Intraoperative Fractions of Inspiratory Oxygen Are Associated With Recurrence-Free Survival After Elective Cancer Surgery

Background: Choice of the fraction of inspiratory oxygen (FiO2) is controversial. The objective of this analysis was to evaluate whether intraoperative FiO2 was associated with recurrence-free survival after elective cancer surgery. Methods and Analysis: In this single-center, retrospective study, we analyzed 1,084 patients undergoing elective resection of pancreatic (n = 652), colorectal (n = 405), or hepatic cancer (n = 27) at Heidelberg University Hospital between 2009 and 2016. Intraoperative mean FiO2 values were calculated. For unstratified analyses, the study cohort was equally divided into a low- and a high-FiO2 group. For cancer-stratified analyses, this division was done within cancer-strata. The primary outcome measure was recurrence-free survival until the last known follow-up. Groups were compared using Kaplan–Meier analysis. A stratified log rank test was used to control for different FiO2 levels and survival times between the cancer strata. Cox-regression analyses were used to control for covariates. Sepsis, reoperations, surgical-site infections, and cardiovascular events during hospital stay and overall survival were secondary outcomes. Results: Median FiO2 was 40.9% (Q1–Q3, 38.3–42.9) in the low vs. 50.4% (Q1–Q3, 47.4–54.7) in the high-FiO2 group. Median follow-up was 3.28 (Q1–Q3, 1.68–4.97) years. Recurrence-free survival was considerable higher in the high-FiO2 group (p < 0.001). This effect was also confirmed when stratified for the different tumor entities (p = 0.007). In colorectal cancer surgery, increased FiO2 was independently associated with increased recurrence-free survival. The hazard for the primary outcome decreased by 3.5% with every 1% increase in FiO2. The effect was not seen in pancreatic cancer surgery and we did not find differences in any of the secondary endpoints. Conclusions: Until definite evidence from large-scale trials is available and in the absence of relevant clinical conditions warranting specific FiO2 values, perioperative care givers should aim for an intraoperative FiO2 of 50% in abdominal cancer surgery as this might benefit oncological outcomes.


INTRODUCTION
An increased fraction of inspiratory oxygen (FiO 2 ) is frequently used in anesthesia, intensive care, and emergency medicine. A concept of using 80 vs. 30% FiO 2 has been proposed for the prevention of surgical site infections (SSI) more than 20 years ago (1) and has been evaluated in several subsequent randomized controlled trials (RCTs) (1)(2)(3). Based on a systematic review and meta-analysis (4), the WHO and other societies (5-7) issued guidelines advocating the use of 80% FiO 2 during surgery. These recommendations have caused considerable concern, as the authors of other systematic reviews and meta-analyses arrive at opposing conclusions (8,9) and because the relevance of potential oxygen-mediated side effects is still under debate (9,10). In a post-hoc analysis of the PROXI trial, administration of 80% oxygen in the perioperative period was associated with increased long-term mortality in patients undergoing cancer surgery (11), and after a median follow-up of 3.9 years, cancer-free survival was significantly reduced in the 80% oxygen group compared with patients randomized to 30% FiO 2 (12). Outside of clinical trials and before the first WHO recommendation regarding supplemental oxygen (13) was published, intraoperative FiO 2 was typically used in the range of 30% to more than 90%, and the choice for the actual FiO 2 did depend on patients' requirements, type of surgical intervention, and anesthesiologist's preferences (14).
Currently, it is unknown whether the choice of higher or lower FiO 2 in a real-world setting outside of clinical trials is associated with long-term outcomes after cancer surgery. Therefore, we conducted a retrospective and exploratory study to evaluate the association of perioperative FiO 2 and recurrencefree survival after elective cancer surgery was performed before the WHO recommendation advocating supplemental oxygen for prevention of SSI was published.

