Friday, May 31, 2013

Over Easy? Trials of Prone Positioning in ARDS

Published May 20 in the  NEJM to coincide with the ATS meeting is the (latest) Guerin et al study of Prone Positioning in ARDS.  The editorialist was impressed.  He thinks that we should start proning patients similar to those in the study.  Indeed, the study results are impressive:  a 16.8% absolute reduction in mortality between the study groups with a corresponding P-value of less than 0.001.  But before we switch our tastes from sunny side up to over easy (or in some cases, over hard - referred to as the "turn of death" in ICU vernacular) we should consider some general principles as well as about a decade of other studies of prone positioning in ARDS.

First, a general principle:  regression to the mean.  Few, if any, therapies in critical care (or in medicine in general) confer a mortality benefit this large.  I refer the reader (again) to our study of delta inflation which tabulated over 30 critical care trials in the top 5 medical journals over 10 years and showed that few critical care trials show mortality deltas (absolute mortality differences) greater than 10%.   Almost all those that do are later refuted.  Indeed it was our conclusion that searching for deltas greater than or equal to 10% is akin to a fool's errand, so unlikely is the probability of finding such a difference.  Jimmy T. Sylvester, my attending at JHH in late 2001 had already recognized this.  When the now infamous sentinel trail of intensive insulin therapy (IIT) was published, we discussed it at our ICU pre-rounds lecture and he said something like "Either these data are faked, or this is revolutionary."  We now know that there was no revolution (although many ICUs continue to practice as if there had been one).  He could have just as easily said that this is an anomaly that will regress to the mean, that there is inherent bias in this study, or that "trials stopped early for benefit...."

But in the case of the current prone positioning study, we may be smitten by the very small P-value, which means literally that "if the null hypothesis is true and there is in reality no mortality difference between prone and supine, the probability of finding a difference as large or larger than the result the authors found is less than 0.1%" - a very small probability indeed.  So the difference is not only large, but statistically robust.  Have we seen this before?

In 2001 the PROWESS trial of drotrecogin-alfa for severe sepsis was published showing that the agent conferred a 6.1% mortality benefit with a P-value of 0.005 - a largish and statistically robust effect.  As is well known (for now - this dialectic will soon be forgotten), subsequent trials all failed to show benefit and manufacture of the agent was discontinued.  Thus, even a largish benefit can (must?) be dismissed when other data fail to support it.  So, let's inspect the other major trials of prone positioning in ARDS over the last decade or so:
  • The 2001 Gattinoni study of proning in ALI/ARDS showed a mortality difference at 10 days of 3.9% in favor of proning, but with a P-value of 0.41 (my calculation).  Among the many critiques of the study were that the patients were not sick enough and were not proned for enough hours each day and for enough days.
  • If being less selective in your enrollment criteria doesn't work, you can always try to be more selective.  In answer to the critiques of the Gattinoni study, namely that the patients weren't sick enough or proned for long enough, Mancebo et al reportedin 2006 a study of prolonged proning in severe ARDS.  Intensive care unit mortality (the primary endpoint) was 43% in the prone group and 58% in the supine group.  (I just realized that I cannot reproduce these numbers with the data reported.  I get mortality to be 41.6% and 56.6% with a P-value of 0.08.  There must be some adjustment somewhere that was not transparently reported in the Methods or the Results.)  I remember being impressed by this study when it came out.  There is a large and almost statistically significant benefit of proning in this study.
  • In yet another response to the critiques and limitations of prior trials, Taccone et al (JAMA, 2009) reported the results of the Prone-Supine II (PSII) trial.  In the overall cohort, mortality was 31% for prone and 32.8% for supine, P-value 0.72.  Is anyone else seeing a pattern here?  That's a rhetorical question, and the answer depends on who's answering it.  The true-believers see the subset of patients with severe hypoxemia in Table 2 of the article where the mortality difference is in favor of prone by about 8% with a P-value of 0.31.  The disinterested among us see the overall negative results in trial after trial after trial.
That is, until the current study.  As I said, the results are indeed impressive.  But I cannot look at this trial without peering through it to the stack of reports that preceded it.  So what's going on beyond the possibilities of regression to the mean or a new panacea for ARDS with twice the effect of low tidal volume ventilation?

