FiO2 requirements during general anesthesia with dual-lung a prospective pilot study

Authors: Jie-feng Sun, Jia-bao Chen, Wei-long Wang, Hong-fa Wang, Jin-tao Liu, Mei Cheng, Zhen-feng Zhou

Abstract

Background

Hyperoxemia and prolonged oxygen exposure were common during general anesthesia. However, the relationship between FiO2 and PaO2 in patients undergoing general anesthesia with dual lung ventilation remained unclear. This prospective pilot study aimed to explore this relationship.

Methods

A cohort of 50 patients was recruited for this self-controlled, prospective pilot study. A standardized volume-controlled ventilation strategy was applied, with FiO2 initially set to 0.3 immediately after tracheal intubation. FiO2 was then increased in steps of 0.1 until it reached 0.6, followed by an increase to 0.8. Each FiO2 step was maintained for at least 30 min before blood samples were drawn for blood gas analysis at each point.

Results

Rmcorr analyses revealed a significant correlation between FiO2 and PaO2 (p < 0.001). The correlation coefficient (r) was 0.967 and Model convergence was robust, with a gradient value of 0.00003. At an FiO2 of 40, 78.0% of patients maintained PaO2 between 100 and 200 mmHg, while more than 68.0% exceeded 200 mmHg at 50% FiO2 levels. Rmcorr analysis revealed a weak but statistically significant correlation between FiO2 and PaO2/FiO2 (r = 0.290; p < 0.001).

Conclusion

A significant linear correlation was identified between FiO2 and PaO2 during general anesthesia with dual-lung ventilation. In this prospective pilot study, our findings suggested that maintaining 40% FiO2 was generally sufficient for most patients to achieve PaO2 levels of 100–200 mmHg, unless specific clinical conditions require otherwise. Importantly, FiO2 need not exceed 50% in most patients undergoing general anesthesia with dual-lung ventilation.

Clinical trial registration

http://www.chictr.org.cn, identifier ChiCTR2500095383.

Introduction

Pulmonary atelectasis occurred in over 90% of patients undergoing general anesthesia, with high inspiratory oxygen fraction (FiO2) identified as a major contributing factor (1). Recent randomized clinical trial (2) and retrospective study (1) have highlighted the association between elevated intraoperative FiO2 and an increased risk of major respiratory complications and mortality, often in a dose-dependent manner. These findings suggested that intraoperative FiO2 should be carefully adjusted to just above the level required to maintain adequate arterial oxygen saturation (2).

The British Thoracic Society guidelines (3) recommend titrating FiO2 to achieve normal PaO2 levels, as supraphysiological oxygen concentrations were commonly observed in patients unless a clear clinical benefit from high FiO2 was evident. Further studies suggested that intraoperative PaO2 should be maintained within the range of 100–200 mmHg (4). However, oxygen saturation (SpO2) monitoring, the most commonly used method in clinical practice, provided limited insight into arterial blood oxygen levels (PaO2), particularly when SpO2 was ≥98% during general anesthesia (4). Additionally, arterial blood gas (ABG) analysis, though informative, was invasive and not routinely required for all patients.

Several theoretical models (5–10) have been proposed to predict the relationship between FiO2 and PaO2. However, most of these models have not been widely accepted in clinical practice and are often tailored to ICU patients receiving various modes of ventilatory support.

A multicenter, cross-sectional observational study (11) also found that hyperoxemia and excessive oxygen exposure are common during general anesthesia. Despite this, the precise relationship between FiO2 and PaO2 in patients undergoing general anesthesia with dual-lung ventilation remained unclear. Therefore, we conducted this prospective pilot study to further investigate this relationship.

Methods

Study design

This prospective pilot study was approved by the Ethics Committee of Zhejiang Provincial People’s Hospital and conducted between January 07 and February 18, 2025 at Zhejiang Provincial People’s Hospital, China.

Eligible patients were aged 18 years or older, with an American Society of Anesthesiologists (ASA) physical status of I-III, scheduled for elective surgery under general anesthesia expected to last longer than 3 h. All patients required radial artery puncture, central venous catheterization, and had preoperative PaO2 > 60 mmHg.

