Combinations of predictors of exercise-induced oxygen desaturation after lung resection: a retrospective observational study

Introduction

Background

Lung cancer ranks among the most prevalent cancers worldwide (1). Lung resection is an effective, widely applied treatment for this disease (2). Postoperative pulmonary complications after lung resection, including the need for postoperative home oxygen therapy, are associated with poor patient outcomes (3,4). One such complication is exercise-induced oxygen desaturation (EID), with incidence rates ranging from 15.4% to 22.9% (5-7). EID results from reduced tidal volume and ventilation-perfusion mismatch due to resection extent and coexisting respiratory diseases such as chronic obstructive pulmonary disease (8). Postoperative EID can lead to dyspnea (9), decreased exercise tolerance (10,11), and reduced health-related quality of life after discharge (12). It is also linked to increased mortality among patients with chronic respiratory disease (13,14). Therefore, accurately predicting postoperative EID in lung resection patients and selecting preventive treatment methods are crucial. Perioperative pulmonary rehabilitation may benefit patients who develop EID, and oxygen therapy should be considered when indicated (15-17). Together, identifying predictors of postoperative EID in this population is imperative to facilitate prompt intervention.

Rationale and knowledge gap

Few studies have examined the predictors of postoperative EID. Specifically, factors that may inform physical therapy interventions for EID remain unclear. Most assessment methods for EID in patients undergoing lung resection involve stair-climbing tests (18), which may be difficult to perform in some patients due to excessively high workloads or environmental factors. Contrastingly, the 6-minute walk test (6MWT) does not require specialized equipment or expertise; further, it can be performed in various institutional settings and across a wide range of patients (19). Decision-tree analysis, which is a form of data mining, yields cutoff values for predictors and a tree diagram (20). This method allows proximity to the decision patterns in actual diagnostic situations and thus is a useful tool for building predictive models.

Objective

We aimed to examine combinations of predictors for EID in patients who underwent lung resection using decision-tree analysis. We present this article in accordance with the STROBE reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1046/rc).

Methods

Study design and participants

This singlecenter, retrospective, observational study included patients scheduled for lung resection for neoplastic lung disease at the Second Department of Surgery, University of the Ryukyus Hospital, between October 2017 and April 2023. Inclusion criteria were: ability to walk unaided at admission and discharge; absence of orthopedic or cerebrovascular disease interfering with daily life; receipt of preoperative physical therapy (21); and cognitive ability to understand test instructions. Exclusion criteria were: missing preoperative or postoperative 6MWT results; unstable saturation of peripheral oxygen (SpO₂) measurements during the 6MWT; requirement for oxygen therapy; missing data; and refusal to participate by the patient or their representative.

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by University of the Ryukyus “Ethics Review Committee for Medical Research Involving Human Subjects” (approval No. 22-1922-02-01-00). Due to the retrospective nature of this study and the full anonymization of patient data, the requirement for individual informed consent was waived by the ethics committee; instead, we adopted an opt-out approach by publicly disclosing information about the study.

Clinical parameters

We collected basic sociodemographic information from medical records: age, sex, body mass index, smoking history, presence and identity of comorbidities, American Society of Anesthesiologists physical status, and clinical stage of lung cancer. Additionally, we investigated preoperative respiratory function with respect to forced vital capacity (FVC% predicted value), forced expiratory volume in 1 s (FEV_1.0% predicted value), FEV_1.0/FVC, restrictive ventilatory defect, and obstructive ventilatory defect. Surgery-related variables included the surgery type, intraoperative blood loss (IBL), surgery duration, and resection area. Postoperative course variables included postoperative hospital stay, chest drainage period, presence or absence of postoperative complications, and postoperative hemoglobin levels. Postoperative complications were defined as those with Clavien-Dindo classification II or higher. Postoperative hemoglobin levels were collected on postoperative day 1.

