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 FiO₂ 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 follow-up 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.

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 FiO₂ was calculated on the basis of the FiO₂ 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 (paO₂) 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.

The primary outcome measure was recurrence-free survival in the period from index surgery until the last known follow-up. 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.

The entire patient cohort was sorted based on ascending mean intraoperative FiO₂ values and was then divided into two equal sized groups. The low-FiO₂ group and the high-FiO₂ 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-FiO₂ 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 FiO₂ values before patients were equally divided into a low- and high-FiO₂ group within each entity. Kaplan–Meier curves of these groups were compared within entities. Then patients from the three–tumor entity specific low-FiO₂ groups were combined into one-entity-controlled low-FiO₂ group, and patients from the three-tumor entity specific high-FiO₂ groups were combined in one entity-controlled high-FiO₂ 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 FiO₂ 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 FiO₂, 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 1-year 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.

Statistical analyses were performed using IBM SPSS Statistics 26.0 (SPSS, Chicago, IL) and Prism 9.0.0 (GraphPad Prism Software, Inc, San Diego, CA).

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-FiO₂-groups of patients with CRC and PC (CRC: 88.45 μg ± 45.04 vs. 80.12 μg ± 32.37; low vs. high-FiO₂, p = 0.012; PC: 80.86 μg ± 51.16 vs. 72.22 μg ± 42.85; low vs. high-FiO₂, p = 0.002; HC: 73.21 μg ± 42.23 vs. 66.92 μg ± 27.80; low- vs. high-FiO₂, p = 1.0). Epidural anesthesia was more common in the high-FiO₂-groups of patients with CRC and PC (CRC: 21(10%) vs. 39 (19%); low- vs. high-FiO₂, p = 0.011; PC: 230 (71%) vs. 258 (79%); low- vs. high-FiO₂, p = 0.011; HC: 10 (71%) vs. 10 (77%), p = 0.745). In total, 111 (10.2%) patients were transfused intraoperatively and 284 (26.2%) during their entire hospital stay. There was no difference between the entity-controlled low-FiO₂ and entity-controlled high-FiO₂ groups (intraoperative transfusion: 54 (9.9%) vs. 57 (10.5%); entity-controlled low-FiO₂ vs. entity-controlled high-FiO₂; p = 0.784; transfusions during entire hospital stay: 139 (25.6%) vs. 145 (26.8%); entity-controlled low-FiO₂ vs. entity-controlled high-FiO₂; p = 0.652).

The average age was higher in the low-FiO₂ group (64 ± 10.5 vs. 62 ± 11 years, low- vs. high-FiO₂,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-FiO₂ group (81 vs. 72%, p = 0.009; low- vs. high-FiO₂). There was no difference in CRC patients with regard to tumor localization between both groups. In CRC surgery, open surgery was more common in the high-FiO₂ group (85 vs. 92%, p = 0.029; low- vs. high-FiO₂). PC and HC surgery were always performed as open surgery. Robotic surgery was not performed in our patient cohort. The proportions of the three tumor entities differed between the low and high-FiO₂ group (PC: 72.1%, CRC: 26.8%, HC: 1.1% vs. PC 48.2%, CRC 48.0%, HC 3.9%, p < 0.001; low- vs. high-FiO₂). The average duration of PC surgery was longer than the duration of CRC and HC surgery (PC:313 min ± 103, CRC 221 min ± 83, HC: 168 min ± 69, p < 0.001).

