Perioperative ketamine administration to prevent delirium and neurocognitive disorders after surgery: a systematic review and meta-analysis

Abstract

Background:

Surgery induces high rates of cognitive disorders, persisting for up to 12 months in elderly adults. This review aimed to assess the currently debated preventive effect of perioperative ketamine on postoperative delirium and postoperative neurocognitive disorders (POND).

Materials and methods:

Systematic review and meta-analysis including all randomized controlled trials investigating the effects of perioperative ketamine administration in adult patients compared to placebo or no intervention on postoperative delirium and/or POND between January 2007 and April 2022. Database searches were conducted in PubMed, Medline, Embase, Scopus, and Central. Random effects models were used to pool overall estimates. The GRADE approach was used to assess the quality of the evidence.

Results:

From 1379 records screened, 14 randomized controlled trials with 1618 patients randomized met our inclusion criteria with a high level of consensus among reviewers, amongst whom 50% were at low-moderate risk of bias. There was no between-group difference in postoperative delirium [8 trials, 1265 patients, odds ratio (OR) 0.93, 95% CI (0.51–1.70), I ²=28%] and POND [5 trials, 494 patients, OR 0.52, 95% CI (0.15–1.80); I ²=78%]. There was no significant between-group difference in postoperative psychological adverse effects, level of pain, hospital length of stay, or mortality. Between-group subgroup analyses showed no difference in delirium or POND incidence according to surgical setting, ketamine dose, mode of administration, combination or not with other drug(s), and assessment timing or definition of cognitive disorders.

Conclusion:

Perioperative ketamine does not prevent postoperative delirium or POND. Significant study heterogeneity suggests that standardized measures for POND assessment and a specific focus on patients at high risk for POND should be used to improve the comparability of future studies.

Introduction

Highlights

• Perioperative ketamine infusion does not prevent postoperative delirium or cognitive decline.

• High heterogeneity between studies highlights the need for standardized evaluations of cognition.

• Perioperative intravenous administration of ketamine is safe.

Postoperative neurocognitive dysfunction accounts for one of the major complications of surgical procedures¹. They are associated not only with an increased mortality, but also with increased risk of premature leave of the labor market and dependence on social transfer payments for active people². The definition of postoperative neurocognitive dysfunction has evolved over time and now encompasses two components: immediate postoperative events such as delirium, which occurs in 5–50% of elderly patients³, and postoperative neurocognitive disorders (POND), which include delayed neurocognitive recovery observed up to 1 year after surgery and reported in 25–55% of patients undergoing major procedures⁴. Currently, there is no treatment available for postoperative neurocognitive dysfunctions, emphasizing the need for identifying specific preventive interventions.

Considering that one of the main risk factors of POND is age, and the increasing prevalence of surgical procedures in the elderly, who currently account for 25–30% of the 234 million major surgeries performed worldwide each year⁵, expected to quadruple in the next 40 years, it appears urgent to highlight pharmacological approaches that can prevent or reduce postoperative cognitive disorders.

Several arguments favor the potential protective effects of ketamine on delirium and POND. A neuro-inflammatory process driven by microglial cells has been recently hypothesized to be the main pathophysiological mechanism leading to POND⁶. Ketamine, an N-methyl-D-aspartic acid (NMDA) receptor antagonist, displays well-established neuroprotective effects, including the prevention of excitotoxic injury and modulation of microglial reactivity^7,8. In experimental⁸ and clinical⁹ studies, subanesthetic administration of ketamine reduces postoperative markers of inflammation, postoperative pain, and opioid consumption. Ketamine has been repeatedly shown to be an innovative and effective drug for treating psychiatric conditions, such as major depressive disorders, in which a key role for microglial cell activation has been demonstrated^8,9.

However, despite an increasing number of recently published major clinical trials, there is still no consensus on the preventive effects of ketamine on postoperative cognitive impairment. A meta-analysis published in 2018¹⁰ reported that the use of ketamine during anesthesia induction would reduce the risk of the occurrence of postoperative cognitive impairment, but without decreasing the incidence of delirium. Since then, several major clinical trials have been published but with contradictory results. For this reason, an update on the potential preventive effect of ketamine on POND seemed justifiable.

The main objective of this meta-analysis was to assess the preventive effect of perioperative ketamine administration compared to placebo or nonintervention on postoperative delirium and neurocognitive disorders separately, in elective surgeries, using all new additional data. Secondary objectives were to evaluate the occurrence of noncognitive adverse events associated with perioperative ketamine administration in these settings.

Materials and methods

Protocol publication

This systematic review and meta-analysis, conducted in accordance with the Cochrane Collaboration methods¹¹, was reported using the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) list¹² (Supplemental Digital Content 1, http://links.lww.com/JS9/A793), (Supplemental Digital Content 2, http://links.lww.com/JS9/A794), the AMSTAR (Assessing the methodological quality of systematic reviews) guidelines¹³ (Supplemental Digital Content 3, http://links.lww.com/JS9/A795) and registered in PROSPERO (International Prospective Register of Systematic Reviews).

Eligibility criteria

We included all randomized controlled trials (RCTs) that investigated the effect of perioperative ketamine administration in adult patients (≥18 years old) undergoing any type of elective surgery under general anesthesia in comparison to placebo administration or nonintervention and reporting postoperative delirium and/or neurocognitive disorders as primary or secondary outcomes. No restriction was applied for surgical settings, ketamine doses, delivery scheme, and duration of administration. We did not apply any language restrictions in our search strategy, provided that the titles were in English.

