥ 18 years) under general anesthesia. We used the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) framework to assess the quality of evidence. Results Five studies were included (three studies/855 participants in dexamethasone group and 538 participants in placebo group for the incidence of POCD, and two studies/410 participants in dexamethasone group and 420 participants in placebo group for the incidence of POD). There was no signicant difference between the dexamethasone group and the placebo group in terms of the incidence of POCD in 30 days after surgery (RR 1.00; 95% CI [0.51, 1.96], P = 1.00, I2 = 77%) or in the incidence of POD (RR 0.96; 95% CI [0.68, 1.35], P = 0.80, I2 = 0%). However, both analyses had some limitations since evidence remains limited and clinical heterogeneity, and we considered the quality of the evidence for the postoperative incidence of POCD and POD to be very low. Conclusions This meta-analysis revealed that prophylactic dexamethasone did not reduce the incidence of POCD and POD. Trials of alternative prevent strategies for POCD/POD, and a better understanding of the pathophysiology of those complex syndrome, are still needed to make progress of this eld. Trial registration number: This study is registered with PROSPERO, 23 October 2018, number CRD42018114552. Our meta-analysis is the rst systematic review to conrm the effects of dexamethasone on the incidence of POCD/POD in adults following general anesthesia. This meta-analysis revealed that prophylactic dexamethasone did not reduce the incidence of POCD and POD. Trials of alternative prevent strategies for POCD/POD, and a better understanding of the pathophysiology of those complex syndrome, are still needed to make progress of this eld.


Background
Postoperative cognitive dysfunction (POCD) and postoperative delirium (POD) are neuropsychological disorders that can occur following the administration of general anesthesia. POCD is one of the most common complications in both young and elderly patients [1]. Evered et al reported that the incidence of POCD at day 7 post-surgery was 17% for total hip joint replacement surgery, and 43% for coronary artery bypass graft (CABG) surgery; The incidence of POCD at 3 months post-surgery for both groups combined was 17% [2]. POCD is subtle and can only be detected by several neuropsychological tests, which are performed before and after surgery [3]. POD is a transient disturbance of a patient's consciousness, attention, cognition and perception, which can last from a few hours to a few days and can uctuate in severity [4]. POCD and POD are serious complications that are associated with longer hospital stays, delayed functional recovery, decreased quality of life, an increased risk of further complications, and mortality [5][6][7].
Unfortunately, there are still many gaps in our knowledge regarding the pathophysiology of POCD and POD, which hinder our attempts in their prevention and treatment. At present, the prevention of POCD/POD is mainly based on non-pharmacological measures and comprehensive geriatric assessments focusing on risk factors [8,9]. But the high prevalence of POCD/POD persists despite these attempts, and giving the lack of human resources in hospitals, the idea that a medication could prevent POCD/POD is interesting and potentially time-saving. There is growing evidence that the brain's reaction to a peripheral in ammatory process may play a role in the development of POD/POCD. Dillon et al found that elevated preoperative and postoperative levels of C-reactive protein are associated with POD [10]. A recent meta-analysis has suggested that high concentrations of in ammatory markers in peripheral circulation and cerebrospinal uid (CSF) are associated with POCD/POD [11]. Dexamethasone is a long-acting glucocorticoid widely used in various in ammatory diseases [12,13]. If the in ammatory response plays a role in the occurrence of POCD/POD, then inhibiting the in ammatory response through dexamethasone may prevent POCD/POD. In addition, several studies have investigated the effects of dexamethasone on POCD/POD; however, their conclusions have not been consistent. Therefore, we have conducted a meta-analysis to evaluate the effects of prophylactic dexamethasone administration on the incidence of POCD/POD. Although the focus is on older adults after surgery, we concern that the existing study was too limited; hence, the search included all adults under general anesthesia.

Methods And Materials
We adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [14], and we used the Grading of Recommendations, Assessment, Development and Evaluations (GRADE) framework to assess the quality of evidence [15]. This study is registered with PROSPERO, 23 October 2018, number CRD42018114552.

Eligibility and exclusion criteria
We selected all of the studies that met the following eligibility criteria: (1) randomized controlled trials (RCT); (2) adults (≥ 18 years old) who underwent general anesthesia; (3) perioperative administration of intravenous dexamethasone in order to prevent POCD/POD (including administration during the preoperative, intraoperative, and postoperative periods) versus no interventions (no drug administered or placebo group), regardless of the dose administered; (4) the incidence of POCD/POD as a primary or secondary outcome; and (5) availability of the full text in English. We excluded studies in which administration of another drug was used in the control group, dexamethasone was administered by another route, and those with no available assessment tools to evaluate the incidence of POCD/POD.