Study Design and Population
We performed a retrospective and exploratory study in patients receiving general anesthesia for elective abdominal cancer surgery in the period from January 1, 2009, to December 31, 2016, at the Department of General-, Visceral-, and Transplant Surgery, Heidelberg University Hospital, Heidelberg, Germany. The study protocol conformed to the "Strengthening the Reporting of Observational studies in Epidemiology (STROBE)" guidelines (15), and the principles of the Declaration of Helsinki and was approved by the local Ethics Committee of the Medical Faculty of the Ruprecht-Karls-University, Heidelberg (S-326/2019, August 9, 2019). We assessed whether perioperative FiO 2 was associated with recurrence-free survival after surgery for three exemplary abdominal cancer entities. Patients with elective surgery of the pancreas, colon or rectum, or liver were included. Subgroup analysis was performed for the three tumor entities. Data were retrieved for all patients ≥ 18 years of age who underwent elective resection of pancreatic (PC), colorectal (CRC,) or hepatic cancer (HC) with R0 (no residual tumor) or R1 (microscopic residual tumor) resection, without distant metastases at the time of surgery, and with followup for at least 180 days. Exclusion criteria were intraoperative detection of peritoneal carcinomatosis or if histology analysis could not reveal cancer tissue i.e., after neoadjuvant chemo-or radiotherapy. Patients were also excluded if they suffered familial adenomatous polyposis or did receive home oxygen therapy in the preoperative period.

Data Collection
Baseline data retrieved from the prospectively maintained electronic pancreas, colorectal, and liver databases of the Department of Surgery at Heidelberg, University Hospital and from the electronic patient file were: demographic data, weight, height, body mass index (BMI), American Society of Anesthesiologists physical status classification (ASA), preexisting diseases including diabetes mellitus and chronic obstructive pulmonary disease (COPD), history of smoking, duration of surgery, need for postoperative ventilation, reintubation, use of epidural anesthesia, intraoperative dose of sufentanil, units of red blood cells (RBC), fresh frozen plasma (FFP), and platelet concentrates (PLT) transfused during surgery and during the entire hospital stay, use of intraoperative radiation therapy (IORT), neoadjuvant and adjuvant radio-, and chemotherapy. For colorectal and pancreatic cancer, tumor site was differentiated as follows: pancreatic head vs. body/tail cancer and rectal vs. colon cancer. Resection margin status, TNM (tumor, node, metastasis) classification, and tumor grading were retrieved from pathology reports. For every patient mean FiO 2 was calculated on the basis of the FiO 2 levels documented in the anesthesia protocol at 15-min intervals after induction of anesthesia (first 45 min) until extubation. Furthermore, the measurements of partial pressure of oxygen in arterial blood gases (paO 2 ) during surgery were collected. For outcome analyses, intensive care unit and hospital length of stay, date of local recurrence or occurrence of distant metastases, and date and cause of death were collected.

Outcome Analysis
The primary outcome measure was recurrence-free survival in the period from index surgery until the last known followup. Recurrence-free survival was defined as the time from index surgery to the first documented event of local cancer recurrence, newly diagnosed metastases, or death. Computer tomography, abdominal ultrasound, physical examination, and blood sampling were conducted in follow-up examinations at regular intervals or were prompted by new symptoms. If there was no diagnosis of cancer recurrence, new metastases, or death documented, the last date of follow-up or doctor-patient contact with negative findings was recorded. Secondary outcomes were overall survival, sepsis, reoperations due to surgical complications, SSI (superficial incisional, deep incisional, and organ space) and cardiovascular events (myocardial infarction, cerebral infarction, or transitory ischemic attack) during hospital stay were recorded.

Statistical Analysis
The entire patient cohort was sorted based on ascending mean intraoperative FiO 2 values and was then divided into two equal sized groups. The low-FiO 2 group and the high-FiO 2 group comprised 542 patients each. Descriptive analyses comprised calculation of mean, standard deviation (SD), median, and first and third quartile for continuous variables and absolute and relative proportions for categorical variables. The distribution of categorical variables in the different groups was compared using the chi-square test. Differences in continuous variables were evaluated using the Mann-Whitney U test. The primary survival analysis for the prespecified primary endpoint recurrence-free survival was performed using the Kaplan-Meier method (16) and groups were compared by means of the log-rank test (17). As the three tumor entities were not equally distributed over the low-and high-FiO 2 groups, and because the tumor entities analyzed in our study differed regarding long-term survival rates, we performed an additional analysis to control for tumor entity. First, within each of the three tumor entities, patients were sorted by ascending intraoperative mean FiO 2 values before patients were equally divided into a low-and high-FiO 2 group within each entity. Kaplan-Meier curves of these groups were compared within entities. Then patients from the three-tumor entity specific low-FiO 2 groups were combined into one-entity-controlled low-FiO 2 group, and patients from the three-tumor entity specific high-FiO 2 groups were combined in one entity-controlled high-FiO 2 group. These two groups were compared using a stratified log-rank test with the different tumor entities as strata. Thereafter, the primary outcome was analyzed within the tumor entities PC and CRC by the Cox proportional hazard model (18) in which the effect of mean FiO 2 on recurrence-free survival adjusted for the following covariates was estimated: gender, age, body mass index, nicotine use, diabetes mellitus, UICC (Tumor classification according to the Union for International Cancer Control) stage, tumor localization, tumor grading, resection margin status, use of epidural anesthesia, intraoperative dose of sufentanil, units of RBC, FFP, and PLT transfused during the entire hospital stay, intraoperative, neoadjuvant and adjuvant radio-and chemotherapy, and laparoscopic surgery in patients with CRC. If fentanyl was used as the opiate instead of sufentanil, the fentanyl dose was multiplied by a factor of 0.1 to calculate the equivalent sufentanil dose (19).
Mean FiO 2 , age, PLT transfusion, intraoperative dose of sufentanil, and BMI are continuous variables, and the other variables are categorial. P-values of regression coefficients were obtained by the Wald-Test. Hazard ratios (HRs) estimated from the Cox analysis were reported with corresponding 95% CIs and a two-sided p < 0.05 was denoted as considerable. The HC-stratum was too small to fit a covariate-adjusted cox regression. The 1year and 5-year overall survival was estimated by the Kaplan-Meier method. P-value is referred to the stratified log rank test with entities as strata.

