The review protocol was registered in the PROSPERO database (https://www.crd.york.ac.uk/PROSPERO, ID: CRD42024567615) and conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines [13].

We systematically searched Cochrane CENTRAL, MEDLINE (via PubMed), and Embase to identify studies examining the effects of intravenous ketamine or esketamine on PONV. The search strategy developed by an experienced medical librarian used synonyms, paraphrases, and free-text terms (Supplementary Table 1). We included studies published after 2000 to reflect the changes in anesthesia practice. In the late 1990s, nitrous oxide was identified as a contributor to PONV, leading to changes in anesthesia protocols [14]. The early 2000s marked the widespread adoption of total intravenous anesthesia with propofol and remifentanil [15], along with the establishment of standardized approaches to PONV research [2] that may have influenced subsequent studies. Studies involving adult patients (aged ≥ 18 years) who underwent general anesthesia for any type of surgery were eligible.

Randomized controlled trials (RCTs), cohort studies, and retrospective studies comparing perioperative intravenous ketamine or esketamine with placebo, opioid-based protocols, or non-opioid-based protocols were included. Studies involving neuraxial anesthesia, ketamine, or esketamine mixed with patient-controlled analgesia; ketamine or esketamine administered in the ward postoperatively; or experimental groups with confounding co-interventions that were not present in the control group were excluded.

The primary outcomes included the incidence of pooled PONV (combined data for PONV and postoperative nausea [PON]), PONV, PON, and postoperative vomiting (POV). Studies lacking incidence data for PON, POV, or PONV were excluded.

The titles and abstracts of the identified records were screened independently by two reviewers (KHS, SCC) using the EndNote software (version X21; Clarivate). Full-text articles were retrieved from the relevant studies. Any discrepancies during the screening were resolved by a consensus or consultation with a third reviewer (JK). The reference lists of the included studies were screened manually to identify additional eligible studies. Email alerts via Embase were established and monitored throughout the review process to capture newly published data.

The primary outcome was the effect of intravenous ketamine and esketamine on the incidence of pooled PONV, derived from studies reporting either PONV as a single endpoint or PON as a separate outcome. In studies that reported PON and POV separately, only PON data were included in the pooled analysis, as vomiting events were likely to overlap with nausea and could lead to overestimation. PONV, PON, and POV were also analyzed separately as individual outcomes. The secondary outcome was the effect of ketamine or esketamine dose on the incidence of pooled PONV. The incidence of adverse effects related to ketamine or esketamine administration, such as hallucinations, dizziness, and drowsiness/sedation, was assessed. In studies reporting multiple postoperative time points, the earliest time point was selected to evaluate the immediate effects of the intervention.

Data extraction was independently performed by two reviewers (KHS, JK) using a standardized pre-piloted form. The extracted information included study details (author, country, year of publication, and study design), patient demographics (mean age, sex distribution, and type of surgery), intervention details (type of ketamine, dosage, route of administration, and timing), and outcome data (incidence of PON, POV, and PONV, along with adverse effects).

In studies with multiple experimental groups, each group was analyzed separately from the control group. When ketamine/esketamine was coadministered with other anesthetic agents in the control group, it was categorized as a placebo in the subgroup analysis. For example, Aveline et al.’s group KM (morphine 0.1 mg/kg + ketamine 0.15 mg/kg) vs. group M (morphine 0.1 mg/kg), Ghadami et al.’s group PK (oral pregabalin 300 mg + ketamine 0.3 mg/kg) vs. group P (pregabalin 300 mg), and Lee et al.’s group KD (dexamethasone 0.1 mg/kg + ketamine 0.5 mg/kg) vs. group D (dexamethasone 0.1 mg/kg) were categorized as placebo in the subgroup analysis [7,16,17]. However, if another agent was administered simultaneously with ketamine in the experimental group and if that agent was not administered in the control group, it was excluded from the analysis. For example, in Ghadami et al.’s study, the subgroup receiving combined pregabalin and ketamine (group PK) was excluded from the analysis comparing it to the saline group to avoid the confounding effects of multiple drug interactions [16]. Studies comparing ketamine plus propofol to propofol alone during induction were classified under the non-opioid group, as the propofol-only arms received an increased propofol dose to match the sedative effects of ketamine [18–20]. For studies with multiple interventions or control groups, the patient numbers were proportionally allocated for comparison with each corresponding group to prevent repetition. However, if there were fewer than five or the total group size was less than 30, we used duplicate comparisons without proportional division, except when proportional division resulted in integers [6,11,21–26]. We counted the number of patients receiving PONV treatment as the incidence of PONV and emesis as vomiting in the study by Vallejo et al. [27]. Delusions were counted as hallucinations in a study by Vosoughin et al. [28].

