The effect of high perioperative inspiratory oxygen fraction for abdominal surgery on surgical site infection: a systematic review and meta-analysis

Guidelines from the World Health Organization strongly recommend the use of a high fraction of inspired oxygen (FiO2) in adult patients undergoing general anesthesia to reduce surgical site infection (SSI). However, previous meta-analyses reported inconsistent results. We aimed to address this controversy by focusing specifically on abdominal surgery with relatively high risk of SSI. Medline, EMBASE, and Cochrane CENTRAL databases were searched. Randomized trials of abdominal surgery comparing high to low perioperative FiO2 were included, given that the incidence of SSI was reported as an outcome. Meta-analyses of risk ratios (RR) were performed using a fixed effects model. Subgroup analysis and meta-regression were employed to explore sources of heterogeneity. We included 27 trials involving 15977 patients. The use of high FiO2 significantly reduced the incidence of SSI (n = 27, risk ratio (RR): 0.87; 95% confidence interval (CI): 0.79, 0.95; I2 = 49%, Z = 3.05). Trial sequential analysis (TSA) revealed that z-curve crossed the trial sequential boundary and data are sufficient. This finding held true for the subgroup of emergency operations (n = 2, RR: 0.54; 95% CI: 0.35, 0.84; I2 = 0%, Z = 2.75), procedures using air as carrier gas (n = 9, RR: 0.79; 95% CI: 0.69, 0.91; I2 = 60%, Z = 3.26), and when a high level of FiO2 was maintained for a postoperative 6 h or more (n = 9, RR: 0.68; 95% CI: 0.56, 0.83; I2 = 46%, Z = 3.83). Meta-regression revealed no significant interaction between SSI with any covariates including age, sex, body-mass index, diabetes mellitus, duration of surgery, and smoking. Quality of evidence was assessed to be moderate to very low. Our pooled analysis revealed that the application of high FiO2 reduced the incidence of SSI after abdominal operations. Although TSA demonstrated sufficient data and cumulative analysis crossed the TSA boundary, our results should be interpreted cautiously given the low quality of evidence. Registration: https://www.crd.york.ac.uk/prospero (CRD42022369212) on October 2022.


SSI
Surgical site infection TSA Trial sequential analysis WHO World Health Organization Millions of patients undergo surgery under general anesthesia each year 1 .Anesthesiologists optimize ventilator setting to reduce postoperative morbidity and mortality and the fraction of inspired oxygen (FiO 2 ) is one of the key settings.Surgical site infection (SSI) is a common and serious complication after abdominal surgery 2 .World Health Organization (WHO) implemented guidelines recommending the use of high fraction of inspired oxygen (FiO 2 ) during the perioperative period to reduce the risk of surgical site infection (SSI) 3 .Poor quality of evidence and failure to address potential harms 4 led to an updated analysis 5 , limiting the intervention to adult patients undergoing general anesthesia.Currently, the US Centers for Disease Control and Prevention (CDC) and other healthcare organizations have adopted the revised WHO standards 6,7 .The consensus is that administering high FiO 2 and thereby increasing tissue oxygen tension 8 could lower SSI by mechanisms such as facilitating neutrophil bacterial killing 9 .Additional benefits such as the reduction of postoperative nausea and vomiting have been debated 10,11 .Nonetheless, concerns still prevail.The current recommendations yet undermine the known adverse effects of high FiO 2 -among the notable drawbacks lies absorptive atelectasis 12 , which potentially results in decreased lung compliance and impaired oxygenation, as well as pneumonia 13 .It is further hypothesized that FiO 2 above physiological range imposes systematic oxidative stress; this may lead to respiratory and cardiovascular complications 13,14 , neurological manifestations 15 , and death.Acutely ill patients fare better with conservative oxygen therapy 16,17 , accordingly, high FiO 2 is not recommended in emergency operations or for critical care patients 18,19 .
There have been attempts to tackle this debate.While older studies focused on the prevention of either SSI or postoperative nausea and vomiting 20 , recent meta-analyses reported diverse clinical effects such as length of hospital stay and mortality [21][22][23] .Despite the increase in scope, previous reviews have neglected to analyze the substantial differences arising from surgical or patient characteristics.For instance, limiting included trials to those operated under general anesthesia prohibited the exclusion of the effect of general anesthetics on wound infection.General anesthesia corresponded to a higher risk of SSI 24 .Furthermore, open surgical approach and emergency operations 25 , male sex, and length of the procedure 26 were documented as independent risk factors of SSI.
