Continuous ketamine infusion for surgical patients in the intensive care unit: a systematic review and meta-analysis of randomized controlled trials with GRADE assessment

Abstract

Background

Optimal pain and sedation management in intensive care unit (ICU) remains challenging. While opioids and benzodiazepines are widely used, their adverse effects highlight the need for alternative or adjunctive strategies. Ketamine, with its analgesic and opioid-sparing has been proposed as an adjuvant. However, current evidence regarding its efficacy in ICU patients, particularly within the surgical population, remains inconclusive. We evaluated whether continuous low-dose ketamine infusion in postoperative surgical ICU patients reduces opioid consumption and improves pain intensity, and whether it affects opioid-related adverse events and key ICU recovery outcomes.

Methods

We conducted a systematic review and meta-analysis of randomized controlled trials (RCTs) comparing continuous postoperative ketamine infusion versus non-ketamine based infusion for analgesia & sedation in patients admitted to the surgical ICU. Literature searches were performed in PubMed, Cochrane Central, and ClinicalTrials.gov from inception until July 2025. The primary endpoints were cumulative opioid consumption in the first 24–48 h (standardized to IV morphine equivalents where applicable) and rest pain intensity at prespecified early postoperative timepoints. Secondary endpoints included PONV, psychotomimetic events (e.g., hallucinations), and ICU recovery outcomes (e.g., ICU length of stay and duration of mechanical ventilation) when reported. We performed random-effects meta-analyses and assessed certainty using GRADE.

Results

Nine RCTs comprising 666 patients were included. Ketamine significantly reduced opioid consumption at 24 h in adults (mean difference [MD] − 5.77 mg morphine equivalents; 95% CI − 6.32 to − 5.23; p < 0.0001), although the reduction at 48 h and in pediatric subgroups was not statistically significant. Pain scores at 12, 24, and 48 h were comparable between the groups. Ketamine did not significantly shorten the mechanical ventilation time (MD − 0.84 h; 95% CI − 1.93 to 0.24) or ICU length of stay (MD − 0.97 h; 95% CI − 1.95 to 0.02). However, ketamine use was associated with a lower incidence of PONV (RR 0.59; 95% CI 0.36–0.98; p = 0.0399). Psychotomimetic adverse events were infrequent and not significantly increased. GRADE evidence was moderate for opioids in adults at 24 h and low to very low for the remaining outcomes.

Conclusions

In postoperative surgical ICU patients, continuous low-dose ketamine infusion provides modest opioid sparing and is associated with lower PONV, without clear improvements in pain scores or ICU recovery outcomes. Large-scale trials with different dosing protocols are needed to explore the optimal role of ketamine beyond analgesia and PONV reduction.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-026-05858-5.

Introduction

The optimal management of pain and sedation in critically ill patients is a central objective in intensive care medicine, aimed at providing patient comfort while supporting necessary medical interventions. Unaddressed pain can provoke harmful physiological stress responses, which could compromise hemodynamic stability and weaken immune defenses. Meanwhile, oversedation can delay the process of weaning from mechanical ventilation and prolong intensive care unit (ICU) admissions [1]. Over time, growing evidence has highlighted the shortcomings and adverse effects linked to commonly used analgesic and sedative agents such as opioids and benzodiazepines.

In contemporary ICU practice, benzodiazepines are increasingly avoided when possible due to associations with oversedation and delirium, and sedation strategies commonly emphasize non-benzodiazepine agents such as propofol and dexmedetomidine, alongside analgesia-first approaches [2]. Within this landscape, ketamine is sometimes selected because it can provide analgesia with relative hemodynamic stability and may reduce opioid exposure, although its optimal role and outcome impact in postoperative surgical ICU care remain uncertain. The 2018 SCCM PADIS guidelines suggest low-dose ketamine as an adjunct to opioid therapy in postsurgical adults admitted to the ICU when the goal is to reduce opioid consumption (conditional recommendation; very low certainty) [1]. The suggested regimen is 0.5 mg/kg IV bolus followed by 1–2 µg/kg/min infusion, which is consistent with the sub-anesthetic infusion strategies evaluated in this review.

In this context, a growing number of publications focusing on ketamine have drawn attention because of its distinctive pharmacological characteristics, notably its capacity to offer analgesia and sedation while potentially reducing the reliance on conventional opioids and sedatives [2]. Acting chiefly through blocking of the N-methyl-D-aspartate (NMDA) receptors in the central nervous system (CNS), providing analgesia and modulating central sensitization of pain pathways, a feature that can especially help those patients with persistent pain or those exhibiting opioid tolerance [2]. In addition, ketamine has also demonstrated anti-inflammatory properties, which could be beneficial in critically ill patients who frequently suffer systemic inflammation secondary to their underlying conditions [3]. These attributes of ketamine have spiked its growing role as an adjuvant therapy in the intensive care setting, where effectively balancing pain control with minimizing side effects poses a substantial clinical challenge [4].

Despite its potential benefits, the use of ketamine in critically ill patients remains controversial due to inconsistent findings in the literature. For example, an early meta-analysis of both observational and randomized controlled trials reported no reduction in opioid use and a significant increase in ICU length of stay [5]. In contrast, a subsequent meta-analysis of randomized controlled trials revealed a significant reduction in opioid consumption, with no significant differences in mechanical ventilation duration or ICU length of stay [6]. Most recently, another review of randomized controlled trials reported no benefit for opioid consumption but did report fewer cases of delirium [7]. Multimodal analgesia aims to improve pain control while limiting opioid exposure by combining agents with complementary mechanisms (e.g., acetaminophen, NSAIDs when appropriate, gabapentinoids, regional techniques) [5]. Ketamine, at low continuous doses, is proposed as an adjunct within multimodal pathways to reduce opioid requirements while maintaining acceptable analgesia [3].