Physical function assessment

EID was assessed using the 6MWT preoperatively and postoperatively (1–3 days before surgery and discharge, respectively). The 6MWT was performed on a flat floor without obstacles on a 9-m-long straight path. During the test, participants were instructed to walk the greatest distance possible over 6 minutes. The difference between the SpO₂ value before the test and the lowest SpO₂ value during the test was defined as ΔSpO₂. Accordingly, a decrease in ΔSpO₂ of more than 4% was defined as EID, and a decrease in ΔSpO₂ of less than 4% was defined as non-EID (22,23), with participants being grouped accordingly. Additionally, we assessed postoperative changes in exercise tolerance as the 6-minute walk distance (6MWD) recovery rate, which was calculated as the percentage of the postoperative 6MWD with respect to the preoperative 6MWD. We performed the 6MWT using ATP-W03 (Fukuda Denshi Co. Ltd., Tokyo, Japan). Preoperative handgrip strength was bilaterally measured twice using Grip D (Takei Machinery Co., Ltd., Niigata, Japan), with participants seated in a stable chair. We calculated the hand grip strength (kg) as the average of the maximum values obtained on each side. Preoperative quadriceps force (QF) was bilaterally measured twice at maximum effort, with the knee joint flexed at 90°. The average maximum value [kilogram-force (kgf)] on each side was normalized to body weight (kgf/BW) multiplied by one hundred was used as the QF variable (%BW). Measurements were performed using µTas F-1 (Anima Co., Ltd., Tokyo, Japan). Participants performed the five-repetition sit-to-stand test while seated on a stable chair with their arms crossed. Initially, participants were instructed to stand up and sit down as quickly as possible without using their arms for five cycles. The test was then repeated, and the fastest completion time was recorded. Participants also performed the 4-meter gait speed test on flat, obstacle-free ground. They were instructed to walk at a comfortable pace. Timing stopped when one foot crossed the finish line. The fastest time from two trials was used to calculate gait speed (m/s), which was used as a variable for the 4-meter gait speed test.

Perioperative rehabilitation program

Preoperatively, patients received instructions on respiratory training methods, including incentive spirometry, diaphragmatic breathing exercises, and expectoration techniques, and on early postoperative ambulation. Postoperatively, patients received instructions regarding early ambulation, respiratory training, and expectoration methods once a day from the day after surgery. Depending on the pain severity and the presence or absence of chest drainage or oxygen therapy, walking practice was performed using assisted walking or a walking aid.

Statistical analysis

Continuous variables are presented as mean [standard deviation (SD)] or median [interquartile range (IQR)]. Categorical variables are presented as n (%). Welch’s t-test or the Wilcoxon rank-sum test was used for between-group comparisons of continuous variables. The Chi-squared test or Fisher’s exact test was used for between-group comparisons of categorical variables. Cohen’s d and r (calculated using Z statistics) were used as the effect size when continuous variables were described as means and medians, respectively. The effect sizes of categorical variables were described as W for the 2×2 contingency tables and Cramer’s V for the non-2×2 contingency tables. As a guide for effect size, d was set at 0.2 for small, 0.5 for medium, and 0.8 for large, while r, W, and V were set at 0.1 for small, 0.3 for medium, and 0.5 for large. The Wilcoxon signed-rank test was used for between-group comparisons of postoperative changes in ΔSpO₂. In the decision-tree analysis, EID and non-EID were the outcome variables. Predictor variables were selected based on univariate analysis (P<0.20) and clinical significance. We first performed a decision-tree analysis on all predictor variables, followed by separate analyses of preoperative and intraoperative/postoperative factors. We applied the Classification and Regression Tree (CART) algorithm to predict EID. CART recursively splits the data by predictor values to maximize node homogeneity and construct the decision tree (24). We limited the tree to a maximum of three branches to facilitate clinical interpretation (25). Predictive accuracy was assessed by plotting receiver operating characteristic curves and calculating the area under the curve (AUC). Missing data were handled via complete case analysis. Statistical analyses were conducted using JMP Pro v15 (SAS Institute, Cary, NC, USA) and G*Power v3.1.9.6.