Outcome measures

The primary outcomes were the incidences of postoperative delirium and of POND, explored separately.

Delirium was defined using well-recognized scales, namely the Confusion Assessment Method (CAM), the Confusion Assessment Method for the Intensive Care Unit (CAM-ICU), and the Intensive Care Delirium Screening Checklist (ICDSC) that meet the criteria of the DSM-5¹⁴.

The definition of POND has evolved over time, and is characterized by a decline in cognitive functions as measured by preoperative and postoperative cognitive tests analyzing multiple domains of cognition such as the Mini Mental State Examination (MMSE) or Montreal Cognitive Assessment (MoCA), while others used tests that focused on specific cognitive domains such as Trail Making Test A (TMTA), TMT B, verbal, or the Backward Digit Span test for executive function evaluation; all of which were considered in our meta-analysis.

Secondary outcomes included ketamine-related noncognitive adverse events, such as nausea and vomiting, seizure, anaphylaxis, purposeless movements, and transient respiratory depression and psychological adverse effects (anxiety, restlessness, uneasiness, purposeless movement/agitation, hallucinations, and nightmares). Mortality, length of hospital stay, and postoperative pain, evaluated using the patient-reported Visual Analog Scale (VAS) or the numeral rating scale (NRS) in the first 24 h after surgery, were also considered.

Search strategy

We conducted an exhaustive search of data published between 1 January 2007 and 11 April 2022, and available in the following databases: PubMed, Medline, Embase, Scopus, and Cochrane (CENTRAL). An additional search was performed to retrieve gray literature in the International Clinical Trials Registry Platform and ClinicalTrials.gov. The starting date was 2007, firstly because the oldest study included in the most recent meta-analysis was published in 2007, and secondly in order to focus on the last 15 years of clinical practice^10,15.

We initially conducted a preliminary keyword search in PubMed to gather an overview of the literature on perioperative ketamine and cognitive dysfunction/delirium. Following this, a more comprehensive search strategy was implemented using a combination of medical subject headings (MeSH) terms and text words to encompass all relevant topics. The final search was conducted in four databases for the period between 12 April 2022, and 18 April 2022, by two researchers. The detailed equation of the MeSH terms used can be found online. Additionally, we performed a meticulous manual search of the reference lists to identify any additional studies of relevance that may have been missed during the electronic search process.

Study selection process

Publications were selected using COVIDENCE systematic review software (Veritas Health Innovation Ltd.)¹⁶. Once duplicates had been automatically eliminated using COVIDENCE, two independent authors (S.F. and B.D.) screened and selected studies based on titles, abstracts, and keywords to determine their eligibility in terms of interventions, populations, study methods, and outcomes relevant to the review question and specific objectives, while remaining blind to the findings of the other reviewer. Any disagreement was discussed. If no consensus was reached, the eligibility of the studies was decided by a third reviewer (F.V.). Then a full-text review was completed. Interventions, populations, study methods, and outcomes relevant to the review question were reviewed, and a final decision was made about whether to include these studies or not. S.F. and B.D. completed this step independently for all studies. As in the previous step of the research, the included studies were compared and discussed before being assessed by the third reviewer if no consensus was reached (F.V.).

Data extraction and the risk of methodological biases

Data from each article were extracted independently by S.F. and B.D. and blindly reviewed by F.V. These data included the following: first author, year of publication, journal, country, trial period, study population, baseline patient characteristics (age, sex), number of patients, type of surgery, duration of surgery, modalities of administration of the intervention, delirium assessment timing and method, POND assessment timing and method, and duration of follow-up. The quality assessment of the included studies was performed according to the second version of the Cochrane risk-of-bias tool for RCTs (RoB 2.0)¹⁷ and graphically presented using ROBVIS¹⁸. The seven sections of this assessment, that is random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other biases, followed an algorithm to generate a judgment that can be classified as low, moderate, or high risk of bias. Low risk of bias was reached if the study was deemed to have a low risk of bias across all the seven domains; reservations were discussed if the study was judged to raise some concerns in at least one domain, but not judged to be at high risk of bias in any domain, and high risk of bias if the study was deemed to be at high risk of bias in at least one domain or if it was judged to have some concerns for multiple domains, thus significantly reducing confidence in the result¹⁷.

Statistics

Variables

Delirium and POND were treated as dichotomous variables (yes/no).

If a study described the incidence of cognitive impairment at several time points, we analyzed the data acquired just before patients’ discharge from the hospital. Indeed, while no consensus has been established regarding the timing of cognitive assessments after surgery, several studies have suggested that the optimal timing for cognitive assessment is closest to hospital discharge to limit confounding factors^19,20.

All the other studied variables were explored as dichotomous, except for pain scores and length of stay, which were analyzed as continuous variables.

Forest plots

We used a random effects model to calculate all estimates for the incidences of postoperative delirium and of POND in this clinical setting because we anticipated high heterogeneity between trials. To combine results, the Mantel–Haenszel method was used. Results were expressed as an odds ratio (OR) with a 95% CI. Comparisons for rare events were made using the Peto OR method.

