Talk:PROSEVA

Too good to be true?

A 50% relative survival benefit is great, but it's hard to swallow. I need to read the article more closely to see what may have accounted for this enormous improvement. The meta-analysis I referenced in the major points [1] showed only a ~16% risk reduction using prone position. I don't think this should be made usable until we address this huge discrepancy, however. --Dave Iberri (talk) 19:03, 12 June 2013 (PDT)

I added some comments in the major points to help explain. --Dave Iberri (talk) 06:43, 13 June 2013 (PDT)

I think it looks much better. Need published commentary to fill out this section -- will have to revisit in a few months. --Tim Plante (talk) 10:35, 13 June 2013 (PDT)

HR vs. RR

I think that 61% was wrong: The HR for prone positioning on 28-day mortality was 0.39. Unfortunately, you can't simply take 1-HR and get a relative risk reduction, so saying that prone positioning reduces 28-day mortality by 61% (1 - 39%) is not quite right. What you really want is 1-RR, and in this trial, RR was 0.16/0.328 = 0.49, which yields a 51% relative risk reduction. --Dave Iberri (talk) 06:43, 13 June 2013 (PDT)

You are correct! My mistake. Thanks for the correction. --Tim Plante (talk) 10:35, 13 June 2013 (PDT)

The real information is in the supplementary material

I don't like this study because they hide the results which show the problems with their study in the supplementary info.

Even ignoring the huge issues with this study, prone positioning doesn't really make physiological sense if you keep someone in it for their entire ICU stay - if your improvement in oxygenation is due to increased perfusion to non-atelectatic alveoli... then any benefit you see should disappear within 24hr of switching them to a prone position, as by that point, those alveoli that had been open are now collapsed as they're at the bottom.

Anyways...

First of all, look at the full exclusion criteria to prone positioning in the supplemental material. A lot of them boil down to active medical comorbidities (recent pacemaker placement, recent DVT, life expectency <1yr, etc.) as well as hypotension (MAP < 65) - this is a big one to keep in mind going forwards.

Next look at Table S2. Again, the big thing here is comorbidities. Renal failure, liver disease, coronary artery disease, malignancy, COPD, and immunodeficiency, were all more common in the supine group - only diabetes was more common in the prone group. And while diabetes does have important mortality implications in ICU patients, I don't think it's to the same extent as these other ones. Add it all up, and of all the comorbidities present (n will be greater than the number of patients as some px will have more than one comorbidity), and there were 5% more comorbidities in the supine group. Ignore diabetes and that increases to 13%.

Next, in the same table, look at the co-interventions - There as a 10% absolute difference in the number of patients who needed vasopressors (more in the supine group - not surprising based on the exclusion criteria and worse health of this group). 6% less of the prone group also required dialysis. This also fits with worse hypotension in the supine group.

Strangely, 10% less supine patients received paralytics... which is interesting based on the fact paralytics are potentially beneficial in these people (though that is controversial).

How do I interpret this? The supine group was systemically more sick than the prone group... and part of this is likely related to their inclusion criteria. Randomization doesn't matter if what you're putting into the computer is already biased towards one group.

And the supine group wasn't receiving the same care as the prone group - witness the difference in paralytic use - this is likely a consequence of the difficulty of blinding someone to whether the patient is prone or not.

The real kicker here is table S7. What was the actual cause of death in these people? It turns out that almost all of the mortality difference between the two groups was due to a reduction in refractory shock in the prone group. Which is not surprising when hypotension is an exclusion criteria to enter the prone group.

In short, almost all of the significant decrease in mortality was driven by a decrease in deaths due to refractory shock... and hypotension was an exclusion criteria for the prone group.

What this study really shows us is that patients with hypotension have a higher mortality rate.

The actual conclusion to this study should be that with prone positioning there was a small, but statistically significant decrease in PEEP and FiO2 in the first 5days. This disappeared by day 7, and was not associated with an decreased risk of pneumothorax, or death due to hypoxemia. However, prone positioning was associated with a significantly increased risk of accidental extubation, and endotracheal tube obstruction. The abstract and conclusion of this study are terrible and misleading, and I'm surprised that NEJM let them publish this - it's a bit embarrassing.--Cnidos (talk) 10:02, 21 June 2013 (PDT)

I agree with a huge amount of this. In regards to the MAP <65 as an exclusion, it seems that if you were unable to be placed in a prone position at all (including MAP <65 mmHg, ICP >30, hemoptysis, etc), you were excluded from randomization entirely to both groups -- not just the prone group. Another point is that people were kept in prone positioning for a goal of 16 hours at a time, ~70% of this was actually met so people ended up both prone and supine for about 50% of the time throughout every day. The authors don't do a very good job of making that clear. In regards to the Table 1 (a.k.a. table S2), the groups were significantly different based upon their "illnesses," specifically SOFA scores, amount who were septic, and requiring pressors. The others didn't have statistical difference for diabetes, CKD, liver disease, etc. That being said, it's clear that the prone group started off sicker. Not a surprise that they would do better than the supine group! --Tim Plante (talk) 13:36, 21 June 2013 (PDT)