These conflicting results, together with the heterogeneity of populations across all previous reviews, hinder the ability to draw definitive conclusions for clinical practice. This systematic review primarily evaluates continuous low-dose ketamine infusion as an adjunct for postoperative analgesia in surgical ICU patients, with sedation-related outcomes considered when reported.

Methods

Protocol registration

Our systematic review protocol was registered in PROSPERO (registration ID: CRD420251138578). This systematic review and meta-analysis was conducted following the guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement and the Cochrane Handbook for Systematic Reviews and Meta-Analysis [8, 9].

Data sources & search strategy

A systematic search was conducted in PubMed using bulit-in RCT filter, the Cochrane Central Database of Controlled Trials and Systematic Reviews (CDSR), and ClinicalTrials.gov from inception until July 2025. A combination of Medical Subject Headings terms, free text, and Boolean logical operators was used to construct the search strategy. The detailed approach and results are outlined in Supplementary Table 1.

Eligibility criteria

We included randomized controlled trials (RCTs) evaluating continuous ketamine infusion as an adjunct to standard analgosedative regimens in adult or pediatric patients following major surgery that required ICU admission postoperatively. Ketamine was started either during the operation or postoperatively and must be continued during the entire ICU stay. The pediatric age cut-off was < 18 years (infants through adolescents). Studies were required to report either the primary outcomes (opioid consumption, pain score) or at least one of the secondary outcomes: duration of mechanical ventilation (MV), ICU length of stay, pain scores, postoperative nausea and vomiting (PONV), or psychotomimetic adverse effects. We excluded studies involving surgeries without postoperative ICU admission, ketamine administered as a single bolus, or without continuous postoperative infusion during ICU stay. In addition, we also excluded non-human, non-randomized designs (case reports, series, retrospective studies, and observational cohorts), narrative reviews, systematic reviews, letters, clinical guidelines, and book chapters. Studies that presented overlapping datasets where the surgical population couldn’t be extracted, as well as articles published in languages other than English, were also excluded.

Study selection

Titles and abstracts from the database searches were imported into Rayyan software, and duplicates were removed manually by (R.F.A) [10]. Four authors (R.F.A., R.A., G.A., and K.A.) independently screened the remaining articles, and disagreements were resolved by a fifth reviewer (A.Z.). Disagreements mainly involved whether studies truly included postoperative surgical ICU patients, used continuous ICU ketamine infusion (not bolus/intraoperative-only), and allowed surgical ICU–specific data extraction in mixed-population reports. Potentially eligible articles underwent full-text review to confirm inclusion. Additional conflict were resolved through discussion or consultation with a fifth reviewer (A.Z).

Data extraction

Data from the included studies were extracted independently using a standardized Google sheet by four reviewers (R.F.A., R.A., G.A., and K.A.). The extracted data included: (1) Study characteristics (authors, study setting, study blinding, sample size, measured outcomes and key results); (2) Participant demographics (age, sex, and type of surgery) and; (3) Intervention details (ketamine dose, control drug, and co-analgosedatives). The primary outcome was cumulative opioid consumption at (24–48 h) and postoperative pain scores. Secondary outcomes included the duration of MV, ICU length of stay, and adverse effects such as PONV, and psychotomimetic symptoms. For the data presented graphically, we used WebPlotDigitizer Version 5.2 to extract relevant outcomes. Any discrepancies in data extraction were resolved through discussion or by consulting a fifth reviewer (A.Z.). We contacted the authors for any key missing data that is relevant to our analysis.

PONV was assessed as either the cumulative incidence of PONV or as the separate incidence of nausea or vomiting. Psychotomimetic adverse effects were defined as the occurrence of delirium, hallucination, or confusion. When studies reported data as medians and interquartile ranges (IQRs), validated statistical methods were used to estimate means and standard deviations [11, 12]. Duration variables were converted to hours to maintain consistency across studies. Opioid consumption was standardized to intravenous morphine equivalents using established conversion factors [13]. For opioid consumption, a subgroup analysis was performed between adults and pediatric patients, as this outcome is strongly influenced by weight, hemodynamics, and dose differences, unlike the other outcomes.

Risk of bias and certainty of evidence

Four investigators (R.F.A., R.A., G.A., and K.A.) independently appraised the risk of bias in each eligible trial using the Cochrane Risk-of-Bias 2 (RoB 2) framework [14]. The evaluation addressed the five standard domains: (i) bias arising from the randomization process, (ii) bias due to deviations from the intended intervention, (iii) bias owing to missing outcome data, (iv) bias in the measurement of outcomes, and (v) bias in the selection of the reported result. Judgements were made separately for every prespecified outcome, with every decision explicitly justified and logged in a shared template. Disagreements were discussed until consensus was reached; if consensus could not be achieved, fifth reviewer (A.Z.) adjudicated. We followed the Grading Recommendations Assessment, Development, and Evaluation (GRADE) guidelines to assess the certainty of evidence for each outcome [15, 16]. Any differences in judgment were resolved by group discussion involving fifth reviewer (A.Z.)