Exclusion criteria included emergency surgery, prior lung surgery, need for non-invasive oxygen therapy, chronic obstructive pulmonary disease (COPD), acute respiratory failure (including acute lung injury, pneumonia, or acute respiratory distress syndrome), severe cardiac disease or persistent hemodynamic instability, requirement for renal replacement therapy (CRRT), sepsis or septic shock, progressive neuromuscular disorders, pregnancy, participation in another clinical trial, or refusal to participate.

Anesthesia and monitoring

Prior to anesthesia induction, an internal jugular vein catheter and radial artery catheter were inserted. Standard anesthesia monitoring included SpO2, ECG, arterial blood pressure, heart rate (HR), end-tidal carbon dioxide (EtCO2), and bispectral index (BIS), all recorded using a Datex Ohmeda S/5 Avance monitor (GE Healthcare, Helsinki, Finland).

Crystalloid solution (12–15 mL/kg/h) and vasoactive drugs were infused to maintain the mean arterial pressure (MAP) at 80%–90% of the patient’s baseline blood pressure, defined as the blood pressure measured in a calm state before surgery. Blood loss and anesthesia-related vasodilation were compensated with colloid infusion as needed.

Routine anesthesia induction was performed using intravenous dexmedetomidine (1 ug/kg) or midazolam (0.05–0.075 mg/kg), cisatracurium (0.2 mg/kg), propofol (2–3 mg/kg), and sufentanil (0.1–0.3 μg/kg) to facilitate tracheal intubation. Anesthesia was maintained with propofol, sevoflurane, and a continuous infusion of remifentanil to achieve a bispectral index (BIS) value of 40–50 until completion of skin suturing. Cisatracurium (0.1 mg/kg) was administered hourly, with the final dose given at least 1 h before the end of surgery. Sufentanil (0.1 μg/kg) and flurbiprofen axetil (50 mg) were administered before discontinuation of remifentanil.

Blinding

An anesthesiologist, adhering to the study protocol, administered anesthesia and was responsible for blood sampling and testing during the preoperative, intraoperative, and post-anesthesia care unit (PACU) periods. Although aware of the intraoperative ventilation strategy, this anesthesiologist did not participate in data collection or postoperative follow-up. Data collection was performed by full-time follow-up staff, and statistical analysis was conducted by a statistician who was blinded to the data collection process.

Intraoperative respiratory parameter settings and adjustment

A volume-controlled ventilation strategy was used, with a tidal volume set to 8 mL/kg of ideal body weight (IBW) and an inspiratory-to-expiratory ratio of 2. IBW was calculated using the following formulas (12): 45.5 + 0.91× (centimeters of height - 152.4) for females and 50 + 0.91× (centimeters of height - 152.4) for males. Mixed air flow was restricted to 2–3 L/min, and the respiratory rate was adjusted to maintain EtCO2 between 35 and 45 mmHg. Positive end expiratory pressure (PEEP) was not applied, however, lung recruitment maneuvers (RMs) (13, 14) were employed to handle potential atelectasis immediately after tracheal intubation and every time when the ventilator was interrupted during the study.

FiO2 settings

Immediately after tracheal intubation, FiO2 was set to 0.3. It was then incrementally increased by 0.1 steps until it reached 0.6, followed by a final increase to 0.8. Each FiO2 setting was maintained for at least 30 min before blood samples were drawn. Following this, FiO2 was adjusted as necessary to maintain PaO2 within the range of 100–200 mmHg. The final FiO2 value was recorded as dispayed on the anesthesia machine monitor.

Blood gas analysis (BGA)

Blood samples were collected via the radial artery catheter into 1-mL heparinized syringes and immediately analyzed using a Siemens Rapidpoint 405 Co-oximeter (Siemens, Munich, Germany) at a satellite laboratory located in the operating room. The analyzer was calibrated automatically every 6 h. To ensure accuracy, blood samples were carefully deaerated, mixed, and analyzed promptly after collection. The primary outcome was PaO2, with additional data on the PaO2/FiO2 ratio, SpO2, hemoglobin (Hb), and other relevant parameters collected by an independent researcher.

Sample size calculation

Based on previous similar studies (4, 15), we estimated that 50 patients would be sufficient for this prospective self-controlled pilot study. Although only 50 patients were enrolled, each participant underwent five interventions and outcomes, which was effectively equivalent to including 250 patients.