Search strategy
We performed a systematic search of the Cochrane Central Register of Controlled Trials (CENTRAL, 2018, Issue 11 of 12) in the Cochrane Library (searched November 17, 2018); MEDLINE OvidSP (1946 to November 16, 2018); and Embase OvidSP (1974 to November 16, 2018). The search strategy is shown in Appendix 1. We also manually searched the references of the included studies and reviews for additional studies. The following sources of ongoing and unpublished trials were screened: www.controlled-trials.com and clinicaltrials.gov.

Endpoints
Primary outcomes: The incidence of POCD or POD according to the author's own de nition; however, there is a need for an objective assessment tool for POCD/POD. Secondary outcomes: (1) all-cause mortality at 30 days; (2) any postoperative complications; (3) the level of Creactive protein (CRP) measured within the rst 24 hours postoperatively; and (4) the duration of hospitalization (measured in days) and length of time in the intensive care unit (ICU) (measured in hours).

Study selection
After importing the search results into EndNote X9, two review authors (Li and Wang) independently screened the reports according to the predetermined inclusion criteria. Firstly, duplicate reports were removed, and the studies selected on the basis of the title and abstract. Subsequently, the full text was screened for compliance with the inclusion criteria. Disagreement between Li and Wang was resolved through discussion and consensus with a third reviewer (Fang).

Data extraction and assessment of risk of bias in included studies
Li and Wang extracted data independently from eligible studies, using a pre-designed form. Disagreement between the two review authors was resolved through discussion and consensus with a third reviewer (Fang). Two review authors (Li and Wang) independently assessed risk of bias for each included study using the Cochrane Collaboration's tool [16]. We assessed each study according to the following seven domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other biases. We rated the overall risk of bias of a study as low if all of the domains were "low risk," and high if one or more of the domains were identi ed as "high risk" or "unclear risk."

Synthesis of results
We used the Cochrane Review Manager 5 (RevMan5) for statistical analysis of the data. Only primary and secondary outcomes that we had de ned in advance were used in our analysis. For those studies that did not report a mean and standard deviation, we did not hand and transformate the date, because we did not know if it was normally distributed [17,18]. Dichotomous variable data (e.g., incidence of POCD/POD, all-cause mortality at 30 days, and any postoperative complications) were expressed as relative risk (RR) with 95% con dence intervals (CI), while continuous data (e.g., the level of C-reactive protein (CRP) and length of time in hospital and ICU) were reported as weighted mean differences (WMD) with 95% CI.

Assessment of heterogeneity and data synthesis
We used the Chi-square test and calculated the I² statistic to assess the heterogeneity of the studies. If there was a signi cant heterogeneity between the studies (which was de ned as Chi-square P < 0.05, and I² > 60%), we performed a subgroup analysis of cardiac surgery versus noncardiac surgery, low dose dexamethasone (≦ 0.2 mg/kg) versus high dose (> 0.2 mg/kg) to explore clinical heterogeneity. We expected substantial clinical and methodological heterogeneity, so we used the randomized effect model [19]. A P value < 0.05 was considered statistically signi cant. If it was inappropriate to undertake the meta-analysis, we instead carried out a descriptive analysis of the study.

Assessment of publication bias and sensitivity analysis
Where the number of included studies was more than 10, we assessed the risk of publication bias among the included studies based on a funnel plot. We performed the following sensitivity analysis to assess the stability of POCD/POD: (1) excluding reports with a high risk of bias; (2) using different models (the randomized effect model and the xed effect model).

