This study was approved by the Institutional Review Board and registered at http://clinicaltrials.gov. This study was conducted according to the Good Clinical Practice guidelines and principles of the Declaration of Helsinki, as revised in 2013. Written informed consent was obtained from all the participants.

The inclusion criteria included age > 18 years, an American Society of Anesthesiologists physical status of I–III, and requiring OLV for elective lung resection surgery. Patients with previous lung resection, baseline partial pressure of oxygen (PaO2) < 70 mmHg, preoperative supplemental oxygen treatment, tracheostomy, or chronic obstructive pulmonary disease with forced expiratory volume in 1 s over a forced vital capacity < 70% were excluded. Each patient underwent preoperative arterial blood gas analysis (ABGA), pulmonary function tests, and chest radiography.

The patients were randomly allocated to three groups according to OLV strategies as follows: those ventilated at 4 mL/kg of predicted body weight (PBW) (n = 29, low VT [LV] group), those ventilated at 6 mL/kg of PBW (n = 29, medium VT [MV] group), and those ventilated at 8 mL/kg of PBW (n = 29, HV group). PEEP of 5 cmH2O was applied to all patients.

Randomization was conducted using a computer-generated program (http://www.randomization.com) at a 1:1:1 allocation ratio. An anesthetic nurse not involved in the study generated a random allocation sequence using sealed opaque envelopes. The patients, investigators, surgeons, and data collectors were blinded to the group allocation. Only the bedside anesthesiologists who regulated the ventilator and maintained intraoperative anesthesia were not blinded.

All the patients were monitored using 3-lead electrocardiography, noninvasive arterial blood pressure measurement, peripheral oxygen saturation (SpO2), bispectral index, and acceleromyography. General anesthesia was induced and maintained with intravenous propofol and remifentanil using an effect-site target-controlled infusion pump (Module DPS Orchestra®, Brezins, France). Propofol was adjusted to 3–5 µg/mL, while remifentanil was adjusted to 2–6 ng/mL according to the effect-site concentrations.

After confirming the patients’ loss of consciousness, a double-lumen tube (DLT) was intubated after injecting 0.6 mg/kg of rocuronium. When the bispectral index was < 60 and the train-of-four (TOF) count was 0, the patients were intubated with a DLT (Mallinckrodt endobronchial tube; Covidien, Mansfield, MA, USA) using direct or video laryngoscopy. After confirming the appropriate position of the DLT using a fiberoptic bronchoscope (FOB) (LF-GP; Olympus Optical Co., Tokyo, Japan), the tracheal and bronchial cuffs were inflated at a pressure of 20–25 cmH2O using a capnometer for lung isolation.

During the surgery, 0.2–0.3 mg/kg of rocuronium was intermittently administered for muscle relaxation. The anesthetic depth was adjusted to achieve a target bispectral index of 40–60. Lactated Ringer’s solution was infused at a rate of 3–5 mL/kg/h as maintenance fluid.

The radial artery was catheterized to monitor continuous arterial blood pressure and for intraoperative blood sampling. During arterial and central line catheterization, both lungs were ventilated at a VT of 10 mL/kg of PBW with a PEEP of 5 cmH2O and an inspiratory to expiratory ratio of 1:2 with a fresh gas flow of 2 L/min in FiO2 1.0 using a mechanical ventilator (Primus; Dräger, Lübeck, Germany). End-tidal carbon dioxide (EtCO2) or arterial tension of carbon dioxide (PaCO2) was maintained at 35–45 mmHg by adjusting the ventilatory respiratory rate (RR).

After the patient was placed in the lateral decubitus position and the final position of the DLT was confirmed using an FOB for the operation, OLV was initiated at a VT of 4, 6, or 8 mL/kg of PBW according to assigned group, while other ventilatory settings were maintained. FiO2 was downregulated to 0.8 after alveolar recruitment with an inspiratory pressure of 20 cmH2O for 15–20 s. During OLV, FiO2 was reduced stepwise from 0.8 to 0.5 if the patients did not show hypoxemia, which was defined as a decrease in the arterial tension of oxygen (PaO2) to < 80 mmHg or SpO2 to < 95%. Each FiO2 level was maintained for 5 min. In patients showing hypoxemia during OLV, FiO2 was increased stepwise to 1.0. If hypoxemia could not be corrected by increasing the FiO2 to 1.0, we considered endobronchial suctioning followed by alveolar recruitment, reassessing the DLT position with an FOB, applying continuous positive airway pressure at 2 cmH2O, and applying intermittent two-lung ventilation (TLV) as rescue interventions. The rescue interventions were recorded.

