Sedation with Sevoflurane versus Propofol in COVID-19 Patients with Acute Respiratory Distress Syndrome: Results from a Randomized Clinical Trial

Background: Acute respiratory distress syndrome (ARDS) related to COVID-19 (coronavirus disease 2019) led to intensive care units (ICUs) collapse. Amalgams of sedative agents (including volatile anesthetics) were used due to the clinical shortage of intravenous drugs (mainly propofol and midazolam). Methods: A multicenter, randomized 1:1, controlled clinical trial was designed to compare sedation using propofol and sevoflurane in patients with ARDS associated with COVID-19 infection in terms of oxygenation and mortality. Results: Data from a total of 17 patients (10 in the propofol arm and 7 in the sevoflurane arm) showed a trend toward PaO2/FiO2 improvement and the sevoflurane arm’s superiority in decreasing the likelihood of death (no statistical significance was found). Conclusions: Intravenous agents are the most-used sedative agents in Spain, even though volatile anesthetics, such as sevoflurane and isoflurane, have shown beneficial effects in many clinical conditions. Growing evidence demonstrates the safety and potential benefits of using volatile anesthetics in critical situations.


Introduction
Coronavirus disease (COVID-19) has become a significant worldwide challenge for health care providers. The wide variety of COVID-19 symptoms developed have ranged from mild headache or isolated cough to severe respiratory failure. Many patients affected by COVID-19 developed ARDS and required ICU admission for invasive mechanical ventilation (IMV). Oxygenation impairment and increased mortality rates have characterized this worldwide health crisis [1,2]. Impaired oxygenation in the ARDS context usually requires IMV support, and sedation is an integral part of therapy for this kind of patient. Titratable light-to-deep sedation may control neurological manifestations and help optimize ventilatory settings and endotracheal tube tolerance, although sometimes it is also necessary to use neuromuscular relaxants. Within the variety of sedative agents used in ICUs, there are two large groups classified according to administration route: intravenous and inhaled agents. Intravenous drugs represent routine clinical practice worldwide [3,4]: benzodiazepines (midazolam, lorazepam, and diazepam), propofol, and ketamine are commonly combined with opioids to achieve analgo-sedation. In contrast, volatile anesthetics (isoflurane and sevoflurane) agents on oxygenation in patients with ARDS due to COVID-19 infection. The initial idea was to recruit all consecutive patients admitted in participant critical care units, but emergency situations, patients' social problems, lack of organization, and workers on sick leave made it difficult to perform. Our team achieved poor patient recruitment, but we decided to analyze these data so we can draw some conclusions to continue with our study.

