Associations between intraoperative ventilator settings during one-lung ventilation and postoperative pulmonary complications: a prospective observational study

A two-center prospective observational study was conducted from April 2014 to October 2014 in Japan. Participating hospitals included an academic tertiary care hospital and a community hospital. This study was approved by the institutional ethics review board (IRB) of Okayama University Hospital (No. 1922) and Fukuyama City Hospital (No. 182). The requirement for written informed consent was waived by each IRB. We screened consecutive patients over the age of 20 who were scheduled for a thoracic surgical procedure and required general anesthesia with OLV. We excluded emergency surgery, re-operative surgery, and patients who did not receive OLV. There was no specific protocol for perioperative management at the participating hospitals.

We investigated perioperative information, including preoperative characteristics, details of surgery and anesthesia, and postoperative course. Demographics and clinical data were extracted from electronic medical records. The preoperative data included sex, age, Assess Respiratory Risk in Surgical Patients in Catalonia (ARISCAT) score [17], preoperative respiratory function, and preoperative percutaneous oxygen saturation (SpO₂). We collected anesthetic and surgical information, such as surgical procedures, types of general anesthesia, use of epidural anesthesia, and airway management as well as duration of procedure, anesthesia, and OLV. Total blood loss and volume of infusion were also collected. Minimum SpO₂ throughout the course of anesthesia was recorded.

During OLV (0, 30, 60, and 120 min after the start of OLV and at the end of OLV), the following variables were recorded: ventilator mode, F_IO₂, V_T corrected for predicted body weight (PBW), driving pressure (ΔP) (peak inspiratory pressure minus PEEP on both pressure control and volume control ventilation), and PEEP. These data were collected by attending anesthesiologists. PBW was calculated as follows: for men, 50 + 0.91 (height (cm) - 152.4); and for women, 45.5 + 0.91 (height (cm) - 152.4) [18].

To avoid surveillance bias, time weighted average (TWA) of ventilation parameters was calculated for the first 2 h of OLV. TWA was determined by summing the mean value between consecutive time points (0, 30, 60, and 120 min after the start of OLV) multiplied by the period of time between consecutive time points and then divided by the total time. We calculated and assessed TWA of F_IO₂, V_T, ΔP, and PEEP during OLV.

The primary outcome was the incidence of PPCs occurring within 7 days of thoracotomy. PPCs included pneumonia, pleural effusion, atelectasis, prolonged air leakage, pulmonary embolism and respiratory failure diagnosed according to the definitions (Table 1), which referred to previous studies [17, 19, 20]. In each center, a predetermined researcher evaluated all patients in accordance with the definitions of PPCs. To investigate the length of hospital stay (LOS) and mortality, patients were followed-up until hospital discharge or death (whichever occurred first).

Variables were assessed for normality. Categorical data were compared using chi-square tests or Fisher exact tests and reported as n (%). Continuous normally distributed variables were compared using Student t tests and reported as means (standard deviation), while non-normally distributed data were compared using Wilcoxon rank-sum tests and reported as medians (interquartile range). Univariate analysis was performed to compare perioperative characteristics between patients with and without PPCs. A multivariate logistic regression analysis was performed to estimate the associations between intraoperative ventilator settings and PPCs, adjusting for ARISCAT score and all univariate relevant factors that discriminate between the two groups. To explore subgroup differences in associations between the ventilator settings and PPCs, the same multivariate analyses were performed for subgroups classified according to the ARISCAT score, preoperative SpO₂ and surgical procedures, respectively. All analyses were performed using JMP version 8.0.2 (SAS Institute, Cary, NC, USA). P < 0.05 was considered statistically significant. This manuscript adheres to the applicable Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.

Overall, 212 cases underwent thoracic surgery with OLV during the study period. Two patients were younger than 20 years old, and 13 cases underwent thoracic surgeries twice during the study period. Thus, 197 patients met the eligibility criteria (Fig. 1).

Baseline characteristics and intraoperative procedures of all patients are noted in Additional file 1. Most patients (n = 190, 96.4%) had an intermediate or high risk of having PPCs according to the ARISCAT score. More than 80% of patients underwent lung resections; however, there was no patient who underwent pneumonectomy.