Additionally, peak inspiratory pressure (PIP) was monitored to ensure it did not exceed 35 cmH2O during mechanical ventilation. If the PIP was > 35 cmH2O during OLV, we provided endobronchial suctioning followed by alveolar recruitment and reassessed the DLT position.

After lung resection, TLV was resumed after gentle suctioning of both lungs, followed by alveolar recruitment to reexpand the operated lung. The same ventilatory setting as applied before OLV was implemented, except for an FiO2 of 0.4 until the surgery was completed. Patients who underwent open thoracotomy due to unexpected adhesion or received blood transfusion due to an estimated blood loss > 500 mL intraoperatively were also recorded.

After surgery completion, intravenous or epidural patient-controlled analgesia comprising fentanyl, morphine, and ramosetron were administered according to the surgical technique. After placing the patients in the supine position, oral and bronchial secretions were removed. The DLT was extubated after the administration of 2–4 mg/kg sugammadex and the patient showed recovery of adequate spontaneous breathing, reflex, and a TOF ratio of > 0.9.

Patients were transferred to postanesthetic care unit (PACU) or surgical intensive care unit (SICU) at the surgeons’ discretion with consideration of the surgical procedure, intraoperative surgical events, or requirement of close monitoring and appropriate postoperative care according to patient’s underlying disease. During their transfer, the patients wore a Venturi mask (MOW Medical, Wonju, Korea) at FiO2 of 0.4 to ensure consistent FiO2 delivery. Supplemental oxygen was changed stepwise by the surgeon to maintain adequate oxygen saturation; if oxygen supply was not required, the patients were ventilated on room air.

The primary outcome was the mean difference in PaO2/FiO2 ratio among the three groups after surgery. Postoperative ABGA was performed by an investigator immediately after the patients arrived at the PACU or SICU and 6 h after the surgery. The PaO2/FiO2 ratio was calculated in terms of PaO2 values and the supplied oxygen at the measurement time points in ABGA. Chest radiographs were obtained each morning for 3 postoperative days. A chest radiologist blinded to the study analyzed the radiographs to evaluate newly developed lung lesions, including lung infiltrations and atelectasis. The incidence of immediate PPCs, which were represented by acute lung injuries determined by a PaO2/FiO2 ratio of < 300 mmHg and/or radiological findings within 72 h after surgery, was also compared among the groups.

Secondary outcomes included hemodynamic variables; intraoperative ABGA; PaO2/FiO2; and ventilator settings, including VT, RR, FiO2, mean airway pressure (Pmean), airway plateau pressure (Pplat), and PIP recorded at five consecutive time points: TTLV_baseline, after anesthesia induction; TOLV_20min, 20 min after OLV; TOLV_40min, 40 min after OLV; TOLV_60min, 60 min after OLV; and TTLV_20min, 20 min after TLV resumption. PaO2/FiO2 was calculated according to PaO2 and applied FiO2 via a ventilator at the measurement time point.

Additional intraoperative treatment options for hypoxemia, rescue interventions, and respiratory and cardiovascular events were also recorded. Moreover, individual postoperative pain was assessed using a visual analog scale at 30 min after arriving in the PACU or SICU. Additionally, data regarding the length of stay in the PACU or SICU, hospitalization, readmission to the ICU due to PPCs, and mortality at 30 postoperative days were recorded. Data were assessed by an investigator who was not involved in the patient’s anesthetic care or postoperative management.

The sample size was calculated based on previously published data showing a difference in PaO2/FiO2 ratios measured 15 min after resuming TLV at the end of the surgery with the three ventilation strategies [17]. An a priori power analysis of the previous data indicated that 24 patients were required in each group, with an estimated effect size of 0.796 (α error of 0.05) and a power of 90%. Accounting for a dropout rate of 20%, the final sample size was 29 patients per group. Statistical analyses were performed using G Power software (version 3.1, University of Düsseldorf, Germany).

The R language version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria) and T&F program version 4.0 (YooJinBioSoft, Korea) were used for all analyses. Continuous variables are expressed as medians (interquartile ranges). The sample numbers and percentages were computed for categorical variables.

The mean differences among the three groups were analyzed using the Kruskal–Wallis H test. The chi-squared test with continuity correction or Fisher’s exact test was performed to analyze the proportion of sample numbers in the subgroups of categorical variables. The Bonferroni method was used for correction of multiple comparisons. P < 0.05 was considered statistically significant. All data were analyzed according to the intention-to-treat principle.