Results

Patient characteristics

We included 100 patients: 43 (43%) in the EID group and 57 (57%) in the non-EID group (Figure 1). Table 1 summarizes baseline characteristics. The EID group had fewer males (P=0.04, W=0.42), a higher prevalence of chronic respiratory disease (P=0.02, W=0.41), and lower QF (P=0.007, d=0.56) than those of the non-EID group. Intraoperatively and postoperatively, the EID group exhibited longer surgery duration (P=0.02, r=0.24), greater IBL (P=0.009, r=0.26), and more frequent complications (P=0.052, W=0.36) than those of the non‑EID group.

Figure 1 Flow diagram of the study participants. 6MWT, 6-minute walk test; EID, exercise-induced oxygen desaturation; ΔSpO₂, Δ saturation of peripheral oxygen.

Postoperative changes in ΔSpO₂ and the 6MWD recovery rate

In the non‑EID group, preoperative ΔSpO₂ [1.7% (IQR, 1.05–3.75)] did not differ significantly from postoperative ΔSpO₂ [2.6% (IQR, 1.65–3.25); P=0.46]. In contrast, postoperative ΔSpO₂ in the EID group [6.7% (IQR, 4.6–8.8)] was higher than that of preoperative ΔSpO₂ [3.3% (IQR, 1.9–4.1); P<0.01]. Preoperative ΔSpO₂ in the EID group was lower than that of the non-EID group (P=0.02) (Figure 2A). The postoperative 6MWD recovery rate was also lower in the EID group [88.1% (SD, 14.0)] than that in the non-EID group [94.1% (SD, 14.7); P=0.04] (Figure 2B).

Figure 2 Changes in each variable before and after surgery in the EID and non-EID groups. (A) Changes in ΔSpO₂. (B) 6MWD recovery rate. 6MWD, 6-minute walk distance; EID, exercise-induced oxygen desaturation; ΔSpO₂, Δ saturation of peripheral oxygen.

Combination of predictive factors for postoperative EID

EID was observed in 43% of enrolled patients. We included fourteen parameters (preoperative: age, sex, chronic respiratory disease, type of neoplastic lung disease, FVC% predicted, FEV_1.0% predicted, hand grip strength, QF, and ΔSpO₂, intraoperative/postoperative: resection area, surgical duration, IBL, postoperative complications, and chest drainage duration) as predictor variables in the decision-tree analysis model. In the analysis performed on all parameters, QF was selected as the root node factor, with all patients with QF ≥64.7% of BW being in the non-EID group (Profile 1). Preoperative ΔSpO₂ and IBL were selected as the internal node and leaf node factors, respectively. The predicted probability of EID was 84% in the following scenario: (I) QF <64.7% BW; (II) preoperative ΔSpO₂ >1.7%; and (III) IBL ≥36 mL (Profile 4) (Figure 3). The AUC of the constructed model was 0.82 [95% confidence interval (CI): 0.73–0.89] (Figure 4, Table 2). Model 1 (preoperative parameters) used QF as the root node, ΔSpO₂ as the internal node, and chronic respiratory disease as the leaf node, yielding an AUC of 0.797 (95% CI: 0.71–0.86). Model 2 (intraoperative and postoperative parameters) used complications as the root node, IBL as the internal node, and resection area as the leaf node, with an AUC of 0.696 (95% CI: 0.59–0.78) (Table 3).

Figure 3 Decision tree model for the prediction of EID. BW, body weight; EID, exercise-induced desaturation; IBL, intraoperative blood loss; QF, quadriceps force.

Figure 4 Receiver operating characteristic curves for individual profiles. Profile 1: decision tree model for preoperative quadriceps force only. Profile 2: decision tree model for preoperative quadriceps force and preoperative ΔSpO₂. Profiles 3 and 4: decision tree model for preoperative quadriceps force, preoperative ΔSpO₂, and intraoperative blood loss. ΔSpO₂, Δ saturation of peripheral oxygen.

Discussion

In this study, we identified QF, preoperative ΔSpO₂, and IBL as the predictors of postoperative EID, with cutoff values of 64.7% BW, 1.7%, and 36 mL, respectively. The combination of these variables predicted postoperative EID with a probability of 84%. Furthermore, the AUC of the constructed predictive model was 0.82 (95% CI: 0.73–0.89).