The risk of bias in the RCTs was assessed for PONV outcomes using the Cochrane Collaboration’s Risk of Bias 2 (ROB2) tool [29]. Observational and retrospective studies were evaluated using the Cochrane Collaboration’s risk of bias in non-randomized studies of interventions (ROBINS-I) [30]. The paired reviewers (KHS, SCC) evaluated the risk of bias, and any discrepancies were resolved through discussion or consultation with a third reviewer (JK). Studies rated as high-risk in one domain or unclear in two or more domains were classified as having a high overall risk of bias for RCTs. For non-RCTs, the overall risk of bias was determined by the highest domain rating, with any ‘Serious’ or ‘Critical’ rating resulting in a high risk of bias.

The quality of evidence for each outcome was assessed using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) framework to evaluate the domains of risk of bias, inconsistency, indirectness, imprecision, and publication bias [31]. The outcomes were rated as high, moderate, low, or very low certainty.

Meta-analyses were conducted using R software, version 4.4.1. The DerSimonian-Laird random-effects model was used to account for heterogeneity across studies, whereas fixed-effect models were employed for comparison.The risk ratios (RRs) with 95% CIs were calculated for pooled effect sizes. For zero-incidence cases, a continuity correction of 0.5 was applied to enable meta-regression analysis.

For all analyses, except for adverse effects, we conducted subgroup analyses to explore the potential heterogeneity among studies by the type of control group: placebo (saline), opioid, or non-opioid adjuvant. These subgroups were analyzed separately to determine whether the effects of ketamine/esketamine on PONV differed according to the comparator.

Meta-regression was performed to analyze changes in the log RR of pooled PONV as a function of the total administered dose (mg/kg) to evaluate the dose-dependent effects of ketamine and esketamine. For studies with continuous drug infusions, the total infused doses were calculated based on the mean anesthesia time, and the doses were standardized by the average patient body weight, where applicable. Additionally, a meta-analysis was conducted to divide esketamine and ketamine into finer dose categories to determine the effective dose for preventing PONV. We categorized studies into three dose groups in each drug for pooled PONV: low-dose (ketamine ≤ 0.4 mg/kg, esketamine ≤ 0.2 mg/kg), intermediate-dose (0.4 mg/kg < ketamine ≤ 0.8 mg/kg, 0.2 mg/kg < esketamine ≤ 0.5 mg/kg), and high-dose (ketamine > 0.8 mg/kg, esketamine > 0.5 mg/kg). To minimize potential heterogeneity, additional meta-regression and dose-category subgroup analyses were conducted using only placebo-controlled studies. This approach aimed at enhancing consistency by reducing variability introduced by different control groups. Statistical significance for subgroup analyses was defined as P < 0.05, unless otherwise specified.

For each outcome, heterogeneity was evaluated using the Q statistic and I² statistic, with significant heterogeneity indicated by a Q P-value < 0.05 or I² > 25%. For subgroup analyses, heterogeneity was assessed using τ² that quantifies between-study variance in a random-effects model. τ² was reported for each subgroup and interpreted as follows: values below 0.04 indicated low heterogeneity, values between 0.04 and 0.16 indicated moderate heterogeneity, and values ≥ 0.16 indicated high heterogeneity.