We believe that surgery type is also a critical factor associated with SSI incidence and that different types should be investigated separately.Previous trials of abdominal surgery indicated a high incidence of 15-25% 27,28 .SSI remains an unrelenting source of morbidity for colorectal operations with an incidence rate of 9.34% 29 , or pancreatoduodenectomy with an incidence of 6-17% 30 .A recent trial of emergency abdominal surgery showed that perioperative administration of 80% FiO 2 significantly decreased SSI incidence 31 .In addition, previous randomized trials reported the significant positive effect of high FiO 2 on SSI in abdominal surgeries 32,33 .However, a meta-analysis focusing solely on abdominal surgery has rarely been conducted to the best of our knowledge.' Therefore, we carried out a meta-analysis on the effect of high FiO 2 on SSI and other clinical outcomes focusing only on abdominal surgeries; we aimed to discover causes of heterogeneity unexposed by previous reviews by performing subgroup analysis or meta-regression for the important risk factors previously reported.

Methods
Following the registration on PROSPERO (https:// www.crd.york.ac.uk/ prosp ero, registration number: CRD42022369212) on October 2022, the present study was conducted according to the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions.The reporting of this review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement guidelines.
A search strategy was developed and applied to Medline, Embase, and Cochrane CENTRAL, and is provided in the protocol (Supplemental Text S1).The inclusion and exclusion criteria of our review was reported in Supplemental Text S2.The last search was executed on October 21st, 2022.Only full-text articles in English were considered as candidates.
Two authors (KJH and WHK) independently screened titles and abstracts for relevant trials.The full text of manuscripts that passed the first level of screening was scrutinized to determine eligibility.Studies were qualified for inclusion if they: (1) mentioned surgical site infection (SSI) incidence following abdominal surgery; (2) compared high FiO 2 of at least 80% with standard levels of at most 40%; and (3) were randomized controlled trials.The references of previous meta-analyses and related articles were manually inspected to incorporate any studies omitted in the original search.Any disagreements were resolved via discussion.
We piloted a standardized extraction form which one author (JHK) filled out and another (WHK) confirmed.Data was reexamined several times to amend any remaining errors.The following data were extracted: study design, inclusion and exclusion criteria, surgery type, the urgency of surgery, open or laparoscopic methods, definition of SSI, size of groups, demographics (age; sex; body-mass index (BMI); smoking; American Society of Anesthesiologists physical status classification; history of hypertension, diabetes mellitus (DM), coronary artery disease, myocardial injury, chronic pulmonary diseases, cerebrovascular accidents; preoperative hemoglobin; and blood glucose levels), intraoperative and perioperative parameters (percentage of acute operations and transfusions; duration of surgery; and estimated blood loss), and outcomes.
The incidence of SSI was evaluated as the primary outcome, which was defined by either the CDC guidelines 6 or the ASEPSIS scoring system 34 .Secondary outcomes were as follows: short-term mortality, myocardial injury, atelectasis, organ-space SSI, anastomotic leakage, pneumonia, and reoperation.
Data synthesis and analysis.Review Manager 5.4 software (RevMan, The Cochrane Collaboration, Oxford, United Kingdom) and STATA/SE version 14.0 (StataCorp, College Station, Texas, USA) were utilized for data synthesis and analysis.Binary data were analyzed via the Mantel-Haenszel method with a fixed-effects approach to calculate pooled risk ratios (RR) and 95% confidence intervals (CI), while the Inverse Variance method was applied to measure mean difference (MD) and 95% CI from continuous variables.Fixed effect model assumes one true effect size underlies all the studies in the meta-analysis.However, we also performed random effects approach to allow for the expected heterogeneity across the studies.The results of meta-analysis were depicted by forest plots for the primary outcome and hospital length of stay.
Subgroup analyses were conducted according to demographic data and surgical characteristics based on a priori analysis plan.Moderators were considered for analysis if they were known risk factors of SSI or if their clinical significance was acknowledged by all authors.Only those reported in a sufficient number of trials (at least 18) were ultimately selected, including mean BMI, percentage of patients with DM, urgency of surgery, type of surgery, carrier gas, durationof postoperative oxygenation supplementation, and percentage of current smokers.Continuous data such as mean BMI and duration of oxygen supplementation were converted into dichotomous categories.
Meta-regression was performed to assess the relationship between continuous covariates and our primary outcome.We applied criteria identical as mentioned above, resulting in age, percentage of males, mean BMI, percentage of patients with DM, duration of surgery, and percentage of smokers.Results are reported in bubble plots and corresponding p values.Publication bias was primarily assessed with funnel plots.Egger's linear regression test was additionally performed for outcomes reported in 10 or more trials.
