Introduction
One-lung ventilation (OLV) is an essential component of thoracoscopic surgery, providing optimal exposure. However, OLV can rapidly increase the risk of hypoxemia due to the development of intrapulmonary shunts and increased dead spaces [1]. A key physiological mechanism to counteract this effect is hypoxic pulmonary vasoconstriction (HPV), which redistributes pulmonary blood flow away from the non-ventilated lung toward the ventilated lung, thereby reducing the shunt fraction and preserving arterial oxygenation [2]. The HPV response is modulated by several factors, including anesthetic agents, cardiac output (CO), acid-base balance, and body temperature [3].
Among anesthetic agents, volatile anesthetics such as sevoflurane have been shown to inhibit HPV in a dose-dependent manner [4–6], whereas intravenous agents such as propofol tend to preserve HPV [4]. Nonetheless, clinical studies have reported that the impact on oxygenation may not differ significantly between these anesthetic agents when used at standard clinical doses, particularly at 1 minimum alveolar concentration (MAC) of volatile agents [7–9].
Remimazolam, a novel ultra-short-acting benzodiazepine, has emerged as a promising intravenous anesthetic with favorable pharmacokinetics, including rapid recovery and hemodynamic stability [10,11]. These features may be particularly advantageous during thoracic surgery, where fluctuations in hemodynamics can secondarily affect oxygenation. Given its similarity to propofol in terms of intravenous administration and potential preservation of HPV, remimazolam may have comparable effects on oxygenation during OLV. However, no prospective clinical studies have directly compared the effects of remimazolam and sevoflurane on arterial oxygenation during thoracic surgery.
Therefore, this study compared the effects of remimazolam and sevoflurane on arterial oxygenation during OLV in patients undergoing thoracoscopic surgery. The primary outcome was the arterial partial pressure of oxygen (PaO2) 30 min after OLV initiation. The secondary outcomes included time-dependent changes in PaO2 and the associated hemodynamic variables during the first 60 min of OLV.
Materials and Methods
Patients
This prospective, randomized, controlled study was approved by the Institutional Review Board of Pusan National University Yangsan Hospital (approval number: 05-2023-133, 14/06/2023). Written informed consent was obtained from all patients in the study. The study was registered before patient enrollment at ClinicalTrials.gov (NCT05907525) and conducted in accordance with the principles of the Helsinki Declaration of 1975 (revised 2013). This manuscript adheres to the Consolidated Standards of Reporting Trials (CONSORT) reporting guidelines for randomized clinical trials.
In this study, we enrolled patients aged 19–79 years with American Society of Anesthesiologists physical status classifications I to III scheduled for thoracoscopic surgery under general anesthesia between July 2023 and January 2024. Both conventional video-assisted thoracoscopic surgery (VATS) and robot-assisted thoracoscopic surgery (RATS) were included. The following exclusion criteria were applied: known allergy to benzodiazepines or propofol, patients with galactose intolerance, Lapp lactase deficiency or glucose-galactose malabsorption, hypersensitivity to dextran 40, end-stage renal disease requiring hemodialysis, history of acute angle glaucoma, emergency surgery, unstable hemodynamics requiring inotropic support, preoperative oxygen supplementation, preoperative forced expiratory volume in 1 s < 40% of predicted, and preoperative ejection fraction < 50%.
Eligible patients were randomly allocated to the sevoflurane or remimazolam groups using a computer-generated randomization sequence with a 1:1 ratio and a block size of two. Randomization was performed using sealed opaque envelopes. On the day of surgery, an investigator (A), who was not involved in patient care, opened the sealed envelope to reveal the group assignment. The attending anesthesiologists were not blinded to the group assignment; however, they did not participate in the patient allocation or data analysis. Another investigator (B), who was responsible for the data analysis, remained blinded to the group assignments and collected the data by reviewing blinded study logs.