Funnel plots and Egger’s tests were used to assess the publication bias. Sensitivity analyses (leave-one-out approach) were used to evaluate the robustness of the findings. For outcomes including two or more non-RCTs, sensitivity analyses were conducted by excluding non-RCTs to assess their impact on the overall results.

Our initial database search retrieved 1333 non-duplicate references, of which 301 were excluded by Endnote, leaving 1032 articles for screening. Titles and abstracts were reviewed, and 984 records were excluded from the study. As shown in the Fig. 1, a total of 48 full-text articles were reviewed and 42 studies were included in the initial meta-analysis: 39 RCTs [5,7,11,12,16,18–25,27,28,32–55], one non-randomized intervention study [56], one cohort study [57], and one retrospective study [58]. The remaining six studies were excluded for the following reasons: two lacked incidence data, one did not involve patients undergoing general anesthesia, and three administered ketamine exclusively postoperatively. Following the initial search until August 20, 2024, an additional 13 RCTs were identified and included through supplementary methods: One study via Google Search [59]; three studies through keyword alerts on Embase between August and November 30, 2024 [8,60,61], and nine studies through backward snowballing of the reference lists [6,17,26,62–67]. A total of 55 studies (n = 6676 participants) were included in the final analysis (Fig. 1), and the study characteristics are detailed in Supplementary Table 2.

The opioid control group included morphine [7,40], meperidine [11,22,24,41,51], remifentanil [8], tramadol [12], fentanyl [27], sufentanil [37], and alfentanil [38]. The non-opioid control group comprised dexmedetomidine [5], dexamethasone [17], lidocaine [23,43,61], propofol [18–20], pregabalin [16,34], doxapram [51], magnesium sulfate [44], and diclofenac [28]. The studies included in subgroup analyses for different outcomes are listed in Supplementary Table 3 to provide a structured reference for subgroup comparisons.

The assessment of the risk of bias is summarized in Supplementary Fig. 1. Among the included RCTs, 33 had a low overall risk of bias, whereas 17 and two were rated as having unclear and high risks, respectively. Three non-randomized studies were categorized as having a moderate risk of bias owing to confounding factors, selection of participants [56–58], and bias in the measurement of outcomes [56].

Egger’s regression test and funnel plot analyses were conducted to assess publication bias for pooled PONV, PONV, PON, and POV, and ketamine and esketamine studies (Supplementary Fig. 2). No significant asymmetry was detected for pooled PONV (t = −1.51, df = 84, P = 0.134), PONV (t = −1.87, df = 52, P = 0.068), PON (t = −0.14, df = 30, P = 0.887), or POV (t = 0.63, df = 21, P = 0.537). Similarly, no evidence of publication bias was found for ketamine (t = −1.58, df = 58, P = 0.118) or esketamine (t = −0.16, df = 24, P = 0.876) studies. These findings suggested that publication bias did not significantly affect the results.

The meta-analysis included 53 studies with 86 comparisons (n = 6592, Fig. 2) [5–8,11,12,16–18,20–28,32–55,57–67]. Ketamine and esketamine did not significantly reduce pooled PONV compared to controls (RR = 0.95, 95% CI [0.87–1.04], P = 0.274, I² = 0%). The prediction interval was 0.86 to 1.05.

Subgroup analyses revealed significant differences between the comparator groups (Q = 9.93, P = 0.007). Compared with placebo, ketamine/esketamine did not significantly reduce pooled PONV (RR = 0.94, 95% CI [0.86–1.02], P = 0.114, τ² = 0). In opioid comparisons, ketamine/esketamine showed an RR of 0.69 (95% CI [0.43–1.08], P = 0.107, τ² = 0.234), but this was not significant. In contrast, in non-opioid comparisons, ketamine/esketamine significantly increased the risk of PONV (RR = 1.46, 95% CI [1.03–2.05], P = 0.032, τ² = 0.035).

