Is volatile sedation truly associated with increased mortality in mechanically ventilated critically ill adults compared to intravenous sedation? Moving beyond pairwise meta-analysis to individual agent assessment via bayesian network meta-analysis

Abstract

Dear editor,

ICU sedation represents a cornerstone of critical care management, balancing patient comfort and safety against the risks of oversedation, prolonged mechanical ventilation, and drug-related adverse effects. We read with great interest the recent meta-analysis by Yamamoto et al. examining volatile sedation in critically ill adults receiving mechanical ventilation [1]. Surprisingly, their findings demonstrated that volatile sedation was associated with increased mortality compared with intravenous sedation (relative risk [RR]: 1.17; 95% confidence interval: 1.02–1.35).

The authors conducted rigorous analyses to explore whether this result, which contradicted their original hypothesis, might reflect methodological limitations. Trial sequential analysis suggested possible false-positive findings, as the Z-curve crossed conventional but not α-spending boundaries, with the sample size well below the required information size. Bayesian analysis yielded more conservative estimates (RR: 1.16; 95% credible interval [CrI]: 0.94–1.42) but indicated a 92.8% probability of increased mortality. They also performed numerous sensitivity analyses to explore potential explanations for their findings. GRADE assessment rated evidence certainty as low, precluding definitive conclusions. This systematic review exemplifies methodological rigor, achieving high quality on AMSTAR2 (A MeaSurement Tool to Assess systematic Reviews, version 2) [2]. We particularly commend their comprehensive statistical approach, which prompts deeper consideration of whether volatile sedation truly increases mortality in this population.

Here, we present an alternative analytical approach—Bayesian network meta-analysis (NMA)—to examine the four sedatives individually: two volatile agents (isoflurane, sevoflurane) and two intravenous agents (propofol, midazolam). Our rationale stems from the distinct pharmacological profiles of these agents. Notable differences exist between volatile agents: isoflurane primarily reduces systemic vascular resistance while maintaining cardiac output, whereas sevoflurane causes more pronounced myocardial depression with less vasodilation [3]. Such differences could impact hemodynamics and potentially influence mortality. Similarly, propofol’s hemodynamic effects [4] and midazolam’s association with increased delirium risk [5] may differentially affect outcomes. Therefore, pooling agents by class in pairwise meta-analysis may introduce substantial heterogeneity. While Yamamoto et al. addressed this through sensitivity analyses, we propose that NMA comparing all four agents individually provides a more granular assessment.

Mesnil et al. (2011) was the sole three-arm study in their review, comparing sevoflurane, propofol, and midazolam [6]. Conducting NMA using only the originally included studies might produce biased results because the network connection between propofol and midazolam relies on a single direct comparison. To strengthen our analysis, we incorporated additional studies from a recent systematic review comparing propofol and midazolam, extracting mortality data to create a more robust network structure [7]. We included only randomized controlled trials (RCTs) with direct comparisons between sedative pairs. Our final analysis comprised 27 RCTs, with direct propofol-midazolam comparisons increasing from one to twelve. We conducted Bayesian NMA using the ‘multinma’ package in R, implementing random-effects models with non-informative priors (warmup: 4000 iterations, sampling: 6000 iterations). Results are presented as odds ratios (ORs) with 95% credible intervals (CrIs) [8]. Markov Chain Monte Carlo trace plots showed good mixing across all chains, confirming model convergence. Network consistency was assessed through deviance information criterion (DIC) comparison of consistency versus unrelated mean effects (UME) models and node-splitting analysis [9]. The consistency model (DIC = 68.2) performed similarly to the UME model (DIC = 69.6), with ΔDIC = 1.4. The difference below 3 indicates no network inconsistency. All omega parameters’ 95% CrIs included zero, and Bayesian p-values ranged from 0.31–0.65, also confirming network consistency [10]. An I² of 22.2% indicated low to moderate and acceptable between-study heterogeneity.