Anesthesia management
In the remimazolam group, anesthesia was induced with remimazolamTM (Byfavo, Hana Pharm) at an infusion rate of 6 mg/kg/h and maintained at a rate of 1–2 mg/kg/h. In the sevoflurane group, anesthesia was induced with a bolus injection of 1% propofol (1.5–2.5 mg/kg) and maintained using sevoflurane at a concentration of 1−2 MAC. Remifentanil was administered to both groups using a target-controlled infusion pump based on the Minto pharmacokinetic model and titrated to maintain an effect-site concentration of 1.0–4.0 ng/ml. Bispectral index (BIS) monitoring was used to maintain the anesthetic depth within a target range of 40–60 under general anesthesia. Following loss of consciousness, 0.8 mg/kg of rocuronium was administered, and a double-lumen endotracheal tube was inserted. Radial artery cannulation was performed to monitor continuous arterial blood pressure. Central venous catheterization was performed in the internal jugular or subclavian vein under ultrasound guidance. After the patients were placed in the lateral decubitus position, the correct placement of the double-lumen tube was reconfirmed via fiberoptic bronchoscopy before initiating OLV. Anesthetic agents were titrated to maintain intraoperative blood pressure within 20% of the baseline values, which was measured upon arrival in the operating room before induction of anesthesia. Vasopressors/inotropes (ephedrine, phenylephrine, or norepinephrine) were administered as appropriate if mean arterial pressure (MAP) dropped below 60 mmHg or more than 20% below baseline. Vasodilators (nicardipine or diltiazem) were administered when the MAP exceeded 100 mmHg or increased by more than 20% from baseline. All patients received intravenous patient-controlled analgesia with postoperative fentanyl.
Ventilation strategies
Patients underwent mechanical ventilation using a tidal volume of 8 ml/kg predicted body weight (PBW) with fraction of inspired oxygen (FiO2) of 0.3 during two-lung ventilation (TLV) and 5 ml/kg PBW with FiO2 of 0.8 during OLV. Positive end-expiratory pressure was maintained at 5 cmH2O. The respiratory rate was initially set at 12 breaths/min and subsequently adjusted to maintain end-tidal carbon dioxide (ETCO2) between 35–40 mmHg. Hypoxemia during OLV was defined as PaO2 < 60 mmHg or peripheral oxygen saturation (SpO2) < 90%. If SpO2 dropped below 90%, FiO2 was increased to 1.0. If SpO2 did not recover within 1 min of administration of 100% oxygen, an alveolar recruitment maneuver was performed, regardless of scheduled arterial blood gas analysis (ABGA) time points. In cases of persistent hypoxemia despite these interventions, TLV was temporarily resumed to restore adequate oxygenation.
Outcome measurements
ABGA was conducted using an Epoc® blood analysis system (Siemens Healthcare) at predefined time points: T0 (10 min after lateral decubitus positioning during TLV [TLV10]), T1 (15 min after initiation of OLV [OLV15]), T2 (30 min into OLV [OLV30]), and T3 (60 min into OLV [OLV60]). CO was continuously assessed via pulse wave analysis using the LiDCOTM Unity hemodynamic monitoring system (Masimo).
The primary outcome was the difference in PaO2 between the sevoflurane and remimazolam groups measured 30 min after the initiation of OLV. Secondary outcomes included changes in PaO2 from baseline (T0) to 60 min (T3) during OLV, expressed both at absolute differences (∆PaO2) and relative percentage changes (%∆PaO2); differences in the ratio of PaO2 to the fraction of inspired oxygen concentration between T0 and T3 (PaO2/FiO2 ratio); and additional ABGA parameters, such as arterial partial pressure of carbon dioxide (PaCO2), and pH. Hemodynamic variables, including MAP, heart rate (HR), serum lactate level, and CO, were recorded at each time point from T0 to T3. The total amount of remifentanil administered during the 60 min of OLV was also recorded.
Sample size estimation
According to a previous study by Cho et al. [12], mean PaO2 at 15 min after initiating OLV was 180 ± 75 mmHg in the desflurane group and 226 ± 84 mmHg in the propofol group. In a separate study, Pruszkowski et al. [8] estimated an effect size of approximately 0.73 based on a 5.3 kPa PaO2 difference and a 7.3 kPa standard deviation. Taking this as a methodological precedent, we conservatively assumed a large effect size (Cohen’s d = 0.8), which we considered both clinically meaningful and feasible for patient recruitment. Based on this assumption, a two-sided α level of 0.05, and a power of 80%, the required sample size was calculated to be 26 patients per group. Accounting for an anticipated dropout rate of 20%, a total sample size of 66 patients (33 per group) was planned. The sample size was calculated using G*Power 3.1.9.7. program.