Statistical analysis

We pooled continuous outcomes as the mean differences (MDs). Pediatric opioid use was analyzed as the MD in mg/kg. Dichotomous outcomes (PONV, psychotomimetic effects) were summarized as risk ratios (RR). Because rest pain was reported using different instruments (e.g., VAS, NRS, MOPS) and non-uniform time windows, and because initial pooling showed substantial heterogeneity, we did not meta-analyze rest pain and instead synthesized these data narratively. All effects were calculated with 95% CIs using inverse-variance weighting. Heterogeneity was assessed with Cochran’s Q (p < 0.10) and quantified with I² and τ². We prespecified random-effects models for primary analyses; when heterogeneity was negligible (I²≤50%) and fixed- and random-effects estimates were concordant, we reported the common-effect (fixed-effect) estimate. Zero-cell trials in binary outcomes used a continuity correction of 0.50. Sensitivity analyses included leave-one-out analyses and outlier removal when I²≥50% (e.g., testing the impact of Abdelfattah et al., 2024 on pain outcomes). Subgrouping by time point (12, 24, 48 h) and population (adult vs. pediatric) was planned; meta-regression was not performed due to the small number of studies per outcome. Given the limited number of eligible studies and the pragmatic clinical question (ketamine-containing continuous infusions used in postoperative surgical ICU practice), we pooled ketamine-based infusion strategies despite regimen variation; however, we explicitly acknowledge that differences in co-administered sedatives/analgesics and titration targets may contribute to heterogeneity and limit regimen-specific inference. Analyses were performed in R 4.3.3. using the meta package and ggplot2; two-sided α = 0.05 was the statistical significance threshold.

Results

Literature search

A comprehensive search identified 860 papers. After removing 164 duplicates, 696 unique records remained for title and abstract screening. Of these, 658 were excluded as irrelevant, resulting in 38 articles for full-text evaluation. A detailed eligibility assessment excluded 29 papers. In total, nine randomized controlled trials met all inclusion criteria and were incorporated into the meta-analysis. The study selection process is illustrated in the PRISMA flow diagram (Fig. 1).

Fig. 1.

PRISMA 2020 flow diagram of study selection

Characteristics of included studies

Nine randomized controlled trials (RCTs) involving 666 patients were included. These studies were conducted between 2003 and 2025 in seven countries: Egypt (n = 3), the United Kingdom (n = 1), Iran (n = 1), France (n = 1), Japan (n = 1), Pakistan (n = 1), and Thailand (n = 1) [17–25]. All trials were described as double-blinded RCTs, except for Rai et al. 2022, which did not specify the blinding protocol [24]. The study populations consisted of postoperative surgical patients admitted to intensive care units. The type of surgery varied, with cardiac surgery being the most common (n = 5), followed by mixed or non-cardiac surgeries (n = 2), major abdominal surgery (n = 1), and scoliosis surgery (n = 1). Sample sizes ranged from 36 to 150 patients. The gender distribution was skewed toward a slight higher proportion of male participants. For adult patients, the mean age ranged from 34.6 ± 18 years to 64.5 ± 17.5 years. Pediatric studies reported mean ages from 6.42 ± 2.36 years to 15 ± 2 years. Ketamine dosing varied across trials. After standardizing all infusion rate to mg/kg/h, all of the trials utilized the sub-anesthetic low dose ketamine regimen ≤ 0.3 mg/kg/h. The corrected infusion rate ranged from ultr-low 0.012 mg/kg/h in Rai et al., 2022 [24] to a max dose of 0.3 mg/kg/h Rashwan et al., 2018 [19]. Ketamine infusion was given for 24–48 h, initiated either during or immediately after surgery. Three studies employed three-arm groups, in which we selected relevant control arms [18–20]. Included trials used heterogeneous ketamine-containing regimens (analgesic ketamine, ketofol, ketamine combined with dexmedetomidine, and ketamine embedded within multimodal protocols), which likely differ in clinical intent (analgesia vs. analgosedation) and co-interventions. Additional details and summaries of the included studies and control arms are shown in Tables 1 and 2.

Table 1.

Study characteristics

Study	Country	Type of Surgery	Timing of infusion	Main Findings	Intervention	Control
Anwar et al. (2019)	UK	Elective cardiac surgery	Postoperative infusion for 48 h	Ketamine with pregabalin reduced morphine use but similar  icu length of stay	N = 50 Age = 64 [54–72] years Ketamine + Pregabalin + Morphine	N = 50 Age = 68 [59–73] years Pregabalin + Morphine
Minoshima et al. (2015)	Japan	Scoliosis Surgery	Intra-operative bolus + continuous infusion for 48 h	Ketamine reduced morphine use and antiemetic requirement.	N = 17 Age = 15 (+/-2) years Ketamine + Morphine	N = 19 Age = 14 (+/- 2) years Placebo (Saline) + Morphine
Hamed et al. (2024)	Egypt	Open Heart Surgery	Postoperative	Ketamine and dexmedetomidine had similar sedation & pain control effect, with superior opioid reduction with ketamine	N = 30 Age = 49 +/-10 years Ketamine + Paracetamol + Intermittent Fentanyl	N = 30 Age = 50.7+/-9.1, 52.1+/-11.5 years Saline + Fentanyl
Moshtaghion et al. (2018)	Iran	CABG Surgery	Postoperative	Ketofol led to reduced morphine use and mv duration. HIgher MAP and delirium	N = 45 Age = 40–60 years Ketamine + Propofol (Ketofol)	N = 45 Age = 40–60 years Propofol Alone
Guillou et al. (2003)	France	Major Abdominal Surgery	Postoperative bolus + infusion for 48 h	Ketamine reduced morphine use with similar sedation, pain and adverse effect	N = 41 Age = 60 +/-16 years Ketamine + Morphine	N = 52 Age = 60+/-15 years Morphine Alone
Abdelfattah et al. (2024)	Egypt	Pediatric Open Heart Surgery	Intra-operative bolus + continuous infusion for 24 h	Ketamine reduced total opioid consumption, similar icu length of stay	N = 35 Age = 6.42 ± 2.36 years Ketamine + Propofol + Fentanyl	N = 34 Age = 5.92 ± 2.13 years Placebo (Saline) + Fentanyl + Propofol
Rai et al. (2022)	Pakistan	CABG Surgery	Postoperative	Ketamine + dexmedetomidine reduced fentanyl use and extubation time	N = 20 Age = 53.65 ± 6.97 years Ketamine + Dexmedetomidine + Fentanyl	N = 20 Age = 55.75 ± 5.65 years Dexmedetomidine + Fentanyl
Kitisin et al. (2025)	Thailand	Non-cardiac/non-neuro surgery	Postoperative for 48 h	Small opioid-sparing effect; increased long-term traumatic memories	N = 60 Age = 64.5 ± 17.5 years Low-dose ketamine infusion + Fentanyl	N = 58 Age = 60.6 ± 17.1 years Placebo (normal saline) + Fentanyl
Rashwan et al. (2018)	Egypt	Mixed (oby&gyn, orthopedics, vascular and abdomen)	Postoperative bolus + infusion for 24 h	Ketofol reduced opioid consumption with higher vitals (MAP, SBP, DBP)	N = 30 Age = 34.6 ± 18 years Ketamine/propofol 1:1 at a rate	N = 30 Age = 40.2 ± 13.4 years Propofol 1:1

