Perioperative oxygen and surgical site infection.

In November 2016 the WHO produced guidelines on the prevention of surgical site infection. Along with some very sensible recommendations about when to give antibiotics and whether patients should have a bath before surgery there was a section on how much oxygen we anaesthetists should administer to patients during and immediately after an operation. (Section 4.12 in the guidelines below).

The guidelines state that the group looked at the evidence and then performed their own meta-analysis, and a further subgroup analysis. Some of you may already know that meta-analyses are one of my personal hobby horses. Every trial has its own biases and then we bunch a load of them together in order to concentrate and distill these biases, then do a subgroup analysis to focus on the areas where those biases may have taken the trial results to statistical significance. I can name you many subjects where a meta-analysis has changed our practice only to be disproven a few years later by a well conducted trial. 

I sent these guidelines out to my consultant colleagues and not one supported it. In fact one of my colleagues who is also a medical statistician produced an excellent piece of work summarised here:

They answered the wrong question, looking only at infection rates and ignoring overall benefit or harm to the patient. In a Cochrane analysis Here the chance of high oxygen concentrations reducing surgical site infection was described as low, and in addition there could be a 20% increase in mortality for those patients with high oxygen concentrations. 

So essentially high oxygen concentration may reduce SSIs a little, or it might cause a 20% increase in mortality. Or neither. In any case the evidence is far too flimsy to put this into international guidelines, and there is a chance they may actually cause harm to the patients we’re treating. I won’t be changing my practice based on these guidelines, it’s a shame it’s been included at all as it reduces the impact of the rest of the document. 



  1. I am writing in response to the criticisms raised by anonymous
    I am the Chair of the WHO Guideline Development Group that performed the many tasks associated with generating the guideline on prevention of surgical site infection this blog addresses. The process WHO uses for guideline development is standardized and is described in various publications findable on PubMed or the WHO website.
    There are several inaccurate statements in this post.

    The author incorrectly describes the purpose of a meta-analysis. The purpose is to combine results of various comparative studies to arrive at a broader statistical assessment. Each trial included, particularly in clinical medicine, represents a small sample of a far larger population. Trials use different entry criteria, somewhat different methods, and occur in different health care environments. By combining studies, a more generalizable statistical analysis is provided.
    Bias is managed separately in a Risk of Bias table derived from by the Cochrane Collaboration and GRADE. We would appreciate the author identifying those meta-analyses he mentions that have changed practice only to be disproved in a single trial. The assessments made through GRADE provide an assessment of the level of evidence, and it would be truly remarkable that a well done meta-analysis would change its conclusion based upon the addition of a single study.
    As regards the statement that we answered the wrong question, the author is simply wrong. Firstly, the purpose of the systematic reviews and meta-analyses was to support recommendations to lower surgical site infection rates. This was therefore used as the primary outcome measure. The statement that the 25 experienced surgeons, infection preventionists, guideline methodologists, and GRADE methodologies should have used some other outcome as the PRIMARY OUTCOME is simply wrong.
    We recorded all outcome measures reported in the reviewed studies. There was no change in mortality in the combined studies, and no study showing a change in mortality. The author of this blog would do a considerable service by providing us the references to those studies supporting the claim. I assume the author has created some kind of a confidence interval that reaches 20% for mortality. This is simply not true. These assessments are provided in the Narrative Summary, which provides the methods, findings, and statistical analyses in detail. This is available on the WHO website.
    In that summary, we stated that none of the clinical trials investigating the administration of 80% FiO2 showed a difference in pulmonary complications or other adverse events, including death.

    The problem with blog posts is that typically rather sweeping statements are made but no information, data, references, or the bases of support are provided. We hope the blog post authors will forward this information to us.


    • There are a number of issues with the guidance.

      The research question only looked at SSI and not overall outcomes for high FiO2. Even so, the meta-analysis, as planned, showed NO benefit until some studies were excluded. The “strong” guidance is actually based on a (weakly) significant subgroup analysis which was not prespecified in the group’s methods.

      The trials, generally, did not specify reliable outcome measures (such as 30 day mortality) in their methods. It is not possible to use these trials for assessing benefit vs harm. The blase dismissal of the increased long term mortality shown in the PROXI followup trial suggests the group were not looking very hard for evidence of harm.

      The guidance is at variance with what was actually studied. When one looks at the individual studies it is clear that the comparators for 80% O2 included O2/N2O (3/15 trials) and a control group allowed to have SpO2 down to 92% (another 3/15 trials). Further, many of the studies excluded “high risk” patients including stable diabetics which restricts the applicability of the guidance.


  2. Thank you for your comment – I will respond in full, but the inference in your comment that the guidelines focus on reducing SSI rate while not taking account of overall risk/benefit is troubling.

    In addition I don’t understand how the recommendation is only focused on patients with an endotracheal tube. There is no mention of patients with an LMA for instance. I don’t think this is a particularly anonymous blog post – it wouldn’t take an investigative journalist to find out where it comes from. My “sweeping statements” are based on the Cochrane review quote, and I’d be interested in how you reconcile what you say with the content of that review.

    My point is that we don’t have enough evidence to make these recommendations yet, especially with emerging evidence of overall harm of hyperoxia.


  3. The premise is old, the argument is fresh, but take a broader outlook, and the recommendations are not so bad. By making these recommendations, WHO ensures at least that plenty of Oxygen be made available, a facility that might be a luxury in some settings.
    But making a strong guideline on an issue that we are hardly sure about shows WHO committee’s low confidence ( sometimes rightly so) on anesthesiologists to provide oxygen during surgeries.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s