Statistical analysis
Variables are reported as frequency and percentage for categorical data and as mean ± standard deviation or median (Q1, Q3) for numerical data. Group differences were tested using the chi-squared test or Fisher’s exact test for categorical variables and the independent t test or Mann–Whitney U test for numerical variables, as appropriate. The Shapiro–Wilk test was used to assess normality. Standardized mean differences (SMDs) were calculated to evaluate baseline balance between groups, with SMD values of 0.2, 0.5, and 0.8 conventionally interpreted as small, medium, and large effect sizes, respectively, based on Cohen’s criteria [13].
A generalized linear mixed model (GLMM) with random intercepts was used to fit the model. The GLMM model included repeated measures of numerical variables as dependent variables; group, time, and group × time interaction as fixed effects; and the subject as a random effect. The statistical significance of group × time interactions is represented as interaction P values. To avoid making any assumptions regarding the covariance structure, we used an unstructured covariance matrix that was allowed to differ across groups for the GLMM analysis. Bonferroni’s post hoc test was used for multiple comparisons between each of the four time points. All statistical analyses were performed using IBM SPSS Statistics for Windows, version 26.0 (IBM Corp.). Statistical significance was set at P < 0.05.
Results
Patient characteristics
A total of 66 patients were recruited and screened for eligibility. One patient was excluded due to age > 80 years, exceeding the upper age limit in the inclusion criteria. Therefore, 65 patients were randomized to receive either sevoflurane (n = 33) or remimazolam (n = 32). Of these, seven patients dropped out due to the following reasons: OLV duration was less than 1 hour in three patients, improper positioning of the double-lumen tube in one patient, a measurement error in one patient, sustained oxygen desaturation below 90% in one patient, and one patient underwent collaborative abdominal surgery prior to thoracic surgery. Finally, 30 and 28 patients in the sevoflurane and remimazolam group, respectively, were included in this study (Fig. 1).
The demographic and perioperative characteristics are shown in Table 1, and the surgical details are presented in Table 2. There were no significant differences between the groups in terms of demographic data, diagnosis, preoperative hemoglobin level, CO, spirometry results, or intraoperative inotropic support. The mean end-tidal concentration of sevoflurane during OLV was approximately 1.8 ± 0.2 vol%, corresponding to approximately 1.0 MAC under standard clinical practice. The total amount of remifentanil administered during OLV was higher in the remimazolam group than in the sevoflurane group (377.7 ± 171.3 vs. 615.3 ± 231.1 μg, 95% CI [131.1–344.0], P < 0.001; SMD = 1.17).
Primary and secondary outcomes
PaO2 at OLV30 was not significantly different between the sevoflurane and remimazolam groups (108.9 ± 37.9 vs. 107.0 ± 37.8 mmHg, respectively; 95% CI [−21.8 to 18.0], P = 0.815 using the Mann–Whitney U test) (Fig. 2A). In both groups, PaO2 initially increased at OLV15, reached its lowest point at OLV30, and subsequently recovered at OLV60 (TLV 10: 118.2 ± 37.4 vs. 102.9 ± 21.0 mmHg, OLV15: 126.5 ± 54.4 vs. 111.3 ± 35.0 mmHg, OLV30: 108.9 ± 37.9 vs. 107.0 ± 37.8 mmHg, OLV60: 120.0 ± 42.9 vs. 121.4 ± 49.7 mmHg in sevoflurane vs. remimazolam group, respectively; interaction P = 0.078) (Fig. 2A). In the remimazolam group, post hoc analysis revealed significant increases at OLV60 compared to TLV10 in PaO2 and ΔPaO2 (95% CI [0.3–36.8]; Bonferroni’s corrected P = 0.044), as was %ΔPaO2 (95% CI [3.2–40.0]; Bonferroni’s corrected P = 0.013) (Figs. 2A–C). Conversely, the sevoflurane group showed no significant changes in PaO2, ΔPaO2, or %ΔPaO2 across the time points.