We performed trial sequential analysis (TSA) with TSA Viewer (Version 0.9.5.10 Beta, Copenhagen Trial Unit, 2016, Copenhagen, Denmark) for SSI, length of hospital stay, short-term mortality, and myocardial injury.It conducts a cumulative meta-analysis, manifested as a Z curve of the pooled observed effect, that reduces the risk of false-positive results from repetitive testing.We used a power of 80% and a 5% alpha error to calculate the required information size, which signifies the threshold for which the effect of the intervention may be confirmed or rejected.A conventional boundary denoting statistical significance (p < 0.05) and the trial sequential boundary (O'Brien-Fleming significance boundary) were also created.Overall, the course of the z-curve in relation to these borders helped us estimate when the effect will be large enough for future studies to be unnecessary.We used a 20% relative risk reduction (RRR) for binary outcomes, except for short-term mortality for which we used 10% RRR considering its clinical importance.For the continuous outcome (length of hospital stay), a mean difference of 0.5 was used.
A sensitivity analysis applying a random-effects approach was carried out to dismiss concerns regarding disparities among operation types and study designs, which may have breached assumptions necessary to undertake a fixed-effects approach.Another sensitivity analysis was done to compare trials with different definitions for SSI, namely per CDC guidelines, per ASEPSIS scoring system, and others.Heterogeneity was represented as the I 2 statistic.We examined potential sources of heterogeneity via subgroup analyses and meta-regression.

Results
We identified a total of 58,576 manuscripts from the initial search of which 13,279 duplicates were removed.After excluding 44,300 irrelevant studies, full texts of the remaining 997 studies were further inspected for eligibility.A total of 970 articles were removed for the reasons shown in Fig. 1.After evaluating trials additionally identified in references (n = 0), a sum of 27 original records comprising 15,977 patients was included in the final analysis.
Most studies showed an overall intermediate risk of bias, except for two determined to be at low risk 41,49 , and two at high risk 39,56 (Supplemental Figure S1).GRADE approach showed that the quality of evidence of our study outcomes is from moderate to very low (Supplemental Table S3).
Regarding organ-space SSI, six trials 31,37,44,47,49,58 including a sum of 2095 patients reported data on the subcategories of SSI, namely superficial, deep, and organ-space.Among these, one trial 58 documented a combination of deep and organ-space SSI, and another only noted deep SSI 31 .Following our decision to assess the incidence of organ-space SSI as a secondary outcome, these two trials were not included in the final analysis.There was no remarkable difference between the two groups (RR: 0.85; 95% CI: 0.57, 1.27, p = 0.43, I 2 = 0%) (Table 2).The funnel plot did not suggest publication bias.

Discussion
Our meta-analysis demonstrated a potential benefit of a FiO 2 of 80% compared to 30% or lower in reducing SSI in abdominal surgeries.TSA revealed that the required sample size was reached and the cumulative z-curve crossed the trial sequential boundary of the preference for high FiO 2 .A sensitivity analysis of the definition of SSI did not alter the results.However, the low quality of evidence still prevents a firm conclusion.
Whilst evidence for WHO guidelines are disputed, the beneficial effect of high FiO 2 on SSI reduction continued to be found in recent studies, albeit in limited populations.It has been previously demonstrated for a subgroup of intubated patients 23 , or trials on a majority of emergency operations 21 , and when a stringent value of 80% versus 30% oxygen was applied 21 .This finding was also supported by a previous meta-analysis when inadequately or ambiguously blinded trials were excluded 61 .
Wound infections are usually established in a "decisive period" of several hours in the postoperative period 9 , during which host defense vigorously functions to remove pathogens, starting with neutrophils.Their key mechanism is the generation of antimicrobial reactive oxidant species 9,42 .In vitro studies have shown that neutrophil oxygen consumption and hence its production of oxidants are hampered at low oxygen tension 8 , which is alleviated by the administration of high levels of oxygen 62 .Therefore, it has been hypothesized that high FiO 2 activates the antimicrobial mechanism of neutrophils and thereby decreases SSI.This theory was true in an animal study which showed that neutrophils in the 80% FiO 2 group were more activated than those in the 30% group, although no variance was found in infiltration amounts 63 .Nonetheless, clinical trials have yielded mixed results due to various confounding factors of the clinical practice.However, the significant result of our subgroup analysis of supplied oxygen for at least postoperative six hours supports the importance of the decisive period.
