A systematic search of studies published over the past 10 years (2011–2021) was carried out in PubMed, Medline, Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library, Google Scholar and the Russian Science Citation Index (RSCI) by four independent researchers (NE, LB, MY and KK). The search was carried out in the form of queries with details available in the supplemental material (Supplemental appendix 1). Additionally, the authors used the backward snowballing method (analysis of references of included articles and retrieved reviews) for further studies. We did not restrict search by language. Medical Subject Headings (MeSH) terms were applied.

The links obtained from the database were first independently examined at the title/annotation level by two researchers (NE and LB). Randomized controlled trials, prospective and retrospective cohort studies comparing inhaled (volatile) versus i/v sedation in the intensive care unit were considered. After removing duplicates, the appropriate publications were selected. The final decision on inclusion in this study was made based on the analysis of full-text articles. Divergences were resolved by consensus.

For this study, the following inclusion criteria were used: adult patients (≥18 years) who underwent inhalation or i/v sedation in the intensive care unit (no restrictions on dose or time of administration). The exclusion criteria were mixed groups (both inhaled and i/v sedation were used), small sample size (less than 25 patients in two groups), duplicate publications, animal studies, clinical guidelines, articles without a comparison group.

Basic information about the study (design, sample size, intervention plan, inclusion criteria), information about the study objects (age, gender, underlying disease and surgery), types of anesthetics and doses, target sedation level, anesthetic-conserving device, ICU treatment outcomes (mortality, duration of mechanical ventilation, troponin level at 1 postoperative day, ventilator-free days, time to extubation, awakening time, length of stay (LOS) in ICU, hospital LOS, catecholamine requirements) was collected. We have not considered the type of anesthetic used in the surgery room. The data were independently extracted by two researchers (NE and MY) and subsequently compared with each other for verification. The primary outcome of this study was all-cause mortality (30-day mortality). Secondary endpoints were: length of stay in ICU and length of hospitalization (days), duration of mechanical ventilation (days) and ventilator-free days, catecholamine requirements, cardiac troponin levels on the first postoperative day, time to extubation (min), and awakening time (min).

The internal validity and risk of bias of the included studies were assessed by two peer reviewers (MY and LB) and peer-reviewed by a third (VL) according to the latest version of the “ROB 2” (the Cochrane tool for assessing risk of bias in randomized trials) and “ROBINS-I” (Risk Of Bias In Non-randomized Studies of Interventions) tools.20,21 Discrepancies in assessment were resolved by consensus. Publication bias was assessed using the Egger's test (MedCalc Statistical Software, version 19.5.6),22,23 and also by visual examination of the “funnel plot” charts. We also used a GRADE systematic approach to rate the certainty of evidence.24 Two review authors (MY and LB) worked independently to assess the quality of evidence, disagreements were resolved by consensus.

To calculate and visualize the results of the meta-analysis in forest plots, the Cochrane tool “RevMan, version 5.3” was used. The heterogeneity of studies was assessed using the Cochran's Q test, and the degree of statistical agreement was measured using the coefficient of heterogeneity I2. The quantitative results of individual studies were brought to the form “mean±standard deviation” and the standardized difference of mean values (SMD) and its 95% confidence interval were calculated. Since many quantitative parameters in this meta-analysis a priori have a deeply skewed distribution, we used the log transformation for summary data of ventilator-free days, time to extubation, time to awakening and ICU/hospital length of stay.25 We used Cochrane handbook recommendations to re-express SMDs using rules of thumb for effect sizes (<0.40=small effect, 0.40–0.70=moderate effect, >0.70=large effect).20 Binary research results were used to calculate the odds ratio with the corresponding 95% confidence interval (CI) using the inverse variance method (Mantel–Haenszel method). For a pooled estimate of the magnitude of the standardized mean difference, two models were used: a fixed-effects model (in the case of low statistical inconsistency, I2<50%) and a random-effects model (I2≥50% and/or p<0.05). Statistical significance was set at 0.05 for hypothesis testing.

Sensitivity analyses were performed by considering additional subgroups – analyzing only studies with a low or moderate risk of bias (Hi-QOL studies), only RCT/non-randomized studies and by sequentially removing each study and reanalyzing the remaining data set (producing a new analysis for each study removed). Additionally, a subgroup of studies which used a prolonged sedation (>12h) regimen was evaluated.

During the initial search, 427 articles were found, of which 31 were eligible. Upon careful reading of the full-text articles, 16 studies were excluded (Supplemental Table 1). Ultimately, 15 full-text articles published between January 2011 and July 2021were included in the meta-analysis. A flowchart illustrating the study selection process is presented in Fig. 1.

Figure 1.

Flowchart used to select the final 15 manuscripts.

(0,43MB).

The characteristics of the included studies are summarized in Table 1.

The meta-analysis included 1520 patients: VA (628 patients); i/v sedation (892 patients). Among the 15 included studies, 9 were RCTs. The following settings were represented: seven studies in sedation in postoperative cardiac surgery patients,27–29,31,34,37,39 two articles in postoperative surgical patients,26,32 one study in patients who were assigned to more than 24h of mechanical ventilation sedation,36 five studies in the intensive care unit critically ill patients (non-traumatic cardiac arrest and successful CPR,33,40 sepsis-associated delirium38 and moderate to severe ARDS30,35).