Statistical analysis

Statistical analysis was conducted using R software (version 4.4.0). Normal variables were expressed as mean ± standard deviation (SD), while non-normally distributed variables were presented as median (interquartile range). Multivariable restricted cubic splines (RCS) were used to assess the linear relationship between continuous variables. Based on the characteristics of repeated measurement data, the repeated-measures correlation (rmcorr) analysis was first used to analyze the correlation between FiO2 and PaO2, calculated the repeated measures correlation coefficient (r), and evaluated the strength of the correlation between the two continuous variables based on the magnitude of the r value. To further quantified the impact of FiO2 on PaO2, linear mixed effects models were constructed. The model taken PO2 as the dependent variable, FiO2, as a fixed effect, the random intercept of patients and the random slope of FiO2 [formula: PO2, FiO2~, + (1 + FiO2, ID)] were included in the random effects section to explain individual differences and heterogeneous responses among patients. To controled the influence of potential confounding factors, clinical related covariates such as age, gender, PCO2, PEEP, temperature, PH, HCT, and laparoscopic surgery or not were gradually introduced on the basis of the basic model to improve model accuracy and reduce residual variation. The model fitting adopted the Restricted Maximum Likelihood (REML) method. The convergence of the model was evaluated by gradient values, and the convergence criterion was set to a maximum gradient absolute value less than 0.002. The statistical significance was set to bilateral a = 0.05. Continuous variables were expressed as mean ± standard deviation or median (interquartile range), while categorical variables were expressed as frequency (percentage).

Results

Baseline clinical parameters

A total of 73 patients were initially recruited for the study, with the final analysis including 50 patients, as shown in Figure 1. The mean age of the participants was 62 years, and 56% were male. The majority of patients (70%) were classified as ASA physical status II. Additionally, 50% of patients underwent laparoscopic surgery. The duration of surgery and mechanical ventilation are detailed in Table 1.

Relationship between FiO2 and PaO2

Rmcorr analyses revealed a significant correlation between FiO2 and PaO2 (p < 0.001) (Table 2). The correlation coefficient (r) was 0.967, indicating a predictive accuracy exceeding 80%. The relationship between FiO2 and PaO2 was described by the linear mixed-effects Y = −18.4 + 4.8**X* (where Y = PaO2 and X = FiO2) (Figure 2 and Table 2). An FiO2 range of 25%–46% was found to maintain PaO2 within the 100–200 mmHg range. Model convergence was robust, with a gradient value of 0.00003. The fixed effects of model covariates were summarized in Supplementary Table S1.

At an FiO2 of 40, 78.0% of patients maintained PaO2 between 100 and 200 mmHg. In contrast, 82.2% of patients had PaO2 values below 100 mmHg at 21% FiO2, while more than 68.0% exceeded 200 mmHg at 50% FiO2 levels (Table 3, Figure 3A).

Comparison of laparoscopic and non-laparoscopic surgery

In patients undergoing laparoscopic surgery, the linear mixed-effects model describing the relationship between FiO2 and PaO2 Y = −3.9 + 4.4X.

An FiO2 range of 24%–46% maintained PaO2 within 100–200 mmHg.

For non-laparoscopic surgery, the relationship Y = −32.4 + 5.2X.

An FiO2 range of 25%–45% achieved similar PaO2 control (Table 2). Both models demonstrated excellent convergence (gradient = 0.00003).

Relationship between FiO2 and PaO2/FiO2

Rmcorr analysis revealed a weak but statistically significant correlation between FiO2 and PaO2/FiO2 (r = 0.290; 95% CI: 0.155–0.414; df = 191; p < 0.001; Figure 4A). Adjusted restricted cubic spline (RCS) analysis showed no significant non-linear association between FiO2 and PaO2/FiO2 (P for non-linearity = 0.255; Figure 4B).

A PaO2/FiO2 ratio below 300 mmHg—indicative of pulmonary gas exchange dysfunction—was observed in 12.2% of patients at 30% FiO2, 6.0% at 50% FiO2, and only 2.0% at 60% FiO2. However, the proportion of patients with PaO2/FiO2 within the normal range (400–500 mmHg) did not significantly increase with rising FiO2 (p > 0.05), remaining between 46.0 and 59.1% across the 40%–80% FiO2 range. When FiO2 was ≥50%, more than 26.0% of patients exhibited PaO2/FiO2 values exceeding 500 mmHg (Table 3, Figure 3B).

Blood gas analysis (BGA)

Detailed results from the blood gas analysis were presented in Supplementary Table S2. Significant differences in PaO2 were observed across the various FiO2 groups.