Our results contrast with the cumulative outcomes of recent meta-analyses that found no significant beneficial or harmful effect of high FiO 2 21-23,61,64,65   .The different result of our meta-analysis is primarily due to the inclusion of only abdominal surgeries while previous meta-analyses included any types of surgeries.Three previous metaanalyses included randomized trials of non-cardiac or any type of surgeries and reported no significant results for SSI 21,61,65 .The other meta-analyses analyzed mortality, length of hospital stay, and the incidence of cardiovascular or respiratory complication but reported no significant difference 22,23,64 , which was consistent with our results.The significant results of our meta-analysis for SSI may be, in part, attributed to certain characteristics specific to abdominal surgery.One possible explanation is a marked variance in pathogens.While Staphylococcus aureus is the most common microorganism causing SSI 66 , Escherichia coli is the most predominant in SSI following abdominal surgery 67 .S. aureus is a notorious foe to neutrophils, armed with mechanisms for evasion such as chemotaxis inhibitory protein 68 and extracellular fibrinogen binding proteins that block complement activation 69 .E. coli is armed with other defense mechanisms such as lipopolysaccharides.As such, discrepancies in bacterial characteristics interact with neutrophil activity differently; perhaps those that infest abdominal wounds are more susceptible to neutrophils, which, in turn, may enhance the effect of high FiO 2 .
The high FiO 2 was also associated with a significantly lower incidence of anastomotic leakage.Ischemia in surrounding tissue is essential for anastomotic leakage development, leading to delayed wound healing, necrosis, and dehiscence 70 .Application of high FiO 2 increases tissue oxygen levels and may thereby prevent the formation of anastomotic leakage.No association was found for organ-space SSI; there is yet no evidence for a beneficial effect of oxygen on severe SSI.Previous studies and ours alike showed no association between FiO 2 and poor clinical outcomes including mortality, myocardial injury, reoperation, pneumonia, and in length of hospital stay [21][22][23] .
However, the adverse effect of high FiO 2 should be acknowledged.Given that one hundred percent oxygen is known to induce absorption atelectasis even if administered for a short time 71 and that another meta-analysis demonstrated detrimental effects of high FiO 2 on oxygen parameters and severity of atelectasis 72 , concerns about pulmonary function cannot be discarded.Atelectasis is not causative of postoperative pneumonia 73 and must be dealt with as an independent outcome.
TSA for SSI revealed that we have reached the required information size and the O'Brien-Fleming boundary had been crossed.Nonetheless, the quality of evidence was judged low; robust evidence is still lacking, and our results are subject to change.Secondary adverse outcomes of high FiO 2 have yet to reach the required information size.www.nature.com/scientificreports/A subgroup analysis showed that high FiO 2 is beneficial only when the mean BMI was below 30.This result lacks sufficient power as revealed by TSA and is contrary to our expectation; obesity is a known risk factor for surgical site infection 74 , especially for colorectal surgery 75 .Hypoperfusion of adipose tissue in obese patients delays wound healing and forms dead space, predisposing patients to SSI 76 .Additionally, poor tissue oxygenation makes it difficult for prophylactic antibiotics to reach sufficient concentrations.
We found high FiO 2 to be beneficial in emergency operations, consistent with the subgroup analysis of a previous meta-analysis 21 yet contrasting with another study on acutely ill adults 16 and a third inconclusive study 77 .These discrepancies may be due to insufficient power and heterogeneity in included surgery types.The previous meta-analysis 21 was based on three trials of 509 patients and our analysis was on only two trials with 388 patients.Studies on acute appendicitis took up the majority with a short operation time duration and a predominantly laparoscopic approach.
Subgroup analysis comparing nitrous oxide to air showed that FiO 2 decreased SSI in only the latter.While a previous meta-analysis claimed that usage of nitrous oxide does not significantly alter SSI rates 78 , its assertion may lack power as it was based on six trials with high heterogeneity.A randomized trial published afterward also found no association between N 2 O and SSI 79 .We believe that N 2 O may potentially serve as a confounding factor; it is known to inhibit methionine production, which leads to a reduction in protein expression which in turn deters the healing process 80 .It is also known to depress chemotactic migration 81 and inhibit methionine synthase 82 .
Subgroup analysis revealed that the beneficial effect of high FiO 2 on SSI was consistent to a subgroup of colorectal surgeries and other abdominal surgeries, but not to trials of Cesarean sections.Caesarean sections are commonly operated under regional anesthesia, which was the case for all trials included in our analysis.It is also a relatively safe procedure with a low SSI rate, hence the effect of FiO 2 may have been statistically trivial even if favorable.