Using midazolam as the reference, the median ORs and 95% CrIs were: propofol 1.14 (0.70–2.00), isoflurane 1.20 (0.53–2.58), and sevoflurane 1.46 (0.72–2.92) (Fig. 1 A). The Surface Under the Cumulative Ranking Curve (SUCRA) values were as follows: midazolam 0.75, propofol 0.56, isoflurane 0.48, and sevoflurane 0.21. Higher SUCRA values indicate better performance in terms of mortality avoidance. These results suggest that midazolam ranked best for mortality outcomes, while sevoflurane ranked worst among the four sedatives. Furthermore, using pooled baseline mortality from midazolam trials, we predicted absolute mortality probabilities for each sedative (Fig. 1B). We stratified risk into three clinically meaningful categories: low (< 5%), moderate (5–15%), and high (>15%) mortality risk. The predicted absolute mortality probabilities showed patterns consistent with our relative effect estimates. Notably, when stratified by mortality risk categories, interesting patterns emerged. Propofol demonstrated the most favorable safety profile, with 95.3% of posterior samples falling within the moderate risk category and only 4.7% exceeding 15% mortality. While midazolam showed a slightly higher probability of high-risk mortality (10.7%) compared to propofol, the difference was modest. In contrast, sevoflurane showed concerning results, with 54.4% of samples indicating mortality risk >15%. Isoflurane demonstrated intermediate risk, with 33.3% probability of high mortality risk. Among the included studies, Jabaudon 2025 enrolled 687 patients with moderate to severe acute respiratory distress syndrome [11]. It is reasonable to suspect that the higher mortality probability in the sevoflurane group might have resulted from the greater weight of this study. We performed a leave-one-out sensitivity analysis, and the OR of sevoflurane decreased when compared to midazolam (OR: 1.26, 95% CrI: 0.53–3.42). As shown in Fig. 1B, the half-eye plots showed minimal changes when excluding Jabaudon 2025. The probability of high mortality risk (>15%) changed to 37.0%, 4.9%, 10.0%, and 31.7% for sevoflurane, propofol, midazolam, and isoflurane groups, respectively. In terms of high-risk mortality probability, sevoflurane clearly remains the most dangerous choice, followed by isoflurane. These findings suggest that while statistical uncertainty persists regarding relative effects, the absolute risk predictions provide clinically actionable insights.

Recently, (network) meta-analyses employing Bayesian methods have gained increasing recognition in critical care medicine [12]. We advocate that even frequentist (network) meta-analyses should incorporate Bayesian approaches as sensitivity analyses. Our Bayesian approach provides probabilistic estimates that facilitate risk-stratified decision-making, moving beyond simple point estimates to quantify uncertainty in treatment effects. This methodology offers clinicians a more nuanced understanding of sedative-associated mortality risks in critically ill patients. Based on our Bayesian NMA, all four sedative agents appear acceptable for mechanically ventilated critically ill adults with estimated mortality risk < 15%. For patients with higher baseline mortality risk, our analysis suggests propofol may offer the most favorable safety profile for hemodynamically stable patients, with midazolam as a reasonable alternative. Among volatile agents, while our data show some uncertainty in mortality outcomes, they remain viable sedative options, particularly in settings with appropriate delivery systems and monitoring capabilities. Our analysis suggests isoflurane may be preferable to sevoflurane based on the mortality probability distributions, though sevoflurane remains justified in specific clinical scenarios such as severe bronchospasm requiring bronchodilation [13].

Ultimately, sedative selection should be individualized based on patient-specific factors, drug characteristics, contraindications, intended sedation duration, and institutional expertise, rather than following a rigid hierarchical protocol. More importantly, future investigations should include guideline-recommended agents such as dexmedetomidine [14] and promising newer agents like remimazolam [15] for comprehensive sedation comparisons in mechanically ventilated critically ill adults. An updated and rigorous NMA incorporating these agents would provide more robust evidence to guide daily clinical practice.