Statistical framework

To assess the inter-rater reliability of the screening process, we calculated Cohen’s kappa coefficient using R software. We applied Cohen’s kappa to evaluate the agreement between the reviewers during both stages of the screening process: the initial screening of titles and abstracts and the subsequent screening of full-text articles.

For dichotomous data, the ORs with 95% CI were calculated using random-effect models. As pain scores were measured using different scales between trials, we calculated the standardized mean difference (SMD) for this variable; the mean difference (MD) was calculated for length of stay. Using RevMan 5.3 software, all available data were represented by forest plots.

The χ ² test and the I ² statistic were used to assess heterogeneity. Cochrane guidelines were followed for heterogeneity grades: a I ²<25% represented low heterogeneity, between 25 and 50% moderate heterogeneity, between 50 and 75% substantial heterogeneity, and >75% considerable heterogeneity. A P-value <0.05 was considered statistically significant.

We assessed publication bias with funnel plots and statistical software R to perform the Egger test and trim and fill analysis. The Egger test evaluated the funnel plot’s asymmetry by analyzing the correlation between effect size and precision. A significant P-value from the Egger test would indicate the presence of publication bias. In addition, we conducted a trim and fill analysis to estimate the number of potentially missing studies due to publication bias and to adjust the results accordingly. To address potential heterogeneity across the included studies, we performed subgroup analyses regarding both primary outcomes, according to ketamine dose, mode of administration (single bolus, repeated bolus, or continuous infusion), the timing of measurement and definition of cognitive disorders, and surgical setting.

We conducted a sensitivity analysis focusing on the risk of bias in the included studies. This analysis involved removing studies with a high risk of bias from the dataset for both delirium and POND. The aim of this analysis was to assess the potential impact of these high-risk bias studies on the overall results of our study. High-risk bias studies were identified using the Cochrane Risk of Bias Tool. The results of this sensitivity analysis were presented using forest plots for delirium and POND.

Grading the quality of evidence

To evaluate the quality of evidence for each outcome, we employed the GRADE method. We classified the evidence as high, moderate, low, or very low quality, considering factors such as bias risk, consistency, precision, directness, and publication bias risk²¹. Furthermore, we utilized the GRADEpro software (version 3, McMaster University and Evidence Prime) to generate a summary of findings table. We also graded the level of evidence using the Cochrane risk of bias tool.

Missing data management

If full-text could not be found, or if some data from the studies was missing, the corresponding authors were contacted via e-mail. When the primary outcomes (incidences of postoperative delirium and of POND) were not reported, or reported as continuous outcome (mean), the corresponding authors were contacted to obtain data in binary format, so that the results could be included in this meta-analysis.

Data were considered definitively missing if the corresponding authors did not respond after three e-mails over a 3-month period. For continuous data, if a result was expressed as a median and the author did not provide us with the mean, we used the method given by the Tiejun Tong group²² to calculate the closest approximation to the mean and standard deviation (SD) of the median and interquartile range.

Results

Research strategy

The selection process is summarized in the PRISMA 2020 flow diagram (Fig. 1). A total of 1379 RCTs were identified and, after excluding duplicates and ineligible studies, 14 RCTs with 1618 patients were finally included in the meta-analysis review.

Figure 1.

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analysis) flowchart.

Inter-rater reliability of reviewers

The inter-rater reliability analysis using Cohen’s kappa coefficient yielded the following results: For the titles and abstracts screening step involving 912 studies, the kappa coefficients were 0.79 (unweighted) and 0.79 (weighted). Additionally, for the full-text screening step involving 49 studies, the kappa coefficients were 0.80 (unweighted) and 0.80 (weighted).

The disagreements that occurred between reviewers (S.F. and B.D.) for 12 studies during the ‘title and abstract’ screening and three studies during the ‘full-text’ screening, were efficiently resolved by a third reviewer (F.V.). These disagreements accounted for less than 3% of the total number of studies screened, indicating a high level of consensus among the reviewers.

Missing data

Five corresponding authors were contacted to retrieve missing data^23–27. Only one of these supplied the relevant information²⁷. The rate of missing data is therefore 11.74% (190 patients with missing data,1618 patients included) regarding the primary outcomes of the current meta-analysis.

Study characteristics and risk of bias

Study characteristics are summarized in (SDC, Table 1, Supplemental Digital Content 4, http://links.lww.com/JS9/A796). The manuscripts were all published between 2008 and 2021. Among the 14 studies included in our review, five studies took place in the United States of America^15,28–31 (n=844 patients), four in Europe^24,32–34 (n=479 patients), two in Turkey (n=100 patients), two in China^23,25–27 (n=144), and one in South Korea³⁵ (n=51 patients). The sample size ranged from 24¹⁵ to 654 patients²⁸. Four studies were multicenter RCTs^28,30,33,34 (n=1062 patients) and the 10 others were monocentric. Ten studies were coordinated in university hospitals^{23,24,26–28,30,32–35} (n=1444 patients) and four were led from regional hospitals^15,25,29,31 (n=174 patients). No author reported any conflicts of interest.

The risk of bias assessment for individual studies is presented in Figure 2 and a Cochrane risk of bias table was added to provide more detailed information (SDC, Table 2, Supplemental Digital Content 4, http://links.lww.com/JS9/A796). Among the included trials, six were at low risk of bias^{27,28,30,32–34}, two studies raised some concerns^26,31, five studies^{15,24,25,29,35} were at high risk of bias, and one study²³ lacked sufficient information to draw any conclusions.