High FiO 2 appeared to be useful only when the duration of exposure was 6 h or longer.This starkly contrasts with the current WHO guidelines, which advise oxygen administration for 2-6 h postoperatively 5 .The suggested duration is not based on physiological evidence and is rather attributable to additional factors such as resource use.As far as we know, the optimal duration of exposure has not been meticulously studied to date.The decisive period for oxygen to benefit a patient is unknown; and though a need for a direct comparison between exposure duration has been suggested 33 , we found no trials delving into this issue.Further trials comparing the effect of oxygenation for different durations are required to better understand the effect of oxygen on SSI reduction.
Our meta-analysis provided timely analysis, including recently published studies 31,43 and a trial previously left out for unknown causes 55 , as well as re-incorporating studies that have been omitted from conservative metaanalyses due to concerns of authors with retracted articles [52][53][54] .Despite several previous meta-analyses on broader classes of surgery, an inspection of abdominal surgery has not been conducted.It has been analyzed as a subgroup analysis with insignificant results 21 .Including recently published trials and trials with regional anesthesia as well have resulted in a contrasting conclusion.We also performed a subgroup analysis on diabetes mellitus and BMI, well-known risk factors of SSI.Our findings that diabetes and obesity may hinder or counterbalance the positive effects of oxygen may contribute to stratifying the intervention to relevant patients in future operations.While concerns have been raised on the vast range of postoperative oxygenation duration 16 , no analysis has been done on the topic; we found a significant benefit of oxygenation administered for at least 6 h.Heterogeneous definitions of the endpoint were also brought up as a source of potential imprecision, yet our sensitivity analysis showed that trials on CDC or ASEPSIS definitions benefited from the intervention alike.A final sensitivity analysis including only the population under general anesthesia with tracheal intubation as indicated by WHO guidelines revealed that high FiO 2 does indeed reduce SSI.This finding effectively diminishes the risks of imprecision arising from our attempt to include more studies than previous reviews.Meta-regression found that age, mean BMI, percentage of patients with DM, and percentage of smokers did not significantly alter our results.Male sex is a disputed risk factor; a recently published multicenter study found its effect non-significant 83 .
There are several important limitations in our study.Firstly, the quality of evidence is low.Only two 41,49 included trials were deemed at low risk of bias, while two trials judged to be at high risk 39,56 impose concerns on the possible bias.We rated high-risk trials due to the concerns with deviations from intended interventions 39 and outcome measurement 56 .Regarding ethical consideration, the studies by Schietroma et al. should be dealt with caution due to questionable methodology based on some retractions 60 .Some of other studies by Schietroma et al. were retracted, although the studies included in our analyses were not retracted or are under investigation 60 .Our sensitivity analysis removing the works of Schietroma yielded no significant effect of FiO 2 [52][53][54] .Secondly, heterogeneity among data is also a downfall.The I 2 value for the main analysis and notable subgroup analyses was high.We attempted to explore major potential sources of heterogeneity with subgroup analyses and meta-regression; unresolved factors may include variations in anesthetic regimes and protocols of prophylactic antibiotics.Aspects of the study population such as comorbidities or selectioncriteria may also have played a role.Study settings and the baseline quality of performance of each research center were also discrepant.As surgical and anesthetic techniques have evolved, older trials may be outdated; the year of inclusion showed indeed a significant influence 21 .Variance in sample size is another potential source of heterogeneity, as was found in the aforementioned study.
In conclusion, this meta-analysis found that high FiO 2 reduced the incidence of SSI and anastomotic leakage after abdominal surgery, a viewpoint distinct from the current consensus that perioperative high FiO 2 does not benefit patients.This difference may be attributable to our focus on solely abdominal operations.Taking together the variance in subgroups, we suggest that the administration of high FiO 2 should be individualized based on patient and surgery characteristics rather than being standardized.Based on our significant subgroup analyses for colorectal surgery and other abdominal surgeries but not for Caesarean section, further studies for these specific types of surgeries are required.The TSA for SSI revealed that the cumulative analysis crossed the trial

Figure 3 .
Figure 3. Trial sequential analysis for surgical site infection.Pc = Probability in the control group, RRR = relative risk reduction, a = alpha error, b = beta-error.The blue line means the cumulative z-score curve.The boundaries favoring high or low FiO2 or area of futility are shown in red lines. https://doi.org/10.1038/s41598-023-41300-4

Table 1 .
Baseline characteristics of the included trials.Data are presented as mean (SD), median (1st-3rd quartile), or median [range].FiO 2 = inspired oxygen concentration, NR = not reported.*5-95% percentile.General anesthesia was assumed when sevoflurane or isoflurane were used, or based on standard technique for the type of surgery in question.

Table 3 .
Summary of the results of subgroup analysis for surgical site infection.FiO 2 = inspired oxygen concentration, 95% CI = 95% confidence interval.