RESULTS
Data from 1,214 patients were retrieved from databases. Datasets from 100 patients could not be assessed for eligibility because of incomplete or missing anesthesia records since the patient ID retrieved from the database did not match with a case in the hospital information system, or because the type of surgery was not eligible, i.e., non-tumor surgery. Considering exclusion criteria, 30 patients had to be excluded. Therefore, 1,084 patients were included in the final analysis set (Figure 1) with a median follow-up of 3.28 (Q1-Q3, 1.68-4.97) years.

Patient Characteristics
Main clinical and demographical baseline characteristics are presented in Table 1 and the Supplementary Material. A total of 1,084 patients underwent elective PC (n = 652), CRC (n = 405), or HC surgery (n = 27). The mean age was 63 ± 11 years. Male participants were slightly more than female participants. At the time of surgery, the most common cancer stage was UICC 2 and the most common tumor grades were 1 and 2. Around 45% of patients' tumor resection margin status in the pathology report was classified as R1. In the majority of patients with PC, "pancreatic head" was found as the tumor localization; in the majority of patients with CRC "rectum" was found as the tumor localization. Patients received neoadjuvant chemotherapy or neoadjuvant radiotherapy in 15% of cases and 70% received adjuvant chemotherapy. IORT was conducted in 3%. General anesthesia was performed as balanced anesthesia except in four patients. In a small proportion (four patients (0.3%)), fentanyl was used instead of sufentanil and the corresponding equivalent dose was calculated. Intraoperative dose of sufentanil was slightly higher in the low-  The average age was higher in the low-FiO 2 group (64 ± 10.5 vs. 62 ± 11 years, low-vs. high-FiO 2, p = 0.003). For none of the tumor entities, differences with regard to diabetes, history of smoking, or gender were detected. Also, for neither of the tumor entities we observed differences regarding tumor grading status, UICC stage, resection margin status, neoadjuvant, adjuvant therapy, or IORT. In PC surgery, tumor localization in the pancreatic head was more common in the low-FiO 2 group (81 vs. 72%, p = 0.009; low-vs. high-FiO 2  (Figure 2B).
The mean paO 2 during surgery was significantly higher in the high-   (Figure 3B).

Survival Analysis
In total, cancer recurrence between index surgery and last follow-up was diagnosed in 588 patients (54%). Among those patients, 136 (23.1%) experienced local recurrence and no   (Figure 4). This effect was preserved in the entity-controlled groups. Recurrence-free survival was lower in the entity-controlled -low-FiO 2 group compared to the entitycontrolled -high-FiO 2 group (stratified log rank test: p = 0.007) ( Figure 5). The cancer entity itself plays a relevant role in the average survival time of a patient (20)(21)(22). Therefore, we performed subgroup analyses for PC and CRC patients respectively using a Cox proportional hazard model, considering factors influencing the observed effect. These analyses revealed an independent association for mean FiO 2 and recurrence-free survival in patients undergoing CRC surgery (B = −0.035; p = 0.025). Recurrence-free survival was longer in patients with higher mean FiO 2 . The hazard ratio (HR) for the variable mean FiO 2 was 0.965, and thus, HR is 0.965 5 = 0.837 between two patients whose mean FiO 2 differ by 5%; per 1% rise of mean FiO 2 the hazard for the primary outcome decreases by 3.5% (Exp(B): 0.965; 95% CI:0.936-0.996) ( Table 2). There was no evidence for an association of FiO 2 mean with recurrence-free survival in patients undergoing PC surgery (B = -0.01; Exp(B) = 0.990; CI:0.975-1.01; p = 0.217) ( Table 3). Because the HC group consists of 27 patients only, it was not considered suitable for an individual Cox regression analysis. FIGURE 4 | FiO 2 and recurrence-free survival. Patients were divided into low-FiO 2 group and high-FiO 2 group. The p-value was evaluated using the log-rank test (p < 0.001). FiO 2 , Fraction of inspired oxygen.