Study selection
Details of the ow of retrieved results and study selection are shown in Figure 1. According to the pre-de ned search strategy, we retrieved 4850 studies. After removing the duplicate studies, we screened the remaining 3230 studies based on the title and abstract. We then readed the full text of 23 studies for further assessment according to the inclusion criteria. No eligible studies were found by manual retrieval. Finally, 5 studies were included in our analysis. After searching the clinical trials registration platform, we identi ed two ongoing studies that will be assessed after they have been completed [20,21]. Table 1 shows the characteristics of the included studies. Four of the included studies involved cardiac surgery [22][23][24][25], while the remaining one involved microvascular decompression surgery [26]. A larger, multicenter placebocontrolled randomized clinical trial was excluded because it did not report the occurrence of POCD, and not all patients had been assessed by the available delirium assessment tools [27]. The participants' mean age in all of the included studies was over 60 years, except for 1 study [26]. Four studies used a two-arm design in comparing dexamethasone to placebo except 1, which used a three-arm design [26]. Three studies assessed the effects of dexamethasone on POCD [22,24,26], while two studies investigated the effects of dexamethasone on POD [23,25]. Dexamethasone doses and the time of administration varied, as shown in Table 2. The de nition and assessment tools for POCD/POD were also different ( Table 2).

Risk of bias in included trials
The risk of bias in included studies is summarized in Figure 2 and Figure 3. One study was identi ed as "low risk" in all domains, three studies had an unclear risk of bias in one of the seven domains due to reporting bias, and one study had a high or unclear risk of bias in ve of the seven domains.

Incidence of POCD
Three studies reported the incidence of POCD, in which data was reported as the number of participants [22,24,26]. Fang et al reported the incidence of POCD at ve days postoperatively [26], Glumac et al at six days postoperatively [22]. Ottens et al assessed the POCD at 1 month and 12 months postoperatively [24], in order to avoid double counting, we only selected the results reported 1 month after surgery Fang et al used neurocognitive tests to assess the incidence of POCD, while Glumac et al and Ottens et al both used a battery of 5 neuropsychological tests, making their de nitions of POCD different. This meta-analysis included 855 participants in dexamethasone group and 538 participants in placebo group, which accounted for 95% of the total 1460 enrolled participants. There was no signi cant difference in the incidence of POCD in 30 days after surgery between the dexamethasone group and the control group (RR 1.00; 95% CI [0.51, 1.96], P = 1.00, I2 = 77%; Figure 4). We judged the quality of the evidence to be very low based on the GRADE framework: (1) two studies had a high risk of bias; (2) the de nitions of the outcome and the assessment tools were different; (3) results were inconsistent; and (4) imprecision of result.

Incidence of POD
Two studies reported the incidence of POD, in which the data was reported as the number of participants [23,25]. Mardani et al reported the incidence of POD at the preoperative day, followed by the rst, second, and third postoperative days, in order to avoid double counting and given that POD typically occurs 2 to 3 days after surgery, we only selected the results reported on the third postoperative day [23]. Sauer et al reported the incidence of the rst 4 postoperative days [25]. Mardani et al used the Mini-Mental State Examination (MMSE) to assess POD [23], Sauer et al used the Confusion Assessment Method (CAM) adapted for the ICU (CAM-ICU) [25]. Meta-analysis showed that there was no signi cant difference in the incidence of POD between the dexamethasone group and placebo (RR 0.96; 95% CI [0.68, 1.35], P = 0.80, I2 = 0%; Figure 5). We deemed the quality of the evidence to be very low because: (1) the two studies had an "unclear risk" of reporting bias; and (2) imprecision of result.

Secondary outcomes
No studies reported the all-cause mortality at 30 days post surgery. One of the ve included studies reported postoperative complications including: deep sternal wound infection, leg infection, sepsis, and pneumonia in addition to cardiac, cerebrovascular, respiratory, and renal complications [23]. The study showed that there was no signi cant difference between the dexamethasone and control groups in postoperative complication rates. We consider the quality of the evidence for this outcome to be very low because of the limited evidence available, and the fact that the study was at high risk of bias. Only one study reported the level of CRP [22], which was measured 12 hours after surgery and on the each of the rst 3 postoperative days; the CRP level was lower in the dexamethasone group compared with the placebo group at all time points (P < 0.001). We deem the evidence for this outcome to be of very low quality based on the limited evidence available.
Two studies reported the duration of hospitalization (measured in days) [22,23]. Our meta-analysis showed that the use of dexamethasone reduced the length of hospital stay (WMD -0.57 d; 95% CI [-1.08, -0.07], P =0.03, I2 = 0%, Figure 6), but the difference was so small that it did not has a clinical signi cance. We considered the quality of the evidence for this outcome to be very low because of the limited evidence available, and imprecision of result.
Three studies reported the length of ICU stay [22,23,25] Figure 7). We deemed the quality of the evidence for this outcome to be very low because of the limited evidence available, and imprecision of result.