Discussion

We observed a significant linear correlation between FiO2 and PaO2 during general anesthesia with dual-lung ventilation. We observed that at an FiO2 of 40, 78.0% of patients maintained PaO2 between 100 and 200 mmHg, while more than 68.0% exceeded 200 mmHg at 50% FiO2 levels. Our findings suggested that maintaining 40% FiO2 was generally sufficient for most patients to achieve PaO2 levels of 100–200 mmHg, unless specific clinical conditions require otherwise. Importantly, FiO2 need not exceed 50% in most patients undergoing general anesthesia with dual-lung ventilation.

The normal PaO2 for healthy adults typically ranged from 80 to 100 mmHg and served as a critical indicator for detecting tissue and organ hypoxia during surgery. According to the oxygen dissociation curve, when SpO2 reached 99%–100%, PaO2 was approximately 160 mmHg. The oxygenation index for normal human lungs was 400–500 (PaO2/FiO2), which implied that a theoretical FiO2 of 32%–40% would be needed to achieve a PaO2 of approximately 160 mmHg. Our findings further suggested that maintaining FiO2 between 25% and 46% was effective in sustaining PaO2 within the 100–200 mmHg range.

A multicenter cross-sectional study (11) highlighted that hyperoxemia and excessive oxygen exposure were prevalent during general anesthesia. The WHO and CDC have recently recommended an FiO2 of 80% during and immediately after surgery to reduce the risk of surgical site infections (SSIs) (16). However, concerns have been raised about the potential adverse effects of high FiO2, and some guidelines now recommend maintaining FiO2 ≤ 0.4 (17). Miller’s Anesthesiology (18) notes that anesthesia, whether delivered with spontaneous or mechanical ventilation, typically impaired lung function. A conventional practice has been to maintain FiO2 between 0.3 and 0.4 during mechanical ventilation, as higher concentrations might contribute to atelectasis.

Our study supported this approach, showing that maintaining FiO2 within a moderate range prevents excessive oxygen exposure and optimizes oxygenation. Our analysis revealed a significant linear correlation between FiO2 and PaO2, with R values of 0.967 in rmcorr analyses. Several theoretical models (5–8) have attempted to predict the relationship between FiO2 and PaO2, but most of these models were designed for ICU patients receiving various ventilatory modes (e.g., volume-controlled, pressure-controlled ventilation). These models might not be directly applicable to intraoperative settings, where different ventilation strategies could influence ventilation/perfusion matching, cardiac output, and venous oxygen saturation (SvO2), which could impact the accuracy of the predictions. Additionally, models such as those by Al-Otaibi and Hardman focus on narrower FiO2 ranges (40%–60%) (9, 10), and no studies have systematically explored a conservative approach to intraoperative oxygen therapy. Our clinical findings provided valuable data for this context, confirming the linear relationship between FiO2 and PaO2 during general anesthesia. We also noted despite changes in FiO2, SpO2 remained high and clustered near saturation, highlighting the limitation of SpO2 monitoring but also suggesting that reliance on SpO2 alone could mask hyperoxemia.

In this study, we observed the weak correlation between FiO2 and PaO2/FiO2 in both repeated-measures and restricted cubic spline analyses. First, the PaO2/FiO2 ratio remained more informative for evaluating pulmonary gas exchange, particularly in patients with ventilation–perfusion (V/Q) mismatch or diffusion impairment. In such cases, PaO2 might not rise in proportion to FiO2, indicating diffusion limitation or alveolar shunting rather than insufficient oxygen delivery. Secondly, the absence of a linear increase in PaO2/FiO2 at higher FiO2 levels likely reflected physiological saturation effects, including the plateau of the oxygen–hemoglobin dissociation curve. Similar findings have been reported in both perioperative and critical care studies, which demonstrate that increasing FiO2 often fails to improve PaO2 proportionally and may even exacerbate V/Q mismatch (19, 20).

Finally, prior research has shown that a period of 5–10 min was typically sufficient for PaO2 to equilibrate after a change in FiO2 in ICU patients (21). In this study, each FiO2 level was maintained for at least 30 min, providing ample time for PaO2 to stabilize following changes in FiO2.