Study Population, Setting and Data Collection
After review and approval by the ethics committee of the INCLIVA Health Research Institute, our team developed a multicenter, national, randomized 1:1, controlled, parallel, open study registered as NCT04359862, in which patients with ARDS due to COVID-19 were included in the first 24 h after diagnosis (Figure 1). The participant centers in this trial were four third-level hospitals in Spain which were referents during the COVID pandemic: Hospital Clínic Universitari of Valencia (Valencia), Consorcio Hospital General Universitario of Valencia (Valencia), Hospital Universitario Ramón y Cajal (Madrid), and Hospital Universitario La Paz (Madrid). The initial sample size took into consideration the high number of daily admissions in our units, but organizational problems made the recruitment rate much lower than expected. This trial was carried out during the year 2020. Each patient who met the inclusion criteria and none of the exclusion criteria was proposed to participate in the study. When the informed consent was signed by the patient (or relatives when patient could not sign), randomization was performed to a treatment arm: sedation using sevoflurane (SEV) or propofol (PROP). In the PROP group, propofol was administered with volumetric pumps (Alaris GW and GP Plus), and sedation levels were evaluated using BIS ® technology (Medtronic Covidien, Spain) . In the SEV group, sevoflurane was administered through an AnaConDa ® device coupled to a Con-traFluran ® scavenging device, and CAM was measured with a SedLine ® monitor. Inclusion criteria were: age ≥ 18 years old, need for sedation, ARDS due to COVID-19, and accepted informed consent by the patient or a relative. Exclusion criteria included intracranial hypertension, allergy to any sedative agent, tidal volume < 250 mL, previous malignant hyperthermia or risk of developing malignant hyperthermia, hepatic failure, neutropenia, pregnancy, or chemotherapy in the previous month. All randomized patients received remifentanil as an analgesic and cisatracurium as a neuromuscular relaxant. A lung-protective ventilation strategy was carried out: VT 6 mL/kg, PEEP > 5 cmH2O, Plateau pressure < 30 cmH2O, respiratory rate < 35 rpm, and I:E ratio ≤ 1:2 [24]. This study protocol is publicly available for verification [25].
Patients were anonymized using a numerical code for data collection. Variables recorded included: anthropometric values, respiratory and hemodynamic parameters, and blood and bronchoalveolar fluid samples at day of randomization (D0), at 24 h (D1), and Each patient who met the inclusion criteria and none of the exclusion criteria was proposed to participate in the study. When the informed consent was signed by the patient (or relatives when patient could not sign), randomization was performed to a treatment arm: sedation using sevoflurane (SEV) or propofol (PROP). In the PROP group, propofol was administered with volumetric pumps (Alaris GW and GP Plus), and sedation levels were evaluated using BIS ® technology (Medtronic Covidien, Spain). In the SEV group, sevoflurane was administered through an AnaConDa ® device coupled to a ContraFluran ® scavenging device, and CAM was measured with a SedLine ® monitor. Inclusion criteria were: age ≥ 18 years old, need for sedation, ARDS due to COVID-19, and accepted informed consent by the patient or a relative. Exclusion criteria included intracranial hypertension, allergy to any sedative agent, tidal volume < 250 mL, previous malignant hyperthermia or risk of developing malignant hyperthermia, hepatic failure, neutropenia, pregnancy, or chemotherapy in the previous month. All randomized patients received remifentanil as an analgesic and cisatracurium as a neuromuscular relaxant. A lung-protective ventilation strategy was carried out: V T 6 mL/kg, PEEP > 5 cmH 2 O, Plateau pressure < 30 cmH 2 O, respiratory rate < 35 rpm, and I:E ratio ≤ 1:2 [24]. This study protocol is publicly available for verification [25].
Patients were anonymized using a numerical code for data collection. Variables recorded included: anthropometric values, respiratory and hemodynamic parameters, and blood and bronchoalveolar fluid samples at day of randomization (D0), at 24 h (D1), and at 48 h (D2) (see Table 1). Thirty days after randomization (D30), the following data were collected: duration and control of mechanical ventilation (MV), ventilator-free days (VFD), length of ICU stay, and mortality at D2 and D30.

Hypothesis and Objectives
The hypothesis of this clinical trial was that sedation using sevoflurane in patients with ARDS associated with COVID-19 infection improves oxygenation. The primary objective included the evaluation of oxygenation in randomized patients during the first 48 h, measured via PaO 2 /FiO 2 . The secondary objective included assessment of mortality rates at D30.

Statistical Analysis
All statistical comparisons were based on the intention-to-treat principle. Sample size was calculated using results of previous works [7,14]. The Student's t-test and the Mann-Whitney test were used as appropriate. Chi-square and Fisher's tests were used for categorical variables. Differences in PaO 2 /FiO 2 for D1 and D2 between the two treatment arms were determined using mixed regression analysis for repeated measures. This method also included an ANCOVA-type design; baseline values of the PaO 2 /FiO 2 variable were added to the regression model as covariates, interacting with the treatment variable. The analysis was also adjusted for the potential of autocorrelation between repeated measures in the same subject and for the nesting effect (due to the study center) through "random intercept" effects. The Kaplan-Meier method, Cox regression, and restricted median survival time (RMST) were used to compare D30 survival rates between the two study groups. Statistical analysis was performed using Stata version 16.1 (StataCorp. 2021. Stata Statistical Software: Release 16. College Station, TX, USA: StataCorp LP).