Fourteen included studies used the AnaConDa device in the ICU for inhaled sedation, and only one study used MIRUS.28 Among the 15 included studies, 10 compared sevoflurane with propofol, 4 compared isoflurane with propofol and midazolam, and one study had multiple comparisons.

Four studies30,32,36,38 used long-term sedation (12h or more). Five of 9 RCTs and two of six non-randomized studies had medium or high quality (Supplemental Figures 1 and 2).

Fourteen studies (8 RCTs) including 1453 patients reported mortality data which were not different between groups 112/609 (18.4%) in the VA group versus 202/844 (23.9%) in the i/v group (Odds Ratio (OR)=0.82 [0.60–1.12]; p-value for effect is 0.20; p-value for heterogeneity=0.15; I2=34%, Fig. 2; Table 2).

Figure 2.

Forest plot for hospital mortality representing the odd's ratio for volatile vs. intravenous sedation effects on all-cause mortality for the included studies. The plot displays the study, sample size, odds ratio (OR), confidence interval (CI), and p-value. The size of the squares indicates the weight of the studies (taking into account sample size and standard deviations); the diamond represents the pooled OR with CI. Hi-QOL – studies with low–moderate risk of bias.

(0,47MB).

Analysis of subgroups of RCTs, non-randomized, Hi-QOL studies (studies with low–moderate risk of bias) and studies with long-term sedation did not reveal an association of sedation with mortality. Funnel plot for mortality is presented in Supplement (Supplemental Figure 3).

A pooled analysis of data from five studies (606 patients, four studies with short-term sedation, two RCTs) showed that patients on VA had an increase in ventilator-free days (moderate effect size: SMD=0.46 [0.28–0.64]; p-value for the effect <0.001; p-value for heterogeneity=0.79; I2=0%, Fig. 3; Table 2). The results were confirmed in all sub-analyses: RCTs (p=0.003), non-randomized (p<0.001) and Hi-QOL studies (2 trials, p=0.004).

Figure 3.

Forest plot for ventilator-free days. The plot displays the study, sample size, log-transformed standardized mean difference (SMD), confidence interval (CI), and p-value. The size of the squares indicates the weight of the studies (taking into account sample size and standard deviations); the diamond represents the pooled SMD with CI. Hi-QOL – studies with low–moderate risk of bias.

(0,28MB).

A pooled analysis of 987 patients (9 studies, 5 RCTs) showed that sedation with VA was associated with a reduction in duration of mechanical ventilation (SMD=−0.46 [−0.88 to −0.04] – moderate effect size; p-value for effect=0.03; p-value for heterogeneity <0.001; I2=88%, Supplemental Figure 4, Table 2). The results were confirmed in Hi-QOL studies (5 trials, p<0.001), but not when considering only RCTs (p=0.09) or non-randomized studies (p=0.29) (Table 2). When considering only studies using prolonged sedation there was also no significant effect (p=0.51).

Four studies with short-term sedation using sevoflurane (3 RCTs, 350 patients) reported cardiac troponin levels (1 day after surgery) with postoperative cardiac surgery patients on VA having a statistically significantly lower troponin levels (moderate effect size: SMD=−0.52 [from −0.84 to −0.20]; p-value for effect 0.001; p-value for heterogeneity=0.10; I2=52%, Supplemental Figure 5; Table 2). The result was robust when considering RCTs (p=0.04) but not Hi-QOL studies (2 trials, p=0.21).

Data from 4 studies (all RCTs, 3 with short-term sedation) in 368 patients showed that sedation with VA was associated with a decrease in time to extubation in both medical and surgical patients (large effect size: SMD=−1.59 [−2.26 to −0.91]; p-value for effect <0.001; p-value for heterogeneity <0.001; I2=87%, Supplemental Figure 6; Table 2). Results on awakening time (large effect size: SMD=−1.30 [−2.54 to −0.06]; p-value for effect 0.04; p-value for heterogeneity <0.001; I2=95%) were confirmed when non-randomized studies were excluded (p=0.04, Supplemental Figure 7 and Table 2). The difference in the awakening time was established only for surgical patients, not for medical ones. However, only one study was of high quality. Moreover, the time to extubation and the awakening time were less in VA group than in i\v group, regardless of the duration of sedation (p<0.05). Despite the depth of sedation across the majority of the studies were approximately similar (from RASS=−1 to RASS=−3), these results were difficult to compare them to the ones in the other studies where the depth of sedation were measured according to other scales.

No differences were observed in length of ICU stay (p=0.93), length of hospital stay (p=0.69), and need for catecholamines (p=0.5, Table 2 and Supplemental Figures 8–10).

Certainty of evidence for all studied outcomes was qualified using GRADE approach for RCTs. Overall, very low quality of evidence shows that volatile sedation has no impact on hospital mortality. Certainty of evidence for other outcomes ranged from very low to high. The reasons for the decrease in the quality of evidence are summarized and presented in Table 3.