Figure 2.

Risk of bias assessment.

The funnel plots for publication bias are shown in (SDC, Fig. 1A, Supplemental Digital Content 4, http://links.lww.com/JS9/A796) and (SDC, Fig. 1B, Supplemental Digital Content 4, http://links.lww.com/JS9/A796) and indicate a low risk of publication bias. The Egger test results showed that there was no evidence of publication bias for either the delirium outcome (P=0.90) or the POND outcome (P=0.58). The trim and fill analysis showed low heterogeneity (I ²=28.1%) with an OR of 0.9610 and a 95% CI of (0.68–1.36) for delirium. It showed high heterogeneity (I ²=84%) with an OR of 0.86 and a 95% CI of (0.20–3.67) for POND. Thus, while there is no evidence of publication bias for the delirium outcome, high heterogeneity was observed for the POND outcome.

The sensitivity analysis revealed that the exclusion of the two high risk of bias studies resulted in a change from an OR of 0.93 [95% CI (0.51–1.70), I ²=28%] to an OR of 1.01 [95% CI (0.71–1.45), I ²=0%] for delirium (SDC, Fig. 2A, Supplemental Digital Content 4, http://links.lww.com/JS9/A796). Removing the high and unknown risk of bias studies and keeping the low and moderate risk of bias studies resulted in an OR of 0.56 [95% CI (0.14–2.27), I ²=85%), which is not significantly different from the original OR of 0.52 [95% CI (0.15–1.80), I ²=78%] for POND (SDC, Fig. 2 B, Supplemental Digital Content 4, http://links.lww.com/JS9/A796).

Population

Population characteristics are summarized in SDC Table 1 (Supplemental Digital Content 4, http://links.lww.com/JS9/A796). The mean age of the patients included in the analysis of main outcomes was 71.1±9 years. It was <65 years in six studies^{15,24,26,27,30,32} (n=351 patients) whereas it was >70 years in four studies^23,28,33,34 (n=1066 patients) and between 65 and 70 years in four others^25,29,31,35 (n= 201 patients). The distribution of surgical procedures in the 14 studies was as follows: orthopedic (3/14, n=336 patients)^15,34,35, digestive (5/14, n=287)^23–26,32, cardiac (2/14, n=110 patients)^29,31, neurosurgery (1/14, n=84 patients)²⁷, ENT surgery (1/14, n=56 patients)³⁰. Two studies included different types of surgery (n=745 patients)^28,33, one with a case-mix of major surgeries (n=654 patients)²⁸ and the other with digestive, orthopedic, vascular, gynecological, cardiac, or thoracic surgeries (n=91 patients)³³. Two studies^29,34 examined the relationship between serum sodium levels, initial hospitalization, and medical history (specifically dementia) in relation to delirium, involving 319 patients.

Intervention

The characteristics of ketamine administration are summarized in SDC Table 1 (Supplemental Digital Content 4, http://links.lww.com/JS9/A796). Ketamine was used at a subanesthetic dose of 0.5 mg/kg in nine studies (n=537 patients)^{23,25,27–31,34,35}, at a dose <0.5 mg/kg in four studies (n=211 patients)^15,24,26,32, and at the dose of 1 mg/kg in three studies (n=801 patients)^28,30,33. In two studies, two doses of ketamine (0.5 mg/kg and 1 mg/kg) were explored separately^28,30. No study evaluated delirium or POND outcomes with perioperative continuous intravenous ketamine injection.

As a comparator, 13 studies used placebo saline 0.9% (n=1594 patients)^23–35. Only one study had no intervention (n=24 patients)¹⁵.

Ketamine or placebo was administered before induction of general anesthesia in two studies^26,35 (n=111 patients), at the time of induction in 10 studies (n=1333 patients)^{15,23,25,28–34}, after induction and before the incision in one study²⁴ (n=90 patients) and after incision (dural opening during neurosurgery) in one study²⁷ (n=84 patients). Two studies also tested the effect of bolus followed by continuous intravenous injection of ketamine during the surgical procedure^15,32 (n=30 patients).

Primary outcome

Five studies examined the incidence of postoperative delirium (5/14, n=857)^{15,27–29,32}, six studies examined the incidence of POND (6/14, n=353)^{23–26,31,35}, and three examined the incidences of both individually (3/14, n=408)^30,33,34.

Delirium was assessed using the Confusion Assessment Method (CAM) in five studies (n=1079)^{15,27,28,30,34} or the Intensive Care Delirium Screening Checklist (ICDSC) in three others (n=186)^29,32,33. The evaluation was performed within a time range of 2 h (n=654)²⁸ to 7 days (n=84)²⁷ after surgery.

Regarding POND diagnosis, the Mini Mental State Examination (MMSE) was used in six studies (n=388)^{23–25,30,33,35}. Five studies combined different tests, including the Montreal Cognitive Assessment (MOCA), Trail Making Test (TMT), Digit Substitution Test (DST), and Cognitive Failures Questionnaire (CFQ), to make a positive diagnosis of POND (n=480)^{23,30,31,34,35}. The evaluations were performed within a time range of 1 day (n=51 patients)³⁵ to 90 days after surgery (n=261 patients)³⁴.