Subgroup analysis
For POCD, one study describes noncardiac surgery [26], while two studies describe cardiac surgery [22,24]. We performed a subgroup analysis of cardiac surgery versus noncardiac surgery (as a subgroup of cardiac surgery only). We noted no difference in incidence of POCD between the dexamethasone and control groups when noncardiac surgery was excluded (RR 0.90, 95% CI [0.21, 3.77], P = 0.89, I2 = 87%; 439 participants, Figure 8). We also found no signi cant difference between subgroups (P = 0.73) as shown in Figure 8. For POD, the participants of all studies underwent cardiac surgery.
For POCD, two studies administered a low dose of dexamethasone (≦ 0.2 mg/kg) [22,26], one study administered dexamethasone at a dose of 1 mg/kg (maximum 100 mg) [24]. We also performed a subgroup analysis of low dose (≦ 0.2 mg/kg) versus high dose (> 0.2 mg/kg) dexamethasone (as a subgroup of low dose ≤ 0.2 mg/kg only). This subgroup analysis showed no signi cant difference in the incidence of POCD between the dexamethasone of low dose (≦ 0.2 mg/kg) and control groups (RR 0.76, 95% CI [0.29, 1.98]; 1115 participants, Figure 9). We noted no signi cant difference between subgroups (P = 0.14) as shown in Figure 9. For POD, one study administered 8 mg dexamethasone before induction of general anesthesia, followed by 8 mg every 8 hours [23]. In another study, dexamethasone was administered at a dose of 1 mg/kg (maximum 100 mg) [25]; thus, we did not conduct a subgroup analysis.

Assessment of publication bias and sensitivity analysis
Considering that the number of included studies was small, we did not conduct an assessment of publication bias. Based on the prior de nition, there was only one study with a low risk of bias, so we did not conduct the sensitivity analysis based on the risk of bias. Sensitivity analyses using the xed-effect HYPERLINK "E:/Dict/7.