A recent large observational cohort study involving 350,647 patients demonstrated that exposure to supraphysiological oxygen concentrations during surgery was associated with an increased incidence of kidney, myocardial, and pulmonary injury (22). Numerous studies have also linked high intraoperative FiO2 with increased risk of both cardiac (23) and pulmonary complications (24), while others have found no such association (16, 25). Even routine preoxygenation with 100% oxygen during anesthesia induction could lead to atelectasis (26), prompting recommendations to decrease FiO2 after induction to mitigate further atelectasis formation (27). High perioperative FiO2 might mask deteriorating oxygenation if it exceeded the levels required to maintain adequate SpO2 (28). Several retrospective studies have also suggested a dose-dependent relationship between intraoperative FiO2 and postoperative respiratory complications or 30-day mortality (1, 29). However, the present study did not investigate the impact of FiO2 on clinical outcomes, as this was beyond the scope of the current investigation.

The study focuses on the FiO2–PaO2 correlation but did not evaluate patient-centered outcomes (e.g., postoperative atelectasis, SSI, or respiratory mortality), thus providing limited guidance for clinical oxygen therapy decisions. However, recent randomized clinical trial (2) and retrospective study (1) have highlighted the association between elevated intraoperative FiO2 and an increased risk of major respiratory complications and mortality, often in a dose-dependent manner. Thus this prospective pilot study aimed to explore physiological correlation between FiO2 and PaO2 in patients undergoing general anesthesia with dual lung ventilation by volume-controlled ventilation strategy.

Clinically, these results highlighted the need for individualized oxygen therapy rather than the routine administration of high FiO2 during anesthesia and mechanical ventilation. Targeting a PaO2 between 100 and 200 mmHg might provide adequate oxygenation while minimizing hyperoxia-related injury. Future research should aim to develop real-time, physiology-based FiO2 titration strategies that integrate continuous gas exchange monitoring, dynamic compliance, and shunt estimation to optimize oxygen delivery during the perioperative period.

Limitations

This study has several limitations. First, it was conducted at a single center and included a relatively specific patient population (ASA physical status I–III, elective surgery, and dual-lung ventilation), which limits the generalizability of the findings to broader clinical settings, including emergency surgery or critically ill patients. Moreover, because patients with pulmonary disease were excluded, the results cannot be extrapolated to individuals with compromised lung function, who are often at increased risk of oxygenation abnormalities. Future studies should include such populations to determine whether the relationship between inspired oxygen fraction (FiO2) and arterial oxygen tension (PaO2) differs in patients with impaired pulmonary function. Notably, arterial blood gas (ABG) analysis is routinely performed in critically ill patients or those with pulmonary disease, which may facilitate more precise titration of FiO2 to achieve normoxemia.

Second, although adjustments were made for body temperature, hemoglobin concentration, and pH, other important physiological variables—such as cardiac output, lung compliance, and ventilation–perfusion mismatch—were not assessed and may have influenced PaO2 measurements. Nevertheless, this was a self-controlled study, and each patient served as their own control under different FiO2 conditions, thereby minimizing interindividual variability. Consequently, baseline clinical characteristics, including smoking history, dynamic lung compliance, intrapulmonary shunt fraction, and the intraoperative anesthesia settings, were remained consistent within each FiO2 conditions.

Thirdly, the results might differ with other ventilatory modes (pressure-controlled, spontaneous breathing), however, a volume-controlled ventilation strategy was the most commonly used mode. Fourth, in this study the FiO2 was altered stepwise in steps from 0.3 to 0.8. Therefore, the relationship between FiO2 and PaO2 could not be evaluated beyond this range. Finally, as a single-center pilot study with 50 patients, the external validity of the results may be limited. Despite enrolling only 50 participants, each underwent six blood gas evaluations, yielding a total of 300 data points, which provided adequate statistical power for constructing a reliable model. Multi-center studies with larger cohorts were needed to confirm the FiO2–PaO2 relationship across diverse patient populations and clinical settings, and add postoperative respiratory complications, duration of mechanical ventilation, or markers of lung injury to connect physiological findings with clinical relevance.

Conclusion

A significant linear correlation was identified between FiO2 and PaO2 during general anesthesia with dual-lung ventilation. In this prospective pilot study, our findings suggested that maintaining 40% FiO2 was generally sufficient for most patients to achieve PaO2 levels of 100–200 mmHg, unless specific clinical conditions require otherwise. Importantly, FiO2 need not exceed 50% in most patients undergoing general anesthesia with dual-lung ventilation.