Incidence of postoperative delirium

In total, eight studies collected data on the incidence of postoperative delirium^{15,27–30,32–34} allowing for analysis in 1265 patients (SDC1, Table 1, Supplemental Digital Content 4, http://links.lww.com/JS9/A796). Among these eight studies, only one found a significant reduction in the incidence of postoperative delirium in favor of the ketamine group²⁹; all the others showed no difference^{15,27,28,30,32–35}. Overall, no statistical difference was observed between the ketamine group and the placebo group [OR 0.93, 95% CI (0.51–1.70), I²=28%] (Fig. 3A).

Figure 3.

Forest plot of incidence of postoperative delirium (A) and Forest plot of incidence of postoperative neurocognitive disorders (POND) (B). Random effects model. df, degrees of freedom.

Incidence of POND

Data on the incidence of POND were available from five of nine studies^{23,31,33–35}, allowing for analysis in 494 patients (Fig. 3B). Among these five studies, two found a significant reduction of incidence of POND in favor of the ketamine group^23,31, all the others showed no difference^33–35. Overall, no statistical difference was observed between the ketamine group and the placebo group on the incidence of POND [OR 0.52, 95% CI (0.15–1.80); I ²=78%) (Fig. 3B).

Secondary outcomes

Side effects

No significant difference was found for nausea and vomiting [OR 1.14, 95% CI (0.79–1.64), I ²=22%; five studies, n=1138]^{15,24,26,28,34} (Fig. 4A), respiratory depression [OR 0.54, 95% CI (0.04–7.43), I ²=63%; two studies, n=161]^24,30 (Fig. 4B), and psychological adverse effects [OR 1.19, 95% CI (0.91–1.55), I ²=0%; six studies, n=1260]^{24,27,28,30,34,35} (Fig. 4C).

Figure 4.

Postoperative side effects in randomized controlled trials. Forest plot of nausea/vomiting (A). Forest plot of respiratory depression (B). Forest plot of psychological adverse events (anxiety, restlessness, uneasiness, purposeless movement/agitation, hallucinations, nightmares) (C). df, degrees of freedom.

Pain evaluation

Three studies^28,30,34 reported postoperative pain intensity within the first 24 h after surgery. Outcomes of pain intensity were measured by validated scales, with two studies using the VAS^30,34 and one study using the NRS²⁸. There was no significant difference in pain intensity between ketamine and placebo at postoperative day (POD)1 [MD 0.06, 95% CI (−0.33–0.44), I ²=0%; n=1035 patients, three studies]. The forest plots of this analysis are presented in SDC Figure 3A (Supplemental Digital Content 4, http://links.lww.com/JS9/A796).

Length of stay

Four studies measured the impact of intravenous ketamine administration on the total length of hospital stay, with 486 patients analyzed^27,29,31,34 (SDC, Fig. 3B, Supplemental Digital Content 4, http://links.lww.com/JS9/A796). Only one study found a significant reduction in total length of stay²⁷ in favor of ketamine. Overall, no statistical difference was observed between the two groups [MD 0.01 days; 95% CI (−1.14–1.16); I ²= 67%] (SDC, Fig. 3B, Supplemental Digital Content 4, http://links.lww.com/JS9/A796).

Mortality

Two studies explored mortality as a secondary outcome^30,34 (n=363 patients): no difference was found between the two groups [OR 1.19, 95% CI (0.14–9.78), I ²=0%] (SDC, Fig. 3C, Supplemental Digital Content 4, http://links.lww.com/JS9/A796).

Subgroup analysis

According to the ketamine dose

No statistically significant difference according to the ketamine dose was observed between groups regarding delirium incidence: neither for a ketamine dose <0.5 mg/kg [n=61, OR 3.27 95% CI (0.46–23.48); I ²=0%]^15,32, nor for a ketamine dose of 0.5 mg/kg [n=875, OR 0.78, 95% CI (0.32–1.87); I ²=53%]^27–30,34 nor for a ketamine dose of 1 mg/kg [n=563, OR 1.10, 95% CI (0.71–1.70); I ²=0%]^28,30,33 (SDC, Fig. 4A, Supplemental Digital Content 4, http://links.lww.com/JS9/A796).

No statistically significant difference according to the ketamine dose was observed between groups regarding POND incidence: neither for a ketamine dose of 0.5 mg/kg [n=403, OR 0.37, 95% CI (0.08–1.84); I ²=81%]^23,31,34,35 nor for a ketamine dose of 1 mg/kg [n=91, one study³³, OR 1.71, 95% CI (0.56–5.19)] (SDC, Fig. 4B, Supplemental Digital Content 4, http://links.lww.com/JS9/A796).

According to administration mode of ketamine

No statistically significant difference according to mode of administration was found in the incidence of delirium, either in the subgroup of bolus injection [n=1204, OR 0.82; 95% CI (0.42–1.59), I ²=38%]^{27–30,33,34} or in the subgroup of bolus injection followed by continuous injection [n=61, OR 3.27 95% CI (0.46–23.48), I ²=0%]^15,32 (SDC, Figure 5, Supplemental Digital Content 4, http://links.lww.com/JS9/A796).