To determine whether the efficacy of ketamine and esketamine in reducing the incidence of pooled PONV varied according to dose, a meta-regression analysis was performed for each drug (Fig. 4). For ketamine, the 36 included studies administered doses ranging from 0.1 to 1.5 mg/kg [5–8,11,12,16–18,21–24,27,28,32–35,39–41,44,46,49,51,53–55,57,59–63,66], while for esketamine, 18 studies analyzed doses from 0.05 to 1.46 mg/kg [20,25,26,36–38,42,43,45,47,48,50,52,58,63–65,67]. Meta-regression analysis revealed no significant dose-response relationship for ketamine (coefficient = −0.0108, P = 0.965). For esketamine, the meta-regression analysis showed a negative association between dose and pooled PONV, but this relationship was not significant (coefficient = −0.7027, P = 0.153).

A separate meta-regression of placebo-controlled studies was conducted to further evaluate dose-dependent effects. In 29 ketamine studies (0.1–1.5 mg/kg) [5–7,11,16,17,21–24,32–35,39,41,44,46,49,53–55,57,59–63,66], no significant association between dose and pooled PONV was found (coefficient = 0.1702, P = 0.549). Similarly, in 15 esketamine studies (0.125–0.73 mg/kg) [25,26,36,42,43,45,47,48,50,52,58,63–65,67], no significant dose-response relationship was observed (coefficient = −0.2377, P = 0.766).

These findings indicate that no significant dose-response relationship was found for either drug, and further research is needed to clarify whether higher doses of esketamine may have a meaningful impact on PONV reduction.

Adverse effects such as hallucinations, dizziness, and drowsiness were analyzed (Fig. 5).

Leave-one-out sensitivity analysis demonstrated consistent results across all analyses (Supplementary Table 5). For pooled PONV, effect sizes ranged from −0.07 to −0.04, with negligible heterogeneity (I² = 0%). For PONV, the estimates varied between −0.14 and −0.10, showing stability with no significant changes in heterogeneity. PON outcomes remained consistent, with effect sizes between 0.03 and 0.07 (I² = 0%). The POV results showed moderate variability, with estimates ranging from 0.23 to 0.38 and moderate heterogeneity (I² = 25.1% to 27.7%). For ketamine, effect sizes ranged from −0.07 to −0.03 with no heterogeneity (I² = 0%), demonstrating a stable and robust outcome. Esketamine also exhibited consistent results, with effect sizes varying between −0.08 and −0.02 and negligible heterogeneity (I² = 0%), indicating that no single study significantly affected the overall effect. These findings confirm the robustness and reliability of the meta-analysis results.

A sensitivity analysis excluding non-RCTs was conducted to assess their impact on the pooled PONV and POV (Supplementary Table 6). For pooled PONV, the studies by Schotola et al. [57] and Fei et al. [58] were excluded, while for POV, the studies by Moradkhani et al. [56] and Schotola et al. [57] were excluded. The results remained consistent, with effect sizes and heterogeneity estimates showing minimal variation compared with the primary analysis.

The GRADE assessment revealed varying levels of evidence certainty across outcomes (Supplementary Table 7). For pooled PONV and POV, the certainty was low and was downgraded owing to imprecision, risk of bias (two high-risk RCTs and non-RCTs), and inconsistency in subgroup analyses. PONV demonstrated moderate certainty driven by the risk of bias (some studies had some concerns or a high risk). For PON, the certainty was moderate owing to wide CIs crossing the null effect and substantial uncertainty in the subgroup results. Ketamine studies have shown low certainty owing to imprecision, risk of bias, and lack of a clear dose-dependent effect. In contrast, esketamine studies achieved moderate certainty, with high doses showing a trend toward reduced PONV, although imprecision and dose variability remain as concerns.