Secondary Endpoints
We examined 1-year and 5-year survival rates in patients with CRC, PC, and HC separately. In CRC patients, 1-year survival was 100% in both the low-and the high-  both the low-and the high-FiO 2 group. Neither 1-nor 5-year survival was different between the entity-controlled low-FiO 2 and entity-controlled high-FiO 2 group (p = 0.525) ( Table 4).

DISCUSSION
In this retrospective exploratory study, we demonstrated in a real-world setting that higher intraoperative FiO 2 during abdominal cancer surgery was associated with better recurrencefree survival. In CRC surgery, higher FiO 2 was independently associated with increased recurrence-free survival. The hazard for the primary outcome decreased by 3.5% with every 1% increase in FiO 2 . In PC surgery, we did not observe a significant effect and we did not find differences in any of the secondary endpoints.
The patient cohort under investigation had undergone surgery before the first WHO recommendation advocating supplemental oxygen for prevention of SSI (13) was published. Therefore, the decision for a specific FiO 2 was a clinical decision mainly based on the patients' respiratory function and the anesthetists' preference.
Oxygen has known pro-and anticarcinogenic effects. Hyperoxia may not only cause direct cell damage through the generation of reactive oxygen species (ROS) but also mediates cancerous effects (12,23,24), such as promotion of proliferation, invasiveness, angiogenesis, and metastasis. Interestingly, ROS generation may be accelerated under hyperoxia and also hypoxia (14). In addition, hypoxia can lead to organ dysfunction, lactic acidosis, cell death (25), and an increased expression of the hypoxia-inducible factor 1 (HIF-1) (26,27) an intrinsic survival factor of tumor cells. Therefore, hypoxia can induce a vast number of gene products that control neovascularization, cell survival, energy metabolism, intracellular pH, and cell migration and are often associated with increased tumor aggressiveness, therapeutic resistance, and mortality (26,(28)(29)(30)(31).
In clinical studies, evidence for FiO 2 -mediated effects associated with long-term outcome after abdominal cancer surgery is sparse and controversial. In a post-hoc analysis of the PROXI trial, cancer-free survival was significantly shorter in the high oxygen group (12). In line, after a median followup of 2.3 years, a subsequent analysis of the same trial found increased long-term mortality in patients receiving 80% oxygen during cancer surgery (11). Contrarily, authors of other post-hoc analysis, combining mortality data from different RCTs, found that there was neither a difference in long-term mortality nor did they find differences in overall survival for patients randomly assigned to 30% vs. 80% FiO 2 (32,33). The median follow-up was 12.8 (3.8, 13.6) years in the trial by Greif,4.3 (3.6,4.8) years in the trial by Kurz, and 3.2 (0.5, 4.9) years in the study by Jiang (1,33,34). This is in line with our finding that overall survival did not differ between high and low FiO 2 groups with a median follow-up of 3.28 (1.68, 4.97) years. Although FiO 2 appears to affect cancer recurrence, follow-up in our study might have been too short to observe effects on mortality after recurrence. Also, overall survival did not differ between the two groups investigated here, probably because overall survival depends on a variety of other factors not assessed in this study.
In most prospective studies, comparing different FiO 2 concentrations, patients were randomized for either 30 or 80% FiO 2 . However, both targets have been accused to cause considerable side effects. Outside of clinical trials and before guidelines advocating high levels of supplemental oxygen were published, most anesthetists chose moderate FiO 2 levels between the high and low extremes. As a result, in our retrospective study, mean FiO 2 levels in the low-and high-FiO 2 groups were closer together than in other studies. Importantly, our institutional standards suggest avoiding not only hypoxia but also hyperoxia as both conditions might cause adverse events.
Pro and anticarcinogenic effects of oxygen might explain why, on one hand, 80% FiO 2 has adverse effects over 30% FiO 2 in some studies, but on the other hand, in our study, FiO 2 in the range of 50% compared with lower FiO 2 concentrations was beneficial with regard to recurrence-free survival. It is conceivable that the optimal FiO 2 for tumor surgery patients is neither met with an FiO 2 of 30% nor 80%. In fact, the association of FiO 2 or paO 2 respectively and recurrence-free survival may follow a V-shaped curve.