According to the POND and delirium evaluation score measurement timing and their definitions

No statistically significant difference was found according to the timing of delirium assessment, that is within the 6 first hours following surgery [n=56 patients, OR 2.95 95% CI (0.58–15.07)] or within the first 3 days after surgery [n=1209 patients, OR 0.82 95% CI (0.45–1.51), I ²=24%].

No statistically significant difference was found according to the timing of POND assessment, that is from POD1 to POD3 following surgery [n=151 patients, OR 0.53 95% CI (0.05–6.13), I ²=84%]^23,33 or for POD6 to POD7 [n=1395 patients, OR 0.47 95% CI (0.07–2.97), I ²=82%]^31,34,35.

No statistically significant difference was found according to the score used for the diagnosis of delirium, with CAM [n=1079, OR 1.03, 95% CI (0.71–1.48); I ²=0%]^{15,27,28,30,34} and with the ICDSC [n=186, OR 0.55, 95% CI (0.07–4.46); I ²=63%]^29,32,33 (SDC, Figure 6A, Supplemental Digital Content 4, http://links.lww.com/JS9/A796).

No statistically significant differences were found according to the POND assessment score either with MMSE or MoCA without TMT [n=91, OR 1.71, 95% CI (0.56–5.19)]³³ or with MMSE/MoCA and TMT [n=351, OR 0.64, 95% CI (0.13–3.07); I ²=65%]^23,34,35 (SDC, Fig. 6B, Supplemental Digital Content 4, http://links.lww.com/JS9/A796).

According to the type of surgery

No statistically significant difference according to type of surgery was observed between groups regarding delirium incidence (SDC, Fig. 7A, Supplemental Digital Content 4, http://links.lww.com/JS9/A796), that is in orthopedic surgery [n=285, OR 1.17 95% CI (0.41–3.35), I ²=0%] or case-mix surgeries [n=980, OR 0.83 95% CI (0.35–1.97), I ²=47%] and regarding POND incidence, that is in orthopedic surgery [n=312, OR 0.94 95% CI (0.58–1.54), I ²=0%] or case-mix surgeries [n=203, OR 0.29 95% CI (0.04–2.11), I ²=85%] (SDC, Figure 7B, Supplemental Digital Content 4, http://links.lww.com/JS9/A796).

According to age

No statistically significant difference was observed between groups regarding delirium incidence (SDC, Figure 8A, Supplemental Digital Content 4, http://links.lww.com/JS9/A796) or POND (SDC, Figure 8B, Supplemental Digital Content 4, http://links.lww.com/JS9/A796) in the subgroup analysis according to the age of the patients.

Grading the quality of evidence

Our meta-analysis revealed that the level of evidence for the impact of ketamine on postoperative delirium, pain, mortality, nausea and vomiting, and psychological adverse events, was moderate according to GRADE standards. The level of evidence on POND and length of stay was low, while evidence for respiratory depression was very low. Please refer to Table 1 for detailed GRADE evidence profiles.

Table 1.

Summary of findings in studies comparing ketamine to placebo or no intervention on postoperative delirium and/or POND after surgery.

[Perioperative ketamine] compared to [placebo or nonintervention] for [postoperative delirium and neurocognitive disorders] Patient or population: [postoperative delirium and neurocognitive disorders] Setting: Any type of elective surgery under general anesthesia Intervention: [Perioperative ketamine] Comparison: [placebo or nonintervention]
				Anticipated absolute effects*
Outcomes	No of participants (studies) Follow-up	Certainty of the evidence (GRADE)	Relative effect (95% CI)	Risk with [placebo or nonintervention]	Risk difference with [Perioperative ketamine]
postoperative delirium assessed with: CAM/ICDSC follow-up: range 2 h to 7 days	1265 (8 RCTs)	⊕⊕⊕◯ Moderatea	OR 0.93 (0.51–1.70)	133 per 1000	8 fewer per 1000 (60 fewer to 74 more)
Postoperative neurocognitive disorders (POND) assessed with: MMSE/MOCA/TMT/DST/CFQ follow-up: range 1 day to 90 days	494 (5 RCTs)	⊕⊕◯◯ Lowa , b , c	OR 0.52 (0.15–1.80)	249 per 1000	102 fewer per 1000 (202 fewer to 125 more)
Pain evaluation assessed with: VAS/NRS	1035 (3 RCTs)	⊕⊕⊕◯ Moderatec	–		MD 0.06 VAS more (0.33 fewer to 0.44 more)
Length of stay assessed with: days	486 (4 RCTs)	⊕⊕◯◯ Lowb,c	–		MD 0.01 Days more (1.14 fewer to 1.16 more)
Mortality	363 (2 RCTs)	⊕⊕⊕◯ Moderatec , d	OR 1.19 (0.14–9.78)	6 per 1 000	1 more per 1000 (5 fewer to 50 more)
Side effects: Nausea; Vomiting	1138 (5 RCTs)	⊕⊕⊕◯ Moderate	OR 1.14 (0.79–1.64)	316 per 1000	29 more per 1000 (49 fewer to 115 more)
Side effects: Respiratory Depression	161 (2 RCTs)	⊕◯◯◯ Very lowb , c , d	OR 0.54 (0.04–7.43)	101 per 1000	44 fewer per 1000 (97 fewer to 355 more)
Side effects: Psychological adverse events	1260 (6 RCTs)	⊕⊕⊕◯ Moderatea	OR 1.19 (0.91–1.55)	294 per 1000	37 more per 1000 (19 fewer to 98 more)

^*The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).