The strength of this study is that it demonstrates a specific combination of predictors (QF, ΔSpO₂, and IBL) to identify patients at high risk of EID. Moreover, using decision‑tree analysis to establish specific cut‑off values and develop a prediction model is a novel approach. However, the study has some limitations. First, potential selection bias may have occurred because patients were required to complete the 6MWT both preoperatively and postoperatively. Those discharged early after an uneventful postoperative course did not undergo the postoperative 6MWT and were excluded, limiting generalizability to patients with favorable outcomes. Second, we defined EID as the difference between baseline SpO₂ before the 6MWT and the maximum SpO₂ during the test (26). In contrast, other studies calculate EID as the difference between SpO₂ before and after the 6MWT. Consequently, the proportion of patients classified with EID may be larger in our study than that in previous reports. Third, the relatively small sample size (n=100) and single-center design may limit the generalizability of our decision-tree model to broad patient populations. We also did not collect pulmonary diffusion capacity data, a key confounding factor in predicting EID (23). Therefore, further multicenter studies with larger cohorts and comprehensive respiratory assessments are warranted.

We selected the preoperative QF and ΔSpO₂ as the root and internal node factors, respectively. Decreased QF is associated with decreased oxygen utilization and ventilatory efficiency during exercise, which results in dyspnea (27-29). Furthermore, a decrease in SpO₂ during exercise reduces tissue oxygen saturation in the quadriceps muscle, which causes quadriceps fatigue and muscle dysfunction (30). This exacerbates fatigue in the lower extremities with extended exercise and therefore contributes toward decreased exercise tolerance (31). This vicious cycle of impaired exercise tolerance reduces physical activity and affects the function of muscles throughout the body, including the quadriceps muscles, which in turn affects exercise tolerance (32,33). Taken together, this can adversely affect the health-related quality of life. Furthermore, the reduced ventilatory range following lung resection may contribute to postoperative EID by reducing the lung oxygenation capacity, leading to an imbalance between oxygen demand and supply, which may lead to decreased SpO₂ (34). Accordingly, a simultaneous decrease in both the preoperative QF and ΔSpO₂ may have been a predictor of postoperative EID due to the low tolerance to surgery among the selected patients.

In addition, IBL was selected as the leaf node factor for predicting postoperative EID. In this study, one-lung ventilation was used for intraoperative respiratory management during lung resection. During one-lung ventilation, the lungs are simultaneously affected by hypoxia, hyperoxia, and pressure injuries during lung re-expansion (35). This is the main contributor to ventilator-related lung injury, which is exacerbated in a time-dependent manner (36). Given that the surgery duration is associated with IBL (37), increased IBL may be associated with an increased risk of ventilator-associated lung injury.

Our findings suggest that assessing QF and ΔSpO₂ before lung resection and determining the IBL are crucial for predicting postoperative EID. This may facilitate early detection of patients at risk of postoperative EID and appropriate implementation of interventions, including respiratory rehabilitation and oxygen therapy.

Conclusions

The combination of preoperative QF and ΔSpO₂, as well as IBL, facilitated the prediction of postoperative EID. Further studies are warranted to verify the diagnostic validity of the prediction models established in the present study.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1046/rc

Data Sharing Statement: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1046/dss