N = number of patient, RCT = Randomized controlled trial, CABG = Coronary Artery Bypass Grafting; MAP = Mean Arterial Pressure, MV = Mechanical ventilation, ICU = Intensive care unit, UK: United kingdom

Table 2.

Measured outcomes

Study	Outcomes	Reported ketamine dose	Standerized ketamine dose to MG/KG/H
Anwar et al. (2019)	1- Total postoperative morphine consumption at 24 h, mg 2- Extubation time, min 3- Length of stay in cardiac intensive care, hours	0.1 mg/kg/h	0.1
Minoshima et al. (2015)	1- Total postoperative morphine consumption at 24 h, mg/kg 2- NRS pain score at rest 12 h, 24 h and 48 h measured by a nurse blinded to the study groups. 3- PONV score ≥ 1, defined as the incidince of PONV	2 µg/kg/min	0.12
Hamed et al. (2024)	1- Total postoperative fentanyl consumption at 48 h, µg 2- Postoperative ventilation time, hours 3- Length of stay in ICU, days 4- Incidence of PONV	1–2 µg/kg/min	0.06–0.12
Moshtaghion et al. (2018)	1- incideince of delirium using CAM-ICU criteria	3 µg/kg/min	0.18
Guillou et al. (2003)	1- Total postoperative morphine consumption at 24 h and 48 h, mg 2- VAS pain score at rest 12 h, 24 h and 48 h measured by measured by a blinded observer every 4 h 3- Incidence of nausea 4- Incidence of confusion & hallucinations	1–2 µg/kg/min	0.06–0.12
Abdelfattah et al. (2024)	1- Total postoperative fentanyl consumption at 24 h, µg/kg 2- MOPS pain score at rest 12 h and 24 h measured by a blinded resident anesthetist 3- Length of stay in ICU, hours 4- Incidence of vomiting	0.1 mg/kg/h	0.1
Rai et al. (2022)	1- Extubation time, min 2- Length of stay in ICU, hours	0.2–0.7 µg/kg/min	0.012–0.042
Kitisin et al. (2025)	1- Total postoperative fentanyl consumption at 24 h, µg 2- Duration of mechanical ventilation, hours 3- Length of stay in ICU, hours 4- Incidence of delirium using CAM-ICU criteria	1.5 µg/kg/min	0.09
Rashwan et al. (2018)	1- Total postoperative fentanyl consumption at 24 h, µg	10 µg/kg/min (Ketofol) 1:1	0.3

h = hours, NRS = Numerical pain scale, CAM-ICU = Confusion Assessment Method for the ICU, PONV = Postoperative nausea and vomiting, VAS = Visual Analogue Scale, MOPS = Modified Objective Pain Scale

Risk of bias assessment

Of the nine randomized trials, four were judged to have a low overall risk of bias (Kitisin 2025, Abdelfattah 2024, Minoshima 2015, and Hamed 2024) [17, 18, 23, 25]. Three studies raised some concerns (Kitisin 2025, Abdelfattah 2024, Minoshima 2015, and Hamed 2024) [19–21], mainly due to incomplete reporting of sequence generation/allocation concealment, unclear blinding, and/or analysis/attrition considerations, which are particularly relevant for subjective outcomes such as pain or sedation scores. Two studies were rated high overall risk: Rai 2022 [24] because allocation/blinding procedures were insufficiently reported for subjective outcomes, and Guillou 2003 [22] due to post-randomization exclusions/attrition that could bias outcome estimates. Overall, many key endpoints (e.g., opioid consumption and ventilation-related timings) were relatively objective, and several trials used blinded drug preparation and/or blinded staff for outcome assessment, limiting detection bias. The risk-of-bias assessment is illustrated in (Supplementary Fig. 1), and the certainty of evidence for each outcome is summarized in Supplementary Table 2.