^aHigh clinical heterogeneity.

^bHigh methodological heterogeneity.

^cSmall sample size.

^dWide CIs.

GRADE Working Group grades of evidence.

High certainty: we are very confident that the true effect lies close to that of the estimate of the effect.

Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect.

Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

CAM, Confusion Assessment Method; CAM-ICU, Confusion Assessment Method for the Intensive Care Unit; ICDSC, Intensive Care Delirium Screening Checklist Disorders; MMSE, Mini mental State Examination; MoCA, Montreal Cognitive Assessment; MD, mean difference; OR, odds ratio; POND, Postoperative Neurocognitive Disorder; TMT, Trail Making Test.

Discussion

In this meta-analysis, there is no reduction in the incidence of postoperative delirium and POND in patients treated with ketamine perioperatively. Subgroup analyses according to the ketamine dose or mode of administration, to the type of score used to diagnose delirium and POND and to its timing report similar results, with low (delirium) and considerable (POND) heterogeneity. To our knowledge, this meta-analysis is the first to gather data on the effect of ketamine dose and mode of administration on the incidence of delirium and POND. Our methodology rigorously followed the international recommendations for conducting a systematic review and meta-analysis³⁶.

Our results differ from those of the meta-analysis conducted by Hovaguimian et al.¹⁰, published in 2018, which concluded that perioperative administration of ketamine had no effect on postoperative delirium, but appeared to have a protective effect on POND [relative risk (RR) 0.34, 95% CI (0.15–0.73), I ²=17%]. However, the authors themselves recognize that their results should be interpreted with caution, because the evidence at that time remained limited and of low quality. Indeed, in the study by Hovaguimian et al.¹⁰, only 8% of the required information was available and so the risk of spurious results was significant. Consequently, the positive effect of ketamine reported in the previous meta-analysis could have only come about by chance. The discrepancy between these two meta-analyses could also be explained by the major weight of the positive Hudetz et al.³¹ [RR 0.33 95% CI (0.17–0.65)] published in 2009, including 52 patients and accounting for 70% of all the patients included in the meta-analysis from Hovaguimian et al. Our meta-analysis included eight additional studies with 682 patients overall. This greatly improves the precision in the estimate of the effect, with four additional studies focusing on delirium^27,30,33,34 (n=492) and two more on POND^33,34 (n=351). The addition of these studies increases the power of the analysis and equilibrates the weight of each study included.

The largest study included in our meta-analysis is the POCK study³⁴, a French RCT conducted from 2017 to 2019, including 292 patients undergoing major orthopedic surgery. This trial failed to demonstrate a benefit of a 0.5 mg/kg bolus of ketamine at the time of induction of anesthesia on the reduction of POND [OR (95% CI) 0.92 (0.56–1.51), P=0.73] (primary endpoint) seven days after surgery or on postoperative delirium (secondary endpoint) [OR (95% CI) 0.80 (0.26–2.47), P=0.698]³⁴. In contrast, in a very similar field in terms of pathophysiological hypotheses, a recent meta-analysis performed by Wang et al.³⁷ published in 2022, investigated the effect of intravenous ketamine on depressive symptoms after surgery in nine RCTs (n=2468 patients). The authors found that ketamine provided a significant reduction of postoperative depression scale scores, by a SMD of −0.51 [95% CI (−0.99 to −0.04), <0.001, I ²=93%; four studies] on POD3. However, this effect was not found to be sustained after POD3. The method of administration and the ketamine dose are essentially the same with those used in our meta-analysis but there are several notable differences, such as type of surgery (orthopedic and gynecological) or anesthesia (predominance of loco-regional anesthesia), lower postoperative pain and younger population (mean age 40 years), all of which could explain the differences in our results. Additionally, Wang et al. found a significant difference in terms of mean score difference but did not examine the overall prevalence of postoperative depression, which also needs to be considered. Interestingly, outside of the perioperative setting, an RCT published in 2019 by Phillips et al.³⁸ concluded that repeated ketamine infusions (six thrice-weekly open-label ketamine infusions) have cumulative and sustained antidepressant effects for treatment-resistant depression. In contrast, in the current meta-analysis, the vast majority of studies tested the impact of a single bolus of ketamine. These results clearly raise the question of what the duration and the repetition of ketamine infusion during the perioperative period should be to achieve a clinically positive impact. In clinical practice, the findings of the current meta-analysis provide compelling evidence in favor of restricting the extensive utilization of ketamine, particularly with regards to mitigating POND and acute pain. Perioperative administration of ketamine should be limited to patients identified as being at risk for postoperative depression or chronic postoperative pain.

In our meta-analysis, the analysis of side effects (nausea, vomiting, and psychological adverse events) showed no significant difference between groups. These results are similar to those of the meta-analysis by Hovaguimian et al.¹⁰ but contradict those of Wang et al.³⁷ who concluded that ketamine administration resulted in increased risks of nausea and vomiting, headache, and hallucination. A study reported that dysphoric effects of ketamine in analgesic doses administered under general anesthesia are not detectable but are observed in 1/21 patients when ketamine is administered in awake patients without benzodiazepine³⁹. A shorter surgery length, an early assessment of side effects, and the predominance of loco-regional anesthesia could explain the fact that ketamine had increased side effects in the meta-analysis by Wang as compared to the present meta-analysis. Finally, no difference was observed in length of hospital stay and mortality in our study. These results confirm that overall, perioperative intravenous ketamine administration is safe.