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1046/prf

Funding: This work was supported by JSPS KAKENHI (grant No. JP22K11460, to N.Y.).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1046/coif). N.Y. reports that this study was supported by JSPS KAKENHI (grant No. JP22K11460). The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by University of the Ryukyus “Ethics Review Committee for Medical Research Involving Human Subjects” (approval No. 22-1922-02-01-00). Due to the retrospective nature of this study and the full anonymization of patient data, the requirement for individual informed consent was waived by the ethics committee; instead, we adopted an opt-out approach by publicly disclosing information about the study.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Committee for Scientific Affairs, The Japanese Association for Thoracic Surgery. Thoracic and cardiovascular surgeries in Japan during 2021: Annual report by the Japanese Association for Thoracic Surgery. Gen Thorac Cardiovasc Surg 2024;72:254-91. [Crossref] [PubMed]
3. Shinohara S, Kobayashi K, Kasahara C, et al. Long-term impact of complications after lung resections in non-small cell lung cancer. J Thorac Dis 2019;11:2024-33. [Crossref] [PubMed]
4. Oslock WM, Xie R, Donahue J, et al. Lung resection and new postoperative home oxygen requirement: a systematic review. J Thorac Dis 2024;16:8796-806. [Crossref] [PubMed]
5. Takeichi H, Urabe N, Asai K. The relation between the frequency of exercise oxygen desaturation following major lung resection and the predictions of postoperative pulmonary function. Kyobu Geka 2008;61:327-30.
6. Varela G, Cordovilla R, Jiménez MF, et al. Utility of standardized exercise oximetry to predict cardiopulmonary morbidity after lung resection. Eur J Cardiothorac Surg 2001;19:351-4. [Crossref] [PubMed]
7. Brunelli A, Al Refai M, Monteverde M, et al. Predictors of exercise oxygen desaturation following major lung resection. Eur J Cardiothorac Surg 2003;24:145-8. [Crossref] [PubMed]
8. Dempsey JA, Wagner PD. Exercise-induced arterial hypoxemia. J Appl Physiol (1985) 1999;87:1997-2006. [Crossref] [PubMed]
9. Dilektasli AG, Porszasz J, Stringer WW, et al. Physiologic Effects of Oxygen Supplementation During Exercise in Chronic Obstructive Pulmonary Disease. Clin Chest Med 2019;40:385-95. [Crossref] [PubMed]
10. Jones LW, Eves ND, Haykowsky M, et al. Exercise intolerance in cancer and the role of exercise therapy to reverse dysfunction. Lancet Oncol 2009;10:598-605. [Crossref] [PubMed]
11. Misu S, Kaneko M, Sakai H, et al. Exercise-Induced Oxygen Desaturation as a Predictive Factor for Longitudinal Decline in 6-Minute Walk Distance in Subjects With COPD. Respir Care 2019;64:145-52. [Crossref] [PubMed]
12. Handy JR Jr, Asaph JW, Skokan L, et al. What happens to patients undergoing lung cancer surgery? Outcomes and quality of life before and after surgery. Chest 2002;122:21-30. [Crossref] [PubMed]
13. Lama VN, Flaherty KR, Toews GB, et al. Prognostic value of desaturation during a 6-minute walk test in idiopathic interstitial pneumonia. Am J Respir Crit Care Med 2003;168:1084-90. [Crossref] [PubMed]
14. Waatevik M, Johannessen A, Gomez Real F, et al. Oxygen desaturation in 6-min walk test is a risk factor for adverse outcomes in COPD. Eur Respir J 2016;48:82-91. [Crossref] [PubMed]
15. Maeda K, Higashimoto Y, Honda N, et al. Effect of a postoperative outpatient pulmonary rehabilitation program on physical activity in patients who underwent pulmonary resection for lung cancer. Geriatr Gerontol Int 2016;16:550-5. [Crossref] [PubMed]
16. Edvardsen E, Skjønsberg OH, Holme I, et al. High-intensity training following lung cancer surgery: a randomised controlled trial. Thorax 2015;70:244-50. [Crossref] [PubMed]
17. Kim SK, Ahn YH, Yoon JA, et al. Efficacy of Systemic Postoperative Pulmonary Rehabilitation After Lung Resection Surgery. Ann Rehabil Med 2015;39:366-73. [Crossref] [PubMed]