After dividing the cohort into low and high FiO₂ patients, the cut off value between the two groups was 45.1%, the median of patients' intraoperative FiO₂ mean [Q1; Q3] in the low-FiO₂ group was 40.9 [38.3; 42.9] vs. 50.4% [47.4; 54.7] in the high-FiO₂ group (Figure 2A). In CRC surgery, the calculated cut off value was 48.1%, the median of patients' FiO₂ means in the low-FiO₂ group was 42.8 [40.0; 45.6] vs. 53.1% [50.3; 58.6] in the high-FiO₂ group. In PC surgery, the calculated cut off value was 43.6%, the median of patients' FiO₂ means in the low-FiO₂ group was 39.8 [37.7; 41.8] vs. 47.9% [45.5; 52.1] in the high-FiO₂ group. In HC surgery, the calculated cut off value was 54.0%, the median of patients' FiO₂ means in the low-FIO₂ group was 45.5 [41.0; 47.9] vs. 60.2% [56.1; 62.8] in the high-FiO₂ group (Figure 2B).

The mean paO₂ during surgery was significantly higher in the high-FiO₂ group (median of patients' intraoperative paO₂ mean: 172.7 mmHg [152.15; 193.05] vs. 200.7 mmHg [173.2; 235.8], low-vs. high-FiO₂) (Figure 3A). After stratifying for tumor entities, median paO₂ values were also significantly lower in the low-FiO₂ groups compared to the high-FiO₂ groups (CRC: Median of patients' intraoperative paO₂ mean: 167.9 mmHg [146.5; 188.0] vs. 190.0 mmHg [147.0; 217.5], low-vs.high-FiO₂; PC: Median of patients' intraoperative paO₂ mean: 172.38 mmHg [152.6; 190.3] vs. 203.4 mmHg [174.1; 230.0], low-vs. high-FiO₂; HC: Median of patients' intraoperative paO₂ mean: 197.0 mmHg [162.3; 214.3] vs. 238.33 mmHg [198.3; 296.4], low-vs. high-FiO₂) (Figure 3B).

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 distant metastases, 309 (52.5%) had distant metastases without local recurrence, and 133 patients (22.6%) suffered from both local relapse and distant metastases. For 10 patients (1.7%), information about the type of recurrence was not available. During the observation period, 402 patients (37.1%) died, mostly (381 cases) due to their malignant disease. Two patients died due to postoperative complications; four patients because of cardiovascular events after hospital discharge. In 15 patients, the cause of death was unknown. Recurrence-free survival was considerably higher in the high-FiO₂ group (log-rank-test: p < 0.001) (Figure 4). This effect was preserved in the entity-controlled groups. Recurrence-free survival was lower in the entity-controlled -low-FiO₂ group compared to the entity-controlled -high-FiO₂ 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 (20–22). 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 FiO₂ 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 FiO₂. The hazard ratio (HR) for the variable mean FiO₂ was 0.965, and thus, HR is 0.965⁵ = 0.837 between two patients whose mean FiO₂ differ by 5%; per 1% rise of mean FiO₂ 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 FiO₂ 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.

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-FiO₂ group. After 5 years, five out of 203 patients had died (5-year survival 96.2%) in the low-FiO₂ group, whereas eight out of 202 patients had died in the high-FiO₂ group (5-year survival 94.6%). In PC patients, 39 out of 326 patients had died after 1 year (1-year survival 87.9%) in the low-FiO₂ group, whereas in the high-FiO₂ group 43 out of 326 patients had died (1-year survival 86.7%). After 5 years, 188 patients in the low-FiO₂ group had died (5-year survival 36.0%) vs. 183 patients in the high-FiO₂ group (5-year survival 38.5%). In HC patients, 1-year and 5-year survival was 100% in both the low-and the high-FiO₂ group. Neither 1- nor 5-year survival was different between the entity-controlled low-FiO₂ and entity-controlled high-FiO₂ group (p = 0.525) (Table 4).

Cardiovascular events (5 vs. 5, low-vs. high-FiO₂, p = 1) and incidence of sepsis (5 vs. 6, low- vs. high-FiO₂, p = 0.76) did not differ between groups. There was no difference in the rate of SSI (61 vs. 55, low-vs. high-FiO₂, p = 0.556), and the occurrence of reoperations was not different (40 vs. 36, p = 0.634, low-vs. high-FiO₂) (Table 4).