Hemodynamic parameters, including HR, MAP, and CO, changed significantly over time (time effect; all P < 0.05); however, no significant group × time interactions were observed (interaction P = 0.583, MAP P = 0.257, CO P = 0.281), indicating similar hemodynamic stability between the groups throughout surgery (Figs. 3A–C). The BIS values remained within the target range of 40−60 during surgery in both groups. Although BIS was consistently higher in the remimazolam group at OLV15, OLV30, and OLV60, no significant group × time interaction was observed (Table 3). Regarding ventilation and gas exchange variables, significant time-dependent increases were observed in ETCO2 and PaCO2 with corresponding decreases in pH (all time effects, P < 0.001) (Table 3 and Supplementary Table 1). However, no significant group × time interactions were detected for these variables (interaction P values; ETCO2 P = 0.763, PaCO2 P = 0.077, and pH, P = 0.183) (Table 3 and Supplementary Table S1). The PaO2/FiO2 ratio showed significant changes over time (P < 0.001), demonstrating a nadir at OLV30, followed by recovery at OLV60; however, the differences between groups were not statistically significant (interaction P = 0.158) (Table 3). Serum lactate levels decreased more in the remimazolam group than in the sevoflurane group during 60 min of OLV, with a significant group × time interaction (P = 0.021) (Fig. 3D). Six patients (three from each group) experienced hypoxemia during OLV, necessitating FiO2 elevation to 1.0 and an alveolar recruitment maneuver to the ventilated lung. Hypoxemia was successfully resolved in all patients without requiring conversion to TLV, allowing for continuous OLV throughout the surgery. In one patient in the remimazolam group, SpO2 decreased to 88% shortly after intubation but increased to 93% following an alveolar recruitment maneuver. During OLV, SpO2 remained above 96% at OLV30 with FiO2 maintained at 0.8. However, a second desaturation episode occurred at 55 minutes into OLV, prompting additional recruitment maneuvers. At OLV60, with FiO2 increased to 1.0, SpO2 was 90%, and PaO2 was recorded as 60 mmHg.
Discussion
This study comparing sevoflurane and remimazolam during thoracoscopic surgery with OLV found no significant difference in PaO2 at 30 min after OLV initiation, although a trend towards a time-by-group interaction was noted (P = 0.078). Notably, serum lactate levels were slightly but significantly lower in the remimazolam group during the OLV60.
The impact of intravenous and inhalational anesthetics on oxygenation during OLV has been extensively investigated, with earlier work focusing on propofol versus volatile agents. Volatile anesthetics attenuate HPV dose-dependently by activating potassium channels and inhibiting calcium influx in pulmonary artery smooth muscle cells [14]. In contrast, propofol appears to preserve or enhance HPV by not interfering with pulmonary oxygen-sensing mechanisms, making it potentially more favorable for oxygenation during OLV [4,15]. However, large trials have shown no significant oxygenation differences at clinically relevant concentrations (≤ 1 MAC), even accounting for the approximately 5% higher shunt fraction associated with sevoflurane [16,17]. Our findings are consistent with these reports, as the sevoflurane concentration was maintained at around 1 MAC, and no significant intergroup difference in oxygenation was observed.
Remimazolam, an intravenous anesthetic agent, has consistently demonstrated more stable hemodynamics than sevoflurane or propofol in various surgeries, including thoracic cases [11,18–21]. While our study did not include a propofol group, we observed comparable PaO2 levels and hemodynamic stability between the remimazolam and sevoflurane groups, which support the pharmacological similarity between remimazolam and established intravenous agents. Future research should directly compare remimazolam and propofol to elucidate their effects on oxygenation during OLV.
Both groups exhibited an initial increase in oxygenation at OLV15 compared to the TLV baseline, primarily due to the marked increase in FiO2 (from 0.3 to 1.0). However, the PaO2/FiO2 ratio showed an expected decrease at OLV15 compared to that during TLV, reflecting the expected development of a pulmonary shunt during OLV. At OLV30, the remimazolam group maintained PaO2 levels above baseline, while the sevoflurane group exhibited a transient decrease below baseline. By OLV60, PaO2 in the remimazolam group exceeded baseline levels, whereas the sevoflurane group showed minimal recovery. Although the mechanisms underlying these intra-group trends remain unclear, one possible explanation is a more rapid or sustained activation of HPV in the remimazolam group during the early phase of OLV. HPV is known to occur in two phases: an acute phase initiated within 2–5 min of hypoxia that stabilizes for 20–30 min and a sustained second phase starting after 30–40 min and persisting for several hours [2,14,22,23]. These time-dependent trends raise the hypothesis that remimazolam may facilitate earlier restoration of HPV. This hypothesis is indirectly supported by a study in elderly patients undergoing pulmonary lobectomy, where remimazolam, compared to propofol, significantly improved PaO2 and reduced intrapulmonary shunt fraction during OLV [20]. However, both groups in our study maintained a PaO2 > 100 mmHg, suggesting broadly preserved pulmonary gas exchange. While remimazolam may subtly enhance systemic oxygen transport, such interpretations should be approached with caution, since this study did not directly measure shunt fraction or pulmonary perfusion markers. Further studies should confirm this hypothesis using shunt fraction and mixed venous oxygen saturation (SvO2).