These studies are all coming from Europe, specifically Italy, Spain, and France.  One can't help but surmise that the European respirologists have a bit of a fetish for the prone position.  And, like so many other investigators, it is possible that they are not disinterested scientists trying to falsify hypotheses.  They believe it works.  They want it to work.  And they will study it until they find that it works.

In my own practice, which entails very minimal sedation (and even less analgesia - what is the basis for the oft repeated adage that pain control is necessary in all ventilated patients?  From whence the pain?) and VERY aggressive weaning from mechanical ventilation, I will use the prone position very infrequently because I worry that the additional sedation it will doubtless entail (dose totals not reported in the recent trial) compared to my usual practice will prolong the course of illness in order to institute a therapy with uncertain benefits.  My patients will remain sunny side up.

[Addendum 6/2/2013:  Several trusted colleagues have commented that my supposition that sedation will be higher is unfounded.  I agree.  I wanted the post to focus on the evidence, but end with my summary position, namely that I will not immediately adopt this practice for routine care, and I used the sedation example as one reason why.  The other major reasons are:
  • There is no on-site intensivist during the full 24 hours in the ICUs in which I work.  Proning and supining cannot therefore be intensivist supervised routinely, and ETT malposition incidents will be more disruptive.  The ED physician will have to be called emergently to respond to any tube reposition needs. 
  • Our ICUs sometimes have a very low census (the reason for that would be a good post for the status iatrogenicus blog).  Some nights there are only two nurses in the ICU.  With those low staffing levels, it is simply not feasible to prone and supine a patient for nursing care and repositioning.
  • Note the restrictive inclusion criteria in the current study:  ARDS for less than 36 hours, PF ratio less than 150 on at least FiO2 0.6 and PEEP greater than or equal to 5, and these criteria had to be reconfirmed (that is met again) after 12-24 hours in the ICU.  These patients are a mionority subset of the total ALI/ARDS cohort.  In my practice, we will have perhaps a dozen patients per year that meet inclusion criteria.
  • The exclusion criteria are available here for subscribers.  Note that MAP less than 65, non-invasive ventilation for 24 hours prior to inclusion, chronic oxygen use, and various other criteria led to exclusion from enrollment.  I mention this not only in relation to the bullet point above, but also because, if you believe the historical legacy of trials of prone positioning, and you believe that investigators have refined their inclusion and exclusion criteria to "hone in" on exactly which patients benefit and thusly finally got a positive result, then strict attention to inclusion/exclusion criteria is warranted, even more than in the usual case.
Finally, I worry that this is going to be a lot of work and a distraction in our ICUs from other important programs such as early mobility, and given my skepticism of this one positive study's results I just don't believe that a new program of routine prone positioning is warranted at this juncture.  When I have a patient who remains severely hypoxemic, I will certainly consider this therapy.  But mostly, I will continue to use it as I always have, as a "rescue therapy."  I am not steadfast in this position and am certainly open to further debate - so feel free to comment below.]

[Addendum 6/7/2013:  In this week's JAMA, Murad and Montori discuss the imperative to consider not just single studies in isolation, but the totality of the evidence, as I attempted to do above.]


  1. Hi Scott

    I disagree with your conclusion that this paper is not inline with previous publications.

    This paper is consistent with a 2011 meta-analysis suggesting a mortality benefit for patients with severe ARDS (PF ratio <100), with an NNT of 11 ( It makes sense that proning may benefit this group, but not less severe forms of ARDS.

    Luciano Gattinoni recently commented at the CICM ASM 2013 that the NNT is likely inflated due to this study being performed during the influenza 'epidemic'. I doubt that the NNT is externally valid, but I think that proning is sufficiently safe, cheap and likely to be effective that we should be performing it on patients with PF ratios <100.


    1. Chris - The 95% CI in the meta-analysis you referenced is 0.5-0.99 (P-value 0.048). That's just not a robust result. Note also that these studies are not and cannot be blinded, so differences in care are possible. It reminds me of the HACA studies, which are also statistically marginal, but also have a softer endpoint. And I would bet that if we did a really large multinational multicenter trial of hypothermia after cardiac arrest, that we would find that it doesn't work. That's just my guess, but it's a guess based on the experience of seeing Type 1 errors revealed time and again as trials are repeated. Especially in those that have other methodological shortcomings, marginal P-values, and effect sizes that are "too good to be true."

    2. I did too and saw dramatic improvements in oxygenation. PAtient had dense basal consolidations from ARDS!