Primary outcomes

Opioid consumption

Four RCTs (n = 371) reported 24-h IV morphine use in adults. Notably, the 24 h infustion duration for Rashwan et al., 2018 was obtained via contacting the author. The fixed-effect meta-analysis showed a significant reduction with ketamine (MD − 5.77 mg; 95% CI − 6.32 to − 5.23; I² = 0%; p < 0.0001; high certainty), driven largely by Rashwan 2018, which carried most of the weight using 0.3 mg/kg/h corrected dose (Fig. 2a). Two RCTs (n = 153) reported 48-h IV morphine use in adults. Although both trials favoured ketamine, the random-effects model was not statistically significant overall (MD − 12.52 mg; 95% CI − 27.45 to 2.41; I² = 75.3%; p = 0.1002; low certainty) (Fig. 2b). Two paediatric RCTs (n = 105) reported 24-h IV morphine use normalized by weight. The pooled random-effects estimate was not significant (MD − 0.52 mg/kg; 95% CI − 1.25 to 0.20; I² = 98.1%; p = 0.1571; very low certainty) (Fig. 2c).

Fig. 2.

Forest plots of opioid consumption (IV morphine equivalents). (a) Adults, 24 h (4 RCTs; N = 371), mean difference (MD) in mg; common-effect (fixed-effect) model. (b) Adults, 48 h (2 RCTs; N = 153), MD in mg; random-effects model. (c) Pediatric patients, 24 h (2 RCTs; N = 105), MD in mg/kg; random-effects model. Negative MD favors ketamine (lower opioid consumption)

Postoperative pain

In the adult surgical ICU evidence, only one trial (Guillou et al., 2003) reported rest pain, and it was assessed with a 100-mm VAS at rest and during mobilization every 4 h for 48 h under a morphine PCA protocol titrated to VAS < 30. Ketamine was given as a 0.5 mg/kg bolus followed by infusion (2 µg/kg/min for 24 h, then 1 µg/kg/min for 24 h). Despite opioid sparing, VAS pain scores at rest and mobilization were similar between groups throughout follow-up.

The pediatric evidence included two different settings and pain measures. In adolescents after scoliosis surgery (Minoshima 2015), NRS pain at rest and on motion was measured up to 48 h; all patients received morphine PCA with background infusion plus diclofenac. Ketamine (2 µg/kg/min to 48 h) reduced morphine use, but NRS scores were not different. In younger children after open cardiac surgery (Abdelfattah 2024), pain was assessed with MOPS after extubation with fentanyl rescue when MOPS > 3, on a protocol of paracetamol and intermittent IV morphine. Low-dose ketamine (0.3 mg/kg bolus plus perioperative infusion) was linked to lower pain scores and reduced opioid exposure in the first 24 h.

ICU and mechanical ventilation duration

Five RCTs (n = 387) reported ICU length of stay. The common-effect model suggested a borderline, non-significant reduction with ketamine use (MD − 0.97 h; 95% CI − 1.95 to 0.02; I² = 0%; p = 0.0543; moderate certainty) (Fig. 3a). Four RCTs (n = 318) reported mechanical ventilation duration. The random-effects meta-analysis was not significant overall (MD − 0.84 h; 95% CI − 1.93 to 0.24; p = 0.1280; low certainty) with substantial heterogeneity (I² = 74.3%) (Fig. 3b). Leave-one-out sensitivity analyses could not resolve this heterogeneity (Supplementary Fig. 2).

Fig. 3.

Forest plots of ICU recovery outcomes (hours). (a) ICU length of stay (5 RCTs; N = 387), MD in hours; common-effect (fixed-effect) model. (b) Duration of mechanical ventilation (4 RCTs; N = 318), MD in hours; random-effects model. Negative MD favors ketamine (shorter duration)

Adverse events

Four RCTs (n = 258) reported PONV. The pooled common-effect estimate showed a significant reduction with ketamine use (RR 0.59; 95% CI 0.36–0.98; I² = 0%; p = 0.0399; moderate certainty) (Fig. 4a). Three RCTs (n = 301) reported psychotomimetic adverse effects. Events were infrequent and the overall effect was not significant between ketamine and control groups (RR 1.40; 95% CI 0.40–4.92; I² = 0%; p = 0.5992; low certainty) (Fig. 4b).

Fig. 4.

Forest plots of adverse events. (a) Postoperative nausea and vomiting (PONV) (4 RCTs; N = 258), risk ratio (RR); common-effect (fixed-effect) model. (b) Psychotomimetic adverse events (3 RCTs; N = 301), RR; common-effect (fixed-effect) model. RR < 1 favors ketamine

Discussion

Our meta-analysis demonstrates that adjunctive low-dose ketamine infusions can significantly reduce postoperative opioid requirements in surgical ICU patients. At 24 h post-surgery, ketamine lowered IV morphine consumption by roughly 5.77 mg on average compared to controls in adult patients. Regarding clinical importance, a minimally important difference of approximately 5 mg IV morphine equivalents over 0–24 h has been previously reported [26]. In this context, the observed effect just exceeds this threshold, supporting a clinically small opioid-sparing benefit. However, this magnitude should not be interpreted as trajectory-changing, particularly because we did not observe corresponding improvements in ICU recovery outcomes (mechanical ventilation duration, ICU length of stay). The most consistent downstream signal was reduced PONV, plausibly mediated by lower opioid exposure. Collectively, these updated data support and strengthen the 2018 PADIS conditional recommendation for low-dose ketamine as an opioid-sparing adjunct in postsurgical ICU adults, while also underscoring that clinically relevant benefits may manifest as fewer opioid-related symptoms (e.g., PONV) rather than large reductions in recorded pain intensity [1].