Ketamine and esketamine did not provide a significant benefit in preventing pooled PONV. However, they significantly reduced the risk of PONV (RR = 0.89, 95% CI [0.80–0.99]) and reduced PONV incidence by 50% in opioid-based comparisons. In contrast, their use was associated with a significantly higher risk of pooled PONV (RR = 1.45, 95% CI [1.03–2.00]) and POV (RR = 4.87, 95% CI [1.09–21.90]) when compared with non-opioid adjuvants. Dose-response analysis found no significant effect of either drug within the evaluated dose ranges (ketamine: 0.1–1.5 mg/kg; esketamine: 0.05–1.46 mg/kg).

The variability in efficacy across comparator groups suggests context-dependent effects. The benefits observed in opioid-based comparisons likely reflect the opioid-sparing properties of these agents, as demonstrated in several studies [7,8,37]. However, compared with non-opioid adjuvants, including propofol [19], dexamethasone [17], dexmedetomidine [5], pregabalin [16], magnesium [44], and diclofenac [28], ketamine and esketamine showed inferior outcomes, with a higher risk of pooled PONV or POV. These non-opioid comparators possess established antiemetic properties and can improve perioperative conditions in ways that indirectly reduce the risk of PONV [3,68], potentially explaining their superior performance over ketamine and esketamine. Notably, several non-opioid comparisons have been conducted in female-only cohorts [5,16,28] or patients undergoing high-risk surgery for PONV [4,19], which represent a high-risk population for PONV. Although the limited number of events and sample sizes in these studies resulted in wide CIs and reduced the overall certainty of the evidence, our findings consistently suggest that non-opioid adjuvants remain a more reliable choice for PONV prevention, particularly in high-risk populations, whereas ketamine and esketamine may offer benefits over opioids.

Ketamine is a racemic mixture of S(+) and R(-) isomers, whereas esketamine consists solely of the S(+) isomer that has an approximately four-fold greater affinity for the phencyclidine site on NMDA receptors [9]. This results in esketamine exhibiting approximately twice the analgesic potency of ketamine [9]. The established subanesthetic doses for analgesia range from 0.1 to 0.5 mg/kg for esketamine and 0.2–0.8 mg/kg for ketamine [4,9], forming the basis for our dose categorization. Doses exceeding these ranges were classified as high dose.

In this study, meta-regression showed no clear dose-response relationship for either drug. Although higher doses of esketamine (> 0.5 mg/kg) were associated with a lower RR for PONV, this association was not significant. Notably, several studies using high dose esketamine have reported significant reductions in perioperative analgesic requirements [26,37,47,50,67]. These findings are consistent with a previous meta-analysis by Hung et al. [69] that reported no significant effect of ketamine on PONV (RR = 1.07; 95% CI [0.78–1.45]) but a meaningful reduction with esketamine (RR = 0.72; 95% CI [0.56–0.92]) across doses ranging from 0.2 to 1.1 mg/kg. The favorable PONV profile of esketamine may be partially attributed to its lower cholinergic inhibition [9]. While dose-optimization for esketamine may be beneficial, its risk of adverse effects at higher doses must be considered in clinical practice.

The dual antiemetic and emetic effects of ketamine and esketamine on PONV stem from their multifaceted pharmacological profiles. Their antiemetic effects may result from NMDA receptor antagonism and opioid-sparing properties, whereas their emetic potential may be related to the modulation of multiple neurotransmitter systems, including the serotonergic, dopaminergic, and cholinergic pathways, along with catecholamine release [4,70,71]. The previously reported reduction in PONV risk with esketamine compared to ketamine [69] and the negative association observed in our meta-regression analysis may be attributed to esketamine’s enhanced receptor affinity and selectivity compared to ketamine [9]. However, further investigation is needed to determine whether these pharmacological differences translate into clinically meaningful effects on PONV.