Primary Results
Analyses were based on a total of 17 patients: 10 patients in the PROP arm and 7 patients in the SEV arm. In Valencia, HCUV recruited seven patients and Consorcio Hospital General Universitario recruited three. In Madrid, Ramón y Cajal Hospital recruited one patient and La Paz Hospital recruited six. For each patient, multiple variables were collected on D1 and D2, and mortality was noted on D30. Regarding the patients' baseline characteristics, no statistically significant differences were found in any of the variables listed in Table 2. No statistically significant difference was found regarding several comorbidities (stroke, arterial hypertension, diabetes mellitus, dyslipemia, smoking habit, chronic kidney injury (CKI), and previous corticosteroid therapy). The calculated severity index (via SAPS-II [26] and LIS [27]) did not show differences between groups. Patients in the PROP arm spent an average 16.4 days in the ICU, and patients in the SEV arm spent an average 20.6 in the ICU (p = 0.563); the average days under IMV were 13 (PROP) vs. 14.6 (SEV), with the average IMV-free days in the ICU being 5.1 vs. 5.2, respectively. The mean number of days from randomization to death was 28 vs. 30, respectively (p = 0.495). Regarding ventilatory settings, no significant differences were found at baseline (Table 3) or at D1 (Table 4). At D2 (Table 5), statistically significant differences among groups were found; respiratory acidosis developed in the SEV group, probably related to differences in end-expiratory lung volumes.   Figure 2 illustrates the changes in PaO 2 /FiO 2 between the two treatment arms at the baseline (before treatment), and at D1 and D2 (post-treatment), as previously shown in Tables 2-5. As shown in Figure 2, there were differences between the groups before randomization, with higher PaO 2 /FiO 2 values in the SEV group (p = 0.246). Differences persisted at D1 and minimized at D2.  Figure 3 presents the core of the analysis: the effect of the randomized treatment on PaO2/FiO2 (at D1 and D2, post-randomization). It was adjusted by the baseline value of PaO2/FiO2 according to the ANCOVA design. Using mixed regression analysis in the context of ANCOVA, the results showed an improvement in the PaO2/FiO2 ratio in the SEV arm at D1 and D2; however, results were statistically significant only at D1** (Figure 3).  Figure 3 presents the core of the analysis: the effect of the randomized treatment on PaO 2 /FiO 2 (at D1 and D2, post-randomization). It was adjusted by the baseline value of PaO 2 /FiO 2 according to the ANCOVA design. Using mixed regression analysis in the context of ANCOVA, the results showed an improvement in the PaO 2 /FiO 2 ratio in the SEV arm at D1 and D2; however, results were statistically significant only at D1** (Figure 3).

Effect over 30-Day Mortality
Kaplan-Meier analysis (Figure 4) showed an intersection of survival curves along the trace, which made it difficult to interpret and find the meaning of the Log-rank test (p = 0.584). Therefore, we tested a time-dependent effect for the PaO 2 /FiO 2 -mortality relationship. On average, patients treated using sevoflurane survived 1.66 days longer than those treated using propofol when patients were followed up at D30 ( Figure 5). Unfortunately, this difference did not achieve statistical significance (95% CIs = −11.00 to 14.33). Figure 6 shows a time-dependent effect with a higher mortality risk at the beginning of the study for the SEV arm, but this risk decreased to become protective in the following period. Unfortunately, confidence intervals reached the 1-line on the Y-edge, so the analysis did not achieve statistical significance. Figure 3 presents the core of the analysis: the effect of the randomized treatment on PaO2/FiO2 (at D1 and D2, post-randomization). It was adjusted by the baseline value of PaO2/FiO2 according to the ANCOVA design. Using mixed regression analysis in the context of ANCOVA, the results showed an improvement in the PaO2/FiO2 ratio in the SEV arm at D1 and D2; however, results were statistically significant only at D1** (Figure 3). FOR PEER REVIEW 8 of 14

Effect over 30-Day Mortality
Kaplan-Meier analysis ( Figure 4) showed an intersection of survival curves along the trace, which made it difficult to interpret and find the meaning of the Log-rank test (p = 0.584). Therefore, we tested a time-dependent effect for the PaO2/FiO2-mortality relationship. On average, patients treated using sevoflurane survived 1.66 days longer than those treated using propofol when patients were followed up at D30 ( Figure 5). Unfortunately, this difference did not achieve statistical significance (95% CIs = −11.00 to 14.33). Figure 6 shows a time-dependent effect with a higher mortality risk at the beginning of the study for the SEV arm, but this risk decreased to become protective in the following period. Unfortunately, confidence intervals reached the 1-line on the Y-edge, so the analysis did not achieve statistical significance.    The prognostic effect of PaO2/FiO2 differences between D1 and D2 compared to the basal values for the SEV and PROP arms was tested using sensitivity analysis. Figure 7 shows Cox regression analysis results for D30 mortality, in which no statistical significance was found for changes in PaO2/FiO2. This showed that the effect of treatment on mortality did not depend on changes in PaO2/FiO2 measured at D1 and D2 compared to basal values.   The prognostic effect of PaO2/FiO2 differences between D1 and D2 compared to t basal values for the SEV and PROP arms was tested using sensitivity analysis. Figur shows Cox regression analysis results for D30 mortality, in which no statistical sign cance was found for changes in PaO2/FiO2. This showed that the effect of treatment mortality did not depend on changes in PaO2/FiO2 measured at D1 and D2 compared basal values. The prognostic effect of PaO 2 /FiO 2 differences between D1 and D2 compared to the basal values for the SEV and PROP arms was tested using sensitivity analysis. Figure 7 shows Cox regression analysis results for D30 mortality, in which no statistical significance was found for changes in PaO 2 /FiO 2 . This showed that the effect of treatment on mortality did not depend on changes in PaO 2 /FiO 2 measured at D1 and D2 compared to basal values.
x FOR PEER REVIEW 10 of 14