Missing data is a common limitation in meta-analyses that can affect the validity and generalizability of the results. In our meta-analysis, the missing data were mostly reported in studies that did not clearly define patients with POND³, which prevented their inclusion in the current meta-analysis. Future studies should aim to improve reporting and minimize missing data to enhance the quality and reliability of systematic reviews and meta-analyses. Another limitation of our study is the limited information available for younger patients (between 18 and 60 years old) with delirium or postoperative neurocognitive dysfunction (POND). Only one study including 60 patients out of the 1618 patients included in our analysis concerned young patients. Most studies in our analysis primarily focused on elderly people, restricting the generalizability of our findings to a broader population. Furthermore, the mean age of the patients included in the analysis of main outcomes was 71.1±9 years, indicating a significant emphasis on older patients. Indeed, age is clearly recognized as an important risk factor for postoperative cognitive disorders. To achieve a more comprehensive understanding of delirium and POND across different age groups, future research should include a large range of ages.

Other potential limitations should be considered when interpreting the results from this meta-analysis. First, there were clinical and methodological heterogeneities regarding ketamine dose and mode of administration. We conducted analyses of three subgroups representing, respectively, a dose <0.5 mg/kg, 0.5 mg/kg, and 1 mg/kg, which did not allow for a focus on dose or administration effect, while it is now recognized that the microglial effect is dependent on the injected dose of ketamine, useful for the treatment of depression⁸. Similarly, subgroup analysis according to the mode of administration of ketamine did not show any difference in our study, but it seems necessary to investigate this hypothesis further in the future. The recommended dosage of 0.5 mg/kg or less⁴⁰ by the French Society of Anesthesiology and Intensive Care Medicine is meant to prevent chronic postoperative pain, but does not seem to be effective in achieving the goal of preventing delirium or POND. Second, there were also clinical and methodological heterogeneities regarding the tools for assessing POND. All studies included in the present meta-analysis were designed before 2018, when the new Recommendations for the Nomenclature of Cognitive Change Associated with Anesthesia and Surgery³ was released. Since 2018, POND has been defined by an objective decline in cognition as measured by neuropsychological testing, usually by associating multiple tests. The criteria for the occurrence of POND are now a decline of 1–2 SDs from a group of reference for mild POND and more than 2 SDs for the occurrence of major POND³. These two specific points may explain the value of the I ² which showed a considerable heterogeneity (I ²>75%) of studies for the evaluation of POND. Our subgroup approach partially reduced heterogeneity between studies without demonstrating any benefit of ketamine. Third, all the included studies, except the POCK³⁴ and PODCAST studies²⁸, are composed of small numbers of patients, leading to a reduced statistical power. Eventually, the level of evidence for the impact of ketamine on all outcomes ranged from moderate to very low according to GRADE standards, limiting the possibility of definitive conclusions.

We acknowledge the importance of further research to uncover the underlying factors contributing to the limited impact observed with ketamine on these outcomes. Subsequent studies could explore several potential factors that might influence the efficacy of ketamine, including patient demographics, comorbidities, and other perioperative variables. Such information has the potential to identify specific patient populations that may experience the greatest advantages from ketamine administration, paving the way for targeted interventions, and improved patient care. Moreover, several large RCTs using standardized diagnostic tools and focusing on patients at risk for developing POND seem to be indispensable to decrease the risk of overestimation or underestimation of true treatment effects.

In conclusion, this meta-analysis did not support the use of perioperative ketamine administration to prevent postoperative delirium and neurocognitive disorders. This conclusion was reached regardless of the dose, the administration mode used or the surgical setting and contrasts with previous meta-analyses on the same topic although these included fewer patients. However, moderate to low levels of evidence and significant heterogeneity between studies highlight the need for more standardized evaluations of neurocognitive disorders and for further clinical trials focusing on patients at high risk for developing POND. Finally, our findings revealed no significant effects of ketamine on postoperative psychological adverse events, pain levels, hospital length of stay, or mortality, suggesting that ketamine is safe for use in the perioperative setting.

Ethical approval

Review was prospectively registered on Prospero CRD42022324856. As it is a review, no ethical approval is required.

Sources of funding

None.

Author contribution

S.F.: methodology, software, formal analysis, investigation, and writing – original draft; B.D.: methodology, investigation, formal analysis, and writing – original draft; A.C.: writing, review, and editing; M.-P.B.: supervision ,validation, writing – review and editing; F.V.: conceptualization; methodology, resources, investigation, writing – review and editing, supervision, project administration.

Conflicts of interest disclosure

The authors declare that there are no conflicts of interest.

Guarantor

Franck Verdonk.

Data availability statement

Full data are available upon request to the corresponding author.

Provenance and peer review

Not commissioned, externally peer-reviewed.

Supplementary Material

js9-109-3555-s001.docx ^{(30.5KB, docx)}

js9-109-3555-s002.docx ^{(35.3KB, docx)}

js9-109-3555-s003.pdf ^{(325.3KB, pdf)}

js9-109-3555-s004.docx ^{(4MB, docx)}