Pressure control ventilation (PCV) was utilized in most cases (n = 181, 92%). At the start of OLV, median F_IO₂ was 1.0 (0.8-1.0). Specifically, an F_IO₂ of 1.0 was applied as an initial setting for more than 60% of all patients. In other initial settings, median V_T was 6.1 (5.2-7.3) ml/kg, and median ΔP was 16 (14-20) cm H₂O. PEEP was applied in 171 patients (87%) at a median level of 4 (4-5) cm H₂O. The distributions of ventilator settings throughout OLV are shown as TWA values in Fig. 2. Median TWA F_IO₂ was 0.8 (0.65-0.94), and 83% of patients received TWA F_IO₂ ≥ 0.6. Other median TWA values, such as V_T, ΔP, and PEEP, were at almost similar levels as the initial settings (V_T, 6.1 (5.3-7.0) ml/kg; ΔP, 17 (15-20) cm H₂O; and PEEP, 4 (4-5) cm H₂O). As a rescue therapy, oxygen therapy to the non-ventilated lung was adopted in only five cases.

PPCs occurred in 51 of 197 cases (25.9%). Atelectasis developed in 35 patients (17.8%), prolonged air leakage in 10 (5.1%), pneumonia in 3 (1.5%), pleural effusion in 3 (1.5%), and respiratory failure in 2 (1.0%). Two cases with respiratory failure occurred with atelectasis or pleural effusion. None of the patients were diagnosed with pulmonary embolism in this period. Only one patient died during hospital stay, and overall mortality was 0.5%. Baseline characteristics and intraoperative procedures in patients with and without PPCs were shown in Table 2. There were no significant differences in preoperative baseline characteristics, surgical procedures, and intraoperative management regarding anesthesia.

Among ventilator settings, only TWA F_IO₂ in patients with PPCs was significantly higher than that in patients without PPCs (0.85 (0.73-1.0) vs. 0.77 (0.63-0.89); P = 0.0032) (Table 3). There was no significant difference in TWA V_T, TWA ΔP, and TWA PEEP between the two groups. Throughout the anesthesia, minimum SpO₂ in patients with PPCs was significantly lower than that in patients without PPCs (94 (91-96) % vs. 95.5 (93-97) %; P = 0.0053). Finally, the postoperative LOS was longer in patients with PPCs (13 (8-16) days vs. 8 (7-11) days; P < 0.001).

In multivariate logistic regression model (Table 4), which was adjusted for ventilator settings (TWA F_IO₂, TWA ΔP, and TWA PEEP), ARISCAT score, and minimum SpO₂, only TWA F_IO₂ during OLV was independently associated with the occurrence of PPCs. Odds ratio (OR) per TWA F_IO₂ increase of 0.1 was 1.30 (95% confidence interval (CI): 1.04-1.65, P = 0.0195). Other variables (TWA ΔP, TWA PEEP, ARISCAT score, and minimum SpO₂) were not related to the occurrence of PPCs in this model.

There were significant associations between F_IO₂ and PPCs in patients with low or intermediate risk of having PPCs according to the ARISCAT score (OR, 1.48; 95% CI, 1.00-2.40; P = 0.0496), or undergoing lung resection (OR, 1.31; 95% CI, 1.03-1.70; P = 0.0278) (Additional file 2). Other subgroups including patients with high risk for PPCs and high or low preoperative SpO₂, also indicated that higher F_IO₂ tended to be associated with higher incidence of PPCs.

We conducted a prospective observational study to investigate the current practice of intraoperative ventilation and to evaluate the associations between ventilator settings during OLV and PPCs in patients undergoing thoracic surgery. We found that F_IO₂ of ≥0.8, V_T of approximately 6 ml/kg, and PEEP of approximately 4 cm H₂O were common. Patients with PPCs received higher F_IO₂ during OLV, while they had lower minimum SpO₂ than those without PPCs. However, in multivariate logistic regression analysis adjusting for ventilator settings, ARISCAT score, and minimum SpO₂, only TWA F_IO₂ was associated with the occurrence of PPCs, and the adjusted OR per F_IO₂ increase of 0.1 was 1.30. Therefore, an increase in oxygen concentration of 10% was associated with approximately 30% increase in the risk of PPCs.