Discussion
This meta-analysis aimed to assess the effects of dexamethasone on POCD and POD in adults following general anesthesia. We found that prophylactic intravenous administration of dexamethasone did not reduce the incidence of POCD in the 30 days following surgery or POD. Furthermore, our meta-analysis showed that the use of dexamethasone reduced the length of hospital stay, but the difference was so small that it did not has a clinical signi cance. Nevertheless, the results of our meta-analysis suggested that the length of ICU stay was shorter in the dexamethasone group than in the placebo group.
Although we included adults aged over 18, the participants' mean age in all of the included studies was over 60 years, except for 1 study [26]. It is likely that older patients are at higher risk of POCD/POD than younger patients, therefore evidence which is related to the incidence of POCD/POD in younger patients (< 60 years old) is still lacking. All studies recruited patients that were scheduled for cardiac surgery except for 1 [26], which included participants that underwent microvascular decompression; this may limit the applicability of the evidence. Consequently, we should be careful not to extrapolate its use to patients undergoing other types of surgery. However, when we excluded this study from the meta-analysis, we found that the direction of the evidence did not change. There are two ongoing studies that were found through the clinical trials registration platform that need to be followed-up once they have been completed [20,21]. This may change the conclusion of this meta-analysis.
Only one study was found to have a "low risk" of bias in all domains [22], while the other four studies had a high or unclear risk of bias in at least one of the seven domains 23,24,25,26]. Two studies did not register their clinical trials or have a published study protocol, so the risks of selective reporting bias were unclear [23,26]. We used the GRADE framework to assess the quality of evidence. We considered the quality of the evidence to be very low for the incidence of POCD because of inconsistencies in results and imprecision of result. The inconsistencies might be explained by differences in types of surgery performed, the dose of dexamethasone administered, and the de nitions and assessment tools used in the diagnosis of POCD. However, we found no reduction in heterogeneity when we conducted the subgroup analyses of cardiac surgery versus noncardiac surgery, and low dose (≦ 0.2 mg/kg) versus high dose (> 0.2 mg/kg) dexamethasone. We also deemed the quality of the evidence for the incidence of POD to be very low because: (1) the two studies were at "unclear risk" of reporting bias; (2) imprecision of the result. As the result relating to the incidence of POD mainly came from one study (weight 98.4%), and the with wide con dence intervals of result.
It is of note that three studies [22,24,26] used neuropsychological tests recommended in the Statement of Consensus on the Assessment of Neurobehavioral Outcomes after Cardiac Surgery to assess the incidence of POCD [28], which has proved to be of immense value in providing a sensitive means of assessing the change and in detecting bene cial results associated with speci c interventions [29][30][31]. The incidence of POCD in the control group of our study is 19%, Johnson found that the incidence of POCD in those aged more than 60 years was 19.2% by using a battery of neuropsychological tests [32], it was comparable with our study. Mardani et al used MMSE and Sauer et al used CAM-ICU to assess the occurrence of POD [23,25]. Unfortunately MMSE scores was not a very reliable diagnostic test of POD, the diagnostic sensitivity and speci city were 96% and 38%, respectively [33]. CAM-ICU had sensitivities of 93-100%, speci cities of 98 to 100%, and high interrater reliability in the detection of delirium [34]. The incidence of POD in control group of our study was 13%. However, the reported incidence of POD varied widely depending on the clinical setting, A systematic review found that the incidence of POD was about 11-51%, and the incidence increased with age [35] , hence, it was di cult to compare the incidence of POD.
It was disappointed that dexamethasone could not reduce the occurrence of POCD/POD in our meta-analysis, particularly since increased studies have shown that in ammatory cytokines are associated with POCD/POD [10,11,36]. The reasons for this phenomenon may be as follows: 1. The occurrence of POCD/POD results from the interaction of many predisposing factors and susceptible factors [37], and a single intervention cannot fully in uence the incidence of POCD/POD; 2. Although there is some evidence that in ammation is an important mechanism for POCD/POD, other factors that have not been identi ed might play a greater role in the development of POCD/POD.
Orena EF et al performed a systematic review to explore the effect of anesthesia on POD, and found that dexamesasone may reduce the risk of POD [38]. However, dexamethasone did not reduce the incidence of POD in our meta-analysis, The reason for this difference was that Orena only included one study [23], where the results of rst postoperative day were exhibited a signi cant reduction of delirium incidence in dexamethasone compared with the control group, but not in the second and third postoperative day. However, the number of patients included in this study was small (43 in the dexamethasone group and 50 in the control group) and the MMSE was used to evaluate the occurrence of POD, which was not a very reliable diagnostic test of POD, and the speci city was just 38% Toner et al assessed the safety of glucocorticoids in noncardiac surgery patients and found no increase in the risk of infection, a clinically unimportant increase in the glucose value, and a lower CRP concentration, but no difference in length of hospitalization [39]. In our meta-analysis, only one study reported the level of CRP, and found it to be lower in the dexamethasone group compared with the placebo group (P < 0.001). Two studies reported the length of hospitalization (measured in days), which revealed that the use of dexamethasone reduced the length of hospital stay (MD -0.57 d; 95% CI [-1.08, -0.07], P =0.03, I2 = 0%), but the difference was so small that it did not has a clinical signi cance.
There are some limitations in this study. Firstly, we conducted a comprehensive search about the effect of dexamethasone on POCD/POD in several databases, however, these conclusions need to be interpreted with caution because of the limited number of evidence and the in uence of the heterogeneity, especially the clinical heterogeneity exist, such as: the type of surgery, the time and dose of dexamethasone administrated, the de nition and evaluation tool of POCD/POD were different, may limit the precision and reliability of the results; Secondly, since POCD/POD was our primary outcome, RCTS that did not contain POCD/POD data were excluded. These studies may evaluation the secondary outcomes, so our meta-analysis of secondary outcomes raises no claim to completeness. Most studies excluded patients at high risk of POCD/POD, so the application of conclusions may be limited.

Conclusion
Our meta-analysis is the rst systematic review to con rm the effects of dexamethasone on the incidence of POCD/POD in adults following general anesthesia. This meta-analysis revealed that prophylactic dexamethasone did not reduce the incidence of POCD and POD. Trials of alternative prevent strategies for POCD/POD, and a better understanding of the pathophysiology of those complex syndrome, are still needed to make progress of this eld.

Funding
There was no funding source in this study.

Availability of data and material
All data generated or analysed during this study are included in this published article.
Authors' contributions LLL participated in the design, collected the data, performed the quality assessment and statistical analyses. ZHZ participated in the design and draft the manuscript. WC and FMD collected the data, performed the quality assessment, and helped to draft the manuscript. XHY, LHL helped to perform statistical analyses and search strategy. All authors read and approved the nal manuscript.
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Competing interests
The authors declare that they have no competing interests. Tables   Table 1 the

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.