Fig. 1

Bayesian network meta-analysis of mortality associated with isoflurane, sevoflurane, propofol, and midazolam in mechanically ventilated adults, presented as half-eye plots. (A) Posterior odds ratios relative to midazolam. Shaded areas represent full posterior distributions; the vertical line marks an odds ratio (OR) of 1. Blue shading indicates OR > 1 (increased mortality versus midazolam); pink shading indicates OR < 1 (decreased mortality versus midazolam). Black dots mark posterior median ORs with black lines indicating 95% credible intervals. The probability of each agent having lower mortality than midazolam (OR < 1) was: propofol 30.5%, isoflurane 31.2%, and sevoflurane 13.0%. (B) Absolute mortality probabilities based on pooled baseline risk from midazolam trials. Ridgeline density plots illustrate posterior distributions, color-coded by mortality risk categories: <5% (green), 5–15% (orange), and > 15% (red). Percentages indicate the proportion of samples in each category. Blue dots mark posterior median probabilities with blue lines indicating 95% credible intervals. Gray dashed/dotted lines and gray text represent the sensitivity analysis excluding Jabaudon 2025

Full size image

The original data were derived from the manuscript provided by the journal.

1. Yamamoto T, Kotani Y, Akutagawa K, Nagayama T, Tomimatsu M, Tonai M, Karumai T, Hayashi Y. Volatile sedation in critically ill adults undergoing mechanical ventilation: a meta-analysis of randomized controlled trials. Crit Care. 2025;29(1):227.

   Article PubMed PubMed Central Google Scholar

2. Shea BJ, Reeves BC, Wells G, Thuku M, Hamel C, Moran J, Moher D, Tugwell P, Welch V, Kristjansson E, et al. AMSTAR 2: a critical appraisal tool for systematic reviews that include randomised or non-randomised studies of healthcare interventions, or both. BMJ. 2017;358:j4008.

   Article PubMed PubMed Central Google Scholar

3. Dale O, Brown BR Jr. Clinical pharmacokinetics of the inhalational anaesthetics. Clin Pharmacokinet. 1987;12(3):145–67.

4. de Wit F, van Vliet AL, de Wilde RB, Jansen JR, Vuyk J, Aarts LP, et al. The effect of Propofol on haemodynamics: cardiac output, venous return, mean systemic filling pressure, and vascular resistances. Br J Anaesth. 2016;116(6):784–9.

   Article PubMed Google Scholar

5. Celis-Rodriguez E, Diaz Cortes JC, Cardenas Bolivar YR, Carrizosa Gonzalez JA, Pinilla DI, Ferrer Zaccaro LE, et al. Evidence-based clinical practice guidelines for the management of sedoanalgesia and delirium in critically ill adult patients. Med Intensiva (Engl Ed). 2020;44(3):171–84.

   CAS PubMed Google Scholar

6. Mesnil M, Capdevila X, Bringuier S, Trine PO, Falquet Y, Charbit J, et al. Long-term sedation in intensive care unit: a randomized comparison between inhaled sevoflurane and intravenous propofol or midazolam. Intensive Care Med. 2011;37(6):933–41.

   Article CAS PubMed Google Scholar

7. Garcia R, Salluh JIF, Andrade TR, Farah D, da Silva PSL, Bastos DF, et al. A systematic review and meta-analysis of Propofol versus Midazolam sedation in adult intensive care (ICU) patients. J Crit Care. 2021;64:91–9.

   Article PubMed Google Scholar

8. Phillippo DM. multinma: Bayesian Network Meta-Analysis of Individual and Aggregate Data. In., R package version 0.8.1 edn; 2025.

9. Spiegelhalter David J, Best Nicola G, Carlin Bradley P, van der Linde A. Bayesian measures of model complexity and fit. J R Stat Soc Series B Stat Methodol. 2002;64(4):583–639.

