hyperoxic ventilation and SSI

In this months Journal of Anesthesiology, Thibon et al. [1] published a study that was unable to find a benefit to intraop hyperoxia in the reduction of SSI in patients undergoing gyenecologic, breast and abdominal surgery.  The authors also included  their data in a new meta analysis looking at the reduction of SSI with the use of hyperoxia and concluded that with their data, there was no benefit to routine hyperoxic ventilation.  I've posted twice previously on this issue.  See a General overview of the topic and the discussion of the PROXI trial.
The concept of hyperoxic ventilation has been debated a number of years.  Hofp et al. was able to show that wound oxygen tension was a better predictor of SSI than a scoring system.  In this study, patients who had oxygen tension lower than 60 mmHg had much higher infection rates compared to patients whose wound oxygen tension was above 90 mmHg.  Furthermore, it has been shown that hyperoxia can be as effective as antibiotics in infected wounds in rats.  Two well done RCTs also showed a reduction of SSI in colon surgeries.  After this many anesthesiologists began to believe that by providing a high FiO2 they might be preventing SSI.  However, in medicine, things are not always quite as simple as they might seem.  In the PROXI trial they found no benefit to oxygen therapy. Harriet Hopf included an editorial on this study noting in particular that the protocol used for fluid management was questionable in allowing for sufficient volume to ensure adequate transport of the elevated oxygen levels to the peripheral tissues.  This is based on a number of studies that have shown that tissue oxygen tension does not increase in response to elevated FiO2 if not accompanied by euvolemia.
This study, therefore, should address these important issues.  1) Should we expect a difference with oxygen when patients with breast surgery were included (typically low infection rates).  2)  infection rates were only 7.2%, a very low rate making it difficult to detect a difference.  3)  What were the infection rates in the different groups (gynecolgoic vs. breast vs. general)?  Was their a trend to improvement in the subgroups with higher infection rates? 4)  What was the fluid regimen? Was euvolemia ensured? 
This study aknowledged a few important flaws.  First, the study was likely underpowered to detect a difference.  First, the power analysis was completed with an expected infection rate of 12% in the control group, but the infection rate in the control group was only 7.2%.  Second, the recruitment was difficult and the time elapsed prior to enrolling sufficient patients.  The power analysis called for 270 patients per arm (total 540) when the infection rate was expected to be 12%.  The enrolled only 414 patients. 
The authors rightly aknowledge that in the two studies that demonstrated that hyperoxia significantly reduced SSI the protocol required post operative oxygen therapy for 2 hrs or 6 hrs, while this study did not utilize post op oxygen because, "...is not routine practice in France..." It is hard to know how important this is, but there are several reasons to expect it to be important.  First, this was the protocol in place when oxygen did reduce SSI.  Second, intraoperatively, oxygen is delivered via an ETT in mechanically ventilated patients; thus, lung volumes are typically optimized to some degree by the anesthesiologist in such a way that even FiO2 of 30% may achieve adequate PaO2.  However, in the PACU, patients spontaneously breathing, sedated, and not managed as intensively as they are during the course of surgery by an anesthesiology provider.  This allows for the potential for brief episodes of oxygen desaturation, hypoventilation with subsequent atelectasis.  Patient splinting from pain may also reduce lung excursion.  The potential end result is decreased PaO2 if supplemental is not provided.
Infections vary by the type of infection and therefore, oxygen therapy may be beneficial in some types of surgery but not helpful in others.  The authors did not include a breakdown of infections based on type of surgery.  They do note that nearly have of the surgeries included in the trial were "breast" surgery.  However, infection rates vary depending on the type of breast surgery from as high as 12% for implant surgery to as low as 1% for breast reductions.  Simple lumpectomies are also likely very low risk. Oxygen therapy is likely not as beneficial in healthy patients undergoing routine breast surgery.
Lastly, This study looked at infections that occurred up to 30 days post operatively.  Interestingly, the 80% group had a median time of infection diagnosis of 16.9 days post op while the 30% groups had a diagnosis made at 13.9 days.  It is plausible to conclude that after 15 days post op infections at the site of surgery may arise due to factors not related to oxygen tension at the wound during the time of surgery.  While this theory has not been studies direclty, it is certainly likely.  Other well done studies where benefit has been demonstrated had a cut off of 14 days.  Interestingly, using the criteria of 14 days as a cut off would have dramatically altered the results of this study because the diagnosis of SSI would have occurred after the cut off period for a more of the patients in the 80% group.  Said in another way, it would have decreased the capture rate of infection in the 80% compared to the 30% group.
The metanalysis included the study by Pryor et al.  I believe this study should be excluded as the study was flawed.  The authors included sicker and more obese patients in the "high oxygen" group. 

Lastly, so far, all previous studies have compared high FiO2 (i.e. 80%) vs. low FiO2 (30%).  In reality, tissue oxygen tension is better correlated with PaO2 and even then not perfectly so particularly in obese patients.  Studies in populations with infection rates that average higher than 10% should be carried out, and the end points should be PaO2 with wound oxygen tension measurements as well.  Studies show that not all patients given a set FiO2 have the same PaO2. Furthermore, not all patients with the same PaO2 will have the same wound oxygen tension.   Since SSI rates vary based upon wound oxygen tension, it seems hazardous to attempt to draw conclusions about infection rate when varying only delivered oxygen concentration.
Studies looking at SSI are critical as SSI increase healthcare speding tremendously. However, SSI remain a bigger problem in certain subsets of patients undergoing certain types of surgeries.  It is important to determine if we can improve outcome with improving PaO2 (one step beyond FiO2).

1. Thibon P, et al. Anesthesiology. 2012; 117(3): 504.