Limitations
The main limitation of this clinical trial was its sample size. The health emergency situation in which this project was carried out led to the loss of recruitment and clinical data (due to work overload and difficulties in obtaining informed consent). In addition, assessment of oxygenation, levels of inflammatory mediators, and mortality in patients admitted to the ICU for ARDS associated with COVID-19 were probably influenced by unknown factors due to the incipient development of infection by this virus. Moreover, the study's initial protocol included cytokine level measurements; however, these could not be assessed due to laboratory issues (overwhelmed by COVID-19 tests) and storage limitations.

Generalizability and Interpretation
Intravenous agents (mainly propofol and benzodiazepines (3)) have been the mostemployed deep sedation drugs in ICUs worldwide. That is surprising, as the use of benzodiazepines has been associated with decreased ventilator-free days, increased risk of delirium, and worse long-term outcomes [28,29], so non-benzodiazepine strategies should be preferred for ICU sedation.
The COVID-19 pandemic emptied all hospital sedative stocks in just a few months. The main reason for this was that ICU capacities were overrun with an increase in invasively mechanically ventilated ARDS patients who required deep sedation combined with muscle relaxation to achieve a depth of sedation sufficient to avoid patient-ventilator dys-

Limitations
The main limitation of this clinical trial was its sample size. The health emergency situation in which this project was carried out led to the loss of recruitment and clinical data (due to work overload and difficulties in obtaining informed consent). In addition, assessment of oxygenation, levels of inflammatory mediators, and mortality in patients admitted to the ICU for ARDS associated with COVID-19 were probably influenced by unknown factors due to the incipient development of infection by this virus. Moreover, the study's initial protocol included cytokine level measurements; however, these could not be assessed due to laboratory issues (overwhelmed by COVID-19 tests) and storage limitations.