   Article Google Scholar

10. Meng X-L. Posterior predictive p-values. Ann Stat. 1994;22(3):1142–60.

    Article Google Scholar

11. Jabaudon M, Quenot JP, Badie J, Audard J, Jaber S, Rieu B, Varillon C, Monsel A, Thouy F, Lorber J, et al. Inhaled sedation in acute respiratory distress syndrome: the SESAR randomized clinical trial. JAMA. 2025;333(18):1608–17.

    Article CAS PubMed PubMed Central Google Scholar

12. Patel S, Green A. Death by p-value: the overreliance on p-values in critical care research. Crit Care. 2025;29(1):73.

    Article PubMed PubMed Central Google Scholar

13. Ho GWK, Thaarun T, Ee NJ, Boon TC, Ning KZ, Cove ME, et al. A systematic review on the use of Sevoflurane in the management of status asthmaticus in adults. Crit Care. 2024;28(1):334.

    Article PubMed PubMed Central Google Scholar

14. Lewis K, Alshamsi F, Carayannopoulos KL, Granholm A, Piticaru J, Al Duhailib Z, Chaudhuri D, Spatafora L, Yuan Y, Centofanti J, et al. Dexmedetomidine vs other sedatives in critically ill mechanically ventilated adults: a systematic review and meta-analysis of randomized trials. Intensive Care Med. 2022;48(7):811–40.

    Article PubMed Google Scholar

15. Tang Y, Gao X, Xu J, Ren L, Qi H, Li R, et al. Remimazolam besylate versus propofol for deep sedation in critically ill patients: a randomized pilot study. Crit Care. 2023;27(1):474.

    Article PubMed PubMed Central Google Scholar

Download references

We express our gratitude to Professor Yu-Kang Tu from the Institute of Epidemiology and Preventive Medicine, College of Public Health, National Taiwan University, Taipei, Taiwan, whose workshop provided us with the expertise to conduct Bayesian meta-analysis using the multinma package.

This research received no external funding.

Authors and Affiliations

1. Division of infection disease, Department of Medicine, Chi-Mei Medical Center, Tainan City, Taiwan

   Po-An Su

2. Education Center, College of Medicine, National Cheng Kung University Hospital, National Cheng Kung University, Tainan City, Taiwan

   Pei‑Chun Lai

3. Department of Pediatrics, College of Medicine, National Cheng Kung University Hospital, National Cheng Kung University, Tainan City, Taiwan

   Pei‑Chun Lai

4. Department of Surgery, College of Medicine, National Cheng Kung University Hospital, National Cheng Kung University, No. 138, Shengli Road, Tainan City, 701, Taiwan

   Yen-Ta Huang

Authors

1. Po-An SuView author publications

   Search author on:PubMed Google Scholar

2. Pei‑Chun LaiView author publications

   Search author on:PubMed Google Scholar

3. Yen-Ta HuangView author publications

   Search author on:PubMed Google Scholar

Contributions

Methodology: YT Huang; Original draft writing: PA Su; Formal analysis: PC Lai; Writing—review and editing: YT Huang; Project administration: YT Huang. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Yen-Ta Huang.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

Cite this article

Su, PA., Lai, P. & Huang, YT. Is volatile sedation truly associated with increased mortality in mechanically ventilated critically ill adults compared to intravenous sedation? Moving beyond pairwise meta-analysis to individual agent assessment via bayesian network meta-analysis. Crit Care 29, 435 (2025). https://doi.org/10.1186/s13054-025-05681-4

Download citation

• Received: 29 June 2025

• Accepted: 22 September 2025

• Published: 15 October 2025

• DOI: https://doi.org/10.1186/s13054-025-05681-4

Share this article

Anyone you share the following link with will be able to read this content:

Get shareable link

Sorry, a shareable link is not currently available for this article.

Copy shareable link to clipboard

Provided by the Springer Nature SharedIt content-sharing initiative