Both ketamine and esketamine were associated with a significantly increased risk of hallucinations and drowsiness. Esketamine had a higher RR for drowsiness (RR = 3.43) and dizziness (RR = 1.31) compared with ketamine (RR = 1.50 and 0.90, respectively); the differences were not significant. These findings align with previous meta-analyses that reported higher rates of neuropsychiatric and vestibular symptoms in patients receiving ketamine or esketamine [10,69,71]. Given the variability and limited reporting of safety outcomes across studies, a more systematic and standardized evaluation of adverse effects is warranted. Clinicians should carefully weigh the potential benefits of ketamine and esketamine in opioid-sparing anesthesia against their increased neuropsychiatric and sedative risks.

Our findings suggest that the role of ketamine and esketamine in PONV management requires careful consideration in the broader context of perioperative care strategies. These agents may be most valuable in clinical scenarios where opioid reduction is a primary goal, when traditional antiemetics are contraindicated or unavailable, or when their potential benefits outweigh their known adverse effects. However, our analysis demonstrated that non-opioid alternatives should generally be preferred when available, given their favorable safety profile and superior efficacy in preventing PONV.

This meta-analysis improves upon previous work by focusing on PONV as a primary endpoint rather than a secondary outcome, and by incorporating a larger dataset than earlier studies [69,72]. It employs robust statistical techniques, including random-effects modeling, subgroup analyses, and meta-regression, to address heterogeneity and dose-response relationships. In particular, the opioid subgroup analysis quantified the potential benefits of these agents in reducing PONV risk, whereas a comparison with non-opioid adjuvants highlighted their relative limitations. These methodological strengths contribute to a more refined understanding of the role of ketamine in perioperative PONV prevention.

However, some limitations must be acknowledged. First, moderate heterogeneity across studies may reflect differences in the study design, dosing regimens, comparator types, and timing of outcome assessments, potentially influencing pooled estimates. The non-opioid comparator group included pharmacologically diverse agents that were analyzed as a single category. This heterogeneity may limit comparability and should be considered when interpreting the findings. Second, the inclusion of non-RCTs aimed to provide a comprehensive synthesis of available evidence. Although sensitivity analyses demonstrated that excluding these studies did not significantly impact the results, their inclusion should be recognized as a potential limitation in interpreting the overall findings. Third, the limited number of studies explicitly targeting PONV as a primary endpoint [27,38,59] restricts the certainty of conclusions regarding prevention strategies. Additionally, the analysis of low-incidence events sometimes necessitates the repeated use of certain control or experimental groups, raising concerns about the potential bias or overrepresentation of particular studies [6,11,21–26]. Such methodological constraints may contribute to variability and reduced precision of effect size estimates. Finally, our use of a combined outcome measure for pooled PONV, integrating PONV and PON while excluding POV, was intended to mitigate the risk of overinflation due to overlapping cases of nausea and vomiting. However, this approach may have underestimated the true incidence by not fully accounting for vomiting events.

Further research is needed to address the knowledge gaps identified in this analysis. Priority should be given to conducting direct head-to-head comparisons between ketamine and esketamine for PONV prevention, along with prospective studies evaluating optimal dosing strategies, particularly for esketamine. Furthermore, the investigation of specific patient subgroups that might derive greater benefits from these agents is required. Future studies should implement standardized protocols for adverse effects, with longer follow-up periods to better characterize the safety profiles of these interventions. Such refinement is essential to translate current evidence into more precise, context-specific recommendations, ultimately enhancing postoperative patient care. Taken together, the findings of this analysis highlight the need for caution when incorporating ketamine or esketamine into clinical practice. Neither ketamine nor esketamine consistently reduced the overall PONV relative to all comparators, despite demonstrating a potential benefit over opioid-based regimens. However, their efficacy was inferior to non-opioid agents, emphasizing the importance of careful selection of anesthetic adjuncts in opioid-minimized protocols. The lack of a clear dose-response relationship and increased risk of neuropsychiatric and sedation-related side effects further underscore the complexity of incorporating ketamine or esketamine into routine PONV prevention strategies. Given these findings, further high-quality targeted research is warranted to clarify the optimal dosing, identify high-benefit subpopulations, and refine the risk-benefit profile of these agents.