This finding aligns with prior evidence in overall critically ill populations, as a recent meta-analysis of six RCTs (494 patients) likewise showed ketamine regimens reduced opioid use (approximately 13.2 µg/kg/h less morphine equivalent, p = 0.004) [7]. Although the opioid-sparing in our analysis is modest, it is broadly consistent with earlier trials reporting reduced opioid requirements with ketamine [21, 23, 24]. In contrast, the largest ICU ketamine trial (Perbet et al. 2018) found no opioid-sparing benefit, potentially because aggressive nurse-driven sedation with low-dose remifentanil reduced the opportunity to detect an incremental effect [27]. These discrepancies underscore that population differences and co-analgesic protocols can materially influence observed benefit. Mechanistically, the overall opioid-sparing signal is consistent with NMDA-receptor antagonism, which may attenuate central sensitization and opioid tolerance [28], and even modest reductions in opioid exposure may be clinically desirable given associations between higher opioid dosing and complications such as respiratory depression, ileus, and hyperalgesia [29].

Interpretation of adjunct analgesics depends on the analgesic protocol. If opioid dosing is held constant, an effective adjunct typically appears as lower pain intensity. In contrast, if opioids are actively titrated downward (e.g., via rescue-analgesia algorithms or clinician titration) to maintain similar pain targets, the same analgesic effect will appear as similar pain scores with reduced opioid exposure and potentially fewer opioid-related adverse effects. This may explain why pooled rest pain scores did not consistently improve despite reduced opioid consumption.

Despite reduced opioid consumption, rest pain intensity in our analysis remained similar between groups across early timepoints (12–48 h). This pattern is consistent with the RCT by Martinez et al., where ketamine reduced 48-h morphine use without improving pain ratings [30], and with perioperative ketamine literature showing opioid sparing without large changes in pain scores when background analgesia is actively titrated [31]. A possible explanation is protocolized titration: clinicians adjust opioids and rescue analgesia to achieve acceptable pain targets, so an effective adjunct may manifest primarily as lower opioid requirements rather than lower recorded pain intensity. In addition, heterogeneity in pain reporting and settings likely contributes to variable effects, and ketamine’s benefit may be more apparent for dynamic or refractory pain (e.g., during physiotherapy or in opioid-tolerant patients), which resting pain measures may not capture [32]. Overall, our findings support ketamine’s role as an opioid-sparing adjunct within multimodal regimens rather than a strategy that reliably reduces rest pain beyond conventional protocols [33].

We found no compelling evidence that ketamine shortened ICU recovery time in this surgical population. ICU length of stay was only about 1 h shorter on average with ketamine (mean difference ≈ − 0.97 h) and this trend narrowly missed statistical significance (p ≈ 0.05). A reduction of under one hour is trivial in clinical terms, and overall our analysis agrees with prior studies indicating ketamine does not significantly affect ICU or hospital LOS [6]. Similarly, duration of mechanical ventilation was not significantly reduced in patients treated with ketamine. A pooled analysis of four RCTs in our review showed a non-significant ~ 0.8 h decrease in ventilator time (with substantial between-study variability). These null findings also mirror those of earlier meta-analyses: Chan et al. (2022) reported no difference in mechanical ventilation days or ICU stay between ketamine and control groups [6]. It appears that adjunctive ketamine, at least at the doses studied, is not a major determinant of ICU course length. This is not entirely surprising in our study context, given that factors such as severity of illness, surgical complexity, and sedation weaning protocols dominate ICU outcomes [27]. Even if ketamine provides some opioid-sparing and hemodynamic stability, these benefits did not translate into measurably faster extubation or ICU discharge in the aggregated data. Our findings reinforce that routine ketamine use in postoperative ICU care should not be expected to dramatically alter ICU length of stay or ventilator time, outcomes that likely require more substantial systematic changes to improve. Rather, ketamine’s benefits may manifest in other areas, like opioid sparing and side-effect profile. Future studies focusing on specific subsets (e.g., patients with difficult-to-control pain or opioid tolerance) might clarify if any niche exists where ketamine shortens ICU recovery.

Encouragingly, the addition of low-dose ketamine did not introduce major safety issues in our analysis. In fact, the incidence of postoperative nausea and vomiting (PONV) was significantly lower in the ketamine group (pooled RR ~ 0.59), indicating roughly a halving of PONV risk compared to control. This finding is clinically meaningful and aligns with the concept that reducing opioid consumption lowers PONV incidence. Opioid dose is one of the strongest independent risk factors for PONV in surgical patients [28], and by sparing opioids, ketamine likely mitigates opioid-related nausea and vomiting. Indeed, in a recent RCT of thoracic surgery patients, perioperative S-ketamine infusion reduced opioid use by 40–75% and correspondingly reduced PONV from 30% down to 9.7% [29]. Our results echo this benefit across multiple studies, and a broad umbrella review confirms that IV ketamine adjuncts are associated with a lower risk of nausea/vomiting while improving analgesia [30]. PONV is not only an adverse event but also a recovery-relevant outcome, as it can limit oral intake, delay mobilization, and affect discharge readiness. Therefore, a reduction in PONV may represent a clinically meaningful functional benefit even when ICU length of stay is unchanged. Several included trials administered routine antiemetic prophylaxis to all participants; despite this, the ketamine group still showed a lower incidence of PONV, consistent with an opioid-sparing mechanism. For ICU patients who often receive high-dose opioids, ketamine’s ability to curb PONV is an attractive ancillary benefit.

