Hyperbaric oxygen therapy is easily as controversial a topic in Emergency Medicine as any. Many physicians and scientists believe HBOT is an essential treatment for carbon monoxide poisoning, with a goal toward restoring normal intracellular physiology as rapidly as possible. Other skeptics, however, point to the paucity of high-quality evidence in support of a logistically complex and expensive intervention.
This is a retrospective review from Taiwan evaluating outcomes of 25,737 patients recorded in their national health database as suffering from carbon monoxide poisoning. Of these, 7,278 patients received HBOT while the remaining 18,459. There were many significant and relevant differences between cohorts, with those not receiving HBOT tending to be older and have more medical comorbidities. On this substrate, unsurprisingly, the authors find a survival advantage – persisting through multivariate statistical adjustment – to receiving HBOT, with an adjusted hazard ratio of 0.74 (95% CI 0.67-0.81).
Despite the size of their sample, it is unlikely these data reflect a true treatment effect from HBOT. In a retrospective cohort such as this, the pervasive differences between groups almost certainly suggests confounding features influencing treatment decisions. Off the limited structured data recorded in this database, it is unlikely any statistical adjustment or matching technique will provide a better reliable estimate of any true mortality benefit – nor is a mortality benefit one of the expected outcomes of HBOT.
The authors also spend some time reporting the survival advantages associated with receiving HBOT more than once over the first month following the poisoning event. These positive findings are, effectively, the definition of survivorship bias – mortality directly affects the ability to receive multiple treatments. You can’t dive the dead, of course, so simply surviving to undergo additional treatments is erroneously associated with a benefit.
The authors eventually state “The results provide important references for decision making in the treatment of COP” – but, unfortunately, they tell us very little. The level of evidence supporting or refuting treatment with HBOT remains poor until an RCT of sufficient scale can be performed.
“Hyperbaric oxygen therapy is associated with lower short- and long-term mortality in patients with carbon monoxide poisoning”
https://www.ncbi.nlm.nih.gov/pubmed/28427969
So the reduction in mortality seems too good to be true. And no doubt there was selection bias. But what do you make of the fact that benefit from HBOT remained after results were adjusted for “age, sex, underlying comorbidities, monthly income, and concomitant conditions.” And those with respiratory failure had the most benefit, meaning more severely ill patients got the treatment? It seems that would make the results tend toward the null or worse. An RCT in the NEJM was done in 2002 (which was not looking at mortality) but was stopped early due to markedly improved cognitive outcome in those receiving HBOT. http://pmid.us/12362006
So this Taiwanese retrospective approach isn’t the study to say HBOT definitely improves mortality, but should we just toss this out as low quality drivel quite so quickly? It deserves more consideration than that, in my opinion.
I don’t discount these data, but I caution against assigning it a higher level of evidence than what it is: a retrospective cohort, not even from a registry designed to reflect any aspect of the individual patient encounter: vital signs, GCS, CO level, etc. I don’t appreciate any substantial value from their statistical judgment due to the limitations of their underlying data set.
Regarding the NEJM HBOT RCT, it has been the subject of enough criticism in other forums not to reproduce it here. To say the least, the most recent Cochrane review equivocated on the strength of the evidence in the same manner as me: https://www.ncbi.nlm.nih.gov/pubmed/21491385/