18. Nikolić I, Majerić-Kogler V, Plavec D, et al. Stairs climbing test with pulse oximetry as predictor of early postoperative complications in functionally impaired patients with lung cancer and elective lung surgery: prospective trial of consecutive series of patients. Croat Med J 2008;49:50-7. [Crossref] [PubMed]
19. Kalinov R, Marinov B, Vladimirova-Kitova L, et al. The six-minute walk test - a reliable test for detection of exercise-related desaturation in patients with chronic obstructive pulmonary disease. Folia Med (Plovdiv) 2023;65:569-76. [Crossref] [PubMed]
20. Podgorelec V, Kokol P, Stiglic B, et al. Decision trees: an overview and their use in medicine. J Med Syst 2002;26:445-63. [Crossref] [PubMed]
21. Alzahrani M, Mehta R, Kadiri S, et al. Effect of pulmonary rehabilitation on lung cancer surgery outcomes: a matched-case analysis. Perioper Med (Lond) 2025;14:35. [Crossref] [PubMed]
22. Chang CH, Lin HC, Yang CH, et al. Factors Associated with Exercise-Induced Desaturation in Patients with Chronic Obstructive Pulmonary Disease. Int J Chron Obstruct Pulmon Dis 2020;15:2643-52. [Crossref] [PubMed]
23. Gupta R, Ruppel GL, Espiritu JRD. Exercise-Induced Oxygen Desaturation during the 6-Minute Walk Test. Med Sci (Basel) 2020;8:8. [Crossref] [PubMed]
24. Mienye ID, Jere N. A survey of decision trees: Concepts, algorithms, and applications. IEEE Access 2024;12:86716-27.
25. Harada N, Gotoh M, Ishitani E, et al. Combination of risk factors affecting retear after arthroscopic rotator cuff repair: a decision tree analysis. J Shoulder Elbow Surg 2021;30:9-15. [Crossref] [PubMed]
26. Chuang ML, Lin IF, Chen SP. Kinetics of changes in oxyhemoglobin saturation during walking and cycling tests in COPD. Respir Care 2014;59:353-62. [Crossref] [PubMed]
27. Clark A, Rafferty D, Arbuthnott K. Relationship between isokinetic muscle strength and exercise capacity in chronic heart failure. Int J Cardiol 1997;59:145-8. [Crossref] [PubMed]
28. Gagnon P, Saey D, Vivodtzev I, et al. Impact of preinduced quadriceps fatigue on exercise response in chronic obstructive pulmonary disease and healthy subjects. J Appl Physiol (1985) 2009;107:832-40. [Crossref] [PubMed]
29. Aimo A, Saccaro LF, Borrelli C, et al. The ergoreflex: how the skeletal muscle modulates ventilation and cardiovascular function in health and disease. Eur J Heart Fail 2021;23:1458-67. [Crossref] [PubMed]
30. Brunton NM, Barbour DJ, Gelinas JC, et al. Lower-limb resistance training reduces exertional dyspnea and intrinsic neuromuscular fatigability in individuals with chronic obstructive pulmonary disease. J Appl Physiol (1985) 2023;134:1105-14. [Crossref] [PubMed]
31. Romer LM, Haverkamp HC, Lovering AT, et al. Effect of exercise-induced arterial hypoxemia on quadriceps muscle fatigue in healthy humans. Am J Physiol Regul Integr Comp Physiol 2006;290:R365-75. [Crossref] [PubMed]
32. Edbrooke L, Granger CL, Clark RA, et al. Physical Activity Levels Are Low in Inoperable Lung Cancer: Exploratory Analyses from a Randomised Controlled Trial. J Clin Med 2019;8:1288. [Crossref] [PubMed]
33. Avancini A, Sartori G, Gkountakos A, et al. Physical Activity and Exercise in Lung Cancer Care: Will Promises Be Fulfilled? Oncologist 2020;25:e555-69. [Crossref] [PubMed]
34. Nicastri DG, Alpert N, Liu B, et al. Oxygen Use After Lung Cancer Surgery. Ann Thorac Surg 2018;106:1548-55. [Crossref] [PubMed]
35. Shum S, Huang A, Slinger P. Hypoxaemia during one lung ventilation. BJA Educ 2023;23:328-36. [Crossref] [PubMed]
36. Lohser J, Slinger P. Lung Injury After One-Lung Ventilation: A Review of the Pathophysiologic Mechanisms Affecting the Ventilated and the Collapsed Lung. Anesth Analg 2015;121:302-18. [Crossref] [PubMed]
37. Urabe M, Ohkura Y, Haruta S, et al. Factors Affecting Blood Loss During Thoracoscopic Esophagectomy for Esophageal Carcinoma. J Chest Surg 2021;54:466-72. [Crossref] [PubMed]