Oxygen Therapy during Surgery: Decisions Need Reliable Data

In 2016, the World Health Organization (Geneva, Switzerland) recommended that adult patients undergoing general anesthesia with tracheal intubation for surgical procedures receive a fraction of inspired oxygen (Fio₂) of 0.80 during surgery to reduce the risk of surgical site infection.  Studies from the United States and other parts of the world suggest the majority of anesthesiologists are ignoring these guidelines and instead typically use Fio₂ 0.4 to 0.5.  The reasons for this disconnect are largely unknown.

In this month’s issue of Anesthesiology, Billings et al. undertook a large retrospective cohort study to describe the patterns of oxygen administration among patients attending medical centers across the United States.  In contrast to previous retrospective analyses that focused on associations between intraoperative oxygen exposure and outcomes, this study was designed to assess the patient, provider, and institutional determinants of oxygen administration. The main finding of interest was not only that very few U.S. anesthesiologists follow the World Health Organization recommendation, but that patient and perioperative factors could not explain more than a small proportion of the variability. Although seemingly simple and perhaps not surprising, this study raises very important questions about the most prevalent drug we use in anesthetic practice, and why World Health Organization and other institutional guidelines are ignored. These deeper questions are worthy of serious consideration.

Billings et al. analyzed intraoperative data from the Multicenter Perioperative Outcomes Group patient registry, including adult patients who underwent prolonged (more than 120 min) surgery during general anesthesia with tracheal intubation during a period of 3 yr from January 2016 across the United States. After exclusion criteria were applied, 367,841 patient events administered by a total of 3,836 anesthesiologists and 15,381 in-room anesthesia providers from 42 medical centers were analyzed. The median (interquartile range) Fio₂ among all cases was 0.55 (95% CI, 0.48 to 0.61). Overall, only 8.7% of patients received a median ± SD Fio₂ of 0.8 ± 0.05. Hence, these data and those from other studies across different jurisdiction provide compelling evidence that a large majority of anesthesiologists around the world do not administer intraoperative oxygen in accordance with the 2016 World Health Organization recommendations, and that most anesthesiologists tend toward administering an intermediate level of intraoperative Fio₂. It is also worth noting that the median Fio₂ values reported by Billings et al. and other similar studies inherently and systematically overestimate the Fio₂ administered during the maintenance phase of anesthesia due to the periods of higher Fio₂ conventionally used during induction and emergence phases of anesthesia. The maintenance phase represents the period of anesthesia care when the induction and emergence preoxygenation with high Fio₂ is typically not used and may better reflect the intent of the anesthesia provider with regard to intraoperative oxygen therapy.

Importantly, Billings et al. then explored the degree of variability of oxygen administration, undertaking multivariable linear mixed-effects regression analyses to determine associations between a range of patient, procedure, clinician, and medical center factors on the level of oxygen administered to patients. The medical center at which anesthesia was administered was reported to be the strongest determinant of Fio₂ administered, accounting for 23.3% of the observed variability in Fio₂. This was followed by in-room anesthesia provider (8.8%), anesthesiologists (7.7%), procedure factors (4.4%), and lastly patient factors (3.5%). Most strikingly, 53% of the variability in intraoperative Fio₂ administration was not accounted for by any measured factor. The authors have noted the possibility that one or more important factors determining intraoperative oxygen administration may not have been included in their analysis. While this is possible, a previous survey of anesthesiologists that asked participants to identify indications for a specific intraoperative oxygen regimen did not identify any specific indications that were not included in the analysis by Billings et al. Indeed, in addition to the large number of participant data sets, a key strength of this study was the inclusion of a large number of baseline covariates relating in particular to patient and procedural factors. Therefore, the more probable cause for the unexplained variability in intraoperative oxygen administration is that it has a large arbitrary component that is unguided and non–patient-centered. That is, the chosen dose (here, Fio₂) of the most essential and prevalent drug we use in our practice is basically idiosyncratic.

Many might argue that an Fio₂ of 0.50 to 0.60 during surgery adds a degree of safety and might still reduce the risk of infections (so sufficiently compliant with the World Health Organization guideline). However, higher concentrations of oxygen administration during surgery or in critical care have been associated with organ injury and death. Furthermore, the 2022 updated surgical site infection prevention guidelines sponsored by the Society for Healthcare Epidemiology of America (Arlington, Virginia) in collaboration with the Infectious Diseases Society of America (Arlington, Virginia), the American Hospital Association (Washington, D.C.), and The Joint Commission (Oakbrook Terrace, Illinois) considered the role of tissue oxygenation “unresolved,” highlighting contradictory evidence from the most recent systematic reviews.  Other concerns regarding the veracity and totality of the evidence have been highlighted.

The finding of unexplained variability in oxygen administration during general anesthesia is consistent with the current relative lack of high-quality randomized clinical trials that could inform the development of consistent perioperative oxygen guidelines; as a result, the current recommendations from international organizations are contradictory. For the individual anesthesia provider, there is uncertainty both about whether the amount of oxygen that is administered to patients undergoing general anesthesia during surgery is important for patient recovery, and if it is considered to be important, what level of oxygen is optimal. Furthermore, as appears to be the case in critically ill adults receiving invasive mechanical ventilation it is possible that the optimal dose of oxygen might vary between individual patients according to pre-existing comorbidities, procedural factors, or intraoperative events.

All of this underscores the need for more work to determine the optimal oxygen concentrations to be administered to surgical patients. Given that the administration of supranormal oxygen is almost universal during anesthesia, the absolute benefit from even a small treatment effect would be substantial. Accordingly, larger and more efficient trials that can be powered to detect small treatment effects are likely to be needed in order to generate the evidence required to not only identify the perioperative oxygen regimen that may be most advantageous for the average patient, but ideally to be able to inform individualized decision-making with regard to oxygen administration to individual patients having surgery during general anesthesia. We submit that there is an ethical imperative to transform current idiosyncratic oxygen therapy practice into randomized practice. Embracing the learning healthcare system paradigm, enrolling nearly all eligible patients into an adaptive randomized clinical trial that adjusts treatment group assignment (here, Fio₂) as outcome data accrue is the ideal way to fuse clinical trial science and quality improvement for the benefit of our patients.