Oxygen related effects in tumor biology may also differ for distinct types of tumors. The association of tumor-associated transcription factors and tumor progress varies among different  (32,33). In the post-hoc analysis of the PROXI trial, about half the patients underwent colorectal surgery (12). A subgroup analysis for non-colorectal procedures was not conducted, although cancer histology was considered (12). Additional adverse effects of high-FiO 2 affecting the cardiovascular and respiratory systems have been reported in the literature. Hyperoxia induces increased peripheral vascular resistance, promotes reduced cardiac output, and mediates coronary vasoconstriction (14,35). A post-hoc analysis of the PROXI-trial showed that patients with a FiO 2 of 80% had a significantly increased risk of myocardial infarction, acute coronary syndrome, or death (36). Contrary, the authors of a systematic review and meta-analysis concluded, based on 17 RCTs and two non-randomized studies, that perioperative supplemental oxygen was not associated with relevant complications (10). Consistent with this report, our analysis did not reveal differences for secondary cardiovascular endpoints or redo surgery.
WHO, ACS (American College of Surgeons), and CDC (Centers for Disease Control and Prevention) recommends supplemental oxygen with the aim to prevent SSI. In our study, SSI did not differ between the high-and low-FiO 2 groups. One could argue that the FiO 2 in the high-FiO 2 group was not sufficiently high to prevent SSI. However, the incidence of SSI reported in this study is lower than reported in most of the RCTs testing different oxygen levels in abdominal cancer surgery patients. Our study has some limitations that need to be addressed. We performed a retrospective, single-center study with a limited set of exemplary tumor entities. The number of individuals included in the analysis was limited by the availability of digitalized anesthesia records, and only patients recruited in the surgical databases could be analyzed. The choice of FiO 2 values was not standardized and the patients were equally split into the low-and high-FiO 2 groups. The patients' intraoperative FiO 2 levels were compared by the calculation of the individual mean FiO 2 , but the duration of exposure and individual dose during surgery were not included in this calculation. A consequence of the retrospective nature of our study is that we do not have full control over confounders. As tumor entity itself affects outcome, and we conducted an analysis stratified for entities and performed a cox regression analysis to control for confounders. However, we cannot fully exclude, that additional confounders affected findings reported in this study.
In conclusion, we demonstrate that, within a moderate range of FiO 2 , higher FiO 2 during abdominal cancer surgery was associated with longer recurrence-free survival. In colorectal cancer surgery, increased FiO 2 was independently associated with increased recurrence-free survival. The findings will be instrumental for designing prospective studies delineating effects of certain FiO 2 on cancer development and progression for specific patient populations. Until definite evidence from largescale randomized controlled trials is available, anesthesiologist, in the absence of relevant clinical conditions warranting specific FiO 2 values, should aim for an intraoperative FiO 2 of 50% in abdominal cancer surgery as this might benefit oncological outcome.

DATA AVAILABILITY STATEMENT
The data analyzed in this study is subject to the following licenses/restrictions: Datenschutzgrundverordnung. Requests to access these datasets should be directed to jan.larmann@med.uni-heidelberg.de.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Ethics Committee of the Medical Faculty of the Ruprecht-Karls-University, Heidelberg. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

AUTHOR CONTRIBUTIONS
SD, VS, RK, and JL designed research and study protocol. SD, VS, LK, SK, and JL formal analysis. JL, SD, and VS statistical analysis. SK methodology and statistical consultation. SD, VS, RK, LK, KH, MS, TH, MB, and MW contributed to the data curation. SD, VS, and JL wrote the first draft of the manuscript. All authors critically reviewed and revised the manuscript and approved the final work.

FUNDING
This study was funded by institutional resources from the Departments of Anesthesiology and General, Visceral and Transplant Surgery. The position of LK was funded by the German Research Foundation (LA 2343/7-1) to JL. This research did not receive any other specific grants from funding agencies in the public, commercial, or not-for-profit sectors.