Generalizability and Interpretation
Intravenous agents (mainly propofol and benzodiazepines (3)) have been the mostemployed deep sedation drugs in ICUs worldwide. That is surprising, as the use of benzodiazepines has been associated with decreased ventilator-free days, increased risk of delirium, and worse long-term outcomes [28,29], so non-benzodiazepine strategies should be preferred for ICU sedation.
The COVID-19 pandemic emptied all hospital sedative stocks in just a few months. The main reason for this was that ICU capacities were overrun with an increase in invasively mechanically ventilated ARDS patients who required deep sedation combined with muscle relaxation to achieve a depth of sedation sufficient to avoid patient-ventilator dyssynchrony. At the beginning of the pandemic, some groups emphasized that affected patients required high sedative doses; although some possible underlying reasons for this were unknown, rational reasons included that patients were younger, without co-morbidities, and most of them needed to toggle prone position.
This situation necessitated opening the therapeutic arsenal to other options, such as volatile anesthetics and multimodal sedative approaches, to avoid adverse effects related to propofol (such as hyperlipidemia or propofol infusion syndrome (PRIS)) or the abuse of neuromuscular relaxants or analgesics (mainly opioids). Inhaled sedation (using isoflurane or sevoflurane) had already demonstrated faster and improved recovery after prolonged sedation [30,31], minimized delirium, an analgesic sparing effect [32][33][34], decreased pulmonary inflammation [14,35], improved oxygenation in patients with ARDS [5], and decreased mortality in long-term ventilated patients [36]. Sevoflurane has been used in many ICUs since AnaConDa© was designed and the accuracy of the pharmacokinetic model was published [37]. Some guidelines included its use in critically ill patients with ARDS for a moderate-to-deep sedation more than ten years ago [38,39]. Still, many Spanish ICUs did not switch to inhaled sedation until they ran out of propofol; the main reasons provided were that staff were unfamiliar with the volatile agent or its specific device (AnaConDa©), even though it was widely used for intraoperative anesthetic maintenance.
The primary role that propofol and midazolam have in ICUs increases the complexity of changing the routine from intravenous sedation in order to assess the spectrum of sedative agents that we currently use. Profiling the potential benefit of a hypnotic agent over oxygenation is complicated given the interference of multiple variables that probably act as confounding factors in a critically ill patient. However, it is important to keep in mind that the profile of these drugs (that act at so many levels) must be recognized in order to adapt decisions based on the patient being treated [40]. In addition to its function as sedative agent, sevoflurane has many intrinsic characteristics with potential therapeutic benefits that could be especially relevant to ICU patients: it is an easy-totitrate drug with shorter wake-up times, it has enhanced effect over analgesics (decreased use of opioids) and neuromuscular relaxants, and it has fewer vasopressor requirements compared to midazolam and propofol. All of these potential benefits should be taken into consideration [34,41].
Our study had inherent design limitations that made it difficult to draw categorical conclusions. However, this study highlights the feasibility of using sevoflurane as a primary sedative agent in ARDS patients. Both PROP and SEV treatment arms were comparable regarding patients' characteristics, co-morbidities, and ventilatory settings. Regarding the PaO 2 /FiO 2 ratio, it tended to improve in the SEV group both at D1 and D2; however, results were statistically significant only for D1. The SEV arm experienced longer ICU stays and longer days under MV; however, the number of days from randomization to death was longer in the SEV group. Moreover, patients on sevoflurane survived longer than those on propofol when patients were followed up at 30 days, even if results were not statistically significant. The reason for not achieving significance could be both low sample size and days under inhaled sedation; a retrospective study in surgical ICU ventilated patients (n = 128) with inhaled drugs used for more than 96 h demonstrated more ventilator-free days at day 60, more hospital-free days at 6 months, and decreased mortality compared with patients under intravenous sedation receiving midazolam or propofol [36]. Our results are consistent with previous analysis [30,[42][43][44][45], which found no differences between inhaled and intravenous sedation in deaths or length of ICU stay. An international retrospective study including 10 ICUs published in 2022 [46] found no association between inhaled sedation in COVID-19 patients and the number of ventilator-free days through to day 28; this suggests that the effect of treatment on mortality probably does not depend on the resulting changes in the PaO 2 /FiO 2 ratio.
Regarding oxygenation, studies performed in mice, rat, and pig models of ARDS found that inhaled agents reduced alveolar and systemic levels of pro-inflammatory cytokines [7,14,[47][48][49], improved arterial oxygenation, and decreased lung alveolar oedema [7,50]. Results of this work agree with previous publications in which the potential benefit of sevoflurane over oxygenation was observed [15,40]; however, more studies recruiting a higher number of patients are needed to support the use of inhaled agents. Few studies registered on the ICH GCP website include objectives regarding the study of sevoflurane in patients with moderate to severe ARDS diagnoses. Even so, volatile agents (sevoflurane and isoflurane) are used as alternatives to intravenous sedation in ICUs by an increasing number of physicians [51] as monotherapy or as part of a combined therapy [52]. There are some detractors because of volatile agents' potential adverse events [53], but there is relevant literature that supports their feasibility and safety of use, without the risk of tolerance or effects on renal or liver function [30,34,45,54,55].
In our unit's experience, managing one more drugs, whether or not they are superior to another agent, allows us to provide alternatives that can be beneficial to our various patients. Therefore, while waiting for new studies, the use of inhaled sedation with sevoflurane as the first line in patients affected by ARDS must be considered, and not only as second or third-line treatment, as recommended recently [28].

Conclusions
This study has demonstrated that sedation using sevoflurane improved oxygenation and increased survival times in patients affected by ARDS due to COVID-19 infection compared to propofol. Hence, in patients with ARDS who require sedation, sevoflurane is a safe and effective option that, in addition to its main purpose, has a beneficial effect on oxygenation and survival. Therefore, it could be considered as a first-choice strategy for this patient profile.  Informed Consent Statement: Informed consent was obtained from all subjects, or relatives in charge, involved in the study. All signed consents are attached to each patient's medical records.
Data Availability Statement: Data supporting reported results are available from the corresponding author at any time.

Conflicts of Interest:
The authors declare no conflict of interest.