Anesthetic effects on oxygenation are influenced by hemodynamics and oxygen balance [9,24]. CO plays an important role in exerting dual effects on PaO2 [8]. A decrease in CO can lower SvO2, thereby reducing the oxygen content of the shunted blood and worsening hypoxemia. Conversely, reduced CO may decrease pulmonary blood flow to the non-ventilated lung, reducing the shunt fraction and partially compensating for hypoxemia [3,25]. CO typically increases during OLV following surgical incision, which helps maintain the balance between oxygen consumption and delivery [24,26–28]. In our study, both groups exhibited a gradual increase in CO, and minor fluctuations (approximately 0.6 L/min, approximately 10%–15% of normal adult CO) were unlikely to be clinically relevant for oxygenation [3]. These findings suggest that hemodynamic stability, rather than direct anesthetic effects on HPV may play a more dominant role in maintaining oxygenation during OLV.
Other variables that may indirectly affect the HPV response during OLV, such as acid-base status and temperature, were well controlled and did not differ between groups, suggesting minimal influence on oxygenation. The higher remifentanil dose in the remimazolam group was probably due to hemodynamic management, as elevated blood pressure required titration. However, previous studies have shown that remifentanil dose variation does not significantly affect PaO2 during OLV [12,29]. Given the comparable BIS values and stable hemodynamics, remifentanil dosage differences were unlikely to impact arterial oxygenation.
Serum lactate levels decreased over time during OLV, with a significantly greater reduction observed in the remimazolam group despite similar PaO2 levels between groups. Although lactate is a marker of tissue hypoxia, it can also rise under normoxic conditions due to inflammatory or catecholamine-driven metabolic stress [30,31]. Remimazolam has been reported to modulate inflammation by downregulating cytokine production and preserving antioxidant enzyme activity [32–34]. One possible explanation for this effect is that the observed lactate reduction reflects attenuated inflammatory and metabolic responses. Higher remifentanil use and more stable hemodynamics in the remimazolam group might also have contributed to this difference by reducing catecholamine-mediated stress. However, in the absence of biomarker data, these interpretations remain speculative and should be viewed with caution. Further studies should investigate inflammatory and oxidative stress markers to clarify the clinical relevance of this difference under OLV conditions.
This study has some limitations. First, as the relatively small sample size might lead to a type II error and limit the generalizability of our findings, our results should be interpreted with caution. However, the observed difference in PaO2 at the primary time point (OLV30) was minimal (Cohen’s d = 0.05), suggesting that the lack of significance likely reflects a true absence of meaningful intergroup difference rather than underpowering. Second, the need for recruitment maneuvers during the study period might have influenced the outcomes. Transient atelectasis after bronchoscopy in the lateral position might have contributed to ventilation–perfusion mismatch, yet no recruitment maneuvers were applied immediately thereafter. Meanwhile, intraoperative recruitment maneuvers were performed based on clinical protocols for patient safety and might have affected PaO2 measurements. Future studies should standardize the timing and frequency of recruitment maneuvers to reduce confounding and improve the reliability of oxygenation outcomes. Third, pulmonary artery catheterization was not used to measure CO or SvO2, although LiDCO monitoring, which correlates well with catheter-derived values, was employed [35,36]. The absence of SvO2 data limited shunt physiology analysis. Fourth, differences in insufflation and incision timing between RATS and VATS cases could not be accounted for due to small subgroup sizes. Nevertheless, physiological variables were comparable, supporting internal validity. Despite these limitations, this study contributes to the broader understanding of whether the effects of remimazolam on pulmonary oxygenation during thoracic surgery differ from that of volatile anesthetics, offering a basis for further characterization of its clinical role.
In conclusion, remimazolam provided oxygenation and hemodynamics comparable to those of sevoflurane during OLV. While a reduction in serum lactate was observed in the remimazolam group, its underlying mechanism remains unclear and warrants further investigation using inflammatory and metabolic biomarkers.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement
Print