The most well-known concern with ketamine is its potential for psychotomimetic adverse effects, such as hallucinations, vivid dreams, or delirium [31]. In our meta-analysis, however, the incidence of these psychotropic side effects was low and did not differ significantly between ketamine and control groups (RR ~ 1.4, 95% CI spanning 0.4–4.9). In absolute terms, few patients in either group experienced emergent reactions or disturbing psychological symptoms, likely reflecting the low, sub-anesthetic dosing used and the ICU setting where patients often received concomitant sedatives. These findings are consistent with other reports that low-dose, continuous-infusion ketamine does not appreciably increase adverse neuropsychiatric events [30, 32]. An analysis of multiple acute pain studies found no significant difference in hallucinations, sedation, or nightmares with sub-anesthetic ketamine compared to placebo [30]. Importantly, co-administration of benzodiazepines or propofol, which is common in ICU sedation protocols, can attenuate ketamine’s psychoactive effects; subgroup data have shown that when midazolam is given or ketamine is infused slowly, the risk of hallucinations is not elevated [33]. In our included trials, patients were generally deeply sedated postoperatively or had concurrent analgesics, which may have blunted any perceptual symptoms. Additionally, ketamine’s pharmacologic profile at low doses (e.g. 0.1–0.3 mg/kg/h) produces more analgesia than dissociation [34]. Another safety consideration is whether ketamine might increase delirium in ICU patients, a pertinent issue since delirium is multifactorial. Interestingly, emerging evidence suggests the opposite; a meta-analysis of 4 ICU trials found that ketamine sedation was associated with a lower incidence of delirium (21% vs. 37% with standard sedation), possibly by reducing the doses of other sedatives and opioids [7]. While this finding must be interpreted cautiously due to the limited sample and heterogeneity, it aligns with ketamine’s lack of depressant effect on respiration and its anti-inflammatory properties, which may protect against delirium [35]. In our review, we did not formally assess ICU delirium, but the lack of excess psychotomimetic events is reassuring. Likewise, ketamine had no significant effect on other adverse events like respiratory depression in the studies, consistent with its known safety at low doses. Overall, our results support that continuous low-dose ketamine, when used appropriately, has a tolerable safety profile in postoperative surgical ICU patients. The evidence suggests a reduction in PONV across the studies that reported it. Other opioid-related adverse events (e.g., ileus and pruritus) were reported inconsistently and could not be pooled, so any suggestion of benefit for these outcomes remains uncertain.

Pediatric surgical ICU evidence

In pediatric postoperative surgical ICU patients, evidence remains limited and heterogeneous, and findings should not be extrapolated from the adult data. In our meta-analysis, only two pediatric RCTs (n = 105 total) were included, and the pooled result did not show a significant opioid-sparing effect at 24 h in children (mean difference − 0.52 mg/kg morphine, p = 0.16). However, heterogeneity was very high (I² ≈ 98%), implying the two trials had markedly differing results. Indeed, one recent pediatric study by Abdelfattah et al. (2024) found clear benefits of ketamine; in children undergoing open-heart surgery, a low-dose ketamine infusion (0.1 mg/kg/h postoperatively) reduced intra- and postoperative fentanyl requirements and led to lower pain scores, without any hallucinations noted [17]. In that high-intensity pain setting (sternotomy for cardiac surgery), ketamine proved to be an effective adjunct for analgesia. Another large 2011 meta-analysis on perioperative ketamine in children found that while ketamine decreased immediate postoperative pain in the PACU, whether surgical patients or not, it “failed to exhibit a postoperative opioid-sparing effect” beyond the first few hours [36]. Our results at 24 h concur with that conclusion, ketamine’s impact on opioid consumption in the pediatric surgical population is not clearly proven, except perhaps in certain contexts. Children undergoing surgery often receive regional anesthesia (e.g., caudal blocks or epidurals) and different opioid-sparing adjuncts, which could dilute the observable effect of ketamine [37]. Moreover, pediatric pain assessment and opioid dosing differ from those of adults, making standardization challenging. Safety in children is also paramount: encouragingly, pediatric studies reported no ketamine-related psychotomimetic events or other major side effects [17, 23], aligning with the notion that low-dose ketamine is generally well tolerated in infants and children [38]. Nonetheless, due to the paucity of trials, the role of ketamine in pediatric surgical ICU analgesia remains an open question. It may be most beneficial in cases of very painful procedures, such as cardiac or major orthopedic surgeries, or opioid-tolerant adolescents, but further pediatric-specific RCTs are needed. In practice, clinicians should individualize decisions, balancing potential opioid-sparing benefits with the limited high-quality data in children.

Notably, the 2025 SCCM focused update (PAADIS) added Anxiety as a new guideline domain, reflecting an expanded emphasis on comprehensive symptom management, consistent with our framing of outcomes beyond pain alone (e.g., PONV and neuropsychiatric effects) [39].