We found that V_T was around 6 ml/kg, and PEEP was set around 4 cm H₂O in most patients. These findings were consistent with recent studies or textbook oriented lung protective strategy [15, 21, 22]. We also found that high F_IO₂ was frequently used during OLV. These findings, however, were inconsistent with recent recommended management [22]. An F_IO₂ of 1.0 was classically a routine component of OLV [15, 23]. However, the incidence of hypoxemia during OLV has been decreasing [15, 22], and the harmful effects of high F_IO₂, including absorption atelectasis [24‐27], production of reactive oxygen species, and increased lung injury [28, 29], have been reported. Therefore, this classic practice has been questioned and avoidance of excessive F_IO₂ has been proposed [15]. The latest textbook suggests that F_IO₂ should be titrated to maintain a stable saturation level above 92-94% during OLV [22]. However, some reports revealed that relatively high F_IO₂ was still applied as a common practice during both two-lung ventilation [30, 31] and OLV [13‐16]. In our survey, intraoperative minimum SpO₂ was ≥95% in 111 patients (56%), with 83% of them receiving TWA F_IO₂ of ≥0.6 (Additional file 3). These findings indicated that almost half of the patients may have received excessive oxygen regardless of their SpO₂. There was low compliance with recommended standards to maintain a SpO₂ above 92-94% during OLV.

According to our results, high F_IO₂ during OLV was independently associated with the increasing incidence of PPCs, and patients with PPCs had a longer LOS in the hospital. Worse clinical outcomes due to high F_IO₂ were previously reported in critically ill adults, including patients with chronic obstructive pulmonary disease, myocardial infarction, cardiac arrest, stroke, and traumatic brain injury [32‐35]. Given the above concern, a conservative oxygenation strategy has been shown to be feasible, safe, and effective for mechanically ventilated patients in recent decades [36, 37]. Notably, conservative oxygen therapy could be associated with decreased evidence of atelectasis as well as earlier weaning from mandatory ventilation in the ICU [38]. Additionally, a recent randomized control trial of conservative oxygen therapy in ICU showed lower mortality [39].

Only a few studies investigated the effect of intraoperative F_IO₂ on clinical outcomes in thoracic surgery with OLV. Yang et al. reported a lower incidence of postoperative lung dysfunction and satisfactory gas exchange was provided by the lung protective strategy using F_IO₂ of 0.5 compared to the conventional strategy using F_IO₂ of 1.0 during OLV [40]. However, F_IO₂ was one of components in this lung protective strategy, because V_T, PEEP, and mode of mechanical ventilation were also different between the groups. Thus, it remains uncertain whether a conservative approach to oxygen therapy during OLV is beneficial or not. To our knowledge, this is the first study to demonstrate an association between high F_IO₂ during OLV and the occurrence of PPCs. To confirm and dissect these findings, additional studies should be performed in different settings. Moreover, our findings support the need for randomized control trials to evaluate the safety and feasibility of conservative oxygen therapy during OLV.

There were several limitations in this study. First, because this was an observational study, causality was not determined. It should be noted that higher F_IO₂ might be confounded by the incidence of hypoxemia, which could cause PPCs. Thus, the role of F_IO₂ is difficult to differentiate between “unnecessary use” and “need for higher support.” However, after adjusting by ARISCAT score, minimum SpO₂, ΔP, and PEEP to reduce potential confounding, only higher F_IO₂ remained statistically significant as an independent risk factor for PPCs. In subgroup analyses, F_IO₂ has been associated with the incidence of PPCs even in patients with comparatively lower risk for PPCs. Additionally, the present study indicated that patients might receive excessive oxygen during OLV. Therefore, we believe that intraoperative F_IO₂ could be titrated safely even during OLV.

Second, the incidence of PPCs could have heavily depended on our definition. There are various definitions of PPCs. For instance, pneumonia was diagnosed based on radiologic images, symptoms, laboratory findings, or antimicrobial treatment used. The diagnosis of atelectasis was based on images or bronchoscopy. In our study, we used definitions of PPCs from previous studies [17, 20] and CDC guidelines [19] as shown in Fig. 1. As a result, the incidence of PPCs in our study (25.9%) was similar to that of previous works [17, 20].