Limitations

Several limitations of our review warrant consideration. First, the number of trials included and patients for some outcomes was modest, limiting statistical power. Many RCTs were single-center with relatively small sample sizes, raising concerns about imprecision and publication bias. The significant heterogeneity observed in key outcomes (e.g., pain scores at 12–24 h, 48-h opioid consumption) suggests substantial variability in patient populations, ketamine dosing strategies, and co-interventions across studies. A key limitation is that the pooled estimates combine ketamine regimens with different clinical goals and co-interventions (e.g., ketofol, ketamine–dexmedetomidine, pregabalin-based multimodal pathways). Therefore, the summary effects should be interpreted as reflecting a broad ketamine-based infusion strategy rather than the isolated effect of any single regimen, and confounding by concurrent agents and titration targets cannot be excluded. This heterogeneity led us to downgrade the certainty of evidence for several outcomes. Secondly, our analysis focused on short-term outcomes in the ICU and did not capture long-term endpoints such as chronic pain, opioid dependence after discharge, or 28-day mortality. It remains unknown if ketamine’s peri-ICU use confers any longer-term benefits or harms. Thirdly, adverse event reporting was suboptimal in several studies, the definition and assessment of psychotomimetic effects, for example, were not uniform, and rare events may have been missed due to the sample size. We primarily assessed nausea/vomiting and acute hallucinations; other potential issues like neurocognitive outcomes, liver toxicity, or urinary symptoms were infrequently reported. Several opioid-sparing–relevant outcomes, particularly ileus, pruritus, sleep quality, and delirium, were infrequently and inconsistently reported, often without standardized definitions or time windows, precluding meta-analysis. Future postoperative surgical ICU trials of ketamine should predefine and systematically capture these patient-centered outcomes alongside standardized co-analgesic pathways and sedation targets. Because pediatric studies were few and clinically distinct, pediatric findings were analyzed separately and should be interpreted cautiously; they are not directly generalizable to adult postoperative surgical ICU practice. Important co-interventions, particularly opioid selection (e.g., fentanyl vs. morphine vs. remifentanil/sufentanil) and the use of regional/local anesthesia, were incompletely reported and could not be accounted for analytically, which may confound analgesic outcomes and contribute to heterogeneity. In addition, ketamine formulation (racemic vs. S-ketamine) was seldom stated; thus, applicability to settings using S-ketamine may be limited. Finally, the optimal dosing regimen of ketamine remains unclear: the included studies used varying bolus doses, or none, and a variable range of infusion rates. We were unable to determine a dose-response relationship or the ideal duration of infusion. This variability, along with the lack of standardized concurrent analgesia protocols, highlights the need for further research before strong clinical recommendations can be made.

Future implications

Our findings have several implications for practice and research. For clinicians, they underscore that ketamine can be a valuable adjunct in the ICU for managing postoperative pain, primarily by reducing opioid requirements and opioid-related side effects, especially PONV. This opioid-sparing quality is especially pertinent amid efforts to minimize opioid exposure in the hospital. ICU protocols might consider incorporating low-dose ketamine as part of multimodal analgesia for suitable surgical patients, for example, those with high opioid needs or contraindications to excessive opioids, such as risk of respiratory depression [40]. However, clinicians should implement ketamine cautiously, monitoring any psychiatric side effects and being mindful that not all patients will experience dramatic pain relief from it. In patients with hemodynamic instability, ketamine’s sympathomimetic properties can be advantageous, suggesting a role in analgesia for patients in shock or after cardiac surgery [41].

From a research perspective, our review highlights the need for larger, high-quality RCTs to resolve the current uncertainties. Future trials should aim to enroll broader multicenter patient samples to improve generalizability and power. Key questions include: Does ketamine improve meaningful patient-centered outcomes? While surrogate outcomes like opioid consumption are consistently reduced, future studies should assess whether ketamine translates into better pain relief during movement, faster functional recovery, reduced incidence of chronic pain, or shorter rehabilitation time. Trials with long-term follow-up (beyond ICU discharge) would be valuable to see if ketamine’s NMDA blockade confers lasting benefits in reducing central sensitization or opioid dependence. Moreover, research should focus on identifying sub-populations that benefit most. It is plausible that patients with severe postoperative pain, such as open abdominal or cardiac surgery, or those tolerant to opioids, might derive greater analgesic and anti-hyperalgesic effects from ketamine. This hypothesis should be tested in subgroup analyses or targeted trials to confirm its validity. Conversely, patients on mechanical ventilation with heavy sedation might not benefit as much, since pain is less of an issue under deep sedation and ketamine’s effect could be masked [42]. Optimal dosing and combination regimens also warrant investigation. Future studies could compare different infusion rates or the addition of a bolus vs. infusion alone to determine the regimen that maximizes analgesia while minimizing side effects. Given suggestions that co-medication with benzodiazepines reduces hallucinations [33], protocols including a low-dose midazolam or alpha-2 agonist with ketamine might be explored for synergy in ICU sedation. In pediatric patients, dedicated trials are needed to confirm whether ketamine can consistently reduce opioid use and improve pain outcomes, as our data there were inconclusive. Finally, ongoing and future research should continue to evaluate ketamine’s impact on ICU delirium and cognitive outcomes, since preliminary data suggest a possible reduction in delirium, a finding that, if confirmed, could position ketamine as a unique sedative-analgesic with cognitive benefits in critical care [7].

Conclusion

Continuous low-dose ketamine may be considered as an adjunct for postoperative analgesia in adult surgical ICU patients, but its benefit appears small in absolute terms. Across studies, it was associated with modest opioid sparing (adult 24-h MD − 5.77 mg IV morphine equivalents, just above the 5-mg MID) and lower PONV, without a clear increase in hallucinations or other psychotomimetic effects. Effects on rest pain and ICU recovery endpoints (ICU length of stay, duration of mechanical ventilation) were inconsistent or not evident. This pattern likely reflects analgesic protocols that titrate opioids to achieve similar pain targets, whereby the primary measurable benefit of an effective adjunct is opioid-sparing and fewer opioid-related adverse effects rather than lower recorded pain intensity. Other opioid-sparing–related outcomes (e.g., ileus, pruritus, sleep, delirium) were insufficiently reported for synthesis, highlighting priorities for future trials.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

A.Z. led the study design, literature review and drafting of the manuscript. A.S performed the statistical analysis, results writing and prepared the figures. R.F.A., R.A., G.A. and K.A. supported data collection and  F.A., N.Z. revised and validated the data. L.A. and S.A. wrote the discussion part. Z.A. did the characteristics part of the results. A.S.A. provided overall supervision and approved the final version. All authors read and approved the final manuscript. All authors have read and agreed to the final version of the manuscript.

Funding

None.

Data availability

Data is available on reasonable request.

Declarations

Ethics approval

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.